diff --git a/en_US.ISO8859-1/articles/ipsec-must/article.sgml b/en_US.ISO8859-1/articles/ipsec-must/article.sgml index 34edb9c5c5..912364faf1 100644 --- a/en_US.ISO8859-1/articles/ipsec-must/article.sgml +++ b/en_US.ISO8859-1/articles/ipsec-must/article.sgml @@ -1,297 +1,297 @@ Independent Verification of IPSec Functionality in FreeBSD

Independent Verification of IPsec Functionality Under FreeBSD 3.0

You installed IPsec and it seems to be working. How do you know? I describe a method for experimentally verifying that IPsec is working

The Problem

First, let's assume you have installed IPsec. How do you know its working? Sure, your connection won't work if its misconfigured, and it will work when you finally get it right. Netstat will list it. But can you independently confirm it?

The Solution

First, some crypto-relevent info theory:

Encrypted data is uniformly distributed, ie, has maximal entropy per symbol.
Raw, uncompressed data is typically redundant, i.e., has sub-maximal entropy.

Suppose you could measure the entropy of the data to- and from- your network interface. Then you could see the difference between unencrypted data and encrypted data. This would be true even if some of the data in "encrypted mode" was not encrypted ---as the outermost IP header must be, if the packet is to be routable.

MUST

Ueli Maurer's "Universal Statistical Test for Random Bit Generators" ("MUST") quickly measures the entropy of a sample. It uses a compression-like algorithm. The code is given below for a variant which measures successive (~quarter megabyte) chunks of a file.

Tcpdump

We also need a way to capture the raw network data. A program called "tcpdump" lets you do this, if you have enabled the Berkeley Packet Filter interface in your kernel's config file.

The command

tcpdump -c 4000 -s 10000 -w dumpfile.bin

will capture 4000 raw packets to dumpfile.bin. Up to 10,000 bytes per packet will be captured in this example.

The Experiment

Here's the experiment. Open a window to an IPsec host and another window to an insecure host.

Now start capturing packets.

In the "secure" window, run the unix command "yes", which will stream the "y" character. After a while, stop this. Switch to the insecure window, and repeat. After a while, stop.

Now run MUST on the - captured packets. You should see something like the the following. + captured packets. You should see something like the following. The important thing to note is that the secure connection has 93% (6.7) of the expected value (7.18), and the "normal" connection has 29% (2.1) of the expected value.

% tcpdump -c 4000 -s 10000 -w ipsecdemo.bin
 % uliscan ipsecdemo.bin
 
 Uliscan 21 Dec 98
 L=8 256 258560
 Measuring file ipsecdemo.bin
 Init done
 Expected value for L=8 is 7.1836656
 6.9396 --------------------------------------------------------
 6.6177 -----------------------------------------------------
 6.4100 ---------------------------------------------------
 2.1101 -----------------
 2.0838 -----------------
 2.0983 -----------------

Caveat

This experiment shows that IPsec does seem to be distributing the payload data uniformly, as encryption should. However, the experiment described here can not detect many possible flaws in a system (none of which do I have any evidence for). These include poor key generation or exchange, data or keys being visible to others, use of weak algorithms, kernel subversion, etc. Study the source; know the code.

IPsec -Definition

Internet Protocol security extensions to IP v 4; required for IP v6. A protocol for negotiating encryption and authentication at the IP (host-to-host) level. SSL secures only one application socket; SSH secures only a login; PGP secures only a specified file or message. IPsec encrypts everything between two hosts.

Installing IPsec

Starting from the BSD 3.0 stable release,

install IPsec v0.04, rebuild, reinstall
run the administration tools (e.g, ipsecadm) and distribute keys (or use Photuris for key exchange)
set the routes (rt) up appropriately

You may want to make an "ipsec_setup" script containing the ipsecadm and rt commands which establish your IPsec tunnel. You can run this script automatically at boottime from your /etc/rc.local The ipsec_setup script will have to contain at least two ipsecadm commands and one rt command to be useful.

usr/src/sys/i386/conf/KERNELNAME

This needs to be present in the kernel config file in order to run IPsec. After adding it, run config, etc. and rebuild and reinstall.

# The `bpfilter' pseudo-device enables the Berkeley Packet Filter. Be
 # aware of the legal and administrative consequences of enabling this
 # option. Heh heh. The number of devices determines the maximum number of
 # simultaneous BPF clients programs runnable.
 pseudo-device bpfilter 2 #Berkeley packet filter
 
 # IPSEC
 options IPSEC
 options "MD5"
 pseudo-device enc 1

Maurer's Universal Statistical Test (for block size=8 bits)


 #include 
 
 int main(argc, argv)
 int argc;
 char **argv;
 {
   FILE *fptr;
   int i,j;
   int b, c;
   int table[V];
   double sum = 0.0;
   int iproduct = 1;
   int run;
 
   extern double   log(/* double x */);
 
   printf("Uliscan 21 Dec 98 \nL=%d %d %d \n", L, V, MAXSAMP);
 
   if (argc < 2) {
     printf("Usage: Uliscan filename\n");
     exit(-1);
   } else {
     printf("Measuring file %s\n", argv[1]);
   }
 
   fptr = fopen(argv[1],"rb");
 
   if (fptr == NULL) {
     printf("Can't find %s\n", argv[1]);
     exit(-1);
   }
 
   for (i = 0; i < V; i++) {
     table[i] = 0;
   }
 
   for (i = 0; i < Q; i++) {
     b = fgetc(fptr);
     table[b] = i;
   }
 
   printf("Init done\n");
 
   printf("Expected value for L=8 is 7.1836656\n");
 
   run = 1;
 
   while (run) {
     sum = 0.0;
     iproduct = 1;
 
     if (run)
       for (i = Q; run && i < Q + K; i++) {
         j = i;
         b = fgetc(fptr);
 
         if (b < 0)
           run = 0;
 
         if (run) {
           if (table[b] > j)
             j += K;
 
           sum += log((double)(j-table[b]));
 
           table[b] = i;
         }
       }
 
     if (!run)
       printf("Premature end of file; read %d blocks.\n", i - Q);
 
     sum = (sum/((double)(i - Q))) /  log(2.0);
     printf("%4.4f ", sum);
 
     for (i = 0; i < (int)(sum*8.0 + 0.50); i++)
       printf("-");
 
     printf("\n");
 
     /* refill initial table */
     if (0) {
       for (i = 0; i < Q; i++) {
         b = fgetc(fptr);
         if (b < 0) {
           run = 0;
         } else {
           table[b] = i;
         }
       }
     }
   }
 }]]>

diff --git a/en_US.ISO8859-1/books/arch-handbook/scsi/chapter.sgml b/en_US.ISO8859-1/books/arch-handbook/scsi/chapter.sgml index 63b8d5475d..72ee58a453 100644 --- a/en_US.ISO8859-1/books/arch-handbook/scsi/chapter.sgml +++ b/en_US.ISO8859-1/books/arch-handbook/scsi/chapter.sgml @@ -1,1983 +1,1983 @@ Common Access Method SCSI Controllers This chapter was written by &a.babkin; Modifications for the handbook made by &a.murray;. Synopsis This document assumes that the reader has a general understanding of device drivers in FreeBSD and of the SCSI protocol. Much of the information in this document was extracted from the drivers : ncr (/sys/pci/ncr.c) by Wolfgang Stanglmeier and Stefan Esser sym (/sys/pci/sym.c) by Gerard Roudier aic7xxx (/sys/dev/aic7xxx/aic7xxx.c) by Justin T. Gibbs and from the CAM code itself (by Justing T. Gibbs, see /sys/cam/*). When some solution looked the most logical and was essentially verbatim extracted from the code by Justin Gibbs, I marked it as "recommended". The document is illustrated with examples in pseudo-code. Although sometimes the examples have many details and look like real code, it's still pseudo-code. It was written to demonstrate the concepts in an understandable way. For a real driver other approaches may be more modular and efficient. It also abstracts from the hardware details, as well as issues that would cloud the demonstrated concepts or that are supposed to be described in the other chapters of the developers handbook. Such details are commonly shown as calls to functions with descriptive names, comments or pseudo-statements. Fortunately real life full-size examples with all the details can be found in the real drivers. General architecture CAM stands for Common Access Method. It's a generic way to address the I/O buses in a SCSI-like way. This allows a separation of the generic device drivers from the drivers controlling the I/O bus: for example the disk driver becomes able to control disks on both SCSI, IDE, and/or any other bus so the disk driver portion does not have to be rewritten (or copied and modified) for every new I/O bus. Thus the two most important active entities are: Peripheral Modules - a driver for peripheral devices (disk, tape, CDROM, etc.) SCSI Interface Modules (SIM) - a Host Bus Adapter drivers for connecting to an I/O bus such as SCSI or IDE. A peripheral driver receives requests from the OS, converts them to a sequence of SCSI commands and passes these SCSI commands to a SCSI Interface Module. The SCSI Interface Module is responsible for passing these commands to the actual hardware (or if the actual hardware is not SCSI but, for example, IDE then also converting the SCSI commands to the native commands of the hardware). Because we are interested in writing a SCSI adapter driver here, from this point on we will consider everything from the SIM standpoint. A typical SIM driver needs to include the following CAM-related header files: #include <cam/cam.h> #include <cam/cam_ccb.h> #include <cam/cam_sim.h> #include <cam/cam_xpt_sim.h> #include <cam/cam_debug.h> #include <cam/scsi/scsi_all.h> The first thing each SIM driver must do is register itself with the CAM subsystem. This is done during the driver's xxx_attach() function (here and further xxx_ is used to denote the unique driver name prefix). The xxx_attach() function itself is called by the system bus auto-configuration code which we don't describe here. This is achieved in multiple steps: first it's necessary to allocate the queue of requests associated with this SIM: struct cam_devq *devq; if(( devq = cam_simq_alloc(SIZE) )==NULL) { error; /* some code to handle the error */ } Here SIZE is the size of the queue to be allocated, maximal number of requests it could contain. It's the number of requests that the SIM driver can handle in parallel on one SCSI card. Commonly it can be calculated as: SIZE = NUMBER_OF_SUPPORTED_TARGETS * MAX_SIMULTANEOUS_COMMANDS_PER_TARGET Next we create a descriptor of our SIM: struct cam_sim *sim; if(( sim = cam_sim_alloc(action_func, poll_func, driver_name, softc, unit, max_dev_transactions, max_tagged_dev_transactions, devq) )==NULL) { cam_simq_free(devq); error; /* some code to handle the error */ } Note that if we are not able to create a SIM descriptor we free the devq also because we can do nothing else with it and we want to conserve memory. If a SCSI card has multiple SCSI buses on it then each bus requires its own cam_sim structure. An interesting question is what to do if a SCSI card has more than one SCSI bus, do we need one devq structure per card or per SCSI bus? The answer given in the comments to the CAM code is: either way, as the driver's author prefers. The arguments are : action_func - pointer to the driver's xxx_action function. static void xxx_action struct cam_sim *sim, union ccb *ccb poll_func - pointer to the driver's xxx_poll() static void xxx_poll struct cam_sim *sim driver_name - the name of the actual driver, such as "ncr" or "wds" softc - pointer to the driver's internal descriptor for this SCSI card. This pointer will be used by the driver in future to get private data. unit - the controller unit number, for example for controller "wds0" this number will be 0 max_dev_transactions - maximal number of simultaneous transactions per SCSI target in the non-tagged mode. This value will be almost universally equal to 1, with possible exceptions only for the non-SCSI cards. Also the drivers that hope to take advantage by preparing one transaction while another one is executed may set it to 2 but this does not seem to be worth the complexity. max_tagged_dev_transactions - the same thing, but in the tagged mode. Tags are the SCSI way to initiate multiple transactions on a device: each transaction is assigned a unique tag and the transaction is sent to the device. When the device completes some transaction it sends back the result together with the tag so that the SCSI adapter (and the driver) can tell which transaction was completed. This argument is also known as the maximal tag depth. It depends on the abilities of the SCSI adapter. Finally we register the SCSI buses associated with our SCSI adapter: if(xpt_bus_register(sim, bus_number) != CAM_SUCCESS) { cam_sim_free(sim, /*free_devq*/ TRUE); error; /* some code to handle the error */ } If there is one devq structure per SCSI bus (i.e. we consider a card with multiple buses as multiple cards with one bus each) then the bus number will always be 0, otherwise each bus on the SCSI card should be get a distinct number. Each bus needs its own separate structure cam_sim. After that our controller is completely hooked to the CAM system. The value of devq can be discarded now: sim will be passed as an argument in all further calls from CAM and devq can be derived from it. CAM provides the framework for such asynchronous events. Some events originate from the lower levels (the SIM drivers), some events originate from the peripheral drivers, some events originate from the CAM subsystem itself. Any driver can register callbacks for some types of the asynchronous events, so that it would be notified if these events occur. A typical example of such an event is a device reset. Each transaction and event identifies the devices to which it applies by the means of "path". The target-specific events normally occur during a transaction with this device. So the path from that transaction may be re-used to report this event (this is safe because the event path is copied in the event reporting routine but not deallocated nor passed anywhere further). Also it's safe to allocate paths dynamically at any time including the interrupt routines, although that incurs certain overhead, and a possible problem with this approach is that there may be no free memory at that time. For a bus reset event we need to define a wildcard path including all devices on the bus. So we can create the path for the future bus reset events in advance and avoid problems with the future memory shortage: struct cam_path *path; if(xpt_create_path(&path, /*periph*/NULL, cam_sim_path(sim), CAM_TARGET_WILDCARD, CAM_LUN_WILDCARD) != CAM_REQ_CMP) { xpt_bus_deregister(cam_sim_path(sim)); cam_sim_free(sim, /*free_devq*/TRUE); error; /* some code to handle the error */ } softc->wpath = path; softc->sim = sim; As you can see the path includes: ID of the peripheral driver (NULL here because we have none) ID of the SIM driver (cam_sim_path(sim)) SCSI target number of the device (CAM_TARGET_WILDCARD means "all devices") SCSI LUN number of the subdevice (CAM_LUN_WILDCARD means "all LUNs") If the driver can't allocate this path it won't be able to work normally, so in that case we dismantle that SCSI bus. And we save the path pointer in the softc structure for future use. After that we save the value of sim (or we can also discard it on the exit from xxx_probe() if we wish). That's all for a minimalistic initialization. To do things right there is one more issue left. For a SIM driver there is one particularly interesting event: when a target device is considered lost. In this case resetting the SCSI negotiations with this device may be a good idea. So we register a callback for this event with CAM. The request is passed to CAM by requesting CAM action on a CAM control block for this type of request: struct ccb_setasync csa; xpt_setup_ccb(&csa.ccb_h, path, /*priority*/5); csa.ccb_h.func_code = XPT_SASYNC_CB; csa.event_enable = AC_LOST_DEVICE; csa.callback = xxx_async; csa.callback_arg = sim; xpt_action((union ccb *)&csa); Now we take a look at the xxx_action() and xxx_poll() driver entry points. static void xxx_action struct cam_sim *sim, union ccb *ccb Do some action on request of the CAM subsystem. Sim describes the SIM for the request, CCB is the request itself. CCB stands for "CAM Control Block". It is a union of many specific instances, each describing arguments for some type of transactions. All of these instances share the CCB header where the common part of arguments is stored. CAM supports the SCSI controllers working in both initiator ("normal") mode and target (simulating a SCSI device) mode. Here we only consider the part relevant to the initiator mode. There are a few function and macros (in other words, methods) defined to access the public data in the struct sim: cam_sim_path(sim) - the path ID (see above) cam_sim_name(sim) - the name of the sim cam_sim_softc(sim) - the pointer to the softc (driver private data) structure cam_sim_unit(sim) - the unit number cam_sim_bus(sim) - the bus ID To identify the device, xxx_action() can get the unit number and pointer to its structure softc using these functions. The type of request is stored in ccb->ccb_h.func_code. So generally xxx_action() consists of a big switch: struct xxx_softc *softc = (struct xxx_softc *) cam_sim_softc(sim); struct ccb_hdr *ccb_h = &ccb->ccb_h; int unit = cam_sim_unit(sim); int bus = cam_sim_bus(sim); switch(ccb_h->func_code) { case ...: ... default: ccb_h->status = CAM_REQ_INVALID; xpt_done(ccb); break; } As can be seen from the default case (if an unknown command was received) the return code of the command is set into ccb->ccb_h.status and the completed CCB is returned back to CAM by calling xpt_done(ccb). xpt_done() does not have to be called from xxx_action(): For example an I/O request may be enqueued inside the SIM driver and/or its SCSI controller. Then when the device would post an interrupt signaling that the processing of this request is complete xpt_done() may be called from the interrupt handling routine. Actually, the CCB status is not only assigned as a return code but a CCB has some status all the time. Before CCB is passed to the xxx_action() routine it gets the status CCB_REQ_INPROG meaning that it's in progress. There are a surprising number of status values defined in /sys/cam/cam.h which should be able to represent the status of a request in great detail. More interesting yet, the status is in fact a "bitwise or" of an enumerated status value (the lower 6 bits) and possible additional flag-like bits (the upper bits). The enumerated values will be discussed later in more detail. The summary of them can be found in the Errors Summary section. The possible status flags are: CAM_DEV_QFRZN - if the SIM driver gets a serious error (for example, the device does not respond to the selection or breaks the SCSI protocol) when processing a CCB it should freeze the request queue by calling xpt_freeze_simq(), return the other enqueued but not processed yet CCBs for this device back to the CAM queue, then set this flag for the troublesome CCB and call xpt_done(). This flag causes the CAM subsystem to unfreeze the queue after it handles the error. CAM_AUTOSNS_VALID - if the device returned an error condition and the flag CAM_DIS_AUTOSENSE is not set in CCB the SIM driver must execute the REQUEST SENSE command automatically to extract the sense (extended error information) data from the device. If this attempt was successful the sense data should be saved in the CCB and this flag set. CAM_RELEASE_SIMQ - like CAM_DEV_QFRZN but used in case there is some problem (or resource shortage) with the SCSI controller itself. Then all the future requests to the controller should be stopped by xpt_freeze_simq(). The controller queue will be restarted after the SIM driver overcomes the shortage and informs CAM by returning some CCB with this flag set. CAM_SIM_QUEUED - when SIM puts a CCB into its request queue this flag should be set (and removed when this CCB gets dequeued before being returned back to CAM). This flag is not used anywhere in the CAM code now, so its purpose is purely diagnostic. The function xxx_action() is not allowed to sleep, so all the synchronization for resource access must be done using SIM or device queue freezing. Besides the aforementioned flags the CAM subsystem provides functions xpt_selease_simq() and xpt_release_devq() to unfreeze the queues directly, without passing a CCB to CAM. The CCB header contains the following fields: path - path ID for the request target_id - target device ID for the request target_lun - LUN ID of the target device timeout - timeout interval for this command, in milliseconds timeout_ch - a convenience place for the SIM driver to store the timeout handle (the CAM subsystem itself does not make any assumptions about it) flags - various bits of information about the request spriv_ptr0, spriv_ptr1 - fields reserved for private use by the SIM driver (such as linking to the SIM queues or SIM private control blocks); actually, they exist as unions: spriv_ptr0 and spriv_ptr1 have the type (void *), spriv_field0 and spriv_field1 have the type unsigned long, sim_priv.entries[0].bytes and sim_priv.entries[1].bytes are byte arrays of the size consistent with the other incarnations of the union and sim_priv.bytes is one array, twice bigger. The recommended way of using the SIM private fields of CCB is to define some meaningful names for them and use these meaningful names in the driver, like: #define ccb_some_meaningful_name sim_priv.entries[0].bytes #define ccb_hcb spriv_ptr1 /* for hardware control block */ The most common initiator mode requests are: XPT_SCSI_IO - execute an I/O transaction The instance "struct ccb_scsiio csio" of the union ccb is used to transfer the arguments. They are: cdb_io - pointer to the SCSI command buffer or the buffer itself cdb_len - SCSI command length data_ptr - pointer to the data buffer (gets a bit complicated if scatter/gather is used) dxfer_len - length of the data to transfer sglist_cnt - counter of the scatter/gather segments scsi_status - place to return the SCSI status sense_data - buffer for the SCSI sense information if the command returns an error (the SIM driver is supposed to run the REQUEST SENSE command automatically in this case if the CCB flag CAM_DIS_AUTOSENSE is not set) sense_len - the length of that buffer (if it happens to be higher than size of sense_data the SIM driver must silently assume the smaller value) resid, sense_resid - if the transfer of data or SCSI sense returned an error these are the returned counters of the residual (not transferred) data. They do not seem to be especially meaningful, so in a case when they are difficult to compute (say, counting bytes in the SCSI controller's FIFO buffer) an approximate value will do as well. For a successfully completed transfer they must be set to zero. tag_action - the kind of tag to use: CAM_TAG_ACTION_NONE - don't use tags for this transaction MSG_SIMPLE_Q_TAG, MSG_HEAD_OF_Q_TAG, MSG_ORDERED_Q_TAG - value equal to the appropriate tag message (see /sys/cam/scsi/scsi_message.h); this gives only the tag type, the SIM driver must assign the tag value itself The general logic of handling this request is the following: The first thing to do is to check for possible races, to make sure that the command did not get aborted when it was sitting in the queue: struct ccb_scsiio *csio = &ccb->csio; if ((ccb_h->status & CAM_STATUS_MASK) != CAM_REQ_INPROG) { xpt_done(ccb); return; } Also we check that the device is supported at all by our controller: if(ccb_h->target_id > OUR_MAX_SUPPORTED_TARGET_ID || cch_h->target_id == OUR_SCSI_CONTROLLERS_OWN_ID) { ccb_h->status = CAM_TID_INVALID; xpt_done(ccb); return; } if(ccb_h->target_lun > OUR_MAX_SUPPORTED_LUN) { ccb_h->status = CAM_LUN_INVALID; xpt_done(ccb); return; } Then allocate whatever data structures (such as card-dependent hardware control block) we need to process this request. If we can't then freeze the SIM queue and remember that we have a pending operation, return the CCB back and ask CAM to re-queue it. Later when the resources become available the SIM queue must be unfrozen by returning a ccb with the CAM_SIMQ_RELEASE bit set in its status. Otherwise, if all went well, link the CCB with the hardware control block (HCB) and mark it as queued. struct xxx_hcb *hcb = allocate_hcb(softc, unit, bus); if(hcb == NULL) { softc->flags |= RESOURCE_SHORTAGE; xpt_freeze_simq(sim, /*count*/1); ccb_h->status = CAM_REQUEUE_REQ; xpt_done(ccb); return; } hcb->ccb = ccb; ccb_h->ccb_hcb = (void *)hcb; ccb_h->status |= CAM_SIM_QUEUED; Extract the target data from CCB into the hardware control block. Check if we are asked to assign a tag and if yes then generate an unique tag and build the SCSI tag messages. The SIM driver is also responsible for negotiations with the devices to set the maximal mutually supported bus width, synchronous rate and offset. hcb->target = ccb_h->target_id; hcb->lun = ccb_h->target_lun; generate_identify_message(hcb); if( ccb_h->tag_action != CAM_TAG_ACTION_NONE ) generate_unique_tag_message(hcb, ccb_h->tag_action); if( !target_negotiated(hcb) ) generate_negotiation_messages(hcb); Then set up the SCSI command. The command storage may be specified in the CCB in many interesting ways, specified by the CCB flags. The command buffer can be contained in CCB or pointed to, in the latter case the pointer may be physical or virtual. Since the hardware commonly needs physical address we always convert the address to the physical one. A NOT-QUITE RELATED NOTE: Normally this is done by a call to vtophys(), but for the PCI device (which account for most of the SCSI controllers now) drivers' portability to the Alpha architecture the conversion must be done by vtobus() instead due to special Alpha quirks. [IMHO it would be much better to have two separate functions, vtop() and ptobus() then vtobus() would be a simple superposition of them.] In case if a physical address is requested it's OK to return the CCB with the status CAM_REQ_INVALID, the current drivers do that. But it's also possible to compile the Alpha-specific piece of code, as in this example (there should be a more direct way to do that, without conditional compilation in the drivers). If necessary a physical address can be also converted or mapped back to a virtual address but with big pain, so we don't do that. if(ccb_h->flags & CAM_CDB_POINTER) { /* CDB is a pointer */ if(!(ccb_h->flags & CAM_CDB_PHYS)) { /* CDB pointer is virtual */ hcb->cmd = vtobus(csio->cdb_io.cdb_ptr); } else { /* CDB pointer is physical */ #if defined(__alpha__) hcb->cmd = csio->cdb_io.cdb_ptr | alpha_XXX_dmamap_or ; #else hcb->cmd = csio->cdb_io.cdb_ptr ; #endif } } else { /* CDB is in the ccb (buffer) */ hcb->cmd = vtobus(csio->cdb_io.cdb_bytes); } hcb->cmdlen = csio->cdb_len; Now it's time to set up the data. Again, the data storage may be specified in the CCB in many interesting ways, specified by the CCB flags. First we get the direction of the data transfer. The simplest case is if there is no data to transfer: int dir = (ccb_h->flags & CAM_DIR_MASK); if (dir == CAM_DIR_NONE) goto end_data; Then we check if the data is in one chunk or in a scatter-gather list, and the addresses are physical or virtual. The SCSI controller may be able to handle only a limited number of chunks of limited length. If the request hits this limitation we return an error. We use a special function to return the CCB to handle in one place the HCB resource shortages. The functions to add chunks are driver-dependent, and here we leave them without detailed implementation. See description of the SCSI command (CDB) handling for the details on the address-translation issues. If some variation is too difficult or impossible to implement with a particular card it's OK to return the status CAM_REQ_INVALID. Actually, it seems like the scatter-gather ability is not used anywhere in the CAM code now. But at least the case for a single non-scattered virtual buffer must be implemented, it's actively used by CAM. int rv; initialize_hcb_for_data(hcb); if((!(ccb_h->flags & CAM_SCATTER_VALID)) { /* single buffer */ if(!(ccb_h->flags & CAM_DATA_PHYS)) { rv = add_virtual_chunk(hcb, csio->data_ptr, csio->dxfer_len, dir); } } else { rv = add_physical_chunk(hcb, csio->data_ptr, csio->dxfer_len, dir); } } else { int i; struct bus_dma_segment *segs; segs = (struct bus_dma_segment *)csio->data_ptr; if ((ccb_h->flags & CAM_SG_LIST_PHYS) != 0) { /* The SG list pointer is physical */ rv = setup_hcb_for_physical_sg_list(hcb, segs, csio->sglist_cnt); } else if (!(ccb_h->flags & CAM_DATA_PHYS)) { /* SG buffer pointers are virtual */ for (i = 0; i < csio->sglist_cnt; i++) { rv = add_virtual_chunk(hcb, segs[i].ds_addr, segs[i].ds_len, dir); if (rv != CAM_REQ_CMP) break; } } else { /* SG buffer pointers are physical */ for (i = 0; i < csio->sglist_cnt; i++) { rv = add_physical_chunk(hcb, segs[i].ds_addr, segs[i].ds_len, dir); if (rv != CAM_REQ_CMP) break; } } } if(rv != CAM_REQ_CMP) { /* we expect that add_*_chunk() functions return CAM_REQ_CMP * if they added a chunk successfully, CAM_REQ_TOO_BIG if * the request is too big (too many bytes or too many chunks), * CAM_REQ_INVALID in case of other troubles */ free_hcb_and_ccb_done(hcb, ccb, rv); return; } end_data: If disconnection is disabled for this CCB we pass this information to the hcb: if(ccb_h->flags & CAM_DIS_DISCONNECT) hcb_disable_disconnect(hcb); If the controller is able to run REQUEST SENSE command all by itself then the value of the flag CAM_DIS_AUTOSENSE should also be passed to it, to prevent automatic REQUEST SENSE if the CAM subsystem does not want it. The only thing left is to set up the timeout, pass our hcb to the hardware and return, the rest will be done by the interrupt handler (or timeout handler). ccb_h->timeout_ch = timeout(xxx_timeout, (caddr_t) hcb, (ccb_h->timeout * hz) / 1000); /* convert milliseconds to ticks */ put_hcb_into_hardware_queue(hcb); return; And here is a possible implementation of the function returning CCB: static void free_hcb_and_ccb_done(struct xxx_hcb *hcb, union ccb *ccb, u_int32_t status) { struct xxx_softc *softc = hcb->softc; ccb->ccb_h.ccb_hcb = 0; if(hcb != NULL) { untimeout(xxx_timeout, (caddr_t) hcb, ccb->ccb_h.timeout_ch); /* we're about to free a hcb, so the shortage has ended */ if(softc->flags & RESOURCE_SHORTAGE) { softc->flags &= ~RESOURCE_SHORTAGE; status |= CAM_RELEASE_SIMQ; } free_hcb(hcb); /* also removes hcb from any internal lists */ } ccb->ccb_h.status = status | (ccb->ccb_h.status & ~(CAM_STATUS_MASK|CAM_SIM_QUEUED)); xpt_done(ccb); } XPT_RESET_DEV - send the SCSI "BUS DEVICE RESET" message to a device There is no data transferred in CCB except the header and the most interesting argument of it is target_id. Depending on the controller hardware a hardware control block just like for the XPT_SCSI_IO request may be constructed (see XPT_SCSI_IO request description) and sent to the controller or the SCSI controller may be immediately programmed to send this RESET message to the device or this request may be just not supported (and return the status CAM_REQ_INVALID). Also on completion of the request all the disconnected transactions for this target must be aborted (probably in the interrupt routine). Also all the current negotiations for the target are lost on reset, so they might be cleaned too. Or they clearing may be deferred, because anyway the target would request re-negotiation on the next transaction. XPT_RESET_BUS - send the RESET signal to the SCSI bus No arguments are passed in the CCB, the only interesting argument is the SCSI bus indicated by the struct sim pointer. A minimalistic implementation would forget the SCSI negotiations for all the devices on the bus and return the status CAM_REQ_CMP. The proper implementation would in addition actually reset the SCSI bus (possible also reset the SCSI controller) and mark all the CCBs being processed, both those in the hardware queue and those being disconnected, as done with the status CAM_SCSI_BUS_RESET. Like: int targ, lun; struct xxx_hcb *h, *hh; struct ccb_trans_settings neg; struct cam_path *path; /* The SCSI bus reset may take a long time, in this case its completion * should be checked by interrupt or timeout. But for simplicity * we assume here that it's really fast. */ reset_scsi_bus(softc); /* drop all enqueued CCBs */ for(h = softc->first_queued_hcb; h != NULL; h = hh) { hh = h->next; free_hcb_and_ccb_done(h, h->ccb, CAM_SCSI_BUS_RESET); } /* the clean values of negotiations to report */ neg.bus_width = 8; neg.sync_period = neg.sync_offset = 0; neg.valid = (CCB_TRANS_BUS_WIDTH_VALID | CCB_TRANS_SYNC_RATE_VALID | CCB_TRANS_SYNC_OFFSET_VALID); /* drop all disconnected CCBs and clean negotiations */ for(targ=0; targ <= OUR_MAX_SUPPORTED_TARGET; targ++) { clean_negotiations(softc, targ); /* report the event if possible */ if(xpt_create_path(&path, /*periph*/NULL, cam_sim_path(sim), targ, CAM_LUN_WILDCARD) == CAM_REQ_CMP) { xpt_async(AC_TRANSFER_NEG, path, &neg); xpt_free_path(path); } for(lun=0; lun <= OUR_MAX_SUPPORTED_LUN; lun++) for(h = softc->first_discon_hcb[targ][lun]; h != NULL; h = hh) { hh=h->next; free_hcb_and_ccb_done(h, h->ccb, CAM_SCSI_BUS_RESET); } } ccb->ccb_h.status = CAM_REQ_CMP; xpt_done(ccb); /* report the event */ xpt_async(AC_BUS_RESET, softc->wpath, NULL); return; Implementing the SCSI bus reset as a function may be a good idea because it would be re-used by the timeout function as a last resort if the things go wrong. XPT_ABORT - abort the specified CCB The arguments are transferred in the instance "struct ccb_abort cab" of the union ccb. The only argument field in it is: abort_ccb - pointer to the CCB to be aborted If the abort is not supported just return the status CAM_UA_ABORT. This is also the easy way to minimally implement this call, return CAM_UA_ABORT in any case. The hard way is to implement this request honestly. First check that abort applies to a SCSI transaction: struct ccb *abort_ccb; abort_ccb = ccb->cab.abort_ccb; if(abort_ccb->ccb_h.func_code != XPT_SCSI_IO) { ccb->ccb_h.status = CAM_UA_ABORT; xpt_done(ccb); return; } Then it's necessary to find this CCB in our queue. This can be done by walking the list of all our hardware control blocks in search for one associated with this CCB: struct xxx_hcb *hcb, *h; hcb = NULL; /* We assume that softc->first_hcb is the head of the list of all * HCBs associated with this bus, including those enqueued for * processing, being processed by hardware and disconnected ones. */ for(h = softc->first_hcb; h != NULL; h = h->next) { if(h->ccb == abort_ccb) { hcb = h; break; } } if(hcb == NULL) { /* no such CCB in our queue */ ccb->ccb_h.status = CAM_PATH_INVALID; xpt_done(ccb); return; } hcb=found_hcb; Now we look at the current processing status of the HCB. It may be either sitting in the queue waiting to be sent to the SCSI bus, being transferred right now, or disconnected and waiting for the result of the command, or actually completed by hardware but not yet marked as done by software. To make sure that we don't get in any races with hardware we mark the HCB as being aborted, so that if this HCB is about to be sent to the SCSI bus the SCSI controller will see this flag and skip it. int hstatus; /* shown as a function, in case special action is needed to make * this flag visible to hardware */ set_hcb_flags(hcb, HCB_BEING_ABORTED); abort_again: hstatus = get_hcb_status(hcb); switch(hstatus) { case HCB_SITTING_IN_QUEUE: remove_hcb_from_hardware_queue(hcb); /* FALLTHROUGH */ case HCB_COMPLETED: /* this is an easy case */ free_hcb_and_ccb_done(hcb, abort_ccb, CAM_REQ_ABORTED); break; If the CCB is being transferred right now we would like to signal to the SCSI controller in some hardware-dependent way that we want to abort the current transfer. The SCSI controller would set the SCSI ATTENTION signal and when the target responds to it send an ABORT message. We also reset the timeout to make sure that the target is not sleeping forever. If the command would not get aborted in some reasonable time like 10 seconds the timeout routine would go ahead and reset the whole SCSI bus. Because the command will be aborted in some reasonable time we can just return the abort request now as successfully completed, and mark the aborted CCB as aborted (but not mark it as done yet). case HCB_BEING_TRANSFERRED: untimeout(xxx_timeout, (caddr_t) hcb, abort_ccb->ccb_h.timeout_ch); abort_ccb->ccb_h.timeout_ch = timeout(xxx_timeout, (caddr_t) hcb, 10 * hz); abort_ccb->ccb_h.status = CAM_REQ_ABORTED; /* ask the controller to abort that HCB, then generate * an interrupt and stop */ if(signal_hardware_to_abort_hcb_and_stop(hcb) < 0) { /* oops, we missed the race with hardware, this transaction * got off the bus before we aborted it, try again */ goto abort_again; } break; If the CCB is in the list of disconnected then set it up as an abort request and re-queue it at the front of hardware queue. Reset the timeout and report the abort request to be completed. case HCB_DISCONNECTED: untimeout(xxx_timeout, (caddr_t) hcb, abort_ccb->ccb_h.timeout_ch); abort_ccb->ccb_h.timeout_ch = timeout(xxx_timeout, (caddr_t) hcb, 10 * hz); put_abort_message_into_hcb(hcb); put_hcb_at_the_front_of_hardware_queue(hcb); break; } ccb->ccb_h.status = CAM_REQ_CMP; xpt_done(ccb); return; That's all for the ABORT request, although there is one more issue. Because the ABORT message cleans all the ongoing transactions on a LUN we have to mark all the other active transactions on this LUN as aborted. That should be done in the interrupt routine, after the transaction gets aborted. Implementing the CCB abort as a function may be quite a good idea, this function can be re-used if an I/O transaction times out. The only difference would be that the timed out transaction would return the status CAM_CMD_TIMEOUT for the timed out request. Then the case XPT_ABORT would be small, like that: case XPT_ABORT: struct ccb *abort_ccb; abort_ccb = ccb->cab.abort_ccb; if(abort_ccb->ccb_h.func_code != XPT_SCSI_IO) { ccb->ccb_h.status = CAM_UA_ABORT; xpt_done(ccb); return; } if(xxx_abort_ccb(abort_ccb, CAM_REQ_ABORTED) < 0) /* no such CCB in our queue */ ccb->ccb_h.status = CAM_PATH_INVALID; else ccb->ccb_h.status = CAM_REQ_CMP; xpt_done(ccb); return; XPT_SET_TRAN_SETTINGS - explicitly set values of SCSI transfer settings The arguments are transferred in the instance "struct ccb_trans_setting cts" of the union ccb: valid - a bitmask showing which settings should be updated: CCB_TRANS_SYNC_RATE_VALID - synchronous transfer rate CCB_TRANS_SYNC_OFFSET_VALID - synchronous offset CCB_TRANS_BUS_WIDTH_VALID - bus width CCB_TRANS_DISC_VALID - set enable/disable disconnection CCB_TRANS_TQ_VALID - set enable/disable tagged queuing flags - consists of two parts, binary arguments and identification of sub-operations. The binary arguments are : CCB_TRANS_DISC_ENB - enable disconnection CCB_TRANS_TAG_ENB - enable tagged queuing the sub-operations are: CCB_TRANS_CURRENT_SETTINGS - change the current negotiations CCB_TRANS_USER_SETTINGS - remember the desired user values sync_period, sync_offset - self-explanatory, if sync_offset==0 then the asynchronous mode is requested bus_width - bus width, in bits (not bytes) Two sets of negotiated parameters are supported, the user settings and the current settings. The user settings are not really used much in the SIM drivers, this is mostly just a piece of memory where the upper levels can store (and later recall) its ideas about the parameters. Setting the user parameters does not cause re-negotiation of the transfer rates. But when the SCSI controller does a negotiation it must never set the values higher than the user parameters, so it's essentially the top boundary. The current settings are, as the name says, current. Changing them means that the parameters must be re-negotiated on the next transfer. Again, these "new current settings" are not supposed to be forced on the device, just they are used as the initial step of negotiations. Also they must be limited by actual capabilities of the SCSI controller: for example, if the SCSI controller has 8-bit bus and the request asks to set 16-bit wide transfers this parameter must be silently truncated to 8-bit transfers before sending it to the device. One caveat is that the bus width and synchronous parameters are per target while the disconnection and tag enabling parameters are per lun. The recommended implementation is to keep 3 sets of negotiated (bus width and synchronous transfer) parameters: user - the user set, as above current - those actually in effect goal - those requested by setting of the "current" parameters The code looks like: struct ccb_trans_settings *cts; int targ, lun; int flags; cts = &ccb->cts; targ = ccb_h->target_id; lun = ccb_h->target_lun; flags = cts->flags; if(flags & CCB_TRANS_USER_SETTINGS) { if(flags & CCB_TRANS_SYNC_RATE_VALID) softc->user_sync_period[targ] = cts->sync_period; if(flags & CCB_TRANS_SYNC_OFFSET_VALID) softc->user_sync_offset[targ] = cts->sync_offset; if(flags & CCB_TRANS_BUS_WIDTH_VALID) softc->user_bus_width[targ] = cts->bus_width; if(flags & CCB_TRANS_DISC_VALID) { softc->user_tflags[targ][lun] &= ~CCB_TRANS_DISC_ENB; softc->user_tflags[targ][lun] |= flags & CCB_TRANS_DISC_ENB; } if(flags & CCB_TRANS_TQ_VALID) { softc->user_tflags[targ][lun] &= ~CCB_TRANS_TQ_ENB; softc->user_tflags[targ][lun] |= flags & CCB_TRANS_TQ_ENB; } } if(flags & CCB_TRANS_CURRENT_SETTINGS) { if(flags & CCB_TRANS_SYNC_RATE_VALID) softc->goal_sync_period[targ] = max(cts->sync_period, OUR_MIN_SUPPORTED_PERIOD); if(flags & CCB_TRANS_SYNC_OFFSET_VALID) softc->goal_sync_offset[targ] = min(cts->sync_offset, OUR_MAX_SUPPORTED_OFFSET); if(flags & CCB_TRANS_BUS_WIDTH_VALID) softc->goal_bus_width[targ] = min(cts->bus_width, OUR_BUS_WIDTH); if(flags & CCB_TRANS_DISC_VALID) { softc->current_tflags[targ][lun] &= ~CCB_TRANS_DISC_ENB; softc->current_tflags[targ][lun] |= flags & CCB_TRANS_DISC_ENB; } if(flags & CCB_TRANS_TQ_VALID) { softc->current_tflags[targ][lun] &= ~CCB_TRANS_TQ_ENB; softc->current_tflags[targ][lun] |= flags & CCB_TRANS_TQ_ENB; } } ccb->ccb_h.status = CAM_REQ_CMP; xpt_done(ccb); return; Then when the next I/O request will be processed it will check if it has to re-negotiate, for example by calling the function target_negotiated(hcb). It can be implemented like this: int target_negotiated(struct xxx_hcb *hcb) { struct softc *softc = hcb->softc; int targ = hcb->targ; if( softc->current_sync_period[targ] != softc->goal_sync_period[targ] || softc->current_sync_offset[targ] != softc->goal_sync_offset[targ] || softc->current_bus_width[targ] != softc->goal_bus_width[targ] ) return 0; /* FALSE */ else return 1; /* TRUE */ } After the values are re-negotiated the resulting values must be assigned to both current and goal parameters, so for future I/O transactions the current and goal parameters would be the same and target_negotiated() would return TRUE. When the card is initialized (in xxx_attach()) the current negotiation values must be initialized to narrow asynchronous mode, the goal and current values must be initialized to the maximal values supported by controller. XPT_GET_TRAN_SETTINGS - get values of SCSI transfer settings This operations is the reverse of XPT_SET_TRAN_SETTINGS. Fill up the CCB instance "struct ccb_trans_setting cts" with data as requested by the flags CCB_TRANS_CURRENT_SETTINGS or CCB_TRANS_USER_SETTINGS (if both are set then the existing drivers return the current settings). Set all the bits in the valid field. XPT_CALC_GEOMETRY - calculate logical (BIOS) geometry of the disk The arguments are transferred in the instance "struct ccb_calc_geometry ccg" of the union ccb: block_size - input, block (A.K.A sector) size in bytes volume_size - input, volume size in bytes cylinders - output, logical cylinders heads - output, logical heads secs_per_track - output, logical sectors per track If the returned geometry differs much enough from what the SCSI controller BIOS thinks and a disk on this SCSI controller is used as bootable the system may not be able to boot. The typical calculation example taken from the aic7xxx driver is: struct ccb_calc_geometry *ccg; u_int32_t size_mb; u_int32_t secs_per_cylinder; int extended; ccg = &ccb->ccg; size_mb = ccg->volume_size / ((1024L * 1024L) / ccg->block_size); extended = check_cards_EEPROM_for_extended_geometry(softc); if (size_mb > 1024 && extended) { ccg->heads = 255; ccg->secs_per_track = 63; } else { ccg->heads = 64; ccg->secs_per_track = 32; } secs_per_cylinder = ccg->heads * ccg->secs_per_track; ccg->cylinders = ccg->volume_size / secs_per_cylinder; ccb->ccb_h.status = CAM_REQ_CMP; xpt_done(ccb); return; This gives the general idea, the exact calculation depends on the quirks of the particular BIOS. If BIOS provides no way set the "extended translation" flag in EEPROM this flag should normally be assumed equal to 1. Other popular geometries are: 128 heads, 63 sectors - Symbios controllers 16 heads, 63 sectors - old controllers Some system BIOSes and SCSI BIOSes fight with each other with variable success, for example a combination of Symbios 875/895 SCSI and Phoenix BIOS can give geometry 128/63 after power up and 255/63 after a hard reset or soft reboot. XPT_PATH_INQ - path inquiry, in other words get the SIM driver and SCSI controller (also known as HBA - Host Bus Adapter) properties The properties are returned in the instance "struct ccb_pathinq cpi" of the union ccb: version_num - the SIM driver version number, now all drivers use 1 hba_inquiry - bitmask of features supported by the controller: PI_MDP_ABLE - supports MDP message (something from SCSI3?) PI_WIDE_32 - supports 32 bit wide SCSI PI_WIDE_16 - supports 16 bit wide SCSI PI_SDTR_ABLE - can negotiate synchronous transfer rate PI_LINKED_CDB - supports linked commands PI_TAG_ABLE - supports tagged commands PI_SOFT_RST - supports soft reset alternative (hard reset and soft reset are mutually exclusive within a SCSI bus) target_sprt - flags for target mode support, 0 if unsupported hba_misc - miscellaneous controller features: PIM_SCANHILO - bus scans from high ID to low ID PIM_NOREMOVE - removable devices not included in scan PIM_NOINITIATOR - initiator role not supported PIM_NOBUSRESET - user has disabled initial BUS RESET hba_eng_cnt - mysterious HBA engine count, something related to compression, now is always set to 0 vuhba_flags - vendor-unique flags, unused now max_target - maximal supported target ID (7 for 8-bit bus, 15 for 16-bit bus, 127 for Fibre Channel) max_lun - maximal supported LUN ID (7 for older SCSI controllers, 63 for newer ones) async_flags - bitmask of installed Async handler, unused now hpath_id - highest Path ID in the subsystem, unused now unit_number - the controller unit number, cam_sim_unit(sim) bus_id - the bus number, cam_sim_bus(sim) initiator_id - the SCSI ID of the controller itself base_transfer_speed - nominal transfer speed in KB/s for asynchronous narrow transfers, equals to 3300 for SCSI sim_vid - SIM driver's vendor id, a zero-terminated string of maximal length SIM_IDLEN including the terminating zero hba_vid - SCSI controller's vendor id, a zero-terminated string of maximal length HBA_IDLEN including the terminating zero dev_name - device driver name, a zero-terminated string of maximal length DEV_IDLEN including the terminating zero, equal to cam_sim_name(sim) The recommended way of setting the string fields is using strncpy, like: strncpy(cpi->dev_name, cam_sim_name(sim), DEV_IDLEN); After setting the values set the status to CAM_REQ_CMP and mark the CCB as done. Polling static void xxx_poll struct cam_sim *sim The poll function is used to simulate the interrupts when the interrupt subsystem is not functioning (for example, when the system has crashed and is creating the system dump). The CAM subsystem sets the proper interrupt level before calling the poll routine. So all it needs to do is to call the interrupt routine (or the other way around, the poll routine may be doing the real action and the interrupt routine would just call the poll routine). Why bother about a separate function then ? Because of different calling conventions. The xxx_poll routine gets the struct cam_sim pointer as its argument when the PCI interrupt routine by common convention gets pointer to the struct xxx_softc and the ISA interrupt routine - gets just the the device unit number. So the poll routine would + gets just the device unit number. So the poll routine would normally look as: static void xxx_poll(struct cam_sim *sim) { xxx_intr((struct xxx_softc *)cam_sim_softc(sim)); /* for PCI device */ } or static void xxx_poll(struct cam_sim *sim) { xxx_intr(cam_sim_unit(sim)); /* for ISA device */ } Asynchronous Events If an asynchronous event callback has been set up then the callback function should be defined. static void ahc_async(void *callback_arg, u_int32_t code, struct cam_path *path, void *arg) callback_arg - the value supplied when registering the callback code - identifies the type of event path - identifies the devices to which the event applies arg - event-specific argument Implementation for a single type of event, AC_LOST_DEVICE, looks like: struct xxx_softc *softc; struct cam_sim *sim; int targ; struct ccb_trans_settings neg; sim = (struct cam_sim *)callback_arg; softc = (struct xxx_softc *)cam_sim_softc(sim); switch (code) { case AC_LOST_DEVICE: targ = xpt_path_target_id(path); if(targ <= OUR_MAX_SUPPORTED_TARGET) { clean_negotiations(softc, targ); /* send indication to CAM */ neg.bus_width = 8; neg.sync_period = neg.sync_offset = 0; neg.valid = (CCB_TRANS_BUS_WIDTH_VALID | CCB_TRANS_SYNC_RATE_VALID | CCB_TRANS_SYNC_OFFSET_VALID); xpt_async(AC_TRANSFER_NEG, path, &neg); } break; default: break; } Interrupts The exact type of the interrupt routine depends on the type of the peripheral bus (PCI, ISA and so on) to which the SCSI controller is connected. The interrupt routines of the SIM drivers run at the interrupt level splcam. So splcam() should be used in the driver to synchronize activity between the interrupt routine and the rest of the driver (for a multiprocessor-aware driver things get yet more interesting but we ignore this case here). The pseudo-code in this document happily ignores the problems of synchronization. The real code must not ignore them. A simple-minded approach is to set splcam() on the entry to the other routines and reset it on return thus protecting them by one big critical section. To make sure that the interrupt level will be always restored a wrapper function can be defined, like: static void xxx_action(struct cam_sim *sim, union ccb *ccb) { int s; s = splcam(); xxx_action1(sim, ccb); splx(s); } static void xxx_action1(struct cam_sim *sim, union ccb *ccb) { ... process the request ... } This approach is simple and robust but the problem with it is that interrupts may get blocked for a relatively long time and this would negatively affect the system's performance. On the other hand the functions of the spl() family have rather high overhead, so vast amount of tiny critical sections may not be good either. The conditions handled by the interrupt routine and the details depend very much on the hardware. We consider the set of "typical" conditions. First, we check if a SCSI reset was encountered on the bus (probably caused by another SCSI controller on the same SCSI bus). If so we drop all the enqueued and disconnected requests, report the events and re-initialize our SCSI controller. It is important that during this initialization the controller won't issue another reset or else two controllers on the same SCSI bus could ping-pong resets forever. The case of fatal controller error/hang could be handled in the same place, but it will probably need also sending RESET signal to the SCSI bus to reset the status of the connections with the SCSI devices. int fatal=0; struct ccb_trans_settings neg; struct cam_path *path; if( detected_scsi_reset(softc) || (fatal = detected_fatal_controller_error(softc)) ) { int targ, lun; struct xxx_hcb *h, *hh; /* drop all enqueued CCBs */ for(h = softc->first_queued_hcb; h != NULL; h = hh) { hh = h->next; free_hcb_and_ccb_done(h, h->ccb, CAM_SCSI_BUS_RESET); } /* the clean values of negotiations to report */ neg.bus_width = 8; neg.sync_period = neg.sync_offset = 0; neg.valid = (CCB_TRANS_BUS_WIDTH_VALID | CCB_TRANS_SYNC_RATE_VALID | CCB_TRANS_SYNC_OFFSET_VALID); /* drop all disconnected CCBs and clean negotiations */ for(targ=0; targ <= OUR_MAX_SUPPORTED_TARGET; targ++) { clean_negotiations(softc, targ); /* report the event if possible */ if(xpt_create_path(&path, /*periph*/NULL, cam_sim_path(sim), targ, CAM_LUN_WILDCARD) == CAM_REQ_CMP) { xpt_async(AC_TRANSFER_NEG, path, &neg); xpt_free_path(path); } for(lun=0; lun <= OUR_MAX_SUPPORTED_LUN; lun++) for(h = softc->first_discon_hcb[targ][lun]; h != NULL; h = hh) { hh=h->next; if(fatal) free_hcb_and_ccb_done(h, h->ccb, CAM_UNREC_HBA_ERROR); else free_hcb_and_ccb_done(h, h->ccb, CAM_SCSI_BUS_RESET); } } /* report the event */ xpt_async(AC_BUS_RESET, softc->wpath, NULL); /* re-initialization may take a lot of time, in such case * its completion should be signaled by another interrupt or * checked on timeout - but for simplicity we assume here that * it's really fast */ if(!fatal) { reinitialize_controller_without_scsi_reset(softc); } else { reinitialize_controller_with_scsi_reset(softc); } schedule_next_hcb(softc); return; } If interrupt is not caused by a controller-wide condition then probably something has happened to the current hardware control block. Depending on the hardware there may be other non-HCB-related events, we just do not consider them here. Then we analyze what happened to this HCB: struct xxx_hcb *hcb, *h, *hh; int hcb_status, scsi_status; int ccb_status; int targ; int lun_to_freeze; hcb = get_current_hcb(softc); if(hcb == NULL) { /* either stray interrupt or something went very wrong * or this is something hardware-dependent */ handle as necessary; return; } targ = hcb->target; hcb_status = get_status_of_current_hcb(softc); First we check if the HCB has completed and if so we check the returned SCSI status. if(hcb_status == COMPLETED) { scsi_status = get_completion_status(hcb); Then look if this status is related to the REQUEST SENSE command and if so handle it in a simple way. if(hcb->flags & DOING_AUTOSENSE) { if(scsi_status == GOOD) { /* autosense was successful */ hcb->ccb->ccb_h.status |= CAM_AUTOSNS_VALID; free_hcb_and_ccb_done(hcb, hcb->ccb, CAM_SCSI_STATUS_ERROR); } else { autosense_failed: free_hcb_and_ccb_done(hcb, hcb->ccb, CAM_AUTOSENSE_FAIL); } schedule_next_hcb(softc); return; } Else the command itself has completed, pay more attention to details. If auto-sense is not disabled for this CCB and the command has failed with sense data then run REQUEST SENSE command to receive that data. hcb->ccb->csio.scsi_status = scsi_status; calculate_residue(hcb); if( (hcb->ccb->ccb_h.flags & CAM_DIS_AUTOSENSE)==0 && ( scsi_status == CHECK_CONDITION || scsi_status == COMMAND_TERMINATED) ) { /* start auto-SENSE */ hcb->flags |= DOING_AUTOSENSE; setup_autosense_command_in_hcb(hcb); restart_current_hcb(softc); return; } if(scsi_status == GOOD) free_hcb_and_ccb_done(hcb, hcb->ccb, CAM_REQ_CMP); else free_hcb_and_ccb_done(hcb, hcb->ccb, CAM_SCSI_STATUS_ERROR); schedule_next_hcb(softc); return; } One typical thing would be negotiation events: negotiation messages received from a SCSI target (in answer to our negotiation attempt or by target's initiative) or the target is unable to negotiate (rejects our negotiation messages or does not answer them). switch(hcb_status) { case TARGET_REJECTED_WIDE_NEG: /* revert to 8-bit bus */ softc->current_bus_width[targ] = softc->goal_bus_width[targ] = 8; /* report the event */ neg.bus_width = 8; neg.valid = CCB_TRANS_BUS_WIDTH_VALID; xpt_async(AC_TRANSFER_NEG, hcb->ccb.ccb_h.path_id, &neg); continue_current_hcb(softc); return; case TARGET_ANSWERED_WIDE_NEG: { int wd; wd = get_target_bus_width_request(softc); if(wd <= softc->goal_bus_width[targ]) { /* answer is acceptable */ softc->current_bus_width[targ] = softc->goal_bus_width[targ] = neg.bus_width = wd; /* report the event */ neg.valid = CCB_TRANS_BUS_WIDTH_VALID; xpt_async(AC_TRANSFER_NEG, hcb->ccb.ccb_h.path_id, &neg); } else { prepare_reject_message(hcb); } } continue_current_hcb(softc); return; case TARGET_REQUESTED_WIDE_NEG: { int wd; wd = get_target_bus_width_request(softc); wd = min (wd, OUR_BUS_WIDTH); wd = min (wd, softc->user_bus_width[targ]); if(wd != softc->current_bus_width[targ]) { /* the bus width has changed */ softc->current_bus_width[targ] = softc->goal_bus_width[targ] = neg.bus_width = wd; /* report the event */ neg.valid = CCB_TRANS_BUS_WIDTH_VALID; xpt_async(AC_TRANSFER_NEG, hcb->ccb.ccb_h.path_id, &neg); } prepare_width_nego_rsponse(hcb, wd); } continue_current_hcb(softc); return; } Then we handle any errors that could have happened during auto-sense in the same simple-minded way as before. Otherwise we look closer at the details again. if(hcb->flags & DOING_AUTOSENSE) goto autosense_failed; switch(hcb_status) { The next event we consider is unexpected disconnect. Which is considered normal after an ABORT or BUS DEVICE RESET message and abnormal in other cases. case UNEXPECTED_DISCONNECT: if(requested_abort(hcb)) { /* abort affects all commands on that target+LUN, so * mark all disconnected HCBs on that target+LUN as aborted too */ for(h = softc->first_discon_hcb[hcb->target][hcb->lun]; h != NULL; h = hh) { hh=h->next; free_hcb_and_ccb_done(h, h->ccb, CAM_REQ_ABORTED); } ccb_status = CAM_REQ_ABORTED; } else if(requested_bus_device_reset(hcb)) { int lun; /* reset affects all commands on that target, so * mark all disconnected HCBs on that target+LUN as reset */ for(lun=0; lun <= OUR_MAX_SUPPORTED_LUN; lun++) for(h = softc->first_discon_hcb[hcb->target][lun]; h != NULL; h = hh) { hh=h->next; free_hcb_and_ccb_done(h, h->ccb, CAM_SCSI_BUS_RESET); } /* send event */ xpt_async(AC_SENT_BDR, hcb->ccb->ccb_h.path_id, NULL); /* this was the CAM_RESET_DEV request itself, it's completed */ ccb_status = CAM_REQ_CMP; } else { calculate_residue(hcb); ccb_status = CAM_UNEXP_BUSFREE; /* request the further code to freeze the queue */ hcb->ccb->ccb_h.status |= CAM_DEV_QFRZN; lun_to_freeze = hcb->lun; } break; If the target refuses to accept tags we notify CAM about that and return back all commands for this LUN: case TAGS_REJECTED: /* report the event */ neg.flags = 0 & ~CCB_TRANS_TAG_ENB; neg.valid = CCB_TRANS_TQ_VALID; xpt_async(AC_TRANSFER_NEG, hcb->ccb.ccb_h.path_id, &neg); ccb_status = CAM_MSG_REJECT_REC; /* request the further code to freeze the queue */ hcb->ccb->ccb_h.status |= CAM_DEV_QFRZN; lun_to_freeze = hcb->lun; break; Then we check a number of other conditions, with processing basically limited to setting the CCB status: case SELECTION_TIMEOUT: ccb_status = CAM_SEL_TIMEOUT; /* request the further code to freeze the queue */ hcb->ccb->ccb_h.status |= CAM_DEV_QFRZN; lun_to_freeze = CAM_LUN_WILDCARD; break; case PARITY_ERROR: ccb_status = CAM_UNCOR_PARITY; break; case DATA_OVERRUN: case ODD_WIDE_TRANSFER: ccb_status = CAM_DATA_RUN_ERR; break; default: /* all other errors are handled in a generic way */ ccb_status = CAM_REQ_CMP_ERR; /* request the further code to freeze the queue */ hcb->ccb->ccb_h.status |= CAM_DEV_QFRZN; lun_to_freeze = CAM_LUN_WILDCARD; break; } Then we check if the error was serious enough to freeze the input queue until it gets proceeded and do so if it is: if(hcb->ccb->ccb_h.status & CAM_DEV_QFRZN) { /* freeze the queue */ xpt_freeze_devq(ccb->ccb_h.path, /*count*/1); /* re-queue all commands for this target/LUN back to CAM */ for(h = softc->first_queued_hcb; h != NULL; h = hh) { hh = h->next; if(targ == h->targ && (lun_to_freeze == CAM_LUN_WILDCARD || lun_to_freeze == h->lun) ) free_hcb_and_ccb_done(h, h->ccb, CAM_REQUEUE_REQ); } } free_hcb_and_ccb_done(hcb, hcb->ccb, ccb_status); schedule_next_hcb(softc); return; This concludes the generic interrupt handling although specific controllers may require some additions. Errors Summary When executing an I/O request many things may go wrong. The reason of error can be reported in the CCB status with great detail. Examples of use are spread throughout this document. For completeness here is the summary of recommended responses for the typical error conditions: CAM_RESRC_UNAVAIL - some resource is temporarily unavailable and the SIM driver can not generate an event when it will become available. An example of this resource would be some intra-controller hardware resource for which the controller does not generate an interrupt when it becomes available. CAM_UNCOR_PARITY - unrecovered parity error occurred CAM_DATA_RUN_ERR - data overrun or unexpected data phase (going in other direction than specified in CAM_DIR_MASK) or odd transfer length for wide transfer CAM_SEL_TIMEOUT - selection timeout occurred (target does not respond) CAM_CMD_TIMEOUT - command timeout occurred (the timeout function ran) CAM_SCSI_STATUS_ERROR - the device returned error CAM_AUTOSENSE_FAIL - the device returned error and the REQUEST SENSE COMMAND failed CAM_MSG_REJECT_REC - MESSAGE REJECT message was received CAM_SCSI_BUS_RESET - received SCSI bus reset CAM_REQ_CMP_ERR - "impossible" SCSI phase occurred or something else as weird or just a generic error if further detail is not available CAM_UNEXP_BUSFREE - unexpected disconnect occurred CAM_BDR_SENT - BUS DEVICE RESET message was sent to the target CAM_UNREC_HBA_ERROR - unrecoverable Host Bus Adapter Error CAM_REQ_TOO_BIG - the request was too large for this controller CAM_REQUEUE_REQ - this request should be re-queued to preserve transaction ordering. This typically occurs when the SIM recognizes an error that should freeze the queue and must place other queued requests for the target at the sim level back into the XPT queue. Typical cases of such errors are selection timeouts, command timeouts and other like conditions. In such cases the troublesome command returns the status indicating the error, the and the other commands which have not be sent to the bus yet get re-queued. CAM_LUN_INVALID - the LUN ID in the request is not supported by the SCSI controller CAM_TID_INVALID - the target ID in the request is not supported by the SCSI controller Timeout Handling When the timeout for an HCB expires that request should be aborted, just like with an XPT_ABORT request. The only difference is that the returned status of aborted request should be CAM_CMD_TIMEOUT instead of CAM_REQ_ABORTED (that's why implementation of the abort better be done as a function). But there is one more possible problem: what if the abort request itself will get stuck? In this case the SCSI bus should be reset, just like with an XPT_RESET_BUS request (and the idea about implementing it as a function called from both places applies here too). Also we should reset the whole SCSI bus if a device reset request got stuck. So after all the timeout function would look like: static void xxx_timeout(void *arg) { struct xxx_hcb *hcb = (struct xxx_hcb *)arg; struct xxx_softc *softc; struct ccb_hdr *ccb_h; softc = hcb->softc; ccb_h = &hcb->ccb->ccb_h; if(hcb->flags & HCB_BEING_ABORTED || ccb_h->func_code == XPT_RESET_DEV) { xxx_reset_bus(softc); } else { xxx_abort_ccb(hcb->ccb, CAM_CMD_TIMEOUT); } } When we abort a request all the other disconnected requests to the same target/LUN get aborted too. So there appears a question, should we return them with status CAM_REQ_ABORTED or CAM_CMD_TIMEOUT ? The current drivers use CAM_CMD_TIMEOUT. This seems logical because if one request got timed out then probably something really bad is happening to the device, so if they would not be disturbed they would time out by themselves. diff --git a/en_US.ISO8859-1/books/developers-handbook/scsi/chapter.sgml b/en_US.ISO8859-1/books/developers-handbook/scsi/chapter.sgml index 63b8d5475d..72ee58a453 100644 --- a/en_US.ISO8859-1/books/developers-handbook/scsi/chapter.sgml +++ b/en_US.ISO8859-1/books/developers-handbook/scsi/chapter.sgml @@ -1,1983 +1,1983 @@ Common Access Method SCSI Controllers This chapter was written by &a.babkin; Modifications for the handbook made by &a.murray;. Synopsis This document assumes that the reader has a general understanding of device drivers in FreeBSD and of the SCSI protocol. Much of the information in this document was extracted from the drivers : ncr (/sys/pci/ncr.c) by Wolfgang Stanglmeier and Stefan Esser sym (/sys/pci/sym.c) by Gerard Roudier aic7xxx (/sys/dev/aic7xxx/aic7xxx.c) by Justin T. Gibbs and from the CAM code itself (by Justing T. Gibbs, see /sys/cam/*). When some solution looked the most logical and was essentially verbatim extracted from the code by Justin Gibbs, I marked it as "recommended". The document is illustrated with examples in pseudo-code. Although sometimes the examples have many details and look like real code, it's still pseudo-code. It was written to demonstrate the concepts in an understandable way. For a real driver other approaches may be more modular and efficient. It also abstracts from the hardware details, as well as issues that would cloud the demonstrated concepts or that are supposed to be described in the other chapters of the developers handbook. Such details are commonly shown as calls to functions with descriptive names, comments or pseudo-statements. Fortunately real life full-size examples with all the details can be found in the real drivers. General architecture CAM stands for Common Access Method. It's a generic way to address the I/O buses in a SCSI-like way. This allows a separation of the generic device drivers from the drivers controlling the I/O bus: for example the disk driver becomes able to control disks on both SCSI, IDE, and/or any other bus so the disk driver portion does not have to be rewritten (or copied and modified) for every new I/O bus. Thus the two most important active entities are: Peripheral Modules - a driver for peripheral devices (disk, tape, CDROM, etc.) SCSI Interface Modules (SIM) - a Host Bus Adapter drivers for connecting to an I/O bus such as SCSI or IDE. A peripheral driver receives requests from the OS, converts them to a sequence of SCSI commands and passes these SCSI commands to a SCSI Interface Module. The SCSI Interface Module is responsible for passing these commands to the actual hardware (or if the actual hardware is not SCSI but, for example, IDE then also converting the SCSI commands to the native commands of the hardware). Because we are interested in writing a SCSI adapter driver here, from this point on we will consider everything from the SIM standpoint. A typical SIM driver needs to include the following CAM-related header files: #include <cam/cam.h> #include <cam/cam_ccb.h> #include <cam/cam_sim.h> #include <cam/cam_xpt_sim.h> #include <cam/cam_debug.h> #include <cam/scsi/scsi_all.h> The first thing each SIM driver must do is register itself with the CAM subsystem. This is done during the driver's xxx_attach() function (here and further xxx_ is used to denote the unique driver name prefix). The xxx_attach() function itself is called by the system bus auto-configuration code which we don't describe here. This is achieved in multiple steps: first it's necessary to allocate the queue of requests associated with this SIM: struct cam_devq *devq; if(( devq = cam_simq_alloc(SIZE) )==NULL) { error; /* some code to handle the error */ } Here SIZE is the size of the queue to be allocated, maximal number of requests it could contain. It's the number of requests that the SIM driver can handle in parallel on one SCSI card. Commonly it can be calculated as: SIZE = NUMBER_OF_SUPPORTED_TARGETS * MAX_SIMULTANEOUS_COMMANDS_PER_TARGET Next we create a descriptor of our SIM: struct cam_sim *sim; if(( sim = cam_sim_alloc(action_func, poll_func, driver_name, softc, unit, max_dev_transactions, max_tagged_dev_transactions, devq) )==NULL) { cam_simq_free(devq); error; /* some code to handle the error */ } Note that if we are not able to create a SIM descriptor we free the devq also because we can do nothing else with it and we want to conserve memory. If a SCSI card has multiple SCSI buses on it then each bus requires its own cam_sim structure. An interesting question is what to do if a SCSI card has more than one SCSI bus, do we need one devq structure per card or per SCSI bus? The answer given in the comments to the CAM code is: either way, as the driver's author prefers. The arguments are : action_func - pointer to the driver's xxx_action function. static void xxx_action struct cam_sim *sim, union ccb *ccb poll_func - pointer to the driver's xxx_poll() static void xxx_poll struct cam_sim *sim driver_name - the name of the actual driver, such as "ncr" or "wds" softc - pointer to the driver's internal descriptor for this SCSI card. This pointer will be used by the driver in future to get private data. unit - the controller unit number, for example for controller "wds0" this number will be 0 max_dev_transactions - maximal number of simultaneous transactions per SCSI target in the non-tagged mode. This value will be almost universally equal to 1, with possible exceptions only for the non-SCSI cards. Also the drivers that hope to take advantage by preparing one transaction while another one is executed may set it to 2 but this does not seem to be worth the complexity. max_tagged_dev_transactions - the same thing, but in the tagged mode. Tags are the SCSI way to initiate multiple transactions on a device: each transaction is assigned a unique tag and the transaction is sent to the device. When the device completes some transaction it sends back the result together with the tag so that the SCSI adapter (and the driver) can tell which transaction was completed. This argument is also known as the maximal tag depth. It depends on the abilities of the SCSI adapter. Finally we register the SCSI buses associated with our SCSI adapter: if(xpt_bus_register(sim, bus_number) != CAM_SUCCESS) { cam_sim_free(sim, /*free_devq*/ TRUE); error; /* some code to handle the error */ } If there is one devq structure per SCSI bus (i.e. we consider a card with multiple buses as multiple cards with one bus each) then the bus number will always be 0, otherwise each bus on the SCSI card should be get a distinct number. Each bus needs its own separate structure cam_sim. After that our controller is completely hooked to the CAM system. The value of devq can be discarded now: sim will be passed as an argument in all further calls from CAM and devq can be derived from it. CAM provides the framework for such asynchronous events. Some events originate from the lower levels (the SIM drivers), some events originate from the peripheral drivers, some events originate from the CAM subsystem itself. Any driver can register callbacks for some types of the asynchronous events, so that it would be notified if these events occur. A typical example of such an event is a device reset. Each transaction and event identifies the devices to which it applies by the means of "path". The target-specific events normally occur during a transaction with this device. So the path from that transaction may be re-used to report this event (this is safe because the event path is copied in the event reporting routine but not deallocated nor passed anywhere further). Also it's safe to allocate paths dynamically at any time including the interrupt routines, although that incurs certain overhead, and a possible problem with this approach is that there may be no free memory at that time. For a bus reset event we need to define a wildcard path including all devices on the bus. So we can create the path for the future bus reset events in advance and avoid problems with the future memory shortage: struct cam_path *path; if(xpt_create_path(&path, /*periph*/NULL, cam_sim_path(sim), CAM_TARGET_WILDCARD, CAM_LUN_WILDCARD) != CAM_REQ_CMP) { xpt_bus_deregister(cam_sim_path(sim)); cam_sim_free(sim, /*free_devq*/TRUE); error; /* some code to handle the error */ } softc->wpath = path; softc->sim = sim; As you can see the path includes: ID of the peripheral driver (NULL here because we have none) ID of the SIM driver (cam_sim_path(sim)) SCSI target number of the device (CAM_TARGET_WILDCARD means "all devices") SCSI LUN number of the subdevice (CAM_LUN_WILDCARD means "all LUNs") If the driver can't allocate this path it won't be able to work normally, so in that case we dismantle that SCSI bus. And we save the path pointer in the softc structure for future use. After that we save the value of sim (or we can also discard it on the exit from xxx_probe() if we wish). That's all for a minimalistic initialization. To do things right there is one more issue left. For a SIM driver there is one particularly interesting event: when a target device is considered lost. In this case resetting the SCSI negotiations with this device may be a good idea. So we register a callback for this event with CAM. The request is passed to CAM by requesting CAM action on a CAM control block for this type of request: struct ccb_setasync csa; xpt_setup_ccb(&csa.ccb_h, path, /*priority*/5); csa.ccb_h.func_code = XPT_SASYNC_CB; csa.event_enable = AC_LOST_DEVICE; csa.callback = xxx_async; csa.callback_arg = sim; xpt_action((union ccb *)&csa); Now we take a look at the xxx_action() and xxx_poll() driver entry points. static void xxx_action struct cam_sim *sim, union ccb *ccb Do some action on request of the CAM subsystem. Sim describes the SIM for the request, CCB is the request itself. CCB stands for "CAM Control Block". It is a union of many specific instances, each describing arguments for some type of transactions. All of these instances share the CCB header where the common part of arguments is stored. CAM supports the SCSI controllers working in both initiator ("normal") mode and target (simulating a SCSI device) mode. Here we only consider the part relevant to the initiator mode. There are a few function and macros (in other words, methods) defined to access the public data in the struct sim: cam_sim_path(sim) - the path ID (see above) cam_sim_name(sim) - the name of the sim cam_sim_softc(sim) - the pointer to the softc (driver private data) structure cam_sim_unit(sim) - the unit number cam_sim_bus(sim) - the bus ID To identify the device, xxx_action() can get the unit number and pointer to its structure softc using these functions. The type of request is stored in ccb->ccb_h.func_code. So generally xxx_action() consists of a big switch: struct xxx_softc *softc = (struct xxx_softc *) cam_sim_softc(sim); struct ccb_hdr *ccb_h = &ccb->ccb_h; int unit = cam_sim_unit(sim); int bus = cam_sim_bus(sim); switch(ccb_h->func_code) { case ...: ... default: ccb_h->status = CAM_REQ_INVALID; xpt_done(ccb); break; } As can be seen from the default case (if an unknown command was received) the return code of the command is set into ccb->ccb_h.status and the completed CCB is returned back to CAM by calling xpt_done(ccb). xpt_done() does not have to be called from xxx_action(): For example an I/O request may be enqueued inside the SIM driver and/or its SCSI controller. Then when the device would post an interrupt signaling that the processing of this request is complete xpt_done() may be called from the interrupt handling routine. Actually, the CCB status is not only assigned as a return code but a CCB has some status all the time. Before CCB is passed to the xxx_action() routine it gets the status CCB_REQ_INPROG meaning that it's in progress. There are a surprising number of status values defined in /sys/cam/cam.h which should be able to represent the status of a request in great detail. More interesting yet, the status is in fact a "bitwise or" of an enumerated status value (the lower 6 bits) and possible additional flag-like bits (the upper bits). The enumerated values will be discussed later in more detail. The summary of them can be found in the Errors Summary section. The possible status flags are: CAM_DEV_QFRZN - if the SIM driver gets a serious error (for example, the device does not respond to the selection or breaks the SCSI protocol) when processing a CCB it should freeze the request queue by calling xpt_freeze_simq(), return the other enqueued but not processed yet CCBs for this device back to the CAM queue, then set this flag for the troublesome CCB and call xpt_done(). This flag causes the CAM subsystem to unfreeze the queue after it handles the error. CAM_AUTOSNS_VALID - if the device returned an error condition and the flag CAM_DIS_AUTOSENSE is not set in CCB the SIM driver must execute the REQUEST SENSE command automatically to extract the sense (extended error information) data from the device. If this attempt was successful the sense data should be saved in the CCB and this flag set. CAM_RELEASE_SIMQ - like CAM_DEV_QFRZN but used in case there is some problem (or resource shortage) with the SCSI controller itself. Then all the future requests to the controller should be stopped by xpt_freeze_simq(). The controller queue will be restarted after the SIM driver overcomes the shortage and informs CAM by returning some CCB with this flag set. CAM_SIM_QUEUED - when SIM puts a CCB into its request queue this flag should be set (and removed when this CCB gets dequeued before being returned back to CAM). This flag is not used anywhere in the CAM code now, so its purpose is purely diagnostic. The function xxx_action() is not allowed to sleep, so all the synchronization for resource access must be done using SIM or device queue freezing. Besides the aforementioned flags the CAM subsystem provides functions xpt_selease_simq() and xpt_release_devq() to unfreeze the queues directly, without passing a CCB to CAM. The CCB header contains the following fields: path - path ID for the request target_id - target device ID for the request target_lun - LUN ID of the target device timeout - timeout interval for this command, in milliseconds timeout_ch - a convenience place for the SIM driver to store the timeout handle (the CAM subsystem itself does not make any assumptions about it) flags - various bits of information about the request spriv_ptr0, spriv_ptr1 - fields reserved for private use by the SIM driver (such as linking to the SIM queues or SIM private control blocks); actually, they exist as unions: spriv_ptr0 and spriv_ptr1 have the type (void *), spriv_field0 and spriv_field1 have the type unsigned long, sim_priv.entries[0].bytes and sim_priv.entries[1].bytes are byte arrays of the size consistent with the other incarnations of the union and sim_priv.bytes is one array, twice bigger. The recommended way of using the SIM private fields of CCB is to define some meaningful names for them and use these meaningful names in the driver, like: #define ccb_some_meaningful_name sim_priv.entries[0].bytes #define ccb_hcb spriv_ptr1 /* for hardware control block */ The most common initiator mode requests are: XPT_SCSI_IO - execute an I/O transaction The instance "struct ccb_scsiio csio" of the union ccb is used to transfer the arguments. They are: cdb_io - pointer to the SCSI command buffer or the buffer itself cdb_len - SCSI command length data_ptr - pointer to the data buffer (gets a bit complicated if scatter/gather is used) dxfer_len - length of the data to transfer sglist_cnt - counter of the scatter/gather segments scsi_status - place to return the SCSI status sense_data - buffer for the SCSI sense information if the command returns an error (the SIM driver is supposed to run the REQUEST SENSE command automatically in this case if the CCB flag CAM_DIS_AUTOSENSE is not set) sense_len - the length of that buffer (if it happens to be higher than size of sense_data the SIM driver must silently assume the smaller value) resid, sense_resid - if the transfer of data or SCSI sense returned an error these are the returned counters of the residual (not transferred) data. They do not seem to be especially meaningful, so in a case when they are difficult to compute (say, counting bytes in the SCSI controller's FIFO buffer) an approximate value will do as well. For a successfully completed transfer they must be set to zero. tag_action - the kind of tag to use: CAM_TAG_ACTION_NONE - don't use tags for this transaction MSG_SIMPLE_Q_TAG, MSG_HEAD_OF_Q_TAG, MSG_ORDERED_Q_TAG - value equal to the appropriate tag message (see /sys/cam/scsi/scsi_message.h); this gives only the tag type, the SIM driver must assign the tag value itself The general logic of handling this request is the following: The first thing to do is to check for possible races, to make sure that the command did not get aborted when it was sitting in the queue: struct ccb_scsiio *csio = &ccb->csio; if ((ccb_h->status & CAM_STATUS_MASK) != CAM_REQ_INPROG) { xpt_done(ccb); return; } Also we check that the device is supported at all by our controller: if(ccb_h->target_id > OUR_MAX_SUPPORTED_TARGET_ID || cch_h->target_id == OUR_SCSI_CONTROLLERS_OWN_ID) { ccb_h->status = CAM_TID_INVALID; xpt_done(ccb); return; } if(ccb_h->target_lun > OUR_MAX_SUPPORTED_LUN) { ccb_h->status = CAM_LUN_INVALID; xpt_done(ccb); return; } Then allocate whatever data structures (such as card-dependent hardware control block) we need to process this request. If we can't then freeze the SIM queue and remember that we have a pending operation, return the CCB back and ask CAM to re-queue it. Later when the resources become available the SIM queue must be unfrozen by returning a ccb with the CAM_SIMQ_RELEASE bit set in its status. Otherwise, if all went well, link the CCB with the hardware control block (HCB) and mark it as queued. struct xxx_hcb *hcb = allocate_hcb(softc, unit, bus); if(hcb == NULL) { softc->flags |= RESOURCE_SHORTAGE; xpt_freeze_simq(sim, /*count*/1); ccb_h->status = CAM_REQUEUE_REQ; xpt_done(ccb); return; } hcb->ccb = ccb; ccb_h->ccb_hcb = (void *)hcb; ccb_h->status |= CAM_SIM_QUEUED; Extract the target data from CCB into the hardware control block. Check if we are asked to assign a tag and if yes then generate an unique tag and build the SCSI tag messages. The SIM driver is also responsible for negotiations with the devices to set the maximal mutually supported bus width, synchronous rate and offset. hcb->target = ccb_h->target_id; hcb->lun = ccb_h->target_lun; generate_identify_message(hcb); if( ccb_h->tag_action != CAM_TAG_ACTION_NONE ) generate_unique_tag_message(hcb, ccb_h->tag_action); if( !target_negotiated(hcb) ) generate_negotiation_messages(hcb); Then set up the SCSI command. The command storage may be specified in the CCB in many interesting ways, specified by the CCB flags. The command buffer can be contained in CCB or pointed to, in the latter case the pointer may be physical or virtual. Since the hardware commonly needs physical address we always convert the address to the physical one. A NOT-QUITE RELATED NOTE: Normally this is done by a call to vtophys(), but for the PCI device (which account for most of the SCSI controllers now) drivers' portability to the Alpha architecture the conversion must be done by vtobus() instead due to special Alpha quirks. [IMHO it would be much better to have two separate functions, vtop() and ptobus() then vtobus() would be a simple superposition of them.] In case if a physical address is requested it's OK to return the CCB with the status CAM_REQ_INVALID, the current drivers do that. But it's also possible to compile the Alpha-specific piece of code, as in this example (there should be a more direct way to do that, without conditional compilation in the drivers). If necessary a physical address can be also converted or mapped back to a virtual address but with big pain, so we don't do that. if(ccb_h->flags & CAM_CDB_POINTER) { /* CDB is a pointer */ if(!(ccb_h->flags & CAM_CDB_PHYS)) { /* CDB pointer is virtual */ hcb->cmd = vtobus(csio->cdb_io.cdb_ptr); } else { /* CDB pointer is physical */ #if defined(__alpha__) hcb->cmd = csio->cdb_io.cdb_ptr | alpha_XXX_dmamap_or ; #else hcb->cmd = csio->cdb_io.cdb_ptr ; #endif } } else { /* CDB is in the ccb (buffer) */ hcb->cmd = vtobus(csio->cdb_io.cdb_bytes); } hcb->cmdlen = csio->cdb_len; Now it's time to set up the data. Again, the data storage may be specified in the CCB in many interesting ways, specified by the CCB flags. First we get the direction of the data transfer. The simplest case is if there is no data to transfer: int dir = (ccb_h->flags & CAM_DIR_MASK); if (dir == CAM_DIR_NONE) goto end_data; Then we check if the data is in one chunk or in a scatter-gather list, and the addresses are physical or virtual. The SCSI controller may be able to handle only a limited number of chunks of limited length. If the request hits this limitation we return an error. We use a special function to return the CCB to handle in one place the HCB resource shortages. The functions to add chunks are driver-dependent, and here we leave them without detailed implementation. See description of the SCSI command (CDB) handling for the details on the address-translation issues. If some variation is too difficult or impossible to implement with a particular card it's OK to return the status CAM_REQ_INVALID. Actually, it seems like the scatter-gather ability is not used anywhere in the CAM code now. But at least the case for a single non-scattered virtual buffer must be implemented, it's actively used by CAM. int rv; initialize_hcb_for_data(hcb); if((!(ccb_h->flags & CAM_SCATTER_VALID)) { /* single buffer */ if(!(ccb_h->flags & CAM_DATA_PHYS)) { rv = add_virtual_chunk(hcb, csio->data_ptr, csio->dxfer_len, dir); } } else { rv = add_physical_chunk(hcb, csio->data_ptr, csio->dxfer_len, dir); } } else { int i; struct bus_dma_segment *segs; segs = (struct bus_dma_segment *)csio->data_ptr; if ((ccb_h->flags & CAM_SG_LIST_PHYS) != 0) { /* The SG list pointer is physical */ rv = setup_hcb_for_physical_sg_list(hcb, segs, csio->sglist_cnt); } else if (!(ccb_h->flags & CAM_DATA_PHYS)) { /* SG buffer pointers are virtual */ for (i = 0; i < csio->sglist_cnt; i++) { rv = add_virtual_chunk(hcb, segs[i].ds_addr, segs[i].ds_len, dir); if (rv != CAM_REQ_CMP) break; } } else { /* SG buffer pointers are physical */ for (i = 0; i < csio->sglist_cnt; i++) { rv = add_physical_chunk(hcb, segs[i].ds_addr, segs[i].ds_len, dir); if (rv != CAM_REQ_CMP) break; } } } if(rv != CAM_REQ_CMP) { /* we expect that add_*_chunk() functions return CAM_REQ_CMP * if they added a chunk successfully, CAM_REQ_TOO_BIG if * the request is too big (too many bytes or too many chunks), * CAM_REQ_INVALID in case of other troubles */ free_hcb_and_ccb_done(hcb, ccb, rv); return; } end_data: If disconnection is disabled for this CCB we pass this information to the hcb: if(ccb_h->flags & CAM_DIS_DISCONNECT) hcb_disable_disconnect(hcb); If the controller is able to run REQUEST SENSE command all by itself then the value of the flag CAM_DIS_AUTOSENSE should also be passed to it, to prevent automatic REQUEST SENSE if the CAM subsystem does not want it. The only thing left is to set up the timeout, pass our hcb to the hardware and return, the rest will be done by the interrupt handler (or timeout handler). ccb_h->timeout_ch = timeout(xxx_timeout, (caddr_t) hcb, (ccb_h->timeout * hz) / 1000); /* convert milliseconds to ticks */ put_hcb_into_hardware_queue(hcb); return; And here is a possible implementation of the function returning CCB: static void free_hcb_and_ccb_done(struct xxx_hcb *hcb, union ccb *ccb, u_int32_t status) { struct xxx_softc *softc = hcb->softc; ccb->ccb_h.ccb_hcb = 0; if(hcb != NULL) { untimeout(xxx_timeout, (caddr_t) hcb, ccb->ccb_h.timeout_ch); /* we're about to free a hcb, so the shortage has ended */ if(softc->flags & RESOURCE_SHORTAGE) { softc->flags &= ~RESOURCE_SHORTAGE; status |= CAM_RELEASE_SIMQ; } free_hcb(hcb); /* also removes hcb from any internal lists */ } ccb->ccb_h.status = status | (ccb->ccb_h.status & ~(CAM_STATUS_MASK|CAM_SIM_QUEUED)); xpt_done(ccb); } XPT_RESET_DEV - send the SCSI "BUS DEVICE RESET" message to a device There is no data transferred in CCB except the header and the most interesting argument of it is target_id. Depending on the controller hardware a hardware control block just like for the XPT_SCSI_IO request may be constructed (see XPT_SCSI_IO request description) and sent to the controller or the SCSI controller may be immediately programmed to send this RESET message to the device or this request may be just not supported (and return the status CAM_REQ_INVALID). Also on completion of the request all the disconnected transactions for this target must be aborted (probably in the interrupt routine). Also all the current negotiations for the target are lost on reset, so they might be cleaned too. Or they clearing may be deferred, because anyway the target would request re-negotiation on the next transaction. XPT_RESET_BUS - send the RESET signal to the SCSI bus No arguments are passed in the CCB, the only interesting argument is the SCSI bus indicated by the struct sim pointer. A minimalistic implementation would forget the SCSI negotiations for all the devices on the bus and return the status CAM_REQ_CMP. The proper implementation would in addition actually reset the SCSI bus (possible also reset the SCSI controller) and mark all the CCBs being processed, both those in the hardware queue and those being disconnected, as done with the status CAM_SCSI_BUS_RESET. Like: int targ, lun; struct xxx_hcb *h, *hh; struct ccb_trans_settings neg; struct cam_path *path; /* The SCSI bus reset may take a long time, in this case its completion * should be checked by interrupt or timeout. But for simplicity * we assume here that it's really fast. */ reset_scsi_bus(softc); /* drop all enqueued CCBs */ for(h = softc->first_queued_hcb; h != NULL; h = hh) { hh = h->next; free_hcb_and_ccb_done(h, h->ccb, CAM_SCSI_BUS_RESET); } /* the clean values of negotiations to report */ neg.bus_width = 8; neg.sync_period = neg.sync_offset = 0; neg.valid = (CCB_TRANS_BUS_WIDTH_VALID | CCB_TRANS_SYNC_RATE_VALID | CCB_TRANS_SYNC_OFFSET_VALID); /* drop all disconnected CCBs and clean negotiations */ for(targ=0; targ <= OUR_MAX_SUPPORTED_TARGET; targ++) { clean_negotiations(softc, targ); /* report the event if possible */ if(xpt_create_path(&path, /*periph*/NULL, cam_sim_path(sim), targ, CAM_LUN_WILDCARD) == CAM_REQ_CMP) { xpt_async(AC_TRANSFER_NEG, path, &neg); xpt_free_path(path); } for(lun=0; lun <= OUR_MAX_SUPPORTED_LUN; lun++) for(h = softc->first_discon_hcb[targ][lun]; h != NULL; h = hh) { hh=h->next; free_hcb_and_ccb_done(h, h->ccb, CAM_SCSI_BUS_RESET); } } ccb->ccb_h.status = CAM_REQ_CMP; xpt_done(ccb); /* report the event */ xpt_async(AC_BUS_RESET, softc->wpath, NULL); return; Implementing the SCSI bus reset as a function may be a good idea because it would be re-used by the timeout function as a last resort if the things go wrong. XPT_ABORT - abort the specified CCB The arguments are transferred in the instance "struct ccb_abort cab" of the union ccb. The only argument field in it is: abort_ccb - pointer to the CCB to be aborted If the abort is not supported just return the status CAM_UA_ABORT. This is also the easy way to minimally implement this call, return CAM_UA_ABORT in any case. The hard way is to implement this request honestly. First check that abort applies to a SCSI transaction: struct ccb *abort_ccb; abort_ccb = ccb->cab.abort_ccb; if(abort_ccb->ccb_h.func_code != XPT_SCSI_IO) { ccb->ccb_h.status = CAM_UA_ABORT; xpt_done(ccb); return; } Then it's necessary to find this CCB in our queue. This can be done by walking the list of all our hardware control blocks in search for one associated with this CCB: struct xxx_hcb *hcb, *h; hcb = NULL; /* We assume that softc->first_hcb is the head of the list of all * HCBs associated with this bus, including those enqueued for * processing, being processed by hardware and disconnected ones. */ for(h = softc->first_hcb; h != NULL; h = h->next) { if(h->ccb == abort_ccb) { hcb = h; break; } } if(hcb == NULL) { /* no such CCB in our queue */ ccb->ccb_h.status = CAM_PATH_INVALID; xpt_done(ccb); return; } hcb=found_hcb; Now we look at the current processing status of the HCB. It may be either sitting in the queue waiting to be sent to the SCSI bus, being transferred right now, or disconnected and waiting for the result of the command, or actually completed by hardware but not yet marked as done by software. To make sure that we don't get in any races with hardware we mark the HCB as being aborted, so that if this HCB is about to be sent to the SCSI bus the SCSI controller will see this flag and skip it. int hstatus; /* shown as a function, in case special action is needed to make * this flag visible to hardware */ set_hcb_flags(hcb, HCB_BEING_ABORTED); abort_again: hstatus = get_hcb_status(hcb); switch(hstatus) { case HCB_SITTING_IN_QUEUE: remove_hcb_from_hardware_queue(hcb); /* FALLTHROUGH */ case HCB_COMPLETED: /* this is an easy case */ free_hcb_and_ccb_done(hcb, abort_ccb, CAM_REQ_ABORTED); break; If the CCB is being transferred right now we would like to signal to the SCSI controller in some hardware-dependent way that we want to abort the current transfer. The SCSI controller would set the SCSI ATTENTION signal and when the target responds to it send an ABORT message. We also reset the timeout to make sure that the target is not sleeping forever. If the command would not get aborted in some reasonable time like 10 seconds the timeout routine would go ahead and reset the whole SCSI bus. Because the command will be aborted in some reasonable time we can just return the abort request now as successfully completed, and mark the aborted CCB as aborted (but not mark it as done yet). case HCB_BEING_TRANSFERRED: untimeout(xxx_timeout, (caddr_t) hcb, abort_ccb->ccb_h.timeout_ch); abort_ccb->ccb_h.timeout_ch = timeout(xxx_timeout, (caddr_t) hcb, 10 * hz); abort_ccb->ccb_h.status = CAM_REQ_ABORTED; /* ask the controller to abort that HCB, then generate * an interrupt and stop */ if(signal_hardware_to_abort_hcb_and_stop(hcb) < 0) { /* oops, we missed the race with hardware, this transaction * got off the bus before we aborted it, try again */ goto abort_again; } break; If the CCB is in the list of disconnected then set it up as an abort request and re-queue it at the front of hardware queue. Reset the timeout and report the abort request to be completed. case HCB_DISCONNECTED: untimeout(xxx_timeout, (caddr_t) hcb, abort_ccb->ccb_h.timeout_ch); abort_ccb->ccb_h.timeout_ch = timeout(xxx_timeout, (caddr_t) hcb, 10 * hz); put_abort_message_into_hcb(hcb); put_hcb_at_the_front_of_hardware_queue(hcb); break; } ccb->ccb_h.status = CAM_REQ_CMP; xpt_done(ccb); return; That's all for the ABORT request, although there is one more issue. Because the ABORT message cleans all the ongoing transactions on a LUN we have to mark all the other active transactions on this LUN as aborted. That should be done in the interrupt routine, after the transaction gets aborted. Implementing the CCB abort as a function may be quite a good idea, this function can be re-used if an I/O transaction times out. The only difference would be that the timed out transaction would return the status CAM_CMD_TIMEOUT for the timed out request. Then the case XPT_ABORT would be small, like that: case XPT_ABORT: struct ccb *abort_ccb; abort_ccb = ccb->cab.abort_ccb; if(abort_ccb->ccb_h.func_code != XPT_SCSI_IO) { ccb->ccb_h.status = CAM_UA_ABORT; xpt_done(ccb); return; } if(xxx_abort_ccb(abort_ccb, CAM_REQ_ABORTED) < 0) /* no such CCB in our queue */ ccb->ccb_h.status = CAM_PATH_INVALID; else ccb->ccb_h.status = CAM_REQ_CMP; xpt_done(ccb); return; XPT_SET_TRAN_SETTINGS - explicitly set values of SCSI transfer settings The arguments are transferred in the instance "struct ccb_trans_setting cts" of the union ccb: valid - a bitmask showing which settings should be updated: CCB_TRANS_SYNC_RATE_VALID - synchronous transfer rate CCB_TRANS_SYNC_OFFSET_VALID - synchronous offset CCB_TRANS_BUS_WIDTH_VALID - bus width CCB_TRANS_DISC_VALID - set enable/disable disconnection CCB_TRANS_TQ_VALID - set enable/disable tagged queuing flags - consists of two parts, binary arguments and identification of sub-operations. The binary arguments are : CCB_TRANS_DISC_ENB - enable disconnection CCB_TRANS_TAG_ENB - enable tagged queuing the sub-operations are: CCB_TRANS_CURRENT_SETTINGS - change the current negotiations CCB_TRANS_USER_SETTINGS - remember the desired user values sync_period, sync_offset - self-explanatory, if sync_offset==0 then the asynchronous mode is requested bus_width - bus width, in bits (not bytes) Two sets of negotiated parameters are supported, the user settings and the current settings. The user settings are not really used much in the SIM drivers, this is mostly just a piece of memory where the upper levels can store (and later recall) its ideas about the parameters. Setting the user parameters does not cause re-negotiation of the transfer rates. But when the SCSI controller does a negotiation it must never set the values higher than the user parameters, so it's essentially the top boundary. The current settings are, as the name says, current. Changing them means that the parameters must be re-negotiated on the next transfer. Again, these "new current settings" are not supposed to be forced on the device, just they are used as the initial step of negotiations. Also they must be limited by actual capabilities of the SCSI controller: for example, if the SCSI controller has 8-bit bus and the request asks to set 16-bit wide transfers this parameter must be silently truncated to 8-bit transfers before sending it to the device. One caveat is that the bus width and synchronous parameters are per target while the disconnection and tag enabling parameters are per lun. The recommended implementation is to keep 3 sets of negotiated (bus width and synchronous transfer) parameters: user - the user set, as above current - those actually in effect goal - those requested by setting of the "current" parameters The code looks like: struct ccb_trans_settings *cts; int targ, lun; int flags; cts = &ccb->cts; targ = ccb_h->target_id; lun = ccb_h->target_lun; flags = cts->flags; if(flags & CCB_TRANS_USER_SETTINGS) { if(flags & CCB_TRANS_SYNC_RATE_VALID) softc->user_sync_period[targ] = cts->sync_period; if(flags & CCB_TRANS_SYNC_OFFSET_VALID) softc->user_sync_offset[targ] = cts->sync_offset; if(flags & CCB_TRANS_BUS_WIDTH_VALID) softc->user_bus_width[targ] = cts->bus_width; if(flags & CCB_TRANS_DISC_VALID) { softc->user_tflags[targ][lun] &= ~CCB_TRANS_DISC_ENB; softc->user_tflags[targ][lun] |= flags & CCB_TRANS_DISC_ENB; } if(flags & CCB_TRANS_TQ_VALID) { softc->user_tflags[targ][lun] &= ~CCB_TRANS_TQ_ENB; softc->user_tflags[targ][lun] |= flags & CCB_TRANS_TQ_ENB; } } if(flags & CCB_TRANS_CURRENT_SETTINGS) { if(flags & CCB_TRANS_SYNC_RATE_VALID) softc->goal_sync_period[targ] = max(cts->sync_period, OUR_MIN_SUPPORTED_PERIOD); if(flags & CCB_TRANS_SYNC_OFFSET_VALID) softc->goal_sync_offset[targ] = min(cts->sync_offset, OUR_MAX_SUPPORTED_OFFSET); if(flags & CCB_TRANS_BUS_WIDTH_VALID) softc->goal_bus_width[targ] = min(cts->bus_width, OUR_BUS_WIDTH); if(flags & CCB_TRANS_DISC_VALID) { softc->current_tflags[targ][lun] &= ~CCB_TRANS_DISC_ENB; softc->current_tflags[targ][lun] |= flags & CCB_TRANS_DISC_ENB; } if(flags & CCB_TRANS_TQ_VALID) { softc->current_tflags[targ][lun] &= ~CCB_TRANS_TQ_ENB; softc->current_tflags[targ][lun] |= flags & CCB_TRANS_TQ_ENB; } } ccb->ccb_h.status = CAM_REQ_CMP; xpt_done(ccb); return; Then when the next I/O request will be processed it will check if it has to re-negotiate, for example by calling the function target_negotiated(hcb). It can be implemented like this: int target_negotiated(struct xxx_hcb *hcb) { struct softc *softc = hcb->softc; int targ = hcb->targ; if( softc->current_sync_period[targ] != softc->goal_sync_period[targ] || softc->current_sync_offset[targ] != softc->goal_sync_offset[targ] || softc->current_bus_width[targ] != softc->goal_bus_width[targ] ) return 0; /* FALSE */ else return 1; /* TRUE */ } After the values are re-negotiated the resulting values must be assigned to both current and goal parameters, so for future I/O transactions the current and goal parameters would be the same and target_negotiated() would return TRUE. When the card is initialized (in xxx_attach()) the current negotiation values must be initialized to narrow asynchronous mode, the goal and current values must be initialized to the maximal values supported by controller. XPT_GET_TRAN_SETTINGS - get values of SCSI transfer settings This operations is the reverse of XPT_SET_TRAN_SETTINGS. Fill up the CCB instance "struct ccb_trans_setting cts" with data as requested by the flags CCB_TRANS_CURRENT_SETTINGS or CCB_TRANS_USER_SETTINGS (if both are set then the existing drivers return the current settings). Set all the bits in the valid field. XPT_CALC_GEOMETRY - calculate logical (BIOS) geometry of the disk The arguments are transferred in the instance "struct ccb_calc_geometry ccg" of the union ccb: block_size - input, block (A.K.A sector) size in bytes volume_size - input, volume size in bytes cylinders - output, logical cylinders heads - output, logical heads secs_per_track - output, logical sectors per track If the returned geometry differs much enough from what the SCSI controller BIOS thinks and a disk on this SCSI controller is used as bootable the system may not be able to boot. The typical calculation example taken from the aic7xxx driver is: struct ccb_calc_geometry *ccg; u_int32_t size_mb; u_int32_t secs_per_cylinder; int extended; ccg = &ccb->ccg; size_mb = ccg->volume_size / ((1024L * 1024L) / ccg->block_size); extended = check_cards_EEPROM_for_extended_geometry(softc); if (size_mb > 1024 && extended) { ccg->heads = 255; ccg->secs_per_track = 63; } else { ccg->heads = 64; ccg->secs_per_track = 32; } secs_per_cylinder = ccg->heads * ccg->secs_per_track; ccg->cylinders = ccg->volume_size / secs_per_cylinder; ccb->ccb_h.status = CAM_REQ_CMP; xpt_done(ccb); return; This gives the general idea, the exact calculation depends on the quirks of the particular BIOS. If BIOS provides no way set the "extended translation" flag in EEPROM this flag should normally be assumed equal to 1. Other popular geometries are: 128 heads, 63 sectors - Symbios controllers 16 heads, 63 sectors - old controllers Some system BIOSes and SCSI BIOSes fight with each other with variable success, for example a combination of Symbios 875/895 SCSI and Phoenix BIOS can give geometry 128/63 after power up and 255/63 after a hard reset or soft reboot. XPT_PATH_INQ - path inquiry, in other words get the SIM driver and SCSI controller (also known as HBA - Host Bus Adapter) properties The properties are returned in the instance "struct ccb_pathinq cpi" of the union ccb: version_num - the SIM driver version number, now all drivers use 1 hba_inquiry - bitmask of features supported by the controller: PI_MDP_ABLE - supports MDP message (something from SCSI3?) PI_WIDE_32 - supports 32 bit wide SCSI PI_WIDE_16 - supports 16 bit wide SCSI PI_SDTR_ABLE - can negotiate synchronous transfer rate PI_LINKED_CDB - supports linked commands PI_TAG_ABLE - supports tagged commands PI_SOFT_RST - supports soft reset alternative (hard reset and soft reset are mutually exclusive within a SCSI bus) target_sprt - flags for target mode support, 0 if unsupported hba_misc - miscellaneous controller features: PIM_SCANHILO - bus scans from high ID to low ID PIM_NOREMOVE - removable devices not included in scan PIM_NOINITIATOR - initiator role not supported PIM_NOBUSRESET - user has disabled initial BUS RESET hba_eng_cnt - mysterious HBA engine count, something related to compression, now is always set to 0 vuhba_flags - vendor-unique flags, unused now max_target - maximal supported target ID (7 for 8-bit bus, 15 for 16-bit bus, 127 for Fibre Channel) max_lun - maximal supported LUN ID (7 for older SCSI controllers, 63 for newer ones) async_flags - bitmask of installed Async handler, unused now hpath_id - highest Path ID in the subsystem, unused now unit_number - the controller unit number, cam_sim_unit(sim) bus_id - the bus number, cam_sim_bus(sim) initiator_id - the SCSI ID of the controller itself base_transfer_speed - nominal transfer speed in KB/s for asynchronous narrow transfers, equals to 3300 for SCSI sim_vid - SIM driver's vendor id, a zero-terminated string of maximal length SIM_IDLEN including the terminating zero hba_vid - SCSI controller's vendor id, a zero-terminated string of maximal length HBA_IDLEN including the terminating zero dev_name - device driver name, a zero-terminated string of maximal length DEV_IDLEN including the terminating zero, equal to cam_sim_name(sim) The recommended way of setting the string fields is using strncpy, like: strncpy(cpi->dev_name, cam_sim_name(sim), DEV_IDLEN); After setting the values set the status to CAM_REQ_CMP and mark the CCB as done. Polling static void xxx_poll struct cam_sim *sim The poll function is used to simulate the interrupts when the interrupt subsystem is not functioning (for example, when the system has crashed and is creating the system dump). The CAM subsystem sets the proper interrupt level before calling the poll routine. So all it needs to do is to call the interrupt routine (or the other way around, the poll routine may be doing the real action and the interrupt routine would just call the poll routine). Why bother about a separate function then ? Because of different calling conventions. The xxx_poll routine gets the struct cam_sim pointer as its argument when the PCI interrupt routine by common convention gets pointer to the struct xxx_softc and the ISA interrupt routine - gets just the the device unit number. So the poll routine would + gets just the device unit number. So the poll routine would normally look as: static void xxx_poll(struct cam_sim *sim) { xxx_intr((struct xxx_softc *)cam_sim_softc(sim)); /* for PCI device */ } or static void xxx_poll(struct cam_sim *sim) { xxx_intr(cam_sim_unit(sim)); /* for ISA device */ } Asynchronous Events If an asynchronous event callback has been set up then the callback function should be defined. static void ahc_async(void *callback_arg, u_int32_t code, struct cam_path *path, void *arg) callback_arg - the value supplied when registering the callback code - identifies the type of event path - identifies the devices to which the event applies arg - event-specific argument Implementation for a single type of event, AC_LOST_DEVICE, looks like: struct xxx_softc *softc; struct cam_sim *sim; int targ; struct ccb_trans_settings neg; sim = (struct cam_sim *)callback_arg; softc = (struct xxx_softc *)cam_sim_softc(sim); switch (code) { case AC_LOST_DEVICE: targ = xpt_path_target_id(path); if(targ <= OUR_MAX_SUPPORTED_TARGET) { clean_negotiations(softc, targ); /* send indication to CAM */ neg.bus_width = 8; neg.sync_period = neg.sync_offset = 0; neg.valid = (CCB_TRANS_BUS_WIDTH_VALID | CCB_TRANS_SYNC_RATE_VALID | CCB_TRANS_SYNC_OFFSET_VALID); xpt_async(AC_TRANSFER_NEG, path, &neg); } break; default: break; } Interrupts The exact type of the interrupt routine depends on the type of the peripheral bus (PCI, ISA and so on) to which the SCSI controller is connected. The interrupt routines of the SIM drivers run at the interrupt level splcam. So splcam() should be used in the driver to synchronize activity between the interrupt routine and the rest of the driver (for a multiprocessor-aware driver things get yet more interesting but we ignore this case here). The pseudo-code in this document happily ignores the problems of synchronization. The real code must not ignore them. A simple-minded approach is to set splcam() on the entry to the other routines and reset it on return thus protecting them by one big critical section. To make sure that the interrupt level will be always restored a wrapper function can be defined, like: static void xxx_action(struct cam_sim *sim, union ccb *ccb) { int s; s = splcam(); xxx_action1(sim, ccb); splx(s); } static void xxx_action1(struct cam_sim *sim, union ccb *ccb) { ... process the request ... } This approach is simple and robust but the problem with it is that interrupts may get blocked for a relatively long time and this would negatively affect the system's performance. On the other hand the functions of the spl() family have rather high overhead, so vast amount of tiny critical sections may not be good either. The conditions handled by the interrupt routine and the details depend very much on the hardware. We consider the set of "typical" conditions. First, we check if a SCSI reset was encountered on the bus (probably caused by another SCSI controller on the same SCSI bus). If so we drop all the enqueued and disconnected requests, report the events and re-initialize our SCSI controller. It is important that during this initialization the controller won't issue another reset or else two controllers on the same SCSI bus could ping-pong resets forever. The case of fatal controller error/hang could be handled in the same place, but it will probably need also sending RESET signal to the SCSI bus to reset the status of the connections with the SCSI devices. int fatal=0; struct ccb_trans_settings neg; struct cam_path *path; if( detected_scsi_reset(softc) || (fatal = detected_fatal_controller_error(softc)) ) { int targ, lun; struct xxx_hcb *h, *hh; /* drop all enqueued CCBs */ for(h = softc->first_queued_hcb; h != NULL; h = hh) { hh = h->next; free_hcb_and_ccb_done(h, h->ccb, CAM_SCSI_BUS_RESET); } /* the clean values of negotiations to report */ neg.bus_width = 8; neg.sync_period = neg.sync_offset = 0; neg.valid = (CCB_TRANS_BUS_WIDTH_VALID | CCB_TRANS_SYNC_RATE_VALID | CCB_TRANS_SYNC_OFFSET_VALID); /* drop all disconnected CCBs and clean negotiations */ for(targ=0; targ <= OUR_MAX_SUPPORTED_TARGET; targ++) { clean_negotiations(softc, targ); /* report the event if possible */ if(xpt_create_path(&path, /*periph*/NULL, cam_sim_path(sim), targ, CAM_LUN_WILDCARD) == CAM_REQ_CMP) { xpt_async(AC_TRANSFER_NEG, path, &neg); xpt_free_path(path); } for(lun=0; lun <= OUR_MAX_SUPPORTED_LUN; lun++) for(h = softc->first_discon_hcb[targ][lun]; h != NULL; h = hh) { hh=h->next; if(fatal) free_hcb_and_ccb_done(h, h->ccb, CAM_UNREC_HBA_ERROR); else free_hcb_and_ccb_done(h, h->ccb, CAM_SCSI_BUS_RESET); } } /* report the event */ xpt_async(AC_BUS_RESET, softc->wpath, NULL); /* re-initialization may take a lot of time, in such case * its completion should be signaled by another interrupt or * checked on timeout - but for simplicity we assume here that * it's really fast */ if(!fatal) { reinitialize_controller_without_scsi_reset(softc); } else { reinitialize_controller_with_scsi_reset(softc); } schedule_next_hcb(softc); return; } If interrupt is not caused by a controller-wide condition then probably something has happened to the current hardware control block. Depending on the hardware there may be other non-HCB-related events, we just do not consider them here. Then we analyze what happened to this HCB: struct xxx_hcb *hcb, *h, *hh; int hcb_status, scsi_status; int ccb_status; int targ; int lun_to_freeze; hcb = get_current_hcb(softc); if(hcb == NULL) { /* either stray interrupt or something went very wrong * or this is something hardware-dependent */ handle as necessary; return; } targ = hcb->target; hcb_status = get_status_of_current_hcb(softc); First we check if the HCB has completed and if so we check the returned SCSI status. if(hcb_status == COMPLETED) { scsi_status = get_completion_status(hcb); Then look if this status is related to the REQUEST SENSE command and if so handle it in a simple way. if(hcb->flags & DOING_AUTOSENSE) { if(scsi_status == GOOD) { /* autosense was successful */ hcb->ccb->ccb_h.status |= CAM_AUTOSNS_VALID; free_hcb_and_ccb_done(hcb, hcb->ccb, CAM_SCSI_STATUS_ERROR); } else { autosense_failed: free_hcb_and_ccb_done(hcb, hcb->ccb, CAM_AUTOSENSE_FAIL); } schedule_next_hcb(softc); return; } Else the command itself has completed, pay more attention to details. If auto-sense is not disabled for this CCB and the command has failed with sense data then run REQUEST SENSE command to receive that data. hcb->ccb->csio.scsi_status = scsi_status; calculate_residue(hcb); if( (hcb->ccb->ccb_h.flags & CAM_DIS_AUTOSENSE)==0 && ( scsi_status == CHECK_CONDITION || scsi_status == COMMAND_TERMINATED) ) { /* start auto-SENSE */ hcb->flags |= DOING_AUTOSENSE; setup_autosense_command_in_hcb(hcb); restart_current_hcb(softc); return; } if(scsi_status == GOOD) free_hcb_and_ccb_done(hcb, hcb->ccb, CAM_REQ_CMP); else free_hcb_and_ccb_done(hcb, hcb->ccb, CAM_SCSI_STATUS_ERROR); schedule_next_hcb(softc); return; } One typical thing would be negotiation events: negotiation messages received from a SCSI target (in answer to our negotiation attempt or by target's initiative) or the target is unable to negotiate (rejects our negotiation messages or does not answer them). switch(hcb_status) { case TARGET_REJECTED_WIDE_NEG: /* revert to 8-bit bus */ softc->current_bus_width[targ] = softc->goal_bus_width[targ] = 8; /* report the event */ neg.bus_width = 8; neg.valid = CCB_TRANS_BUS_WIDTH_VALID; xpt_async(AC_TRANSFER_NEG, hcb->ccb.ccb_h.path_id, &neg); continue_current_hcb(softc); return; case TARGET_ANSWERED_WIDE_NEG: { int wd; wd = get_target_bus_width_request(softc); if(wd <= softc->goal_bus_width[targ]) { /* answer is acceptable */ softc->current_bus_width[targ] = softc->goal_bus_width[targ] = neg.bus_width = wd; /* report the event */ neg.valid = CCB_TRANS_BUS_WIDTH_VALID; xpt_async(AC_TRANSFER_NEG, hcb->ccb.ccb_h.path_id, &neg); } else { prepare_reject_message(hcb); } } continue_current_hcb(softc); return; case TARGET_REQUESTED_WIDE_NEG: { int wd; wd = get_target_bus_width_request(softc); wd = min (wd, OUR_BUS_WIDTH); wd = min (wd, softc->user_bus_width[targ]); if(wd != softc->current_bus_width[targ]) { /* the bus width has changed */ softc->current_bus_width[targ] = softc->goal_bus_width[targ] = neg.bus_width = wd; /* report the event */ neg.valid = CCB_TRANS_BUS_WIDTH_VALID; xpt_async(AC_TRANSFER_NEG, hcb->ccb.ccb_h.path_id, &neg); } prepare_width_nego_rsponse(hcb, wd); } continue_current_hcb(softc); return; } Then we handle any errors that could have happened during auto-sense in the same simple-minded way as before. Otherwise we look closer at the details again. if(hcb->flags & DOING_AUTOSENSE) goto autosense_failed; switch(hcb_status) { The next event we consider is unexpected disconnect. Which is considered normal after an ABORT or BUS DEVICE RESET message and abnormal in other cases. case UNEXPECTED_DISCONNECT: if(requested_abort(hcb)) { /* abort affects all commands on that target+LUN, so * mark all disconnected HCBs on that target+LUN as aborted too */ for(h = softc->first_discon_hcb[hcb->target][hcb->lun]; h != NULL; h = hh) { hh=h->next; free_hcb_and_ccb_done(h, h->ccb, CAM_REQ_ABORTED); } ccb_status = CAM_REQ_ABORTED; } else if(requested_bus_device_reset(hcb)) { int lun; /* reset affects all commands on that target, so * mark all disconnected HCBs on that target+LUN as reset */ for(lun=0; lun <= OUR_MAX_SUPPORTED_LUN; lun++) for(h = softc->first_discon_hcb[hcb->target][lun]; h != NULL; h = hh) { hh=h->next; free_hcb_and_ccb_done(h, h->ccb, CAM_SCSI_BUS_RESET); } /* send event */ xpt_async(AC_SENT_BDR, hcb->ccb->ccb_h.path_id, NULL); /* this was the CAM_RESET_DEV request itself, it's completed */ ccb_status = CAM_REQ_CMP; } else { calculate_residue(hcb); ccb_status = CAM_UNEXP_BUSFREE; /* request the further code to freeze the queue */ hcb->ccb->ccb_h.status |= CAM_DEV_QFRZN; lun_to_freeze = hcb->lun; } break; If the target refuses to accept tags we notify CAM about that and return back all commands for this LUN: case TAGS_REJECTED: /* report the event */ neg.flags = 0 & ~CCB_TRANS_TAG_ENB; neg.valid = CCB_TRANS_TQ_VALID; xpt_async(AC_TRANSFER_NEG, hcb->ccb.ccb_h.path_id, &neg); ccb_status = CAM_MSG_REJECT_REC; /* request the further code to freeze the queue */ hcb->ccb->ccb_h.status |= CAM_DEV_QFRZN; lun_to_freeze = hcb->lun; break; Then we check a number of other conditions, with processing basically limited to setting the CCB status: case SELECTION_TIMEOUT: ccb_status = CAM_SEL_TIMEOUT; /* request the further code to freeze the queue */ hcb->ccb->ccb_h.status |= CAM_DEV_QFRZN; lun_to_freeze = CAM_LUN_WILDCARD; break; case PARITY_ERROR: ccb_status = CAM_UNCOR_PARITY; break; case DATA_OVERRUN: case ODD_WIDE_TRANSFER: ccb_status = CAM_DATA_RUN_ERR; break; default: /* all other errors are handled in a generic way */ ccb_status = CAM_REQ_CMP_ERR; /* request the further code to freeze the queue */ hcb->ccb->ccb_h.status |= CAM_DEV_QFRZN; lun_to_freeze = CAM_LUN_WILDCARD; break; } Then we check if the error was serious enough to freeze the input queue until it gets proceeded and do so if it is: if(hcb->ccb->ccb_h.status & CAM_DEV_QFRZN) { /* freeze the queue */ xpt_freeze_devq(ccb->ccb_h.path, /*count*/1); /* re-queue all commands for this target/LUN back to CAM */ for(h = softc->first_queued_hcb; h != NULL; h = hh) { hh = h->next; if(targ == h->targ && (lun_to_freeze == CAM_LUN_WILDCARD || lun_to_freeze == h->lun) ) free_hcb_and_ccb_done(h, h->ccb, CAM_REQUEUE_REQ); } } free_hcb_and_ccb_done(hcb, hcb->ccb, ccb_status); schedule_next_hcb(softc); return; This concludes the generic interrupt handling although specific controllers may require some additions. Errors Summary When executing an I/O request many things may go wrong. The reason of error can be reported in the CCB status with great detail. Examples of use are spread throughout this document. For completeness here is the summary of recommended responses for the typical error conditions: CAM_RESRC_UNAVAIL - some resource is temporarily unavailable and the SIM driver can not generate an event when it will become available. An example of this resource would be some intra-controller hardware resource for which the controller does not generate an interrupt when it becomes available. CAM_UNCOR_PARITY - unrecovered parity error occurred CAM_DATA_RUN_ERR - data overrun or unexpected data phase (going in other direction than specified in CAM_DIR_MASK) or odd transfer length for wide transfer CAM_SEL_TIMEOUT - selection timeout occurred (target does not respond) CAM_CMD_TIMEOUT - command timeout occurred (the timeout function ran) CAM_SCSI_STATUS_ERROR - the device returned error CAM_AUTOSENSE_FAIL - the device returned error and the REQUEST SENSE COMMAND failed CAM_MSG_REJECT_REC - MESSAGE REJECT message was received CAM_SCSI_BUS_RESET - received SCSI bus reset CAM_REQ_CMP_ERR - "impossible" SCSI phase occurred or something else as weird or just a generic error if further detail is not available CAM_UNEXP_BUSFREE - unexpected disconnect occurred CAM_BDR_SENT - BUS DEVICE RESET message was sent to the target CAM_UNREC_HBA_ERROR - unrecoverable Host Bus Adapter Error CAM_REQ_TOO_BIG - the request was too large for this controller CAM_REQUEUE_REQ - this request should be re-queued to preserve transaction ordering. This typically occurs when the SIM recognizes an error that should freeze the queue and must place other queued requests for the target at the sim level back into the XPT queue. Typical cases of such errors are selection timeouts, command timeouts and other like conditions. In such cases the troublesome command returns the status indicating the error, the and the other commands which have not be sent to the bus yet get re-queued. CAM_LUN_INVALID - the LUN ID in the request is not supported by the SCSI controller CAM_TID_INVALID - the target ID in the request is not supported by the SCSI controller Timeout Handling When the timeout for an HCB expires that request should be aborted, just like with an XPT_ABORT request. The only difference is that the returned status of aborted request should be CAM_CMD_TIMEOUT instead of CAM_REQ_ABORTED (that's why implementation of the abort better be done as a function). But there is one more possible problem: what if the abort request itself will get stuck? In this case the SCSI bus should be reset, just like with an XPT_RESET_BUS request (and the idea about implementing it as a function called from both places applies here too). Also we should reset the whole SCSI bus if a device reset request got stuck. So after all the timeout function would look like: static void xxx_timeout(void *arg) { struct xxx_hcb *hcb = (struct xxx_hcb *)arg; struct xxx_softc *softc; struct ccb_hdr *ccb_h; softc = hcb->softc; ccb_h = &hcb->ccb->ccb_h; if(hcb->flags & HCB_BEING_ABORTED || ccb_h->func_code == XPT_RESET_DEV) { xxx_reset_bus(softc); } else { xxx_abort_ccb(hcb->ccb, CAM_CMD_TIMEOUT); } } When we abort a request all the other disconnected requests to the same target/LUN get aborted too. So there appears a question, should we return them with status CAM_REQ_ABORTED or CAM_CMD_TIMEOUT ? The current drivers use CAM_CMD_TIMEOUT. This seems logical because if one request got timed out then probably something really bad is happening to the device, so if they would not be disturbed they would time out by themselves. diff --git a/en_US.ISO8859-1/books/developers-handbook/secure/chapter.sgml b/en_US.ISO8859-1/books/developers-handbook/secure/chapter.sgml index 8b4710336d..84ac07b233 100644 --- a/en_US.ISO8859-1/books/developers-handbook/secure/chapter.sgml +++ b/en_US.ISO8859-1/books/developers-handbook/secure/chapter.sgml @@ -1,513 +1,513 @@ Secure Programming This chapter was written by &a.murray;. Synopsis This chapter describes some of the security issues that have plagued Unix programmers for decades and some of the new tools available to help programmers avoid writing exploitable code. Secure Design Methodology Writing secure applications takes a very scrutinous and pessimistic outlook on life. Applications should be run with the principle of least privilege so that no process is ever running with more than the bare minimum access that it needs to accomplish its function. Previously tested code should be reused whenever possible to avoid common mistakes that others may have already fixed. One of the pitfalls of the Unix environment is how easy it is to make assumptions about the sanity of the environment. Applications should never trust user input (in all its forms), system resources, inter-process communication, or the timing of events. Unix processes do not execute synchronously so logical operations are rarely atomic. Buffer Overflows Buffer Overflows have been around since the very beginnings of the Von-Neuman architecture. buffer overflow Von-Neuman They first gained widespread notoriety in 1988 with the Morris Internet worm. Unfortunately, the same basic attack remains Morris Internet worm effective today. Of the 17 CERT security advisories of 1999, 10 CERTsecurity advisories of them were directly caused by buffer-overflow software bugs. By far the most common type of buffer overflow attack is based on corrupting the stack. stack arguments Most modern computer systems use a stack to pass arguments to procedures and to store local variables. A stack is a last in first out (LIFO) buffer in the high memory area of a process image. When a program invokes a function a new "stack frame" is LIFO process image stack pointer created. This stack frame consists of the arguments passed to the function as well as a dynamic amount of local variable space. The "stack pointer" is a register that holds the current stack frame stack pointer location of the top of the stack. Since this value is constantly changing as new values are pushed onto the top of the stack, many implementations also provide a "frame pointer" that is located near the beginning of a stack frame so that local variables can more easily be addressed relative to this value. The return address for function frame pointer process image frame pointer return address stack-overflow calls is also stored on the stack, and this is the cause of stack-overflow exploits since overflowing a local variable in a function can overwrite the return address of that function, potentially allowing a malicious user to execute any code he or she wants. Although stack-based attacks are by far the most common, it would also be possible to overrun the stack with a heap-based (malloc/free) attack. The C programming language does not perform automatic bounds checking on arrays or pointers as many other languages do. In addition, the standard C library is filled with a handful of very dangerous functions. strcpy(char *dest, const char *src) May overflow the dest buffer strcat(char *dest, const char *src) May overflow the dest buffer getwd(char *buf) May overflow the buf buffer gets(char *s) May overflow the s buffer [vf]scanf(const char *format, ...) May overflow its arguments. realpath(char *path, char resolved_path[]) May overflow the path buffer [v]sprintf(char *str, const char *format, ...) May overflow the str buffer. Example Buffer Overflow The following example code contains a buffer overflow designed to overwrite the return address and skip the instruction immediately following the function call. (Inspired by ) #include stdio.h void manipulate(char *buffer) { char newbuffer[80]; strcpy(newbuffer,buffer); } int main() { char ch,buffer[4096]; int i=0; while ((buffer[i++] = getchar()) != '\n') {}; i=1; manipulate(buffer); i=2; printf("The value of i is : %d\n",i); return 0; } Let us examine what the memory image of this process would look like if we were to input 160 spaces into our little program before hitting return. [XXX figure here!] Obviously more malicious input can be devised to execute actual compiled instructions (such as exec(/bin/sh)). Avoiding Buffer Overflows The most straightforward solution to the problem of stack-overflows is to always use length restricted memory and string copy functions. strncpy and strncat are part of the standard C library. string copy functions strncpy string copy functions strncat These functions accept a length value as a parameter which should be no larger than the size of the destination buffer. These functions will then copy up to `length' bytes from the source to the destination. However there are a number of problems with these functions. Neither function guarantees NUL termination if the size of the input buffer is as large as the NUL termination destination. The length parameter is also used inconsistently between strncpy and strncat so it is easy for programmers to get confused as to their proper usage. There is also a significant performance loss compared to strcpy when copying a short string into a large buffer since - strncpy NUL fills up the the size + strncpy NUL fills up the size specified. In OpenBSD, another memory copy implementation has been OpenBSD created to get around these problem. The strlcpy and strlcat functions guarantee that they will always null terminate the destination string when given a non-zero length argument. For more information about these functions see . The OpenBSD strlcpy and strlcat instructions have been in FreeBSD since 3.3. string copy functions strlcpy string copy functions strlcat Compiler based run-time bounds checking bounds checking compiler-based Unfortunately there is still a very large assortment of code in public use which blindly copies memory around without using any of the bounded copy routines we just discussed. Fortunately, there is another solution. Several compiler add-ons and libraries exist to do Run-time bounds checking in C/C++. StackGuard gcc StackGuard is one such add-on that is implemented as a small patch to the gcc code generator. From the StackGuard website, http://immunix.org/stackguard.html :

"StackGuard detects and defeats stack smashing attacks by protecting the return address on the stack from being altered. StackGuard places a "canary" word next to the return address when a function is called. If the canary word has been altered when the function returns, then a stack smashing attack has been attempted, and the program responds by emitting an intruder alert into syslog, and then halts."

"StackGuard is implemented as a small patch to the gcc code generator, specifically the function_prolog() and function_epilog() routines. function_prolog() has been enhanced to lay down canaries on the stack when functions start, and function_epilog() checks canary integrity when the function exits. Any attempt at corrupting the return address is thus detected before the function returns."

buffer overflow Recompiling your application with StackGuard is an effective means of stopping most buffer-overflow attacks, but it can still be compromised. Library based run-time bounds checking bounds checking library-based Compiler-based mechanisms are completely useless for binary-only software for which you cannot recompile. For these situations there are a number of libraries which re-implement the unsafe functions of the C-library (strcpy, fscanf, getwd, etc..) and ensure that these functions can never write past the stack pointer. libsafe libverify libparnoia Unfortunately these library-based defenses have a number of shortcomings. These libraries only protect against a very small set of security related issues and they neglect to fix the actual problem. These defenses may fail if the application was compiled with -fomit-frame-pointer. Also, the LD_PRELOAD and LD_LIBRARY_PATH environment variables can be overwritten/unset by the user. SetUID issues seteuid There are at least 6 different IDs associated with any given process. Because of this you have to be very careful with the access that your process has at any given time. In particular, all seteuid applications should give up their privileges as soon as it is no longer required. user IDs real user ID user IDs effective user ID The real user ID can only be changed by a superuser process. The login program sets this when a user initially logs in and it is seldom changed. The effective user ID is set by the exec() functions if a program has its seteuid bit set. An application can call seteuid() at any time to set the effective user ID to either the real user ID or the saved set-user-ID. When the effective user ID is set by exec() functions, the previous value is saved in the saved set-user-ID. Limiting your program's environment chroot() The traditional method of restricting a process is with the chroot() system call. This system call changes the root directory from which all other paths are referenced for a process and any child processes. For this call to succeed the process must have execute (search) permission on the directory being referenced. The new environment does not actually take effect until you chdir() into your new environment. It should also be noted that a process can easily break out of a chroot environment if it has root privilege. This could be accomplished by creating device nodes to read kernel memory, attaching a debugger to a process outside of the jail, or in many other creative ways. The behavior of the chroot() system call can be controlled somewhat with the kern.chroot_allow_open_directories sysctl variable. When this value is set to 0, chroot() will fail with EPERM if there are any directories open. If set to the default value of 1, then chroot() will fail with EPERM if there are any directories open and the process is already subject to a chroot() call. For any other value, the check for open directories will be bypassed completely. FreeBSD's jail functionality jail The concept of a Jail extends upon the chroot() by limiting the powers of the superuser to create a true `virtual server'. Once a prison is setup all network communication must take place through the specified IP address, and the power of "root privilege" in this jail is severely constrained. While in a prison, any tests of superuser power within the kernel using the suser() call will fail. However, some calls to suser() have been changed to a new interface suser_xxx(). This function is responsible for recognizing or denying access to superuser power for imprisoned processes. A superuser process within a jailed environment has the power to : Manipulate credential with setuid, seteuid, setgid, setegid, setgroups, setreuid, setregid, setlogin Set resource limits with setrlimit Modify some sysctl nodes (kern.hostname) chroot() Set flags on a vnode: chflags, fchflags Set attributes of a vnode such as file permission, owner, group, size, access time, and modification time. Bind to privileged ports in the Internet domain (ports < 1024) Jail is a very useful tool for running applications in a secure environment but it does have some shortcomings. Currently, the IPC mechanisms have not been converted to the suser_xxx so applications such as MySQL can not be run within a jail. Superuser access may have a very limited meaning within a jail, but there is no way to specify exactly what "very limited" means. POSIX.1e Process Capabilities POSIX.1e Process Capabilities TrustedBSD Posix has released a working draft that adds event auditing, access control lists, fine grained privileges, information labeling, and mandatory access control. This is a work in progress and is the focus of the TrustedBSD project. Some of the initial work has been committed to FreeBSD-current (cap_set_proc(3)). Trust An application should never assume that anything about the users environment is sane. This includes (but is certainly not limited to) : user input, signals, environment variables, resources, IPC, mmaps, the file system working directory, file descriptors, the # of open files, etc. positive filtering data validation You should never assume that you can catch all forms of invalid input that a user might supply. Instead, your application should use positive filtering to only allow a specific subset of inputs that you deem safe. Improper data validation has been the cause of many exploits, especially with CGI scripts on the world wide web. For filenames you need to be extra careful about paths ("../", "/"), symbolic links, and shell escape characters. Perl Taint mode Perl has a really cool feature called "Taint" mode which can be used to prevent scripts for using data derived outside the program in an unsafe way. This mode will check command line arguments, environment variables, locale information, the results of certain syscalls (readdir(), readlink(), getpwxxx(), and all file input. Race Conditions A race condition is anomalous behavior caused by the unexpected dependence on the relative timing of events. In other words, a programmer incorrectly assumed that a particular event would always happen before another. race conditions signals race conditions access checks race conditions file opens Some of the common causes of race conditions are signals, access checks, and file opens. Signals are asynchronous events by nature so special care must be taken in dealing with them. Checking access with access(2) then open(2) is clearly non-atomic. Users can move files in between the two calls. Instead, privileged applications should seteuid() and then call open() directly. Along the same lines, an application should always set a proper umask before open() to obviate the need for spurious chmod() calls. diff --git a/en_US.ISO8859-1/books/faq/book.sgml b/en_US.ISO8859-1/books/faq/book.sgml index 35e0ee4a5b..7b03e9cb8f 100644 --- a/en_US.ISO8859-1/books/faq/book.sgml +++ b/en_US.ISO8859-1/books/faq/book.sgml @@ -1,12531 +1,12531 @@ %man; %authors; %bookinfo; ]> Frequently Asked Questions for FreeBSD 2.X, 3.X and 4.X The FreeBSD Documentation Project - $FreeBSD: doc/en_US.ISO8859-1/books/faq/book.sgml,v 1.209 2001/06/08 03:42:15 chris Exp $ + $FreeBSD: doc/en_US.ISO8859-1/books/faq/book.sgml,v 1.210 2001/06/11 01:16:41 ache Exp $ 1995 1996 1997 1998 1999 2000 2001 The FreeBSD Documentation Project &bookinfo.legalnotice; This is the FAQ for FreeBSD versions 2.X, 3.X, and 4.X. All entries are assumed to be relevant to FreeBSD 2.0.5 and later, unless otherwise noted. Any entries with a <XXX> are under construction. If you are interested in helping with this project, send email to the FreeBSD documentation project mailing list freebsd-doc@FreeBSD.org. The latest version of this document is always available from the FreeBSD World Wide Web server. It may also be downloaded as one large HTML file with HTTP or as plain text, postscript, PDF, etc. from the FreeBSD FTP server. You may also want to Search the FAQ. Preface Welcome to the FreeBSD 2.X-4.X FAQ! As is usual with Usenet FAQs, this document aims to cover the most frequently asked questions concerning the FreeBSD operating system (and of course answer them!). Although originally intended to reduce bandwidth and avoid the same old questions being asked over and over again, FAQs have become recognized as valuable information resources. Every effort has been made to make this FAQ as informative as possible; if you have any suggestions as to how it may be improved, please feel free to mail them to the &a.faq;. What is FreeBSD? Briefly, FreeBSD is a UN*X-like operating system for the i386 and Alpha/AXP platforms based on U.C. Berkeley's 4.4BSD-Lite release, with some 4.4BSD-Lite2 enhancements. It is also based indirectly on William Jolitz's port of U.C. Berkeley's Net/2 to the i386, known as 386BSD, though very little of the 386BSD code remains. A fuller description of what FreeBSD is and how it can work for you may be found on the FreeBSD home page. FreeBSD is used by companies, Internet Service Providers, researchers, computer professionals, students and home users all over the world in their work, education and recreation. See some of them in the FreeBSD Gallery. For more detailed information on FreeBSD, please see the FreeBSD Handbook. What are the goals of FreeBSD? The goals of the FreeBSD Project are to provide software that may be used for any purpose and without strings attached. Many of us have a significant investment in the code (and project) and would certainly not mind a little financial compensation now and then, but we're definitely not prepared to insist on it. We believe that our first and foremost mission is to provide code to any and all comers, and for whatever purpose, so that the code gets the widest possible use and provides the widest possible benefit. This is, we believe, one of the most fundamental goals of Free Software and one that we enthusiastically support. That code in our source tree which falls under the GNU General Public License (GPL) or GNU Library General Public License (LGPL) comes with slightly more strings attached, though at least on the side of enforced access rather than the usual opposite. Due to the additional complexities that can evolve in the commercial use of GPL software, we do, however, endeavor to replace such software with submissions under the more relaxed BSD copyright whenever possible. Why is it called FreeBSD? It may be used free of charge, even by commercial users. Full source for the operating system is freely available, and the minimum possible restrictions have been placed upon its use, distribution and incorporation into other work (commercial or non-commercial). Anyone who has an improvement and/or bug fix is free to submit their code and have it added to the source tree (subject to one or two obvious provisos). For those of our readers whose first language is not English, it may be worth pointing out that the word free is being used in two ways here, one meaning at no cost, the other meaning you can do whatever you like. Apart from one or two things you cannot do with the FreeBSD code, for example pretending you wrote it, you really can do whatever you like with it. What is the latest version of FreeBSD? Version 4.3 is the latest STABLE version; it was released in April, 2001. This is also the latest RELEASE version. Briefly explained, -STABLE is aimed at the ISP or other corporate user who wants stability and a low change count over the wizzy new features of the latest -CURRENT snapshot. Releases can come from either branch, but you should only use -CURRENT if you're sure that you're prepared for its increased volatility (relative to -STABLE, that is). Releases are only made every few months. While many people stay more up-to-date with the FreeBSD sources (see the questions on FreeBSD-CURRENT and FreeBSD-STABLE) than that, doing so is more of a commitment, as the sources are a moving target. What is FreeBSD-CURRENT? FreeBSD-CURRENT is the development version of the operating system, which will in due course become 5.0-RELEASE. As such, it is really only of interest to developers working on the system and die-hard hobbyists. See the relevant section in the handbook for details on running -CURRENT. If you are not familiar with the operating system or are not capable of identifying the difference between a real problem and a temporary problem, you should not use FreeBSD-CURRENT. This branch sometimes evolves quite quickly and can be un-buildable for a number of days at a time. People that use FreeBSD-CURRENT are expected to be able to analyze any problems and only report them if they are deemed to be mistakes rather than glitches. Questions such as make world produces some error about groups on the -CURRENT mailing list are sometimes treated with contempt. Every day, snapshot releases are made based on the current state of the -CURRENT and -STABLE branches. Nowadays, distributions of the occasional snapshot are now being made available. The goals behind each snapshot release are: To test the latest version of the installation software. To give people who would like to run -CURRENT or -STABLE but who don't have the time and/or bandwidth to follow it on a day-to-day basis an easy way of bootstrapping it onto their systems. To preserve a fixed reference point for the code in question, just in case we break something really badly later. (Although CVS normally prevents anything horrible like this happening :) To ensure that any new features in need of testing have the greatest possible number of potential testers. No claims are made that any -CURRENT snapshot can be considered production quality for any purpose. If you want to run a stable and fully tested system, you will have to stick to full releases, or use the -STABLE snaphosts. Snapshot releases are directly available from ftp://current.FreeBSD.org/pub/FreeBSD/ for 5.0-CURRENT and releng4.FreeBSD.org for 4-STABLE snapshots. 3-STABLE snapshots are not being produced at the time of this writing (May 2000). Snapshots are generated, on the average, once a day for all actively developed branches. What is the FreeBSD-STABLE concept? Back when FreeBSD 2.0.5 was released, we decided to branch FreeBSD development into two parts. One branch was named -STABLE, with the intention that only well-tested bug fixes and small incremental enhancements would be made to it (for Internet Service Providers and other commercial enterprises for whom sudden shifts or experimental features are quite undesirable). The other branch was -CURRENT, which essentially has been one unbroken line leading towards 5.0-RELEASE (and beyond) since 2.0 was released. If a little ASCII art would help, this is how it looks: 2.0 | | | [2.1-STABLE] *BRANCH* 2.0.5 -> 2.1 -> 2.1.5 -> 2.1.6 -> 2.1.7.1 [2.1-STABLE ends] | (Mar 1997) | | | [2.2-STABLE] *BRANCH* 2.2.1 -> 2.2.2-RELEASE -> 2.2.5 -> 2.2.6 -> 2.2.7 -> 2.2.8 [end] | (Mar 1997) (Oct 97) (Apr 98) (Jul 98) (Dec 98) | | 3.0-SNAPs (started Q1 1997) | | 3.0-RELEASE (Oct 1998) | | [3.0-STABLE] *BRANCH* 3.1-RELEASE (Feb 1999) -> 3.2 -> 3.3 -> 3.4 -> 3.5 -> 3.5.1 | (May 1999) (Sep 1999) (Dec 1999) (June 2000) (July 2000) | | [4.0-STABLE] *BRANCH* 4.0 (Mar 2000) -> 4.1 -> 4.1.1 -> 4.2 -> 4.3 -> ... future 4.x releases ... | | (July 2000) (Sep 2000) (Nov 2000) \|/ + [5.0-CURRENT continues] The -CURRENT branch is slowly progressing towards 5.0 and beyond, the previous 2.2-STABLE branch having been retired with the release of 2.2.8. 3-STABLE replaced it, with 3.5.1 (the final 3.X release) being released in July 2000. In May 2000 (even though 3.5 came after that), the 3-STABLE branch was more or less replaced by the 4-STABLE branch. 4.3-RELEASE was released in April 2001. 4-STABLE is the actively developed -STABLE branch, although some bugfixes (mostly security-related) are still being committed to 3-STABLE. It is expected that the 3.X branch will be officially obsoleted some time in summer 2000. 5.0-CURRENT is now the current branch, with the no release date planed. When are FreeBSD releases made? As a general principle, the FreeBSD core team only release a new version of FreeBSD when they believe that there are sufficient new features and/or bug fixes to justify one, and are satisfied that the changes made have settled down sufficiently to avoid compromising the stability of the release. Many users regard this caution as one of the best things about FreeBSD, although it can be a little frustrating when waiting for all the latest goodies to become available... Releases are made about every 4 months on average. For people needing (or wanting) a little more excitement, binary snapshots are made every day... see above. Is FreeBSD only available for PCs? Since 3.x, FreeBSD has run on the DEC Alpha as well as the x86 architecture. Some interest has also been expressed in a SPARC, PowerPC and IA64 ports. If your machine has a different architecture and you need something right now, we suggest you look at NetBSD or OpenBSD. Who is responsible for FreeBSD? The key decisions concerning the FreeBSD project, such as the overall direction of the project and who is allowed to add code to the source tree, are made by a core team of 9 people. There is a much larger team of more than 200 committers who are authorized to make changes directly to the FreeBSD source tree. However, most non-trivial changes are discussed in advance in the mailing lists, and there are no restrictions on who may take part in the discussion. Where can I get FreeBSD? Every significant release of FreeBSD is available via anonymous ftp from the FreeBSD FTP site: For the current 3.X-STABLE release, 3.5.1-RELEASE, see the 3.5.1-RELEASE directory. The current 4-STABLE release, 4.3-RELEASE can be found in the 4.3-RELEASE directory. 4.X snapshots are usually made once a day. 5.0 Snapshot releases are made once a day for the -CURRENT branch, these being of service purely to bleeding-edge testers and developers. FreeBSD is also available via CDROM, from the following place(s):

BSDi 4041 Pike Lane, Suite F Concord, CA 94520 USA Orders: +1 800 786-9907 Questions: +1 925 674-0783 FAX: +1 925 674-0821 email: BSDi Orders address WWW: BSDi Home page

In Australia, you may find it at:

Advanced Multimedia Distributors Factory 1/1 Ovata Drive Tullamarine, Melbourne Victoria Australia Voice: +61 3 9338 6777 CDROM Support BBS 17 Irvine St Peppermint Grove, WA 6011 Voice: +61 9 385-3793 Fax: +61 9 385-2360

And in the UK:

The Public Domain & Shareware Library Winscombe House, Beacon Rd Crowborough Sussex. TN6 1UL Voice: +44 1892 663-298 Fax: +44 1892 667-473

Where do I find info on the FreeBSD mailing lists? You can find full information in the Handbook entry on mailing-lists. Where do I find the FreeBSD Y2K info? You can find full information in the FreeBSD Y2K page. What FreeBSD news groups are available? You can find full information in the Handbook entry on newsgroups. Are there FreeBSD IRC (Internet Relay Chat) channels? Yes, most major IRC networks host a FreeBSD chat channel: Channel #FreeBSD on EFNet is a FreeBSD forum, but don't go there for tech support or to try and get folks there to help you avoid the pain of reading man pages or doing your own research. It is a chat channel, first and foremost, and topics there are just as likely to involve sex, sports or nuclear weapons as they are FreeBSD. You Have Been Warned! Available at server irc.chat.org. Channel #FreeBSDhelp on EFNet is a channel dedicated to helping FreeBSD users. They are much more sympathetic to questions then #FreeBSD is. Channel #FreeBSD on DALNET is available at irc.dal.net in the US and irc.eu.dal.net in Europe. Channel #FreeBSD on UNDERNET is available at us.undernet.org in the US and eu.undernet.org in Europe. Since it is a help channel, be prepared to read the documents you are referred to. Channel #FreeBSD on HybNet. This channel is a help channel. A list of servers can be found on the HybNet web site. Each of these channels are distinct and are not connected to each other. Their chat styles also differ, so you may need to try each to find one suited to your chat style. As with *all* types of IRC traffic, if you're easily offended or can't deal with lots of young people (and more than a few older ones) doing the verbal equivalent of jello wrestling, don't even bother with it. What books are there on FreeBSD? There is a FreeBSD Documentation Project which you may contact (or even better, join) at the freebsd-doc mailing list: freebsd-doc@FreeBSD.org. This list is for discussion of the FreeBSD documentation. For actual questions about FreeBSD, there is the freebsd-questions mailing list: freebsd-questions@FreeBSD.org. A FreeBSD handbook is available, and can be found as: the FreeBSD Handbook. Note that this is a work in progress; some parts may be incomplete or out-of-date. The definitive printed guide on FreeBSD is The Complete FreeBSD, written by Greg Lehey and published by BSDi (formerly Walnut Creek CDROM) Books. Now in its third edition, the book contains 773 pages of install & system administration guidance, program setup help, and manual pages. The book (and current FreeBSD release) can be ordered from BSDi, CheapBytes, or at your favorite bookstore. The ISBN is 1-57176-246-9 (this may not be unique). Since FreeBSD is based upon Berkeley 4.4BSD-Lite, most of the 4.4BSD manuals are applicable to FreeBSD. O'Reilly and Associates publishes the following manuals: A description of these can be found via WWW as: 4.4BSD books description. Due to poor sales, however, these manuals may be hard to get a hold of. For a more in-depth look at the 4.4BSD kernel organization, you can't go wrong with . A good book on system administration is . Make sure you get the third edition, with a purple cover, instead of the first edition. This book covers the basics, as well as TCP/IP, DNS, NFS, SLIP/PPP, sendmail, INN/NNTP, printing, etc. It's expensive but worth it. The third edition covers Solaris, HP/UX, FreeBSD, and Linux. How do I access the Problem Report database? The Problem Report database of all user change requests may be queried (or submitted to) by using our web-based PR submission and query interfaces. The &man.send-pr.1; command can also be used to submit problem reports and change requests via electronic mail. Is the documentation available in other formats, such as plain text (ASCII), or Postscript? Yes. The documentation is available in a number of different formats and compression schemes on the FreeBSD FTP site, in the /pub/FreeBSD/doc/ directory. The documentation is categorised in a number of different ways. These include: The document's name, such as faq, or handbook. The document's language and encoding. These are based on the locale names you will find under /usr/share/locale on your FreeBSD system. The current languages and encodings that we have for documentation are as follows: Name Meaning en_US.ISO8859-1 US English de_DE.ISO8859-1 German es_ES.ISO8859-1 Spanish fr_FR.ISO8859-1 French ja_JP.eucJP Japanese (EUC encoding) ru_RU.KOI8-R Russian (KOI8-R encoding) zh_TW.Big5 Chinese (Big5 encoding) Some documents may not be available in all languages. The document's format. We produce the documentation in a number of different output formats to try and make it as flexible as possible. The current formats are; Format Meaning html-split A collection of small, linked, HTML files. html One large HTML file containing the entire document pdb Palm Pilot database format, for use with the iSilo reader. pdf Adobe's Portable Document Format ps Postscript rtf Microsoft's Rich Text Format Page numbers are not automatically updated when loading this format in to Word. Press CTRL+A, CTRL+END, F9 after loading the document, to update the page numbers. txt Plain text The compression and packaging scheme. There are three of these currently in use. Where the format is html-split, the files are bundled up using &man.tar.1;. The resulting .tar file is then compressed using the compression schemes detailed in the next point. All the other formats generate one file, called book.format (i.e., book.pdb, book.html, and so on). These files are then compressed using three compression schemes. Scheme Description zip The Zip format. If you want to uncompress this on FreeBSD you will need to install the archivers/unzip port first. gz The GNU Zip format. Use &man.gunzip.1; to uncompress these files, which is part of FreeBSD. bz2 The BZip2 format. Less widespread than the others, but generally gives smaller files. Install the archivers/bzip2 port to uncompress these files. So the Postscript version of the Handbook, compressed using BZip2 will be stored in a file called book.sgml.bz2 in the handbook/ directory. The formatted documentation is also available as a FreeBSD package, of which more later. After choosing the format and compression mechanism that you want to download, you must then decide whether or not you want to download the document as a FreeBSD package. The advantage of downloading and installing the package is that the documentation can then be managed using the normal FreeBSD package management comments, such as &man.pkg.add.1; and &man.pkg.delete.1;. If you decide to download and install the package then you must know the filename to download. The documentation-as-packages files are stored in a directory called packages. Each package file looks like document-name.lang.encoding.format.tgz. For example, the FAQ, in English, formatted as PDF, is in the package called faq.en_US.ISO8859-1.pdf.tgz. Knowing this, you can use the following command to install the English PDF FAQ package. &prompt.root; pkg_add ftp://ftp.FreeBSD.org/pub/FreeBSD/doc/packages/faq.en_US.ISO8859-1.pdf.tgz Having done that, you can use &man.pkg.info.1; to determine where the file has been installed. &prompt.root; pkg_info -f faq.en_US.ISO8859-1.pdf Information for faq.en_US.ISO8859-1.pdf: Packing list: Package name: faq.en_US.ISO8859-1.pdf CWD to /usr/share/doc/en_US.ISO8859-1/books/faq File: book.pdf CWD to . File: +COMMENT (ignored) File: +DESC (ignored) As you can see, book.pdf will have been installed in to /usr/share/doc/en_US.ISO8859-1/books/faq. If you do not want to use the packages then you will have to download the compressed files yourself, uncompress them, and then copy the appropriate documents in to place. For example, the split HTML version of the FAQ, compressed using &man.gzip.1;, can be found in the en_US.ISO8859-1/books/faq/book.html-split.tar.gz file. To download and uncompress that file you would have to do this. &prompt.root; fetch ftp://ftp.freebsd.org/pub/FreeBSD/doc/en_US.ISO8859-1/books/faq/book.html-split.tar.gz &prompt.root; gzip -d book.html-split.tar.gz &prompt.root; tar xvf book.html-split.tar You will be left with a collection of .html files. The main one is called index.html, which will contain the table of contents, introductory material, and links to the other parts of the document. You can then copy or move these to their final location as necessary. How do I become a FreeBSD Web mirror? Certainly! There are multiple ways to mirror the Web pages. Using CVSup: You can retrieve the formatted files using CVSup, and connecting to a CVSup server. To retrieve the webpages, please look at the example supfile, which can be found in /usr/share/examples/cvsup/www-supfile. Using ftp mirror: You can download the FTP server's copy of the web site sources using your favorite ftp mirror tool. Keep in mind that you have to build these sources before publishing them. Simply start at ftp://ftp.FreeBSD.org/pub/FreeBSD/FreeBSD-current/www. What do I need to do to translate a FreeBSD document into my native language? Well, we can't pay, but we might arrange a free CD or T-shirt and a Contributor's Handbook entry if you submit a translation of the documentation. Before you begin translating please contact the freebsd-doc mailing list at freebsd-doc@FreeBSD.org; you may find somebody to help with the translation effort. You may also find out there is already a team translating the docs into your chosen language, who surely wouldn't turn down your help. What other sources of information are there? The following newsgroups contain pertinent discussion for FreeBSD users: comp.unix.bsd.freebsd.announce (moderated) comp.unix.bsd.freebsd.misc comp.unix.bsd.misc Web resources: The FreeBSD Home Page. If you have a laptop, be sure and see Tatsumi Hosokawa's Mobile Computing page in Japan. For information on SMP (Symmetric MultiProcessing), please see the SMP support page. For information on FreeBSD multimedia applications, please see the multimedia page. If you're interested specifically in the Bt848 video capture chip, then follow that link. The FreeBSD handbook also has a fairly complete bibliography section which is worth reading if you're looking for actual books to buy. Installation Which file do I download to get FreeBSD? Prior to release 3.1, you only needed one floppy image to install FreeBSD, namely floppies/boot.flp. However, since release 3.1 the Project has added base support for a wide variety of hardware which needed more space, and thus for 3.x and 4.x we now use two floppy images, namely floppies/kernel.flp and floppies/mfsroot.flp. These images need to be copied onto floppies by tools like fdimage or &man.dd.1;. If you need to download the distributions yourself (for a DOS filesystem install, for instance), below are some recommendations for distributions to grab: bin/ manpages/ compat*/ doc/ src/ssys.* Full instructions on this procedure and a little bit more about installation issues in general can be found in the Handbook entry on installing FreeBSD. What do I do if the floppy images does not fit on a single floppy? A 3.5 inch (1.44MB) floppy can accomodate 1474560 bytes of data. The boot image is exactly 1474560 bytes in size. Common mistakes when preparing the boot floppy are: Not downloading the floppy image in binary mode when using FTP. Some FTP clients default their transfer mode to ascii and attempt to change any end-of-line characters received to match the conventions used by the client's system. This will almost invariably corrupt the boot image. Check the size of the downloaded boot image: if it is not exactly that on the server, then the download process is suspect. To workaround: type binary at the FTP command prompt after getting connected to the server and before starting the download of the image. Using the DOS copy command (or equivalent GUI tool) to transfer the boot image to floppy. Programs like copy will not work as the boot image has been created to be booted into directly. The image has the complete content of the floppy, track for track, and is not meant to be placed on the floppy as a regular file. You have to transfer it to the floppy raw, using the low-level tools (e.g. fdimage or rawrite) described in the installation guide to FreeBSD. Where are the instructions for installing FreeBSD? Installation instructions can be found in the Handbook entry on installing FreeBSD. What do I need in order to run FreeBSD? You'll need a 386 or better PC, with 5 MB or more of RAM and at least 60 MB of hard disk space. It can run with a low end MDA graphics card but to run X11R6, a VGA or better video card is needed. See also the section on I have only 4 MB of RAM. Can I install FreeBSD? FreeBSD 2.1.7 was the last version of FreeBSD that could be installed on a 4MB system. Newer versions of FreeBSD, like 2.2, need at least 5MB to install on a new system. All versions of FreeBSD, including 3.0, will run in 4MB of RAM, they just can't run the installation program in 4MB. You can add extra memory for the install process, if you like, and then after the system is up and running, go back to 4MB. Or you could always just swap your disk into a system which has >4MB, install onto it and then swap it back. There are also situations in which FreeBSD 2.1.7 will not install in 4 MB. To be exact: it does not install with 640 kB base + 3 MB extended memory. If your motherboard can remap some of the lost memory out of the 640kB to 1MB region, then you may still be able to get FreeBSD 2.1.7 up. Try to go into your BIOS setup and look for a remap option. Enable it. You may also have to disable ROM shadowing. It may be easier to get 4 more MB just for the install. Build a custom kernel with only the options you need and then get the 4MB out again. You may also install 2.0.5 and then upgrade your system to 2.1.7 with the upgrade option of the 2.1.7 installation program. After the installation, if you build a custom kernel, it will run in 4 MB. Someone has even succeeded in booting with 2 MB (the system was almost unusable though :-)) How can I make my own custom install floppy? Currently there's no way to just make a custom install floppy. You have to cut a whole new release, which will include your install floppy. To make a custom release, follow the instructions here. Can I have more than one operating system on my PC? Have a look at The multi-OS page. Can Windows 95/98 co-exist with FreeBSD? Install Windows 95/98 first, after that FreeBSD. FreeBSD's boot manager will then manage to boot Win95/98 and FreeBSD. If you install Windows 95/98 second, it will boorishly overwrite your boot manager without even asking. If that happens, see the next section. Windows 95/98 killed my boot manager! How do I get it back? You can reinstall the boot manager FreeBSD comes with in one of three ways: Running DOS, go into the tools/ directory of your FreeBSD distribution and look for bootinst.exe. You run it like so: ...\TOOLS> bootinst.exe boot.bin and the boot manager will be reinstalled. Boot the FreeBSD boot floppy again and go to the Custom installation menu item. Choose Partition. Select the drive which used to contain your boot manager (likely the first one) and when you come to the partition editor for it, as the very first thing (e.g. do not make any changes) select (W)rite. This will ask for confirmation, say yes, and when you get the Boot Manager selection prompt, be sure to select Boot Manager. This will re-write the boot manager to disk. Now quit out of the installation menu and reboot off the hard disk as normal. Boot the FreeBSD boot floppy (or CD-ROM) and choose the Fixit menu item. Select either the Fixit floppy or CD-ROM #2 (the live file system option) as appropriate and enter the fixit shell. Then execute the following command: Fixit# fdisk -B -b /boot/boot0 bootdevice substituting bootdevice for your real boot device such as ad0 (first IDE disk), ad4 (first IDE disk on auxiliary controller), da0 (first SCSI disk), etc. My A, T, or X series IBM Thinkpad locks up when I first booted up my FreeBSD installation. How can I solve this? A bug in early revisions of IBM's BIOS on these machines mistakenly identifies the FreeBSD partition as a potential FAT suspend-to-disk partition. When the BIOS tries to parse the FreeBSD partition it hangs. According to IBMIn an e-mail from Keith Frechette kfrechet@us.ibm.com., the following model/BIOS release numbers incorporate the fix. Model BIOS revision T20 IYET49WW or later T21 KZET22WW or later A20p IVET62WW or later A20m IWET54WW or later A21p KYET27WW or later A21m KXET24WW or later A21e KUET30WW It has been reported that later IBM BIOS revisions may have reintroduced the bug. This message from Jacques Vidrine to the mobile@freebsd.org mailing list describes a procedure which may work if your newer IBM laptop does not boot FreeBSD properly, and you can upgrade or downgrade the BIOS.. If you have an earlier BIOS, and upgrading is not an option a workaround is to install FreeBSD, change the partition ID FreeBSD uses, and install new boot blocks that can handle the different partition ID. First, you'll need to restore the machine to a state where it can get through its self-test screen. Doing this requires powering up the machine without letting it find a FreeBSD partition on its primary disk. One way is to remove the hard disk and temporarily move it to an older ThinkPad (such as a ThinkPad 600) or a desktop PC with an appropriate conversion cable. Once it's there, you can delete the FreeBSD partition and move the hard disk back. The ThinkPad should now be in a bootable state again. With the machine functional again, you can use the workaround procedure described here to get a working FreeBSD installation. Download boot1 and boot2 from http://people.freebsd.org/~bmah/ThinkPad/. Put these files somewhere you will be able to retrieve them later. Install FreeBSD as normal on to the ThinkPad. Do not use Dangerously Dedicated mode. Do not reboot when the install has finished. Either switch to the Emergency Holographic Shell (ALT F4) or start a fixit shell. Use &man.fdisk.8; to change the FreeBSD partition ID from 165 to 166 (this is the type used by OpenBSD). Bring the boot1 and boot2 files to the local filesystem. Use &man.disklabel.8; to write boot1 and boot2 to your FreeBSD slice. &prompt.root; disklabel -B -b boot1 -s boot2 ad0sn n is the number of the slice where you installed FreeBSD. Reboot. At the boot prompt you will be given the option of booting OpenBSD. This will actually boot FreeBSD. Getting this to work in the case where you want to dual boot OpenBSD and FreeBSD on the same laptop is left as an exercise for the reader. Can I install on a disk with bad blocks? Prior to 3.0, FreeBSD included a utility known as bad144, which automatically remapped bad blocks. Because modern IDE drives perform this function themselves, bad144 has been removed from the FreeBSD source tree. If you wish to install FreeBSD 3.0 or later, we strongly suggest you purchase a newer disk drive. If you do not wish to do this, you must run FreeBSD 2.x. If you are seeing bad block errors with a modern IDE drive, chances are the drive is going to die very soon (the drive's internal remapping functions are no longer sufficient to fix the bad blocks, which means the disk is heavily corrupted); we suggest you by a new hard drive. If you have a SCSI drive with bad blocks, see this answer. Strange things happen when I boot the install floppy! What is happening? If you're seeing things like the machine grinding to a halt or spontaneously rebooting when you try to boot the install floppy, here are three questions to ask yourself:- Did you use a new, freshly-formatted, error-free floppy (preferably a brand-new one straight out of the box, as opposed to the magazine coverdisk that's been lying under the bed for the last three years)? Did you download the floppy image in binary (or image) mode? (don't be embarrassed, even the best of us have accidentally downloaded a binary file in ASCII mode at least once!) If you're using Windows95 or Win98 did you run fdimage or rawrite in pure DOS mode? These OS's can interfere with programs that write directly to hardware, which the disk creation program does; even running it inside a DOS shell in the GUI can cause this problem. There have also been reports of Netscape causing problems when downloading the boot floppy, so it's probably best to use a different FTP client if you can. I booted from my ATAPI CD-ROM, but the install program says no CD-ROM is found. Where did it go? The usual cause of this problem is a mis-configured CD-ROM drive. Many PCs now ship with the CD-ROM as the slave device on the secondary IDE controller, with no master device on that controller. This is illegal according to the ATAPI specification, but Windows plays fast and loose with the specification, and the BIOS ignores it when booting. This is why the BIOS was able to see the CD-ROM to boot from it, but why FreeBSD can not see it to complete the install. Reconfigure your system so that the CD-ROM is either the master device on the IDE controller it is attached to, or make sure that it is the slave on an IDE controller that also has a master device. Why can I not install from tape? If you are installing 2.1.7R from tape, you must create the tape using a tar blocksize of 10 (5120 bytes). The default tar blocksize is 20 (10240 bytes), and tapes created using this default size cannot be used to install 2.1.7R; with these tapes, you will get an error that complains about the record size being too big. How do I connect two FreeBSD systems over a parallel line using PLIP? Get a laplink cable. Make sure both computer have a kernel with lpt driver support. &prompt.root; dmesg | grep lp lpt0 at 0x378-0x37f irq 7 on isa lpt0: Interrupt-driven lp0: TCP/IP capable interface Plug in the laplink cable into the parallel interface. Configure the network interface parameters for lp0 on both sites as root. For example, if you want connect the host max with moritz max <-----> moritz IP Address 10.0.0.1 10.0.0.2 on max start &prompt.root; ifconfig lp0 10.0.0.1 10.0.0.2 on moritz start &prompt.root; ifconfig lp0 10.0.0.2 10.0.0.1 Thats all! Please read also the manpages &man.lp.4; and &man.lpt.4; . You should also add the hosts to /etc/hosts. 127.0.0.1 localhost.my.domain localhost 10.0.0.1 max.my.domain max 10.0.0.2 moritz.my.domain To check if it works do: on max: &prompt.root; ifconfig lp0 lp0: flags=8851<UP,POINTOPOINT,RUNNING,SIMPLEX,MULTICAST> mtu 1500 inet 10.0.0.1 --> 10.0.0.2 netmask 0xff000000 &prompt.root; netstat -r Routing tables Internet: Destination Gateway Flags Refs Use Netif Expire moritz max UH 4 127592 lp0 &prompt.root; ping -c 4 moritz PING moritz (10.0.0.2): 56 data bytes 64 bytes from 10.0.0.2: icmp_seq=0 ttl=255 time=2.774 ms 64 bytes from 10.0.0.2: icmp_seq=1 ttl=255 time=2.530 ms 64 bytes from 10.0.0.2: icmp_seq=2 ttl=255 time=2.556 ms 64 bytes from 10.0.0.2: icmp_seq=3 ttl=255 time=2.714 ms --- moritz ping statistics --- 4 packets transmitted, 4 packets received, 0% packet loss round-trip min/avg/max/stddev = 2.530/2.643/2.774/0.103 ms Can I install on my laptop over PLIP (Parallel Line IP)? Connect the two computers using a Laplink parallel cable to use this feature: Wiring a parallel cable for networking A-name A-End B-End Descr. Post/Bit DATA0 -ERROR 2 15 15 2 Data 0/0x01 1/0x08 DATA1 +SLCT 3 13 13 3 Data 0/0x02 1/0x10 DATA2 +PE 4 12 12 4 Data 0/0x04 1/0x20 DATA3 -ACK 5 10 10 5 Strobe 0/0x08 1/0x40 DATA4 BUSY 6 11 11 6 Data 0/0x10 1/0x80 GND 18-25 18-25 GND -

See also this note on the Mobile Computing page. Which geometry should I use for a disk drive? By the geometry of a disk, we mean the number of cylinders, heads and sectors/track on a disk - I'll refer to this as C/H/S for convenience. This is how the PC's BIOS works out which area on a disk to read/write from. This seems to cause a lot of confusion for some reason. First of all, the physical geometry of a SCSI drive is totally irrelevant, as FreeBSD works in term of disk blocks. In fact, there is no such thing as the physical geometry, as the sector density varies across the disk - what manufacturers claim is the quote physical geometry is usually the geometry that they've worked out results in the least wasted space. For IDE disks, FreeBSD does work in terms of C/H/S, but all modern drives will convert this into block references internally as well. All that matters is the logical geometry - the answer that the BIOS gets when it asks what is your geometry? and then uses to access the disk. As FreeBSD uses the BIOS when booting, it's very important to get this right. In particular, if you have more than one operating system on a disk, they must all agree on the geometry, otherwise you will have serious problems booting! For SCSI disks, the geometry to use depends on whether extended translation support is turned on in your controller (this is often referred to as support for DOS disks >1GB or something similar). If it's turned off, then use N cylinders, 64 heads and 32 sectors/track, where N is the capacity of the disk in MB. For example, a 2GB disk should pretend to have 2048 cylinders, 64 heads and 32 sectors/track. If it is turned on (it's often supplied this way to get around certain limitations in MSDOS) and the disk capacity is more than 1GB, use M cylinders, 63 sectors per track (*not* 64), and 255 heads, where 'M' is the disk capacity in MB divided by 7.844238 (!). So our example 2GB drive would have 261 cylinders, 63 sectors per track and 255 heads. If you are not sure about this, or FreeBSD fails to detect the geometry correctly during installation, the simplest way around this is usually to create a small DOS partition on the disk. The correct geometry should then be detected (and you can always remove the DOS partition in the partition editor if you don't want to keep it, or leave it around for programming network cards and the like). Alternatively, there is a freely available utility distributed with FreeBSD called pfdisk.exe (located in the tools subdirectory on the FreeBSD CDROM or on the various FreeBSD ftp sites) which can be used to work out what geometry the other operating systems on the disk are using. You can then enter this geometry in the partition editor. Any restrictions on how I divide the disk up? Yes. You must make sure that your root partition is below 1024 cylinders so the BIOS can boot the kernel from it. (Note that this is a limitation in the PC's BIOS, not FreeBSD). For a SCSI drive, this will normally imply that the root partition will be in the first 1024MB (or in the first 4096MB if extended translation is turned on - see previous question). For IDE, the corresponding figure is 504MB. Is FreeBSD compatible with any disk managers? FreeBSD recognizes the Ontrack Disk Manager and makes allowances for it. Other disk managers are not supported. If you just want to use the disk with FreeBSD you don't need a disk manager. Just configure the disk for as much space as the BIOS can deal with (usually 504 megabytes), and FreeBSD should figure out how much space you really have. If you're using an old disk with an MFM controller, you may need to explicitly tell FreeBSD how many cylinders to use. If you want to use the disk with FreeBSD and another operating system, you may be able to do without a disk manager: - just make sure the the FreeBSD boot partition and the slice for + just make sure the FreeBSD boot partition and the slice for the other operating system are in the first 1024 cylinders. If you're reasonably careful, a 20 megabyte boot partition should be plenty. When I boot FreeBSD I get Missing Operating System. What is happening? This is classically a case of FreeBSD and DOS or some other OS conflicting over their ideas of disk geometry. You will have to reinstall FreeBSD, but obeying the instructions given above will almost always get you going. Why can I not get past the boot manager's F? prompt? This is another symptom of the problem described in the preceding question. Your BIOS geometry and FreeBSD geometry settings do not agree! If your controller or BIOS supports cylinder translation (often marked as >1GB drive support), try toggling its setting and reinstalling FreeBSD. Do I need to install the complete sources? In general, no. However, we would strongly recommend that you install, at a minimum, the base source kit, which includes several of the files mentioned here, and the sys (kernel) source kit, which includes sources for the kernel. There is nothing in the system which requires the presence of the sources to operate, however, except for the kernel-configuration program &man.config.8;. With the exception of the kernel sources, our build structure is set up so that you can read-only mount the sources from elsewhere via NFS and still be able to make new binaries. (Because of the kernel-source restriction, we recommend that you not mount this on /usr/src directly, but rather in some other location with appropriate symbolic links to duplicate the top-level structure of the source tree.) Having the sources on-line and knowing how to build a system with them will make it much easier for you to upgrade to future releases of FreeBSD. To actually select a subset of the sources, use the Custom menu item when you are in the Distributions menu of the system installation tool. Do I need to build a kernel? Building a new kernel was originally pretty much a required step in a FreeBSD installation, but more recent releases have benefited from the introduction of a much friendlier kernel configuration tool. When at the FreeBSD boot prompt (boot:), use the flag and you will be dropped into a visual configuration screen which allows you to configure the kernel's settings for most common ISA cards. It's still recommended that you eventually build a new kernel containing just the drivers that you need, just to save a bit of RAM, but it's no longer a strict requirement for most systems. Should I use DES passwords, or MD5, and how do I specify which form my users receive? The default password format on FreeBSD is to use MD5-based passwords. These are believed to be more secure than the traditional UNIX password format, which used a scheme based on the DES algorithm. DES passwords are still available if you need to share your password file with legacy operating systems which still use the less secure password format (they are available if you choose to install the crypto distribution in sysinstall, or by installing the crypto sources if building from source). Which password format to use for new passwords is controlled by the passwd_format login capability in /etc/login.conf, which takes values of either des (if available) or md5. See the &man.login.conf.5; manpage for more information about login capabilities. Why does the boot floppy start, but hang at the Probing Devices... screen? If you have a IDE Zip or Jaz drive installed, remove it and try again. The boot floppy can get confused by the drives. After the system is installed you can reconnect the drive. Hopefully this will be fixed in a later release. Why do I get a panic: can't mount root error when rebooting the system after installation? This error comes from confusion between the boot block's and the kernel's understanding of the disk devices. The error usually manifests on two-disk IDE systems, with the hard disks arranged as the master or single device on separate IDE controllers, with FreeBSD installed on the secondary IDE controller. The boot blocks think the system is installed on wd1 (the second BIOS disk) while the kernel assigns the first disk on the secondary controller device wd2. After the device probing, the kernel tries to mount what the boot blocks think is the boot disk, wd1, while it is really wd2, and fails. To fix the problem, do one of the following: For FreeBSD 3.3 and later, reboot the system and hit Enter at the Booting kernel in 10 seconds; hit [Enter] to interrupt prompt. This will drop you into the boot loader. Then type set root_disk_unit="disk_number" . disk_number will be 0 if FreeBSD is installed on the master drive on the first IDE controller, 1 if it is installed on the slave on the first IDE controller, 2 if it is installed on the master of the second IDE controller, and 3 if it is installed on the slave of the second IDE controller. Then type boot, and your system should boot correctly. To make this change permanent (ie so you don't have to do this everytime you reboot or turn on your FreeBSD machine), put the line root_disk_unit="disk_number" in /boot/loader.conf.local . If using FreeBSD 3.2 or earlier, at the Boot: prompt, enter 1:wd(2,a)kernel and press Enter. If the system starts, then run the command echo "1:wd(2,a)kernel" > /boot.config to make it the default boot string. Move the FreeBSD disk onto the primary IDE controller, so the hard disks are consecutive. Rebuild your kernel, modify the wd configuration lines to read: controller wdc0 at isa? port "IO_WD1" bio irq 14 vector wdintr disk wd0 at wdc0 drive 0 # disk wd1 at wdc0 drive 1 # comment out this line controller wdc1 at isa? port "IO_WD2" bio irq 15 vector wdintr disk wd1 at wdc1 drive 0 # change from wd2 to wd1 disk wd2 at wdc1 drive 1 # change from wd3 to wd2 Install the new kernel. If you moved your disks and wish to restore the previous configuration, replace the disks in the desired configuration and reboot. Your system should boot successfully. What are the limits for memory? For memory, the limit is 4 gigabytes. This configuration has been tested, see wcarchive's configuration for more details. If you plan to install this much memory into a machine, you need to be careful. You'll probably want to use ECC memory and to reduce capacitive loading use 9 chip memory modules vice 18 chip memory modules. What are the limits for ffs filesystems? For ffs filesystems, the maximum theoretical limit is 8 terabytes (2G blocks), or 16TB for the default block size of 8K. In practice, there is a soft limit of 1 terabyte, but with modifications filesystems with 4 terabytes are possible (and exist). The maximum size of a single ffs file is approximately 1G blocks (4TB) if the block size is 4K. Maximum file sizes fs block size 2.2.7-stable 3.0-current works should work 4K 4T-1 4T-1 4T-1 >4T 8K >32G 8T-1 >32G 32T-1 16K >128G 16T-1 >128G 32T-1 32K >512G 32T-1 >512G 64T-1 64K >2048G 64T-1 >2048G 128T-1

When the fs block size is 4K, triple indirect blocks work and everything should be limited by the maximum fs block number that can be represented using triple indirect blocks (approx. 1K^3 + 1K^2 + 1K), but everything is limited by a (wrong) limit of 1G-1 on fs block numbers. The limit on fs block numbers should be 2G-1. There are some bugs for fs block numbers near 2G-1, but such block numbers are unreachable when the fs block size is 4K. For block sizes of 8K and larger, everything should be limited by the 2G-1 limit on fs block numbers, but is actually limited by the 1G-1 limit on fs block numbers, except under -STABLE triple indirect blocks are unreachable, so the limit is the maxiumum fs block number that can be represented using double indirect blocks (approx. (blocksize/4)^2 + (blocksize/4)), and under -CURRENT exceeding this limit may cause problems. Using the correct limit of 2G-1 blocks does cause problems. How can I put 1TB files on my floppy? I keep several virtual ones on floppies :-). The maxiumum file size is not closely related to the maximum disk size. The maximum disk size is 1TB. It is a feature that the file size can be larger than the disk size. The following example creates a file of size 8T-1 using a whole 32K of disk space (3 indirect blocks and 1 data block) on a small root partition. The dd command requires a dd that works with large files. &prompt.user; cat foo df . dd if=/dev/zero of=z bs=1 seek=`echo 2^43 - 2 | bc` count=1 ls -l z du z df . &prompt.user; sh foo Filesystem 1024-blocks Used Avail Capacity Mounted on /dev/da0a 64479 27702 31619 47% / 1+0 records in 1+0 records out 1 bytes transferred in 0.000187 secs (5346 bytes/sec) -rw-r--r-- 1 bde bin 8796093022207 Sep 7 16:04 z 32 z Filesystem 1024-blocks Used Avail Capacity Mounted on /dev/da0a 64479 27734 31587 47% / Bruce Evans, September 1998 Why do I get an error message, archsw.readin.failed after compiling and booting a new kernel? You can boot by specifying the kernel directly at the second stage, pressing any key when the | shows up before loader is started. More specifically, you have upgraded the source for your kernel, and installed a new kernel builtin from them without making world. This is not supported. Make world. How do I upgrade from 3.X -> 4.X? We strongly recommend that you use binary snapshots to do this. 4-STABLE snapshots are available at releng4.FreeBSD.org. If you wish to upgrade using source, please see the FreeBSD Handbook for more information. Upgrading via source is never recommended for new users, and upgading from 3.X to 4.X is even less so; make sure you have read the instructions carefully before attempting to upgrade via source. What are these security profiles? A security profile is a set of configuration options that attempts to achieve the desired ratio of security to convenience by enabling and disabling certain programs and other settings. The more severe the security profile, the less programs will be enabled by default; this is one of the basic principles of security: do not run anything except what you must. Please note that the security profile is just a default setting. All programs can be enabled and disabled after you've installed FreeBSD by editing or adding the appropriate line(s) to /etc/rc.conf. For more information on the latter, please see the &man.rc.conf.5; manual page. Following is a table that describes what each security profile does. The columns are the choices you have for a security profile, and the rows are the program or feature that is enabled or disabled. Possible security profiles Extreme High Moderate Low &man.inetd.8; NO NO YES YES &man.sendmail.8; NO YES YES YES &man.sshd.8; NO YES YES YES &man.portmap.8; NO NO MAYBE The portmapper is enabled if the machine has been configured as an NFS client or server earlier in the installation. YES NFS server NO NO YES YES &man.securelevel.8; YES (2) If you choose a security profile that sets the securelevel (Extreme or High), you must be aware of the implications. Please read the &man.init.8; manual page and pay particular attention to the meanings of the security levels, or you may have significant trouble later! YES (1) NO NO

The security profile is not a silver bullet! Setting it high does not mean you don't have to keep up with security issues by reading an appropriate mailing list, using good passwords and passphrases, and generally adhering to good security practices. It simply sets up the desired security to convenience ration out of the box. The security profile mechanism is meant to be used when you first install FreeBSD. If you already have FreeBSD installed, it would probably be more beneficial to simply enable or disable the desired functionality. If you really want to use a security profile, you can re-run &man.sysinstall.8; to set it. Hardware compatibility What kind of hard drives does FreeBSD support? FreeBSD supports EIDE and SCSI drives (with a compatible controller; see the next section), and all drives using the original Western Digital interface (MFM, RLL, ESDI, and of course IDE). A few ESDI controllers that use proprietary interfaces may not work: stick to WD1002/3/6/7 interfaces and clones. Which SCSI controllers are supported? See the complete list in the Handbook. Which CD-ROM drives are supported by FreeBSD? Any SCSI drive connected to a supported controller is supported. The following proprietary CD-ROM interfaces are also supported: Mitsumi LU002 (8bit), LU005 (16bit) and FX001D (16bit 2x Speed). Sony CDU 31/33A Sound Blaster Non-SCSI CD-ROM Matsushita/Panasonic CD-ROM ATAPI compatible IDE CD-ROMs All non-SCSI cards are known to be extremely slow compared to SCSI drives, and some ATAPI CDROMs may not work. As of 2.2 the FreeBSD CDROM from BSDi supports booting directly from the CD. Which CD-RW drives are supported by FreeBSD? FreeBSD supports any ATAPI-compatible IDE CD-R or CD-RW drive. For FreeBSD versions 4.0 and later, see the man page for &man.burncd.8;. For earlier FreeBSD versions, see the examples in /usr/share/examples/atapi. FreeBSD also supports any SCSI CD-R or CD-RW drives. Install and use the cdrecord command from the ports or packages system, and make sure that you have the pass device compiled in your kernel. Does FreeBSD support ZIP drives? FreeBSD supports the SCSI ZIP drive out of the box, of course. The ZIP drive can only be set to run at SCSI target IDs 5 or 6, but if your SCSI host adapter's BIOS supports it you can even boot from it. It is not clear which host adapters support booting from targets other than 0 or 1, so you will have to consult your adapter's documentation if you'd like to use this feature. ATAPI (IDE) Zip drives are supported in FreeBSD 2.2.6 and later releases. FreeBSD has contained support for Parallel Port Zip Drives since version 3.0. If you are using a sufficiently up to date version, then you should check that your kernel contains the scbus0, da0, ppbus0, and vp0 drivers (the GENERIC kernel contains everything except vp0). With all these drivers present, the Parallel Port drive should be available as /dev/da0s4. Disks can be mounted using mount /dev/da0s4 /mnt OR (for dos disks) mount_msdos /dev/da0s4 /mnt as appropriate. Also check out this note on removable drives, and this note on formatting. Does FreeBSD support JAZ, EZ and other removable drives? Apart from the IDE version of the EZ drive, these are all SCSI devices, so the should all look like SCSI disks to FreeBSD, and the IDE EZ should look like an IDE drive. I'm not sure how well FreeBSD supports changing the media out while running. You will of course need to dismount the drive before swapping media, and make sure that any external units are powered on when you boot the system so FreeBSD can see them. See this note on formatting. Which multi-port serial cards are supported by FreeBSD? There is a list of these in the Miscellaneous devices section of the handbook. Some unnamed clone cards have also been known to work, especially those that claim to be AST compatible. Check the &man.sio.4; man page to get more information on configuring such cards. Does FreeBSD support my USB keyboard? USB device support was added to FreeBSD 3.1. However, it is still in preliminary state and may not always work as of version 3.2. If you want to experiment with the USB keyboard support, follow the procedure described below. Use FreeBSD 3.2 or later. Add the following lines to your kernel configuration file, and rebuild the kernel. device uhci device ohci device usb device ukbd options KBD_INSTALL_CDEV In versions of FreeBSD before 4.0, use this instead: controller uhci0 controller ohci0 controller usb0 controller ukbd0 options KBD_INSTALL_CDEV Go to the /dev directory and create device nodes as follows: &prompt.root; cd /dev &prompt.root; ./MAKEDEV kbd0 kbd1 Edit /etc/rc.conf and add the following lines: usbd_enable="YES" usbd_flags="" After the system is rebooted, the AT keyboard becomes /dev/kbd0 and the USB keyboard becomes /dev/kbd1, if both are connected to the system. If there is the USB keyboard only, it will be /dev/ukbd0. If you want to use the USB keyboard in the console, you have to explicitly tell the console driver to use the existence of the USB keyboard. This can be done by running the following command as a part of system initialization. &prompt.root; kbdcontrol -k /dev/kbd1 < /dev/ttyv0 > /dev/null Note that if the USB keyboard is the only keyboard, it is accessed as /dev/kbd0, thus, the command should look like: &prompt.root; kbdcontrol -k /dev/kbd0 < /dev/ttyv0 > /dev/null /etc/rc.i386 is a good place to add the above command. Once this is done, the USB keyboard should work in the X environment as well without any special settings. Hot-plugging and unplugging of the USB keyboard may not work quite right yet. It is a good idea to connect the keyboard before you start the system and leave it connected until the system is shutdown to avoid troubles. See the &man.ukbd.4; man page for more information. I have an unusual bus mouse. How do I set it up? FreeBSD supports the bus mouse and the InPort bus mouse from such manufactures as Microsoft, Logitech and ATI. The bus device driver is compiled in the GENERIC kernel by default in FreeBSD versions 2.X, but not included in version 3.0 or later. If you are building a custom kernel with the bus mouse driver, make sure to add the following line to the kernel config file In FreeBSD 3.0 or before, add: device mse0 at isa? port 0x23c tty irq5 vector mseintr In FreeBSD 3.X, the line should be: device mse0 at isa? port 0x23c tty irq5 And in FreeBSD 4.X and later, the line should read: device mse0 at isa? port 0x23c irq5 Bus mice usually comes with dedicated interface cards. These cards may allow you to set the port address and the IRQ number other than shown above. Refer to the manual of your mouse and the &man.mse.4; man page for more information. How do I use my PS/2 (mouse port or keyboard) mouse? If you're running a post-2.2.5 version of FreeBSD, the necessary driver, psm, is included and enabled in the kernel. The kernel should detect your PS/2 mouse at boot time. If you're running a previous but relatively recent version of FreeBSD (2.1.x or better) then you can simply enable it in the kernel configuration menu at installation time, otherwise later with at the boot: prompt. It is disabled by default, so you will need to enable it explicitly. If you're running an older version of FreeBSD then you'll have to add the following lines to your kernel configuration file and compile a new kernel. In FreeBSD 3.0 or earlier, the line should be: device psm0 at isa? port "IO_KBD" conflicts tty irq 12 vector psmintr In FreeBSD 3.1 or later, the line should be: device psm0 at isa? tty irq 12 In FreeBSD 4.0 or later, the line should be: device psm0 at atkbdc? irq 12 See the Handbook entry on configuring the kernel if you've no experience with building kernels. Once you have a kernel detecting psm0 correctly at boot time, make sure that an entry for psm0 exists in /dev. You can do this by typing: &prompt.root; cd /dev; sh MAKEDEV psm0 when logged in as root. Is it possible to make use of a mouse in any way outside the X Window system? If you are using the default console driver, syscons, you can use a mouse pointer in text consoles to cut & paste text. Run the mouse daemon, moused, and turn on the mouse pointer in the virtual console: &prompt.root; moused -p /dev/xxxx -t yyyy &prompt.root; vidcontrol -m on Where xxxx is the mouse device name and yyyy is a protocol type for the mouse. See the &man.moused.8; man page for supported protocol types. You may wish to run the mouse daemon automatically when the system starts. In version 2.2.1, set the following variables in /etc/sysconfig. mousedtype="yyyy" mousedport="xxxx" mousedflags="" In versions 2.2.2 to 3.0, set the following variables in /etc/rc.conf. moused_type="yyyy" moused_port="xxxx" moused_flags="" In 3.1 and later, assuming you have a PS/2 mouse, all you need to is add moused_enable="YES" to /etc/rc.conf. In addition, if you would like to be able to use the mouse daemon on all virtual terminals instead of just console at boot-time, add the following to /etc/rc.conf. allscreens_flags="-m on" Staring from FreeBSD 2.2.6, the mouse daemon is capable of determining the correct protocol type automatically unless the mouse is a relatively old serial mouse model. Specify auto the protocol to invoke automatic detection. When the mouse daemon is running, access to the mouse needs to be coordinated between the mouse daemon and other programs such as the X Window. Refer to another section on this issue. How do I cut and paste text with mouse in the text console? Once you get the mouse daemon running (see previous section), hold down the button 1 (left button) and move the mouse to select a region of text. Then, press the button 2 (middle button) or the button 3 (right button) to paste it at the text cursor. In versions 2.2.6 and later, pressing the button 2 will paste the text. Pressing the button 3 will extend the selected region of text. If your mouse does not have the middle button, you may wish to emulate it or remap buttons using moused options. See the &man.moused.8; man page for details. Does FreeBSD support any USB mice? USB device support was added to FreeBSD 3.1. However, it is still in a preliminary state and may not always work as of version 3.2. If you want to experiment with the USB mouse support, follow the procedure described below. Use FreeBSD 3.2 or later. Add the following lines to your kernel configuration file, and rebuild the kernel. device uhci device ohci device usb device ums In versions of FreeBSD before 4.0, use this instead: controller uhci0 controller ohci0 controller usb0 device ums0 Go to the /dev directory and create a device node as follows: &prompt.root; cd /dev &prompt.root; ./MAKEDEV ums0 Edit /etc/rc.conf and add the following lines: moused_enable="YES" moused_type="auto" moused_port="/dev/ums0" moused_flags="" usbd_enable="YES" usbd_flags="" See the previous section for more detailed discussion on moused. In order to use the USB mouse in the X session, edit XF86Config. If you are using XFree86 3.3.2 or later, be sure to have the following lines in the Pointer section: Device "/dev/sysmouse" Protocol "Auto" If you are using earlier versions of XFree86, be sure to have the following lines in the Pointer section: Device "/dev/sysmouse" Protocol "SysMouse" Refer to another section on the mouse support in the X environment. Hot-plugging and unplugging of the USB mouse may not work quite right yet. It is a good idea connect the mouse before you start the system and leave it connected until the system is shutdown to avoid trouble. My mouse has a fancy wheel and buttons. Can I use them in FreeBSD? The answer is, unfortunately, It depends. These mice with additional features require specialized driver in most cases. Unless the mouse device driver or the user program has specific support for the mouse, it will act just like a standard two, or three button mouse. For the possible usage of wheels in the X Window environment, refer to that section. Why does my wheel-equipped PS/2 mouse cause my mouse cursor to jump around the screen? The PS/2 mouse driver psm in FreeBSD versions 3.2 or earlier has difficulty with some wheel mice, including Logitech model M-S48 and its OEM siblings. Apply the following patch to /sys/i386/isa/psm.c and rebuild the kernel. Index: psm.c =================================================================== RCS file: /src/CVS/src/sys/i386/isa/Attic/psm.c,v retrieving revision 1.60.2.1 retrieving revision 1.60.2.2 diff -u -r1.60.2.1 -r1.60.2.2 --- psm.c 1999/06/03 12:41:13 1.60.2.1 +++ psm.c 1999/07/12 13:40:52 1.60.2.2 @@ -959,14 +959,28 @@ sc->mode.packetsize = vendortype[i].packetsize; /* set mouse parameters */ +#if 0 + /* + * A version of Logitech FirstMouse+ won't report wheel movement, + * if SET_DEFAULTS is sent... Don't use this command. + * This fix was found by Takashi Nishida. + */ i = send_aux_command(sc->kbdc, PSMC_SET_DEFAULTS); if (verbose >= 2) printf("psm%d: SET_DEFAULTS return code:%04x\n", unit, i); +#endif if (sc->config & PSM_CONFIG_RESOLUTION) { sc->mode.resolution = set_mouse_resolution(sc->kbdc, - (sc->config & PSM_CONFIG_RESOLUTION) - 1); + (sc->config & PSM_CONFIG_RESOLUTION) - 1); + } else if (sc->mode.resolution >= 0) { + sc->mode.resolution + = set_mouse_resolution(sc->kbdc, sc->dflt_mode.resolution); + } + if (sc->mode.rate > 0) { + sc->mode.rate = set_mouse_sampling_rate(sc->kbdc, sc->dflt_mode.rate); } + set_mouse_scaling(sc->kbdc, 1); /* request a data packet and extract sync. bits */ if (get_mouse_status(sc->kbdc, stat, 1, 3) < 3) { Versions later than 3.2 should be all right. How do I use the mouse/trackball/touchpad on my laptop? Please refer to the answer to the previous question. And check out this note on the Mobile Computing page. What types of tape drives are supported? FreeBSD supports SCSI and QIC-36 (with a QIC-02 interface). This includes 8-mm (aka Exabyte) and DAT drives. Some of the early 8-mm drives are not quite compatible with SCSI-2, and may not work well with FreeBSD. Does FreeBSD support tape changers? FreeBSD 2.2 supports SCSI changers using the &man.ch.4; device and the &man.chio.1; command. The details of how you actually control the changer can be found in the &man.chio.1; man page. If you're not using AMANDA or some other product that already understands changers, remember that they're only know how to move a tape from one point to another, so you need to keep track of which slot a tape is in, and which slot the tape currently in the drive needs to go back to. Which sound cards are supported by FreeBSD? FreeBSD supports the SoundBlaster, SoundBlaster Pro, SoundBlaster 16, Pro Audio Spectrum 16, AdLib and Gravis UltraSound sound cards. There is also limited support for MPU-401 and compatible MIDI cards. Cards conforming to the Microsoft Sound System specification are also supported through the pcm driver. This is only for sound! This driver does not support CD-ROMs, SCSI or joysticks on these cards, except for the SoundBlaster. The SoundBlaster SCSI interface and some non-SCSI CDROMS are supported, but you can't boot off this device. Workarounds for no sound from es1370 with pcm driver? You can run the following command everytime the machine booted up: &prompt.root; mixer pcm 100 vol 100 cd 100 Which network cards does FreeBSD support? See the Ethernet cards section of the handbook for a more complete list. I don't have a math co-processor - is that bad? This will only affect 386/486SX/486SLC owners - other machines will have one built into the CPU. In general this will not cause any problems, but there are circumstances where you will take a hit, either in performance or accuracy of the math emulation code (see the section on FP emulation). In particular, drawing arcs in X will be VERY slow. It is highly recommended that you buy a math co-processor; it's well worth it. Some math co-processors are better than others. It pains us to say it, but nobody ever got fired for buying Intel. Unless you're sure it works with FreeBSD, beware of clones. What other devices does FreeBSD support? See the Handbook for the list of other devices supported. Does FreeBSD support power management on my laptop? FreeBSD supports APM on certain machines. Please look in the LINT kernel config file, searching for the APM keyword. Further information can be found in &man.apm.4;. Why does my Micron system hang at boot time? Certain Micron motherboards have a non-conforming PCI BIOS implementation that causes grief when FreeBSD boots because PCI devices don't get configured at their reported addresses. Disable the Plug and Play Operating System flag in the BIOS to work around this problem. More information can be found at http://cesdis.gsfc.nasa.gov/linux/drivers/vortex.html#micron Why does FreeBSD not recognize my Adaptec SCSI controller card? The newer AIC789x series Adaptec chips are supported under the CAM SCSI framework which made it's debut in 3.0. Patches against 2.2-STABLE are in ftp://ftp.FreeBSD.org/pub/FreeBSD/development/cam/. A CAM-enhanced boot floppy is available at http://people.FreeBSD.org/~abial/cam-boot/. In both cases read the README before beginning. How come FreeBSD cannot find my internal Plug & Play modem? You will need to add the modem's PnP ID to the PnP ID list in the serial driver. To enable Plug & Play support, compile a new kernel with controller pnp0 in the configuration file, then reboot the system. The kernel will print the PnP IDs of all the devices it finds. Copy the PnP ID from the modem to the table in /sys/i386/isa/sio.c, at about line 2777. Look for the string SUP1310 in the structure siopnp_ids[] to find the table. Build the kernel again, install, reboot, and your modem should be found. You may have to manually configure the PnP devices using the pnp command in the boot-time configuration with a command like pnp 1 0 enable os irq0 3 drq0 0 port0 0x2f8 to make the modem show. How do I get the boot: prompt to show on the serial console? Build a kernel with options COMCONSOLE. Create /boot.config and place as the only text in the file. Unplug the keyboard from the system. See /usr/src/sys/i386/boot/biosboot/README.serial for information. Why doesn't my 3Com PCI network card work with my Micron computer? Certain Micron motherboards have a non-conforming PCI BIOS implementation that does not configure PCI devices at the addresses reported. This causes grief when FreeBSD boots. To work around this problem, disable the Plug and Play Operating System flag in the BIOS. More information on this problem is available at URL: http://cesdis.gsfc.nasa.gov/linux/drivers/vortex.html#micron Does FreeBSD support Symmetric Multiprocessing (SMP)? SMP is supported in 3.0-STABLE and later releases only. SMP is not enabled in the GENERIC kernel, so you will have to recompile your kernel to enable SMP. Take a look at /sys/i386/conf/LINT to figure out what options to put in your kernel config file. The boot floppy hangs on a system with an ASUS K7V motherboard. How do I fix this? Go in to the BIOS setup and disable the boot virus protection. Troubleshooting What do I do when I have bad blocks on my hard drive? With SCSI drives, the drive should be capable of re-mapping these automatically. However, many drives are shipped with this feature disabled, for some mysterious reason... To enable this, you'll need to edit the first device page mode, which can be done on FreeBSD by giving the command (as root) &prompt.root; scsi -f /dev/rsd0c -m 1 -e -P 3 and changing the values of AWRE and ARRE from 0 to 1:- AWRE (Auto Write Reallocation Enbld): 1 ARRE (Auto Read Reallocation Enbld): 1 The following paragraphs were submitted by Ted Mittelstaedt: For IDE drives, any bad block is usually a sign of potential trouble. All modern IDE drives come with internal bad-block remapping turned on. All IDE hard drive manufacturers today offer extensive warranties and will replace drives with bad blocks on them. If you still want to attempt to rescue an IDE drive with bad blocks, you can attempt to download the IDE drive manufacturer's IDE diagnostic program, and run this against the drive. Sometimes these programs can be set to force the drive electronics to rescan the drive for bad blocks and lock them out. For ESDI, RLL and MFM drives, bad blocks are a normal part of the drive and are no sign of trouble, generally. With a PC, the disk drive controller card and BIOS handle the task of locking out bad sectors. This is fine for operating systems like DOS that use BIOS code to access the disk. However, FreeBSD's disk driver does not go through BIOS, therefore a mechanism, bad144, exists that replaces this functionality. bad144 only works with the wd driver (which means it is not supported in FreeBSD 4.0), it is NOT able to be used with SCSI. bad144 works by entering all bad sectors found into a special file. One caveat with bad144 - the bad block special file is placed on the last track of the disk. As this file may possibly contain a listing for a bad sector that would occur near the beginning of the disk, where the /kernel file might be located, it therefore must be accessible to the bootstrap program that uses BIOS calls to read the kernel file. This means that the disk with bad144 used on it must not exceed 1024 cylinders, 16 heads, and 63 sectors. This places an effective limit of 500MB on a disk that is mapped with bad144. To use bad144, simply set the Bad Block scanning to ON in the FreeBSD fdisk screen during the initial install. This works up through FreeBSD 2.2.7. The disk must have less than 1024 cylinders. It is generally recommended that the disk drive has been in operation for at least 4 hours prior to this to allow for thermal expansion and track wandering. If the disk has more than 1024 cylinders (such as a large ESDI drive) the ESDI controller uses a special translation mode to make it work under DOS. The wd driver understands about these translation modes, IF you enter the translated geometry with the set geometry command in fdisk. You must also NOT use the dangerously dedicated mode of creating the FreeBSD partition, as this ignores the geometry. Also, even though fdisk will use your overridden geometry, it still knows the true size of the disk, and will attempt to create a too large FreeBSD partition. If the disk geometry is changed to the translated geometry, the partition MUST be manually created with the number of blocks. A quick trick to use is to set up the large ESDI disk with the ESDI controller, boot it with a DOS disk and format it with a DOS partition. Then, boot the FreeBSD install and in the fdisk screen, read off and write down the blocksize and block numbers for the DOS partition. Then, reset the geometry to the same that DOS uses, delete the DOS partition, and create a cooperative FreeBSD partition using the blocksize you recorded earlier. Then, set the partition bootable and turn on bad block scanning. During the actual install, bad144 will run first, before any filesystems are created. (you can view this with an Alt-F2) If it has any trouble creating the badsector file, you have set too large a disk geometry - reboot the system and start all over again (including repartitioning and reformatting with DOS). If remapping is enabled and you are seeing bad blocks, consider replacing the drive. The bad blocks will only get worse as time goes on. How come FreeBSD does not recognize my Bustek 742a EISA SCSI controller? This info is specific to the 742a but may also cover other Buslogic cards. (Bustek = Buslogic) There are 2 general versions of the 742a card. They are hardware revisions A-G, and revisions H - onwards. The revision letter is located after the Assembly number on the edge of the card. The 742a has 2 ROM chips on it, one is the BIOS chip and the other is the Firmware chip. FreeBSD doesn't care what version of BIOS chip you have but it does care about what version of firmware chip. Buslogic will send upgrade ROMS out if you call their tech support dept. The BIOS and Firmware chips are shipped as a matched pair. You must have the most current Firmware ROM in your adapter card for your hardware revision. The REV A-G cards can only accept BIOS/Firmware sets up to 2.41/2.21. The REV H- up cards can accept the most current BIOS/Firmware sets of 4.70/3.37. The difference between the firmware sets is that the 3.37 firmware supports round robin The Buslogic cards also have a serial number on them. If you have a old hardware revision card you can call the Buslogic RMA department and give them the serial number and attempt to exchange the card for a newer hardware revision. If the card is young enough they will do so. FreeBSD 2.1 only supports Firmware revisions 2.21 onward. If you have a Firmware revision older than this your card will not be recognized as a Buslogic card. It may be recognized as an Adaptec 1540, however. The early Buslogic firmware contains an AHA1540 emulation mode. This is not a good thing for an EISA card, however. If you have an old hardware revision card and you obtain the 2.21 firmware for it, you will need to check the position of jumper W1 to B-C, the default is A-B. How come FreeBSD does not detect my HP Netserver's SCSI controller? This is basically a known problem. The EISA on-board SCSI controller in the HP Netserver machines occupies EISA slot number 11, so all the true EISA slots are in front of it. Alas, the address space for EISA slots >= 10 collides with the address space assigned to PCI, and FreeBSD's auto-configuration currently cannot handle this situation very well. So now, the best you can do is to pretend there is no address range clash :), by bumping the kernel option EISA_SLOTS to a value of 12. Configure and compile a kernel, as described in the Handbook entry on configuring the kernel. Of course, this does present you with a chicken-and-egg problem when installing on such a machine. In order to work around this problem, a special hack is available inside UserConfig. Do not use the visual interface, but the plain command-line interface there. Simply type eisa 12 quit at the prompt, and install your system as usual. While it's recommended you compile and install a custom kernel anyway. Hopefully, future versions will have a proper fix for this problem. You can not use a dangerously dedicated disk with an HP Netserver. See this note for more info. What is going on with my CMD640 IDE controller? It's broken. It cannot handle commands on both channels simultaneously. There's a workaround available now and it is enabled automatically if your system uses this chip. For the details refer to the manual page of the disk driver (man 4 wd). If you're already running FreeBSD 2.2.1 or 2.2.2 with a CMD640 IDE controller and you want to use the second channel, build a new kernel with options "CMD640" enabled. This is the default for 2.2.5 and later. I keep seeing messages like ed1: timeout. What do these messages mean? This is usually caused by an interrupt conflict (e.g., two boards using the same IRQ). FreeBSD prior to 2.0.5R used to be tolerant of this, and the network driver would still function in the presence of IRQ conflicts. However, with 2.0.5R and later, IRQ conflicts are no longer tolerated. Boot with the -c option and change the ed0/de0/... entry to match your board. If you're using the BNC connector on your network card, you may also see device timeouts because of bad termination. To check this, attach a terminator directly to the NIC (with no cable) and see if the error messages go away. Some NE2000 compatible cards will give this error if there is no link on the UTP port or if the cable is disconnected. Why do I get Incorrect super block when mounting a CDROM? You have to tell &man.mount.8; the type of the device that you want to mount. By default, &man.mount.8; will assume the filesystem is of type ufs. You want to mount a CDROM filesystem, and you do this by specifying the option to &man.mount.8;. This does, of course, assume that the CDROM contains an ISO 9660 filesystem, which is what most CDROMs have. As of 1.1R, FreeBSD automatically understands the Rock Ridge (long filename) extensions as well. As an example, if you want to mount the CDROM device, /dev/cd0c, under /mnt, you would execute: &prompt.root; mount -t cd9660 /dev/cd0c /mnt Note that your device name (/dev/cd0c in this example) could be different, depending on the CDROM interface. Note that the option just causes the &man.mount.cd9660.8; command to be executed, and so the above example could be shortened to: &prompt.root; mount_cd9660 /dev/cd0c /mnt Why do I get Device not configured when mounting a CDROM? This generally means that there is no CDROM in the CDROM drive, or the drive is not visible on the bus. Feed the drive something, and/or check its master/slave status if it is IDE (ATAPI). It can take a couple of seconds for a CDROM drive to notice that it's been fed, so be patient. Sometimes a SCSI CD-ROM may be missed because it hadn't enough time to answer the bus reset. If you have a SCSI CD-ROM please try to add the following symbol into your kernel configuration file and recompile. options "SCSI_DELAY=15" Why do all non-English characters in filenames show up as ? on my CDs when mounted in FreeBSD? Most likely your CDROM uses the Joilet extention for storing information about files and directories. This extension specifies that all filenames are stored using Unicode two-byte characters. Currently, efforts are under way to introduce a generic Unicode interface into the FreeBSD kernel, but since that isn't ready yet, the CD9660 driver does not have the ability to decode the characters in the filenames. As a temporary solution, starting with FreeBSD 4.3, a special hook has been added into the CD9660 driver to allow the user to load an appropriate conversion table on the fly. Modules for some of the common encodings are available via the sysutils/cd9660_unicode port. My printer is ridiculously slow. What can I do? If it's parallel, and the only problem is that it's terribly slow, try setting your printer port into polled mode: &prompt.root; lptcontrol -p Some newer HP printers are claimed not to work correctly in interrupt mode, apparently due to some (not yet exactly understood) timing problem. Why do my programs occasionally die with Signal 11 errors? Signal 11 errors are caused when your process has attempted to access memory which the operating system has not granted it access to. If something like this is happening at seemingly random intervals then you need to start investigating things very carefully. These problems can usually be attributed to either: If the problem is occurring only in a specific application that you are developing yourself it is probably a bug in your code. If it's a problem with part of the base FreeBSD system, it may also be buggy code, but more often than not these problems are found and fixed long before us general FAQ readers get to use these bits of code (that's what -current is for). In particular, a dead giveaway that this is *not* a FreeBSD bug is if you see the problem when you're compiling a program, but the activity that the compiler is carrying out changes each time. For example, suppose you're running make buildworld, and the compile fails while trying to compile ls.c in to ls.o. If you next run make buildworld again, and the compile fails in the same place then this is a broken build -- try updating your sources and try again. If the compile fails elsewhere then this is almost certainly hardware. What you should do: In the first case you can use a debugger e.g. gdb to find the point in the program which is attempting to access a bogus address and then fix it. In the second case you need to verify that it's not your hardware at fault. Common causes of this include: Your hard disks might be overheating: Check the fans in your case are still working, as your disk (and perhaps other hardware might be overheating). The processor running is overheating: This might be because the processor has been overclocked, or the fan on the processor might have died. In either case you need to ensure that you have hardware running at what it's specified to run at, at least while trying to solve this problem. i.e. Clock it back to the default settings. If you are overclocking then note that it's far cheaper to have a slow system than a fried system that needs replacing! Also the wider community is not often sympathetic to problems on overclocked systems, whether you believe it's safe or not. Dodgy memory: If you have multiple memory SIMMS/DIMMS installed then pull them all out and try running the machine with each SIMM or DIMM individually and narrow the problem down to either the problematic DIMM/SIMM or perhaps even a combination. Over-optimistic Motherboard settings: In your BIOS settings, and some motherboard jumpers you have options to set various timings, mostly the defaults will be sufficient, but sometimes, setting the wait states on RAM too low, or setting the RAM Speed: Turbo option, or similar in the BIOS will cause strange behaviour. A possible idea is to set to BIOS defaults, but it might be worth noting down your settings first! Unclean or insufficient power to the motherboard. If you have any unused I/O boards, hard disks, or CDROMs in your system, try temporarily removing them or disconnecting the power cable from them, to see if your power supply can manage a smaller load. Or try another power supply, preferably one with a little more power (for instance, if your current power supply is rated at 250 Watts try one rated at 300 Watts). You should also read the SIG11 FAQ (listed below) which has excellent explanations of all these problems, albeit from a Linux viewpoint. It also discusses how memory testing software or hardware can still pass faulty memory. Finally, if none of this has helped it is possible that you've just found a bug in FreeBSD, and you should follow the instructions to send a problem report. There's an extensive FAQ on this at the SIG11 problem FAQ Why does the screen go black and lose sync when I boot? This is a known problem with the ATI Mach 64 video card. The problem is that this card uses address 2e8, and the fourth serial port does too. Due to a bug (feature?) in the &man.sio.4; driver it will touch this port even if you don't have the fourth serial port, and even if you disable sio3 (the fourth port) which normally uses this address. Until the bug has been fixed, you can use this workaround: Enter at the bootprompt. (This will put the kernel into configuration mode). Disable sio0, sio1, sio2 and sio3 (all of them). This way the sio driver doesn't get activated -> no problems. Type exit to continue booting. If you want to be able to use your serial ports, you'll have to build a new kernel with the following modification: in /usr/src/sys/i386/isa/sio.c find the one occurrence of the string 0x2e8 and remove that string and the preceding comma (keep the trailing comma). Now follow the normal procedure of building a new kernel. Even after applying these workarounds, you may still find that the X Window System does not work properly. If this is the case, make sure that the XFree86 version you are using is at least XFree86 3.3.3 or higher. This version and upwards has built-in support for the Mach64 cards and even a dedicated X server for those cards. How come FreeBSD uses only 64 MB of RAM when my system has 128 MB of RAM installed? Due to the manner in which FreeBSD gets the memory size from the BIOS, it can only detect 16 bits worth of Kbytes in size (65535 Kbytes = 64MB) (or less... some BIOSes peg the memory size to 16M). If you have more than 64MB, FreeBSD will attempt to detect it; however, the attempt may fail. To work around this problem, you need to use the kernel option specified below. There is a way to get complete memory information from the BIOS, but we don't have room in the bootblocks to do it. Someday when lack of room in the bootblocks is fixed, we'll use the extended BIOS functions to get the full memory information...but for now we're stuck with the kernel option. options "MAXMEM=n" Where n is your memory in Kilobytes. For a 128 MB machine, you'd want to use 131072. Why does FreeBSD 2.0 panic with kmem_map too small!? The message may also be mb_map too small! The panic indicates that the system ran out of virtual memory for network buffers (specifically, mbuf clusters). You can increase the amount of VM available for mbuf clusters by adding: options "NMBCLUSTERS=n" to your kernel config file, where n is a number in the range 512-4096, depending on the number of concurrent TCP connections you need to support. I'd recommend trying 2048 - this should get rid of the panic completely. You can monitor the number of mbuf clusters allocated/in use on the system with netstat -m (see &man.netstat.1;). The default value for NMBCLUSTERS is 512 + MAXUSERS * 16. Why do I get an error reading CMAP busy when rebooting with a new kernel? The logic that attempts to detect an out of date /var/db/kvm_*.db files sometimes fails and using a mismatched file can sometimes lead to panics. If this happens, reboot single-user and do: &prompt.root; rm /var/db/kvm_*.db What does the message ahc0: brkadrint, Illegal Host Access at seqaddr 0x0 mean? This is a conflict with an Ultrastor SCSI Host Adapter. During the boot process enter the kernel configuration menu and disable uha0, which is causing the problem. Why does Sendmail give me an error reading mail loops back to myself? This is answered in the sendmail FAQ as follows:- * I'm getting "Local configuration error" messages, such as: 553 relay.domain.net config error: mail loops back to myself 554 <user@domain.net>... Local configuration error How can I solve this problem? You have asked mail to the domain (e.g., domain.net) to be forwarded to a specific host (in this case, relay.domain.net) by using an MX record, but the relay machine doesn't recognize itself as domain.net. Add domain.net to /etc/sendmail.cw (if you are using FEATURE(use_cw_file)) or add "Cw domain.net" to /etc/sendmail.cf. The current version of the sendmail FAQ is no longer maintained with the sendmail release. It is however regularly posted to comp.mail.sendmail, comp.mail.misc, comp.mail.smail, comp.answers, and news.answers. You can also receive a copy via email by sending a message to mail-server@rtfm.mit.edu with the command send usenet/news.answers/mail/sendmail-faq as the body of the message. Why do full screen applications on remote machines misbehave? The remote machine may be setting your terminal type to something other than the cons25 terminal type required by the FreeBSD console. There are a number of possible work-arounds for this problem: After logging on to the remote machine, set your TERM shell variable to ansi or sco if the remote machine knows about these terminal types. Use a VT100 emulator like screen at the FreeBSD console. screen offers you the ability to run multiple concurrent sessions from one terminal, and is a neat program in its own right. Each screen window behaves like a VT100 terminal, so the TERM variable at the remote end should be set to vt100. Install the cons25 terminal database entry on the remote machine. The way to do this depends on the operating system on the remote machine. The system administration manuals for the remote system should be able to help you here. Fire up an X server at the FreeBSD end and login to the remote machine using an X based terminal emulator such as xterm or rxvt. The TERM variable at the remote host should be set to xterm or vt100. Why does my machine print calcru: negative time...? This can be caused by various hardware and/or software ailments relating to interrupts. It may be due to bugs but can also happen by nature of certain devices. Running TCP/IP over the parallel port using a large MTU is one good way to provoke this problem. Graphics accelerators can also get you here, in which case you should check the interrupt setting of the card first. A side effect of this problem are dying processes with the message SIGXCPU exceeded cpu time limit. For FreeBSD 3.0 and later from Nov 29, 1998 forward: If the problem cannot be fixed otherwise the solution is to set this sysctl variable: &prompt.root; sysctl -w kern.timecounter.method=1 This means a performance impact, but considering the cause of this problem, you probably will not notice. If the problem persists, keep the sysctl set to one and set the NTIMECOUNTER option in your kernel to increasingly large values. If by the time you have reached NTIMECOUNTER=20 the problem isn't solved, interrupts are too hosed on your machine for reliable timekeeping. I see pcm0 not found or my sound card is found as pcm1 but I have device pcm0 in my kernel config file. What is going on? This occurs in FreeBSD 3.x with PCI sound cards. The pcm0 device is reserved exclusively for ISA-based cards so, if you have a PCI card, then you will see this error, and your card will appear as pcm1. You cannot remove the warning by simply changing the line in the kernel config file to device pcm1 as this will result in pcm1 being reserved for ISA cards and your PCI card being found as pcm2 (along with the warning pcm1 not found). If you have a PCI sound card you will also have to make the snd1 device rather than snd0: &prompt.root; cd /dev &prompt.root; ./MAKEDEV snd1 This situation does not arise in FreeBSD 4.x as has a lot of work has been done to make the it more PnP-centric and the pcm0 device is no longer reserved exclusively fo ISA cards Why is my PnP card no longer found (or found as unknown) since upgrading to FreeBSD 4.x? FreeBSD 4.x is now much more PnP-centric and this has had the side effect of some PnP devices (e.g. sound cards and internal modems) not working even though they worked under FreeBSD 3.x. The reasons for this behaviour are explained by the following e-mail, posted to the freebsd-questions mailing list by Peter Wemm, in answer to a question about an internal modem that was no longer found after an upgrade to FreeBSD 4.x (the comments in [] have been added to clarify the context.

The PNP bios preconfigured it [the modem] and left it laying around in port space, so [in 3.x] the old-style ISA probes found it there. Under 4.0, the ISA code is much more PnP-centric. It was possible [in 3.x] for an ISA probe to find a stray device and then for the PNP device id to match and then fail due to resource conflicts. So, it disables the programmable cards first so this double probing cannot happen. It also means that it needs to know the PnP id's for supported PnP hardware. Making this more user tweakable is on the TODO list.

To get the device working again requires finding its PnP id and adding it to the list that the ISA probes use to identify PnP devices. This is obtained using &man.pnpinfo.8; to probe the device, for example this is the output from &man.pnpinfo.8; for an internal modem: &prompt.root; pnpinfo Checking for Plug-n-Play devices... Card assigned CSN #1 Vendor ID PMC2430 (0x3024a341), Serial Number 0xffffffff PnP Version 1.0, Vendor Version 0 Device Description: Pace 56 Voice Internal Plug & Play Modem Logical Device ID: PMC2430 0x3024a341 #0 Device supports I/O Range Check TAG Start DF I/O Range 0x3f8 .. 0x3f8, alignment 0x8, len 0x8 [16-bit addr] IRQ: 4 - only one type (true/edge) [more TAG lines elided] TAG End DF End Tag Successfully got 31 resources, 1 logical fdevs -- card select # 0x0001 CSN PMC2430 (0x3024a341), Serial Number 0xffffffff Logical device #0 IO: 0x03e8 0x03e8 0x03e8 0x03e8 0x03e8 0x03e8 0x03e8 0x03e8 IRQ 5 0 DMA 4 0 IO range check 0x00 activate 0x01 The information you require is in the Vendor ID line at the start of the output. The hexadecimal number in parentheses (0x3024a341 in this example) is the PnP id and the string immediately before this (PMC2430) is a unique ASCII id. This information needs adding to the file /usr/src/sys/isa/sio.c. You should first make a backup of sio.c just in case things go wrong. You will also need it to make the patch to submit with your PR (you are going to submit a PR, aren't you?) then edit sio.c and search for the line static struct isa_pnp_id sio_ids[] = { then scroll down to find the correct place to add the entry for your device. The entries look like this, and are sorted on the ASCII Vendor ID string which should be included in the comment to the right of the line of code along with all (if it will fit) or part of the Device Description from the output of &man.pnpinfo.8;: {0x0f804f3f, NULL}, /* OZO800f - Zoom 2812 (56k Modem) */ {0x39804f3f, NULL}, /* OZO8039 - Zoom 56k flex */ {0x3024a341, NULL}, /* PMC2430 - Pace 56 Voice Internal Modem */ {0x1000eb49, NULL}, /* ROK0010 - Rockwell ? */ {0x5002734a, NULL}, /* RSS0250 - 5614Jx3(G) Internal Modem */ Add the hexadecimal Vendor ID for your device in the correct place, save the file, rebuild your kernel, and reboot. Your device should now be found as an sio device as it was under FreeBSD 3.x Why do I get the error nlist failed when running, for example, top or systat? The problem is that the application you are trying to run is looking for a specific kernel symbol, but, for whatever reason, cannot find it; this error stems from one of two problems: Your kernel and userland are not synchronized (i.e., you built a new kernel but did not do an installworld, or vice versa), and thus the symbol table is different from what the user application thinks it is. If this is the case, simply complete the upgrade process (see /usr/src/UPDATING for the correct sequence). You are not using /boot/loader to load your kernel, but doing it directly from boot2 (see &man.boot.8;). While there is nothing wrong with bypassing /boot/loader, it generally does a better job of making the kernel symbols available to user applications. Why does it take so long to connect to my computer via ssh or telnet? The symptom: there is a long delay between the time the TCP connection is established and the time when the client software asks for a password (or, in &man.telnet.1;'s case, when a login prompt appears). The problem: more likely than not, the delay is caused by the server software trying to resolve the client's IP address into a hostname. Many servers, including the Telnet and SSH servers that come with FreeBSD, do this in order to, among other things, store the hostname in a log file for future reference by the administrator. The remedy: if the problem occurs whenever you connect from your computer (the client) to any server, the problem is with the client; likewise, if the problem only occurs when someone connects to your computer (the server) the problem is with the server. If the problem is with the client, the only remedy is to fix the DNS so the server can resolve it. If this is on a local network, consider it a server problem and keep reading; conversely, if this is on the global Internet, you will most likely need to contact your ISP and ask them to fix it for you. If the problem is with the server, and this is on a local network, you need to configure the server to be able to resolve address-to-hostname queries for your local address range. See the &man.hosts.5; and &man.named.8; manual pages for more information. If this is on the global Internet, the problem may be that your server's resolver is not functioning correctly. To check, try to look up another host--say, www.yahoo.com. If it doesn't work, that's your problem. Why does file: table is full show up repeatedly in dmesg? This error is caused when you have exhausted the number of available file descriptors on your system. The file table in memory is full. The solution: Manually adjust the kern.maxfiles kernel limit setting. &prompt.root; sysctl -w kern.maxfiles=n Adjust n according to your system needs. Each open file, socket, or fifo uses one file descriptor. A large-scale server may easily require tens of thousands of file descriptors (10,000+), depending on the kind and number of services running concurrently. The number of default file descriptors set in the kernel is dictated by the maxusers 32 maxusers line in your kernel config file. Increasing this will proportionally increase kern.maxfiles. You can see what kern.maxfiles is currently set to by: &prompt.root; sysctl kern.maxfiles kern.maxfiles: 1064 Why does the clock on my laptop keep incorrect time? Your laptop has two or more clocks, and FreeBSD has chosen to use the wrong one. Run &man.dmesg.8;, and check for lines that contain Timecounter. The last line printed is the one that FreeBSD chose, and will almost certainly be TSC. &prompt.root; dmesg | grep Timecounter Timecounter "i8254" frequency 1193182 Hz Timecounter "TSC" frequency 595573479 Hz You can confirm this by checking the kern.timecounter.hardware &man.sysctl.3;. &prompt.root; sysctl kern.timecounter.hardware kern.timecounter.hardware: TSC The BIOS may modify the TSC clock—perhaps to change the speed of the processor when running from batteries, or going in to a power saving mode, but FreeBSD is unaware of these adjustments, and appears to gain or lose time. In this example, the i8254 clock is also available, and can be selected by writing its name to the kern.timecounter.hardware &man.sysctl.3;. &prompt.root; sysctl -w kern.timecounter.hardware=i8254 kern.timecounter.hardware: TSC -> i8254 Your laptop should now start keeping more accurate time. To have this change automatically run at boot time, add the following line to /etc/sysctl.conf. kern.timecounter.hardware=i8254 Why does FreeBSD's boot loader display Read error and stop after the BIOS screen? FreeBSD's boot loader is incorrectly recognizing the hard drive's geometry. This must be manually set within fdisk when creating or modifying FreeBSD's slice. The correct drive geometry values can be found within the machine's BIOS. Look for the number of cylinders, heads and sectors for the particular drive. Within &man.sysinstall.8;'s fdisk, hit G to set the drive geometry. A dialog will pop up requesting the number of cylinders, heads and sectors. Type the numbers found from the BIOS seperates by forward slashes. 5000 cylinders, 250 sectors and 60 sectors would be entered as 5000/250/60 Press enter to set the values, and hit W to write the new partition table to the drive. Another operating system destroyed my Boot Manager. How do I get it back? Enter &man.sysinstall.8; and choose Configure, then Fdisk. Select the disk the Boot Manager resided on with the space key. Press W to write changes to the drive. A prompt will appear asking which boot loader to install. Select this, and it will be restored. Commercial Applications This section is still very sparse, though we're hoping, of course, that companies will add to it! :) The FreeBSD group has no financial interest in any of the companies listed here but simply lists them as a public service (and feels that commercial interest in FreeBSD can have very positive effects on FreeBSD's long-term viability). We encourage commercial software vendors to send their entries here for inclusion. See the Vendors page for a longer list. Where can I get an Office Suite for FreeBSD? BSDi offers a FreeBSD native version of VistaSource ApplixWare 5. ApplixWare is a rich full-featured, commercial Office Suite for FreeBSD containing a word processor, spreadsheet, presentation program, vector drawing package, and other applications. You can purchase ApplixWare for FreeBSD here. The Linux version of StarOffice works flawlessly on FreeBSD. The easiest way to install the Linux version of StarOffice is through the FreeBSD Ports collection. Future versions of the open-source OpenOffice suite should work as well. Where can I get Motif for FreeBSD? The Open Group has released the source code to Motif 2.1.30. You can install the open-motif package, or compile it from ports. Refer to the ports section of the Handbook for more information on how to do this. The Open Motif distribution only allows redistribution if it is running on an open source operating system. In addition, there are commercial distributions of the Motif software available. These, however, are not for free, but their license allows them to be used in closed-source software. Contact Apps2go for the least expensive ELF Motif 2.1.20 distribution for FreeBSD (either i386 or Alpha). There are two distributions, the developement edition and the runtime edition (for much less). These distributions includes: OSF/Motif manager, xmbind, panner, wsm. Development kit with uil, mrm, xm, xmcxx, include and Imake files. Static and dynamic ELF libraries (for use with FreeBSD 3.0 and above). Demonstration applets. Be sure to specify that you want the FreeBSD version of Motif when ordering (don't forget to mention the architecture you want too)! Versions for NetBSD and OpenBSD are also sold by Apps2go. This is currently a FTP only download. More info Apps2go WWW page or Sales or Support email addresses. or phone (817) 431 8775 or +1 817 431-8775 Contact Metro Link for an either ELF or a.out Motif 2.1 distribution for FreeBSD. This distribution includes: OSF/Motif manager, xmbind, panner, wsm. Development kit with uil, mrm, xm, xmcxx, include and Imake files. Static and dynamic libraries (specify ELF for use with FreeBSD 3.0 and later; or a.out for use with FreeBSD 2.2.8 and eariler). Demonstration applets. Preformatted man pages. Be sure to specify that you want the FreeBSD version of Motif when ordering! Versions for Linux are also sold by Metro Link. This is available on either a CDROM or for FTP download. Contact Xi Graphics for an a.out Motif 2.0 distribution for FreeBSD. This distribution includes: OSF/Motif manager, xmbind, panner, wsm. Development kit with uil, mrm, xm, xmcxx, include and Imake files. Static and dynamic libraries (for use with FreeBSD 2.2.8 and eariler). Demonstration applets. Preformatted man pages. Be sure to specify that you want the FreeBSD version of Motif when ordering! Versions for BSDI and Linux are also sold by Xi Graphics. This is currently a 4 diskette set... in the future this will change to a unified CD distribution like their CDE. Where can I get CDE for FreeBSD? Xi Graphics used to sell CDE for FreeBSD, but no longer do. KDE is an open source X11 desktop which is similar to CDE in many respects. You might also like the look and feel of xfce. KDE and xfce are both in the ports system. Are there any commercial high-performance X servers? Yes, Xi Graphics and Metro Link sells Accelerated-X product for FreeBSD and other Intel based systems. The Metro Link offering is a high performance X Server that offers easy configuration using the FreeBSD Package suite of tools, support for multiple concurrent video boards and is distributed in binary form only, in a convienent FTP download. Not to mention the Metro Link offering is available at the very reasonable price of $39. Metro Link also sells both ELF and a.out Motif for FreeBSD (see above). More info Metro Link WWW page or Sales or Support email addresses. or phone (954) 938-0283 or +1 954 938-0283 The Xi Graphics offering is a high performance X Server that offers easy configuration, support for multiple concurrent video boards and is distributed in binary form only, in a unified diskette distribution for FreeBSD and Linux. Xi Graphics also offers a high performance X Server taylored for laptop support. There is a free compatibility demo of version 5.0 available. Xi Graphics also sells Motif and CDE for FreeBSD (see above). More info Xi Graphics WWW page or Sales or Support email addresses. or phone (800) 946 7433 or +1 303 298-7478. Are there any Database systems for FreeBSD? Yes! See the Commercial Vendors section of FreeBSD's Web site. Also see the Databases section of the Ports collection. Can I run Oracle on FreeBSD? Yes. The following pages tell you exactly how to setup Linux-Oracle on FreeBSD: http://www.scc.nl/~marcel/howto-oracle.html http://www.lf.net/lf/pi/oracle/install-linux-oracle-on-freebsd User Applications So, where are all the user applications? Please take a look at the ports page for info on software packages ported to FreeBSD. The list currently tops 3400 and is growing daily, so come back to check often or subscribe to the freebsd-announce mailing list for periodic updates on new entries. Most ports should be available for the 2.2, 3.x and 4.x branches, and many of them should work on 2.1.x systems as well. Each time a FreeBSD release is made, a snapshot of the ports tree at the time of release in also included in the ports/ directory. We also support the concept of a package, essentially no more than a gzipped binary distribution with a little extra intelligence embedded in it for doing whatever custom installation work is required. A package can be installed and uninstalled again easily without having to know the gory details of which files it includes. Use the package installation menu in /stand/sysinstall (under the post-configuration menu item) or invoke the &man.pkg.add.1; command on the specific package files you're interested in installing. Package files can usually be identified by their .tgz suffix and CDROM distribution people will have a packages/All directory on their CD which contains such files. They can also be downloaded over the net for various versions of FreeBSD at the following locations: for 2.2.8-RELEASE/2.2.8-STABLE ftp://ftp.FreeBSD.org/pub/FreeBSD/ports/i386/packages-2.2.8/ for 3.X-RELEASE/3.X-STABLE ftp://ftp.FreeBSD.org/pub/FreeBSD/ports/i386/packages-3-stable/ for 4.X-RELEASE/4-STABLE ftp://ftp.FreeBSD.org/pub/FreeBSD/ports/i386/packages-4-stable/ for 5.X-CURRENT ftp://ftp.FreeBSD.org/pub/FreeBSD/ports/i386/packages-5-current or your nearest local mirror site. Note that all ports may not be available as packages since new ones are constantly being added. It is always a good idea to check back periodically to see which packages are available at the ftp.FreeBSD.org master site. Why is /bin/sh so minimal? Why doesn't FreeBSD use bash or another shell? Because POSIX says that there shall be such a shell. The more complicated answer: many people need to write shell scripts which will be portable across many systems. That's why POSIX specifies the shell and utility commands in great detail. Most scripts are written in Bourne shell, and because several important programming interfaces (&man.make.1;, &man.system.3;, &man.popen.3;, and analogues in higher-level scripting languages like Perl and Tcl) are specified to use the Bourne shell to interpret commands. Because the Bourne shell is so often and widely used, it is important for it to be quick to start, be deterministic in its behavior, and have a small memory footprint. The existing implementation is our best effort at meeting as many of these requirements simultaneously as we can. In order to keep /bin/sh small, we have not provided many of the convenience features that other shells have. That's why the Ports Collection includes more featureful shells like bash, scsh, tcsh, and zsh. (You can compare for yourself the memory utilization of all these shells by looking at the VSZ and RSS columns in a ps -u listing.) Where do I find libc.so.3.0? You are trying to run a package built on 2.2 and later on a 2.1.x system. Please take a look at the previous section and get the correct port/package for your system. Why do I get a message reading Error: can't find libc.so.4.0? You accidently downloaded packages meant for 4.X and 5.X systems and attempted to install them on your 2.X or 3.X FreeBSD system. Please download the correct version of the packages. Why does ghostscript give lots of errors with my 386/486SX? You don't have a math co-processor, right? You will need to add the alternative math emulator to your kernel; you do this by adding the following to your kernel config file and it will be compiled in. options GPL_MATH_EMULATE You will need to remove the MATH_EMULATE option when you do this. Why do SCO/iBCS2 applications bomb on socksys? (FreeBSD 3.0 and older only). You first need to edit the /etc/sysconfig (or /etc/rc.conf, see &man.rc.conf.5;) file in the last section to change the following variable to YES: # Set to YES if you want ibcs2 (SCO) emulation loaded at startup ibcs2=NO It will load the ibcs2 kernel module at startup. You'll then need to set up /compat/ibcs2/dev to look like: lrwxr-xr-x 1 root wheel 9 Oct 15 22:20 X0R@ -> /dev/null lrwxr-xr-x 1 root wheel 7 Oct 15 22:20 nfsd@ -> socksys -rw-rw-r-- 1 root wheel 0 Oct 28 12:02 null lrwxr-xr-x 1 root wheel 9 Oct 15 22:20 socksys@ -> /dev/null crw-rw-rw- 1 root wheel 41, 1 Oct 15 22:14 spx You just need socksys to go to /dev/null (see &man.null.4;) to fake the open & close. The code in -CURRENT will handle the rest. This is much cleaner than the way it was done before. If you want the spx driver for a local socket X connection, define SPX_HACK when you compile the system. How do I configure INN (Internet News) for my machine? After installing the inn package or port, an excellent place to start is Dave Barr's INN Page where you'll find the INN FAQ. What version of Microsoft FrontPage should I get? Use the Port, Luke! A pre-patched version of Apache is available in the ports tree. Does FreeBSD support Java? Yes. Please see http://www.FreeBSD.org/java/. Why can't I build this port on my 3.X-STABLE machine? If you're running a FreeBSD version that lags significantly behind -CURRENT or -STABLE, you may need a ports upgrade kit from http://www.FreeBSD.org/ports/. If you are up to date, then someone might have committed a change to the port which works for -CURRENT but which broke the port for -STABLE. Please submit a bug report on this with the &man.send-pr.1; command, since the ports collection is supposed to work for both the -CURRENT and -STABLE branches. Where do I find ld.so? If you want to run some aout applications like Netscape Navigator on an Elf'ened machine such as 3.1-R or later, it would need /usr/libexec/ld.so and some aout libs. They are included in the compat22 distribution. Use /stand/sysinstall or install.sh in the compat22 subdirectory and install it. Also read ERRATAs for 3.1-R and 3.2-R. I updated the sources, now how do I update my installed ports? Unfortunately, there is no easy way to update installed ports. The &man.pkg.version.1; command can be used to generate a script that will update the installed ports with a newer version in the ports tree: &prompt.root; pkg_version > /tmp/myscript The output script must be edited by hand before you use it. Current versions of &man.pkg.version.1; force this by inserting an &man.exit.1; at the beginning of the script. You should save the output of the script, as it will note packages that depend on the one that has been udpated. These may or may not need to be updated as well. The usual case where they need to be updated is that a shared library has changed version numbers, so the ports that used that library need to be rebuilt to use the new version. If your system is up full time, the &man.periodic.8 system can be used to generate a weekly list of ports that might need updating by setting weekly_status_pkg_enable="YES" in /etc/periodic.conf. Kernel Configuration I'd like to customize my kernel. Is it difficult? Not at all! Check out the kernel config section of the Handbook. It is recommended that you make a dated snapshot of your kernel in kernel.YYMMDD after you get it all working, that way if you do something dire the next time you play with your configuration you can boot that kernel instead of having to go all the way back to kernel.GENERIC. This is particularly important if you're now booting off a controller that isn't supported in the GENERIC kernel. My kernel compiles fail because _hw_float is missing. How do I solve this problem? Let me guess. You removed npx0 (see &man.npx.4;) from your kernel configuration file because you don't have a math co-processor, right? Wrong! :-) The npx0 is MANDATORY. Even if you don't have a mathematic co-processor, you must include the npx0 device. Why is my kernel so big (over 10MB)? Chances are, you compiled your kernel in debug mode. Kernels built in debug mode contain many symbols that are used for debugging, thus greatly increasing the size of the kernel. Note that if you running a FreeBSD 3.0 or later system, there will be little or no performance decrease from running a debug kernel, and it is useful to keep one around in case of a system panic. However, if you are running low on disk space, or you simply don't want to run a debug kernel, make sure that both of the following are true: You do not have a line in your kernel configuration file that reads: makeoptions DEBUG=-g You are not running &man.config.8; with the option. Both of the above situations will cause your kernel to be built in debug mode. As long as you make sure you follow the steps above, you can build your kernel normally, and you should notice a fairly large size decrease; most kernels tend to be around 1.5MB to 2MB. Why do I get interrupt conflicts with multi-port serial code? When I compile a kernel with multi-port serial code, it tells me that only the first port is probed and the rest skipped due to interrupt conflicts. How do I fix this? The problem here is that FreeBSD has code built-in to keep the kernel from getting trashed due to hardware or software conflicts. The way to fix this is to leave out the IRQ settings on all but one port. Here is a example: # # Multiport high-speed serial line - 16550 UARTS # device sio2 at isa? port 0x2a0 tty irq 5 flags 0x501 vector siointr device sio3 at isa? port 0x2a8 tty flags 0x501 vector siointr device sio4 at isa? port 0x2b0 tty flags 0x501 vector siointr device sio5 at isa? port 0x2b8 tty flags 0x501 vector siointr Why does every kernel I try to build fail to compile, even GENERIC? There are a number of possible causes for this problem. They are, in no particular order: You are not using the new make buildkernel and make installkernel targets, and your source tree is different from the one used to build the currently running system (e.g., you're compiling 4.3-RELEASE on a 4.0-RELEASE system). If you're attempting an upgrade, please read the /usr/src/UPDATING file, paying particular attention to the COMMON ITEMS section at the end. You are using the new make buildkernel and make installkernel targets, but you failed to assert the completion of the make buildworld target. The make buildkernel target relies on files generated by the make buildworld target to complete its job correctly. Even if you are trying to build FreeBSD-STABLE, it is possible that you fetched the source tree at a time when it was either being modified, or broken for other reasons; only releases are absolutly guaranteed to be buildable, although FreeBSD-STABLE builds fine the majority of the time. If you have not already done so, try re-fetching the source tree and see if the problem goes away. Try using a different server in case the one you're using is having problems. System Administration Where are the system start-up configuration files? From 2.0.5R to 2.2.1R, the primary configuration file is /etc/sysconfig. All the options are to be specified in this file and other files such as /etc/rc (see &man.rc.8;) and /etc/netstart just include it. Look in the /etc/sysconfig file and change the value to match your system. This file is filled with comments to show what to put in there. In post-2.2.1 and 3.0, /etc/sysconfig was renamed to a more self-describing &man.rc.conf.5; file and the syntax cleaned up a bit in the process. /etc/netstart was also renamed to /etc/rc.network so that all files could be copied with a cp /usr/src/etc/rc* /etc command. And, in 3.1 and later, /etc/rc.conf has been moved to /etc/defaults/rc.conf. Do not edit this file! Instead, if there is any entry in /etc/defaults/rc.conf that you want to change, you should copy the line into /etc/rc.conf and change it there. For example, if you wish to start named, the DNS server included with FreeBSD in FreeBSD 3.1 or later, all you need to do is: &prompt.root; echo named_enable="YES" >> /etc/rc.conf To start up local services in FreeBSD 3.1 or later, place shell scripts in the /usr/local/etc/rc.d directory. These shell scripts should be set executable, and end with a .sh. In FreeBSD 3.0 and earlier releases, you should edit the /etc/rc.local file. The /etc/rc.serial is for serial port initialization (e.g. locking the port characteristics, and so on.). The /etc/rc.i386 is for Intel-specifics settings, such as iBCS2 emulation or the PC system console configuration. How do I add a user easily? Use the &man.adduser.8; command. For more complicated usage, the &man.pw.8; command. To remove the user again, use the &man.rmuser.8; command. Once again, &man.pw.8; will work as well. How can I add my new hard disk to my FreeBSD system? See the Disk Formatting Tutorial at www.FreeBSD.org. I have a new removable drive, how do I use it? Whether it's a removable drive like a ZIP or an EZ drive (or even a floppy, if you want to use it that way), or a new hard disk, once it's installed and recognized by the system, and you have your cartridge/floppy/whatever slotted in, things are pretty much the same for all devices. (this section is based on Mark Mayo's ZIP FAQ) If it's a ZIP drive or a floppy , you've already got a DOS filesystem on it, you can use a command like this: &prompt.root; mount -t msdos /dev/fd0c /floppy if it's a floppy, or this: &prompt.root; mount -t msdos /dev/da2s4 /zip for a ZIP disk with the factory configuration. For other disks, see how they're laid out using &man.fdisk.8; or &man.sysinstall.8;. The rest of the examples will be for a ZIP drive on da2, the third SCSI disk. Unless it's a floppy, or a removable you plan on sharing with other people, it's probably a better idea to stick a BSD file system on it. You'll get long filename support, at least a 2X improvement in performance, and a lot more stability. First, you need to redo the DOS-level partitions/filesystems. You can either use &man.fdisk.8; or /stand/sysinstall, or for a small drive that you don't want to bother with multiple operating system support on, just blow away the whole FAT partition table (slices) and just use the BSD partitioning: &prompt.root; dd if=/dev/zero of=/dev/rda2 count=2 &prompt.root; disklabel -Brw da2 auto You can use disklabel or /stand/sysinstall to create multiple BSD partitions. You'll certainly want to do this if you're adding swap space on a fixed disk, but it's probably irrelevant on a removable drive like a ZIP. Finally, create a new file system, this one's on our ZIP drive using the whole disk: &prompt.root; newfs /dev/rda2c and mount it: &prompt.root; mount /dev/da2c /zip and it's probably a good idea to add a line like this to /etc/fstab (see &man.fstab.5;) so you can just type mount /zip in the future: /dev/da2c /zip ffs rw,noauto 0 0 Why do I keep getting messages like root: not found after editing my crontab file? This is normally caused by editing the system crontab (/etc/crontab) and then using &man.crontab.1; to install it: &prompt.root; crontab /etc/crontab This is not the correct way to do things. The system crontab has a different format to the per-user crontabs which &man.crontab.1; updates (the &man.crontab.5; manual page explains the differences in more detail). If this is what you did, the extra crontab is simply a copy of /etc/crontab in the wrong format it. Delete it with the command: &prompt.root; crontab -r Next time, when you edit /etc/crontab, you should not do anything to inform &man.cron.8; of the changes, since it will notice them automatically. If you want something to be run once per day, week, or month, it's probably better to add shell scripts /usr/local/etc/periodic, and let the &man.periodic.8; command run from the system cron schedule it with the other periodic system tasks. The actual reason for the error is that the system crontab has an extra field, specifying which user to run the command as. In the default system crontab provided with FreeBSD, this is root for all entries. When this crontab is used as the root user's crontab (which is not the same as the system crontab), &man.cron.8; assumes the string root is the first word of the command to execute, but no such command exists. Why do I get the error, you are not in the correct group to su root when I try to su to root? This is a security feature. In order to su to root (or any other account with superuser privileges), you must be in the wheel group. If this feature were not there, anybody with an account on a system who also found out root's password would be able to gain superuser level access to the system. With this feature, this isn't strictly true; &man.su.1; will prevent them from even trying to enter the password if they aren't in wheel. To allow someone to su to root, simply put them in the wheel group. I made a mistake in rc.conf, or another startup file, and now I can't edit it because the filesystem is read-only. What should I do? When you get the prompt to enter the shell pathname, simply press ENTER, and run mount / to re-mount the root filesystem in read/write mode. You may also need to run mount -a -t ufs to mount the filesystem where your favourite editor is defined. If your favourite editor is on a network filesystem, you will need to either configure the network manually before you can mount network filesystems, or use an editor which resides on a local filesystem, such as &man.ed.1;. If you intend to use a full screen editor such as &man.vi.1; or &man.emacs.1;, you may also need to run export TERM=cons25 so that these editors can load the correct data from the &man.termcap.5; database. Once you have performed these steps, you can edit /etc/rc.conf as you usually would to fix the syntax error. The error message displayed immediately after the kernel boot messages should tell you the number of the line in the file which is at fault. How do I mount a secondary DOS partition? The secondary DOS partitions are found after ALL the primary partitions. For example, if you have an E partition as the second DOS partition on the second SCSI drive, you need to create the special files for slice 5 in /dev, then mount /dev/da1s5: &prompt.root; cd /dev &prompt.root; sh MAKEDEV da1s5 &prompt.root; mount -t msdos /dev/da1s5 /dos/e Can I mount other foreign filesystems under FreeBSD? Digital UNIX UFS CDROMs can be mounted directly on FreeBSD. Mounting disk partitions from Digital UNIX and other systems that support UFS may be more complex, depending on the details of the disk partitioning for the operating system in question. Linux: 2.2 and later have support for ext2fs partitions. See &man.mount.ext2fs.8; for more information. NT: A read-only NTFS driver exists for FreeBSD. For more information, see this tutorial by Mark Ovens at http://ukug.uk.freebsd.org/~mark/ntfs_install.html. Any other information on this subject would be appreciated. How can I use the NT loader to boot FreeBSD? This procedure is slightly different for 2.2.x and 3.x (with the 3-stage boot) systems. The general idea is that you copy the first sector of your native root FreeBSD partition into a file in the DOS/NT partition. Assuming you name that file something like c:\bootsect.bsd (inspired by c:\bootsect.dos), you can then edit the c:\boot.ini file to come up with something like this: [boot loader] timeout=30 default=multi(0)disk(0)rdisk(0)partition(1)\WINDOWS [operating systems] multi(0)disk(0)rdisk(0)partition(1)\WINDOWS="Windows NT" C:\BOOTSECT.BSD="FreeBSD" C:\="DOS" For 2.2.x systems this procedure assumes that DOS, NT, FreeBSD, or whatever have been installed into their respective fdisk partitions on the same disk. This example was tested on a system where DOS & NT were on the first fdisk partition, and FreeBSD on the second. FreeBSD was also set up to boot from its native partition, not the disk's MBR. Mount a DOS-formatted floppy (if you've converted to NTFS) or the FAT partition, under, say, /mnt. &prompt.root; dd if=/dev/rda0a of=/mnt/bootsect.bsd bs=512 count=1 Reboot into DOS or NT. NTFS users copy the bootsect.bsd and/or the bootsect.lnx file from the floppy to C:\. Modify the attributes (permissions) on boot.ini with: C:\> attrib -s -r c:\boot.ini Edit to add the appropriate entries from the example boot.ini above, and restore the attributes: C:\> attrib +s +r c:\boot.ini If FreeBSD is booting from the MBR, restore it with the DOS fdisk command after you reconfigure them to boot from their native partitions. For FreeBSD 3.x systems the procedure is somewhat simpler. If FreeBSD is installed on the same disk as the NT boot partition simply copy /boot/boot1 to C:\BOOTSECT.BSD However, if FreeBSD is installed on a different disk /boot/boot1 will not work, /boot/boot0 is needed. DO NOT SIMPLY COPY /boot/boot0 INSTEAD OF /boot/boot1, YOU WILL OVERWRITE YOUR PARTITION TABLE AND RENDER YOUR COMPUTER UN-BOOTABLE! /boot/boot0 needs to be installed using sysinstall by selecting the FreeBSD boot manager on the screen which asks if you wish to use a boot manager. This is because /boot/boot0 has the partition table area filled with NULL characters but sysinstall copies the partition table before copying /boot/boot0 to the MBR. When the FreeBSD boot manager runs it records the last OS booted by setting the active flag on the partition table entry for that OS and then writes the whole 512-bytes of itself back to the MBR so if you just copy /boot/boot0 to C:\BOOTSECT.BSD then it writes an empty partition table, with the active flag set on one entry, to the MBR. How do I boot FreeBSD and Linux from LILO? If you have FreeBSD and Linux on the same disk, just follow LILO's installation instructions for booting a non-Linux operating system. Very briefly, these are: Boot Linux, and add the following lines to /etc/lilo.conf: other=/dev/hda2 table=/dev/hda label=FreeBSD (the above assumes that your FreeBSD slice is known to Linux as /dev/hda2; tailor to suit your setup). Then, run lilo as root and you should be done. If FreeBSD resides on another disk, you need to add loader=/boot/chain.b to the LILO entry. For example: other=/dev/dab4 table=/dev/dab loader=/boot/chain.b label=FreeBSD In some cases you may need to specify the BIOS drive number to the FreeBSD boot loader to successfully boot off the second disk. For example, if your FreeBSD SCSI disk is probed by BIOS as BIOS disk 1, at the FreeBSD boot loader prompt you need to specify: Boot: 1:da(0,a)/kernel On FreeBSD 2.2.5 and later, you can configure &man.boot.8; to automatically do this for you at boot time. The Linux+FreeBSD mini-HOWTO is a good reference for FreeBSD and Linux interoperability issues. How do I boot FreeBSD and Linux using BootEasy? Install LILO at the start of your Linux boot partition instead of in the Master Boot Record. You can then boot LILO from BootEasy. If you're running Windows-95 and Linux this is recommended anyway, to make it simpler to get Linux booting again if you should need to reinstall Windows95 (which is a Jealous Operating System, and will bear no other Operating Systems in the Master Boot Record). Will a dangerously dedicated disk endanger my health? The installation procedure allows you to chose two different methods in partitioning your harddisk(s). The default way makes it compatible with other operating systems on the same machine, by using fdisk table entries (called slices in FreeBSD), with a FreeBSD slice that employs partitions of its own. Optionally, one can chose to install a boot-selector to switch between the possible operating systems on the disk(s). The alternative uses the entire disk for FreeBSD, and makes no attempt to be compatible with other operating systems. So why it is called dangerous? A disk in this mode doesn't contain what normal PC utilities would consider a valid fdisk table. Depending on how well they have been designed, they might complain at you once they are getting in contact with such a disk, or even worse, they might damage the BSD bootstrap without even asking or notifying you. In addition, the dangerously dedicated disk's layout is known to confuse many BIOSsen, including those from AWARD (eg. as found in HP Netserver and Micronics systems as well as many others) and Symbios/NCR (for the popular 53C8xx range of SCSI controllers). This isn't a complete list, there are more. Symptoms of this confusion include the read error message printed by the FreeBSD bootstrap when it can't find itself, as well as system lockups when booting. Why have this mode at all then? It only saves a few kbytes of disk space, and it can cause real problems for a new installation. Dangerously dedicated mode's origins lie in a desire to avoid one of the most common problems plaguing new FreeBSD installers - matching the BIOS geometry numbers for a disk to the disk itself. Geometry is an outdated concept, but one still at the heart of the PC's BIOS and its interaction with disks. When the FreeBSD installer creates slices, it has to record the location of these slices on the disk in a fashion that corresponds with the way the BIOS expects to find them. If it gets it wrong, you won't be able to boot. Dangerously dedicated mode tries to work around this by making the problem simpler. In some cases, it gets it right. But it's meant to be used as a last-ditch alternative - there are better ways to solve the problem 99 times out of 100. So, how do you avoid the need for DD mode when you're installing? Start by making a note of the geometry that your BIOS claims to be using for your disks. You can arrange to have the kernel print this as it boots by specifying at the boot: prompt, or using boot -v in the loader. Just before the installer starts, the kernel will print a list of BIOS geometries. Don't panic - wait for the installer to start and then use scrollback to read the numbers. Typically the BIOS disk units will be in the same order that FreeBSD lists your disks, first IDE, then SCSI. When you're slicing up your disk, check that the disk geometry displayed in the FDISK screen is correct (ie. it matches the BIOS numbers); if it's wrong, use the g key to fix it. You may have to do this if there's absolutely nothing on the disk, or if the disk has been moved from another system. Note that this is only an issue with the disk that you're going to boot from; FreeBSD will sort itself out just fine with any other disks you may have. Once you've got the BIOS and FreeBSD agreeing about the geometry of the disk, your problems are almost guaranteed to be over, and with no need for DD mode at all. If, however, you are still greeted with the dreaded read error message when you try to boot, it's time to cross your fingers and go for it - there's nothing left to lose. To return a dangerously dedicated disk for normal PC use, there are basically two options. The first is, you write enough NULL bytes over the MBR to make any subsequent installation believe this to be a blank disk. You can do this for example with &prompt.root; dd if=/dev/zero of=/dev/rda0 count=15 Alternatively, the undocumented DOS feature C:\> fdisk /mbr will to install a new master boot record as well, thus clobbering the BSD bootstrap. How can I add more swap space? The best way is to increase the size of your swap partition, or take advantage of this convenient excuse to add another disk. The general rule of thumb is to have around 2x the swap space as you have main memory. However, if you have a very small amount of main memory you may want to configure swap beyond that. It is also a good idea to configure sufficient swap relative to anticipated future memory upgrades so you do not have to futz with your swap configuration later. Adding swap onto a separate disk makes things faster than simply adding swap onto the same disk. As an example, if you are compiling source located on one disk, and the swap is on another disk, this is much faster than both swap and compile on the same disk. This is true for SCSI disks specifically. When you have several disks, configuring a swap partition on each one is usually beneficial, even if you wind up putting swap on a work disk. Typically, each fast disk in your system should have some swap configured. FreeBSD supports up to 4 interleaved swap devices by default. When configuring multiple swap partitions you generally want to make them all about the same size, but people sometimes make their primary swap parition larger in order to accomodate a kernel core dump. Your primary swap partition must be at least as large as main memory in order to be able to accomodate a kernel core. IDE drives are not able to allow access to both drives on the same channel at the same time (FreeBSD doesn't support mode 4, so all IDE disk I/O is programmed). It is still suggested that you put your swap partition on a separate driver, however: the drives are so cheap, it is not worth worrying about. Swapping over NFS is only recommended if you do not have a local disk to swap to. Swapping over NFS is slow and inefficient in FreeBSD releases prior to 4.x, but reasonably fast in releases greater or equal to 4.0. Even so, it will be limited to the network bandwidth available and puts an additional burden on the NFS server. Here is an example for 64Mb vn-swap (/usr/swap0, though of course you can use any name that you want). Make sure your kernel was built with the line pseudo-device vn 1 #Vnode driver (turns a file into a device) in your config-file. The GENERIC kernel already contains this. create a vn-device &prompt.root; cd /dev &prompt.root; sh MAKEDEV vn0 create a swapfile (/usr/swap0) &prompt.root; dd if=/dev/zero of=/usr/swap0 bs=1024k count=64 set proper permissions on (/usr/swap0) &prompt.root; chmod 0600 /usr/swap0 enable the swap file in /etc/rc.conf swapfile="/usr/swap0" # Set to name of swapfile if aux swapfile desired. reboot the machine To enable the swap file immediately, type &prompt.root; vnconfig -e /dev/vn0b /usr/swap0 swap Why am I having trouble setting up my printer? Please have a look at the Handbook entry on printing. It should cover most of your problem. See the Handbook entry on printing. Some printers require a host-based driver to do any kind of printing. These so-called WinPrinters are not natively supported by FreeBSD. If your printer does not work in DOS or Windows NT 4.0, it is probably a WinPrinter. Your only hope of getting one of these to work is to check if the ports/print/pnm2ppa port supports it. From its package description:

This software creates output using the PPA (printer performance architecture) protocol. This protocol is used by some HP "Windows-only" printers, including the HP Deskjet 820C series, the HP DeskJet 720 series, and the HP DeskJet 1000 series. [...] WWW: http://pnm2ppa.sourceforge.net/

How can I correct the keyboard mappings for my system? The kbdcontrol program has an option to load a keyboard map file. Under /usr/share/syscons/keymaps are a number of map files. Choose the one relevant to your system and load it. &prompt.root; kbdcontrol -l uk.iso Both the /usr/share/syscons/keymaps and the .kbd extension are assumed by &man.kbdcontrol.1;. This can be configured in /etc/sysconfig (or &man.rc.conf.5;). See the appropriate comments in this file. In 2.0.5R and later, everything related to text fonts, keyboard mapping is in /usr/share/examples/syscons. The following mappings are currently supported: Belgian ISO-8859-1 Brazilian 275 keyboard Codepage 850 Brazilian 275 keyboard ISO-8859-1 Danish Codepage 865 Danish ISO-8859-1 French ISO-8859-1 German Codepage 850 German ISO-8859-1 Italian ISO-8859-1 Japanese 106 Japanese 106x Latin American Norwegian ISO-8859-1 Polish ISO-8859-2 (programmer's) Russian Codepage 866 (alternative) Russian koi8-r (shift) Russian koi8-r Spanish ISO-8859-1 Swedish Codepage 850 Swedish ISO-8859-1 Swiss-German ISO-8859-1 United Kingdom Codepage 850 United Kingdom ISO-8859-1 United States of America ISO-8859-1 United States of America dvorak United States of America dvorakx Why do I get messages like: unknown: <PNP0303> can't assign resources on boot? The following is an excerpt from a post to the freebsd-current mailing list.

&a.wollman;, 24 April 2001 The can't assign resources messages indicate that the devices are legacy ISA devices for which a non-PnP-aware driver is compiled into the kernel. These include devices such as keyboard controllers, the programmable interrupt controller chip, and several other bits of standard infrastructure. The resources can't be assigned because there is already a driver using those addresses.

How come I cannot get user quotas to work properly? Don't turn on quotas on /, Put the quota file on the file system that the quotas are to be enforced on. ie: Filesystem Quota file /usr /usr/admin/quotas /home /home/admin/quotas … … What's inappropriate about my ccd? The symptom of this is: &prompt.root; ccdconfig -C ccdconfig: ioctl (CCDIOCSET): /dev/ccd0c: Inappropriate file type or format This usually happens when you are trying to concatenate the c partitions, which default to type unused. The ccd driver requires the underlying partition type to be FS_BSDFFS. Edit the disklabel of the disks you are trying to concatenate and change the types of partitions to 4.2BSD. Why can't I edit the disklabel on my ccd? The symptom of this is: &prompt.root; disklabel ccd0 (it prints something sensible here, so let's try to edit it) &prompt.root; disklabel -e ccd0 (edit, save, quit) disklabel: ioctl DIOCWDINFO: No disk label on disk; use "disklabel -r" to install initial label This is because the disklabel returned by ccd is actually a fake one that is not really on the disk. You can solve this problem by writing it back explicitly, as in: &prompt.root; disklabel ccd0 > /tmp/disklabel.tmp &prompt.root; disklabel -Rr ccd0 /tmp/disklabel.tmp &prompt.root; disklabel -e ccd0 (this will work now) Does FreeBSD support System V IPC primitives? Yes, FreeBSD supports System V-style IPC. This includes shared memory, messages and semaphores. You need to add the following lines to your kernel config to enable them. options SYSVSHM options SYSVSHM # enable shared memory options SYSVSEM # enable for semaphores options SYSVMSG # enable for messaging In FreeBSD 3.2 and later, these options are already part of the GENERIC kernel, which meansthey should already be compiled into your system. Recompile and install your kernel. How do I use sendmail for mail delivery with UUCP? The sendmail configuration that ships with FreeBSD is suited for sites that connect directly to the Internet. Sites that wish to exchange their mail via UUCP must install another sendmail configuration file. Tweaking /etc/sendmail.cf manually is considered something for purists. Sendmail version 8 comes with a new approach of generating config files via some &man.m4.1; preprocessing, where the actual hand-crafted configuration is on a higher abstraction level. You should use the configuration files under /usr/src/usr.sbin/sendmail/cf If you didn't install your system with full sources, the sendmail config stuff has been broken out into a separate source distribution tarball just for you. Assuming you've got your CD-ROM mounted, do: &prompt.root; cd /cdrom/src &prompt.root; cat scontrib.?? | tar xzf - -C /usr/src contrib/sendmail Don't panic, this is only a few hundred kilobytes in size. The file README in the cf directory can serve as a basic introduction to m4 configuration. For UUCP delivery, you are best advised to use the mailertable feature. This constitutes a database that sendmail can use to base its routing decision upon. First, you have to create your .mc file. The directory /usr/src/usr.sbin/sendmail/cf/cf is the home of these files. Look around, there are already a few examples. Assuming you have named your file foo.mc, all you need to do in order to convert it into a valid sendmail.cf is: &prompt.root; cd /usr/src/usr.sbin/sendmail/cf/cf &prompt.root; make foo.cf &prompt.root; cp foo.cf /etc/sendmail.cf A typical .mc file might look like: include(`../m4/cf.m4') VERSIONID(`Your version number') OSTYPE(bsd4.4) FEATURE(nodns) FEATURE(nocanonify) FEATURE(mailertable) define(`UUCP_RELAY', your.uucp.relay) define(`UUCP_MAX_SIZE', 200000) MAILER(local) MAILER(smtp) MAILER(uucp) Cw your.alias.host.name Cw youruucpnodename.UUCP The nodns and nocanonify features will prevent any usage of the DNS during mail delivery. The UUCP_RELAY clause is needed for bizarre reasons, don't ask. Simply put an Internet hostname there that is able to handle .UUCP pseudo-domain addresses; most likely, you will enter the mail relay of your ISP there. Once you've got this, you need this file called /etc/mailertable. A typical example of this gender again: # # makemap hash /etc/mailertable.db < /etc/mailertable # horus.interface-business.de uucp-dom:horus .interface-business.de uucp-dom:if-bus interface-business.de uucp-dom:if-bus .heep.sax.de smtp8:%1 horus.UUCP uucp-dom:horus if-bus.UUCP uucp-dom:if-bus . uucp-dom: As you can see, this is part of a real-life file. The first three lines handle special cases where domain-addressed mail should not be sent out to the default route, but instead to some UUCP neighbor in order to shortcut the delivery path. The next line handles mail to the local Ethernet domain that can be delivered using SMTP. Finally, the UUCP neighbors are mentioned in the .UUCP pseudo-domain notation, to allow for a uucp-neighbor !recipient override of the default rules. The last line is always a single dot, matching everything else, with UUCP delivery to a UUCP neighbor that serves as your universal mail gateway to the world. All of the node names behind the uucp-dom: keyword must be valid UUCP neighbors, as you can verify using the command uuname. As a reminder that this file needs to be converted into a DBM database file before being usable, the command line to accomplish this is best placed as a comment at the top of the mailertable. You always have to execute this command each time you change your mailertable. Final hint: if you are uncertain whether some particular mail routing would work, remember the option to sendmail. It starts sendmail in address test mode; simply enter 0, followed by the address you wish to test for the mail routing. The last line tells you the used internal mail agent, the destination host this agent will be called with, and the (possibly translated) address. Leave this mode by typing Control-D. &prompt.user; sendmail -bt ADDRESS TEST MODE (ruleset 3 NOT automatically invoked) Enter <ruleset> <address> > 0 foo@interface-business.de rewrite: ruleset 0 input: foo @ interface-business . de ... rewrite: ruleset 0 returns: $# uucp-dom $@ if-bus $: foo \ < @ interface-business . de > > ^D How do I set up mail with a dialup connection to the 'net? If you've got a statically assigned IP number, you should not need to adjust anything from the default. Set your host name up as your assigned internet name and sendmail will do the rest. If you've got a dynamically assigned IP number and use a dialup ppp connection to the internet, you will probably be given a mailbox on your ISPs mail server. Lets assume your ISPs domain is myISP.com, and that your user name is user. Lets also assume you've called your machine bsd.home and that your ISP has told you that you may use relay.myISP.com as a mail relay. In order to retrieve mail from your mailbox, you'll need to install a retrieval agent. Fetchmail is a good choice as it supports many different protocols. Usually, POP3 will be provided by your ISP. If you've chosen to use user-ppp, you can automatically fetch your mail when a connection to the 'net is established with the following entry in /etc/ppp/ppp.linkup: MYADDR: !bg su user -c fetchmail If you are using sendmail (as shown below) to deliver mail to non-local accounts, put the command !bg su user -c "sendmail -q" after the above shown entry. This forces sendmail to process your mailqueue as soon as the connection to the 'net is established. I'm assuming that you have an account for user on bsd.home. In the home directory of user on bsd.home, create a .fetchmailrc file: poll myISP.com protocol pop3 fetchall pass MySecret Needless to say, this file should not be readable by anyone except user as it contains the password MySecret. In order to send mail with the correct from: header, you must tell sendmail to use user@myISP.com rather than user@bsd.home. You may also wish to tell sendmail to send all mail via relay.myISP.com, allowing quicker mail transmission. The following .mc file should suffice: VERSIONID(`bsd.home.mc version 1.0') OSTYPE(bsd4.4)dnl FEATURE(nouucp)dnl MAILER(local)dnl MAILER(smtp)dnl Cwlocalhost Cwbsd.home MASQUERADE_AS(`myISP.com')dnl FEATURE(allmasquerade)dnl FEATURE(masquerade_envelope)dnl FEATURE(nocanonify)dnl FEATURE(nodns)dnl define(`SMART_HOST', `relay.myISP.com') Dmbsd.home define(`confDOMAIN_NAME',`bsd.home')dnl define(`confDELIVERY_MODE',`deferred')dnl Refer to the previous section for details of how to turn this .mc file into a sendmail.cf file. Also, don't forget to restart sendmail after updating sendmail.cf. What's this UID 0 toor account? Have I been compromised? Don't worry. toor is an alternative superuser account (toor is root spelt backwards). Previously it was created when the &man.bash.1; shell was installed but now it is created by default. It is intended to be used with a non-standard shell so you don't have to change root's default shell. This is important as shells which aren't part of the base distribution (for example a shell installed from ports or packages) are likely be to be installed in /usr/local/bin which, by default, resides on a different filesystem. If root's shell is located in /usr/local/bin and /usr (or whatever filesystem contains /usr/local/bin) isn't mounted for some reason, root won't be able to log in to fix a problem (although if you reboot into single user mode you'll be prompted for the path to a shell). Some people use toor for day-to-day root tasks with a non-standard shell, leaving root, with a standard shell, for single user mode or emergencies. By default you can't log in using toor as it doesn't have a password, so log in as root and set a password for toor if you want to use it. I've fogotten the root password! What do I do? Don't Panic! Simply restart the system, type boot -s at the Boot: prompt (just -s for FreeBSD releases before 3.2) to enter Single User mode. At the question about the shell to use, hit ENTER. You'll be dropped to a &prompt.root; prompt. Enter mount -u / to remount your root filesystem read/write, then run mount -a to remount all the filesystems. Run passwd root to change the root password then run &man.exit.1; to continue booting. How do I keep Control-Alt-Delete from rebooting the system? If you are using syscons (the default console driver) in FreeBSD 2.2.7-RELEASE or later, build and install a new kernel with the line options SC_DISABLE_REBOOT in the configuration file. If you use the PCVT console driver in FreeBSD 2.2.5-RELEASE or later, use the following kernel configuration line instead: options PCVT_CTRL_ALT_DEL For older versions of FreeBSD, edit the keymap you are using for the console and replace the boot keywords with nop. The default keymap is /usr/share/syscons/keymaps/us.iso.kbd. You may have to instruct /etc/rc.conf to load this keymap explicitly for the change to take effect. Of course if you are using an alternate keymap for your country, you should edit that one instead. How do I reformat DOS text files to UNIX ones? Simply use this perl command: &prompt.user; perl -i.bak -npe 's/\r\n/\n/g' file ... file is the file(s) to process. The modification is done in-place, with the original file stored with a .bak extension. Alternatively you can use the &man.tr.1; command: &prompt.user; tr -d '\r' < dos-text-file > unix-file dos-text-file is the file containing DOS text while unix-file will contain the converted output. This can be quite a bit faster than using perl. How do I kill processes by name? Use &man.killall.1;. Why is su bugging me about not being in root's ACL? The error comes from the Kerberos distributed authentication system. The problem isn't fatal but annoying. You can either run su with the -K option, or uninstall Kerberos as described in the next question. How do I uninstall Kerberos? To remove Kerberos from the system, reinstall the bin distribution for the release you are running. If you have the CDROM, you can mount the cd (we'll assume on /cdrom) and run &prompt.root; cd /cdrom/bin &prompt.root; ./install.sh Alternately, you can remove all "MAKE_KERBEROS" options from /etc/make.conf and rebuild world. How do I add pseudoterminals to the system? If you have lots of telnet, ssh, X, or screen users, you'll probably run out of pseudoterminals. Here's how to add more: Build and install a new kernel with the line pseudo-device pty 256 in the configuration file. Run the commands &prompt.root; cd /dev &prompt.root; sh MAKEDEV pty{1,2,3,4,5,6,7} to make 256 device nodes for the new terminals. Edit /etc/ttys and add lines for each of the 256 terminals. They should match the form of the existing entries, i.e. they look like ttyqc none network The order of the letter designations is tty[pqrsPQRS][0-9a-v], using a regular expression. Reboot the system with the new kernel and you're ready to go. How come I cannot create the snd0 device? There is no snd device. The name is used as a shorthand for the various devices that make up the FreeBSD sound driver, such as mixer, sequencer, and dsp. To create these devices you should &prompt.root; cd /dev &prompt.root; sh MAKEDEV snd0 How do I re-read /etc/rc.conf and re-start /etc/rc without a reboot? Go into single user mode and than back to multi user mode. On the console do: &prompt.root; shutdown now (Note: without -r or -h) &prompt.root; return &prompt.root; exit What is a sandbox? Sandbox is a security term. It can mean two things: A process which is placed inside a set of virtual walls that are designed to prevent someone who breaks into the process from being able to break into the wider system. The process is said to be able to play inside the walls. That is, nothing the process does in regards to executing code is supposed to be able to breech the walls so you do not have to do a detailed audit of its code to be able to say certain things about its security. The walls might be a userid, for example. This is the definition used in the security and named man pages. Take the ntalk service, for example (see /etc/inetd.conf). This service used to run as userid root. Now it runs as userid tty. The tty user is a sandbox designed to make it more difficult for someone who has successfully hacked into the system via ntalk from being able to hack beyond that user id. A process which is placed inside a simulation of the machine. This is more hard-core. Basically it means that someone who is able to break into the process may believe that he can break into the wider machine but is, in fact, only breaking into a simulation of that machine and not modifying any real data. The most common way to accomplish this is to build a simulated environment in a subdirectory and then run the processes in that directory chroot'd (i.e. / for that process is this directory, not the real / of the system). Another common use is to mount an underlying filesystem read-only and then create a filesystem layer on top of it that gives a process a seemingly writeable view into that filesystem. The process may believe it is able to write to those files, but only the process sees the effects - other processes in the system do not, necessarily. An attempt is made to make this sort of sandbox so transparent that the user (or hacker) does not realize that he is sitting in it. UNIX implements two core sanboxes. One is at the process level, and one is at the userid level. Every UNIX process is completely firewalled off from every other UNIX process. One process can not modify the address space of another. This is unlike Windows where a process can easily overwrite the address space of any other, leading to a crash. A UNIX process is owned by a patricular userid. If the userid is not the root user, it serves to firewall the process off from processes owned by other users. The userid is also used to firewall off on-disk data. What is securelevel? The securelevel is a security mechanism implemented in the kernel. Basically, when the securelevel is positive, the kernel restricts certain tasks; not even the superuser (i.e., root) is allowed to do them. At the time of this writing, the securelevel mechanism is capable of, among other things, limiting the ability to, unset certain file flags, such as schg (the system immutable flag), write to kernel memory via /dev/mem and /dev/kmem, load kernel modules, and alter &man.ipfirewall.4; rules. To check the status of the securelevel on a running system, simply execute the following command: &prompt.root; sysctl kern.securelevel The output will contain the name of the &man.sysctl.8; variable (in this case, kern.securelevel) and a number. The latter is the current value of the securelevel. If it is positive (i.e., greater than 0), at least some of the securelevel's protections are enabled. You cannot lower the securelevel of a running system; being able to do that would defeat its purpose. If you need to do a task that requires that the securelevel be non-positive (e.g., an installworld or changing the date), you will have to change the securelevel setting in /etc/rc.conf (you want to look for the kern_securelevel and kern_securelevel_enable variables) and reboot. For more information on securelevel and the specific things all the levels do, please consult the &man.init.8; manual page. Securelevel is not a silver bullet; it has many known deficiencies. More often than not, it provides a false sense of security. One of its biggest problems is that in order for it to be at all effective, all files used in the boot process up until the securelevel is set must be protected. If an attacker can get the system to execute their code prior to the securelevel being set (which happens quite late in the boot process since some things the system must do at start-up cannot be done at an elevated securelevel), its protections are invalidated. While this task of protecting all files used in the boot process is not technically impossible, if it is achieved, system maintenance will become a nightmare since one would have to take the system down, at least to single-user mode, to modify a configuration file. This point and others are often discussed on the mailing lists, particuarly freebsd-security. Please search the archives here for an extensive discussion. Some people are hopeful that securelevel will soon go away in favor of a more fine-grained mechanism, but things are still hazy in this respect. Consider yourself warned. How do I let ordinary users mount floppies, CDROMs and other removable media? Ordinary users can be permitted to mount devices. Here is how: As root set the sysctl variable vfs.usermount to 1. &prompt.root; sysctl -w vfs.usermount=1 As root assign the appropriate permissions to the block device associated with the removable media. For example, to allow users to mount the first floppy drive, use: &prompt.root; chmod 666 /dev/fd0 To allow users in the group operator to mount the cdrom drive, use: &prompt.root; chgrp operator /dev/cd0c &prompt.root; chmod 640 /dev/cd0c Finally, add the line vfs.usermount=1 to the file /etc/sysctl.conf so that it is reset at system boot time. All users can now mount the floppy /dev/fd0 onto a directory that they own: &prompt.user; mkdir ~/my-mount-point &prompt.user; mount -t msdos /dev/fd0 ~/my-mount-point Users in group operator can now mount the cdrom /dev/cd0c onto a directory that they own: &prompt.user; mkdir ~/my-mount-point &prompt.user; mount -t msdos /dev/cd0c ~/my-mount-point Unmounting the device is simple: &prompt.user; umount ~/my-mount-point Enabling vfs.usermount, however, has negative security implications. A better way to access MSDOS formatted media is to use the mtools package in the ports collection. How do I move my system over to my huge new disk? The best way is to reinstall the OS on the new disk, then move the user data over. This is highly recommended if you've been tracking -stable for more than one release, or have updated a release instead of installing a new one. You can install booteasy on both disks with &man.boot0cfg.8;, and dual boot them until you are happy with the new configuration. Skip the next paragraph to find out how to move the data after doing this. Should you decide not to do a fresh install, you need to partition and label the new disk with either /stand/sysinstall, or &man.fdisk.8; and &man.disklabel.8;. You should also install booteasy on both disks with &man.boot0cfg.8;, so that you can dual boot to the old or new system after the copying is done. See the formatting-media tutorial for details on this process. Now you've got the new disk set up, and are ready to move the data. Unfortunately, you can't just blindly copy the data. Things like device files (in /dev) and symbolic links tend to screw that up. You need to use tools that understand these things, which means &man.dump.8; and &man.tar.1;. Although it is suggested that you move the data in single user mode, it is not required. You should never use anything but &man.dump.8; and &man.restore.8; to move the root file system. The &man.tar.1; command may work - then again, it may not. You should also use &man.dump.8; and &man.restore.8; if you are moving a single partition to another empty partition. The sequence of steps to use dump to move a partitions data to a new partition is: newfs the new partition. mount it on a temporary mount point. cd to that directory. dump the old partition, piping output to the new one. For example, if you are going to move root to /dev/ad1s1a, with /mnt as the temporary mount point, it's: &prompt.root; newfs /dev/ad1s1a &prompt.root; mount /dev/ad1s1a &prompt.root; cd /mnt &prompt.root; dump 0uaf - / | restore xf - If you are going to rearrange your partitions - say, splitting one into two, or combing two into one, you may find yourself needing to move everything under a subdirectory to a new location. Since &man.dump.8; works with file systems, it can't do this. So you use &man.tar.1;. The general command to move /old to /new for &man.tar.1; is: &prompt.root; (cd /old; tar cf - .) | (cd /new; tar xpf -) If /old has file systems mounted on that, and you don't want to move that data or unmount them, you just add the 'l' flag to the first &man.tar.1;: &prompt.root; (cd /old; tar clf - .) | (cd /new; tar xpf -). You might prefer &man.cpio.1;, &man.pax.1;, or cpdup (in ports/sysutils/cpdup) to &man.tar.1;. I tried to update my system to the latest -STABLE, but got -RC or -BETA! What's going on? Short answer: it's just a name. RC stands for Release Candidate. It signifies that a release is imminent. In FreeBSD, -BETA is typically synonymous with the code freeze before a release. Long answer: FreeBSD derives its releases from one of two places. Major, dot-zero, releases, such as 3.0-RELEASE and 4.0-RELEASE, are branched from the head of the development stream, commonly referred to as -CURRENT. Minor releases, such as 3.1-RELEASE or 4.2-RELEASE, have been snapshots of the active -STABLE branch. Starting with 4.3-RELEASE, each release also now has its own branch which can be tracked by people requiring an extremely conservative rate of development (typically only security advisories). When a release is about to be made, the branch from which it will be derived from has to undergo a certain process. Part of this process is a code freeze. When a code freeze is initiated, the name of the branch is changed to reflect that it's about to become a release. For example, if the branch used to be called 4.0-STABLE, its name will be changed to 4.1-BETA to signify the code freeze and signify that extra pre-release testing should be happening. Bug fixes can still be committed to be part of the release. When the source code is in shape for the release the name will be chagned to 4.1-RC to signify that a release is about to be made from it. Once in the RC stage, only the most critical bugs found can be fixed. Once the release, 4.1-RELEASE in this example, has been made, the branch will be renamed to 4.1-STABLE. I tried to install a new kernel, and the chflags failed. How do I get around this? Short answer: You're probably at security level greater than 0. Reboot directly to single user mode to install the kernel. Long answer: FreeBSD disallows changing system flags at security levels greater than 0. You can check your security level with the command: &prompt.root; sysctl kern.securelevel You cannot lower the security level; you have to boot to single mode to install the kernel, or change the security in /etc/rc.conf then reboot. See the &man.init.8; man page for details on securelevel, and see /etc/defaults/rc.conf and the &man.rc.conf.5; man page for more information on rc.conf. I can't change the time on my system by more than one second! How do I get around this? Short answer: You're probably at security level greater than 1. Reboot directly to single user mode to change the date. Long answer: FreeBSD disallows changing the time by more that one second at security levels greater than 1. You can check your security level with the command: &prompt.root; sysctl kern.securelevel You cannot lower the security level; you have to boot to single mode to change the date, or change the security level in /etc/rc.conf then reboot. See the &man.init.8; man page for details on securelevel, and see /etc/defaults/rc.conf and the &man.rc.conf.5; man page for more information on rc.conf. Why is rpc.statd using 256 megabytes of memory? No, there is no memory leak, and it's not using 256 Mbytes of memory. It simply likes to (i.e., always does) map an obscene amount of memory into its address space for convenience. There is nothing terribly wrong with this from a technical standpoint; it just throws off things like &man.top.1; and &man.ps.1;. &man.rpc.statd.8; maps its status file (resident on /var) into its address space; to save worrying about remapping it later when it needs to grow, it maps it with a generious size. This is very evident from the source code, where one can see that the length argument to &man.mmap.2; is 0x10000000, or one sixteenth of the address space on an IA32, or exactly 256MB. Why can't I unset the schg file flag? You're running at an elevated (i.e., greater than 0) securelevel. Lower the securelevel and try again. For more information, see the FAQ entry on securelevel and the &man.init.8; manual page. Why doesn't SSH authentication through .shosts work by default in recent versions of FreeBSD? The reason why .shosts authentication does not work by default in more recent versions of FreeBSD is because &man.ssh.1; is not installed suid root by default. To fix this, you can do one of the following: As a permanent fix, set ENABLE_SUID_SSH to true in /etc/make.conf and rebuild ssh (or run make world). As a temporary fix, change the mode on /usr/bin/ssh to 4555 by running chmod 4755 /usr/bin/ssh as root. Then add ENABLE_SUID_SSH= true to /etc/make.conf so the change takes effect the next time make world is run. The X Window System and Virtual Consoles I want to run X, how do I go about it? The easiest way is to simply specify that you want to run X during the installation process. Then read and follow the documentation on the xf86config tool, which assists you in configuring XFree86(tm) for your particular graphics card/mouse/etc. You may also wish to investigate the Xaccel server. See the section on Xi Graphics or Metro Link for more details. I tried to run X, but I get an KDENABIO failed (Operation not permitted) error when I type startx. What do I do now? Your system is running at a raised securelevel, isn't it? It is, indeed, impossible to start X at a raised securelevel. To see why, look at the &man.init.8; man page. So the question is what else you should do instead, and you basically have two choices: set your securelevel back down to zero (usually from /etc/rc.conf), or run &man.xdm.1; at boot time (before the securelevel is raised). See for more information about running &man.xdm.1; at boot time. Why doesn't my mouse work with X? If you are using syscons (the default console driver), you can configure FreeBSD to support a mouse pointer on each virtual screen. In order to avoid conflicting with X, syscons supports a virtual device called /dev/sysmouse. All mouse events received from the real mouse device are written to the sysmouse device via moused. If you wish to use your mouse on one or more virtual consoles, and use X, see and set up moused. Then edit /etc/XF86Config and make sure you have the following lines. Section Pointer Protocol "SysMouse" Device "/dev/sysmouse" ..... The above example is for XFree86 3.3.2 or later. For earlier versions, the Protocol should be MouseSystems. Some people prefer to use /dev/mouse under X. To make this work, /dev/mouse should be linked to /dev/sysmouse (see &man.sysmouse.4;): &prompt.root; cd /dev &prompt.root; rm -f mouse &prompt.root; ln -s sysmouse mouse My mouse has a fancy wheel. Can I use it in X? Yes. But you need to customize X client programs. See Colas Nahaboo's web page (http://www.inria.fr/koala/colas/mouse-wheel-scroll/) . If you want to use the imwheel program, just follow these simple steps. Translate the Wheel Events The imwheel program works by translating mouse button 4 and mouse button 5 events into key events. Thus, you have to get the mouse driver to translate mouse wheel events to button 4 and 5 events. There are two ways of doing this, the first way is to have &man.moused.8; do the translation. The second way is for the X server itself to do the event translation. Using &man.moused.8; to Translate Wheel Events To have &man.moused.8; perform the event translations, simply add to the command line used to start &man.moused.8;. For example, if you normally start &man.moused.8; via moused -p /dev/psm0 you would start it by entering moused -p /dev/psm0 -z 4 instead. If you start &man.moused.8; automatically during bootup via /etc/rc.conf, you can simply add to the moused_flags variable in /etc/rc.conf. You now need to tell X that you have a 5 button mouse. To do this, simply add the line Buttons 5 to the Pointer section of /etc/XF86Config. For example, you might have the following Pointer section in /etc/XF86Config. <quote>Pointer</quote> Section for Wheeled Mouse in XFree86 3.3.x series XF86Config with moused Translation Section "Pointer" Protocol "SysMouse" Device "/dev/sysmouse" Buttons 5 EndSection <quote>InputDevice</quote> Section for Wheeled Mouse in XFree86 4.x series XF86Config with automatic protocol recognition and button mapping Translation Section "InputDevice" Identifier "Mouse1" Driver "mouse" Option "Protocol" "auto" Option "Device" "/dev/psm0" Option "Buttons" "5" Option "ZAxisMapping" "4 5" EndSection <quote>.emacs</quote> example for naive page scrolling with Wheeled Mouse ;; wheel mouse (global-set-key [mouse-4] 'scroll-down) (global-set-key [mouse-5] 'scroll-up) Using Your X Server to Translate the Wheel Events If you aren't running &man.moused.8;, or if you don't want &man.moused.8; to translate your wheel events, you can have the X server do the event translation instead. This requires a couple of modifications to your /etc/XF86Config file. First, you need to choose the proper protocol for your mouse. Most wheeled mice use the IntelliMouse protocol. However, XFree86 does support other protocols, such as MouseManPlusPS/2 for the Logitech MouseMan+ mice. Once you have chosen the protocol you will use, you need to add a Protocol line to the Pointer section. Secondly, you need to tell the X server to remap wheel scroll events to mouse buttons 4 and 5. This is done with the ZAxisMapping option. For example, if you aren't using &man.moused.8;, and you have an IntelliMouse attached to the PS/2 mouse port you would use the following in /etc/XF86Config. <quote>Pointer</quote> Section for Wheeled Mouse in <filename>XF86Config</filename> with X Server Translation Section "Pointer" Protocol "IntelliMouse" Device "/dev/psm0" ZAxisMapping 4 5 EndSection Install imwheel Next, install imwheel from the Ports collection. It can be found in the x11 category. This program will map the wheel events from your mouse into keyboard events. For example, it might send Page Up to a program when you scroll the wheel forwards. Imwheel uses a configuration file to map the wheel events to keypresses so that it can send different keys to different applications. The default imwheel configuration file is installed in /usr/X11R6/etc/imwheelrc. You can copy it to ~/.imwheelrc and then edit it if you wish to customize imwheel's configuration. The format of the configuration file is documented in &man.imwheel.1;. Configure Emacs to Work with Imwheel (optional) If you use emacs or Xemacs, then you need to add a small section to your ~/.emacs file. For emacs, add the following: <application>Emacs</application> Configuration for <application>Imwheel</application> ;;; For imwheel (setq imwheel-scroll-interval 3) (defun imwheel-scroll-down-some-lines () (interactive) (scroll-down imwheel-scroll-interval)) (defun imwheel-scroll-up-some-lines () (interactive) (scroll-up imwheel-scroll-interval)) (global-set-key [?\M-\C-\)] 'imwheel-scroll-up-some-lines) (global-set-key [?\M-\C-$] 'imwheel-scroll-down-some-lines) ;;; end imwheel section For Xemacs, add the following to your ~/.emacs file instead: <application>Xemacs</application> Configuration for <application>Imwheel</application> ;;; For imwheel (setq imwheel-scroll-interval 3) (defun imwheel-scroll-down-some-lines () (interactive) (scroll-down imwheel-scroll-interval)) (defun imwheel-scroll-up-some-lines () (interactive) (scroll-up imwheel-scroll-interval)) (define-key global-map [(control meta $)] 'imwheel-scroll-up-some-lines) (define-key global-map [(control meta \()] 'imwheel-scroll-down-some-lines) ;;; end imwheel section Run Imwheel You can just type imwheel in an xterm to start it up once it is installed. It will background itself and take effect immediately. If you want to always use imwheel, simply add it to your .xinitrc or .xsession file. You can safely ignore any warnings imwheel displays about PID files. Those warnings only apply to the Linux version of imwheel. Try turning off the Num Lock key. If your Num Lock key is on by default at boot-time, you may add the following line in the Keyboard section of the XF86Config file. # Let the server do the NumLock processing. This should only be # required when using pre-R6 clients ServerNumLock What is a virtual console and how do I make more? Virtual consoles, put simply, enable you to have several simultaneous sessions on the same machine without doing anything complicated like setting up a network or running X. When the system starts, it will display a login prompt on the monitor after displaying all the boot messages. You can then type in your login name and password and start working (or playing!) on the first virtual console. At some point, you will probably wish to start another session, perhaps to look at documentation for a program you are running or to read your mail while waiting for an FTP transfer to finish. Just do Alt-F2 (hold down the Alt key and press the F2 key), and you will find a login prompt waiting for you on the second virtual console! When you want to go back to the original session, do Alt-F1. The default FreeBSD installation has three virtual consoles enabled (8 starting with 3.3-RELEASE), and Alt-F1, Alt-F2, and Alt-F3 will switch between these virtual consoles. To enable more of them, edit /etc/ttys (see &man.ttys.5;) and add entries for ttyv4 to ttyvc after the comment on Virtual terminals: # Edit the existing entry for ttyv3 in /etc/ttys and change # "off" to "on". ttyv3 "/usr/libexec/getty Pc" cons25 on secure ttyv4 "/usr/libexec/getty Pc" cons25 on secure ttyv5 "/usr/libexec/getty Pc" cons25 on secure ttyv6 "/usr/libexec/getty Pc" cons25 on secure ttyv7 "/usr/libexec/getty Pc" cons25 on secure ttyv8 "/usr/libexec/getty Pc" cons25 on secure ttyv9 "/usr/libexec/getty Pc" cons25 on secure ttyva "/usr/libexec/getty Pc" cons25 on secure ttyvb "/usr/libexec/getty Pc" cons25 on secure Use as many or as few as you want. The more virtual terminals you have, the more resources that are used; this can be important if you have 8MB RAM or less. You may also want to change the secure to insecure. If you want to run an X server you must leave at least one virtual terminal unused (or turned off) for it to use. That is to say that if you want to have a login prompt pop up for all twelve of your Alt-function keys, you're out of luck - you can only do this for eleven of them if you also want to run an X server on the same machine. The easiest way to disable a console is by turning it off. For example, if you had the full 12 terminal allocation mentioned above and you wanted to run X, you would change settings for virtual terminal 12 from: ttyvb "/usr/libexec/getty Pc" cons25 on secure to: ttyvb "/usr/libexec/getty Pc" cons25 off secure If your keyboard has only ten function keys, you would end up with: ttyv9 "/usr/libexec/getty Pc" cons25 off secure ttyva "/usr/libexec/getty Pc" cons25 off secure ttyvb "/usr/libexec/getty Pc" cons25 off secure (You could also just delete these lines.) Once you have edited /etc/ttys, the next step is to make sure that you have enough virtualterminal devices. The easiest way to do this is: &prompt.root; cd /dev &prompt.root; sh MAKEDEV vty12 Next, the easiest (and cleanest) way to activate the virtual consoles is to reboot. However, if you really don't want to reboot, you can just shut down the X Window system and execute (as root): &prompt.root; kill -HUP 1 It's imperative that you completely shut down X Window if it is running, before running this command. If you don't, your system will probably appear to hang/lock up after executing the kill command. How do I access the virtual consoles from X? Use Ctrl Alt Fn to switch back to a virtual console. Ctrl Alt F1 would return you to the first virtual console. Once you are back to a text console, you can then use Alt Fn as normal to move between them. To return to the X session, you must switch to the virtual console running X. If you invoked X from the command line, (e.g., using startx) then the X session will attach to the next unused virtual console, not the text console from which it was invoked. If you have eight active virtual terminals then X will be running on the ninth, and you would use Alt F9 to return. How do I start XDM on boot? There are two schools of thought on how to start xdm. One school starts xdm from /etc/ttys (see &man.ttys.5;) using the supplied example, while the other simply runs xdm from rc.local (see &man.rc.8;) or from a X.sh script in /usr/local/etc/rc.d. Both are equally valid, and one may work in situations where the other doesn't. In both cases the result is the same: X will popup a graphical login: prompt. The ttys method has the advantage of documenting which vty X will start on and passing the responsibility of restarting the X server on logout to init. The rc.local method makes it easy to kill xdm if there is a problem starting the X server. If loaded from rc.local, xdm should be started without any arguments (i.e., as a daemon). xdm must start AFTER getty runs, or else getty and xdm will conflict, locking out the console. The best way around this is to have the script sleep 10 seconds or so then launch xdm. If you are to start xdm from /etc/ttys, there still is a chance of conflict between xdm and &man.getty.8;. One way to avoid this is to add the vt number in the /usr/X11R6/lib/X11/xdm/Xservers file. :0 local /usr/X11R6/bin/X vt4 The above example will direct the X server to run in /dev/ttyv3. Note the number is offset by one. The X server counts the vty from one, whereas the FreeBSD kernel numbers the vty from zero. Why do I get Couldn't open console when I run xconsole? If you start X with startx, the permissions on /dev/console will not get changed, resulting in things like xterm -C and xconsole not working. This is because of the way console permissions are set by default. On a multi-user system, one doesn't necessarily want just any user to be able to write on the system console. For users who are logging directly onto a machine with a VTY, the &man.fbtab.5; file exists to solve such problems. In a nutshell, make sure an uncommented line of the form /dev/ttyv0 0600 /dev/console is in /etc/fbtab (see &man.fbtab.5;) and it will ensure that whomever logs in on /dev/ttyv0 will own the console. Before, I was able to run XFree86 as a regular user. Why does it now say that I must be root? All X servers need to be run as root in order to get direct access to your video hardware. Older versions of XFree86 (<= 3.3.6) installed all bundled servers to be automatically run as root (setuid to root). This is obviously a security hazard because X servers are large, complicated programs. Newer versions of XFree86 do not install the servers setuid to root for just this reason. Obviously, running an X server as the root user is not acceptable, nor a good idea security-wise. There are two ways to be able to use X as a regular user. The first is to use xdm or another display manager (e.g., kdm); the second is to use the Xwrapper. xdm is a daemon that handles graphical logins. It is usually started at boot time, and is responsible for authenticating users and starting their sessions; it is essentially the graphical counterpart of &man.getty.8; and &man.login.1;. For more information on xdm see the XFree86 documentation, and the the FAQ entry on it. Xwrapper is the X server wrapper; it is a small utility to enable one to manually run an X server while maintaining reasonable safety. It performs some sanity checks on the command line arguments given, and if they pass, runs the appropriate X server. If you do not want to run a display manger for whatever reason, this is for you. If you have installed the complete ports collection, you can find the port in /usr/ports/x11/wrapper. Why does my PS/2 mouse misbehave under X? Your mouse and the mouse driver may have somewhat become out of synchronization. In versions 2.2.5 and earlier, switching away from X to a virtual terminal and getting back to X again may make them re-synchronized. If the problem occurs often, you may add the following option in your kernel configuration file and recompile it. options PSM_CHECKSYNC See the section on building a kernel if you've no experience with building kernels. With this option, there should be less chance of synchronization problem between the mouse and the driver. If, however, you still see the problem, click any mouse button while holding the mouse still to re-synchronize the mouse and the driver. Note that unfortunately this option may not work with all the systems and voids the tap feature of the ALPS GlidePoint device attached to the PS/2 mouse port. In versions 2.2.6 and later, synchronization check is done in a slightly better way and is standard in the PS/2 mouse driver. It should even work with GlidePoint. (As the check code has become a standard feature, PSM_CHECKSYNC option is not available in these versions.) However, in rare case the driver may erroneously report synchronization problem and you may see the kernel message: psmintr: out of sync (xxxx != yyyy) and find your mouse doesn't seem to work properly. If this happens, disable the synchronization check code by setting the driver flags for the PS/2 mouse driver to 0x100. Enter UserConfig by giving the option at the boot prompt: boot: -c Then, in the UserConfig command line, type: UserConfig> flags psm0 0x100 UserConfig> quit How come my PS/2 mouse from MouseSystems does not seem to work? There have been some reports that certain model of PS/2 mouse from MouseSystems works only if it is put into the high resolution mode. Otherwise, the mouse cursor may jump to the upper-left corner of the screen every so often. Unfortunately there is no workaround for versions 2.0.X and 2.1.X. In versions 2.2 through 2.2.5, apply the following patch to /sys/i386/isa/psm.c and rebuild the kernel. See the section on building a kernel if you've no experience with building kernels. @@ -766,6 +766,8 @@ if (verbose >= 2) log(LOG_DEBUG, "psm%d: SET_DEFAULTS return code:%04x\n", unit, i); + set_mouse_resolution(sc->kbdc, PSMD_RES_HIGH); + #if 0 set_mouse_scaling(sc->kbdc); /* 1:1 scaling */ set_mouse_mode(sc->kbdc); /* stream mode */ In versions 2.2.6 or later, specify the flags 0x04 to the PS/2 mouse driver to put the mouse into the high resolution mode. Enter UserConfig by giving the option at the boot prompt: boot: -c Then, in the UserConfig command line, type: UserConfig> flags psm0 0x04 UserConfig> quit See the previous section for another possible cause of mouse problems. When building an X app, imake can't find Imake.tmpl. Where is it? Imake.tmpl is part of the Imake package, a standard X application building tool. Imake.tmpl, as well as several header files that are required to build X apps, is contained in the X prog distribution. You can install this from sysinstall or manually from the X distribution files. How do I reverse the mouse buttons? Run the command xmodmap -e "pointer = 3 2 1" from your .xinitrc or .xsession. How do I install a splash screen and where do I find them? Just prior to the release of FreeBSD 3.1, a new feature was added to allow the display of splash screens during the boot messages. The splash screens currently must be a 256 color bitmap (*.BMP) or ZSoft PCX (*.PCX) file. In addition, they must have a resolution of 320x200 or less to work on standard VGA adapters. If you compile VESA support into your kernel, then you can use larger bitmaps up to 1024x768. Note that VESA support requires the VM86 kernel option to be compiled into the kernel. The actual VESA support can either be compiled directly into the kernel with the VESA kernel config option or by loading the VESA kld module during bootup. To use a splash screen, you need to modify the startup files that control the boot process for FreeBSD. The files for this changed prior to the release of FreeBSD 3.2, so there are now two ways of loading a splash screen: FreeBSD 3.1 The first step is to find a bitmap version of your splash screen. Release 3.1 only supports Windows bitmap splash screens. Once you've found your splash screen of choice copy it to /boot/splash.bmp. Next, you need to have a /boot/loader.rc file that contains the following lines: load kernel load -t splash_image_data /boot/splash.bmp load splash_bmp autoboot FreeBSD 3.2+ In addition to adding support for PCX splash screens, FreeBSD 3.2 includes a nicer way of configuring the boot process. If you wish, you can use the method listed above for FreeBSD 3.1. If you do and you want to use PCX, replace splash_bmp with splash_pcx. If, on the other hand, you want to use the newer boot configuration, you need to create a /boot/loader.rc file that contains the following lines: include /boot/loader.4th start and a /boot/loader.conf that contains the following: splash_bmp_load="YES" bitmap_load="YES" This assumes you are using /boot/splash.bmp for your splash screen. If you'd rather use a PCX file, copy it to /boot/splash.pcx, create a /boot/loader.rc as instructed above, and create a /boot/loader.conf that contains: splash_pcx_load="YES" bitmap_load="YES" bitmap_name="/boot/splash.pcx" Now all you need is a splash screen. For that you can surf on over to the gallery at http://www.baldwin.cx/splash/. Can I use the Windows(tm) keys on my keyboard in X? Yes. All you need to do is use &man.xmodmap.1; to define what function you wish them to perform. Assuming all Windows(tm) keyboards are standard then the keycodes for the 3 keys are 115 - Windows(tm) key, between the left-hand Ctrl and Alt keys 116 - Windows(tm) key, to the right of the Alt-Gr key 117 - Menu key, to the left of the right-hand Ctrl key To have the left Windows(tm) key print a comma, try this. &prompt.root; xmodmap -e "keycode 115 = comma" You will probably have to re-start your window manager to see the result. To have the Windows(tm) key-mappings enabled automatically everytime you start X either put the xmodmap commands in your ~/.xinitrc file or, preferably, create a file ~/.xmodmaprc and include the xmodmap options, one per line, then add the line xmodmap $HOME/.xmodmaprc to your ~/.xinitrc. For example, you could map the 3 keys top be F13, F14, and F15, respectively. This would make it easy to map them to useful functions within applications or your window manager, as demonstrated further down. To do this put the following in ~/.xmodmaprc. keycode 115 = F13 keycode 116 = F14 keycode 117 = F15 If you use fvwm2, for example, you could map the keys so that F13 iconifies (or de-iconifies) the window the cursor is in, F14 brings the window the cursor is in to the front or, if it is already at the front, pushes it to the back, and F15 pops up the main Workplace (application) menu even if the cursor is not on the desktop, which is useful if you don't have any part of the desktop visible (and the logo on the key matches its functionality). The following entries in ~/.fvwmrc implement the aforementioned setup: Key F13 FTIWS A Iconify Key F14 FTIWS A RaiseLower Key F15 A A Menu Workplace Nop Networking Where can I get information on diskless booting? Diskless booting means that the FreeBSD box is booted over a network, and reads the necessary files from a server instead of its hard disk. For full details, please read the Handbook entry on diskless booting Can a FreeBSD box be used as a dedicated network router? Internet standards and good engineering practice prohibit us from providing packet forwarding by default in FreeBSD. You can however enable this feature by changing the following variable to YES in &man.rc.conf.5;: gateway_enable=YES # Set to YES if this host will be a gateway This option will put the &man.sysctl.8; variable net.inet.ip.forwarding to 1. In most cases, you will also need to run a routing process to tell other systems on your network about your router; FreeBSD comes with the standard BSD routing daemon &man.routed.8; or for more complex situations you may want to try GaTeD (available from http://www.gated.org/) which supports FreeBSD as of 3_5Alpha7. It is our duty to warn you that, even when FreeBSD is configured in this way, it does not completely comply with the Internet standard requirements for routers; however, it comes close enough for ordinary usage. Can I connect my Win95 box to the Internet via FreeBSD? Typically, people who ask this question have two PC's at home, one with FreeBSD and one with Win95; the idea is to use the FreeBSD box to connect to the Internet and then be able to access the Internet from the Windows95 box through the FreeBSD box. This is really just a special case of the previous question. ... and the answer is yes! In FreeBSD 3.x, user-mode ppp contains a option. If you run ppp with the , set gateway_enable to YES in /etc/rc.conf, and configure your Windows machine correctly, this should work fine. More detailed information about setting this up can be found in the Pedantic PPP Primer by Steve Sims. If you are using kernel-mode ppp, or have an Ethernet connection to the Internet, you will have to use &man.natd.8;. Please look at the natd section of this FAQ. Why does recompiling the latest BIND from ISC fail? There is a conflict between the cdefs.h file in the distribution and the one shipped with FreeBSD. Just remove compat/include/sys/cdefs.h. Does FreeBSD support SLIP and PPP? Yes. See the manual pages for &man.slattach.8;, &man.sliplogin.8;, &man.ppp.8;, and &man.pppd.8;. &man.ppp.8; and &man.pppd.8; provide support for both incoming and outgoing connections, while &man.sliplogin.8; deals exlusively with incoming connections, and &man.slattach.8; deals exclusively with outgoing connections. For more information on how to use these, please see the Handbook chapter on PPP and SLIP. If you only have access to the Internet through a shell account, you may want to have a look at the slirp package. It can provide you with (limited) access to services such as ftp and http direct from your local machine. Does FreeBSD support NAT or Masquerading? If you have a local subnet (one or more local machines), but have been allocated only a single IP number from your Internet provider (or even if you receive a dynamic IP number), you may want to look at the &man.natd.8; program. &man.natd.8; allows you to connect an entire subnet to the internet using only a single IP number. The &man.ppp.8; program has similar functionality built in via the switch. The alias library (&man.libalias.3;) is used in both cases. How come I cannot create a /dev/ed0 device? In the Berkeley networking framework, network interfaces are only directly accessible by kernel code. Please see the /etc/rc.network file and the manual pages for the various network programs mentioned there for more information. If this leaves you totally confused, then you should pick up a book describing network administration on another BSD-related operating system; with few significant exceptions, administering networking on FreeBSD is basically the same as on SunOS 4.0 or Ultrix. How can I setup Ethernet aliases? Add netmask 0xffffffff to your &man.ifconfig.8; command-line like the following: &prompt.root; ifconfig ed0 alias 204.141.95.2 netmask 0xffffffff How do I get my 3C503 to use the other network port? If you want to use the other ports, you'll have to specify an additional parameter on the &man.ifconfig.8; command line. The default port is link0. To use the AUI port instead of the BNC one, use link2. These flags should be specified using the ifconfig_* variables in /etc/rc.conf (see &man.rc.conf.5;). Why am I having trouble with NFS and FreeBSD? Certain PC network cards are better than others (to put it mildly) and can sometimes cause problems with network intensive applications like NFS. See the Handbook entry on NFS for more information on this topic. Why can't I NFS-mount from a Linux box? Some versions of the Linux NFS code only accept mount requests from a privileged port; try &prompt.root; mount -o -P linuxbox:/blah /mnt Why can't I NFS-mount from a Sun box? Sun workstations running SunOS 4.X only accept mount requests from a privileged port; try &prompt.root; mount -o -P sunbox:/blah /mnt Why does mountd keep telling me it can't change attributes and that I have a bad exports list on my FreeBSD NFS server? The most frequent problem is not understanding this passage from the &man.exports.5; manual page correctly:

Each line in the file (other than comment lines that begin with a #) specifies the mount point(s) and export flags within one local server filesystem for one or more hosts. A host may be specified only once for each local filesystem on the server and there may be only one default entry for each server filesystem that applies to all other hosts.

This is made more clear by an example of a common mistake. If everything above /usr is part of one filesystem (there are no mounts above /usr) the following exports list is not valid: /usr/src client /usr/ports client There are two lines specifying properties for one filesystem, /usr, exported to the same host, client. The correct format is: /usr/src /usr/ports client To rephrase the passage from the manual page, the properties of one filesystem exported to a given host (world-wide exports are treated like another unique host) must all occur on one line. And yes, this does cause limitiation in how you can export filesystems without ugly workarounds, but for most people, this is not an issue. The following is an example of a valid export list, where /usr and /exports are local filesystems: # Export src and ports to client01 and client02, but only # client01 has root privileges on it /usr/src /usr/ports -maproot=0 client01 /usr/src /usr/ports client02 # The "client" machines have root and can mount anywhere # up /exports. The world can mount /exports/obj read-only /exports -alldirs -maproot=0 client01 client02 /exports/obj -ro Why am I having problems talking PPP to NeXTStep machines? Try disabling the TCP extensions in /etc/rc.conf (see &man.rc.conf.5;) by changing the following variable to NO: tcp_extensions=NO Xylogic's Annex boxes are also broken in this regard and you must use the above change to connect thru them. How do I enable IP multicast support? Multicast host operations are fully supported in FreeBSD 2.0 and later by default. If you want your box to run as a multicast router, you will need to recompile your kernel with the MROUTING option and run &man.mrouted.8;. FreeBSD 2.2 and later will start &man.mrouted.8; at boot time if the flag mrouted_enable is set to "YES" in /etc/rc.conf. MBONE tools are available in their own ports category, mbone. If you are looking for the conference tools vic and vat, look there! For more information, see the Mbone Information Web. Which network cards are based on the DEC PCI chipset? Here is a list compiled by Glen Foster, with some more modern additions: Vendor Model ---------------------------------------------- ASUS PCI-L101-TB Accton ENI1203 Cogent EM960PCI Compex ENET32-PCI D-Link DE-530 Dayna DP1203, DP2100 DEC DE435, DE450 Danpex EN-9400P3 JCIS Condor JC1260 Linksys EtherPCI Mylex LNP101 SMC EtherPower 10/100 (Model 9332) SMC EtherPower (Model 8432) TopWare TE-3500P Znyx (2.2.x) ZX312, ZX314, ZX342, ZX345, ZX346, ZX348 (3.x) ZX345Q, ZX346Q, ZX348Q, ZX412Q, ZX414, ZX442, ZX444, ZX474, ZX478, ZX212, ZX214 (10mbps/hd) Why do I have to use the FQDN for hosts on my site? You will probably find that the host is actually in a different domain; for example, if you are in foo.bar.edu and you wish to reach a host called mumble in the bar.edu domain, you will have to refer to it by the fully-qualified domain name, mumble.bar.edu, instead of just mumble. Traditionally, this was allowed by BSD BIND resolvers. However the current version of bind (see &man.named.8;) that ships with FreeBSD no longer provides default abbreviations for non-fully qualified domain names other than the domain you are in. So an unqualified host mumble must either be found as mumble.foo.bar.edu, or it will be searched for in the root domain. This is different from the previous behavior, where the search continued across mumble.bar.edu, and mumble.edu. Have a look at RFC 1535 for why this was considered bad practice, or even a security hole. As a good workaround, you can place the line search foo.bar.edu bar.edu instead of the previous domain foo.bar.edu into your /etc/resolv.conf file (see &man.resolv.conf.5;). However, make sure that the search order does not go beyond the boundary between local and public administration, as RFC 1535 calls it. Why do I get an error, Permission denied, for all networking operations? If you have compiled your kernel with the IPFIREWALL option, you need to be aware that the default policy as of 2.1.7R (this actually changed during 2.1-STABLE development) is to deny all packets that are not explicitly allowed. If you had unintentionally misconfigured your system for firewalling, you can restore network operability by typing the following while logged in as root: &prompt.root; ipfw add 65534 allow all from any to any You can also set firewall_type="open" in /etc/rc.conf. For further information on configuring a FreeBSD firewall, see the Handbook section. How much overhead does IPFW incur? The answer to this depends mostly on your rule set and processor speed. For most applications dealing with ethernet and small rule sets, the answer is, negligible. For those of you that need actual measurements to satisfy your curiosity, read on. The following measurements were made using 2.2.5-STABLE on a 486-66. IPFW was modified to measure the time spent within the ip_fw_chk routine, displaying the results to the console every 1000 packets. Two rule sets, each with 1000 rules were tested. The first set was designed to demonstrate a worst case scenario by repeating the rule: &prompt.root; ipfw add deny tcp from any to any 55555 This demonstrates worst case by causing most of IPFW's packet check routine to be executed before finally deciding that the packet does not match the rule (by virtue of the port number). Following the 999th iteration of this rule was an allow ip from any to any. The second set of rules were designed to abort the rule check quickly: &prompt.root; ipfw add deny ip from 1.2.3.4 to 1.2.3.4 The nonmatching source IP address for the above rule causes these rules to be skipped very quickly. As before, the 1000th rule was an allow ip from any to any. The per-packet processing overhead in the former case was approximately 2.703ms/packet, or roughly 2.7 microseconds per rule. Thus the theoretical packet processing limit with these rules is around 370 packets per second. Assuming 10Mbps ethernet and a ~1500 byte packet size, we would only be able to achieve a 55.5% bandwidth utilization. For the latter case each packet was processed in approximately 1.172ms, or roughly 1.2 microseconds per rule. The theoretical packet processing limit here would be about 853 packets per second, which could consume 10Mbps ethernet bandwidth. The excessive number of rules tested and the nature of those rules do not provide a real-world scenario -- they were used only to generate the timing information presented here. Here are a few things to keep in mind when building an efficient rule set: Place an established rule early on to handle the majority of TCP traffic. Don't put any allow tcp statements before this rule. Place heavily triggered rules earlier in the rule set than those rarely used (without changing the permissiveness of the firewall, of course). You can see which rules are used most often by examining the packet counting statistics with ipfw -a l. Why is my ipfw fwd rule to redirect a service to another machine not working? Possibly because you want to do network address translation (NAT) and not just forward packets. A fwd rule does exactly what it says; it forwards packets. It does not actually change the data inside the packet. Say we have a rule like: 01000 fwd 10.0.0.1 from any to foo 21 When a packet with a destination address of foo arrives at the machine with this rule, the packet is forwarded to 10.0.0.1, but it still has the destination address of foo! The destination address of the packet is not changed to 10.0.0.1. Most machines would probably drop a packet that they receive with a destination address that is not their own. Therefore, using a fwd rule does not often work the way the user expects. This behavior is a feature and not a bug. See the FAQ about redirecting services, the &man.natd.8; manual, or one of the several port redirecting utilities in the ports collection for a correct way to do this. How can I redirect service requests from one machine to another? You can redirect FTP (and other service) request with the socket package, available in the ports tree in category sysutils. Simply replace the service's commandline to call socket instead, like so: ftp stream tcp nowait nobody /usr/local/bin/socket socket ftp.foo.com ftp where ftp.foo.com and ftp are the host and port to redirect to, respectively. Where can I get a bandwidth management tool? There are two bandwidth management tools available for FreeBSD. ALTQ is available for free; Bandwidth Manager from Emerging Technologies is a commercial product. BIND (named) is listening on port 53 and some other high-numbered port. Has my host been compromised? Probably not. FreeBSD 3.0 and later use a version of BIND that uses a random high-numbered port for outgoing queries. If you want to use port 53 for outgoing queries, either to get past a firewall or to make yourself feel better, you can try the following in /etc/namedb/named.conf: options { query-source address * port 53; }; You can replace the * with a single IP address if you want to tighten things further. Congratulations, by the way. It is good practice to read your &man.sockstat.1; output and notice odd things! Why do I get /dev/bpf0: device not configured? The Berkeley Packet Filter (&man.bpf.4;) driver needs to be enabled before running programs that utilize it. Add this to your kernel config file and build a new kernel: pseudo-device bpfilter # Berkeley Packet Filter Secondly, after rebooting you will have to create the device node. This can be accomplished by a change to the /dev directory, followed by the execution of: &prompt.root; sh MAKEDEV bpf0 Please see the handbook's entry on device nodes for more information on creating devices. Use the sharity light package in the ports collection. What are these messages about icmp-response bandwidth limit 300/200 pps in my log files? This is the kernel telling you that some activity is provoking it to send more ICMP or TCP reset (RST) responses than it thinks it should. ICMP responses are often generated as a result of attempted connections to unused UDP ports. TCP resets are generated as a result of attempted connections to unopened TCP ports. Among others, these are the kinds of activities which may cause these messages: Brute-force denial of service (DoS) attacks (as opposed to single-packet attacks which exploit a specific vulnerability). Port scans which attempt to connect to a large number of ports (as opposed to only trying a few well-known ports). The first number in the message tells you how many packets the kernel would have sent if the limit wasn't in place, and the second number tells you the limit. You can control the limit using the net.inet.icmp.icmplim sysctl variable like this, where 300 is the limit in packets per second: &prompt.root; sysctl -w net.inet.icmp.icmplim=300 If you don't want to see messages about this in your log files, but you still want the kernel to do response limiting, you can use the net.inet.icmp.icmplim_output sysctl variable to disable the output like this: &prompt.root; sysctl -w net.inet.icmp.icmplim_output=0 Finally, if you want to disable response limiting, you can set the net.inet.icmp.icmplim sysctl variable (see above for an example) to 0. Disabling response limiting is discouraged for the reasons listed above. PPP I can't make &man.ppp.8; work. What am I doing wrong? You should first read the &man.ppp.8; man page and the ppp section of the handbook. Enable logging with the command set log Phase Chat Connect Carrier lcp ipcp ccp command This command may be typed at the ppp command prompt or it may be entered in the /etc/ppp/ppp.conf configuration file (the start of the default section is the best place to put it). Make sure that /etc/syslog.conf (see &man.syslog.conf.5;) contains the lines !ppp *.* /var/log/ppp.log and that the file /var/log/ppp.log exists. You can now find out a lot about what's going on from the log file. Don't worry if it doesn't all make sense. If you need to get help from someone, it may make sense to them. If your version of ppp doesn't understand the set log command, you should download the latest version. It will build on FreeBSD version 2.1.5 and higher. Why does &man.ppp.8; hang when I run it? This is usually because your hostname won't resolve. The best way to fix this is to make sure that /etc/hosts is consoluted by your resolver first by editing /etc/host.conf and putting the hosts line first. Then, simply put an entry in /etc/hosts for your local machine. If you have no local network, change your localhost line: 127.0.0.1 foo.bar.com foo localhost Otherwise, simply add another entry for your host. Consult the relevant man pages for more details. You should be able to successfully ping -c1 `hostname` when you're done. Why won't &man.ppp.8; dial in -auto mode? First, check that you've got a default route. By running netstat -rn (see &man.netstat.1;), you should see two entries like this: Destination Gateway Flags Refs Use Netif Expire default 10.0.0.2 UGSc 0 0 tun0 10.0.0.2 10.0.0.1 UH 0 0 tun0 This is assuming that you've used the addresses from the handbook, the man page or from the ppp.conf.sample file. If you haven't got a default route, it may be because you're running an old version of &man.ppp.8; that doesn't understand the word HISADDR in the ppp.conf file. If your version of ppp is from before FreeBSD 2.2.5, change the add 0 0 HISADDR line to one saying add 0 0 10.0.0.2 Another reason for the default route line being missing is that you have mistakenly set up a default router in your /etc/rc.conf (see &man.rc.conf.5;) file (this file was called /etc/sysconfig prior to release 2.2.2), and you have omitted the line saying delete ALL from ppp.conf. If this is the case, go back to the Final system configuration section of the handbook. What does No route to host mean? This error is usually due to a missing MYADDR: delete ALL add 0 0 HISADDR section in your /etc/ppp/ppp.linkup file. This is only necessary if you have a dynamic IP address or don't know the address of your gateway. If you're using interactive mode, you can type the following after entering packet mode (packet mode is indicated by the capitalized PPP in the prompt): delete ALL add 0 0 HISADDR Refer to the PPP and Dynamic IP addresses section of the handbook for further details. Why does my connection drop after about 3 minutes? The default ppp timeout is 3 minutes. This can be adjusted with the line set timeout NNN where NNN is the number of seconds of inactivity before the connection is closed. If NNN is zero, the connection is never closed due to a timeout. It is possible to put this command in the ppp.conf file, or to type it at the prompt in interactive mode. It is also possible to adjust it on the fly while the line is active by connecting to ppps server socket using &man.telnet.1; or &man.pppctl.8;. Refer to the &man.ppp.8; man page for further details. Why does my connection drop under heavy load? If you have Link Quality Reporting (LQR) configured, it is possible that too many LQR packets are lost between your machine and the peer. Ppp deduces that the line must therefore be bad, and disconnects. Prior to FreeBSD version 2.2.5, LQR was enabled by default. It is now disabled by default. LQR can be disabled with the line disable lqr Why does my connection drop after a random amount of time? Sometimes, on a noisy phone line or even on a line with call waiting enabled, your modem may hang up because it thinks (incorrectly) that it lost carrier. There's a setting on most modems for determining how tolerant it should be to temporary losses of carrier. On a USR Sportster for example, this is measured by the S10 register in tenths of a second. To make your modem more forgiving, you could add the following send-expect sequence to your dial string: set dial "...... ATS10=10 OK ......" Refer to your modem manual for details. Why does my connection hang after a random amount of time? Many people experience hung connections with no apparent explaination. The first thing to establish is which side of the link is hung. If you are using an external modem, you can simply try using &man.ping.8; to see if the TD light is flashing when you transmit data. If it flashes (and the RD light doesn't), the problem is with the remote end. If TD doesn't flash, the problem is local. With an internal modem, you'll need to use the set server command in your ppp.conf file. When the hang occurs, connect to ppp using pppctl. If your network connection suddenly revives (ppp was revived due to the activity on the diagnostic socket) or if you can't connect (assuming the set socket command succeeded at startup time), the problem is local. If you can connect and things are still hung, enable local async logging with set log local async and use &man.ping.8; from another window or terminal to make use of the link. The async logging will show you the data being transmitted and received on the link. If data is going out and not coming back, the problem is remote. Having established whether the problem is local or remote, you now have two possibilities: The remote end is not responding. What can I do? There's very little you can do about this. Most ISPs will refuse to help if you're not running a Microsoft OS. You can enable lqr in your ppp.conf file, allowing ppp to detect the remote failure and hang up, but this detection is relatively slow and therefore not that useful. You may want to avoid telling your ISP that you're running user-ppp.... First, try disabling all local compression by adding the following to your configuration: disable pred1 deflate deflate24 protocomp acfcomp shortseq vj deny pred1 deflate deflate24 protocomp acfcomp shortseq vj Then reconnect to ensure that this makes no difference. If things improve or if the problem is solved completely, determine which setting makes the difference through trial and error. This will provide good amunition when you contact your ISP (although it may make it apparent that you're not running a Microsoft product). Before contacting your ISP, enable async logging locally and wait until the connection hangs again. This may use up quite a bit of disk space. The last data read from the port may be of interest. It is usually ascii data, and may even describe the problem (Memory fault, core dumped?). If your ISP is helpful, they should be able to enable logging on their end, then when the next link drop occurs, they may be able to tell you why their side is having a problem. Feel free to send the details to &a.brian;, or even to ask your ISP to contact me directly. &man.ppp.8; has hung. What can I do? Your best bet here is to rebuild ppp by adding CFLAGS+=-g and STRIP= to the end of the Makefile, then doing a make clean && make && make install. When ppp hangs, find the ppp process id with ps ajxww | fgrep ppp and run gdb ppp PID. From the gdb prompt, you can then use bt to get a stack trace. Send the results to brian@Awfulhak.org. Why does nothing happen after the Login OK! message? Prior to FreeBSD version 2.2.5, once the link was established, &man.ppp.8; would wait for the peer to initiate the Line Control Protocol (LCP). Many ISPs will not initiate negotiations and expect the client to do so. To force ppp to initiate the LCP, use the following line: set openmode active It usually does no harm if both sides initiate negotiation, so openmode is now active by default. However, the next section explains when it does do some harm. I keep seeing errors about magic being the same. What does it mean? Occasionally, just after connecting, you may see messages in the log that say magic is the same. Sometimes, these messages are harmless, and sometimes one side or the other exits. Most ppp implementations cannot survive this problem, and even if the link seems to come up, you'll see repeated configure requests and configure acknowledgements in the log file until ppp eventually gives up and closes the connection. This normally happens on server machines with slow disks that are spawning a getty on the port, and executing ppp from a login script or program after login. I've also heard reports of it happening consistently when using slirp. The reason is that in the time taken between getty exiting and ppp starting, the client-side ppp starts sending Line Control Protocol (LCP) packets. Because ECHO is still switched on for the port on the server, the client ppp sees these packets reflect back. One part of the LCP negotiation is to establish a magic number for each side of the link so that reflections can be detected. The protocol says that when the peer tries to negotiate the same magic number, a NAK should be sent and a new magic number should be chosen. During the period that the server port has ECHO turned on, the client ppp sends LCP packets, sees the same magic in the reflected packet and NAKs it. It also sees the NAK reflect (which also means ppp must change its magic). This produces a potentially enormous number of magic number changes, all of which are happily piling into the server's tty buffer. As soon as ppp starts on the server, it's flooded with magic number changes and almost immediately decides it's tried enough to negotiate LCP and gives up. Meanwhile, the client, who no longer sees the reflections, becomes happy just in time to see a hangup from the server. This can be avoided by allowing the peer to start negotiating with the following line in your ppp.conf file: set openmode passive This tells ppp to wait for the server to initiate LCP negotiations. Some servers however may never initiate negotiations. If this is the case, you can do something like: set openmode active 3 This tells ppp to be passive for 3 seconds, and then to start sending LCP requests. If the peer starts sending requests during this period, ppp will immediately respond rather than waiting for the full 3 second period. LCP negotiations continue 'till the connection is closed. What is wrong? There is currently an implementation mis-feature in ppp where it doesn't associate LCP, CCP & IPCP responses with their original requests. As a result, if one ppp implementation is more than 6 seconds slower than the other side, the other side will send two additional LCP configuration requests. This is fatal. Consider two implementations, A and B. A starts sending LCP requests immediately after connecting and B takes 7 seconds to start. When B starts, A has sent 3 LCP REQs. We're assuming the line has ECHO switched off, otherwise we'd see magic number problems as described in the previous section. B sends a REQ, then an ACK to the first of A's REQs. This results in A entering the OPENED state and sending and ACK (the first) back to B. In the meantime, B sends back two more ACKs in response to the two additional REQs sent by A before B started up. B then receives the first ACK from A and enters the OPENED state. A receives the second ACK from B and goes back to the REQ-SENT state, sending another (forth) REQ as per the RFC. It then receives the third ACK and enters the OPENED state. In the meantime, B receives the forth REQ from A, resulting in it reverting to the ACK-SENT state and sending another (second) REQ and (forth) ACK as per the RFC. A gets the REQ, goes into REQ-SENT and sends another REQ. It immediately receives the following ACK and enters OPENED. This goes on 'till one side figures out that they're getting nowhere and gives up. The best way to avoid this is to configure one side to be passive - that is, make one side wait for the other to start negotiating. This can be done with the set openmode passive command. Care should be taken with this option. You should also use the set stopped N command to limit the amount of time that ppp waits for the peer to begin negotiations. Alternatively, the set openmode active N command (where N is the number of seconds to wait before starting negotiations) can be used. Check the manual page for details. Why does &man.ppp.8; lock up shortly after connection? Prior to version 2.2.5 of FreeBSD, it was possible that your link was disabled shortly after connection due to ppp mis-handling Predictor1 compression negotiation. This would only happen if both sides tried to negotiate different Compression Control Protocols (CCP). This problem is now corrected, but if you're still running an old version of ppp, the problem can be circumvented with the line disable pred1 Why does &man.ppp.8; lock up when I shell out to test it? When you execute the shell or ! command, ppp executes a shell (or if you've passed any arguements, ppp will execute those arguements). Ppp will wait for the command to complete before continuing. If you attempt to use the ppp link while running the command, the link will appear to have frozen. This is because ppp is waiting for the command to complete. If you wish to execute commands like this, use the !bg command instead. This will execute the given command in the background, and ppp can continue to service the link. How come &man.ppp.8; over a null-modem cable never exits? There is no way for ppp to automatically determine that a direct connection has been dropped. This is due to the lines that are used in a null-modem serial cable. When using this sort of connection, LQR should always be enabled with the line enable lqr LQR is accepted by default if negotiated by the peer. Why does &man.ppp.8; dial for no reason in -auto mode? If ppp is dialing unexpectedly, you must determine the cause, and set up Dial filters (dfilters) to prevent such dialing. To determine the cause, use the following line: set log +tcp/ip This will log all traffic through the connection. The next time the line comes up unexpectedly, you will see the reason logged with a convenient timestamp next to it. You can now disable dialing under these circumstances. Usually, this sort of problem arises due to DNS lookups. To prevent DNS lookups from establishing a connection (this will not prevent ppp from passing the packets through an established connection), use the following: set dfilter 1 deny udp src eq 53 set dfilter 2 deny udp dst eq 53 set dfilter 3 permit 0/0 0/0 This is not always suitable, as it will effectively break your demand-dial capabilities - most programs will need a DNS lookup before doing any other network related things. In the DNS case, you should try to determine what is actually trying to resolve a host name. A lot of the time, &man.sendmail.8; is the culprit. You should make sure that you tell sendmail not to do any DNS lookups in its configuration file. See the section on Mail Configuration for details on how to create your own configuration file and what should go into it. You may also want to add the following line to your .mc file: define(`confDELIVERY_MODE', `d')dnl This will make sendmail queue everything until the queue is run (usually, sendmail is invoked with , telling it to run the queue every 30 minutes) or until a sendmail -q is done (perhaps from your ppp.linkup file). What do these CCP errors mean? I keep seeing the following errors in my log file: CCP: CcpSendConfigReq CCP: Received Terminate Ack (1) state = Req-Sent (6) This is because ppp is trying to negotiate Predictor1 compression, and the peer does not want to negotiate any compression at all. The messages are harmless, but if you wish to remove them, you can disable Predictor1 compression locally too: disable pred1 Why does &man.ppp.8; lock up during file transfers with IO errors? Under FreeBSD 2.2.2 and before, there was a bug in the tun driver that prevents incoming packets of a size larger than the tun interface's MTU size. Receipt of a packet greater than the MTU size results in an IO error being logged via syslogd. The ppp specification says that an MRU of 1500 should always be accepted as a minimum, despite any LCP negotiations, therefore it is possible that should you decrease the MTU to less than 1500, your ISP will transmit packets of 1500 regardless, and you will tickle this non-feature - locking up your link. The problem can be circumvented by never setting an MTU of less than 1500 under FreeBSD 2.2.2 or before. Why doesn't &man.ppp.8; log my connection speed? In order to log all lines of your modem conversation, you must enable the following: set log +connect This will make &man.ppp.8; log everything up until the last requested expect string. If you wish to see your connect speed and are using PAP or CHAP (and therefore don't have anything to chat after the CONNECT in the dial script - no set login script), you must make sure that you instruct ppp to expect the whole CONNECT line, something like this: set dial "ABORT BUSY ABORT NO\\sCARRIER TIMEOUT 4 \ \"\" ATZ OK-ATZ-OK ATDT\\T TIMEOUT 60 CONNECT \\c \\n" Here, we get our CONNECT, send nothing, then expect a line-feed, forcing ppp to read the whole CONNECT response. Why does &man.ppp.8; ignore the \ character in my chat script? Ppp parses each line in your config files so that it can interpret strings such as set phone "123 456 789" correctly (and realize that the number is actually only one argument. In order to specify a " character, you must escape it using a backslash (\). When the chat interpreter parses each argument, it re-interprets the argument in order to find any special escape sequences such as \P or \T (see the man page). As a result of this double-parsing, you must remember to use the correct number of escapes. If you wish to actually send a \ character to (say) your modem, you'd need something like: set dial "\"\" ATZ OK-ATZ-OK AT\\\\X OK" resulting in the following sequence: ATZ OK AT\X OK or set phone 1234567 set dial "\"\" ATZ OK ATDT\\T" resulting in the following sequence: ATZ OK ATDT1234567 Why does &man.ppp.8; get a seg-fault, but I see no ppp.core file? Ppp (or any other program for that matter) should never dump core. Because ppp runs with an effective user id of 0, the operating system will not write ppps core image to disk before terminating it. If, however ppp is actually termating due to a segmentation violation or some other signal that normally causes core to be dumped, and you're sure you're using the latest version (see the start of this section), then you should do the following: &prompt.user; tar xfz ppp-*.src.tar.gz &prompt.user; cd ppp*/ppp &prompt.user; echo STRIP= >>Makefile &prompt.user; echo CFLAGS+=-g >>Makefile &prompt.user; make clean all &prompt.user; su &prompt.root; make install &prompt.root; chmod 555 /usr/sbin/ppp You will now have a debuggable version of ppp installed. You will have to be root to run ppp as all of its privileges have been revoked. When you start ppp, take a careful note of what your current directory was at the time. Now, if and when ppp receives the segmentation violation, it will dump a core file called ppp.core. You should then do the following: &prompt.user; su &prompt.root; gdb /usr/sbin/ppp ppp.core (gdb) bt ..... (gdb) f 0 .... (gdb) i args .... (gdb) l ..... All of this information should be given alongside your question, making it possible to diagnose the problem. If you're familiar with gdb, you may wish to find out some other bits and pieces such as what actually caused the dump and the addresses & values of the relevant variables. Why does the process that forces a dial in auto mode never connect? This was a known problem with ppp set up to negotiate a dynamic local IP number with the peer in auto mode. It is fixed in the latest version - search the man page for iface. The problem was that when that initial program calls &man.connect.2;, the IP number of the tun interface is assigned to the socket endpoint. The kernel creates the first outgoing packet and writes it to the tun device. ppp then reads the packet and establishes a connection. If, as a result of ppps dynamic IP assignment, the interface address is changed, the original socket endpoint will be invalid. Any subsequent packets sent to the peer will usually be dropped. Even if they aren't, any responses will not route back to the originating machine as the IP number is no longer owned by that machine. There are several theoretical ways to approach this problem. It would be nicest if the peer would re-assign the same IP number if possible :-) The current version of ppp does this, but most other implementations don't. The easiest method from our side would be to never change the tun interface IP number, but instead to change all outgoing packets so that the source IP number is changed from the interface IP to the negotiated IP on the fly. This is essentially what the iface-alias option in the latest version of ppp is doing (with the help of &man.libalias.3; and ppp's switch) - it's maintaining all previous interface addresses and NATing them to the last negotiated address. Another alternative (and probably the most reliable) would be to implement a system call that changes all bound sockets from one IP to another. ppp would use this call to modify the sockets of all existing programs when a new IP number is negotiated. The same system call could be used by dhcp clients when they are forced to re-bind() their sockets. Yet another possibility is to allow an interface to be brought up without an IP number. Outgoing packets would be given an IP number of 255.255.255.255 up until the first SIOCAIFADDR ioctl is done. This would result in fully binding the socket. It would be up to ppp to change the source IP number, but only if it's set to 255.255.255.255, and only the IP number and IP checksum would need to change. This, however is a bit of a hack as the kernel would be sending bad packets to an improperly configured interface, on the assumption that some other mechanism is capable of fixing things retrospectively. Why don't most games work with the -nat switch? The reason games and the like don't work when libalias is in use is that the machine on the outside will try to open a connection or send (unsolicited) UDP packets to the machine on the inside. The NAT software doesn't know that it should send these packets to the interior machine. To make things work, make sure that the only thing running is the software that you're having problems with, then either run tcpdump on the tun interface of the gateway or enable ppp tcp/ip logging (set log +tcp/ip) on the gateway. When you start the offending software, you should see packets passing through the gateway machine. When something comes back from the outside, it'll be dropped (that's the problem). Note the port number of these packets then shut down the offending software. Do this a few times to see if the port numbers are consistent. If they are, then the following line in the relevant section of /etc/ppp/ppp.conf will make the software functional: nat port proto internalmachine:port port where proto is either tcp or udp, internalmachine is the machine that you want the packets to be sent to and port is the destination port number of the packets. You won't be able to use the software on other machines without changing the above command, and running the software on two internal machines at the same time is out of the question - after all, the outside world is seeing your entire internal network as being just a single machine. If the port numbers aren't consistent, there are three more options: Submit support in libalias. Examples of special cases can be found in /usr/src/lib/libalias/alias_*.c (alias_ftp.c is a good prototype). This usually involves reading certain recognised outgoing packets, identifying the instruction that tells the outside machine to initiate a connection back to the internal machine on a specific (random) port and setting up a route in the alias table so that the subsequent packets know where to go. This is the most difficult solution, but it is the best and will make the software work with multiple machines. Use a proxy. The application may support socks5 for example, or (as in the cvsup case) may have a passive option that avoids ever requesting that the peer open connections back to the local machine. Redirect everything to the internal machine using nat addr. This is the sledge-hammer approach. Has anybody made a list of useful port numbers? Not yet, but this is intended to grow into such a list (if any interest is shown). In each example, internal should be replaced with the IP number of the machine playing the game. Asheron's Call nat port udp internal :65000 65000 Manually change the port number within the game to 65000. If you've got a number of machines that you wish to play on assign a unique port number for each (i.e. 65001, 65002, etc) and add a nat port line for each one. Half Life nat port udp internal:27005 27015 PCAnywhere 8.0 nat port udp internal:5632 5632 nat port tcp internal:5631 5631 Quake nat port udp internal:6112 6112 Alternatively, you may want to take a look at www.battle.net for Quake proxy support. Quake 2 nat port udp internal:27901 27910 Red Alert nat port udp internal:8675 8675 nat port udp internal:5009 5009 What are FCS errors? FCS stands for Frame Check Sequence. Each ppp packet has a checksum attached to ensure that the data being received is the data being sent. If the FCS of an incoming packet is incorrect, the packet is dropped and the HDLC FCS count is increased. The HDLC error values can be displayed using the show hdlc command. If your link is bad (or if your serial driver is dropping packets), you will see the occasional FCS error. This is not usually worth worrying about although it does slow down the compression protocols substantially. If you have an external modem, make sure your cable is properly shielded from interference - this may eradicate the problem. If your link freezes as soon as you've connected and you see a large number of FCS errors, this may be because your link is not 8 bit clean. Make sure your modem is not using software flow control (XON/XOFF). If your datalink must use software flow control, use the command set accmap 0x000a0000 to tell ppp to escape the ^Q and ^S characters. Another reason for seeing too many FCS errors may be that the remote end has stopped talking PPP. You may want to enable async logging at this point to determine if the incoming data is actually a login or shell prompt. If you have a shell prompt at the remote end, it's possible to terminate ppp without dropping the line by using the close lcp command (a following term command will reconnect you to the shell on the remote machine. If nothing in your log file indicates why the link might have been terminated, you should ask the remote administrator (your ISP?) why the session was terminated. Why do MacOS and Windows 98 connections freeze when running PPPoE on the gateway? Thanks to Michael Wozniak mwozniak@netcom.ca for figuring this out and Dan Flemming danflemming@mac.com for the Mac solution: This is due to what's called a Black Hole router. MacOS and Windows 98 (and maybe other Microsoft OSs) send TCP packets with a requested segment size too big to fit into a PPPoE frame (MTU is 1500 by default for ethernet) and have the don't fragment bit set (default of TCP) and the Telco router is not sending ICMP must fragment back to the www site you are trying to load. (Alternatively, the router is sending the ICMP packet correctly, but the firewall at the www site is dropping it.) When the www server is sending you frames that don't fit into the PPPoE pipe the Telco router drops them on the floor and your page doesn't load (some pages/graphics do as they are smaller than a MSS.) This seems to be the default of most Telco PPPoE configurations (if only they knew how to program a router... sigh...) One fix is to use regedit on your 95/98 boxes to add the following registry entry... HKEY_LOCAL_MACHINE\System\CurrentControlSet\Services\Class\NetTrans\0000\MaxMTU It should be a string with a value 1450 (more accurately it should be 1464 to fit TCP packets into a PPPoE frame perfectly but the 1450 gives you a margin of error for other IP protocols you may encounter). This registry key is reported to have moved to Tcpip\Parameters\Interfaces\ID for adapter\MTU in Windows 2000. Refer to Microsoft Knowledge Base documents Q158474 - Windows TCPIP Registry Entries and Q120642 - TCPIP & NBT Configuration Parameters for Windows NT for more information on changing Windoze MTU to work with a FreeBSD/NAT/PPPoE router. Unfortunately, MacOS does not provide an interface for changing TCP/IP settings. However, there is commercial software available, such as OTAdvancedTuner (OT for OpenTransport, the MacOS TCP/IP stack) by Sustainable Softworks, that will allow users to customize TCP/IP settings. MacOS NAT users should select ip_interface_MTU from the drop-down menu, enter 1450 instead of 1500 in the box, click the box next to Save as Auto Configure, and click Make Active. The latest version of ppp (2.3 or greater) has an enable tcpmssfixup command that will automatically adjust the MSS to an appropriate value. This facility is enabled by default. If you're stuck with an older version of ppp, you may want to look at the tcpmssd port. None of this helps - I'm desperate! What can I do? If all else fails, send as much information as you can, including your config files, how you're starting ppp, the relevant parts of your log file and the output of the netstat -rn command (before and after connecting) to the freebsd-questions@FreeBSD.org mailing list or the comp.unix.bsd.freebsd.misc news group, and someone should point you in the right direction. Serial Communications This section answers common questions about serial communications with FreeBSD. PPP and SLIP are covered in the section. How do I tell if FreeBSD found my serial ports? As the FreeBSD kernel boots, it will probe for the serial ports in your system for which the kernel was configured. You can either watch your system closely for the messages it prints or run the command &prompt.user; dmesg | grep sio after your system's up and running. Here's some example output from the above command: sio0 at 0x3f8-0x3ff irq 4 on isa sio0: type 16550A sio1 at 0x2f8-0x2ff irq 3 on isa sio1: type 16550A This shows two serial ports. The first is on irq 4, is using port address 0x3f8, and has a 16550A-type UART chip. The second uses the same kind of chip but is on irq 3 and is at port address 0x2f8. Internal modem cards are treated just like serial ports---except that they always have a modem attached to the port. The GENERIC kernel includes support for two serial ports using the same irq and port address settings in the above example. If these settings aren't right for your system, or if you've added modem cards or have more serial ports than your kernel is configured for, just reconfigure your kernel. See section about building a kernel for more details. How do I tell if FreeBSD found my modem cards? Refer to the answer to the previous question. I just upgraded to 2.0.5 and my tty0X are missing! How do I solve this problem? Don't worry, they have been merged with the ttydX devices. You'll have to change any old configuration files you have, though. How do I access the serial ports on FreeBSD? The third serial port, sio2 (see &man.sio.4;, known as COM3 in DOS), is on /dev/cuaa2 for dial-out devices, and on /dev/ttyd2 for dial-in devices. What's the difference between these two classes of devices? You use ttydX for dial-ins. When opening /dev/ttydX in blocking mode, a process will wait for the corresponding cuaaX device to become inactive, and then wait for the carrier detect line to go active. When you open the cuaaX device, it makes sure the serial port isn't already in use by the ttydX device. If the port's available, it steals it from the ttydX device. Also, the cuaaX device doesn't care about carrier detect. With this scheme and an auto-answer modem, you can have remote users log in and you can still dialout with the same modem and the system will take care of all the conflicts. How do I enable support for a multiport serial card? Again, the section on kernel configuration provides information about configuring your kernel. For a multiport serial card, place an &man.sio.4; line for each serial port on the card in the kernel configuration file. But place the irq and vector specifiers on only one of the entries. All of the ports on the card should share one irq. For consistency, use the last serial port to specify the irq. Also, specify the COM_MULTIPORT option. The following example is for an AST 4-port serial card on irq 7: options "COM_MULTIPORT" device sio4 at isa? port 0x2a0 tty flags 0x781 device sio5 at isa? port 0x2a8 tty flags 0x781 device sio6 at isa? port 0x2b0 tty flags 0x781 device sio7 at isa? port 0x2b8 tty flags 0x781 irq 7 vector siointr The flags indicate that the master port has minor number 7 (0x700), diagnostics enabled during probe (0x080), and all the ports share an irq (0x001). Not yet. You'll have to use a different irq for each card. Can I set the default serial parameters for a port? The ttydX (or cuaaX) device is the regular device you'll want to open for your applications. When a process opens the device, it'll have a default set of terminal I/O settings. You can see these settings with the command &prompt.root; stty -a -f /dev/ttyd1 When you change the settings to this device, the settings are in effect until the device is closed. When it's reopened, it goes back to the default set. To make changes to the default set, you can open and adjust the settings of the initial state device. For example, to turn on CLOCAL mode, 8 bits, and XON/XOFF flow control by default for ttyd5, do: &prompt.root; stty -f /dev/ttyid5 clocal cs8 ixon ixoff A good place to do this is in /etc/rc.serial. Now, an application will have these settings by default when it opens ttyd5. It can still change these settings to its liking, though. You can also prevent certain settings from being changed by an application by making adjustments to the lock state device. For example, to lock the speed of ttyd5 to 57600 bps, do &prompt.root; stty -f /dev/ttyld5 57600 Now, an application that opens ttyd5 and tries to change the speed of the port will be stuck with 57600 bps. Naturally, you should make the initial state and lock state devices writable only by root. The &man.MAKEDEV.8; script does NOT do this when it creates the device entries. How can I enable dialup logins on my modem? So you want to become an Internet service provider, eh? First, you'll need one or more modems that can auto-answer. Your modem will need to assert carrier-detect when it detects a carrier and not assert it all the time. It will need to hang up the phone and reset itself when the data terminal ready (DTR) line goes from on to off. It should probably use RTS/CTS flow control or no local flow control at all. Finally, it must use a constant speed between the computer and itself, but (to be nice to your callers) it should negotiate a speed between itself and the remote modem. For many Hayes command-set--compatible modems, this command will make these settings and store them in nonvolatile memory: AT &C1 &D3 &K3 &Q6 S0=1 &W See the section on sending AT commands below for information on how to make these settings without resorting to an MS-DOS terminal program. Next, make an entry in /etc/ttys (see &man.ttys.5;) for the modem. This file lists all the ports on which the operating system will await logins. Add a line that looks something like this: ttyd1 "/usr/libexec/getty std.57600" dialup on insecure This line indicates that the second serial port (/dev/ttyd1) has a modem connected running at 57600 bps and no parity (std.57600, which comes from the file /etc/gettytab, see &man.gettytab.5;). The terminal type for this port is dialup. The port is on and is insecure---meaning root logins on the port aren't allowed. For dialin ports like this one, use the ttydX entry. It's common practice to use dialup as the terminal type. Many users set up in their .profile or .login files a prompt for the actual terminal type if the starting type is dialup. The example shows the port as insecure. To become root on this port, you have to login as a regular user, then &man.su.1; to become root. If you use secure then root can login in directly. After making modifications to /etc/ttys, you need to send a hangup or HUP signal to the &man.init.8; process: &prompt.root; kill -HUP 1 This forces the &man.init.8; process to reread /etc/ttys. The init process will then start getty processes on all on ports. You can find out if logins are available for your port by typing &prompt.user; ps -ax | grep '[t]tyd1' You should see something like: 747 ?? I 0:00.04 /usr/libexec/getty std.57600 ttyd1 How can I connect a dumb terminal to my FreeBSD box? If you're using another computer as a terminal into your FreeBSD system, get a null modem cable to go between the two serial ports. If you're using an actual terminal, see its accompanying instructions. Then, modify /etc/ttys (see &man.ttys.5;), like above. For example, if you're hooking up a WYSE-50 terminal to the fifth serial port, use an entry like this: ttyd4 "/usr/libexec/getty std.38400" wyse50 on secure This example shows that the port on /dev/ttyd4 has a wyse50 terminal connected at 38400 bps with no parity (std.38400 from /etc/gettytab, see &man.gettytab.5;) and root logins are allowed (secure). Why can't I run tip or cu? On your system, the programs &man.tip.1; and &man.cu.1; are probably executable only by uucp and group dialer. You can use the group dialer to control who has access to your modem or remote systems. Just add yourself to group dialer. Alternatively, you can let everyone on your system run &man.tip.1; and &man.cu.1; by typing: &prompt.root; chmod 4511 /usr/bin/cu &prompt.root; chmod 4511 /usr/bin/tip My stock Hayes modem isn't supported---what can I do? Actually, the man page for &man.tip.1; is out of date. There is a generic Hayes dialer already built in. Just use at=hayes in your /etc/remote (see &man.remote.5;) file. The Hayes driver isn't smart enough to recognize some of the advanced features of newer modems---messages like BUSY, NO DIALTONE, or CONNECT 115200 will just confuse it. You should turn those messages off when you use &man.tip.1; (using ATX0&W). Also, the dial timeout for &man.tip.1; is 60 seconds. Your modem should use something less, or else tip will think there's a communication problem. Try ATS7=45&W. Actually, as shipped &man.tip.1; doesn't yet support it fully. The solution is to edit the file tipconf.h in the directory /usr/src/usr.bin/tip/tip. Obviously you need the source distribution to do this. Edit the line #define HAYES 0 to #define HAYES 1. Then make and make install. Everything works nicely after that. How am I expected to enter these AT commands? Make what's called a direct entry in your /etc/remote file (see &man.remote.5;). For example, if your modem's hooked up to the first serial port, /dev/cuaa0, then put in the following line: cuaa0:dv=/dev/cuaa0:br#19200:pa=none Use the highest bps rate your modem supports in the br capability. Then, type tip cuaa0 (see &man.tip.1;) and you'll be connected to your modem. If there is no /dev/cuaa0 on your system, do this: &prompt.root; cd /dev &prompt.root; sh MAKEDEV cuaa0 Or use cu as root with the following command: &prompt.root; cu -lline -sspeed with line being the serial port (e.g. /dev/cuaa0) and speed being the speed (e.g.57600). When you are done entering the AT commands hit ~. to exit. How come the <@> sign for the pn capability does not work? The <@> sign in the phone number capability tells tip to look in /etc/phones for a phone number. But the <@> sign is also a special character in capability files like /etc/remote. Escape it with a backslash: pn=\@ How can I dial a phone number on the command line? Put what's called a generic entry in your /etc/remote file (see &man.remote.5;). For example: tip115200|Dial any phone number at 115200 bps:\ :dv=/dev/cuaa0:br#115200:at=hayes:pa=none:du: tip57600|Dial any phone number at 57600 bps:\ :dv=/dev/cuaa0:br#57600:at=hayes:pa=none:du: Then you can do something like tip -115200 5551234. If you prefer &man.cu.1; over &man.tip.1;, use a generic cu entry: cu115200|Use cu to dial any number at 115200bps:\ :dv=/dev/cuaa1:br#57600:at=hayes:pa=none:du: and type cu 5551234 -s 115200. Do I have to type in the bps rate every time I do that? Put in an entry for tip1200 or cu1200, but go ahead and use whatever bps rate is appropriate with the br capability. &man.tip.1; thinks a good default is 1200 bps which is why it looks for a tip1200 entry. You don't have to use 1200 bps, though. How can I more easily access a number of hosts through a terminal server? Rather than waiting until you're connected and typing CONNECT host each time, use tip's cm capability. For example, these entries in /etc/remote (see &man.remote.5;): pain|pain.deep13.com|Forrester's machine:\ :cm=CONNECT pain\n:tc=deep13: muffin|muffin.deep13.com|Frank's machine:\ :cm=CONNECT muffin\n:tc=deep13: deep13:Gizmonics Institute terminal server:\ :dv=/dev/cuaa2:br#38400:at=hayes:du:pa=none:pn=5551234: will let you type tip pain or tip muffin to connect to the hosts pain or muffin; and tip deep13 to get to the terminal server. Can tip try more than one line for each site? This is often a problem where a university has several modem lines and several thousand students trying to use them... Make an entry for your university in /etc/remote (see &man.remote.5;) and use <\@> for the pn capability: big-university:\ :pn=\@:tc=dialout dialout:\ :dv=/dev/cuaa3:br#9600:at=courier:du:pa=none: Then, list the phone numbers for the university in /etc/phones (see &man.phones.5;): big-university 5551111 big-university 5551112 big-university 5551113 big-university 5551114 &man.tip.1; will try each one in the listed order, then give up. If you want to keep retrying, run &man.tip.1; in a while loop. Why do I have to hit CTRL+P twice to send CTRL+P once? CTRL+P is the default force character, used to tell &man.tip.1; that the next character is literal data. You can set the force character to any other character with the ~s escape, which means set a variable. Type ~sforce=single-char followed by a newline. single-char is any single character. If you leave out single-char, then the force character is the nul character, which you can get by typing CTRL+2 or CTRL+SPACE. A pretty good value for single-char is SHIFT+CTRL+6, which I've seen only used on some terminal servers. You can have the force character be whatever you want by specifying the following in your $HOME/.tiprc file: force=single-char Why is everything I type suddenly in UPPER CASE? You must've pressed CTRL+A, &man.tip.1; raise character, specially designed for people with broken caps-lock keys. Use ~s as above and set the variable raisechar to something reasonable. In fact, you can set it to the same as the force character, if you never expect to use either of these features. Here's a sample .tiprc file perfect for Emacs users who need to type CTRL+2 and CTRL+A a lot: force=^^ raisechar=^^ The ^^ is SHIFT+CTRL+6. How can I do file transfers with tip? If you're talking to another UNIX system, you can send and receive files with ~p (put) and ~t (take). These commands run &man.cat.1; and &man.echo.1; on the remote system to accept and send files. The syntax is: ~p <local-file> [<remote-file>] ~t <remote-file> [<local-file>] There's no error checking, so you probably should use another protocol, like zmodem. How can I run zmodem with tip? First, install one of the zmodem programs from the ports collection (such as one of the two from the comms category, lrzsz or rzsz. To receive files, start the sending program on the remote end. Then, press enter and type ~C rz (or ~C lrz if you installed lrzsz) to begin receiving them locally. To send files, start the receiving program on the remote end. Then, press enter and type ~C sz files (or ~C lsz files) to send them to the remote system. How come FreeBSD cannot seem to find my serial ports, even when the settings are correct? Motherboards and cards with Acer UARTs do not probe properly under the FreeBSD sio probe. Obtain a patch from www.lemis.com to fix your problem. Miscellaneous Questions FreeBSD uses far more swap space than Linux. Why? FreeBSD only appears to use more swap than Linux. In actual fact, it does not. The main difference between FreeBSD and Linux in this regard is that FreeBSD will proactively move entirely idle, unused pages of main memory into swap in order to make more main memory available for active use. Linux tends to only move pages to swap as a last resort. The perceived heavier use of swap is balanced by the more efficient use of main memory. Note that while FreeBSD is proactive in this regard, it does not arbitrarily decide to swap pages when the system is truely idle. Thus you will not find your system all paged out when you get up in the morning after leaving it idle overnight. Why does top show very little free memory even when I have very few programs running? The simple answer is that free memory is wasted memory. Any memory that your programs don't actively allocate is used within the FreeBSD kernel as disk cache. The values shown by &man.top.1; labelled as Inact, Cache, and Buf are all cached data at different aging levels. This cached data means the system does not have to access a slow disk again for data it has accessed recently, thus increasing overall performance. In general, a low value shown for Free memory in &man.top.1; is good, provided it is not very low. Why use (what are) a.out and ELF executable formats? To understand why FreeBSD uses the ELF format, you must first know a little about the 3 currently dominant executable formats for UNIX: Prior to FreeBSD 3.x, FreeBSD used the a.out format. &man.a.out.5; The oldest and classic unix object format. It uses a short and compact header with a magic number at the beginning that's often used to characterize the format (see &man.a.out.5; for more details). It contains three loaded segments: .text, .data, and .bss plus a symbol table and a string table. COFF The SVR3 object format. The header now comprises a section table, so you can have more than just .text, .data, and .bss sections. ELF The successor to COFF, featuring Multiple sections and 32-bit or 64-bit possible values. One major drawback: ELF was also designed with the assumption that there would be only one ABI per system architecture. That assumption is actually quite incorrect, and not even in the commercial SYSV world (which has at least three ABIs: SVR4, Solaris, SCO) does it hold true. FreeBSD tries to work around this problem somewhat by providing a utility for branding a known ELF executable with information about the ABI it's compliant with. See the man page for &man.brandelf.1; for more information. FreeBSD comes from the classic camp and has traditionally used the &man.a.out.5; format, a technology tried and proven through many generations of BSD releases. Though it has also been possible for some time to build and run native ELF binaries (and kernels) on a FreeBSD system, FreeBSD initially resisted the push to switch to ELF as the default format. Why? Well, when the Linux camp made their painful transition to ELF, it was not so much to flee the a.out executable format as it was their inflexible jump-table based shared library mechanism, which made the construction of shared libraries very difficult for vendors and developers alike. Since the ELF tools available offered a solution to the shared library problem and were generally seen as the way forward anyway, the migration cost was accepted as necessary and the transition made. In FreeBSD's case, our shared library mechanism is based more closely on Sun's SunOS-style shared library mechanism and, as such, is very easy to use. However, starting with 3.0, FreeBSD officially supports ELF binaries as the default format. Even though the a.out executable format has served us well, the GNU people, who author the compiler tools we use, have dropped support for the a.out format. This has forced us to maintain a divergent version of the compler and linker, and has kept us from reaping the benefits of the latest GNU development efforts. Also the demands of ISO-C++, notably contstructors and destructors, has also led to native ELF support in future FreeBSD releases. Yes, but why are there so many different formats? Back in the dim, dark past, there was simple hardware. This simple hardware supported a simple, small system. a.out was completely adequate for the job of representing binaries on this simple system (a PDP-11). As people ported unix from this simple system, they retained the a.out format because it was sufficient for the early ports of unix to architectures like the Motorola 68k, VAXen, etc. Then some bright hardware engineer decided that if he could force software to do some sleazy tricks, then he'd be able to shave a few gates off the design and allow his CPU core to run faster. While it was made to work with this new kind of hardware (known these days as RISC), a.out was ill-suited for this hardware, so many formats were developed to get to a better performance from this hardware than the limited, simple a.out format could offer. Things like COFF, ECOFF, and a few obscure others were invented and their limitations explored before things seemed to settle on ELF. In addition, program sizes were getting huge and disks (and physical memory) were still relatively small so the concept of a shared library was born. The VM system also became more sophisticated. While each one of these advancements was done using the a.out format, its usefulness was stretched more and more with each new feature. In addition, people wanted to dynamically load things at run time, or to junk parts of their program after the init code had run to save in core memory and/or swap space. Languages became more sophistocated and people wanted code called before main automatically. Lots of hacks were done to the a.out format to allow all of these things to happen, and they basically worked for a time. In time, a.out wasn't up to handling all these problems without an ever increasing overhead in code and complexity. While ELF solved many of these problems, it would be painful to switch from the system that basically worked. So ELF had to wait until it was more painful to remain with a.out than it was to migrate to ELF. However, as time passed, the build tools that FreeBSD derived their build tools from (the assembler and loader especially) evolved in two parallel trees. The FreeBSD tree added shared libraries and fixed some bugs. The GNU folks that originally write these programs rewrote them and added simpler support for building cross compilers, plugging in different formats at will, etc. Since many people wanted to build cross compilers targeting FreeBSD, they were out of luck since the older sources that FreeBSD had for as and ld weren't up to the task. The new gnu tools chain (binutils) does support cross compiling, ELF, shared libraries, C++ extnensions, etc. In addition, many vendors are releasing ELF binaries, and it is a good thing for FreeBSD to run them. And if it is running ELF binaries, why bother having a.out any more? It is a tired old horse that has proven useful for a long time, but it is time to turn him out to pasture for his long, faithful years of service. ELF is more expressive than a.out and will allow more extensibility in the base system. The ELF tools are better maintained, and offer cross compilation support, which is important to many people. ELF may be a little slower than a.out, but trying to measure it can be difficult. There are also numerous details that are different between the two in how they map pages, handle init code, etc. None of these are very important, but they are differences. In time support for a.out will be moved out of the GENERIC kernel, and eventually removed from the kernel once the need to run legacy a.out programs is past. Why won't chmod change the permissions on symlinks? Symlinks do not have permissions, and by default, &man.chmod.1; will not follow symlinks to change the permissions on the target file. So if you have a file, foo, and a symlink to that file, bar, then this command will always succeed. &prompt.user; chmod g-w bar However, the permissions on foo will not have changed. You have to use either or together with the option to make this work. See the &man.chmod.1; and &man.symlink.7; man pages for more info. The option does a RECURSIVE &man.chmod.1;. Be careful about specifying directories or symlinks to directories to &man.chmod.1;. If you want to change the permissions of a directory referenced by a symlink, use &man.chmod.1; without any options and follow the symlink with a trailing slash (/). For example, if foo is a symlink to directory bar, and you want to change the permissions of foo (actually bar), you would do something like: &prompt.user; chmod 555 foo/ With the trailing slash, &man.chmod.1; will follow the symlink, foo, to change the permissions of the directory, bar. You'd think it'd be easy enough to change UT_NAMESIZE and rebuild the whole world, and everything would just work. Unfortunately there are often scads of applications and utilities (including system tools) that have hard-coded small numbers (not always 8 or 9, but oddball ones like 15 and 20) in structures and buffers. Not only will this get you log files which are trashed (due to variable-length records getting written when fixed records were expected), but it can break Sun's NIS clients and potentially cause other problems in interacting with other UNIX systems. In FreeBSD 3.0 and later, the maximum name length has been increased to 16 characters and those various utilities with hard-coded name sizes have been found and fixed. The fact that this touched so many areas of the system is why, in fact, the change was not made until 3.0. If you're absolutely confident in your ability to find and fix these sorts of problems for yourself when and if they pop up, you can increase the login name length in earlier releases by editing /usr/include/utmp.h and changing UT_NAMESIZE accordingly. You must also update MAXLOGNAME in /usr/include/sys/param.h to match the UT_NAMESIZE change. Finally, if you build from sources, don't forget that /usr/include is updated each time! Change the appropriate files in /usr/src/.. instead. Can I run DOS binaries under FreeBSD? Yes, starting with version 3.0 you can using BSDI's doscmd DOS emulation which has been integrated and enhanced. Send mail to The FreeBSD emulation discussion list if you're interested in joining this ongoing effort! For pre-3.0 systems, there is a neat utility called pcemu in the ports collection which emulates an 8088 and enough BIOS services to run DOS text mode applications. It requires the X Window System (provided as XFree86). Where can I find a free FreeBSD account? While FreeBSD does not provide open access to any of their servers, others do provide open access Unix systems. The charge varies and limited services may be available. Arbornet, Inc, also known as M-Net, has been providing open access to Unix systems since 1983. Starting on an Altos running System III, the site switched to BSD/OS in 1991. In June of 2000, the site switched again to FreeBSD. M-Net can be accessed via telnet and SSH and provides basic access to the entire FreeBSD software suite. However, network access is limited to members and patrons who donate to the system, which is run as a non-proft organization. M-Net also provides an bulletin board system and interactive chat. Grex provides a site very similar to M-Net including the same bulletin board and interactive chat software. However, the machine is a Sun 4M and is running SunOS What is sup, and how do I use it? SUP stands for Software Update Protocol, and was developed by CMU for keeping their development trees in sync. We used it to keep remote sites in sync with our central development sources. SUP is not bandwidth friendly, and has been retired. The current recommended method to keep your sources up to date is Handbook entry on CVSup How cool is FreeBSD? Q. Has anyone done any temperature testing while running FreeBSD? I know Linux runs cooler than dos, but have never seen a mention of FreeBSD. It seems to run really hot. A. No, but we have done numerous taste tests on blindfolded volunteers who have also had 250 micrograms of LSD-25 administered beforehand. 35% of the volunteers said that FreeBSD tasted sort of orange, whereas Linux tasted like purple haze. Neither group mentioned any significant variances in temperature. We eventually had to throw the results of this survey out entirely anyway when we found that too many volunteers were wandering out of the room during the tests, thus skewing the results. We think most of the volunteers are at Apple now, working on their new scratch and sniff GUI. It's a funny old business we're in! Seriously, both FreeBSD and Linux use the HLT (halt) instruction when the system is idle thus lowering its energy consumption and therefore the heat it generates. Also if you have APM (advanced power management) configured, then FreeBSD can also put the CPU into a low power mode. Who's scratching in my memory banks?? Q. Is there anything odd that FreeBSD does when compiling the kernel which would cause the memory to make a scratchy sound? When compiling (and for a brief moment after recognizing the floppy drive upon startup, as well), a strange scratchy sound emanates from what appears to be the memory banks. A. Yes! You'll see frequent references to daemons in the BSD documentation, and what most people don't know is that this refers to genuine, non-corporeal entities that now possess your computer. The scratchy sound coming from your memory is actually high-pitched whispering exchanged among the daemons as they best decide how to deal with various system administration tasks. If the noise gets to you, a good fdisk /mbr from DOS will get rid of them, but don't be surprised if they react adversely and try to stop you. In fact, if at any point during the exercise you hear the satanic voice of Bill Gates coming from the built-in speaker, take off running and don't ever look back! Freed from the counterbalancing influence of the BSD daemons, the twin demons of DOS and Windows are often able to re-assert total control over your machine to the eternal damnation of your soul. Now that you know, given a choice you'd probably prefer to get used to the scratchy noises, no? What does MFC mean? MFC is an acronym for Merged From -CURRENT. It's used in the CVS logs to denote when a change was migrated from the CURRENT to the STABLE branches. What does BSD mean? It stands for something in a secret language that only members can know. It doesn't translate literally but its ok to tell you that BSD's translation is something between, Formula-1 Racing Team, Penguins are tasty snacks, and We have a better sense of humor than Linux. :-) Seriously, BSD is an acronym for Berkeley Software Distribution, which is the name the Berkeley CSRG (Computer Systems Research Group) chose for their Unix distribution way back when. What is a repo-copy? A repo-copy (which is a short form of repository copy) refers to the direct copying of files within the CVS repository. Without a repo-copy, if a file needed to be copied or moved to another place in the repository, the committer would run cvs add to put the file in its new location, and then cvs rm on the old file if the old copy was being removed. The disadvantage of this method is that the history (i.e. the entries in the CVS logs) of the file would not be copied to the new location. As the FreeBSD Project considers this history very useful, a repository copy is often used instead. This is a process where one of the repository meisters will copy the files directly within the repository, rather than using the &man.cvs.1; program. Why should I care what color the bikeshed is? The really, really short answer is that you shouldn't. The somewhat longer answer is that just because you are capable of building a bikeshed doesn't mean you should stop others from building one just because you don't like the color they plan to paint it. This is a metaphor indicating that you need not argue about every little feature just because you know enough to do so. Some people have commented that the amount of noise generated by a change is inversely proportional to the complexity of the change. The longer and more complete answer is that after a very long argument about whether &man.sleep.1; should take fractional second arguments, &a.phk; posted a long message entitled A bike shed (any colour will do) on greener grass.... The appropriate portions of that message are quoted below.

&a.phk; on freebsd-hackers, October 2, 1999 What is it about this bike shed? Some of you have asked me. It's a long story, or rather it's an old story, but it is quite short actually. C. Northcote Parkinson wrote a book in the early 1960'ies, called Parkinson's Law, which contains a lot of insight into the dynamics of management. [snip a bit of commentary on the book] In the specific example involving the bike shed, the other vital component is an atomic power-plant, I guess that illustrates the age of the book. Parkinson shows how you can go in to the board of directors and get approval for building a multi-million or even billion dollar atomic power plant, but if you want to build a bike shed you will be tangled up in endless discussions. Parkinson explains that this is because an atomic plant is so vast, so expensive and so complicated that people cannot grasp it, and rather than try, they fall back on the assumption that somebody else checked all the details before it got this far. Richard P. Feynmann gives a couple of interesting, and very much to the point, examples relating to Los Alamos in his books. A bike shed on the other hand. Anyone can build one of those over a weekend, and still have time to watch the game on TV. So no matter how well prepared, no matter how reasonable you are with your proposal, somebody will seize the chance to show that he is doing his job, that he is paying attention, that he is here. In Denmark we call it setting your fingerprint. It is about personal pride and prestige, it is about being able to point somewhere and say There! I did that. It is a strong trait in politicians, but present in most people given the chance. Just think about footsteps in wet cement.

How many FreeBSD hackers does it take to change a lightbulb? One thousand, one hundred and seventy-two: Twenty-three to complain to -CURRENT about the lights being out; Four to claim that it is a configuration problem, and that such matters really belong on -questions; Three to submit PRs about it, one of which is misfiled under doc and consists only of "it's dark"; One to commit an untested lightbulb which breaks buildworld, then back it out five minutes later; Eight to flame the PR originators for not including patches in their PRs; Five to complain about buildworld being broken; Thirty-one to answer that it works for them, and they must have cvsupped at a bad time; One to post a patch for a new lightbulb to -hackers; One to complain that he had patches for this three years ago, but when he sent them to -CURRENT they were just ignored, and he has had bad experiences with the PR system; besides, the proposed new lightbulb is non-reflexive; Thirty-seven to scream that lightbulbs do not belong in the base system, that committers have no right to do things like this without consulting the Community, and WHAT IS -CORE DOING ABOUT IT!? Two hundred to complain about the color of the bicycle shed; Three to point out that the patch breaks &man.style.9;; Seventeen to complain that the proposed new lightbulb is under GPL; Five hundred and eighty-six to engage in a flame war about the comparative advantages of the GPL, the BSD license, the MIT license, the NPL, and the personal hygiene of unnamed FSF founders; Seven to move various portions of the thread to -chat and -advocacy; One to commit the suggested lightbulb, even though it shines dimmer than the old one; Two to back it out with a furious flame of a commit message, arguing that FreeBSD is better off in the dark than with a dim lightbulb; Forty-six to argue vociferously about the backing out of the dim lightbulb and demanding a statement from -core; Eleven to request a smaller lightbulb so it will fit their Tamagotchi if we ever decide to port FreeBSD to that platform; Seventy-three to complain about the SNR on -hackers and -chat and unsubscribe in protest; Thirteen to post "unsubscribe", "How do I unsubscribe?", or "Please remove me from the list", followed by the usual footer; One to commit a working lightbulb while everybody is too busy flaming everybody else to notice; Thirty-one to point out that the new lightbulb would shine 0.364% brighter if compiled with TenDRA (although it will have to be reshaped into a cube), and that FreeBSD should therefore switch to TenDRA instead of EGCS; One to complain that the new lightbulb lacks fairings; Nine (including the PR originators) to ask "what is MFC?"; Fifty-seven to complain about the lights being out two weeks after the bulb has been changed. &a.nik; adds: I was laughing quite hard at this. And then I thought, "Hang on, shouldn't there be '1 to document it.' in that list somewhere?" And then I was enlightened :-) This entry is Copyright (c) 1999 &a.des;. Please do not reproduce without attribution. For serious FreeBSD hackers only What are SNAPs and RELEASEs? There are currently three active/semi-active branches in the FreeBSD CVS Repository (the RELENG_2 branch is probably only changed twice a year, which is why there are only three active branches of development): RELENG_2_2 AKA 2.2-STABLE RELENG_3 AKA 3.X-STABLE RELENG_4 AKA 4-STABLE HEAD AKA -CURRENT AKA 5.0-CURRENT HEAD is not an actual branch tag, like the other two; it's simply a symbolic constant for the current, non-branched development stream which we simply refer to as -CURRENT. Right now, -CURRENT is the 5.0 development stream and the 4-STABLE branch, RELENG_4, forked off from -CURRENT in Mar 2000. The 2.2-STABLE branch, RELENG_2_2, departed -CURRENT in November 1996, and has pretty much been retired. How do I make my own custom release? To make a release you need to do three things: First, you need to be running a kernel with the &man.vn.4; driver configured in. Add this to your kernel config file and build a new kernel: pseudo-device vn #Vnode driver (turns a file into a device) Second, you have to have the whole CVS repository at hand. To get this you can use CVSUP but in your supfile set the release name to cvs and remove any tag or date fields: *default prefix=/home/ncvs *default base=/a *default host=cvsup.FreeBSD.org *default release=cvs *default delete compress use-rel-suffix ## Main Source Tree src-all src-eBones src-secure # Other stuff ports-all www doc-all Then run cvsup -g supfile to suck all the good bits onto your box... Finally, you need a chunk of empty space to build into. Let's say it's in /some/big/filesystem, and from the example above you've got the CVS repository in /home/ncvs: &prompt.root; setenv CVSROOT /home/ncvs # or export CVSROOT=/home/ncvs &prompt.root; cd /usr/src &prompt.root; make buildworld &prompt.root; cd /usr/src/release &prompt.root; make release BUILDNAME=3.0-MY-SNAP CHROOTDIR=/some/big/filesystem/release Please note that you do not need to build world if you already have a populated /usr/obj. An entire release will be built in /some/big/filesystem/release and you will have a full FTP-type installation in /some/big/filesystem/release/R/ftp when you're done. If you want to build your SNAP along some other branch than -CURRENT, you can also add RELEASETAG=SOMETAG to the make release command line above, e.g. RELEASETAG=RELENG_2_2 would build an up-to-the- minute 2.2-STABLE snapshot. How do I create customized installation disks? The entire process of creating installation disks and source and binary archives is automated by various targets in /usr/src/release/Makefile. The information there should be enough to get you started. However, it should be said that this involves doing a make world and will therefore take up a lot of time and disk space. Why does make world clobber my existing installed binaries? Yes, this is the general idea; as its name might suggest, make world rebuilds every system binary from scratch, so you can be certain of having a clean and consistent environment at the end (which is why it takes so long). If the environment variable DESTDIR is defined while running make world or make install, the newly-created binaries will be deposited in a directory tree identical to the installed one, rooted at ${DESTDIR}. Some random combination of shared libraries modifications and program rebuilds can cause this to fail in make world however. How come when my system boots, it says (bus speed defaulted)? The Adaptec 1542 SCSI host adapters allow the user to configure their bus access speed in software. Previous versions of the 1542 driver tried to determine the fastest usable speed and set the adapter to that. We found that this breaks some users' systems, so you now have to define the TUNE_1542 kernel configuration option in order to have this take place. Using it on those systems where it works may make your disks run faster, but on those systems where it doesn't, your data could be corrupted. Can I follow current with limited Internet access? Yes, you can do this without downloading the whole source tree by using the CTM facility. How did you split the distribution into 240k files? Newer BSD based systems have a option to split that allows them to split files on arbitrary byte boundaries. Here is an example from /usr/src/Makefile. bin-tarball: (cd ${DISTDIR}; \ tar cf - . \ gzip --no-name -9 -c | \ split -b 240640 - \ ${RELEASEDIR}/tarballs/bindist/bin_tgz.) I've written a kernel extension, who do I send it to? Please take a look at The Handbook entry on how to submit code. And thanks for the thought! How are Plug N Play ISA cards detected and initialized? By: Frank Durda IV In a nutshell, there a few I/O ports that all of the PnP boards respond to when the host asks if anyone is out there. So when the PnP probe routine starts, he asks if there are any PnP boards present, and all the PnP boards respond with their model # to a I/O read of the same port, so the probe routine gets a wired-OR yes to that question. At least one bit will be on in that reply. Then the probe code is able to cause boards with board model IDs (assigned by Microsoft/Intel) lower than X to go off-line. It then looks to see if any boards are still responding to the query. If the answer was 0, then there are no boards with IDs above X. Now probe asks if there are any boards below X. If so, probe knows there are boards with a model numbers below X. Probe then asks for boards greater than X-(limit/4) to go off-line. If repeats the query. By repeating this semi-binary search of IDs-in-range enough times, the probing code will eventually identify all PnP boards present in a given machine with a number of iterations that is much lower than what 2^64 would take. The IDs are two 32-bit fields (hence 2ˆ64) + 8 bit checksum. The first 32 bits are a vendor identifier. They never come out and say it, but it appears to be assumed that different types of boards from the same vendor could have different 32-bit vendor ids. The idea of needing 32 bits just for unique manufacturers is a bit excessive. The lower 32 bits are a serial #, ethernet address, something that makes this one board unique. The vendor must never produce a second board that has the same lower 32 bits unless the upper 32 bits are also different. So you can have multiple boards of the same type in the machine and the full 64 bits will still be unique. The 32 bit groups can never be all zero. This allows the wired-OR to show non-zero bits during the initial binary search. Once the system has identified all the board IDs present, it will reactivate each board, one at a time (via the same I/O ports), and find out what resources the given board needs, what interrupt choices are available, etc. A scan is made over all the boards to collect this information. This info is then combined with info from any ECU files on the hard disk or wired into the MLB BIOS. The ECU and BIOS PnP support for hardware on the MLB is usually synthetic, and the peripherals don't really do genuine PnP. However by examining the BIOS info plus the ECU info, the probe routines can cause the devices that are PnP to avoid those devices the probe code cannot relocate. Then the PnP devices are visited once more and given their I/O, DMA, IRQ and Memory-map address assignments. The devices will then appear at those locations and remain there until the next reboot, although there is nothing that says you can't move them around whenever you want. There is a lot of oversimplification above, but you should get the general idea. Microsoft took over some of the primary printer status ports to do PnP, on the logic that no boards decoded those addresses for the opposing I/O cycles. I found a genuine IBM printer board that did decode writes of the status port during the early PnP proposal review period, but MS said tough. So they do a write to the printer status port for setting addresses, plus that use that address + 0x800, and a third I/O port for reading that can be located anywhere between 0x200 and 0x3ff. Does FreeBSD support architectures other than the x86? Several groups of people have expressed interest in working on multi-architecture ports for FreeBSD and the FreeBSD/AXP (ALPHA) port is one such effort which has been quite successful, now available at ftp://ftp.FreeBSD.org/pub/FreeBSD/alpha. The ALPHA port currently runs on a growing number of ALPHA machine types, among them the AlphaStation, AXPpci, PC164, Miata and Multia models. For status information, please join the freebsd-alpha@FreeBSD.org mailing list. Interest has also been expressed in a port of FreeBSD to the SPARC architecture, join the freebsd-sparc@FreeBSD.org mailing list if you are interested in joining that project. Most recent additions to the list of upcoming plaforms are IA-64 and PowerPC, join the freebsd-ia64@FreeBSD.org or/and freebsd-ppc@FreeBSD.org mailing lists for more information. For general discussion on new architectures, join the freebsd-platforms@FreeBSD.org mailing list. Can you assign a major number for a device driver I've written? This depends on whether or not you plan on making the driver publicly available. If you do, then please send us a copy of the driver source code, plus the appropriate modifications to files.i386, a sample configuration file entry, and the appropriate &man.MAKEDEV.8; code to create any special files your device uses. If you do not, or are unable to because of licensing restrictions, then character major number 32 and block major number 8 have been reserved specifically for this purpose; please use them. In any case, we'd appreciate hearing about your driver on freebsd-hackers@FreeBSD.org. What about alternative layout policies for directories? In answer to the question of alternative layout policies for directories, the scheme that is currently in use is unchanged from what I wrote in 1983. I wrote that policy for the original fast filesystem, and never revisited it. It works well at keeping cylinder groups from filling up. As several of you have noted, it works poorly for find. Most filesystems are created from archives that were created by a depth first search (aka ftw). These directories end up being striped across the cylinder groups thus creating a worst possible senario for future depth first searches. If one knew the total number of directories to be created, the solution would be to create (total / fs_ncg) per cylinder group before moving on. Obviously, one would have to create some heuristic to guess at this number. Even using a small fixed number like say 10 would make an order of magnitude improvement. To differentiate restores from normal operation (when the current algorithm is probably more sensible), you could use the clustering of up to 10 if they were all done within a ten second window. Anyway, my conclusion is that this is an area ripe for experimentation. Kirk McKusick, September 1998 How can I make the most of the data I see when my kernel panics? [This section was extracted from a mail written by &a.wpaul; on the freebsd-current mailing list by &a.des;, who fixed a few typos and added the bracketed comments] From: Bill Paul <wpaul@skynet.ctr.columbia.edu> Subject: Re: the fs fun never stops To: ben@rosengart.com Date: Sun, 20 Sep 1998 15:22:50 -0400 (EDT) Cc: current@FreeBSD.org [<ben@rosengart.com> posted the following panic message] > Fatal trap 12: page fault while in kernel mode > fault virtual address = 0x40 > fault code = supervisor read, page not present > instruction pointer = 0x8:0xf014a7e5 ^^^^^^^^^^ > stack pointer = 0x10:0xf4ed6f24 > frame pointer = 0x10:0xf4ed6f28 > code segment = base 0x0, limit 0xfffff, type 0x1b > = DPL 0, pres 1, def32 1, gran 1 > processor eflags = interrupt enabled, resume, IOPL = 0 > current process = 80 (mount) > interrupt mask = > trap number = 12 > panic: page fault [When] you see a message like this, it's not enough to just reproduce it and send it in. The instruction pointer value that I highlighted up there is important; unfortunately, it's also configuration dependent. In other words, the value varies depending on the exact kernel image that you're using. If you're using a GENERIC kernel image from one of the snapshots, then it's possible for somebody else to track down the offending function, but if you're running a custom kernel then only you can tell us where the fault occured. What you should do is this: Write down the instruction pointer value. Note that the 0x8: part at the begining is not significant in this case: it's the 0xf0xxxxxx part that we want. When the system reboots, do the following: &prompt.user; nm -n /kernel.that.caused.the.panic | grep f0xxxxxx where f0xxxxxx is the instruction pointer value. The odds are you will not get an exact match since the symbols in the kernel symbol table are for the entry points of functions and the instruction pointer address will be somewhere inside a function, not at the start. If you don't get an exact match, omit the last digit from the instruction pointer value and try again, i.e.: &prompt.user; nm -n /kernel.that.caused.the.panic | grep f0xxxxx If that doesn't yield any results, chop off another digit. Repeat until you get some sort of output. The result will be a possible list of functions which caused the panic. This is a less than exact mechanism for tracking down the point of failure, but it's better than nothing. I see people constantly show panic messages like this but rarely do I see someone take the time to match up the instruction pointer with a function in the kernel symbol table. The best way to track down the cause of a panic is by capturing a crash dump, then using &man.gdb.1; to generate a stack trace on the crash dump. In any case, the method I normally use is this: Set up a kernel config file, optionally adding options DDB if you think you need the kernel debugger for something. (I use this mainly for setting beakpoints if I suspect an infinite loop condition of some kind.) Use config -g KERNELCONFIG to set up the build directory. cd /sys/compile/ KERNELCONFIG; make Wait for kernel to finish compiling. make install reboot The &man.make.1; process will have built two kernels. kernel and kernel.debug. kernel was installed as /kernel, while kernel.debug can be used as the source of debugging symbols for &man.gdb.1;. To make sure you capture a crash dump, you need edit /etc/rc.conf and set dumpdev to point to your swap partition. This will cause the &man.rc.8; scripts to use the &man.dumpon.8; command to enable crash dumps. You can also run &man.dumpon.8; manually. After a panic, the crash dump can be recovered using &man.savecore.8;; if dumpdev is set in /etc/rc.conf, the &man.rc.8; scripts will run &man.savecore.8; automatically and put the crash dump in /var/crash. FreeBSD crash dumps are usually the same size as the physical RAM size of your machine. That is, if you have 64MB of RAM, you will get a 64MB crash dump. Therefore you must make sure there's enough space in /var/crash to hold the dump. Alternatively, you run &man.savecore.8; manually and have it recover the crash dump to another directory where you have more room. It's possible to limit the size of the crash dump by using options MAXMEM=(foo) to set the amount of memory the kernel will use to something a little more sensible. For example, if you have 128MB of RAM, you can limit the kernel's memory usage to 16MB so that your crash dump size will be 16MB instead of 128MB. Once you have recovered the crash dump, you can get a stack trace with &man.gdb.1; as follows: &prompt.user; gdb -k /sys/compile/KERNELCONFIG/kernel.debug /var/crash/vmcore.0 (gdb) where Note that there may be several screens worth of information; ideally you should use &man.script.1; to capture all of them. Using the unstripped kernel image with all the debug symbols should show the exact line of kernel source code where the panic occured. Usually you have to read the stack trace from the bottom up in order to trace the exact sequence of events that lead to the crash. You can also use &man.gdb.1; to print out the contents of various variables or structures in order to examine the system state at the time of the crash. Now, if you're really insane and have a second computer, you can also configure &man.gdb.1; to do remote debugging such that you can use &man.gdb.1; on one system to debug the kernel on another system, including setting breakpoints, single-stepping through the kernel code, just like you can do with a normal user-mode program. I haven't played with this yet as I don't often have the chance to set up two machines side by side for debugging purposes. [Bill adds: "I forgot to mention one thing: if you have DDB enabled and the kernel drops into the debugger, you can force a panic (and a crash dump) just by typing 'panic' at the ddb prompt. It may stop in the debugger again during the panic phase. If it does, type 'continue' and it will finish the crash dump." -ed] Why has dlsym() stopped working for ELF executables? The ELF toolchain does not, by default, make the symbols defined in an executable visible to the dynamic linker. Consequently dlsym() searches on handles obtained from calls to dlopen(NULL, flags) will fail to find such symbols. If you want to search, using dlsym(), for symbols present in the main executable of a process, you need to link the executable using the option to the ELF linker (&man.ld.1;). How can I increase or reduce the kernel address space? By default, the kernel address space is 256 MB on FreeBSD 3.x and 1 GB on FreeBSD 4.x. If you run a network-intensive server (e.g. a large FTP or HTTP server), you might find that 256 MB is not enough. So how do you increase the address space? There are two aspects to this. First, you need to tell the kernel to reserve a larger portion of the address space for itself. Second, since the kernel is loaded at the top of the address space, you need to lower the load address so it doesn't bump its head against the ceiling. The first goal is achieved by increasing the value of NKPDE in src/sys/i386/include/pmap.h. Here's what it looks like for a 1 GB address space: #ifndef NKPDE #ifdef SMP #define NKPDE 254 /* addressable number of page tables/pde's */ #else #define NKPDE 255 /* addressable number of page tables/pde's */ #endif /* SMP */ #endif To find the correct value of NKPDE, divide the desired address space size (in megabytes) by four, then subtract one for UP and two for SMP. To achieve the second goal, you need to compute the correct load address: simply subtract the address space size (in bytes) from 0x100100000; the result is 0xc0100000 for a 1 GB address space. Set LOAD_ADDRESS in src/sys/i386/conf/Makefile.i386 to that value; then set the location counter in the beginning of the section listing in src/sys/i386/conf/kernel.script to the same value, as follows: OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386") OUTPUT_ARCH(i386) ENTRY(btext) SEARCH_DIR(/usr/lib); SEARCH_DIR(/usr/obj/elf/home/src/tmp/usr/i386-unknown-freebsdelf/lib); SECTIONS { /* Read-only sections, merged into text segment: */ . = 0xc0100000 + SIZEOF_HEADERS; .interp : { *(.interp) } Then reconfig and rebuild your kernel. You will probably have problems with &man.ps.1; &man.top.1; and the like; make world should take care of it (or a manual rebuild of libkvm, &man.ps.1; and &man.top.1; after copying the patched pmap.h to /usr/include/vm/. NOTE: the size of the kernel address space must be a multiple of four megabytes. [&a.dg; adds: I think the kernel address space needs to be a power of two, but I'm not certain about that. The old(er) boot code used to monkey with the high order address bits and I think expected at least 256MB granularity.] Acknowledgments

FreeBSD Core Team If you see a problem with this FAQ, or wish to submit an entry, please mail the &a.faq;. We appreciate your feedback, and cannot make this a better FAQ without your help!

&a.jkh; Occasional fits of FAQ-reshuffling and updating. &a.dwhite; Services above and beyond the call of duty on freebsd-questions &a.joerg; Services above and beyond the call of duty on Usenet &a.wollman; Networking and formatting Jim Lowe Multicast information &a.pds; FreeBSD FAQ typing machine slavey The FreeBSD Team Kvetching, moaning, submitting data And to any others we've forgotten, apologies and heartfelt thanks! Useful Books 4.4BSD System Manager's Manual Computer Systems Research Group, University of California, Berkeley O'Reilly and Associates 1st Edition June 1994 804 pages ISBN 1-56592-080-5 4.4BSD User's Reference Manual Computer Systems Research Group, University of California, Berkeley O'Reilly and Associates 1st Edition June 1994 905 pages ISBN 1-56592-075-9 4.4BSD User's Supplementary Documents Computer Systems Research Group, University of California, Berkeley O'Reilly and Associates 1st Edition June 1994 712 pages ISBN 1-56592-076-7 4.4BSD Programmer's Reference Manual Computer Systems Research Group, University of California, Berkeley O'Reilly and Associates 1st Edition June 1994 866 pages ISBN 1-56592-078-3 4.4BSD Programmer's Supplementary Documents Computer Systems Research Group, University of California, Berkeley O'Reilly and Associates 1st Edition June 1994 596 pages ISBN 1-56592-079-1 The Design and Implementation of the 4.4BSD Operating System M. K. McKusick Kirk Marshall Keith Bostic Michael J Karels John Quarterman Addison-Wesley

Reading MA

1996 ISBN 0-201-54979-4 Unix System Administration Handbook Evi Nemeth Garth Snyder Scott Seebass Trent R. Hein John Quarterman Prentice-Hall 3rd edition 2000 ISBN 0-13-020601-6 The Complete FreeBSD Greg Lehey Walnut Creek 3rd edition June 1999 773 pages ISBN 1-57176-246-9 McKusick et al, 1994 Berkeley Software Architecture Manual, 4.4BSD Edition M. K. McKusick M. J. Karels S. J. Leffler W. N. Joy R. S. Faber 5:1-42 diff --git a/en_US.ISO8859-1/books/fdp-primer/translations/chapter.sgml b/en_US.ISO8859-1/books/fdp-primer/translations/chapter.sgml index 04ae626fa1..85c95ba7ca 100644 --- a/en_US.ISO8859-1/books/fdp-primer/translations/chapter.sgml +++ b/en_US.ISO8859-1/books/fdp-primer/translations/chapter.sgml @@ -1,480 +1,480 @@ Translations This is the FAQ for people translating the FreeBSD documentation (FAQ, Handbook, tutorials, man pages, and others) to different languages. It is very heavily based on the translation FAQ from the FreeBSD German Documentation Project, originally written by Frank Gründer elwood@mc5sys.in-berlin.de and translated back to English by Bernd Warken bwarken@mayn.de. The FAQ maintainer is Nik Clayton nik@FreeBSD.org. Why a FAQ? More and more people are approaching the freebsd-doc mailing list and volunteering to translate FreeBSD documentation to other languages. This FAQ aims to answer their questions so they can start translating documentation as quickly as possible. What do i18n and l10n mean? i18n means internationalisation and l10n means localisation. They are just a convenient shorthand. i18n can be read as “i” followed by 18 letters, followed by “n”. Similarly, l10n is “l” followed by 10 letters, followed by “n”. Is there a mailing list for translators? Yes, freebsd-translate@ngo.org.uk. Subscribe by sending a message to freebsd-translate-request@ngo.org.uk with the word subscribe in the body of the message. You will receive a reply asking you to confirm your subscription - (in exactly the same manner as the the FreeBSD lists at FreeBSD.org). The primary language of the mailing list is English. However, posts in other languages will be accepted. The mailing list is not moderated, but you need to be a member of the list before you can post to it. The mailing list is archived, but they are not currently searchable. Sending the message help to majordomo@ngo.org.uk will send back instructions on how to access the archive. It is expected that the mailing list will transfer to FreeBSD.org and therefore become official in the near future. Are more translators needed? Yes. The more people work on translation the faster it gets done, and the faster changes to the English documentation are mirrored in the translated documents. You do not have to be a professional translator to be able to help. What languages do I need to know? Ideally, you will have a good knowledge of written English, and obviously you will need to be fluent in the language you are translating to. English is not strictly necessary. For example, you could do a Hungarian translation of the FAQ from the Spanish translation. What software do I need to know? It is strongly recommended that you maintain a local copy of the FreeBSD CVS repository (at least the documentation part) either using CTM or CVSup. The "Staying current with FreeBSD" chapter in the Handbook explains how to use these applications. You should be comfortable using CVS. This will allow you to see what has changed between different versions of the files that make up the documentation. [XXX To Do -- write a tutorial that shows how to use CVSup to get just the documentation, check it out, and see what's changed between two arbitrary revisions] How do I find out who else might be translating to the same language? The Documentation Project translations page lists the translation efforts that are currently known about. If others are already working on translating documentation to your language, please don't duplicate their efforts. Instead, contact them to see how you can help. If no one is listed on that page as translating for your language, then send a message to freebsd-doc@FreeBSD.org in case someone else is thinking of doing a translation, but hasn't announced it yet. No one else is translating to my language. What do I do? Congratulations, you have just started the “FreeBSD your-language-here Documentation Translation Project”. Welcome aboard. First, decide whether or not you've got the time to spare. Since you are the only person working on your language at the moment it is going to be your responsibility to publicise your work and coordinate any volunteers that might want to help you. Write an e-mail to the Documentation Project mailing list, announcing that you are going to translate the documentation, so the Documentation Project translations page can be maintained. You should subscribe to the freebsd-translate@ngo.org.uk mailing list (as described earlier). If there is already someone in your country providing FreeBSD mirroring services you should contact them and ask if you can have some webspace for your project, and possibly an e-mail address or mailing list services. Then pick a document and start translating. It is best to start with something fairly small—either the FAQ, or one of the tutorials. I've translated some documentation, where do I send it? That depends. If you are already working with a translation team (such as the Japanese team, or the German team) then they will have their own procedures for handling submitted documentation, and these will be outlined on their web pages. If you are the only person working on a particular language (or you are responsible for a translation project and want to submit your changes back to the FreeBSD project) then you should send your translation to the FreeBSD project (see the next question). I'm the only person working on translating to this language, how do I submit my translation? or We're a translation team, and want to submit documentation that our members have translated for us? First, make sure your translation is organised properly. This means that it should drop in to the existing documentation tree and build straight away. Currently, the FreeBSD documentation is stored in a top level directory called doc/. Directories below this are named according to the language code they are written in, as defined in ISO639 (/usr/share/misc/iso639 on a version of FreeBSD newer than 20th January 1999). If your language can be encoded in different ways (for example, Chinese) then there should be directories below this, one for each encoding format you have provided. Finally, you should have directories for each document. For example, a hypothetical Swedish translation might look like doc/ sv_SE.ISO8859-1/ Makefile books/ faq/ Makefile book.sgml sv_SE.ISO8859-1 is the name of the translation, in lang.encoding form. Note the two Makefiles, which will be used to build the documentation. Use &man.tar.1; and &man.gzip.1; to compress up your documentation, and send it to the project. &prompt.user; cd doc &prompt.user; tar cf swedish-docs.tar sv &prompt.user; gzip -9 swedish-docs.tar Put swedish-docs.tar.gz somewhere. If you do not have access to your own webspace (perhaps your ISP does not let you have any) then you can e-mail Nik Clayton nik@FreeBSD.org, and arrange to e-mail the files when it is convenient. Either way, you should use &man.send-pr.1; to submit a report indicating that you have submitted the documentation. It would be very helpful if you could get other people to look over your translation and double check it first, since it is unlikely that the person committing it will be fluent in the language. Someone (probably the Documentation Project Manager, currently Nik Clayton nik@FreeBSD.org) will then take your translation and confirm that it builds. In particular, the following things will be looked at: Do all your files use RCS strings (such as "ID")? Does make all in the sv_SE.ISO8859-1 directory work correctly? Does make install work correctly? If there are any problems then whoever is looking at the submission will get back to you to try and work them out. If there are no problems your translation will be committed as soon as possible. Can I include language or country specific text in my translation? We would prefer that you did not. For example, suppose that you are translating the Handbook to Korean, and want to include a section about retailers in Korea in your Handbook. There's no real reason why that information should not be in the English (or German, or Spanish, or Japanese, or …) versions as well. It is feasible that an English speaker in Korea might try and pick up a copy of FreeBSD whilst over there. It also helps increase FreeBSD's perceived presence around the globe, which is not a bad thing. If you have country specific information, please submit it as a change to the English Handbook (using &man.send-pr.1;) and then translate the change back to your language in the translated Handbook. Thanks. How should language specific characters be included? Non-ASCII characters in the documentation should be included using SGML entities. Briefly, these look like an ampersand (&), the name of the entity, and a semi-colon (;). The entity names are defined in ISO8879, which is in the ports tree as textproc/iso8879. A few examples include Entity Appearance Description é é Small “e” with an acute accent É É Large “E” with an acute accent ü ü Small “u” with an umlaut After you have installed the iso8879 port, the files in /usr/local/share/sgml/iso8879 contain the complete list. Addressing the reader In the English documents, the reader is addressed as “you”, there is no formal/informal distinction as there is in some languages. If you are translating to a language which does distinguish, use whichever form is typically used in other technical documentation in your language. If in doubt, use a mildly polite form. Do I need to include any additional information in my translations? Yes. The header of the English version of each document will look something like this;  The exact boilerplate may change, but it will always include a $FreeBSD$ line and the phrase The FreeBSD Documentation Project. Note that the $FreeBSD part is expanded automatically by CVS, so it should be empty (just $FreeBSD$) for new files. Your translated documents should include their own $FreeBSD$ line, and change the FreeBSD Documentation Project line to The FreeBSD language Documentation Project. In addition, you should add a third line which indicates which revision of the English text this is based on. So, the Spanish version of this file might start  diff --git a/en_US.ISO8859-1/books/handbook/advanced-networking/chapter.sgml b/en_US.ISO8859-1/books/handbook/advanced-networking/chapter.sgml index ec15f3f6f2..d38cc10648 100644 --- a/en_US.ISO8859-1/books/handbook/advanced-networking/chapter.sgml +++ b/en_US.ISO8859-1/books/handbook/advanced-networking/chapter.sgml @@ -1,3466 +1,3466 @@ Advanced Networking Synopsis The following chapter will cover some of the more frequently used network services on UNIX systems. This, of course, will pertain to configuring said services on your FreeBSD system. Gateways and Routes Contributed by &a.gryphon;. 6 October 1995. For one machine to be able to find another, there must be a mechanism in place to describe how to get from one to the other. This is called Routing. A route is a defined pair of addresses: a destination and a gateway. The pair indicates that if you are trying to get to this destination, send along through this gateway. There are three types of destinations: individual hosts, subnets, and default. The default route is used if none of the other routes apply. We will talk a little bit more about default routes later on. There are also three types of gateways: individual hosts, interfaces (also called links), and ethernet hardware addresses. An example To illustrate different aspects of routing, we will use the following example which is the output of the command netstat -r: Destination Gateway Flags Refs Use Netif Expire default outside-gw UGSc 37 418 ppp0 localhost localhost UH 0 181 lo0 test0 0:e0:b5:36:cf:4f UHLW 5 63288 ed0 77 10.20.30.255 link#1 UHLW 1 2421 foobar.com link#1 UC 0 0 host1 0:e0:a8:37:8:1e UHLW 3 4601 lo0 host2 0:e0:a8:37:8:1e UHLW 0 5 lo0 => host2.foobar.com link#1 UC 0 0 224 link#1 UC 0 0 The first two lines specify the default route (which we will cover in the next section) and the localhost route. The interface (Netif column) that it specifies to use for localhost is lo0, also known as the loopback device. This says to keep all traffic for this destination internal, rather than sending it out over the LAN, since it will only end up back where it started anyway. The next thing that stands out are the 0:e0:... addresses. These are ethernet hardware addresses. FreeBSD will automatically identify any hosts (test0 in the example) on the local ethernet and add a route for that host, directly to it over the ethernet interface, ed0. There is also a timeout (Expire column) associated with this type of route, which is used if we fail to hear from the host in a specific amount of time. In this case the route will be automatically deleted. These hosts are identified using a mechanism known as RIP (Routing Information Protocol), which figures out routes to local hosts based upon a shortest path determination. FreeBSD will also add subnet routes for the local subnet (10.20.30.255 is the broadcast address for the subnet 10.20.30, and foobar.com is the domain name associated with that subnet). The designation link#1 refers to the first ethernet card in the machine. You will notice no additional interface is specified for those. Both of these groups (local network hosts and local subnets) have their routes automatically configured by a daemon called routed. If this is not run, then only routes which are statically defined (ie. entered explicitly) will exist. The host1 line refers to our host, which it knows by ethernet address. Since we are the sending host, FreeBSD knows to use the loopback interface (lo0) rather than sending it out over the ethernet interface. The two host2 lines are an example of what happens when we use an ifconfig alias (see the section of ethernet for reasons why we would do this). The => symbol after the lo0 interface says that not only are we using the loopback (since this is address also refers to the local host), but specifically it is an alias. Such routes only show up on the host that supports the alias; all other hosts on the local network will simply have a link#1 line for such. The final line (destination subnet 224) deals with MultiCasting, which will be covered in a another section. The other column that we should talk about are the Flags. Each route has different attributes that are described in the column. Below is a short table of some of these flags and their meanings: U Up: The route is active. H Host: The route destination is a single host. G Gateway: Send anything for this destination on to this remote system, which will figure out from there where to send it. S Static: This route was configured manually, not automatically generated by the system. C Clone: Generates a new route based upon this route for machines we connect to. This type of route is normally used for local networks. W WasCloned: Indicated a route that was auto-configured based upon a local area network (Clone) route. L Link: Route involves references to ethernet hardware. Default routes When the local system needs to make a connection to remote host, it checks the routing table to determine if a known path exists. If the remote host falls into a subnet that we know how to reach (Cloned routes), then the system checks to see if it can connect along that interface. If all known paths fail, the system has one last option: the default route. This route is a special type of gateway route (usually the only one present in the system), and is always marked with a c in the flags field. For hosts on a local area network, this gateway is set to whatever machine has a direct connection to the outside world (whether via PPP link, or your hardware device attached to a dedicated data line). If you are configuring the default route for a machine which itself is functioning as the gateway to the outside world, then the default route will be the gateway machine at your Internet Service Provider's (ISP) site. Let us look at an example of default routes. This is a common configuration: [Local2] <--ether--> [Local1] <--PPP--> [ISP-Serv] <--ether--> [T1-GW] The hosts Local1 and Local2 are at your site, with the formed being your PPP connection to your ISP's Terminal Server. Your ISP has a local network at their site, which has, among other things, the server where you connect and a hardware device (T1-GW) attached to the ISP's Internet feed. The default routes for each of your machines will be: host default gateway interface Local2 Local1 ethernet Local1 T1-GW PPP A common question is Why (or how) would we set the T1-GW to be the default gateway for Local1, rather than the ISP server it is connected to?. Remember, since the PPP interface is using an address on the ISP's local network for your side of the connection, routes for any other machines on the ISP's local network will be automatically generated. Hence, you will already know how to reach the T1-GW machine, so there is no need for the intermediate step of sending traffic to the ISP server. As a final note, it is common to use the address ...1 as the gateway address for your local network. So (using the same example), if your local class-C address space was 10.20.30 and your ISP was using 10.9.9 then the default routes would be: Local2 (10.20.30.2) --> Local1 (10.20.30.1) Local1 (10.20.30.1, 10.9.9.30) --> T1-GW (10.9.9.1) Dual homed hosts There is one other type of configuration that we should cover, and that is a host that sits on two different networks. Technically, any machine functioning as a gateway (in the example above, using a PPP connection) counts as a dual-homed host. But the term is really only used to refer to a machine that sits on two local-area networks. In one case, the machine as two ethernet cards, each having an address on the separate subnets. Alternately, the machine may only have one ethernet card, and be using ifconfig aliasing. The former is used if two physically separate ethernet networks are in use, the latter if there is one physical network segment, but two logically separate subnets. Either way, routing tables are set up so that each subnet knows that this machine is the defined gateway (inbound route) to the other subnet. This configuration, with the machine acting as a Bridge between the two subnets, is often used when we need to implement packet filtering or firewall security in either or both directions. Routing propagation We have already talked about how we define our routes to the outside world, but not about how the outside world finds us. We already know that routing tables can be set up so that all traffic for a particular address space (in our examples, a class-C subnet) can be sent to a particular host on that network, which will forward the packets inbound. When you get an address space assigned to your site, your service provider will set up their routing tables so that all traffic for your subnet will be sent down your PPP link to your site. But how do sites across the country know to send to your ISP? There is a system (much like the distributed DNS information) that keeps track of all assigned address-spaces, and defines their point of connection to the Internet Backbone. The Backbone are the main trunk lines that carry Internet traffic across the country, and around the world. Each backbone machine has a copy of a master set of tables, which direct traffic for a particular network to a specific backbone carrier, and from there down the chain of service providers until it reaches your network. It is the task of your service provider to advertise to the backbone sites that they are the point of connection (and thus the path inward) for your site. This is known as route propagation. Troubleshooting Sometimes, there is a problem with routing propagation, and some sites are unable to connect to you. Perhaps the most useful command for trying to figure out where a routing is breaking down is the &man.traceroute.8; command. It is equally useful if you cannot seem to make a connection to a remote machine (i.e. &man.ping.8; fails). The &man.traceroute.8; command is run with the name of the remote host you are trying to connect to. It will show the gateway hosts along the path of the attempt, eventually either reaching the target host, or terminating because of a lack of connection. For more information, see the manual page for &man.traceroute.8;. Bridging Written by Steve Peterson steve@zpfe.com. Introduction It is sometimes useful to divide one physical network (i.e., an Ethernet segment) into two separate network segments, without having to create IP subnets and use a router to connect the segments together. A device that connects two networks together in this fashion is called a bridge. and a FreeBSD system with two network interface cards can act as a bridge. The bridge works by learning the MAC layer addresses (i.e., Ethernet addresses) of the devices on each of its network interfaces. It forwards traffic between two networks only when its source and destination are on different networks. In many respects, a bridge is like an Ethernet switch with very few ports. Situations where bridging is appropriate There are two common situations in which a bridge is used today. High traffic on a segment Situation one is where your physical network segment is overloaded with traffic, but you don't want for whatever reason to subnet the network and interconnect the subnets with a router. Let's consider an example of a newspaper where the Editorial and Production departments are on the same subnetwork. The Editorial users all use server A for file service, and the Production users are on server B. An Ethernet is used to connect all users together, and high loads on the network are slowing things down. If the Editorial users could be segregated on one network segment and the Production users on another, the two network segments could be connected with a bridge. Only the network traffic destined for interfaces on the "other" side of the bridge would be sent to the other network, reducing congestion on each network segment. Filtering/traffic shaping firewall The second common situation is where firewall functionality is needed without IP Masquerading (NAT). An example is a small company that is connected via DSL or ISDN to their ISP. They have a 13 address global IP allocation for their ISP and have 10 PCs on their network. In this situation, using a router-based firewall is difficult because of subnetting issues. A bridge-based firewall can be configured and dropped into the path just downstream of their DSL/ISDN router without any IP numbering issues. Configuring a bridge Network interface card selection A bridge requires at least two network cards to function. Unfortunately, not all network interface cards as of FreeBSD 4.0 support bridging. Read &man.bridge.4; for details on the cards that are supported. Install and test the two network cards before continuing. Kernel configuration changes To enable kernel support for bridging, add the options BRIDGE statement to your kernel configuration file, and rebuild your kernel. Firewall support If you are planning to use the bridge as a firewall, you will need to add the IPFIREWALL option as well. Read for general information on configuring the bridge as a firewall. If you need to allow non-IP packets (such as ARP) to flow through the bridge, there is an undocumented firewall option that must be set. This option is IPFIREWALL_DEFAULT_TO_ACCEPT. Note that this changes the default rule for the firewall to accept any packet. Make sure you know how this changes the meaning of your ruleset before you set it. Traffic shaping support If you want to use the bridge as a traffic shaper, you will need to add the DUMMYNET option to your kernel configuration. Read &man.dummynet.4; for further information. Enabling the bridge Add the line net.link.ether.bridge=1 to /etc/sysctl.conf to enable the bridge at runtime. If you want the bridged packets to be filtered by ipfw, you should also add net.link.ether.bridge_ipfw=1 as well. Performance My bridge/firewall is a Pentium 90 with one 3Com 3C900B and one 3C905B. The protected side of the network runs at 10mbps half duplex and the connection between the bridge and my router (a Cisco 675) runs at 100mbps full duplex. With no filtering enabled, I've found that the bridge adds about 0.4 milliseconds of latency to pings from the protected 10mbps network to the Cisco 675. Other information If you want to be able to telnet into the bridge from the network, it is OK to assign one of the network cards an IP address. The consensus is that assigning both cards an address is a bad idea. If you have multiple bridges on your network, there cannot be more than one path between any two workstations. Technically, this means that there is no support for spanning tree link management. NFS Written by &a.unfurl;, 4 March 2000. Among the many different file systems that FreeBSD supports is a very unique type, the Network File System or NFS. NFS allows you to share directories and files on one machine with one or more other machines via the network they are attached to. Using NFS, users and programs can access files on remote systems as if they were local files. NFS has several benefits: Local workstations don't need as much disk space because commonly used data can be stored on a single machine and still remain accessible to everyone on the network. There is no need for users to have unique home directories on every machine on your network. Once they have an established directory that is available via NFS it can be accessed from anywhere. Storage devices such as floppies and CD-ROM drives can be used by other machines on the network eliminating the need for extra hardware. How It Works NFS is composed of two sides – a client side and a server side. Think of it as a want/have relationship. The client wants the data that the server side has. The server shares its data with the client. In order for this system to function properly a few processes have to be configured and running properly. The server has to be running the following daemons: nfsd - The NFS Daemon which services requests from NFS clients. mountd - The NFS Mount Daemon which actually carries out requests that nfsd passes on to it. portmap - The portmapper daemon which allows NFS clients to find out which port the NFS server is using. The client side only needs to run a single daemon: nfsiod - The NFS async I/O Daemon which services requests from its NFS server. Configuring NFS Luckily for us, on a FreeBSD system this setup is a snap. The processes that need to be running can all be run at boot time with a few modifications to your /etc/rc.conf file. On the NFS server make sure you have: portmap_enable="YES" nfs_server_enable="YES" nfs_server_flags="-u -t -n 4" mountd_flags="-r" mountd is automatically run whenever the NFS server is enabled. The and flags to nfsd tell it to serve UDP and TCP clients. The flag tells nfsd to start 4 copies of itself. On the client, make sure you have: nfs_client_enable="YES" nfs_client_flags="-n 4" Like nfsd, the tells nfsiod to start 4 copies of itself. The last configuration step requires that you create a file called /etc/exports. The exports file specifies which file systems on your server will be shared (a.k.a., exported) and with what clients they will be shared. Each line in the file specifies a file system to be shared. There are a handful of options that can be used in this file but only a few will be mentioned here. You can find out about the rest in the &man.exports.5; man page. Here are a few example /etc/exports entries: The following line exports /cdrom to three silly machines that have the same domain name as the server (hence the lack of a domain name for each) or have entries in your /etc/hosts file. The flag makes the shared file system read-only. With this flag, the - remote system will not be able to make any changes to the the + remote system will not be able to make any changes to the shared file system. /cdrom -ro moe larry curly The following line exports /home to three hosts by IP address. This is a useful setup if you have a private network but do not have DNS running. The flag allows all the directories below the specified file system to be exported as well. /home -alldirs 10.0.0.2 10.0.0.3 10.0.0.4 The following line exports /a to two machines that have different domain names than the server. The flag allows the root user on the remote system to write to the shared file system as root. Without the -maproot=0 flag even if someone has root access on the remote system they won't be able to modify files on the shared file system. /a -maproot=0 host.domain.com box.example.com In order for a client to share an exported file system it must have permission to do so. Make sure your client is listed in your /etc/exports file. It's important to remember that you must restart mountd whenever you modify /etc/exports so that your changes take effect. This can be accomplished by sending the hangup signal to the mountd process : &prompt.root; kill -HUP `cat /var/run/mountd.pid` Now that you have made all these changes you can just reboot and let FreeBSD start everything for you at boot time or you can run the following commands as root: On the NFS server: &prompt.root; portmap &prompt.root; nfsd -u -t -n 4 &prompt.root; mountd -r On the NFS client: &prompt.root; nfsiod -n 4 Now you should be ready to actually mount a remote file system. This can be done one of two ways. In these examples the server's name will be server and the client's name will be client. If you just want to temporarily mount a remote file system or just want to test out your config you can run a command like this as root on the client: &prompt.root; mount server:/home /mnt This will mount /home on the server on /mnt on the client. If everything is setup correctly you should be able to go into /mnt on the client and see all the files that are on the server. If you want to permanently (each time you reboot) mount a remote file system you need to add it to your /etc/fstab file. Here is an example line: server:/home /mnt nfs rw 0 0 Read the &man.fstab.5; man page for more options. Practical Uses There are many very cool uses for NFS. Some of the more common ones are listed below. Have several machines on a network and share a CD-ROM or floppy drive among them. This is cheaper and often more convenient. With so many machines on a network, it gets old having your personal files strewn all over the place. You can have a central NFS server that houses all user home directories and shares them with the rest of the machines on the LAN, so no matter where you log in you will have the same home directory. When you get to reinstalling FreeBSD on one of your machines, NFS is the way to go! Just pop your distribution CD-ROM into your file server and away you go! Have a common /usr/ports/distfiles directory that all your machines share. That way, when you go to install a port that you've already installed on a different machine, you do not have to download the source all over again! Problems integrating with other systems Contributed by &a.jlind;. Certain Ethernet adapters for ISA PC systems have limitations which can lead to serious network problems, particularly with NFS. This difficulty is not specific to FreeBSD, but FreeBSD systems are affected by it. The problem nearly always occurs when (FreeBSD) PC systems are networked with high-performance workstations, such as those made by Silicon Graphics, Inc., and Sun Microsystems, Inc. The NFS mount will work fine, and some operations may succeed, but suddenly the server will seem to become unresponsive to the client, even though requests to and from other systems continue to be processed. This happens to the client system, whether the client is the FreeBSD system or the workstation. On many systems, there is no way to shut down the client gracefully once this problem has manifested itself. The only solution is often to reset the client, because the NFS situation cannot be resolved. Though the correct solution is to get a higher performance and capacity Ethernet adapter for the FreeBSD system, there is a simple workaround that will allow satisfactory operation. If the FreeBSD system is the server, include the option on the mount from the client. If the FreeBSD system is the client, then mount the NFS file system with the option . These options may be specified using the fourth field of the fstab entry on the client for automatic mounts, or by using the parameter of the mount command for manual mounts. It should be noted that there is a different problem, sometimes mistaken for this one, when the NFS servers and clients are on different networks. If that is the case, make certain that your routers are routing the necessary UDP information, or you will not get anywhere, no matter what else you are doing. In the following examples, fastws is the host (interface) name of a high-performance workstation, and freebox is the host (interface) name of a FreeBSD system with a lower-performance Ethernet adapter. Also, /sharedfs will be the exported NFS filesystem (see man exports), and /project will be the mount point on the client for the exported file system. In all cases, note that additional options, such as or and may be desirable in your application. Examples for the FreeBSD system (freebox) as the client: in /etc/fstab on freebox: fastws:/sharedfs /project nfs rw,-r=1024 0 0 As a manual mount command on freebox: &prompt.root; mount -t nfs -o -r=1024 fastws:/sharedfs /project Examples for the FreeBSD system as the server: in /etc/fstab on fastws: freebox:/sharedfs /project nfs rw,-w=1024 0 0 As a manual mount command on fastws: &prompt.root; mount -t nfs -o -w=1024 freebox:/sharedfs /project Nearly any 16-bit Ethernet adapter will allow operation without the above restrictions on the read or write size. For anyone who cares, here is what happens when the failure occurs, which also explains why it is unrecoverable. NFS typically works with a block size of 8k (though it may do fragments of smaller sizes). Since the maximum Ethernet packet is around 1500 bytes, the NFS block gets split into multiple Ethernet packets, even though it is still a single unit to the upper-level code, and must be received, assembled, and acknowledged as a unit. The high-performance workstations can pump out the packets which comprise the NFS unit one right after the other, just as close together as the standard allows. On the smaller, lower capacity cards, the later packets overrun the earlier packets of the same unit before they can be transferred to the host and the unit as a whole cannot be reconstructed or acknowledged. As a result, the workstation will time out and try again, but it will try again with the entire 8K unit, and the process will be repeated, ad infinitum. By keeping the unit size below the Ethernet packet size limitation, we ensure that any complete Ethernet packet received can be acknowledged individually, avoiding the deadlock situation. Overruns may still occur when a high-performance workstations is slamming data out to a PC system, but with the better cards, such overruns are not guaranteed on NFS units. When an overrun occurs, the units affected will be retransmitted, and there will be a fair chance that they will be received, assembled, and acknowledged. Diskless Operation Contributed by &a.martin;. netboot.com/netboot.rom allow you to boot your FreeBSD machine over the network and run FreeBSD without having a disk on your client. Under 2.0 it is now possible to have local swap. Swapping over NFS is also still supported. Supported Ethernet cards include: Western Digital/SMC 8003, 8013, 8216 and compatibles; NE1000/NE2000 and compatibles (requires recompile) Setup Instructions Find a machine that will be your server. This machine will require enough disk space to hold the FreeBSD 2.0 binaries and have bootp, tftp and NFS services available. Tested machines: HP9000/8xx running HP-UX 9.04 or later (pre 9.04 doesn't work) Sun/Solaris 2.3. (you may need to get bootp) Set up a bootp server to provide the client with IP, gateway, netmask. diskless:\ :ht=ether:\ :ha=0000c01f848a:\ :sm=255.255.255.0:\ :hn:\ :ds=192.1.2.3:\ :ip=192.1.2.4:\ :gw=192.1.2.5:\ :vm=rfc1048: Set up a TFTP server (on same machine as bootp server) to provide booting information to client. The name of this file is cfg.X.X.X.X (or /tftpboot/cfg.X.X.X.X, it will try both) where X.X.X.X is the IP address of the client. The contents of this file can be any valid netboot commands. Under 2.0, netboot has the following commands: help print help list ip print/set client's IP address server print/set bootp/tftp server address netmask print/set netmask hostname name print/set hostname kernel print/set kernel name rootfs print/set root filesystem swapfs print/set swap filesystem swapsize set diskless swapsize in KBytes diskboot boot from disk autoboot continue boot process trans | turn transceiver on|off flags set boot flags A typical completely diskless cfg file might contain: rootfs 192.1.2.3:/rootfs/myclient swapfs 192.1.2.3:/swapfs swapsize 20000 hostname myclient.mydomain A cfg file for a machine with local swap might contain: rootfs 192.1.2.3:/rootfs/myclient hostname myclient.mydomain Ensure that your NFS server has exported the root (and swap if applicable) filesystems to your client, and that the client has root access to these filesystems A typical /etc/exports file on FreeBSD might look like: /rootfs/myclient -maproot=0:0 myclient.mydomain /swapfs -maproot=0:0 myclient.mydomain And on HP-UX: /rootfs/myclient -root=myclient.mydomain /swapfs -root=myclient.mydomain If you are swapping over NFS (completely diskless configuration) create a swap file for your client using dd. If your swapfs command has the arguments /swapfs and the size 20000 as in the example above, the swapfile for myclient will be called /swapfs/swap.X.X.X.X where X.X.X.X is the client's IP addr, e.g.: &prompt.root; dd if=/dev/zero of=/swapfs/swap.192.1.2.4 bs=1k count=20000 Also, the client's swap space might contain sensitive information once swapping starts, so make sure to restrict read and write access to this file to prevent unauthorized access: &prompt.root; chmod 0600 /swapfs/swap.192.1.2.4 Unpack the root filesystem in the directory the client will use for its root filesystem (/rootfs/myclient in the example above). On HP-UX systems: The server should be running HP-UX 9.04 or later for HP9000/800 series machines. Prior versions do not allow the creation of device files over NFS. When extracting /dev in /rootfs/myclient, beware that some systems (HPUX) will not create device files that FreeBSD is happy with. You may have to go to single user mode on the first bootup (press control-c during the bootup phase), cd /dev and do a sh ./MAKEDEV all from the client to fix this. Run netboot.com on the client or make an EPROM from the netboot.rom file Using Shared <filename>/</filename> and <filename>/usr</filename> filesystems Although this is not an officially sanctioned or supported way of doing this, some people report that it works quite well. If anyone has any suggestions on how to do this cleanly, please tell &a.doc;. Compiling netboot for specific setups Netboot can be compiled to support NE1000/2000 cards by changing the configuration in /sys/i386/boot/netboot/Makefile. See the comments at the top of this file. ISDN Last modified by &a.wlloyd;. A good resource for information on ISDN technology and hardware is Dan Kegel's ISDN Page. A quick simple road map to ISDN follows: If you live in Europe you might want to investigate the ISDN card section. If you are planning to use ISDN primarily to connect to the Internet with an Internet Provider on a dial-up non-dedicated basis, you might look into Terminal Adapters. This will give you the most flexibility, with the fewest problems, if you change providers. If you are connecting two LANs together, or connecting to the Internet with a dedicated ISDN connection, you might consider the stand alone router/bridge option. Cost is a significant factor in determining what solution you will choose. The following options are listed from least expensive to most expensive. ISDN Cards Contributed by &a.hm;. This section is really only relevant to ISDN users in countries where the DSS1/Q.931 ISDN standard is supported. Some growing number of PC ISDN cards are supported under FreeBSD 2.2.x and up by the isdn4bsd driver package. It is still under development but the reports show that it is successfully used all over Europe. The latest isdn4bsd version is available from ftp://isdn4bsd@ftp.consol.de/pub/, the main isdn4bsd ftp site (you have to log in as user isdn4bsd , give your mail address as the password and change to the pub directory. Anonymous ftp as user ftp or anonymous will not give the desired result). Isdn4bsd allows you to connect to other ISDN routers using either IP over raw HDLC or by using synchronous PPP. A telephone answering machine application is also available. Many ISDN PC cards are supported, mostly the ones with a Siemens ISDN chipset (ISAC/HSCX), support for other chipsets (from Motorola, Cologne Chip Designs) is currently under development. For an up-to-date list of supported cards, please have a look at the README file. In case you are interested in adding support for a different ISDN protocol, a currently unsupported ISDN PC card or otherwise enhancing isdn4bsd, please get in touch with hm@kts.org. A majordomo maintained mailing list is available. To join the list, send mail to &a.majordomo; and specify: subscribe freebsd-isdn in the body of your message. ISDN Terminal Adapters Terminal adapters(TA), are to ISDN what modems are to regular phone lines. Most TA's use the standard hayes modem AT command set, and can be used as a drop in replacement for a modem. A TA will operate basically the same as a modem except connection and throughput speeds will be much faster than your old modem. You will need to configure PPP exactly the same as for a modem setup. Make sure you set your serial speed as high as possible. The main advantage of using a TA to connect to an Internet Provider is that you can do Dynamic PPP. As IP address space becomes more and more scarce, most providers are not willing to provide you with a static IP anymore. Most stand-alone routers are not able to accommodate dynamic IP allocation. TA's completely rely on the PPP daemon that you are running for their features and stability of connection. This allows you to upgrade easily from using a modem to ISDN on a FreeBSD machine, if you already have PPP setup. However, at the same time any problems you experienced with the PPP program and are going to persist. If you want maximum stability, use the kernel PPP option, not the user-land iijPPP. The following TA's are know to work with FreeBSD. Motorola BitSurfer and Bitsurfer Pro Adtran Most other TA's will probably work as well, TA vendors try to make sure their product can accept most of the standard modem AT command set. The real problem with external TA's is like modems you need a good serial card in your computer. You should read the serial ports section in the handbook for a detailed understanding of serial devices, and the differences between asynchronous and synchronous serial ports. A TA running off a standard PC serial port (asynchronous) limits you to 115.2Kbs, even though you have a 128Kbs connection. To fully utilize the 128Kbs that ISDN is capable of, you must move the TA to a synchronous serial card. Do not be fooled into buying an internal TA and thinking you have avoided the synchronous/asynchronous issue. Internal TA's simply have a standard PC serial port chip built into them. All this will do, is save you having to buy another serial cable, and find another empty electrical socket. A synchronous card with a TA is at least as fast as a stand-alone router, and with a simple 386 FreeBSD box driving it, probably more flexible. The choice of sync/TA v.s. stand-alone router is largely a religious issue. There has been some discussion of this in the mailing lists. I suggest you search the archives for the complete discussion. Stand-alone ISDN Bridges/Routers ISDN bridges or routers are not at all specific to FreeBSD or any other operating system. For a more complete description of routing and bridging technology, please refer to a Networking reference book. In the context of this page, the terms router and bridge will be used interchangeably. As the cost of low end ISDN routers/bridges comes down, it will likely become a more and more popular choice. An ISDN router is a small box that plugs directly into your local Ethernet network(or card), and manages its own connection to the other bridge/router. It has all the software to do PPP and other protocols built in. A router will allow you much faster throughput that a standard TA, since it will be using a full synchronous ISDN connection. The main problem with ISDN routers and bridges is that interoperability between manufacturers can still be a problem. If you are planning to connect to an Internet provider, you should discuss your needs with them. If you are planning to connect two lan segments together, ie: home lan to the office lan, this is the simplest lowest maintenance solution. Since you are buying the equipment for both sides of the connection you can be assured that the link will work. For example to connect a home computer or branch office network to a head office network the following setup could be used. Branch office or Home network Network is 10 Base T Ethernet. Connect router to network cable with AUI/10BT transceiver, if necessary. ---Sun workstation | ---FreeBSD box | ---Windows 95 (Do not admit to owning it) | Stand-alone router | ISDN BRI line If your home/branch office is only one computer you can use a twisted pair crossover cable to connect to the stand-alone router directly. Head office or other lan Network is Twisted Pair Ethernet. -------Novell Server | H | | ---Sun | | | U ---FreeBSD | | | ---Windows 95 | B | |___---Stand-alone router | ISDN BRI line One large advantage of most routers/bridges is that they allow you to have 2 separate independent PPP connections to 2 separate sites at the same time. This is not supported on most TA's, except for specific(expensive) models that have two serial ports. Do not confuse this with channel bonding, MPP etc. This can be very useful feature, for example if you have an dedicated ISDN connection at your office and would like to tap into it, but don't want to get another ISDN line at work. A router at the office location can manage a dedicated B channel connection (64Kbs) to the internet, as well as a use the other B channel for a separate data connection. The second B channel can be used for dial-in, dial-out or dynamically bond(MPP etc.) with the first B channel for more bandwidth. An Ethernet bridge will also allow you to transmit more than just IP traffic, you can also send IPX/SPX or whatever other protocols you use. NIS/YP Written by &a.unfurl;, 21 January 2000, enhanced with parts and comments from Eric Ogren eogren@earthlink.net and Udo Erdelhoff ue@nathan.ruhr.de in June 2000. What is it? NIS, which stands for Network Information Services, was developed by Sun Microsystems to centralize administration of Unix (originally SunOS) systems. It has now essentially become an industry standard; all major Unices (Solaris, HP-UX, AIX, Linux, NetBSD, OpenBSD, FreeBSD, etc) support NIS. NIS was formerly known as Yellow Pages (or yp), but due to copyright violations, Sun was forced to change the name. It is a RPC-based client/server system that allows a group of machines within an NIS domain to share a common set of configuration files. This permits a system administrator to set up NIS client systems with only minimal configuration data and add, remove or modify configuration data from a single location. It is similar to Windows NT's domain system; although the internal implementation of the two aren't at all similar, the basic functionality can be compared. Terms/processes you should know There are several terms and several important user processes that you will come across when attempting to implement NIS on FreeBSD, whether you are trying to create an NIS server or act an NIS client: The NIS domainname. An NIS master server and all of its clients (including its slave servers) have a NIS domainname. Similar to an NT domain name, the NIS domainname does not have anything to do with DNS. portmap. portmap must be running in order to enable RPC (Remote Procedure Call, a network protocol used by NIS). If portmap is not running, it will be impossible to run an NIS server, or to act as an NIS client. ypbind. ypbind “binds” an NIS client to its NIS server. It will take the NIS domainname from the system, and using RPC, connect to the server. ypbind is the core of client-server communication in an NIS environment; if ypbind dies on a client machine, it will not be able to access the NIS server. ypserv. ypserv, which should only be running on NIS servers, is the NIS server process itself. If ypserv dies, then the server will no longer be able to respond to NIS requests (hopefully, there is a slave server to take over for it). There are some implementations of NIS (but not the FreeBSD one), that don't try to reconnect to another server if the server it used before dies. Often, the only thing that helps in this case is to restart the server process (or even the whole server) or the ypbind process on the client. rpc.yppasswdd. rpc.yppasswdd, another process that should only be running on NIS master servers, is a daemon that will allow NIS clients to change their NIS passwords. If this daemon is not running, users will have to login to the NIS master server and change their passwords there. How does it work? There are three types of hosts in an NIS environment; master servers, slave servers, and clients. Servers act as a central repository for host configuration information. Master servers hold the authoritative copy of this information, while slave servers mirror this information for redundancy. Clients rely on the servers to provide this information to them. Information in many files can be shared in this manner. The master.passwd, group, and hosts files are commonly shared via NIS. Whenever a process on a client needs information that would normally be found in these files locally, it makes a query to the server it is bound to, to get this information. Machine types A NIS master server. This server, analogous to a Windows NT primary domain controller, maintains the files used by all of the NIS clients. The passwd, group, and other various files used by the NIS clients live on the master server. It is possible for one machine to be an NIS master server for more than one NIS domain. However, this will not be covered in this introduction, which assumes a relatively small-scale NIS environment. NIS slave servers. Similar to NT's backup domain controllers, NIS slave servers maintain copies of the NIS master's data files. NIS slave servers provide the redundancy, which is needed in important environments. They also help to balance the load of the master server: NIS Clients always attach to the NIS server whose response they get first, and this includes slave-server-replies. NIS clients. NIS clients, like most NT workstations, authenticate against the NIS server (or the NT domain controller in the NT Workstation case) to log on. Using NIS/YP This section will deal with setting up a sample NIS environment. This section assumes that you are running FreeBSD 3.3 or later. The instructions given here will probably work for any version of FreeBSD greater than 3.0, but there are no guarantees that this is true. Planning Let's assume that you are the administrator of a small university lab. This lab, which consists of 15 FreeBSD machines, currently has no centralized point of administration; each machine has its own /etc/passwd and /etc/master.passwd. These files are kept in sync with each other only through manual intervention; currently, when you add a user to the lab, you must run adduser on all 15 machines. Clearly, this has to change, so you have decided to convert the lab to use NIS, using two of the machines as servers. Therefore, the configuration of the lab now looks something like: Machine name IP address Machine role ellington 10.0.0.2 NIS master coltrane 10.0.0.3 NIS slave basie 10.0.0.4 Faculty workstation bird 10.0.0.5 Client machine cli[1-11] 10.0.0.[6-17] Other client machines If you are setting up a NIS scheme for the first time, it is a good idea to think through how you want to go about it. No matter what the size of your network, there are a few decisions that need to be made. Choosing a NIS Domain Name This might not be the domainname that you are used to. It is more accurately called the NIS domainname. When a client broadcasts its requests for info, it includes the name of the NIS domain that it is part of. This is how multiple servers on one network can tell which server should answer which request. Think of the NIS domainname as the name for a group of hosts that are related in some way. Some organizations choose to use their Internet domainname for their NIS domainname. This is not recommended as it can cause confusion when trying to debug network problems. The NIS domainname should be unique within your network and it is helpful if it describes the group of machines it represents. For example, the Art department at Acme Inc. might be in the "acme-art" NIS domain. For this example, assume you have chosen the name test-domain. However, some operating systems (notably SunOS) use their NIS domain name as their Internet domain name. If one or more machines on your network have this restriction, you must use the Internet domain name as your NIS domain name. Physical Server Requirements There are several things to keep in mind when choosing a machine to use as a NIS server. One of the unfortunate things about NIS is the level of dependency the clients have on the server. If a client cannot contact the server for its NIS domain, very often the machine becomes unusable. The lack of user and group information causes most systems to temporarily freeze up. With this in mind you should make sure to choose a machine that won't be prone to being rebooted regularly, or one that might be used for development. The NIS server should ideally be a stand alone machine whose sole purpose in life is to be an NIS server. If you have a network that is not very heavily used, it is acceptable to put the NIS server on a machine running other services, just keep in mind that if the NIS server becomes unavailable, it will affect all of your NIS clients adversely. NIS Servers The canonical copies of all NIS information are stored on a single machine called the NIS master server. The databases used to store the information are called NIS maps. In FreeBSD, these maps are stored in /var/yp/[domainname] where [domainname] is the name of the NIS domain being served. A single NIS server can support several domains at once, therefore it is possible to have several such directories, one for each supported domain. Each domain will have its own independent set of maps. NIS master and slave servers handle all NIS requests with the ypserv daemon. Ypserv is responsible for receiving incoming requests from NIS clients, translating the requested domain and map name to a path to the corresponding database file and transmitting data from the database back to the client. Setting up a NIS master server Setting up a master NIS server can be relatively straight forward, depending on your needs. FreeBSD comes with support for NIS out-of-the-box. All you need is to add the following lines to /etc/rc.conf, and FreeBSD will do the rest for you. nisdomainname="test-domain" This line will set the NIS domainname to test-domain upon network setup (e.g. after reboot). nis_server_enable="YES" This will tell FreeBSD to start up the NIS server processes when the networking is next brought up. nis_yppasswdd_enable="YES" This will enable the rpc.yppasswdd daemon, which, as mentioned above, will allow users to change their NIS password from a client machine. Now, all you have to do is to run the command /etc/netstart as superuser. It will setup everything for you, using the values you defined in /etc/rc.conf. Initializing the NIS maps The NIS maps are database files, that are kept in the /var/yp directory. They are generated from configuration files in the /etc directory of the NIS master, with one exception: the /etc/master.passwd file. This is for a good reason; you don't want to propagate passwords to your root and other administrative accounts to all the servers in the NIS domain. Therefore, before we initialize the NIS maps, you should: &prompt.root; cp /etc/master.passwd /var/yp/master.passwd &prompt.root; cd /var/yp &prompt.root; vi master.passwd You should remove all entries regarding system accounts (bin, tty, kmem, games, etc), as well as any accounts that you don't want to be propagated to the NIS clients (for example root and any other UID 0 (superuser) accounts). Make sure the /var/yp/master.passwd is neither group nor world readable (mode 600)! Use the chmod command, if appropriate. When you have finished, it's time to initialize the NIS maps! FreeBSD includes a script named ypinit to do this for you (see its man page for more information). Note that this script is available on most UNIX OSs, but not on all. On Digital Unix/Compaq Tru64 Unix it is called ypsetup. Because we are generating maps for an NIS master, we are going to pass the option to ypinit. To generate the NIS maps, assuming you already performed the steps above, run: ellington&prompt.root; ypinit -m test-domain Server Type: MASTER Domain: test-domain Creating an YP server will require that you answer a few questions. Questions will all be asked at the beginning of the procedure. Do you want this procedure to quit on non-fatal errors? [y/n: n] n Ok, please remember to go back and redo manually whatever fails. If you don't, something might not work. At this point, we have to construct a list of this domains YP servers. rod.darktech.org is already known as master server. Please continue to add any slave servers, one per line. When you are done with the list, type a <control D>. master server : ellington next host to add: coltrane next host to add: ^D The current list of NIS servers looks like this: ellington coltrane Is this correct? [y/n: y] y [..output from map generation..] NIS Map update completed. ellington has been setup as an YP master server without any errors. ypinit should have created /var/yp/Makefile from /var/yp/Makefile.dist. When created, this file assumes that you are operating in a single server NIS environment with only FreeBSD machines. Since test-domain has a slave server as well, you must edit /var/yp/Makefile: ellington&prompt.root; vi /var/yp/Makefile You should comment out the line that says `NOPUSH = "True"' (if it is not commented out already). Setting up a NIS slave server Setting up an NIS slave server is even more simple than setting up the master. Log on to the slave server and edit the file /etc/rc.conf as you did before. The only difference is that we now must use the option when running ypinit. The option requires the name of the NIS master be passed to it as well, so our command line looks like: coltrane&prompt.root; ypinit -s ellington test-domain Server Type: SLAVE Domain: test-domain Master: ellington Creating an YP server will require that you answer a few questions. Questions will all be asked at the beginning of the procedure. Do you want this procedure to quit on non-fatal errors? [y/n: n] n Ok, please remember to go back and redo manually whatever fails. If you don't, something might not work. There will be no further questions. The remainder of the procedure should take a few minutes, to copy the databases from ellington. Transferring netgroup... ypxfr: Exiting: Map successfully transferred Transferring netgroup.byuser... ypxfr: Exiting: Map successfully transferred Transferring netgroup.byhost... ypxfr: Exiting: Map successfully transferred Transferring master.passwd.byuid... ypxfr: Exiting: Map successfully transferred Transferring passwd.byuid... ypxfr: Exiting: Map successfully transferred Transferring passwd.byname... ypxfr: Exiting: Map successfully transferred Transferring group.bygid... ypxfr: Exiting: Map successfully transferred Transferring group.byname... ypxfr: Exiting: Map successfully transferred Transferring services.byname... ypxfr: Exiting: Map successfully transferred Transferring rpc.bynumber... ypxfr: Exiting: Map successfully transferred Transferring rpc.byname... ypxfr: Exiting: Map successfully transferred Transferring protocols.byname... ypxfr: Exiting: Map successfully transferred Transferring master.passwd.byname... ypxfr: Exiting: Map successfully transferred Transferring networks.byname... ypxfr: Exiting: Map successfully transferred Transferring networks.byaddr... ypxfr: Exiting: Map successfully transferred Transferring netid.byname... ypxfr: Exiting: Map successfully transferred Transferring hosts.byaddr... ypxfr: Exiting: Map successfully transferred Transferring protocols.bynumber... ypxfr: Exiting: Map successfully transferred Transferring ypservers... ypxfr: Exiting: Map successfully transferred Transferring hosts.byname... ypxfr: Exiting: Map successfully transferred coltrane has been setup as an YP slave server without any errors. Don't forget to update map ypservers on ellington. You should now have a directory called /var/yp/test-domain. Copies of the NIS master server's maps should be in this directory. You will need to make sure that these stay updated. The following /etc/crontab entries on your slave servers should do the job: 20 * * * * root /usr/libexec/ypxfr passwd.byname 21 * * * * root /usr/libexec/ypxfr passwd.byuid These two lines force the slave to sync its maps with the maps on the master server. Although this is not mandatory, because the master server tries to make sure any changes to its NIS maps are communicated to its slaves, the password information is so vital to systems that depend on the server, that it is a good idea to force the updates. This is more important on busy networks where map updates might not always complete. Now, run the command /etc/netstart on the slave server as well, which again starts the NIS server. NIS Clients An NIS client establishes what is called a binding to a particular NIS server using the ypbind daemon. ypbind checks the system's default domain (as set by the domainname command), and begins broadcasting RPC requests on the local network. These requests specify the name of the domain for which ypbind is attempting to establish a binding. If a server that has been configured to serve the requested domain receives one of the broadcasts, it will respond to ypbind, which will record the server's address. If there are several servers available (a master and several slaves, for example), ypbind will use the address of the first one to respond. From that point on, the client system will direct all of its NIS requests to that server. Ypbind will occasionally ping the server to make sure it is still up and running. If it fails to receive a reply to one of its pings within a reasonable amount of time, ypbind will mark the domain as unbound and begin broadcasting again in the hopes of locating another server. Setting up an NIS client Setting up a FreeBSD machine to be a NIS client is fairly straightforward. Edit the file /etc/rc.conf and add the following lines in order to set the NIS domainname and start ypbind upon network startup: nisdomainname="test-domain" nis_client_enable="YES" To import all possible password entries from the NIS server, add this line to your /etc/master.passwd file, using vipw: +::::::::: This line will afford anyone with a valid account in the NIS server's password maps an account. There are many ways to configure your NIS client by changing this line. See the netgroups part below for more information. For more detailed reading see O'Reilly's book on Managing NFS and NIS. To import all possible group entries from the NIS server, add this line to your /etc/group file: +:*:: After completing these steps, you should be able to run ypcat passwd and see the NIS server's passwd map. NIS Security In general, any remote user can issue an RPC to ypserv and retrieve the contents of your NIS maps, provided the remote user knows your domainname. To prevent such unauthorized transactions, ypserv supports a feature called securenets which can be used to restrict access to a given set of hosts. At startup, ypserv will attempt to load the securenets information from a file called /var/yp/securenets. This path varies depending on the path specified with the option. This file contains entries that consist of a network specification and a network mask separated by white space. Lines starting with # are considered to be comments. A sample securenets file might look like this: # allow connections from local host -- mandatory 127.0.0.1 255.255.255.255 # allow connections from any host # on the 192.168.128.0 network 192.168.128.0 255.255.255.0 # allow connections from any host # between 10.0.0.0 to 10.0.15.255 # this includes the machines in the testlab 10.0.0.0 255.255.240.0 If ypserv receives a request from an address that matches one of these rules, it will process the request normally. If the address fails to match a rule, the request will be ignored and a warning message will be logged. If the /var/yp/securenets file does not exist, ypserv will allow connections from any host. The ypserv program also has support for Wietse Venema's tcpwrapper package. This allows the administrator to use the tcpwrapper configuration files for access control instead of /var/yp/securenets. While both of these access control mechanisms provide some security, they, like the privileged port test, are vulnerable to IP spoofing attacks. All NIS-related traffic should be blocked at your firewall. Servers using /var/yp/securenets may fail to serve legitimate NIS clients with archaic TCP/IP implementations. Some of these implementations set all host bits to zero when doing broadcasts and/or fail to observe the subnet mask when calculating the broadcast address. While some of these problems can be fixed by changing the client configuration, other problems may force the retirement of the client systems in question or the abandonment of /var/yp/securenets. Using /var/yp/securenets on a server with such an archaic implementation of TCP/IP is a really bad idea and will lead to loss of NIS functionality for large parts of your network. The use of the tcpwrapper package increases the latency of your NIS server. The additional delay may be long enough to cause timeouts in client programs, especially in busy networks or with slow NIS servers. If one or more of your client systems suffers from these symptoms, you should convert the client systems in question into NIS slave servers and force them to bind to themselves. Barring some users from logging on In our lab, there is a machine basie that is supposed to be a faculty only workstation. We don't want to take this machine out of the NIS domain, yet the passwd file on the master NIS server contains accounts for both faculty and students. What can we do? There is a way to bar specific users from logging on to a machine, even if they are present in the NIS database. To do this, all you must do is add -username to the end of the /etc/master.passwd file on the client machine, where username is the username of the user you wish to bar from logging in. This should preferably be done using vipw, since vipw will sanity check your changes to /etc/master.passwd, as well as automatically rebuild the password database when you finish editing. For example, if we wanted to bar user bill from logging on to basie we would: basie&prompt.root; vipw [add -bill to the end, exit] vipw: rebuilding the database... vipw: done basie&prompt.root; cat /etc/master.passwd root:[password]:0:0::0:0:The super-user:/root:/bin/csh toor:[password]:0:0::0:0:The other super-user:/root:/bin/sh daemon:*:1:1::0:0:Owner of many system processes:/root:/sbin/nologin operator:*:2:5::0:0:System &:/:/sbin/nologin bin:*:3:7::0:0:Binaries Commands and Source,,,:/:/sbin/nologin tty:*:4:65533::0:0:Tty Sandbox:/:/sbin/nologin kmem:*:5:65533::0:0:KMem Sandbox:/:/sbin/nologin games:*:7:13::0:0:Games pseudo-user:/usr/games:/sbin/nologin news:*:8:8::0:0:News Subsystem:/:/sbin/nologin man:*:9:9::0:0:Mister Man Pages:/usr/share/man:/sbin/nologin bind:*:53:53::0:0:Bind Sandbox:/:/sbin/nologin uucp:*:66:66::0:0:UUCP pseudo-user:/var/spool/uucppublic:/usr/libexec/uucp/uucico xten:*:67:67::0:0:X-10 daemon:/usr/local/xten:/sbin/nologin pop:*:68:6::0:0:Post Office Owner:/nonexistent:/sbin/nologin nobody:*:65534:65534::0:0:Unprivileged user:/nonexistent:/sbin/nologin +::::::::: -bill basie&prompt.root; Using netgroups The netgroups part was contributed by Udo Erdelhoff ue@nathan.ruhr.de in July 2000. The method shown in the previous chapter works reasonably well if you need special rules for a very small number of users and/or machines. On larger networks, you will forget to bar some users from logging onto sensitive machines, or you may even have to modify each machine separately, thus losing the main benefit of NIS, centralized administration. The NIS developers' solution for this problem is called netgroups. Their purpose and semantics can be compared to the normal groups used by Unix file systems. The main differences are the lack of a numeric id and the ability to define a netgroup by including both user accounts and other netgroups. Netgroups were developed to handle large, complex networks with hundreds of users and machines. On one hand, this is a Good Thing if you are forced to deal with such a situation. On the other hand, this complexity makes it almost impossible to explain netgroups with really simple examples. The example used in the remainder of this chapter demonstrates this problem. Let us assume that your successful introduction of NIS in your laboratory caught your superiors' interest. Your next job is to extend your NIS domain to cover some of the other machines on campus. The two tables contain the names of the new users and new machines as well as brief descriptions of them. User Name(s) Description alpha, beta Normal employees of the IT department charlie, delta The new apprentices of the IT department echo, foxtrott, golf, ... Ordinary employees able, baker, ... The current interns Machine Name(s) Description war, death, famine, pollution Your most important servers. Only the IT employees are allowed to log onto these machines. pride, greed, envy, wrath, lust, sloth Less important servers. All members of the IT department are allowed to login onto these machines. one, two, three, four, ... Ordinary workstations. Only the real employees are allowed to use these machines. trashcan A very old machine without any critical data. Even the intern is allowed to use this box. If you tried to implement these restrictions by separately blocking each user, you would have to add one -user line to each system's passwd for each user who is not allowed to login onto that system. If you forget just one entry, you could be in trouble. It may be feasible to do this correctly during the initial setup, however you will eventually forget to add the lines for new users during day-to-day operations. After all, Murphy was an optimist. Handling this situation with netgroups offers several advantages. Each user need not be handled separately; you assign a user to one or more netgroups and allow or forbid logins for all members of the netgroup. If you add a new machine, you will only have to define login restrictions for netgroups. If a new user is added, you will only have to add the user to one or more netgroups. Those changes are independent of each other; no more for each combination of user and machine do... If your NIS setup is planned carefully, you will only have to modify exactly one central configuration file to grant or deny access to machines. The first step is the initialization of the NIS map netgroup. FreeBSD's ypinit does not create this map by default, but its NIS implementation will support it once it has been created. To create an empty map, simply type ellington&prompt.root; vi /var/yp/netgroup and start adding content. For our example, we need at least four netgroups: IT employees, IT apprentices, normal employees and interns. IT_EMP (,alpha,test-domain) (,beta,test-domain) IT_APP (,charlie,test-domain) (,delta,test-domain) USERS (,echo,test-domain) (,foxtrott,test-domain) \ (,golf,test-domain) INTERNS (,able,test-domain) (,baker,test-domain) IT_EMP, IT_APP etc. are the names of the netgroups. Each bracketed group adds one or more user accounts to it. The three fields inside a group are: The name of the host(s) where the following items are valid. If you do not specify a hostname, the entry is valid on all hosts. If you do specify a hostname, you will enter a realm of darkness, horror and utter confusion. The name of the account that belongs to this netgroup. The NIS domain for the account. You can import accounts from other NIS domains into your netgroup if you are one of unlucky fellows with more than one NIS domain. Each of these fields can contain wildcards. See &man.netgroup.5; for details. Netgroup names longer than 8 characters should not be used, especially if you have machines running other operating systems within your NIS domain. The names are case sensitive; using capital letters for your netgroup names is an easy way to distinguish between user, machine and netgroup names. Some NIS clients (other than FreeBSD) cannot handle netgroups with a large number of entries. For example, some older versions of SunOS start to cause trouble if a netgroup contains more than 15 entries. You can circumvent this limit by creating several sub-netgroups with 15 users or less and a real netgroup that consists of the sub-netgroups: BIGGRP1 (,joe1,domain) (,joe2,domain) (,joe3,domain) [...] BIGGRP2 (,joe16,domain) (,joe17,domain) [...] BIGGRP3 (,joe31,domain) (,joe32,domain) BIGGROUP BIGGRP1 BIGGRP2 BIGGRP3 You can repeat this process if you need more than 225 users within a single netgroup. Activating and distributing your new NIS map is easy: ellington&prompt.root; cd /var/yp ellington&prompt.root; make This will generate the three NIS maps netgroup, netgroup.byhost and netgroup.byuser. Use &man.ypcat.1; to check if your new NIS maps are available: ellington&prompt.user; ypcat -k netgroup ellington&prompt.user; ypcat -k netgroup.byhost ellington&prompt.user; ypcat -k netgroup.byuser The output of the first command should resemble the contents of /var/yp/netgroup. The second command will not produce output if you have not specified host-specific netgroups. The third command can be used to get the list of netgroups for a user. The client setup is quite simple. To configure the server war, you only have to start &man.vipw.8; and replace the line +::::::::: with +@IT_EMP::::::::: Now, only the data for the users defined in the netgroup IT_EMP is imported into war's password database and only these users are allowed to login. Unfortunately, this limitation also applies to the ~ function of the shell and all routines converting between user names and numerical user ids. In other words, cd ~user will not work, ls -l will show the numerical id instead of the username and find . -user joe -print will fail with No such user. To fix this, you will have to import all user entries without allowing them to login onto your servers. This can be achieved by adding another line to /etc/master.passwd. This line should contain +:::::::::/sbin/nologin, meaning Import all entries but replace the shell with /sbin/nologin in the imported entries. You can replace any field in the passwd entry by placing a default value in your /etc/master.passwd. Make sure that the line +:::::::::/sbin/nologin is placed after +@IT_EMP:::::::::. Otherwise, all user accounts imported from NIS will have /sbin/nologin as their login shell. After this change, you will only have to change one NIS map if a new employee joins the IT department. You could use a similar approach for the less important servers by replacing the old +::::::::: in their local version of /etc/master.passwd with something like this: +@IT_EMP::::::::: +@IT_APP::::::::: +:::::::::/sbin/nologin The corresponding lines for the normal workstations could be: +@IT_EMP::::::::: +@USERS::::::::: +:::::::::/sbin/nologin And everything would be fine until there is a policy change a few weeks later: The IT department starts hiring interns. The IT interns are allowed to use the normal workstations and the less important servers; and the IT apprentices are allowed to login onto the main servers. You add a new netgroup IT_INTERN, add the new IT interns to this netgroup and start to change the config on each and every machine... As the old saying goes: Errors in centralized planning lead to global mess. NIS' ability to create netgroups from other netgroups can be used to prevent situations like these. One possibility is the creation of role-based netgroups. For example, you could create a netgroup called BIGSRV to define the login restrictions for the important servers, another netgroup called SMALLSRV for the less important servers and a third netgroup called USERBOX for the normal workstations. Each of these netgroups contains the netgroups that are allowed to login onto these machines. The new entries for your NIS map netgroup should look like this: BIGSRV IT_EMP IT_APP SMALLSRV IT_EMP IT_APP ITINTERN USERBOX IT_EMP ITINTERN USERS This method of defining login restrictions works reasonably well if you can define groups of machines with identical restrictions. Unfortunately, this is the exception and not the rule. Most of the time, you will need the ability to define login restrictions on a per-machine basis. Machine-specific netgroup definitions are the other possibility to deal with the policy change outlined above. In this scenario, the /etc/master.passwd of each box contains two lines starting with ``+''. The first of them adds a netgroup with the accounts allowed to login onto this machine, the second one adds all other accounts with /sbin/nologin as shell. It is a good idea to use the ALL-CAPS version of the machine name as the name of the netgroup. In other words, the lines should look like this: +@BOXNAME::::::::: +:::::::::/sbin/nologin Once you have completed this task for all your machines, you will not have to modify the local versions of /etc/master.passwd ever again. All further changes can be handled by modifying the NIS map. Here is an example of a possible netgroup map for this scenario with some additional goodies. # Define groups of users first IT_EMP (,alpha,test-domain) (,beta,test-domain) IT_APP (,charlie,test-domain) (,delta,test-domain) DEPT1 (,echo,test-domain) (,foxtrott,test-domain) DEPT2 (,golf,test-domain) (,hotel,test-domain) DEPT3 (,india,test-domain) (,juliet,test-domain) ITINTERN (,kilo,test-domain) (,lima,test-domain) D_INTERNS (,able,test-domain) (,baker,test-domain) # # Now, define some groups based on roles USERS DEPT1 DEPT2 DEPT3 BIGSRV IT_EMP IT_APP SMALLSRV IT_EMP IT_APP ITINTERN USERBOX IT_EMP ITINTERN USERS # # And a groups for a special tasks # Allow echo and golf to access our anti-virus-machine SECURITY IT_EMP (,echo,test-domain) (,golf,test-domain) # # machine-based netgroups # Our main servers WAR BIGSRV FAMINE BIGSRV # User india needs access to this server POLLUTION BIGSRV (,india,test-domain) # # This one is really important and needs more access restrictions DEATH IT_EMP # # The anti-virus-machine mentioned above ONE SECURITY # # Restrict a machine to a single user TWO (,hotel,test-domain) # [...more groups to follow] If you are using some kind of database to manage your user accounts, you should be able to create the first part of the map with your database's report tools. This way, new users will automatically have access to the boxes. One last word of caution: It may not always be advisable to use machine-based netgroups. If you are deploying a couple dozen or even hundreds of identical machines for student labs, you should use role-based netgroups instead of machine-based netgroups to keep the size of the NIS map within reasonable limits. Important things to remember There are still a couple of things that you will need to do differently now that you are in an NIS environment. Every time you wish to add a user to the lab, you must add it to the master NIS server only, and you must remember to rebuild the NIS maps. If you forget to do this, the new user will not be able to login anywhere except on the NIS master. For example, if we needed to add a new user “jsmith” to the lab, we would: &prompt.root; pw useradd jsmith &prompt.root; cd /var/yp &prompt.root; make test-domain You could also run adduser jsmith instead of pw useradd jsmith. Keep the administration accounts out of the NIS maps. You don't want to be propagating administrative accounts and passwords to machines that will have users that shouldn't have access to those accounts. Keep the NIS master and slave secure, and minimize their downtime. If somebody either hacks or simply turns off these machines, they have effectively rendered many people without the ability to login to the lab. This is the chief weakness of any centralized administration system, and it is probably the most important weakness. If you do not protect your NIS servers, you will have a lot of angry users! NIS v1 compatibility FreeBSD's ypserv has some support for serving NIS v1 clients. FreeBSD's NIS implementation only uses the NIS v2 protocol, however other implementations include support for the v1 protocol for backwards compatibility with older systems. The ypbind daemons supplied with these systems will try to establish a binding to an NIS v1 server even though they may never actually need it (and they may persist in broadcasting in search of one even after they receive a response from a v2 server). Note that while support for normal client calls is provided, this version of ypserv does not handle v1 map transfer requests; consequently, it can not be used as a master or slave in conjunction with older NIS servers that only support the v1 protocol. Fortunately, there probably are not any such servers still in use today. NIS servers that are also NIS clients Care must be taken when running ypserv in a multi-server domain where the server machines are also NIS clients. It is generally a good idea to force the servers to bind to themselves rather than allowing them to broadcast bind requests and possibly become bound to each other. Strange failure modes can result if one server goes down and others are dependent upon on it. Eventually all the clients will time out and attempt to bind to other servers, but the delay involved can be considerable and the failure mode is still present since the servers might bind to each other all over again. You can force a host to bind to a particular server by running ypbind with the flag. libscrypt v.s. libdescrypt One of the most common issues that people run into when trying to implement NIS is crypt library compatibility. If your NIS server is using the DES crypt libraries, it will only support clients that are using DES as well. To check which one your server and clients are using look at the symlinks in /usr/lib. If the machine is configured to use the DES libraries, it will look something like this: &prompt.user; ls -l /usr/lib/*crypt* lrwxrwxrwx 1 root wheel 13 Jul 15 08:55 /usr/lib/libcrypt.a@ -> libdescrypt.a lrwxrwxrwx 1 root wheel 14 Jul 15 08:55 /usr/lib/libcrypt.so@ -> libdescrypt.so lrwxrwxrwx 1 root wheel 16 Jul 15 08:55 /usr/lib/libcrypt.so.2@ -> libdescrypt.so.2 lrwxrwxrwx 1 root wheel 15 Jul 15 08:55 /usr/lib/libcrypt_p.a@ -> libdescrypt_p.a -r--r--r-- 1 root wheel 13018 Nov 8 14:27 /usr/lib/libdescrypt.a lrwxr-xr-x 1 root wheel 16 Nov 8 14:27 /usr/lib/libdescrypt.so@ -> libdescrypt.so.2 -r--r--r-- 1 root wheel 12965 Nov 8 14:27 /usr/lib/libdescrypt.so.2 -r--r--r-- 1 root wheel 14750 Nov 8 14:27 /usr/lib/libdescrypt_p.a If the machine is configured to use the standard FreeBSD MD5 crypt libraries they will look something like this: &prompt.user; ls -l /usr/lib/*crypt* lrwxrwxrwx 1 root wheel 13 Jul 15 08:55 /usr/lib/libcrypt.a@ -> libscrypt.a lrwxrwxrwx 1 root wheel 14 Jul 15 08:55 /usr/lib/libcrypt.so@ -> libscrypt.so lrwxrwxrwx 1 root wheel 16 Jul 15 08:55 /usr/lib/libcrypt.so.2@ -> libscrypt.so.2 lrwxrwxrwx 1 root wheel 15 Jul 15 08:55 /usr/lib/libcrypt_p.a@ -> libscrypt_p.a -r--r--r-- 1 root wheel 6194 Nov 8 14:27 /usr/lib/libscrypt.a lrwxr-xr-x 1 root wheel 14 Nov 8 14:27 /usr/lib/libscrypt.so@ -> libscrypt.so.2 -r--r--r-- 1 root wheel 7579 Nov 8 14:27 /usr/lib/libscrypt.so.2 -r--r--r-- 1 root wheel 6684 Nov 8 14:27 /usr/lib/libscrypt_p.a If you have trouble authenticating on an NIS client, this is a pretty good place to start looking for possible problems. If you want to deploy an NIS server for a heterogenous network, you will probably have to use DES on all systems because it is the lowest common standard. DHCP Written by &a.gsutter;, March 2000. What is DHCP? DHCP, the Dynamic Host Configuration Protocol, describes the means by which a system can connect to a network and obtain the necessary information for communication upon that network. FreeBSD uses the ISC (Internet Software Consortium) DHCP implementation, so all implementation-specific information here is for use with the ISC distribution. What This Section Covers This handbook section attempts to describe only the parts of the DHCP system that are integrated with FreeBSD; consequently, the server portions are not described. The DHCP manual pages, in addition to the references below, are useful resources. How it Works When dhclient, the DHCP client, is executed on the client machine, it begins broadcasting requests for configuration information. By default, these requests are on UDP port 68. The server replies on UDP 67, giving the client an IP address and other relevant network information such as netmask, router, and DNS servers. All of this information comes in the form of a DHCP "lease" and is only valid for a certain time (configured by the DHCP server maintainer). In this manner, stale IP addresses for clients no longer connected to the network can be automatically reclaimed. DHCP clients can obtain a great deal of information from the server. An exhaustive list may be found in &man.dhcp-options.5;. FreeBSD Integration FreeBSD fully integrates the ISC DHCP client, dhclient. DHCP client support is provided within both the installer and the base system, obviating the need for detailed knowledge of network configurations on any network that runs a DHCP server. dhclient has been included in all FreeBSD distributions since 3.2. DHCP is supported by sysinstall. When configuring a network interface within sysinstall, the first question asked is, "Do you want to try dhcp configuration of this interface?" Answering affirmatively will execute dhclient, and if successful, will fill in the network configuration information automatically. There are two things you must do to have your system use DHCP upon startup: Make sure that the bpf device is compiled into your kernel. To do this, add pseudo-device bpf to your kernel configuration file, and rebuild the kernel. For more information about building kernels, see . The bpf device is already part of the GENERIC kernel that is supplied with FreeBSD, so if you don't have a custom kernel, you shouldn't need to create one in order to get DHCP working. For those who are particularly security conscious, you should be warned that bpf is also the device that allows packet sniffers to work correctly (although they still have to be run as root). bpf is required to use DHCP, but if you are very sensitive about security, you probably shouldn't add bpf to your kernel in the expectation that at some point in the future you will be using DHCP. Edit your /etc/rc.conf to include the following: ifconfig_fxp0="DHCP" Be sure to replace fxp0 with the designation for the interface that you wish to dynamically configure. If you are using a different location for dhclient, or if you wish to pass additional flags to dhclient, also include the following (editing as necessary): dhcp_program="/sbin/dhclient" dhcp_flags="" The DHCP server, dhcpd, is included as part of the isc-dhcp2 port in the ports collection. This port contains the full ISC DHCP distribution, consisting of client, server, relay agent and documentation. Files /etc/dhclient.conf dhclient requires a configuration file, /etc/dhclient.conf. Typically the file contains only comments, the defaults being reasonably sane. This configuration file is described by the &man.dhclient.conf.5; man page. /sbin/dhclient dhclient is statically linked and resides in /sbin. The &man.dhclient.8; manual page gives more information about dhclient. /sbin/dhclient-script dhclient-script is the FreeBSD-specific DHCP client configuration script. It is described in &man.dhclient-script.8;, but should not need any user modification to function properly. /var/db/dhclient.leases The DHCP client keeps a database of valid leases in this file, which is written as a log. &man.dhclient.leases.5; gives a slightly longer description. Further Reading The DHCP protocol is fully described in RFC 2131. An informational resource has also been set up at dhcp.org. DNS Contributed by &a.chern;, April 12, 2001. Overview FreeBSD utilizes, by default, a version of BIND (Berkeley Internet Name Domain), which is the most common implementation of the DNS protocol. DNS is the protocol through which names are mapped to IPs, and vice versa. For example, a query for www.freebsd.org will send back a reply for the IP address of The FreeBSD Project's webpage, whereas, a query for ftp.freebsd.org will return the IP of the corresponding ftp machine. Likewise, the opposite can happen. A query for an IP address can resolve its hostname. DNS is coordinated across the Internet through a somewhat complex system of authoritative root name servers, and other smaller-scale nameservers who host and relay individual domain information. This document refers to BIND 8.x, as it is the most current, stable version used in FreeBSD. RFC1034 and RFC1035 dictates the DNS protocol. Currently, BIND is maintained by the Internet Software Consortium (www.isc.org) Terminology zone - Each individual domain, subdomain, or 'area' dictated by DNS is considered a zone. Examples of zones: . is the root zone org. is a zone under the root zone foobardomain.org is a zone under the org. zone foo.foobardomain.org. is a subdomain, a zone under the foobardomain.org. zone 1.2.3.in-addr.arpa is a zone referencing all ips which fall under the 3.2.1.* ip space. named, bind, name server - these are all common names for the BIND name server package within FreeBSD. resolver - a network process by which a system queries a nameserver for answers root zone - literally, a '.', refers to the root, or beginning zone. All zones fall under this, as do all files in fall under the root directory. It is the beginning of the Internet zone hierarchy origin - refers to the point of start for the particular zone forward dns - mapping of hostnames to ip addresses reverse dns - the opposite, mapping of ip addresses to hostnames Reasons to run a name server You need your machine to host DNS information to the world An authoritative nameserver replies exclusively to requests. For example, you register foobardomain.org and wish to assign hostnames to the proper IP addresses. A slave nameserver, which replies to queries for a domain when the primary is down or inaccessible. The above two can also be done with in-addr.arpa, IP to hostname entries You wish your machine to act as a local relay of DNS information DNS traffic has been measured to be about 5% or more of the total Internet traffic. A local DNS server may have some added benefit by providing a local cache of DNS information. For example, when one queries for www.freebsd.org, their resolver goes out to (usually) your ISP's name server, and retrieves the query. With a local, caching DNS server, the query only has to be made once to the outside world. Every additional query will not have to go outside of the local network, since the information is cached. How it works A DNS server in FreeBSD relies on the BIND daemon. This daemon is called 'named' for obvious reasons. named - the bind daemon ndc - name daemon control program /etc/namedb - directory where all the bind information resides /etc/namedb/named.conf - daemon configuration file zone files are usually contained within the /etc/namedb directory, and contain the information (query answers from your site) served by your name server. Starting BIND Since bind is installed by default, configuring it all is relatively simple. To ensure the named daemon is started at boot, put the following modifications in your /etc/rc.conf named_enable="YES" To start the daemon manually (after configuring it) &prompt.root; ndc start Configuration files make-localhost Be sure to &prompt.root; cd /etc/namedb &prompt.root; sh make-localhost to properly create your local reverse dns zone file in /etc/namedb/localhost.rev. <filename>/etc/namedb/named.conf</filename> - // $FreeBSD: doc/en_US.ISO_8859-1/books/handbook/advanced-networking/chapter.sgml,v 1.41 2001/05/28 13:41:56 sheldonh Exp $ + // $FreeBSD: doc/en_US.ISO8859-1/books/handbook/advanced-networking/chapter.sgml,v 1.42 2001/06/05 01:37:15 dd Exp $ // // Refer to the named(8) man page for details. If you are ever going // to setup a primary server, make sure you've understood the hairy // details of how DNS is working. Even with simple mistakes, you can // break connectivity for affected parties, or cause huge amount of // useless Internet traffic. options { directory "/etc/namedb"; // In addition to the "forwarders" clause, you can force your name // server to never initiate queries of its own, but always ask its // forwarders only, by enabling the following line: // // forward only; // If you've got a DNS server around at your upstream provider, enter // its IP address here, and enable the line below. This will make you // benefit from its cache, thus reduce overall DNS traffic in the Internet. /* forwarders { 127.0.0.1; }; */ Just as the comment says, if you want to benefit from your uplink's cache, you can enable this section of the config file. Normally, your nameserver will recursively query different nameservers until it finds the answer it is looking for. Having this enabled will have it automatically see if your uplink's (or whatever provided) ns has the requested query. If your uplink has a heavily trafficked, fast nameserver, enabling this properly could work to your advantage. 127.0.0.1 will *NOT* work here; change this to the IP of a nameserver at your uplink. /* * If there is a firewall between you and nameservers you want * to talk to, you might need to uncomment the query-source * directive below. Previous versions of BIND always asked * questions using port 53, but BIND 8.1 uses an unprivileged * port by default. */ // query-source address * port 53; /* * If running in a sandbox, you may have to specify a different * location for the dumpfile. */ // dump-file "s/named_dump.db"; }; // Note: the following will be supported in a future release. /* host { any; } { topology { 127.0.0.0/8; }; }; */ // Setting up secondaries is way easier and the rough picture for this // is explained below. // // If you enable a local name server, don't forget to enter 127.0.0.1 // into your /etc/resolv.conf so this server will be queried first. // Also, make sure to enable it in /etc/rc.conf. zone "." { type hint; file "named.root"; }; zone "0.0.127.IN-ADDR.ARPA" { type master; file "localhost.rev"; }; zone "0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.IP6.INT" { type master; file "localhost.rev"; }; // NB: Do not use the IP addresses below, they are faked, and only // serve demonstration/documentation purposes! // // Example secondary config entries. It can be convenient to become // a secondary at least for the zone where your own domain is in. Ask // your network administrator for the IP address of the responsible // primary. // // Never forget to include the reverse lookup (IN-ADDR.ARPA) zone! // (This is the first bytes of the respective IP address, in reverse // order, with ".IN-ADDR.ARPA" appended.) // // Before starting to setup a primary zone, better make sure you fully // understand how DNS and BIND works, however. There are sometimes // unobvious pitfalls. Setting up a secondary is comparably simpler. // // NB: Don't blindly enable the examples below. :-) Use actual names // and addresses instead. // // NOTE!!! FreeBSD runs bind in a sandbox (see named_flags in rc.conf). // The directory containing the secondary zones must be write accessible // to bind. The following sequence is suggested: // // mkdir /etc/namedb/s // chown bind:bind /etc/namedb/s // chmod 750 /etc/namedb/s /* zone "domain.com" { type slave; file "s/domain.com.bak"; masters { 192.168.1.1; }; }; zone "0.168.192.in-addr.arpa" { type slave; file "s/0.168.192.in-addr.arpa.bak"; masters { 192.168.1.1; }; }; */ These are example slave entries, read below to see more. For each new domain added to your nameserver, you must add one of these entries to your named.conf The simplest zone entry, can look like zone "foobardomain.org" { type master; file "foorbardomain.org"; }; For a master entry with the zone information within foobardomain.org, or zone "foobardomain.org" { type slave; file "foobardomain.org"; }; for a slave. Note that slave zones automatically query the listed master (authoritative) name servers for the zone file. Zone files An example master 'foobardomain.org' (existing within /etc/namedb/foobardomain.org) is as follows: $TTL 3600 foobardomain.org. IN SOA ns1.foobardomain.org. admin.foobardomain.org. ( 5 ; Serial 10800 ; Refresh 3600 ; Retry 604800 ; Expire 86400 ) ; Minimum TTL ; DNS Servers @ IN NS ns1.foobardomain.org. @ IN NS ns2.foobardomain.org. ; Machine Names localhost IN A 127.0.0.1 ns1 IN A 3.2.1.2 ns2 IN A 3.2.1.3 mail IN A 3.2.1.10 @ IN A 3.2.1.30 ; Aliases www IN CNAME @ ; MX Record @ IN MX 10 mail.foobardomain.org. Note that every hostname ending in a '.' is an exact hostname, whereas everything without a trailing '.' is referenced to the origin. For example, www is translated into www + origin. In our fictitious zone file, our origin is foobardomain.org, so www would be www.foobardomain.org. The format of this file follows: recordname IN recordtype value The most commonly used DNS records: SOA - start of zone authority NS - an authoritative nameserver A - A host address CNAME - the canonical name for an alias MX - mail exchange PTR - a domain name pointer (used in reverse dns) foobardomain.org. IN SOA ns1.foobardomain.org. admin.foobardomain.org. ( 5 ; Serial 10800 ; Refresh after 3 hours 3600 ; Retry after 1 hour 604800 ; Expire after 1 week 86400 ) ; Minimum TTL of 1 day foobardomain.org. - the domain name, also the origin for this zone file. ns1.foobardomain.org. - the primary/authoritative nameserver for this zone admin.foobardomain.org. - the responsible person for this zone, e-mail address with @ replaced. (admin@foobardomain.org becomes admin.foobardomain.org) 5 - the serial number of the file. this must be incremented each time the zone file is modified. Nowadays, many admins prefer a yyyymmddrr format for the serial number. 2001041002 would mean last modified 04/10/2001, the latter 02 being the second time the zone file has been modified this day. The serial number is important as it alerts slave nameservers for a zone when it is updated. @ IN NS ns1.foobardomain.org. This is an NS entry. Every nameserver that is going to reply authoritatively for the zone must have one of these entries. The @ as seen here could have been 'foobardomain.org.' The @ transalates to the origin. localhost IN A 127.0.0.1 ns1 IN A 3.2.1.2 ns2 IN A 3.2.1.3 mail IN A 3.2.1.10 @ IN A 3.2.1.30 The A record indicates machine names. As seen above, ns1.foobardomain.org would resolve to 3.2.1.2. Again, the origin symbol, @, is used here, thus meaning foobardomain.org would resolve to 3.2.1.30. www IN CNAME @ The canonical name record is usually used for giving aliases to a machine. In the example, www is aliased to the machine addressed to the origin, or foobardomain.org (3.2.1.30). CNAMEs can be used to provide alias hostnames, or round robin one hostname among multiple machines. @ IN MX 10 mail.foobardomain.org. The MX record indictes which mail servers are responsible for handling incoming mail for the zone. mail.foobardomain.org is the hostname of the mail server, and 10 being the priority of that mailserver. One can have several mailservers, with priorities of 3, 2, 1. A mail server attempting to deliver to foobardomain.org would first try the highest priority MX, then the second highest, etc, until the mail can be properly delivered. For in-addr.arpa zone files (reverse dns), the same format is used, except with PTR entries instead of A or CNAME. $TTL 3600 1.2.3.in-addr.arpa. IN SOA ns1.foobardomain.org. admin.foobardomain.org. ( 5 ; Serial 10800 ; Refresh 3600 ; Retry 604800 ; Expire 3600 ) ; Minimum @ IN NS ns1.foobardomain.org. @ IN NS ns2.foobardomain.org. 2 IN PTR ns1.foobardomain.org. 3 IN PTR ns2.foobardomain.org. 10 IN PTR mail.foobardomain.org. 30 IN PTR foobardomain.org. This file gives the proper IP to hostname mappings of our above ficticious domain. Caching Name Server A caching nameserver is simply a nameserver that is not authoritative for any zones. It simply asks queries of its own, and remembers them for later use. To set one up, just configure the name server as usual, omitting any inclusions of zones. Running named in a Sandbox Contributed by Mike Makonnen mike_makonnen@yahoo.com, May 1, 2001 For added security you may want to run &man.named.8; in a sandox. This will reduce the potential damage should it be compromised. If you include a sandbox directory in its command line, named will &man.chroot.8; into that directory immediately upon finishing processing its command line. It is also a good idea to have named run as a non-priveleged user in the sandbox. The default FreeBSD install contains a user bind with group bind. If we wanted the sandbox in the /etc/namedb/sanbox directory the command line for named would look like this: &prompt.root; /usr/sbin/named -u bind -g bind -t /etc/namedb/sandbox <path_to_named.conf> The following steps should be taken in order to successfully run named in a sandbox. Throughout the following discussion we will assume the path to your sandbox is /etc/namedb/sandox Create the sandbox directory: /etc/namedb/sandbox - Create other necessary directories off of the the sandbox + Create other necessary directories off of the sandbox directory: etc and var/run copy /etc/localtime to sandbox/etc make bind:bind the owner of all files and directories in the sandbox: &prompt.root; chown -R bind:bind /etc/namedb/sandbox &prompt.root; chmod -R 750 /etc/namedb/sandbox There are some issues you need to be aware of when running named in a sandbox. Your &man.named.conf.5; file and all your zone files must be in the sandbox sandbox/etc/localtime is needed in order to have the correct time for your time zone in log messages. &man.named.8; will write its process id to a file in sandbox/var/run The unix socket used for comunication by the &man.ndc.8; utility will be created in sandbox/var/run When using the ndc utility you need to specify the location of the unix socket created in the sandbox, by &man.named.8;, by using the -c switch: &prompt.root; ndc -c /etc/namedb/sandbox/var/run/ndc If you enable logging to file, the log files must be in the sandbox &man.named.8; can be started in a sandbox properly, if the following is in /etc/rc.conf named_flags="-u bind -g bind -t /etc/namedb/sandbox" How to use the nameserver If setup properly, the nameserver should be accessible through the network and locally. /etc/resolv.conf must contain a nameserver entry with the local ip so it will query the local name server first. To access it over the network, the machine must have the nameserver's IP address set properly in its own nameserver configuration options. Security Although BIND is the most common implementation of DNS, there is always the issue of security. Possible and exploitable security holes are sometimes found. It is a good idea to subscribe to CERT and freebsd-announce to stay up to date with the current Internet and FreeBSD security issues. If a problem arises, keeping your sources up to date and having a fresh build of named can't hurt. Further Reading &man.ndc.8; &man.named.8; &man.named.conf.5; Official ISC BIND Page http://www.isc.org/products/BIND/ BIND FAQ http://www.nominum.com/resources/faqs/bind-faqs.html O'Reilly DNS and BIND 4th Edition RFC1034 - Domain Names - Concepts and Facilities RFC1035 - Domain Names - Implementation and Specification diff --git a/en_US.ISO8859-1/books/handbook/x11/chapter.sgml b/en_US.ISO8859-1/books/handbook/x11/chapter.sgml index 8a55b857a0..4caa600c81 100644 --- a/en_US.ISO8859-1/books/handbook/x11/chapter.sgml +++ b/en_US.ISO8859-1/books/handbook/x11/chapter.sgml @@ -1,1615 +1,1615 @@ The X Window System This chapter has been graciously donated by &a.grog; from his book, The Complete FreeBSD, and remains copyright of him. Modifications for the handbook made by &a.jim;. The section on fonts in XFree86 was contributed by &a.murray; and the section on XDM was contributed by &a.sethk;. Synopsis The following chapter will cover installing and configuring X11 on your system. For more information on X11 and to see whether your video card is supported, check the XFree86 web site. Overview FreeBSD comes with XFree86, a port of X11R6 that supports several versions of Intel-based UNIX. This chapter describes how to set up your XFree86 server. It is based on material supplied with the FreeBSD release, specifically the files README.FreeBSD and README.Config in the directory /usr/X11R6/lib/X11/doc. If you find any discrepancy, the material in those files will be more up-to-date than this description. In addition, the file /usr/X11R6/lib/X11/doc/RELNOTES contains OS-independent information about the current release. X uses a lot of memory. In order to run X, your system should have an absolute minimum of 8 MB of memory, but performance will be painful with so little memory. A more practical minimum is 16 MB, and you can improve performance by adding more memory. If you use X intensively, you will continue seeing performance improvement by increasing to as much as 128 MB of RAM. There is lots of useful information in the rest of this chapter, but maybe you are not interested in information right now. You just want to get your X server up and running. However, be warned: An incorrect installation can burn out your monitor or your video board. However, if you know you are in spec, and you have a standard Super VGA board and a good multi-frequency monitor, then you can probably get things up and running without reading this chapter. Installing XFree86 The easiest way to install XFree86 is with the sysinstall program, either when you are installing the system, or later by starting the program /stand/sysinstall. In the rest of this chapter, we will look at what makes up the distribution, and we will also take a look at manually installing X11. The XFree86 Distribution XFree86 is distributed as a bewildering number of archives. In the following section, we will take a look at what you should install. Do not worry too much, though; if you cannot decide what to pick and you have 200MB of disk space free, it's safe to unpack everything. At a minimum you need to unpack the archives in the following table and at least one server that matches your VGA board. You will need 10Mb for the minimum required run-time binaries only, and between 1.7 and 3 MB for the server. Below is a table of the required components. Archive Description Xbin.tgz All the executable X client applications and shared libraries. Xfnts.tgz The misc and 75 dpi fonts. Xlib.tgz Data files and libraries needed at runtime. The X Server In addition to the archives above, you need at least one server, which will take up about 3 MB of disk. The choice depends primarily on what kind of display board you have. The default server name is /usr/X11R6/bin/X, and it is a link to a specific server binary /usr/X11R6/bin/XF86_xxxx. You will find the server archives for the standard PC architecture in /cdrom/XF86336/Servers, and the servers for the Japanese PC98 architecture in /cdrom/XF86336/PC98-Servers if you have the CD set. Alternatively, they are available on our FTP site at ftp://ftp.FreeBSD.org/pub/FreeBSD/releases/i386/&rel.current;-RELEASE/XF86336/Servers/ or ftp://ftp.FreeBSD.org/pub/FreeBSD/releases/i386/&rel.current;-RELEASE/XF86336/PC98-Servers/ Available X servers for the standard PC architecture: Archive Description X8514.tgz 8-bit color for IBM 8514 and true compatibles. XAGX.tgz 8 and 16-bit color for AGX and XGA boards. XI128.tgz 8 and 16-bit color for I128 boards. XMa32.tgz 8 and 16-bit color for ATI Mach32 boards. XMa64.tgz 8, 16, and 32-bit color fot ATI Mach64 boards. XMa8.tgz 8-bit color for ATI Mach8 boards. XMono.tgz 1-bit monochrome for VGA, Super-VGA, Hercules, and others. XP9K.tgz 8, 16, and 32-bit color for Weitek P9000 boards (Diamond Viper). XS3.tgz 8, 16, and 32-bit color for S3 boards. XS3V.tgz 8 and 16-bit color for S3 ViRGE boards. XSVGA.tgz >=8-bit color for Super-VGA cards. XVG16.tgz 4-bit color for VGA and Super-VGA cards. XW32.tgz 8-bit color for ET4000/W32, /W32i, /W32p, and ET6000 cards. Available X servers for the Japanese PC98 architecture: Archive Description X9GAN.tgz 8-bit color for PC98 GA-98NB/WAP boards. X9GA9.tgz 8, 16, and 32-bit color for PC98 S3 GA-968 boards. X9480.tgz 8-bit color for PC98 PEGC X9NKV.tgz 8-bit color for PC98 NEC-CIRRUS/EPSON NKV/NKV2 boards. X9WBS.tgz 8-bit color for PC98 WAB-S boards. X9WEP.tgz 8-bit color for PC98 WAB-EP boards. X9WSN.tgz 8-bit color for PC98 WSN-A2F boards. X9EGC.tgz 4-bit color for PC98 EGC. X9TGU.tgz 8 and 16-bit color for PC98 Trident Cyber9320/9680 boards. X9NS3.tgz 8 and 16-bit color for PC98 NEC S3 boards. X9SPW.tgz 8 and 16-bit color for PC98 S3 PW/PCSKB boards. X9LPW.tgz 8 and 16-bit color for PC98 S3 PW/LB boards. Each of these servers includes a manual page which contains details of supported chipsets and server-specific configuration options. There are also a number of archives are provided for X programmers: Archive Description Xprog.tgz Config, lib*.a, and *.h files needed for compiling clients. Xctrb.tgz Contributed sources. Xlk98.tgz The link kit for building servers, Japanese PC98 version. Xlkit.tgz The link kit for building servers, normal PC architecture. Xsrc-1.tgz Part 1 of the complete sources. Xsrc-2.tgz Part 2 of the complete sources. Xsrc-3.tgz Part 3 of the complete sources. You will need Xprog.tgz if you intend to install ports of X software. XFree86 also includes a number of optional parts, such as documentation, and setup programs. Archive Description Xdoc.tgz READMEs Xjdoc.tgz READMEs in Japanese. Xps.tgz READMEs in PostScript. Xhtml.tgz READMEs in HTML. Xman.tgz Manual pages. Xcfg.tgz Customizable xinit and xdm runtime configuration files. Xset.tgz The X86Setup utility; a graphical version of the xf86config utility. Xjset.tgz The XF86Setup utility, Japanese version, for the normal PC architecture. XF86Setup is a graphical mode setup program for XFree86, and you may prefer it to the standard setup program xf86config. You do not need any special archives for xf86config; it is included in Xbin.tgz. The first time you install, you will need Xcfg.tgz to create your initial configuration files. Do not use it when upgrading; it overwrites your configuration files. There are also additional fonts that are available with XFree86: Archive Description Xf100.tgz 100 dpi fonts. Xfscl.tgz Speedo and Type1 fonts. Xfnon.tgz Japanese, Chinese, and other non-english fonts. Xfcyr.tgz Cyrillic fonts. Unlike the X servers described above, the archives for the following servers are all in the main directory. Archive Description Xfsrv.tgz The font server. Xnest.tgz A nested server running as a client window on another display. Xprt.tgz The print server. Xvfb.tgz The Virtual Framebuffer X server, which renders into memory or an mmapped file. Installing XFree86 Manually If you do not use sysinstall to install X, you need to perform a number of steps: Create the directories and unpack the required archives. Choose and install an X server. Set up the environment to be able to access X. Find a virtual terminal in which to run X. Configure X for your hardware. This sounds like a lot of work, but if you approach it methodically, it is not too bad. In the rest of this section, we will look at each step in turn. Unpacking the Archives You must unpack the archives as root, since a number of the executables are set-user-id (they run as root even when started by other users). If you unpack the server as an ordinary user, it may abort when you try to run it. You must also use a umask value of 022 (permissions rwxr-xr-x), because the X server requires special permissions. &prompt.user; su Password: &prompt.root; umask 022 If you do not have enough space in the /usr file system, create a directory on another partition and symlink it to /usr. For example, if you have a file system /home with adequate space, you could do: &prompt.root; cd /home &prompt.root; mkdir X11R6 &prompt.root; ln -s /home/X11R6 /usr/X11R6 Next, decide which archives you want to install. For a minimal installation, choose Xbin.tgz, Xfnts.tgz, Xlib.tgz, and Xcfg.tgz. If you have already configured X for your hardware, you can omit Xcfg.tgz. If you are using sh, unpack like this: &prompt.root; mkdir -p /usr/X11R6 &prompt.root; cd /usr/X11R6 &prompt.root; for i in bin fnts lib cfg; do &prompt.root; tar xzf X$i.tgz &prompt.root; done If you are using csh, enter: &prompt.root; mkdir -p /usr/X11R6 &prompt.root; cd /usr/X11R6 &prompt.root; foreach i (bin fnts lib cfg) ? tar xzf X$i.tgz ? end Installing the Server Choose a server archive corresponding to your VGA board. If the table in the section above does not give you enough information, check the server man pages, /usr/X11R6/man/man1/XF86_*, which list the VGA chipsets supported by each server. For example, if you have an ET4000 based board you will use the XF86_SVGA server. In this case you would enter: &prompt.root; cd /usr/X11R6 &prompt.root; tar xzf XSVGA.tgz [substitute your server name here] Setting up the environment Next, you may wish to create a symbolic link /usr/X11/bin/X that points to the server that matches your video board. In this example, it is the XF86_SVGA server: &prompt.root; cd /usr/X11R6/bin &prompt.root; rm X &prompt.root; ln -s XF86_SVGA X X needs this symbolic link in order to be able to work correctly, but you have the option of setting it when you run xf86config – see below. Next, check that the directory /usr/X11R6/bin is in the default path for sh in /etc/profile and for csh in /etc/csh.login, and add it if it is not. It is best to do this with an editor, but if you want to take a shortcut, you can enter: &prompt.root; echo 'PATH=$PATH:/usr/X11R6/bin' >>/etc/profile or: &prompt.root; echo 'set path = ($path /usr/X11R6/bin)' >>/etc/csh.login Alternatively, make sure everybody who uses X puts /usr/X11R6/bin in their shell's PATH variable. Next, invoke ldconfig to put the shared libraries in ld.so's cache: &prompt.root; ldconfig -m /usr/X11R6/lib You can omit invoking ldconfig if you plan to reboot before using X. You do not need to uncompress the font files, but if you do, you must run mkfontdir in the corresponding font directory, otherwise your server will abort with the message could not open default font `fixed'. Assigning a virtual terminal to X Next, make sure you have a spare virtual console which is running a getty. First check how many virtual consoles you have: &prompt.root; dmesg | grep virtual sc0: VGA color <16 virtual consoles, flags=0x0> Then check /etc/ttys to make sure there is at least one virtual terminal (ttyvxx device) which does not have a getty enabled. Look for the keyword off: &prompt.root; grep ttyv /etc/ttys ttyv0 "/usr/libexec/getty Pc" cons25 on secure ttyv1 "/usr/libexec/getty Pc" cons25 on secure ttyv2 "/usr/libexec/getty Pc" cons25 on secure ttyv3 "/usr/libexec/getty Pc" cons25 off secure In this case, /dev/ttyv3 is available, if your kernel has least 4 VTs. If not, either disable a getty in /etc/ttys by changing on to off, or build another kernel with more virtual terminals. Configuring X for Your Hardware After installing the X software, you will need to customize the file XF86Config, which tells the X server about your hardware and how you want to run it. In order to set up XF86Config, you will need the following hardware information: Your mouse type, the bit rate if it is a serial mouse, and the name of the device to which it is connected. This will typically be /dev/ttyd0 or /dev/ttyd1 for a serial mouse, /dev/psm0 for a PS/2 mouse, or /dev/mse0 for a bus mouse. The type of the video board and the amount of display memory. If it is a no-name board, establish what VGA chip set it uses. The parameters of your monitor; vertical and horizontal frequency. Identifying the hardware How do you decide what your hardware is? The manufacturer should tell you, but very often the information you get about your display board and monitor is pitiful; Super VGA board with 76 Hz refresh rate and 16,777,216 colors. This tells you the maximum pixel depth (24 bits – - the number of colors is 2(pixel depth)), but it doesn't tell you anything else about the display board. As we will see later, the real parameters you need to know are the maximum horizontal frequency, the dot clock range, the chipset and the amount of display memory. You could be unlucky trying to get some of this information, but you can get some with the SuperProbe program. It should always be able to tell you the chipset and the amount of memory on board. Occasionally SuperProbe can crash your system. Make sure you are not doing anything important when you run it. Running SuperProbe looks like this: &prompt.root; SuperProbe (warnings and acknowledgments omitted) First video: Super-VGA Chipset: Tseng ET4000 (Port Probed) Memory: 1024 Kbytes RAMDAC: Generic 8-bit pseudo-color DAC (with 6-bit wide lookup tables (or in 6-bit mode)) SuperProbe is very finicky about running at all, and you will often get messages like: SuperProbe: Cannot be run while an X server is running SuperProbe: If an X server is not running, unset $DISPLAY and try again SuperProbe: Cannot open video In other words, even if no X server is running, SuperProbe will not work if you have the environment variable DISPLAY set. How do you unset it? With Bourne-style shells, you enter: &prompt.root; unset DISPLAY In the C shell, you enter: &prompt.root; unsetenv DISPLAY Running <command>xf86config</command> The easy way to create your configuration file is with one of the utilities xf86config (note the lower case name) or XF86Setup. Both lead you through the configuration step by step. xf86config runs in character mode, while XF86Setup runs in a graphical mode. XF86Setup can have problems with unusual hardware, so I personally prefer xf86config. You can also use sysinstall, but this does not change much; sysinstall just starts xf86config for you, and it is easier to start it directly. In this section, we will use an example to illustrate configuration via xf86config. We are installing X for an ancient Diamond SpeedStar with 1 MB of display memory, a Logitech MouseMan mouse, and an ADI MicroScan 5AP monitor. The mouse is connected to the system via the first serial port, /dev/ttyd0. To run xf86config, type in the name. If /usr/X11R6/bin is included in your PATH environment variable, you just need to type xf86config. If it is not, you need to type out the full path to xf86config, like so: &prompt.root; /usr/X11R6/bin/xf86config This program will create a basic XF86Configfile, based on menu selections you make. The XF86Config file usually resides in /usr/X11R6/lib/X11 or /etc. A sample XF86Config file is supplied with XFree86; it is configured for a standard VGA card and monitor with 640x480 resolution. This program will ask for a pathname when it is ready to write the file. You can either take the sample XF86Config as a base and edit it for your configuration, or let this program produce a base XF86Config file for your configuration and fine-tune it. Refer to /usr/X11R6/lib/X11/doc/README.Config for a detailed overview of the configuration process. For accelerated servers (including accelerated drivers in the SVGA server), there are many chipset and card-specific options and settings. This program does not know about these. On some configurations some of these settings must be specified. Refer to the server man pages and chipset-specific READMEs. Before continuing with this program, make sure you know the chipset and amount of video memory on your video card. SuperProbe can help with this. It is also helpful if you know what server you want to run. Press enter to continue, or ctrl-c to abort. ENTER First specify a mouse protocol type. Choose one from the following list: 1. Microsoft compatible (2-button protocol) 2. Mouse Systems (3-button protocol) 3. Bus Mouse 4. PS/2 Mouse 5. Logitech Mouse (serial, old type, Logitech protocol) 6. Logitech MouseMan (Microsoft compatible) 7. MM Series 8. MM HitTablet 9. Microsoft IntelliMouse If you have a two-button mouse, it is most likely of type 1, and if you have a three-button mouse, it can probably support both protocol 1 and 2. There are two main varieties of the latter type; mice with a switch to select the protocol, and mice that default to 1 and require a button to be held at boot-time to select protocol 2. Some mice can be convinced to do 2 by sending a special sequence to the serial port (see the ClearDTR/ClearRTS options). Enter a protocol number: 6 Logitech MouseMan You have selected a Logitech MouseMan type mouse. You might want to enable ChordMiddle which could cause the third button to work. Please answer the following question with either 'y' or 'n'. Do you want to enable ChordMiddle? n You definitely want to enable the third button on your mouse, since many X clients use it. With a genuine Logitech mouse, however, you don't need to enable ChordMiddle in order to use the button. If you find that the third button does not work when you start X, you can enable ChordMiddle by editing the configuration file – it is much easier and less error-prone than re-running XF86Setup. Continuing through the setup: If your mouse has only two buttons, it is recommended that you enable Emulate3Buttons. Please answer the following question with either 'y' or 'n'. Do you want to enable Emulate3Buttons? n Now give the full device name that the mouse is connected to, for example /dev/tty00. Just pressing enter will use the default, /dev/mouse. Mouse device: /dev/ttyd1 Be very careful about this entry. You must specify the correct name for the device to which the mouse is connected. xf86config is not specific to FreeBSD, and the suggested example is just plain wrong for FreeBSD. Use the names /dev/ttyd0 through /dev/ttyd3 for serial mice, /dev/psm0 for PS/2 mice or /dev/mse0 for a bus mouse. Continuing, we see: Beginning with XFree86 3.1.2D, you can use the new X11R6.1 XKEYBOARD extension to manage the keyboard layout. If you answer 'n' to the following question, the server will use the old method, and you have to adjust your keyboard layout with xmodmap. Please answer the following question with either 'y' or 'n'. Do you want to use XKB? y The following dialogue will allow you to select from a list of already preconfigured keymaps. If you don't find a suitable keymap in the list, the program will try to combine a keymap from additional information you are asked then. Such a keymap is by default untested and may require manual tuning. Please report success or required changes for such a keymap to XFREE86@XFREE86.ORG for addition to the list of preconfigured keymaps in the future. Press enter to continue, or ctrl-c to abort. List of preconfigured keymaps: 1 Standard 101-key, US encoding 2 Microsoft Natural, US encoding 3 KeyTronic FlexPro, US encoding 4 Standard 101-key, US encoding with ISO9995-3 extensions 5 Standard 101-key, German encoding 6 Standard 101-key, French encoding 7 Standard 101-key, Thai encoding 8 Standard 101-key, Swiss/German encoding 9 Standard 101-key, Swiss/French encoding 10 None of the above Enter a number to choose the keymap. 1 Choose the standard US keyboard Now we want to set the specifications of the monitor. The two critical parameters are the vertical refresh rate, which - is the rate at which the the whole screen is refreshed, and + is the rate at which the whole screen is refreshed, and most importantly the horizontal sync rate, which is the rate at which scanlines are displayed. The valid range for horizontal sync and vertical sync should be documented in the manual of your monitor. If in doubt, check the monitor database /usr/X11R6/lib/X11/doc/Monitors to see if your monitor is there. Press enter to continue, or ctrl-c to abort. ENTER You must indicate the horizontal sync range of your monitor. You can either select one of the predefined ranges below that correspond to industry- standard monitor types, or give a specific range. It is VERY IMPORTANT that you do not specify a monitor type with a horizontal sync range that is beyond the capabilities of your monitor. If in doubt, choose a conservative setting. hsync in kHz; monitor type with characteristic modes 1 31.5; Standard VGA, 640x480 @@ 60 Hz 2 31.5 - 35.1; Super VGA, 800x600 @@ 56 Hz 3 31.5, 35.5; 8514 Compatible, 1024x768 @@ 87 Hz interlaced (no 800x600) 4 31.5, 35.15, 35.5; Super VGA, 1024x768 @@ 87 Hz interlaced, 800x600 @@ 56 Hz 5 31.5 - 37.9; Extended Super VGA, 800x600 @@ 60 Hz, 640x480 @@ 72 Hz 6 31.5 - 48.5; Non-Interlaced SVGA, 1024x768 @@ 60 Hz, 800x600 @@ 72 Hz 7 31.5 - 57.0; High Frequency SVGA, 1024x768 @@ 70 Hz 8 31.5 - 64.3; Monitor that can do 1280x1024 @@ 60 Hz 9 31.5 - 79.0; Monitor that can do 1280x1024 @@ 74 Hz 10 31.5 - 82.0; Monitor that can do 1280x1024 @@ 76 Hz 11 Enter your own horizontal sync range Enter your choice (1-11): Unfortunately, our monitor is not mentioned in the file /usr/X11R6/lib/X11/doc/Monitors, but by chance the manual does specify the frequency range in the Technical Data section. The horizontal frequency range is from 30 to 64 kHz, and the vertical frequency range is from 50 to 100 Hz. The horizontal frequency range is almost exactly covered by choice 8, but that setting threatens to go 0.3 kHz higher in frequency than the technical data state. Do you want to risk it? Doing so will most likely not be a problem, since it is unlikely that the monitor will die at such a small deviation from the specs, and it is also unlikely that your XF86Config will actually generate a horizontal frequency between 64.0 and 64.3 kHz. However, there is no need to take even this slight risk. Just specify the real values: Enter your choice (1-11): 11 Please enter the horizontal sync range of your monitor, in the format used in the table of monitor types above. You can either specify one or more continuous ranges (e.g. 15-25, 30-50), or one or more fixed sync frequencies. Horizontal sync range: 30-64 Next, we select the vertical frequency range: You must indicate the vertical sync range of your monitor. You can either select one of the predefined ranges below that correspond to industry-standard monitor types, or give a specific range. For interlaced modes, the number that counts is the high one (e.g., 87 Hz rather than 43 Hz). 1 50-70 2 50-90 3 50-100 4 40-150 5 Enter your own vertical sync range Enter your choice: 3 exactly the range of the monitor The next step is to specify identification strings. You can think out names if you want, but unless you are juggling a lot of different hardware, you can let xf86config do it for you: You must now enter a few identification/description strings, namely an identifier, a vendor name, and a model name. Just pressing enter will fill in default names. The strings are free-form, spaces are allowed. Enter an identifier for your monitor definition: ENTER Enter the vendor name of your monitor: ENTER Enter the model name of your monitor: ENTER Next comes the choice of the video board. We have an elderly Diamond SpeedStar Plus with an ET4000 chip, and unknown Ramdac and Clock Chip. Let's see how we fare: Now we must configure video card specific settings. At this point you can choose to make a selection out of a database of video card definitions. Because there can be variation in Ramdacs and clock generators even between cards of the same model, it is not sensible to blindly copy the settings (e.g., a Device section). For this reason, after you make a selection, you will still be asked about the components of the card, with the settings from the chosen database entry presented as a strong hint. The database entries include information about the chipset, what server to run, the Ramdac and ClockChip, and comments that will be included in the Device section. However, a lot of definitions only hint about what server to run (based on the chipset the card uses) and are untested. If you can't find your card in the database, there's nothing to worry about. You should only choose a database entry that is exactly the same model as your card; choosing one that looks similar is just a bad idea (e.g. a GemStone Snail 64 may be as different from a GemStone Snail 64+ in terms of hardware as can be). Do you want to look at the card database? y 0 2 the Max MAXColor S3 Trio64V+ S3 Trio64V+ 1 928Movie S3 928 2 AGX (generic) AGX-014/15/16 3 ALG-5434(E) CL-GD5434 4 ASUS 3Dexplorer RIVA128 5 ASUS PCI-AV264CT ATI-Mach64 6 ASUS PCI-V264CT ATI-Mach64 7 ASUS Video Magic PCI V864 S3 864 8 ASUS Video Magic PCI VT64 S3 Trio64 9 AT25 Alliance AT3D 10 AT3D Alliance AT3D 11 ATI 3D Pro Turbo ATI-Mach64 12 ATI 3D Xpression ATI-Mach64 13 ATI 3D Xpression+ PC2TV ATI-Mach64 14 ATI 8514 Ultra (no VGA) ATI-Mach8 15 ATI All-in-Wonder ATI-Mach64 16 ATI Graphics Pro Turbo ATI-Mach64 17 ATI Graphics Pro Turbo 1600 ATI-Mach64 Enter a number to choose the corresponding card definition. Press enter for the next page, q to continue configuration. ENTER Dozens of board definitions come in alphabetic order. Finally we see: 108 DSV3325 S3 ViRGE 109 DSV3326 S3 Trio64V+ 110 DataExpert DSV3325 S3 ViRGE 111 DataExpert DSV3365 S3 Trio64V+ 112 Dell S3 805 S3 801/805 113 Dell onboard ET4000 ET4000 114 Diamond Edge 3D nv1 115 Diamond Multimedia Stealth 3D 2000 S3 ViRGE 116 Diamond Multimedia Stealth 3D 2000 PRO S3 ViRGE/DX 117 Diamond SpeedStar (Plus) ET4000 118 Diamond SpeedStar 24 ET4000 119 Diamond SpeedStar 24X (not fully supported) WD90C31 120 Diamond SpeedStar 64 CL-GD5434 121 Diamond SpeedStar HiColor ET4000 122 Diamond SpeedStar Pro (not SE) CL-GD5426/28 123 Diamond SpeedStar Pro 1100 CL-GD5420/2/4/6/8/9 124 Diamond SpeedStar Pro SE (CL-GD5430/5434) CL-GD5430/5434 125 Diamond SpeedStar64 Graphics 2000/2200 CL-GD5434 Enter a number to choose the corresponding card definition. Press enter for the next page, q to continue configuration. 117 Your selected card definition: Identifier: Diamond SpeedStar (Plus) Chipset: ET4000 Server: XF86_SVGA Press enter to continue, or ctrl-c to abort.ENTER Now you must determine which server to run. Refer to the man pages and other documentation. The following servers are available (they may not all be installed on your system): 1 The XF86_Mono server. This a monochrome server that should work on any VGA-compatible card, in 640x480 (more on some SVGA chipsets). 2 The XF86_VGA16 server. This is a 16-color VGA server that should work on any VGA-compatible card. 3 The XF86_SVGA server. This is a 256 color SVGA server that supports a number of SVGA chipsets. On some chipsets it is accelerated or supports higher color depths. 4 The accelerated servers. These include XF86_S3, XF86_Mach32, XF86_Mach8, XF86_8514, XF86_P9000, XF86_AGX, XF86_W32, XF86_Mach64, XF86_I128 and XF86_S3V. These four server types correspond to the four different "Screen" sections in XF86Config (vga2, vga16, svga, accel). 5 Choose the server from the card definition, XF86_SVGA. Which one of these screen types do you intend to run by default (1-5)? The system already chose XF86_SVGA for us. Do we want to change? We would need a good reason. In this case, we do not have a reason, so we will keep the server from the card definition: Which one of these screen types do you intend to run by default (1-5)? 5 The server to run is selected by changing the symbolic link 'X'. For example, the SVGA server. Please answer the following question with either 'y' or 'n'. Do you want me to set the symbolic link? y All the programs that start X (xinit, startx, and xdm) start a program /usr/X11R6/bin/X. This symbolic link makes /usr/X11R6/bin/X point to your X server. If you don't have a link, you will not be able to start X. Now you must give information about your video card. This will be used for the "Device" section of your video card in XF86Config. You must indicate how much video memory you have. It is probably a good idea to use the same approximate amount as that detected by the server you intend to use. If you encounter problems that are due to the used server not supporting the amount memory you have (e.g. ATI Mach64 is limited to 1024K with the SVGA server), specify the maximum amount supported by the server. How much video memory do you have on your video card: 1 256K 2 512K 3 1024K 4 2048K 5 4096K 6 Other Enter your choice: 3 You must now enter a few identification/description strings, namely an identifier, a vendor name, and a model name. Just pressing enter will fill in default names (possibly from a card definition). Your card definition is Diamond SpeedStar (Plus). The strings are free-form, spaces are allowed. Enter an identifier for your video card definition: ENTER You can simply press enter here if you have a generic card, or want to describe your card with one string. Enter the vendor name of your video card: ENTER Enter the model (board) name of your video card: ENTER Especially for accelerated servers, Ramdac, Dacspeed and ClockChip settings or special options may be required in the Device section. The RAMDAC setting only applies to the S3, AGX, W32 servers, and some drivers in the SVGA servers. Some RAMDAC's are auto-detected by the server. The detection of a RAMDAC is forced by using a Ramdac "identifier" line in the Device section. The identifiers are shown at the right of the following table of RAMDAC types: 1 AT&T 20C490 (S3 and AGX servers, ARK driver) att20c490 2 AT&T 20C498/21C498/22C498 (S3, autodetected) att20c498 3 AT&T 20C409/20C499 (S3, autodetected) att20c409 4 AT&T 20C505 (S3) att20c505 5 BrookTree BT481 (AGX) bt481 6 BrookTree BT482 (AGX) bt482 7 BrookTree BT485/9485 (S3) bt485 8 Sierra SC15025 (S3, AGX) sc15025 9 S3 GenDAC (86C708) (autodetected) s3gendac 10 S3 SDAC (86C716) (autodetected) s3_sdac 11 STG-1700 (S3, autodetected) stg1700 12 STG-1703 (S3, autodetected) stg1703 Enter a number to choose the corresponding RAMDAC. Press enter for the next page, q to quit without selection of a RAMDAC. q We don't need this A Clockchip line in the Device section forces the detection of a programmable clock device. With a clockchip enabled, any required clock can be programmed without requiring probing of clocks or a Clocks line. Most cards don't have a programmable clock chip. Choose from the following list: 1 Chrontel 8391 ch8391 2 ICD2061A and compatibles (ICS9161A, DCS2824) icd2061a 3 ICS2595 ics2595 4 ICS5342 (similar to SDAC, but not completely compatible) ics5342 5 ICS5341 ics5341 6 S3 GenDAC (86C708) and ICS5300 (autodetected) s3gendac 7 S3 SDAC (86C716) s3_sdac 8 STG 1703 (autodetected) stg1703 9 Sierra SC11412 sc11412 10 TI 3025 (autodetected) ti3025 11 TI 3026 (autodetected) ti3026 12 IBM RGB 51x/52x (autodetected) ibm_rgb5xx Just press enter if you don't want a Clockchip setting. What Clockchip setting do you want (1-12)? ENTER For most configurations, a Clocks line is useful since it prevents the slow and nasty sounding clock probing at server start-up. Probed clocks are displayed at server startup, along with other server and hardware configuration info. You can save this information in a file by running imprecise; some clocks may be slightly too high (varies per run). At this point I can run X -probeonly, and try to extract the clock information from the output. It is recommended that you do this yourself and add a clocks line (note that the list of clocks may be split over multiple Clocks lines) to your Device section afterwards. Be aware that a clocks line is not appropriate for drivers that have a fixed set of clocks and don't probe by default (e.g. Cirrus). Also, for the P9000 server you must simply specify clocks line that matches the modes you want to use. For the S3 server with a programmable clock chip you need a 'ClockChip' line and no Clocks line. You must be root to be able to run X -probeonly now. Do you want me to run 'X -probeonly' now? This last question is worth thinking about. You should run X -probeonly at some point, but it requires some extra work. We'll take the recommendation and try it later. Do you want me to run 'X -probeonly' now? n For each depth, a list of modes (resolutions) is defined. The default resolution that the server will start-up with will be the first listed mode that can be supported by the monitor and card. Currently it is set to: "640x480" "800x600" "1024x768" for 8bpp "640x480" "800x600" for 16bpp "640x480" for 24bpp "640x400" for 32bpp Note that 16, 24 and 32bpp are only supported on a few configurations. Modes that cannot be supported due to monitor or clock constraints will be automatically skipped by the server. 1 Change the modes for 8pp (256 colors) 2 Change the modes for 16bpp (32K/64K colors) 3 Change the modes for 24bpp (24-bit color, packed pixel) 4 Change the modes for 32bpp (24-bit color) 5 The modes are OK, continue. Enter your choice: 5 accept the defaults You can have a virtual screen (desktop), which is screen area that is larger than the physical screen and which is panned by moving the mouse to the edge of the screen. If you don't want virtual desktop at a certain resolution, you cannot have modes listed that are larger. Each color depth can have a differently-sized virtual screen Please answer the following question with either 'y' or 'n'. Do you want a virtual screen that is larger than the physical screen? n It is difficult to decide whether you want a virtual screen larger than the physical screen. I find it extremely disturbing, so I suggest you answer n. You might find it useful, especially if your highest resolution is small. Now the configuration is complete, and sysinstall just need to write the configuration file: I am going to write the XF86Config file now. Make sure you don't accidently overwrite a previously configured one. Shall I write it to /etc/XF86Config? y File has been written. Take a look at it before running 'startx'. Note that the XF86Config file must be in one of the directories searched by the server (e.g. /usr/X11R6/lib/X11) in order to be used. Within the server press ctrl, alt and '+' simultaneously to cycle video resolutions. Pressing ctrl, alt and backspace simultaneously immediately exits the server (use if the monitor doesn't sync for a particular mode). For further configuration, refer to /usr/X11R6/lib/X11/doc/README.Config. Once you have completed this configuration, you are ready to start X. Using Fonts in XFree86 TrueType Fonts The default fonts that ship with XFree86 are less than ideal for typical desktop publishing applications. Large presentation fonts show up jagged and unprofessional looking and small fonts in Netscape are almost completely unintelligable. Fortunately, XFree86 can be configured to use TrueType fonts with a minimum of effort. XFree86 4.0 has built in support for rendering TrueType fonts. There are two different modules that can enable this functionality. The "freetype" module is used in this example because it is more consistent with the other font rendering backends. To enable the freetype module just add the following line to the module section of your /etc/X11/XF86Config file. Load "freetype" For XFree86 3.3.X you will need to run a seperate TrueType font server. Xfstt is commonly used for this purpose. To install Xfstt on your FreeBSD system simply install the port from /usr/ports/x11-servers/Xfstt You should now make a directory for your TrueType fonts (e.g. /usr/X11R6/lib/X11/fonts/TrueType) and copy all of your TrueType fonts into this directory. Keep in mind that you can not take TrueType fonts directly from a Macintosh; they must be in Unix/DOS/Windows format for use by XFree86. Once you have copied the files into this directory you need to use ttmkfdir to create a fonts.dir file so that the X font renderer knows that you've installed these new files. There is a FreeBSD port for ttmkfdir in /usr/ports/x11-fonts/ttmkfdir. &prompt.root; cd /usr/X11R6/lib/X11/fonts/TrueType &prompt.root; ttmkfdir > fonts.dir Now you need to add your TrueType directory to your fonts path. The easiest way to do this is to add the following entries into your ~/.xinitrc file. &prompt.user; xset fp+ /usr/X11R6/lib/X11/fonts/TrueType &prompt.user; xset fp rehash That's it. Now Netscape, Gimp, StarOffice, and all of your other X applications should now recognize your installed TrueType fonts. Extremely small fonts (as with text in a high resolution display on a web page) and extremely large fonts (within StarOffice) will look much better now. One Caveat : XFree86 does not currently support anti-aliased font rendering. This is less of an issue at higher screen resolutions but the output is still less than optimal when compared with MacOS or Microsoft Windows. The X Display Manager Overview The X Display Manager (XDM) is an optional part of the X Window System that is used for login session management. This is useful for several types of situations, including minimal X Terminals (see section ), desktops, and large network display servers. Since the X Window System is network and protocol independent, there are a wide variety of possible configurations for running X clients and servers on different machines connected by a network. XDM provides a graphical interface for choosing which display server to connect to, and entering authorization information such as a login and password combination. You may think of XDM as providing the same functionality to the user as the &man.getty.8; utility (see for details). That is, it performs system logins to the display being connected to and then runs a session manager on behalf of the user (usually an X window manager). XDM then waits for this program to exit, signalling that the user is done and should be logged out of the display. At this point, XDM can display the login and display chooser screens for the next user to login. Using XDM The XDM daemon program is located in /usr/X11R6/bin/xdm. You can run this program at any time as root and it will start managing the X display on the local machine. If you want XDM to run in the background every time the machine boots up, a convenient way to do this is by adding an entry to /etc/ttys. For more information about the format and usage of this file, see . There is a line in the default /etc/ttys file for running the xdm daemon on a virtual terminal: ttyv8 "/usr/X11R6/bin/xdm -nodaemon" xterm off secure By default this entry is disabled, and in order to enable it you will need to change field 5 from off to on and then restart &man.init.8; using the directions in . The first field, the name of the terminal this program will manage, is ttyv8. This means that XDM will start running on the 9th virtual terminal. Configuring XDM The XDM configuration directory is located in /usr/X11R6/lib/X11/xdm. In this directory you will see several files used to change the behavior and appearence of XDM. Typically you will find these files: File Description Xaccess Client authorization ruleset. Xresources Default X resource values. Xservers List of remote and local displays to manage. Xsession Default session script for logins. Xsetup_* xdm-config Global configuration for all displays running on this machine. xdm-errors Errors generated by the server program. xdm-pid The process ID of the currently running XDM. Also in this directory are a few scripts and programs used to setup the desktop when XDM is running. In the next few sections I will briefly describe the purpose of each of these files. The exact syntax and usage of all of these files is described in &man.xdm.1; The default configuration is a simple rectangular login window with the hostname of the machine displayed at the top in a large font and Login: and Password: prompts below. This is a good starting point if you are planning to design your own look and feel for the XDM screens. Xaccess The protocol for connecting to XDM controlled displays is called the X Display Manager Connection Protocol (XDMCP). This file is basically just a ruleset for controlling XDMCP connections from remote machines. By default, it allows any client to connect, but you will see this will not matter because the default xdm-config file does not listen for remote connections. Xresources This is an application-defaults file for the display chooser and the login screens. This is where you can customize the appearence of the login program. The format is identical to the app-defaults file described in the XFree86 documentation. Xservers This is a list of the remote displays the chooser should provide as choices. Xsession This is the default session script for XDM to run after a user has logged in. Normally each user will have a customized session script in ~/.xsessionrc that overrides this script. Xsetup_* These files contain scripts that will be run automatically before displaying the chooser or login interfaces. There is a script for each display being used, named Xsetup_followed by the local display number (for instance Xsetup_0). Typically these scripts will run one or two programs in the background such as xconsole. xdm-config This file contains settings in the form of app-defaults that are applicable to every display that this installation manages. xdm-errors This file contains the output of the X servers that XDM is trying to run. If a display that XDM is trying to start hangs for some reason, this is a good place to look for error messages. These messages are also written to the user's ~/.xsession-errors file on a per-session basis Running A Network Display Server In order for other clients to connect to your display server, you will need to edit the access control rules, and enable the connection listener. By default these are set to conservative values, which is a good descision security-wise. To get XDM to listen for connections first comment out a line in the xdm-config file: ! SECURITY: do not listen for XDMCP or Chooser requests ! Comment out this line if you want to manage X terminals with xdm DisplayManager.requestPort: 0 and then restart XDM. Remember that comments in app-defaults files begin with a ! character, not a #. After this, you may need to put more strict access controls in place. Look at the example entries in Xaccess file, and refer to the &man.xdm.1; manual page. diff --git a/en_US.ISO8859-1/books/porters-handbook/book.sgml b/en_US.ISO8859-1/books/porters-handbook/book.sgml index e694c885cc..bd67d8edb7 100644 --- a/en_US.ISO8859-1/books/porters-handbook/book.sgml +++ b/en_US.ISO8859-1/books/porters-handbook/book.sgml @@ -1,4482 +1,4482 @@ %man; %bookinfo; %authors; %mailing-lists; ]> FreeBSD Porter's Handbook The FreeBSD Documentation Project April 2000 2000 The FreeBSD Documentation Project &bookinfo.legalnotice; Making a port yourself So, now you are interested in making your own port or upgrading an existing one? Great! What follows are some guidelines for creating a new port for FreeBSD. If you want to upgrade an existing port, you should read this and then read . When this document is not sufficiently detailed, you should refer to /usr/ports/Mk/bsd.port.mk, which all port Makefiles include. Even if you do not hack Makefiles daily, it is well commented, and you will still gain much knowledge from it. Additionally, you may send specific questions to the &a.ports;. Only a fraction of the variables (VAR) that can be overridden are mentioned in this document. Most (if not all) are documented at the start of bsd.port.mk. This file uses a non-standard tab setting. Emacs and Vim should recognize the setting on loading the file. Both vi and ex can be set to use the correct value by typing :set tabstop=4 once the file has been loaded. Quick Porting This section tells you how to do a quick port. In many cases, it is not enough, but we will see. First, get the original tarball and put it into DISTDIR, which defaults to /usr/ports/distfiles. The following assumes that the software compiled out-of-the-box, i.e., there was absolutely no change required for the port to work on your FreeBSD box. If you needed to change something, you will have to refer to the next section too. Writing the <filename>Makefile</filename> The minimal Makefile would look something like this: # New ports collection makefile for: oneko # Date created: 5 December 1994 # Whom: asami # # $FreeBSD$ # PORTNAME= oneko PORTVERSION= 1.1b CATEGORIES= games MASTER_SITES= ftp://ftp.cs.columbia.edu/archives/X11R5/contrib/ MAINTAINER= asami@FreeBSD.org MAN1= oneko.1 MANCOMPRESSED= yes USE_IMAKE= yes .include <bsd.port.mk> See if you can figure it out. Do not worry about the contents of the $FreeBSD$ line, it will be filled in automatically by CVS when the port is imported to our main ports tree. You can find a more detailed example in the sample Makefile section. Writing the description files There are three description files that are required for any port, whether they actually package or not. They are pkg-comment, pkg-descr, and pkg-plist, and their pkg- prefix distinguishes them from other files. <filename>pkg-comment</filename> This is the one-line description of the port. Please do not include the package name (or version number of the software) in the comment. The comment should begin with a capital, and end without a period. Here is an example: A cat chasing a mouse all over the screen <filename>pkg-descr</filename> This is a longer description of the port. One to a few paragraphs concisely explaining what the port does is sufficient. This is not a manual or an in-depth description on how to use or compile the port! Please be careful if you are copying from the README or manpage; too often they are not a concise description of the port or are in an awkward format (e.g., manpages have justified spacing). If the ported software has an official WWW homepage, you should list it here. Prefix one of the websites with WWW: so that automated tools will work correctly. It is recommended that you sign your name at the end of this file, as in: This is a port of oneko, in which a cat chases a poor mouse all over the screen. : (etc.) WWW: http://www.oneko.org/ - Satoshi asami@cs.berkeley.edu <filename>pkg-plist</filename> This file lists all the files installed by the port. It is also called the “packing list” because the package is generated by packing the files listed here. The pathnames are relative to the installation prefix (usually /usr/local or /usr/X11R6). If you are using the MANn variables (as you should be), do not list any manpages here. Here is a small example: bin/oneko lib/X11/app-defaults/Oneko lib/X11/oneko/cat1.xpm lib/X11/oneko/cat2.xpm lib/X11/oneko/mouse.xpm @dirrm lib/X11/oneko Refer to the &man.pkg.create.1; man page for details on the packing list. You should list all the files, but not the name directories, in the list. Also, if the port creates directories for itself during installation, make sure to add @dirrm lines as necessary to remove them when the port is deleted. It is recommended that you keep all the filenames in this file sorted alphabetically. It will make verifying the changes when you upgrade the port much easier. Creating a packing list manually can be a very tedious task. If the port installs a large numbers of files, creating the packing list automatically might save time. Creating the checksum file Just type make makesum. The ports make rules will automatically generate the file distinfo. Testing the port You should make sure that the port rules do exactly what you want them to do, including packaging up the port. These are the important points you need to verify. pkg-plist does not contain anything not installed by your port pkg-plist contains everything that is installed by your port Your port can be installed multiple times using the reinstall target Your port cleans up after itself upon deinstall Recommended test ordering make install make package make deinstall pkg_add package-name make deinstall make reinstall make package Make sure that there are not any warnings issued in any of the package and deinstall stages. After step 3, check to see if all the new directories are correctly deleted. Also, try using the software after step 4, to ensure that it works correctly when installed from a package. Checking your port with <command>portlint</command> Please use portlint to see if your port conforms to our guidelines. The portlint program is part of the ports collection. In particular, you may want to check if the Makefile is in the right shape and the package is named appropriately. Submitting the port First, make sure you have read the DOs and DON'Ts section. Now that you are happy with your port, the only thing remaining is to put it in the main FreeBSD ports tree and make everybody else happy about it too. We do not need your work directory or the pkgname.tgz package, so delete them now. Next, simply include the output of shar `find port_dir` in a bug report and send it with the &man.send-pr.1; program (see Bug Reports and General Commentary for more information about &man.send-pr.1;. If the uncompressed port is larger than 20KB, you should compress it into a tarfile and use &man.uuencode.1; before including it in the bug report (uuencoded tarfiles are acceptable even if the bug report is smaller than 20KB but are not preferred). Be sure to classify the bug report as category ports and class change-request (Do not mark the report confidential!). Also add a short description of the program you ported to the Description field of the PR and the shar or uuencoded tarfile to the Fix field. The latter one helps the committers a lot, who use scripts for the ports-work. One more time, do not include the original source distfile, the work directory, or the package you built with make package. In the past, we asked you to upload new port submissions in our ftp site (ftp.FreeBSD.org). This is no longer recommended as read access is turned off on the incoming/ directory of that site due to the large amount of pirated software showing up there. We will look at your port, get back to you if necessary, and put it in the tree. Your name will also appear in the list of “Additional FreeBSD contributors” in the FreeBSD Handbook and other files. Isn't that great?!? :-) You can make our work a lot easier, if you use a good description in the synopsis of the problem report. We prefer something like “New port: <short description of the port>” for new ports and “Update port: <category>/<port> <short description of the update>” for port updates. If you stick to this scheme, the chance that one takes a look at your PR soon is much bigger. Slow Porting Ok, so it was not that simple, and the port required some modifications to get it to work. In this section, we will explain, step by step, how to modify it to get it to work with the ports paradigm. How things work First, this is the sequence of events which occurs when the user first types make in your port's directory. You may find that having bsd.port.mk in another window while you read this really helps to understand it. But do not worry if you do not really understand what bsd.port.mk is doing, not many people do... :-> The fetch target is run. The fetch target is responsible for making sure that the tarball exists locally in DISTDIR. If fetch cannot find the required files in DISTDIR it will look up the URL MASTER_SITES, which is set in the Makefile, as well as our main ftp site at ftp://ftp.FreeBSD.org/pub/FreeBSD/ports/distfiles/, where we put sanctioned distfiles as backup. It will then attempt to fetch the named distribution file with FETCH, assuming that the requesting site has direct access to the Internet. If that succeeds, it will save the file in DISTDIR for future use and proceed. The extract target is run. It looks for your port's distribution file (typically a gzip'd tarball) in DISTDIR and unpacks it into a temporary subdirectory specified by WRKDIR (defaults to work). The patch target is run. First, any patches defined in PATCHFILES are applied. Second, if any patch files named patch-* are found in PATCHDIR (defaults to the files subdirectory), they are applied at this time in alphabetical order. The configure target is run. This can do any one of many different things. If it exists, scripts/configure is run. If HAS_CONFIGURE or GNU_CONFIGURE is set, WRKSRC/configure is run. If USE_IMAKE is set, XMKMF (default: xmkmf -a) is run. The build target is run. This is responsible for descending into the port's private working directory (WRKSRC) and building it. If USE_GMAKE is set, GNU make will be used, otherwise the system make will be used. The above are the default actions. In addition, you can define targets pre-something or post-something, or put scripts with those names, in the scripts subdirectory, and they will be run before or after the default actions are done. For example, if you have a post-extract target defined in your Makefile, and a file pre-build in the scripts subdirectory, the post-extract target will be called after the regular extraction actions, and the pre-build script will be executed before the default build rules are done. It is recommended that you use Makefile targets if the actions are simple enough, because it will be easier for someone to figure out what kind of non-default action the port requires. The default actions are done by the bsd.port.mk targets do-something. For example, the commands to extract a port are in the target do-extract. If you are not happy with the default target, you can fix it by redefining the do-something target in your Makefile. The “main” targets (e.g., extract, configure, etc.) do nothing more than make sure all the stages up to that one are completed and call the real targets or scripts, and they are not intended to be changed. If you want to fix the extraction, fix do-extract, but never ever touch extract! Now that you understand what goes on when the user types make, let us go through the recommended steps to create the perfect port. Getting the original sources Get the original sources (normally) as a compressed tarball (foo.tar.gz or foo.tar.Z) and copy it into DISTDIR. Always use mainstream sources when and where you can. If you cannot find a ftp/http site that is well-connected to the net, or can only find sites that have irritatingly non-standard formats, you might want to put a copy on a reliable ftp or http server that you control (e.g., your home page). Make sure you set MASTER_SITES to reflect your choice. If you cannot find somewhere convenient and reliable to put the distfile we can “house” it ourselves on ftp.FreeBSD.org. The distfile must be placed into ~/public_distfiles/ of someone's freefall account. Ask the person who commits your port to do this. This person will also set MASTER_SITES to MASTER_SITE_LOCAL and MASTER_SITE_SUBDIR to their freefall username. If your port's distfile changes all the time for no good reason, consider putting the distfile in your home page and listing it as the first MASTER_SITES. This will prevent users from getting checksum mismatch errors, and also reduce the workload of maintainers of our ftp site. Also, if there is only one master site for the port, it is recommended that you house a backup at your site and list it as the second MASTER_SITES. If your port requires some additional `patches' that are available on the Internet, fetch them too and put them in DISTDIR. Do not worry if they come from a site other than where you got the main source tarball, we have a way to handle these situations (see the description of PATCHFILES below). Modifying the port Unpack a copy of the tarball in a private directory and make whatever changes are necessary to get the port to compile properly under the current version of FreeBSD. Keep careful track of everything you do, as you will be automating the process shortly. Everything, including the deletion, addition, or modification of files should be doable using an automated script or patch file when your port is finished. If your port requires significant user interaction/customization to compile or install, you should take a look at one of Larry Wall's classic Configure scripts and perhaps do something similar yourself. The goal of the new ports collection is to make each port as “plug-and-play” as possible for the end-user while using a minimum of disk space. Unless explicitly stated, patch files, scripts, and other files you have created and contributed to the FreeBSD ports collection are assumed to be covered by the standard BSD copyright conditions. Patching In the preparation of the port, files that have been added or changed can be picked up with a recursive diff for later feeding to patch. Each set of patches you wish to apply should be collected into a file named patch-* where * denotes the sequence in which the patches will be applied — these are done in alphabetical order, thus aa first, ab second and so on. If you wish, you can use names that indicate the pathnames of the files that are patched, such as patch-Imakefile or patch-src-config.h. These files should be stored in PATCHDIR, from where they will be automatically applied. All patches should be relative to WRKSRC (generally the directory your port's tarball unpacks itself into, that being where the build is done). To make fixes and upgrades easier, you should avoid having more than one patch fix the same file (e.g., patch-aa and patch-ab both changing WRKSRC/foobar.c). Configuring Include any additional customization commands in your configure script and save it in the scripts subdirectory. As mentioned above, you can also do this with Makefile targets and/or scripts with the name pre-configure or post-configure. Handling user input If your port requires user input to build, configure, or install, then set IS_INTERACTIVE in your Makefile. This will allow “overnight builds” to skip your port if the user sets the variable BATCH in his environment (and if the user sets the variable INTERACTIVE, then only those ports requiring interaction are built). It is also recommended that if there are reasonable default answers to the questions, you check the PACKAGE_BUILDING variable and turn off the interactive script when it is set. This will allow us to build the packages for CD-ROMs and ftp. Configuring the Makefile Configuring the Makefile is pretty simple, and again we suggest that you look at existing examples before starting. Also, there is a sample Makefile in this handbook, so take a look and please follow the ordering of variables and sections in that template to make your port easier for others to read. Now, consider the following problems in sequence as you design your new Makefile: The original source Does it live in DISTDIR as a standard gzip'd tarball named something like foozolix-1.2.tar.gz? If so, you can go on to the next step. If not, you should look at overriding any of the DISTNAME, EXTRACT_CMD, EXTRACT_BEFORE_ARGS, EXTRACT_AFTER_ARGS, EXTRACT_SUFX, or DISTFILES variables, depending on how alien a format your port's distribution file is. (The most common case is EXTRACT_SUFX=.tar.Z, when the tarball is condensed by regular compress, not gzip.) In the worst case, you can simply create your own do-extract target to override the default, though this should be rarely, if ever, necessary. <makevar>PORTNAME</makevar> and <makevar>PORTVERSION</makevar> You should set PORTNAME to the base name of your port, and PORTVERSION to the version number of the port. <makevar>PORTREVISION</makevar> and <makevar>PORTEPOCH</makevar> <makevar>PORTREVISION</makevar> The PORTREVISION variable is a monotonically increasing value which is reset to 0 with every increase of PORTVERSION (i.e. every time a new official vendor release is made), and appended to the package name if non-zero. PORTREVISION is increased each time a change is made to the FreeBSD port which significantly affects the content or stucture of the derived package. Examples of when PORTREVISION should be bumped: Addition of patches to correct security vulnerabilities, bugs, or to add new functionality to the FreeBSD port. Changes to the port makefile to enable or disable compile-time options in the package. Changes in the packing list or the install-time behaviour of the package (e.g. change to a script which generates initial data for the package, like ssh host keys). Version bump of a port's shared library dependency (in this case, someone trying to install the old package after installing a newer version of the dependency will fail since it will look for the old libfoo.x instead of libfoo.(x+1)). Silent changes to the port distfile which have significant functional differences, i.e. changes to the distfile requiring a correction to distinfo with no corresponding change to PORTVERSION, where a diff -ru of the old and new versions shows non-trivial changes to the code. Examples of changes which do not require a PORTREVISION bump: Style changes to the port skeleton with no functional change to what appears in the resulting package. Changes to MASTER_SITES or other functional changes to the port which do not effect the resulting package. Trivial patches to the distfile such as correction of typos, which are not important enough that users of the package should go to the trouble of upgrading. Build fixes which cause a package to become compilable where it was previously failing (as long as the changes do not introduce any functional change on any other platforms on which the port did previously build). Since PORTREVISION reflects the content of the package, if no package was previously buildable then there is no need to increase PORTREVISION to mark a change. A rule of thumb is to ask yourself whether a change committed to a port is something which someone, somewhere, would benefit from having (either because of an enhancement, fix, or by virtue that the new package will actually work for them). If yes, the PORTREVISION should be bumped so that automated tools (e.g. pkg_version) will highlight the fact that a new package is available. <makevar>PORTEPOCH</makevar> From time to time a software vendor or FreeBSD porter will do something silly and release a version of their software which is actually numerically less than the previous version. An example of this is a port which goes from foo-20000801 to foo-1.0 (the former will be incorrectly treated as a newer version since 20000801 is a numerically greater value than 1). In situations such as this, the PORTEPOCH version should be increased. If PORTEPOCH is nonzero it is appended to the package name as described in section 0 above. PORTEPOCH is never decreased or reset to zero, because that would cause comparison to a package from an earlier epoch to fail (i.e. the package would not be detected as out of date): the new version number (e.g. 1.0,1 in the above example) is still numerically less than the previous version (2000801), but the ,1 suffix is treated specially by automated tools and found to be greater than the implied suffix ",0" on the earlier package) It is expected that PORTEPOCH will not be used for the majority of ports, and that sensible use of PORTVERSION can often pre-empt it becoming necessary if a future release of the software should change the version structure. However, care is needed by FreeBSD porters when a vendor release is made without an official version number - such as a code "snapshot" release. The temptation is to label the release with the release date, which will cause problems as in the example above when a new "official" release is made. For example, if a snapshot release is made on the date 20000917, and the previous version of the software was version 1.2, the snapshot release should be given a PORTVERSION of 1.2.20000917 or similar, not 20000917, so that the succeeding release, say 1.3, is still a numerically greater value. Example of <makevar>PORTREVISION</makevar> and <makevar>PORTEPOCH</makevar> usage The gtkmumble port, version 0.10, is committed to the ports collection. PORTNAME= gtkmumble PORTVERSION= 0.10 PKGNAME becomes gtkmumble-0.10. A security hole is discovered which requires a local FreeBSD patch. PORTREVISION is bumped accordingly. PORTNAME= gtkmumble PORTVERSIOn= 0.10 PORTREVISION= 1 PKGNAME becomes gtkmumble-0.10_1 A new version is released by the vendor, numbered 0.2 (it turns out the author actually intended 0.10 to actually mean 0.1.0, not what comes after 0.9 - oops, too late now). Since the new minor version 2 is numerically less than the previous version 10 the PORTEPOCH must be bumped to manually force the new package to be detected as "newer". Since it is a new vendor release of the code, PORTREVISION is reset to 0 (or removed from the makefile). PORTNAME= gtkmumble PORTVERSION= 0.2 PORTEPOCH= 1 PKGNAME becomes gtkmumble-0.2,1 The next release is 0.3. Since PORTEPOCH never decreases, the version variables are now: PORTNAME= gtkmumble PORTVERSION= 0.3 PORTEPOCH= 1 PKGNAME becomes gtkmumble-0.3,1 If PORTEPOCH were reset to 0 with this upgrade, someone who had installed the gtkmumble-0.10_1 package would not detect the gtkmumble-0.3 package as newer, since 3 is still numerically less than 10. <makevar>PKGNAMEPREFIX</makevar> and <makevar>PKGNAMESUFFIX</makevar> Two optional variables, PKGNAMEPREFIX and PKGNAMESUFFIX, are combined with PORTNAME and PORTVERSION to form PKGNAME as ${PKGNAMEPREFIX}${PORTNAME}${PKGNAMESUFFIX}-${PORTVERSION}. Make sure this conforms to our guidelines for a good package name. In particular, you are not allowed to use a hyphen (-) in PORTVERSION. Also, if the package name has the language- or the compiled.specifics part, use PKGNAMEPREFIX and PKGNAMESUFFIX, respectively. Do not make them part of PORTNAME. <makevar>DISTNAME</makevar> DISTNAME is the name of the port as called by the authors of the software. DISTNAME defaults to ${PORTNAME}-${PORTVERSION}, so override it if necessary. DISTNAME is only used in two places. First, the distribution file list (DISTFILES) defaults to ${DISTNAME}${EXTRACT_SUFX}. Second, the distribution file is expected to extract into a subdirectory named WRKSRC, which defaults to work/${DISTNAME}. PKGNAMEPREFIX and PKGNAMESUFFIX do not affect DISTNAME. Also note that when WRKSRC is equal to work/${PORTNAME}-${PORTVERSION} while the original source archive is named something other than ${PORTNAME}-${PORTVERSION}${EXTRACT_SUFX}, you should probably leave DISTNAME alone— you are better off defining DISTFILES than having to set both DISTNAME and WRKSRC (and possibly EXTRACT_SUFX). <makevar>CATEGORIES</makevar> When a package is created, it is put under /usr/ports/packages/All and links are made from one or more subdirectories of /usr/ports/packages. The names of these subdirectories are specified by the variable CATEGORIES. It is intended to make life easier for the user when he is wading through the pile of packages on the ftp site or the CD-ROM. Please take a look at the existing categories and pick the ones that are suitable for your port. This list also determines where in the ports tree the port is imported. If you put more than one category here, it is assumed that the port files will be put in the subdirectory with the name in the first category. See the categories section for more discussion about how to pick the right categories. If your port truly belongs to something that is different from all the existing ones, you can even create a new category name. In that case, please send mail to the &a.ports; to propose a new category. <makevar>MASTER_SITES</makevar> Record the directory part of the ftp/http-URL pointing at the original tarball in MASTER_SITES. Do not forget the trailing slash (/)! The make macros will try to use this specification for grabbing the distribution file with FETCH if they cannot find it already on the system. It is recommended that you put multiple sites on this list, preferably from different continents. This will safeguard against wide-area network problems, and we are even planning to add support for automatically determining the closest master site and fetching from there! If the original tarball is part of one of the popular archives such as X-contrib, GNU, or Perl CPAN, you may be able refer to those sites in an easy compact form using MASTER_SITE_* (e.g., MASTER_SITE_XCONTRIB and MASTER_SITE_PERL_GNU). Simply set MASTER_SITES to one of these variables and MASTER_SITE_SUBDIR to the path within the archive. Here is an example: MASTER_SITES= ${MASTER_SITE_XCONTRIB} MASTER_SITE_SUBDIR= applications These variables are defined in /usr/ports/Mk/bsd.sites.mk. There are new archives added all the time, so make sure to check the latest version of this file before submitting a port. The user can also set the MASTER_SITE_* variables in /etc/make.conf to override our choices, and use their favorite mirrors of these popular archives instead. <makevar>PATCHFILES</makevar> If your port requires some additional patches that are available by ftp or http, set PATCHFILES to the names of the files and PATCH_SITES to the URL of the directory that contains them (the format is the same as MASTER_SITES). If the patch is not relative to the top of the source tree (i.e., WRKSRC) because it contains some extra pathnames, set PATCH_DIST_STRIP accordingly. For instance, if all the pathnames in the patch have an extra foozolix-1.0/ in front of the filenames, then set PATCH_DIST_STRIP=-p1. Do not worry if the patches are compressed; they will be decompressed automatically if the filenames end with .gz or .Z. If the patch is distributed with some other files, such as documentation, in a gzip'd tarball, you cannot just use PATCHFILES. If that is the case, add the name and the location of the patch tarball to DISTFILES and MASTER_SITES. Then, use the EXTRA_PATCHES variable to point to those files and bsd.port.mk will automatically apply them for you. In particular, do not copy patch files into the PATCHDIR directory—that directory may not be writable. Note that the tarball will have been extracted alongside the regular source by then, so there is no need to explicitly extract it if it is a regular gzip'd or compress'd tarball. If you do the latter, take extra care not to overwrite something that already exists in that directory. Also, do not forget to add a command to remove the copied patch in the pre-clean target. <makevar>MAINTAINER</makevar> Set your mail-address here. Please. :-) For a detailed description of the responsibilities of maintainers, refer to the MAINTAINER on Makefiles section. Dependencies Many ports depend on other ports. There are five variables that you can use to ensure that all the required bits will be on the user's machine. There are also some pre-supported dependency variables for common cases, plus a few more to control the behaviour of dependencies. <makevar>LIB_DEPENDS</makevar> This variable specifies the shared libraries this port depends on. It is a list of lib:dir:target tuples where lib is the name of the shared library, dir is the directory in which to find it in case it is not available, and target is the target to call in that directory. For example, LIB_DEPENDS= jpeg.9:${PORTSDIR}/graphics/jpeg:install will check for a shared jpeg library with major version 9, and descend into the graphics/jpeg subdirectory of your ports tree to build and install it if it is not found. The target part can be omitted if it is equal to DEPENDS_TARGET (which defaults to install). The lib part is an argument given to ldconfig -r | grep -wF. There shall be no regular expressions in this variable. The dependency is checked twice, once from within the extract target and then from within the install target. Also, the name of the dependency is put into the package so that pkg_add will automatically install it if it is not on the user's system. <makevar>RUN_DEPENDS</makevar> This variable specifies executables or files this port depends on during run-time. It is a list of path:dir:target tuples where path is the name of the executable or file, dir is the directory in which to find it in case it is not available, and target is the target to call in that directory. If path starts with a slash (/), it is treated as a file and its existence is tested with test -e; otherwise, it is assumed to be an executable, and which -s is used to determine if the program exists in the user's search path. For example, RUN_DEPENDS= ${PREFIX}/etc/innd:${PORTSDIR}/news/inn \ wish8.0:${PORTSDIR}/x11-toolkits/tk80 will check if the file or directory /usr/local/etc/innd exists, and build and install it from the news/inn subdirectory of the ports tree if it is not found. It will also see if an executable called wish8.0 is in your search path, and descend into the x11-toolkits/tk80 subdirectory of your ports tree to build and install it if it is not found. In this case, innd is actually an executable; if an executable is in a place that is not expected to be in a normal user's search path, you should use the full pathname. The dependency is checked from within the install target. Also, the name of the dependency is put in to the package so that pkg_add will automatically install it if it is not on the user's system. The target part can be omitted if it is the same as DEPENDS_TARGET. <makevar>BUILD_DEPENDS</makevar> This variable specifies executables or files this port requires to build. Like RUN_DEPENDS, it is a list of path:dir:target tuples. For example, BUILD_DEPENDS= unzip:${PORTSDIR}/archivers/unzip will check for an executable called unzip, and descend into the archivers/unzip subdirectory of your ports tree to build and install it if it is not found. “build” here means everything from extraction to compilation. The dependency is checked from within the extract target. The target part can be omitted if it is the same as DEPENDS_TARGET <makevar>FETCH_DEPENDS</makevar> This variable specifies executables or files this port requires to fetch. Like the previous two, it is a list of path:dir:target tuples. For example, FETCH_DEPENDS= ncftp2:${PORTSDIR}/net/ncftp2 will check for an executable called ncftp2, and descend into the net/ncftp2 subdirectory of your ports tree to build and install it if it is not found. The dependency is checked from within the fetch target. The target part can be omitted if it is the same as DEPENDS_TARGET. <makevar>DEPENDS</makevar> If there is a dependency that does not fall into either of the above four categories, or your port requires having the source of the other port extracted in addition to having it installed, then use this variable. This is a list of dir:target, as there is nothing to check, unlike the previous four. The target part can be omitted if it is the same as DEPENDS_TARGET. Common dependency variables Define USE_XLIB=yes if your port requires the X Window System to be installed (it is implied by USE_IMAKE). Define USE_GMAKE=yes if your port requires GNU make instead of BSD make. Define USE_AUTOCONF=yes if your port requires GNU autoconf to be run. Define USE_QT=yes if your port uses the latest qt toolkit. Use USE_PERL5=yes if your port requires version 5 of the perl language. (The last is especially important since some versions of FreeBSD have perl5 as part of the base system while others do not.) Notes on dependencies As mentioned above, the default target to call when a dependency is required is DEPENDS_TARGET. It defaults to install. This is a user variable; it is never defined in a port's Makefile. If your port needs a special way to handle a dependency, use the :target part of the *_DEPENDS variables instead of redefining DEPENDS_TARGET. When you type make clean, its dependencies are automatically cleaned too. If you do not wish this to happen, define the variable NOCLEANDEPENDS in your environment. To depend on another port unconditionally, use the variable ${NONEXISTENT} as the first field of BUILD_DEPENDS or RUN_DEPENDS. Use this only when you need to the to get to the source of the other port. You can often save compilation time by specifying the target too. For instance BUILD_DEPENDS= ${NONEXISTENT}:${PORTSDIR}/graphics/jpeg:extract will always descend to the JPEG port and extract it. Do not use DEPENDS unless there is no other way the behaviour you want can be accomplished. It will cause the other port to always be built (and installed, by default), and the dependency will go into the packages as well. If this is really what you need, you should probably write it as BUILD_DEPENDS and RUN_DEPENDS instead—at least the intention will be clear. Optional dependencies Some large applications can be built in a number of configurations, adding functionality if one of a number of libraries or applications is available. Since not all users want those libraries or applications, the ports system provides hooks that the port author can use to decide which configuration should be built. Supporting these properly will make uses happy, and effectively provide 2 or more ports for the price of one. The easiest of these to use is WITHOUT_X11. If the port can be built both with and without X support, then it should normally be built with X support. If WITHOUT_X11 is defined, then the version that does not have X support should be built. Various parts of GNOME have such knobs, though they are slightly more difficult to use. The variables to use in the Makefile are WANT_* and HAVE_*. If the application can be built both with or without one of the dependencies listed below, then the Makefile should set WANT_PKG, and should build the version that uses PKG if HAVE_PKG is defined. The WANT_* variables currently supported this way are WANT_GLIB, WANT_GTK, WANT_ESOUND, WANT_IMLIB, and WANT_GNOME. Building mechanisms If your package uses GNU make, set USE_GMAKE=yes. If your package uses configure, set HAS_CONFIGURE=yes. If your package uses GNU configure, set GNU_CONFIGURE=yes (this implies HAS_CONFIGURE). If you want to give some extra arguments to configure (the default argument list --prefix=${PREFIX} for GNU configure and empty for non-GNU configure), set those extra arguments in CONFIGURE_ARGS. If your package uses GNU autoconf, set USE_AUTOCONF=yes. This implies GNU_CONFIGURE, and will cause autoconf to be run before configure. If your package is an X application that creates Makefiles from Imakefiles using imake, then set USE_IMAKE=yes. This will cause the configure stage to automatically do an xmkmf -a. If the flag is a problem for your port, set XMKMF=xmkmf. If the port uses imake but does not understand the install.man target, NO_INSTALL_MANPAGES=yes should be set. In addition, the author of the original port should be shot. :-> If your port's source Makefile has something else than all as the main build target, set ALL_TARGET accordingly. Same goes for install and INSTALL_TARGET. Special considerations There are some more things you have to take into account when you create a port. This section explains the most common of those. Shared Libraries If your port installs one or more shared libraries, define a INSTALLS_SHLIB make variable, which will instruct a bsd.port.mk to run ${LDCONFIG} -m on the directory where the new library is installed (usually PREFIX/lib) during post-install target to register it into the shared library cache. This variable, when defined, will also facilitate addition of an appropriate @exec /sbin/ldconfig -m and @unexec /sbin/ldconfig -R pair into your pkg-plist file, so that a user who installed the package can start using the shared library immediately and deinstallation will not cause the system to still believe the library is there. If you need, you can override default location where the new library is installed by defining LDCONFIG_DIRS make variable, which should contain a list of directories into which shared libraries are to be installed. For example if your port installs shared libraries into PREFIX/lib/foo and PREFIX/lib/bar directories you could use the following in your Makefile: INSTALLS_SHLIB= yes LDCONFIG_DIRS= %%PREFIX%%/lib/foo %%PREFIX%%/lib/bar Note that content of LDCONFIG_DIRS is passed through &man.sed.1; just like the rest of pkg-plist, so PLIST_SUB substitutions also apply here. It is recommended that you use %%PREFIX%% for PREFIX, %%LOCALBASE%% for LOCALBASE and %%X11BASE%% for X11BASE. <makevar>MASTERDIR</makevar> If your port needs to build slightly different versions of packages by having a variable (for instance, resolution, or paper size) take different values, create one subdirectory per package to make it easier for users to see what to do, but try to share as many files as possible between ports. Typically you only need a very short Makefile in all but one of the directories if you use variables cleverly. In the sole Makefiles, you can use MASTERDIR to specify the directory where the rest of the files are. Also, use a variable as part of PKGNAMESUFFIX so the packages will have different names. This will be best demonstrated by an example. This is part of japanese/xdvi300/Makefile; PORTNAME= xdvi PORTVERSION= 17 PKGNAMEPREFIX= ja- PKGNAMESUFFIX= ${RESOLUTION} : # default RESOLUTION?= 300 .if ${RESOLUTION} != 118 && ${RESOLUTION} != 240 && \ ${RESOLUTION} != 300 && ${RESOLUTION} != 400 @${ECHO} "Error: invalid value for RESOLUTION: \"${RESOLUTION}\"" @${ECHO} "Possible values are: 118, 240, 300 (default) and 400." @${FALSE} .endif japanese/xdvi300 also has all the regular patches, package files, etc. If you type make there, it will take the default value for the resolution (300) and build the port normally. As for other resolutions, this is the entire xdvi118/Makefile: RESOLUTION= 118 MASTERDIR= ${.CURDIR}/../xdvi300 .include ${MASTERDIR}/Makefile (xdvi240/Makefile and xdvi400/Makefile are similar). The MASTERDIR definition tells bsd.port.mk that the regular set of subdirectories like FILESDIR and SCRIPTDIR are to be found under xdvi300. The RESOLUTION=118 line will override the RESOLUTION=300 line in xdvi300/Makefile and the port will be built with resolution set to 118. Shared library versions Please read our policy on shared library versioning to understand what to do with shared library versions in general. Do not blindly assume software authors know what they are doing; many of them do not. It is very important that these details are carefully considered, as we have quite a unique situation where we are trying to have dozens of potentially incompatible software pairs co-exist. Careless port imports have caused great trouble regarding shared libraries in the past (ever wondered why the port jpeg-6b has a shared library version of 9?). If in doubt, send a message to the &a.ports;. Most of the time, your job ends by determining the right shared library version and making appropriate patches to implement it. Manpages The MAN[1-9LN] variables will automatically add any manpages to pkg-plist (this means you must not list manpages in the pkg-plist—see generating PLIST for more). It also makes the install stage automatically compress or uncompress manpages depending on the setting of NOMANCOMPRESS in /etc/make.conf. If your port tries to install multiple names for manpages using symlinks or hardlinks, you must use the MLINKS variable to identify these. The link installed by your port will be destroyed and recreated by bsd.port.mk to make sure it points to the correct file. Any manpages listed in MLINKS must not be listed in the pkg-plist. To specify whether the manpages are compressed upon installation, use the MANCOMPRESSED variable. This variable can take three values, yes, no and maybe. yes means manpages are already installed compressed, no means they are not, and maybe means the software already respects the value of NOMANCOMPRESS so bsd.port.mk does not have to do anything special. MANCOMPRESSED is automatically set to yes if USE_IMAKE is set and NO_INSTALL_MANPAGES is not set, and to no otherwise. You do not have to explicitly define it unless the default is not suitable for your port. If your port anchors its man tree somewhere other than PREFIX, you can use the MANPREFIX to set it. Also, if only manpages in certain sections go in a non-standard place, such as some Perl modules ports, you can set individual man paths using MANsectPREFIX (where sect is one of 1-9, L or N). If your manpages go to language-specific subdirectories, set the name of the languages to MANLANG. The value of this variable defaults to "" (i.e., English only). Here is an example that puts it all together. MAN1= foo.1 MAN3= bar.3 MAN4= baz.4 MLINKS= foo.1 alt-name.8 MANLANG= "" ja MAN3PREFIX= ${PREFIX}/share/foobar MANCOMPRESSED= yes This states that six files are installed by this port; ${PREFIX}/man/man1/foo.1.gz ${PREFIX}/man/ja/man1/foo.1.gz ${PREFIX}/share/foobar/man/man3/bar.3.gz ${PREFIX}/share/foobar/man/ja/man3/bar.3.gz ${PREFIX}/man/man4/baz.4.gz ${PREFIX}/man/ja/man4/baz.4.gz Additionally ${PREFIX}/man/man8/alt-name.8.gz may or may not be installed by your port. Regardless, a symlink will be made to join the foo(1) manpage and alt-name(8) manpage. Ports that require Motif There are many programs that require a Motif library (available from several commercial vendors, while there is a free clone reported to be able to run many applications in x11-toolkits/lesstif) to compile. Since it is a popular toolkit and their licenses usually permit redistribution of statically linked binaries, we have made special provisions for handling ports that require Motif in a way that we can easily compile binaries linked either dynamically (for people who are compiling from the port) or statically (for people who distribute packages). <makevar>REQUIRES_MOTIF</makevar> If your port requires Motif, define this variable in the Makefile. This will prevent people who do not own a copy of Motif from even attempting to build it. <makevar>MOTIFLIB</makevar> This variable will be set by bsd.port.mk to be the appropriate reference to the Motif library. Please patch the source to use this wherever the Motif library is referenced in the Makefile or Imakefile. There are two common cases: If the port refers to the Motif library as -lXm in its Makefile or Imakefile, simply substitute ${MOTIFLIB} for it. If the port uses XmClientLibs in its Imakefile, change it to ${MOTIFLIB} ${XTOOLLIB} ${XLIB}. Note that MOTIFLIB (usually) expands to -L/usr/X11R6/lib -lXm or /usr/X11R6/lib/libXm.a, so there is no need to add -L or -l in front. X11 fonts If your port installs fonts for the X Window system, put them in X11BASE/lib/X11/fonts/local. This directory is new to XFree86 release 3.3.3. If it does not exist, please create it, and print out a message urging the user to update their XFree86 to 3.3.3 or newer, or at least add this directory to the font path in /etc/XF86Config. Info files The new version of texinfo (included in 2.2.2-RELEASE and onwards) contains a utility called install-info to add and delete entries to the dir file. If your port installs any info documents, please follow these instructions so your port/package will correctly update the user's PREFIX/info/dir file. (Sorry for the length of this section, but is it imperative to weave all the info files together. If done correctly, it will produce a beautiful listing, so please bear with me! First, this is what you (as a porter) need to know &prompt.user; install-info --help install-info [OPTION]... [INFO-FILE [DIR-FILE]] Install INFO-FILE in the Info directory file DIR-FILE. Options: --delete Delete existing entries in INFO-FILE; don't insert any new entries. : --entry=TEXT Insert TEXT as an Info directory entry. : --section=SEC Put this file's entries in section SEC of the directory. : This program will not actually install info files; it merely inserts or deletes entries in the dir file. Here's a seven-step procedure to convert ports to use install-info. editors/emacs will be used as an example. Look at the texinfo sources and make a patch to insert @dircategory and @direntry statements to files that do not have them. This is part of my patch: --- ./man/vip.texi.org Fri Jun 16 15:31:11 1995 +++ ./man/vip.texi Tue May 20 01:28:33 1997 @@ -2,6 +2,10 @@ @setfilename ../info/vip @settitle VIP +@dircategory The Emacs editor and associated tools +@direntry +* VIP: (vip). A VI-emulation for Emacs. +@end direntry @iftex @finalout : The format should be self-explanatory. Many authors leave a dir file in the source tree that contains all the entries you need, so look around before you try to write your own. Also, make sure you look into related ports and make the section names and entry indentations consistent (we recommend that all entry text start at the 4th tab stop). Note that you can put only one info entry per file because of a bug in install-info --delete that deletes only the first entry if you specify multiple entries in the @direntry section. You can give the dir entries to install-info as arguments ( and ) instead of patching the texinfo sources. This probably is not a good idea for ports because you need to duplicate the same information in three places (Makefile and @exec/@unexec of pkg-plist; see below). However, if you have Japanese (or other multibyte encoding) info files, you will have to use the extra arguments to install-info because makeinfo cannot handle those texinfo sources. (See Makefile and pkg-plist of japanese/skk for examples on how to do this). Go back to the port directory and do a make clean; make and verify that the info files are regenerated from the texinfo sources. Since the texinfo sources are newer than the info files, they should be rebuilt when you type make; but many Makefiles do not include correct dependencies for info files. In emacs' case, it was necessary to patch the main Makefile.in so it would descend into the man subdirectory to rebuild the info pages. --- ./Makefile.in.org Mon Aug 19 21:12:19 1996 +++ ./Makefile.in Tue Apr 15 00:15:28 1997 @@ -184,7 +184,7 @@ # Subdirectories to make recursively. `lisp' is not included # because the compiled lisp files are part of the distribution # and you cannot remake them without installing Emacs first. -SUBDIR = lib-src src +SUBDIR = lib-src src man # The makefiles of the directories in $SUBDIR. SUBDIR_MAKEFILES = lib-src/Makefile man/Makefile src/Makefile oldXMenu/Makefile lwlib/Makefile --- ./man/Makefile.in.org Thu Jun 27 15:27:19 1996 +++ ./man/Makefile.in Tue Apr 15 00:29:52 1997 @@ -66,6 +66,7 @@ ${srcdir}/gnu1.texi \ ${srcdir}/glossary.texi +all: info info: $(INFO_TARGETS) dvi: $(DVI_TARGETS) The second hunk was necessary because the default target in the man subdir is called info, while the main Makefile wants to call all. The installation of the info info file was also removed because we already have one with the same name in /usr/share/info (that patch is not shown here). If there is a place in the Makefile that is installing the dir file, delete it. Your port may not be doing it. Also, remove any commands that are otherwise mucking around with the dir file. --- ./Makefile.in.org Mon Aug 19 21:12:19 1996 +++ ./Makefile.in Mon Apr 14 23:38:07 1997 @@ -368,14 +368,8 @@ if [ `(cd ${srcdir}/info && /bin/pwd)` != `(cd ${infodir} && /bin/pwd)` ]; \ then \ (cd ${infodir}; \ - if [ -f dir ]; then \ - if [ ! -f dir.old ]; then mv -f dir dir.old; \ - else mv -f dir dir.bak; fi; \ - fi; \ cd ${srcdir}/info ; \ - (cd $${thisdir}; ${INSTALL_DATA} ${srcdir}/info/dir ${infodir}/dir); \ - (cd $${thisdir}; chmod a+r ${infodir}/dir); \ for f in ccmode* cl* dired-x* ediff* emacs* forms* gnus* info* message* mh-e* sc* vip*; do \ (cd $${thisdir}; \ ${INSTALL_DATA} ${srcdir}/info/$$f ${infodir}/$$f; \ chmod a+r ${infodir}/$$f); \ (This step is only necessary if you are modifying an existing port.) Take a look at pkg-plist and delete anything that is trying to patch up info/dir. They may be in pkg-install or some other file, so search extensively. Index: pkg-plist =================================================================== RCS file: /usr/cvs/ports/editors/emacs/pkg-plist,v retrieving revision 1.15 diff -u -r1.15 pkg-plist --- pkg-plist 1997/03/04 08:04:00 1.15 +++ pkg-plist 1997/04/15 06:32:12 @@ -15,9 +15,6 @@ man/man1/emacs.1.gz man/man1/etags.1.gz man/man1/ctags.1.gz -@unexec cp %D/info/dir %D/info/dir.bak -info/dir -@unexec cp %D/info/dir.bak %D/info/dir info/cl info/cl-1 info/cl-2 Add a post-install target to the Makefile to call install-info with the installed info files. (It is no longer necessary to create the dir file yourself; install-info automatically creates this file if it does not exist.) Index: Makefile =================================================================== RCS file: /usr/cvs/ports/editors/emacs/Makefile,v retrieving revision 1.26 diff -u -r1.26 Makefile --- Makefile 1996/11/19 13:14:40 1.26 +++ Makefile 1997/05/20 10:25:09 1.28 @@ -20,5 +20,8 @@ post-install: .for file in emacs-19.34 emacsclient etags ctags b2m strip ${PREFIX}/bin/${file} .endfor +.for info in emacs vip viper forms gnus mh-e cl sc dired-x ediff ccmode + install-info ${PREFIX}/info/${info} ${PREFIX}/info/dir +.endfor .include <bsd.port.mk> Edit pkg-plist and add equivalent @exec statements and also @unexec for pkg_delete. Index: pkg-plist =================================================================== RCS file: /usr/cvs/ports/editors/emacs/pkg-plist,v retrieving revision 1.15 diff -u -r1.15 pkg-plist --- pkg-plist 1997/03/04 08:04:00 1.15 +++ pkg-plist 1997/05/20 10:25:12 1.17 @@ -16,7 +14,14 @@ man/man1/etags.1.gz man/man1/ctags.1.gz +@unexec install-info --delete %D/info/emacs %D/info/dir : +@unexec install-info --delete %D/info/ccmode %D/info/dir info/cl info/cl-1 @@ -87,6 +94,18 @@ info/viper-3 info/viper-4 +@exec install-info %D/info/emacs %D/info/dir : +@exec install-info %D/info/ccmode %D/info/dir libexec/emacs/19.34/i386--freebsd/cvtmail libexec/emacs/19.34/i386--freebsd/digest-doc The @unexec install-info --delete commands have to be listed before the info files themselves so they can read the files. Also, the @exec install-info commands have to be after the info files and the @exec command that creates the the dir file. Test and admire your work. :-). Check the dir file before and after each step. The <filename>pkg-<replaceable>*</replaceable></filename> files There are some tricks we have not mentioned yet about the pkg-* files that come in handy sometimes. <filename>pkg-message</filename> If you need to display a message to the installer, you may place the message in pkg-message. This capability is often useful to display additional installation steps to be taken after a pkg_add or to display licensing information. The pkg-message file does not need to be added to pkg-plist. Also, it will not get automatically printed if the user is using the port, not the package, so you should probably display it from the post-install target yourself. <filename>pkg-install</filename> If your port needs to execute commands when the binary package is installed with pkg_add you can do this via the pkg-install script. This script will automatically be added to the package, and will be run twice by pkg_add. The first time as ${SH} pkg-install ${PKGNAME} PRE-INSTALL and the second time as ${SH} pkg-install ${PKGNAME} POST-INSTALL. $2 can be tested to determine which mode the script is being run in. The PKG_PREFIX environmental variable will be set to the package installation directory. See &man.pkg.add.1; for additional information. This script is not run automatically if you install the port with make install. If you are depending on it being run, you will have to explicitly call it from your port's Makefile. <filename>pkg-req</filename> If your port needs to determine if it should install or not, you can create a pkg-req “requirements” script. It will be invoked automatically at installation/deinstallation time to determine whether or not installation/deinstallation should proceed. Changing <filename>pkg-plist</filename> based on make variables Some ports, particularly the p5- ports, need to change their pkg-plist depending on what options they are configured with (or version of perl, in the case of p5- ports). To make this easy, any instances in the pkg-plist of %%OSREL%%, %%PERL_VER%%, and %%PERL_VERSION%% will be substituted for appropriately. The value of %%OSREL%% is the numeric revision of the operating system (e.g., 2.2.7). %%PERL_VERSION%% is the full version number of perl (e.g., 5.00502) and %%PERL_VER%% is the perl version number minus the patchlevel (e.g., 5.005). If you need to make other substitutions, you can set the PLIST_SUB variable with a list of VAR=VALUE pairs and instances of %%VAR%%' will be substituted with VALUE in the pkg-plist. For instance, if you have a port that installs many files in a version-specific subdirectory, you can put something like OCTAVE_VERSION= 2.0.13 PLIST_SUB= OCTAVE_VERSION=${OCTAVE_VERSION} in the Makefile and use %%OCTAVE_VERSION%% wherever the version shows up in pkg-plist. That way, when you upgrade the port, you will not have to change dozens (or in some cases, hundreds) of lines in the pkg-plist. This substitution (as well as addition of any man pages) will be done between the do-install and post-install targets, by reading from PLIST and writing to TMPPLIST (default: WRKDIR/.PLIST.mktmp). So if your port builds PLIST on the fly, do so in or before do-install. Also, if your port needs to edit the resulting file, do so in post-install to a file named TMPPLIST. Changing the names of <filename>pkg-<replaceable>*</replaceable></filename> files All the names of pkg-* files are defined using variables so you can change them in your Makefile if need be. This is especially useful when you are sharing the same pkg-* files among several ports or have to write to one of the above files (see writing to places other than WRKDIR for why it is a bad idea to write directly in to the pkg-* subdirectory). Here is a list of variable names and their default values. (PKGDIR defaults to ${MASTERDIR}.) Variable Default value COMMENT ${PKGDIR}/pkg-comment DESCR ${PKGDIR}/pkg-descr PLIST ${PKGDIR}/pkg-plist PKGINSTALL ${PKGDIR}/pkg-install PKGDEINSTALL ${PKGDIR}/pkg-deinstall PKGREQ ${PKGDIR}/pkg-req PKGMESSAGE ${PKGDIR}/pkg-message Please change these variables rather than overriding PKG_ARGS. If you change PKG_ARGS, those files will not correctly be installed in /var/db/pkg upon install from a port. Licensing Problems Some software packages have restrictive licenses or can be in violation of the law in some countries (such as violating a patent). What we can do with them varies a lot, depending on the exact wordings of the respective licenses. It is your responsibility as a porter to read the licensing terms of the software and make sure that the FreeBSD project will not be held accountable for violating them by redistributing the source or compiled binaries either via ftp or CD-ROM. If in doubt, please contact the &a.ports;. There are two variables you can set in the Makefile to handle the situations that arise frequently: If the port has a “do not sell for profit” type of license, set the variable NO_CDROM to a string describing the reason why. We will make sure such ports will not go into the CD-ROM come release time. The distfile and package will still be available via ftp. If the resulting package needs to be built uniquely for each site, or the resulting binary package cannot be distributed due to licensing; set the variable NO_PACKAGE to a string describing the reason why. We will make sure such packages will not go on the ftp site, nor into the CD-ROM come release time. The distfile will still be included on both however. If the port has legal restrictions on who can use it (e.g., patented stuff) or has a “no commercial use” license, set the variable RESTRICTED to be the string describing the reason why. For such ports, the distfiles/packages will not be available even from our ftp sites. The GNU General Public License (GPL), both version 1 and 2, should not be a problem for ports. If you are a committer, make sure you update the ports/LEGAL file too. Upgrading When you notice that a port is out of date compared to the latest version from the original authors, first make sure you have the latest port. You can find them in the ports/ports-current directory of the ftp mirror sites. You may also use CVSup to keep your whole ports collection up-to-date, as described in the Handbook. The next step is to send a mail to the maintainer, if one is listed in the port's Makefile. That person may already be working on an upgrade, or have a reason to not upgrade the port right now (because of, for example, stability problems of the new version). If the maintainer asks you to do the upgrade or there is not any such person to begin with, please make the upgrade and send the recursive diff (either unified or context diff is fine, but port committers appear to prefer unified diff more) of the new and old ports directories to us (e.g., if your modified port directory is called superedit and the original as in our tree is superedit.bak, then send us the result of diff -ruN superedit.bak superedit). Please examine the output to make sure all the changes make sense. The best way to send us the diff is by including it via &man.send-pr.1; (category ports). Please mention any added or deleted files in the message, as they have to be explicitly specified to CVS when doing a commit. If the diff is more than about 20KB, please compress and uuencode it; otherwise, just include it in the PR as is. Once again, please use &man.diff.1; and not &man.shar.1; to send updates to existing ports! <anchor id="porting-dads">Dos and Don'ts Here is a list of common dos and don'ts that you encounter during the porting process.You should check your own port against this list, but you can also check ports in the PR database that others have submitted. Submit any comments on ports you check as described in Bug Reports and General Commentary. Checking ports in the PR database will both make it faster for us to commit them, and prove that you know what you are doing. Strip Binaries Do strip binaries. If the original source already strips the binaries, fine; otherwise you should add a post-install rule to to it yourself. Here is an example: post-install: strip ${PREFIX}/bin/xdl Use the &man.file.1; command on the installed executable to check whether the binary is stripped or not. If it does not say not stripped, it is stripped. INSTALL_* macros Do use the macros provided in bsd.port.mk to ensure correct modes and ownership of files in your own *-install targets. INSTALL_PROGRAM is a command to install binary executables. INSTALL_SCRIPT is a command to install executable scripts. INSTALL_DATA is a command to install sharable data. INSTALL_MAN is a command to install manpages and other documentation (it does not compress anything). These are basically the install command with all the appropriate flags. See below for an example on how to use them. <makevar>WRKDIR</makevar> Do not write anything to files outside WRKDIR. WRKDIR is the only place that is guaranteed to be writable during the port build (see compiling ports from CDROM for an example of building ports from a read-only tree). If you need to modify one of the pkg-* files, do so by redefining a variable, not by writing over it. <makevar>WRKDIRPREFIX</makevar> Make sure your port honors WRKDIRPREFIX. Most ports do not have to worry about this. In particular, if you are referring to a WRKDIR of another port, note that the correct location is WRKDIRPREFIXPORTSDIR/subdir/name/work not PORTSDIR/subdir/name/work or .CURDIR/../../subdir/name/work or some such. Also, if you are defining WRKDIR yourself, make sure you prepend ${WRKDIRPREFIX}${.CURDIR} in the front. Differentiating operating systems and OS versions You may come across code that needs modifications or conditional compilation based upon what version of UNIX it is running under. If you need to make such changes to the code for conditional compilation, make sure you make the changes as general as possible so that we can back-port code to FreeBSD 1.x systems and cross-port to other BSD systems such as 4.4BSD from CSRG, BSD/386, 386BSD, NetBSD, and OpenBSD. The preferred way to tell 4.3BSD/Reno (1990) and newer versions of the BSD code apart is by using the BSD macro defined in <sys/param.h>. Hopefully that file is already included; if not, add the code: #if (defined(__unix__) || defined(unix)) && !defined(USG) #include <sys/param.h> #endif to the proper place in the .c file. We believe that every system that defines these two symbols has sys/param.h. If you find a system that does not, we would like to know. Please send mail to the &a.ports;. Another way is to use the GNU Autoconf style of doing this: #ifdef HAVE_SYS_PARAM_H #include <sys/param.h> #endif Do not forget to add -DHAVE_SYS_PARAM_H to the CFLAGS in the Makefile for this method. Once you have sys/param.h included, you may use: #if (defined(BSD) && (BSD >= 199103)) to detect if the code is being compiled on a 4.3 Net2 code base or newer (e.g. FreeBSD 1.x, 4.3/Reno, NetBSD 0.9, 386BSD, BSD/386 1.1 and below). Use: #if (defined(BSD) && (BSD >= 199306)) to detect if the code is being compiled on a 4.4 code base or newer (e.g. FreeBSD 2.x, 4.4, NetBSD 1.0, BSD/386 2.0 or above). The value of the BSD macro is 199506 for the 4.4BSD-Lite2 code base. This is stated for informational purposes only. It should not be used to distinguish between versions of FreeBSD based only on 4.4-Lite vs. versions that have merged in changes from 4.4-Lite2. The __FreeBSD__ macro should be used instead. Use sparingly: __FreeBSD__ is defined in all versions of FreeBSD. Use it if the change you are making only affects FreeBSD. Porting gotchas like the use of sys_errlist[] vs strerror() are Berkeleyisms, not FreeBSD changes. In FreeBSD 2.x, __FreeBSD__ is defined to be 2. In earlier versions, it is 1. Later versions will bump it to match their major version number. If you need to tell the difference between a FreeBSD 1.x system and a FreeBSD 2.x or 3.x system, usually the right answer is to use the BSD macros described above. If there actually is a FreeBSD specific change (such as special shared library options when using ld) then it is OK to use __FreeBSD__ and #if __FreeBSD__ > 1 to detect a FreeBSD 2.x and later system. If you need more granularity in detecting FreeBSD systems since 2.0-RELEASE you can use the following: #if __FreeBSD__ >= 2 #include <osreldate.h> # if __FreeBSD_version >= 199504 /* 2.0.5+ release specific code here */ # endif #endif Release __FreeBSD_version 2.0-RELEASE 119411 2.1-CURRENT 199501, 199503 2.0.5-RELEASE 199504 2.2-CURRENT before 2.1 199508 2.1.0-RELEASE 199511 2.2-CURRENT before 2.1.5 199512 2.1.5-RELEASE 199607 2.2-CURRENT before 2.1.6 199608 2.1.6-RELEASE 199612 2.1.7-RELEASE 199612 2.2-RELEASE 220000 2.2.1-RELEASE 220000 (no change) 2.2-STABLE after 2.2.1-RELEASE 220000 (no change) 2.2-STABLE after texinfo-3.9 221001 2.2-STABLE after top 221002 2.2.2-RELEASE 222000 2.2-STABLE after 2.2.2-RELEASE 222001 2.2.5-RELEASE 225000 2.2-STABLE after 2.2.5-RELEASE 225001 2.2-STABLE after ldconfig -R merge 225002 2.2.6-RELEASE 226000 2.2.7-RELEASE 227000 2.2-STABLE after 2.2.7-RELEASE 227001 2.2-STABLE after &man.semctl.2; change 227002 2.2.8-RELEASE 228000 2.2-STABLE after 2.2.8-RELEASE 228001 3.0-CURRENT before &man.mount.2; change 300000 3.0-CURRENT after &man.mount.2; change 300001 3.0-CURRENT after &man.semctl.2; change 300002 3.0-CURRENT after ioctl arg changes 300003 3.0-CURRENT after ELF conversion 300004 3.0-RELEASE 300005 3.0-CURRENT after 3.0-RELEASE 300006 3.0-STABLE after 3/4 branch 300007 3.1-RELEASE 310000 3.1-STABLE after 3.1-RELEASE 310001 3.1-STABLE after C++ constructor/destructor order change 310002 3.2-RELEASE 320000 3.2-STABLE 320001 3.2-STABLE after binary-incompatible IPFW and socket changes 320002 3.3-RELEASE 330000 3.3-STABLE 330001 3.3-STABLE after adding &man.mkstemp.3; to libc 330002 3.4-RELEASE 340000 3.4-STABLE 340001 4.0-CURRENT after 3.4 branch 400000 4.0-CURRENT after change in dynamic linker handling 400001 4.0-CURRENT after C++ constructor/destructor order change 400002 4.0-CURRENT after functioning &man.dladdr.3; 400003 4.0-CURRENT after __deregister_frame_info dynamic linker bug fix (also 4.0-CURRENT after EGCS 1.1.2 integration) 400004 4.0-CURRENT after &man.suser.9; API change (also 4.0-CURRENT after newbus) 400005 4.0-CURRENT after cdevsw registration change 400006 4.0-CURRENT after the addition of so_cred for socket level credentials 400007 4.0-CURRENT after the addition of a poll syscall wrapper to libc_r 400008 4.0-CURRENT after the change of the kernel's dev_t type to struct specinfo pointer 400009 4.0-CURRENT after fixing a hole in &man.jail.2; 400010 4.0-CURRENT after the sigset_t datatype change 400011 4.0-CURRENT after the cutover to the GCC 2.95.2 compiler 400012 4.0-CURRENT after adding pluggable linux-mode ioctl handlers 400013 4.0-CURRENT after importing OpenSSL 400014 4.0-CURRENT after the C++ ABI change in GCC 2.95.2 from -fvtable-thunks to -fno-vtable-thunks by default 400015 4.0-CURRENT after importing OpenSSH 400016 4.0-RELEASE 400017 4.0-STABLE after 4.0-RELEASE 400018 4.0-STABLE after merging libxpg4 code into libc. 400020 4.0-STABLE after upgrading Binutils to 2.10.0, ELF branding changes, and tcsh in the base system. 400021 4.1-RELEASE 410000 4.1-STABLE after 4.1-RELEASE 410001 4.1-STABLE after &man.setproctitle.3; moved from libutil to libc. 410002 4.1.1-RELEASE 411000 4.1.1-STABLE after 4.1.1-RELEASE 411001 4.2-RELEASE 420000 4.2-STABLE after combining libgcc.a and libgcc_r.a, and associated GCC linkage changes. 420001 5.0-CURRENT 500000 5.0-CURRENT after adding addition ELF header fields, and changing our ELF binary branding method. 500001 5.0-CURRENT after kld metadata changes. 500002 5.0-CURRENT after buf/bio changes. 500003 5.0-CURRENT after binutils upgrade. 500004 5.0-CURRENT after merging libxpg4 code into libc and after TASKQ interface introduction. 500005 5.0-CURRENT after the addition of AGP interfaces. 500006 5.0-CURRENT after Perl upgrade to 5.6.0 500007 5.0-CURRENT after the update of KAME code to 2000/07 sources. 500008 5.0-CURRENT after ether_ifattach() and ether_ifdetach() changes. 500009 5.0-CURRENT after changing mtree defaults back to original variant, adding -L to follow symlinks. 500010 5.0-CURRENT after kqueue API changed. 500011 5.0-CURRENT after &man.setproctitle.3; moved from libutil to libc. 500012 5.0-CURRENT after the first SMPng commit. 500013 5.0-CURRENT after <sys/select.h> moved to <sys/selinfo.h>. 500014 5.0-CURRENT after combining libgcc.a and libgcc_r.a, and associated GCC linkage changes. 500015 5.0-CURRENT after change allowing libc and libc_r to be linked together, deprecating -pthread option. 500016 5.0-CURRENT after switch from struct ucred to struct xucred to stabilize kernel-exported API for mountd et al. 500017 5.0-CURRENT after addition of CPUTYPE make variable for controlling CPU-specific optimizations. 500018 Note that 2.2-STABLE sometimes identifies itself as “2.2.5-STABLE” after the 2.2.5-RELEASE. The pattern used to be year followed by the month, but we decided to change it to a more straightforward major/minor system starting from 2.2. This is because the parallel development on several branches made it infeasible to classify the releases simply by their real release dates. If you are making a port now, you do not have to worry about old -CURRENTs; they are listed here just for your reference. In the hundreds of ports that have been done, there have only been one or two cases where __FreeBSD__ should have been used. Just because an earlier port screwed up and used it in the wrong place does not mean you should do so too. Writing something after <filename>bsd.port.mk</filename> Do not write anything after the .include <bsd.port.mk> line. It usually can be avoided by including bsd.port.pre.mk somewhere in the middle of your Makefile and bsd.port.post.mk at the end. You need to include either the pre.mk/post.mk pair or bsd.port.mk only; do not mix these two. bsd.port.pre.mk only defines a few variables, which can be used in tests in the Makefile, bsd.port.post.mk defines the rest. Here are some important variables defined in bsd.port.pre.mk (this is not the complete list, please read bsd.port.mk for the complete list). Variable Description ARCH The architecture as returned by uname -m (e.g., i386) OPSYS The operating system type, as returned by uname -s (e.g., FreeBSD) OSREL The release version of the operating system (e.g., 2.1.5 or 2.2.7) OSVERSION The numeric version of the operating system, same as __FreeBSD_version. PORTOBJFORMAT The object format of the system (aout or elf) LOCALBASE The base of the “local” tree (e.g., /usr/local/) X11BASE The base of the “X11” tree (e.g., /usr/X11R6) PREFIX Where the port installs itself (see more on PREFIX). If you have to define the variables USE_IMAKE, USE_X_PREFIX, or MASTERDIR, do so before including bsd.port.pre.mk. Here are some examples of things you can write after bsd.port.pre.mk: # no need to compile lang/perl5 if perl5 is already in system .if ${OSVERSION} > 300003 BROKEN= perl is in system .endif # only one shlib version number for ELF .if ${PORTOBJFORMAT} == "elf" TCL_LIB_FILE= ${TCL_LIB}.${SHLIB_MAJOR} .else TCL_LIB_FILE= ${TCL_LIB}.${SHLIB_MAJOR}.${SHLIB_MINOR} .endif # software already makes link for ELF, but not for a.out post-install: .if ${PORTOBJFORMAT} == "aout" ${LN} -sf liblinpack.so.1.0 ${PREFIX}/lib/liblinpack.so .endif Install additional documentation If your software has some documentation other than the standard man and info pages that you think is useful for the user, install it under PREFIX/share/doc. This can be done, like the previous item, in the post-install target. Create a new directory for your port. The directory name should reflect what the port is. This usually means PORTNAME. However, if you think the user might want different versions of the port to be installed at the same time, you can use the whole PKGNAME. Make the installation dependent to the variable NOPORTDOCS so that users can disable it in /etc/make.conf, like this: post-install: .if !defined(NOPORTDOCS) ${MKDIR} ${PREFIX}/share/doc/xv ${INSTALL_MAN} ${WRKSRC}/docs/xvdocs.ps ${PREFIX}/share/doc/xv .endif Do not forget to add them to pkg-plist too. (Do not worry about NOPORTDOCS here; there is currently no way for the packages to read variables from /etc/make.conf.) You can also use the pkg-message file to display messages upon installation. See the using pkg-message section for details. pkg-message does not need to be added to pkg-plist. <makevar>DIST_SUBDIR</makevar> Do not let your port clutter /usr/ports/distfiles. If your port requires a lot of files to be fetched, or contains a file that has a name that might conflict with other ports (e.g., Makefile), set DIST_SUBDIR to the name of the port (${PORTNAME} or ${PKGNAMEPREFIX}${PORTNAME} should work fine). This will change DISTDIR from the default /usr/ports/distfiles to /usr/ports/distfiles/DIST_SUBDIR, and in effect puts everything that is required for your port into that subdirectory. It will also look at the subdirectory with the same name on the backup master site at ftp.FreeBSD.org. (Setting DISTDIR explicitly in your Makefile will not accomplish this, so please use DIST_SUBDIR.) This does not affect the MASTER_SITES you define in your Makefile. Package information Do include package information, i.e. pkg-comment, pkg-descr, and pkg-plist. Note that these files are not used only for packaging anymore, and are mandatory now, even if NO_PACKAGE is set. RCS strings Do not put RCS strings in patches. CVS will mangle them when we put the files into the ports tree, and when we check them out again, they will come out different and the patch will fail. RCS strings are surrounded by dollar ($) signs, and typically start with $Id or $RCS. Recursive diff Using the recurse () option to diff to generate patches is fine, but please take a look at the resulting patches to make sure you do not have any unnecessary junk in there. In particular, diffs between two backup files, Makefiles when the port uses Imake or GNU configure, etc., are unnecessary and should be deleted. If you had to edit configure.in and run autoconf to regenerate configure, do not take the diffs of configure (it often grows to a few thousand lines!); define USE_AUTOCONF=yes and take the diffs of configure.in. Also, if you had to delete a file, then you can do it in the post-extract target rather than as part of the patch. Once you are happy with the resulting diff, please split it up into one source file per patch file. <makevar>PREFIX</makevar> Do try to make your port install relative to PREFIX. (The value of this variable will be set to LOCALBASE (default /usr/local), unless USE_X_PREFIX or USE_IMAKE is set, in which case it will be X11BASE (default /usr/X11R6).) Not hard-coding /usr/local or /usr/X11R6 anywhere in the source will make the port much more flexible and able to cater to the needs of other sites. For X ports that use imake, this is automatic; otherwise, this can often be done by simply replacing the occurrences of /usr/local (or /usr/X11R6 for X ports that do not use imake) in the various scripts/Makefiles in the port to read PREFIX, as this variable is automatically passed down to every stage of the build and install processes. Make sure your application isn't installing things in /usr/local instead of PREFIX. A quick test for this is to do this is: &prompt.root; make clean; make package PREFIX=/var/tmp/port-name If anything is installed outside of PREFIX, making the package creation process will complain that it can't find the files. This does not test for the existence of internal references, or correct use of LOCALBASE for references to files from other ports. Testing the installation in /var/tmp/port-name to do that that while you have it installed would do that. Do not set USE_X_PREFIX unless your port truly requires it (i.e., it links against X libs or it needs to reference files in X11BASE). The variable PREFIX can be reassigned in your Makefile or in the user's environment. However, it is strongly discouraged for individual ports to set this variable explicitly in the Makefiles. Also, refer to programs/files from other ports with the variables mentioned above, not explicit pathnames. For instance, if your port requires a macro PAGER to be the full pathname of less, use the compiler flag: -DPAGER=\"${PREFIX}/bin/less\" or -DPAGER=\"${LOCALBASE}/bin/less\" if this is an X port, instead of -DPAGER=\"/usr/local/bin/less\". This way it will have a better chance of working if the system administrator has moved the whole `/usr/local' tree somewhere else. Subdirectories Try to let the port put things in the right subdirectories of PREFIX. Some ports lump everything and put it in the subdirectory with the port's name, which is incorrect. Also, many ports put everything except binaries, header files and manual pages in the a subdirectory of lib, which does not bode well with the BSD paradigm. Many of the files should be moved to one of the following: etc (setup/configuration files), libexec (executables started internally), sbin (executables for superusers/managers), info (documentation for info browser) or share (architecture independent files). See man &man.hier.7; for details, the rules governing /usr pretty much apply to /usr/local too. The exception are ports dealing with USENET “news”. They may use PREFIX/news as a destination for their files. Cleaning up empty directories Do make your ports clean up after themselves when they are deinstalled. This is usually accomplished by adding @dirrm lines for all directories that are specifically created by the port. You need to delete subdirectories before you can delete parent directories. : lib/X11/oneko/pixmaps/cat.xpm lib/X11/oneko/sounds/cat.au : @dirrm lib/X11/oneko/pixmaps @dirrm lib/X11/oneko/sounds @dirrm lib/X11/oneko However, sometimes @dirrm will give you errors because other ports also share the same subdirectory. You can call rmdir from @unexec to remove only empty directories without warning. @unexec rmdir %D/share/doc/gimp 2>/dev/null || true This will neither print any error messages nor cause pkg_delete to exit abnormally even if PREFIX/share/doc/gimp is not empty due to other ports installing some files in there. UIDs If your port requires a certain user to be on the installed system, let the pkg-install script call pw to create it automatically. Look at net/cvsup-mirror for an example. If your port must use the same user/group ID number when it is installed as a binary package as when it was compiled, then you must choose a free UID from 50 to 99 and register it below. Look at japanese/Wnn for an example. Make sure you do not use a UID already used by the system or other ports. This is the current list of UIDs between 50 and 99. majordom:*:54:54:Majordomo Pseudo User:/usr/local/majordomo:/nonexistent cyrus:*:60:60:the cyrus mail server:/nonexistent:/nonexistent gnats:*:61:1:GNATS database owner:/usr/local/share/gnats/gnats-db:/bin/sh uucp:*:66:66:UUCP pseudo-user:/var/spool/uucppublic:/usr/libexec/uucp/uucico xten:*:67:67:X-10 daemon:/usr/local/xten:/nonexistent pop:*:68:6:Post Office Owner (popper):/nonexistent:/nonexistent wnn:*:69:7:Wnn:/nonexistent:/nonexistent ifmail:*:70:66:Ifmail user:/nonexistent:/nonexistent pgsql:*:70:70:PostgreSQL pseudo-user:/usr/local/pgsql:/bin/sh ircd:*:72:72:IRCd hybrid:/nonexistent:/nonexistent alias:*:81:81:QMail user:/var/qmail/alias:/nonexistent qmaill:*:83:81:QMail user:/var/qmail:/nonexistent qmaild:*:82:81:QMail user:/var/qmail:/nonexistent qmailq:*:85:82:QMail user:/var/qmail:/nonexistent qmails:*:87:82:QMail user:/var/qmail:/nonexistent qmailp:*:84:81:QMail user:/var/qmail:/nonexistent qmailr:*:86:82:QMail user:/var/qmail:/nonexistent msql:*:87:87:mSQL-2 pseudo-user:/var/db/msqldb:/bin/sh mysql:*:88:88:MySQL Daemon:/var/db/mysql:/sbin/nologin vpopmail:*:89:89::0:0:User &:/usr/local/vpopmail:/nonexistent Please include a notice when you submit a port (or an upgrade) that reserves a new UID or GID in this range. This allows us to keep the list of reserved IDs up to date. Do things rationally The Makefile should do things simply and reasonably. If you can make it a couple of lines shorter or more readable, then do so. Examples include using a make .if construct instead of a shell if construct, not redefining do-extract if you can redefine EXTRACT* instead, and using GNU_CONFIGURE instead of CONFIGURE_ARGS += --prefix=${PREFIX}. Respect <makevar>CFLAGS</makevar> The port should respect the CFLAGS variable. If it does not, please add NO_PACKAGE=ignores cflags to the Makefile. An example of a Makefile respecting the CFLAGS variable follows. Note the +=: CFLAGS += -Wall -Werror Here is an example which does not respect the CFLAGS variable: CFLAGS = -Wall -Werror The CFLAGS variable is defined on FreeBSD systems in /etc/make.conf. The first example appends additional flags to the CFLAGS variable, preserving any system-wide definitions. The second example clobbers anything previously defined. Configuration files If your port requires some configuration files in PREFIX/etc, do not just install them and list them in pkg-plist. That will cause pkg_delete to delete files carefully edited by the user and a new installation to wipe them out. Instead, install sample files with a suffix (filename.sample will work well) and print out a message pointing out that the user has to copy and edit the file before the software can be made to work. Portlint Do check your work with portlint before you submit or commit it. Feedback Do send applicable changes/patches to the original author/maintainer for inclusion in next release of the code. This will only make your job that much easier for the next release. <filename>README.html</filename> Do not include the README.html file. This file is not part of the cvs collection but is generated using the make readme command. Miscellanea The files pkg-comment, pkg-descr, and pkg-plist should each be double-checked. If you are reviewing a port and feel they can be worded better, do so. Do not copy more copies of the GNU General Public License into our system, please. Please be careful to note any legal issues! Do not let us illegally distribute software! If you are stuck… Do look at existing examples and the bsd.port.mk file before asking us questions! ;-) Do ask us questions if you have any trouble! Do not just beat your head against a wall! :-) A Sample <filename>Makefile</filename> Here is a sample Makefile that you can use to create a new port. Make sure you remove all the extra comments (ones between brackets)! It is recommended that you follow this format (ordering of variables, empty lines between sections, etc.). This format is designed so that the most important information is easy to locate. We recommend that you use portlint to check the Makefile. [the header...just to make it easier for us to identify the ports.] # New ports collection makefile for: xdvi [the "version required" line is only needed when the PORTVERSION variable is not specific enough to describe the port.] # Date created: 26 May 1995 [this is the person who did the original port to FreeBSD, in particular, the person who wrote the first version of this Makefile. Remember, this should not be changed when upgrading the port later.] # Whom: Satoshi Asami <asami@FreeBSD.org> # # $FreeBSD$ [ ^^^^^^^^^ This will be automatically replaced with RCS ID string by CVS when it is committed to our repository. If upgrading a port, do not alter this line back to "$FreeBSD$". CVS deals with it automatically.] # [section to describe the port itself and the master site - PORTNAME and PORTVERSION are always first, followed by CATEGORIES, and then MASTER_SITES, which can be followed by MASTER_SITE_SUBDIR. PKGNAMEPREFIX and PKGNAMESUFFIX, if needed, will be after that. Then comes DISTNAME, EXTRACT_SUFX and/or DISTFILES, and then EXTRACT_ONLY, as necessary.] PORTNAME= xdvi PORTVERSION= 18.2 CATEGORIES= print [do not forget the trailing slash ("/")! if you are not using MASTER_SITE_* macros] MASTER_SITES= ${MASTER_SITE_XCONTRIB} MASTER_SITE_SUBDIR= applications PKGNAMEPREFIX= ja- DISTNAME= xdvi-pl18 [set this if the source is not in the standard ".tar.gz" form] EXTRACT_SUFX= .tar.Z [section for distributed patches -- can be empty] PATCH_SITES= ftp://ftp.sra.co.jp/pub/X11/japanese/ PATCHFILES= xdvi-18.patch1.gz xdvi-18.patch2.gz [maintainer; *mandatory*! This is the person (preferably with commit privileges) whom a user can contact for questions and bug reports - this person should be the porter or someone who can forward questions to the original porter reasonably promptly. If you really do not want to have your address here, set it to "ports@FreeBSD.org".] MAINTAINER= asami@FreeBSD.org [dependencies -- can be empty] RUN_DEPENDS= gs:${PORTSDIR}/print/ghostscript LIB_DEPENDS= Xpm.5:${PORTSDIR}/graphics/xpm [this section is for other standard bsd.port.mk variables that do not belong to any of the above] [If it asks questions during configure, build, install...] IS_INTERACTIVE= yes [If it extracts to a directory other than ${DISTNAME}...] WRKSRC= ${WRKDIR}/xdvi-new [If the distributed patches were not made relative to ${WRKSRC}, you may need to tweak this] PATCH_DIST_STRIP= -p1 [If it requires a "configure" script generated by GNU autoconf to be run] GNU_CONFIGURE= yes [If it requires GNU make, not /usr/bin/make, to build...] USE_GMAKE= yes [If it is an X application and requires "xmkmf -a" to be run...] USE_IMAKE= yes [et cetera.] [non-standard variables to be used in the rules below] MY_FAVORITE_RESPONSE= "yeah, right" [then the special rules, in the order they are called] pre-fetch: i go fetch something, yeah post-patch: i need to do something after patch, great pre-install: and then some more stuff before installing, wow [and then the epilogue] .include <bsd.port.mk> Automated package list creation First, make sure your port is almost complete, with only pkg-plist missing. Create an empty pkg-plist. &prompt.root; touch pkg-plist Next, create a new set of directories which your port can be installed, and install any dependencies. &prompt.root; mtree -U -f /etc/mtree/BSD.local.dist -d -e -p /var/tmp/port-name &prompt.root; make depends PREFIX=/var/tmp/port-name Store the directory structure in a new file. &prompt.root; (cd /var/tmp/port-name && find * -type d) > OLD-DIRS If your port honors PREFIX (which it should) you can then install the port and create the package list. &prompt.root; make install PREFIX=/var/tmp/port-name &prompt.root; (cd /var/tmp/port-name && find * \! -type d) > pkg-plist You must also add any newly created directories to the packing list. &prompt.root; (cd /var/tmp/port-name && find * -type d) | comm -13 OLD-DIRS - | sed -e 's#^#@dirrm #' >> pkg-plist Finally, you need to tidy up the packing list by hand; it isn't all automated. Manual pages should be listed in the port's Makefile under MANn, and not in the package list. User configuration files should be removed, or installed as filename.sample. The info/dir file should not be listed and appropriate install-info lines should be added as noted in the info files section. Any libraries installed by the port should be listed as specified in the shared libraries section. Package Names The following are the conventions you should follow in naming your packages. This is to have our package directory easy to scan, as there are already lots and lots of packages and users are going to turn away if they hurt their eyes! The package name should look like language_region-name-compiled.specifics-version.numbers. The package name is defined as ${PKGNAMEPREFIX}${PORTNAME}${PKGNAMESUFFIX}-${PORTVERSION}. Make sure to set the variables to conform to that format. FreeBSD strives to support the native language of its users. The language- part should be a two letter abbreviation of the natural language defined by ISO-639 if the port is specific to a certain language. Examples are ja for Japanese, ru for Russian, vi for Vietnamese, zh for Chinese, ko for Korean and de for German. If the port is specific to a certain region within the language area, add the two letter country code as well. Examples are en_US for US English and fr_CH for Swiss French. The language- part should be set in the PKGNAMEPREFIX variable. The first letter of name part should be lowercase. (The rest of the name can contain capital letters, so use your own discretion when you are converting a software name that has some capital letters in it.) There is a tradition of naming Perl 5 modules by prepending p5- and converting the double-colon separator to a hyphen; for example, the Data::Dumper module becomes p5-Data-Dumper. If the software in question has numbers, hyphens, or underscores in its name, you may include them as well (like kinput2). If the port can be built with different hardcoded defaults (usually part of the directory name in a family of ports), the -compiled.specifics part should state the compiled-in defaults (the hyphen is optional). Examples are papersize and font units. The compiled.specifics part should be set in the PKGNAMESUFFIX variable. The version string should follow a dash (-) and be a period-separated list of integers and single lowercase alphabetics. In particular, it is not permissible to have another dash inside the version string. The only exception is the string pl (meaning `patchlevel'), which can be used only when there are no major and minor version numbers in the software. If the software version has strings like "alpha", "beta", "rc", or "pre", take the first letter and put it immediately after a period. If the version string continues after those names, the numbers should follow the single alphabet without an extra period between them. The idea is to make it easier to sort ports by looking at the version string. In particular, make sure version number components are always delimited by a period, and if the date is part of the string, use the yyyy.mm.dd format, not dd.mm.yyyy or the non-Y2K compliant yy.mm.dd format. Here are some (real) examples on how to convert the name as called by the software authors to a suitable package name: Distribution Name PKGNAMEPREFIX PORTNAME PKGNAMESUFFIX PORTVERSION Reason mule-2.2.2 (empty) mule (empty) 2.2.2 No changes required XFree86-3.3.6 (empty) XFree86 (empty) 3.3.6 No changes required EmiClock-1.0.2 (empty) emiclock (empty) 1.0.2 No uppercase names for single programs rdist-1.3alpha (empty) rdist (empty) 1.3.a No strings like alpha allowed es-0.9-beta1 (empty) es (empty) 0.9.b1 No strings like beta allowed mailman-2.0rc3 (empty) mailman (empty) 2.0.r3 No strings like rc allowed v3.3beta021.src (empty) tiff (empty) 3.3 What the heck was that anyway? tvtwm (empty) tvtwm (empty) pl11 Version string always required piewm (empty) piewm (empty) 1.0 Version string always required xvgr-2.10pl1 (empty) xvgr (empty) 2.10.1 pl allowed only when no major/minor version numbers gawk-2.15.6 ja- gawk (empty) 2.15.6 Japanese language version psutils-1.13 (empty) psutils -letter 1.13 Papersize hardcoded at package build time pkfonts (empty) pkfonts 300 1.0 Package for 300dpi fonts If there is absolutely no trace of version information in the original source and it is unlikely that the original author will ever release another version, just set the version string to 1.0 (like the piewm example above). Otherwise, ask the original author or use the date string (yyyy.mm.dd) as the version. Categories As you already know, ports are classified in several categories. But for this to work, it is important that porters and users understand what each category is for and how we decide what to put in each category. Current list of categories First, this is the current list of port categories. Those marked with an asterisk (*) are virtual categories—those that do not have a corresponding subdirectory in the ports tree. For non-virtual categories, you will find a one-line description in the pkg/COMMENT file in that subdirectory (e.g., archivers/pkg/COMMENT). Category Description afterstep* Ports to support the AfterStep window manager. archivers Archiving tools. astro Astronomical ports. audio Sound support. benchmarks Benchmarking utilities. biology Biology-related software. cad Computer aided design tools. chinese Chinese language support. comms Communication software. Mostly software to talk to your serial port. converters Character code converters. databases Databases. deskutils Things that used to be on the desktop before computers were invented. devel Development utilities. Do not put libraries here just because they are libraries—unless they truly do not belong anywhere else, they should not be in this category. editors General editors. Specialized editors go in the section for those tools (e.g., a mathematical-formula editor will go in math). elisp* Emacs-lisp ports. emulators Emulators for other operating systems. Terminal emulators do not belong here—X-based ones should go to x11 and text-based ones to either comms or misc, depending on the exact functionality. french French language support. ftp FTP client and server utilities. If your port speaks both FTP and HTTP, put it in ftp with a secondary category of www. games Games. german German language support. gnome* Ports from the GNU Object Model Environment (GNOME) Project. graphics Graphics utilities. hebrew Hebrew language support. irc Internet Relay Chat utilities. ipv6* IPv6 related software. japanese Japanese language support. java Java language support. kde* Ports from the K Desktop Environment (KDE) Project. korean Korean language support. lang Programming languages. linux* Linux applications and support utilities. mail Mail software. math Numerical computation software and other utilities for mathematics. mbone MBone applications. misc Miscellaneous utilities—basically things that do not belong anywhere else. This is the only category that should not appear with any other non-virtual category. If you have misc with something else in your CATEGORIES line, that means you can safely delete misc and just put the port in that other subdirectory! net Miscellaneous networking software. news USENET news software. offix* Ports from the OffiX suite. palm Software support for the 3Com Palm(tm) series. perl5* Ports that require perl version 5 to run. picobsd Ports to support PicoBSD. plan9* Various programs from Plan9. print Printing software. Desktop publishing tools (previewers, etc.) belong here too. python* Software written in python. ruby* Software written in ruby. russian Russian language support. science Scientific ports that don't fit into other categories such as astro, biology and math. security Security utilities. shells Command line shells. sysutils System utilities. tcl76* Ports that use Tcl version 7.6 to run. tcl80* Ports that use Tcl version 8.0 to run. tcl81* Ports that use Tcl version 8.1 to run. tcl82* Ports that use Tcl version 8.2 to run. textproc Text processing utilities. It does not include desktop publishing tools, which go to print/. tk42* Ports that use Tk version 4.2 to run. tk80* Ports that use Tk version 8.0 to run. tk81* Ports that use Tk version 8.1 to run. tk82* Ports that use Tk version 8.2 to run. tkstep80* Ports that use TkSTEP version 8.0 to run. ukrainian Ukrainian language support. vietnamese Vietnamese language support. windowmaker* Ports to support the WindowMaker window manager www Software related to the World Wide Web. HTML language support belongs here too. x11 The X window system and friends. This category is only for software that directly supports the window system. Do not put regular X applications here. If your port is an X application, define USE_XLIB (implied by USE_IMAKE) and put it in the appropriate categories. Also, many of them go into other x11-* categories (see below). x11-clocks X11 clocks. x11-fm X11 file managers. x11-fonts X11 fonts and font utilities. x11-servers X11 servers. x11-toolkits X11 toolkits. x11-wm X11 window managers. zope* Zope support. Choosing the right category As many of the categories overlap, you often have to choose which of the categories should be the primary category of your port. There are several rules that govern this issue. Here is the list of priorities, in decreasing order of precedence. Language specific categories always come first. For example, if your port installs Japanese X11 fonts, then your CATEGORIES line would read japanese x11-fonts. Specific categories win over less-specific ones. For instance, an HTML editor should be listed as www editors, not the other way around. Also, you do not need to list net when the port belongs to any of irc, mail, mbone, news, security, or www. x11 is used as a secondary category only when the primary category is a natural language. In particular, you should not put x11 in the category line for X applications. Emacs modes should be placed in the same ports category as the application supported by the mode, not in editors. For example, an Emacs mode to edit source files of some programming language should go into lang. If your port truly does not belong anywhere else, put it in misc. If you are not sure about the category, please put a comment to that effect in your send-pr submission so we can discuss it before we import it. If you are a committer, send a note to the &a.ports; so we can discuss it first—too often new ports are imported to the wrong category only to be moved right away. Changes to this document and the ports system If you maintain a lot of ports, you should consider following the &a.ports;. Important changes to the way ports work will be announced there. You can always find more detailed information on the latest changes by looking at the bsd.port.mk CVS log. That is It, Folks! Boy, this sure was a long tutorial, wasn't it? Thanks for following us to here, really. Now that you know how to do a port, have at it and convert everything in the world into ports! That is the easiest way to start contributing to the FreeBSD Project! :-)