Index: head/sys/conf/NOTES =================================================================== --- head/sys/conf/NOTES (revision 131258) +++ head/sys/conf/NOTES (revision 131259) @@ -1,2420 +1,2428 @@ # $FreeBSD$ # # NOTES -- Lines that can be cut/pasted into kernel and hints configs. # # Lines that begin with 'device', 'options', 'machine', 'ident', 'maxusers', # 'makeoptions', 'hints', etc. go into the kernel configuration that you # run config(8) with. # # Lines that begin with 'hint.' are NOT for config(8), they go into your # hints file. See /boot/device.hints and/or the 'hints' config(8) directive. # # Please use ``make LINT'' to create an old-style LINT file if you want to # do kernel test-builds. # # This file contains machine independent kernel configuration notes. For # machine dependent notes, look in /sys//conf/NOTES. # # # NOTES conventions and style guide: # # Large block comments should begin and end with a line containing only a # comment character. # # To describe a particular object, a block comment (if it exists) should # come first. Next should come device, options, and hints lines in that # order. All device and option lines must be described by a comment that # doesn't just expand the device or option name. Use only a concise # comment on the same line if possible. Very detailed descriptions of # devices and subsystems belong in man pages. # # A space followed by a tab separates 'options' from an option name. Two # spaces followed by a tab separate 'device' from a device name. Comments # after an option or device should use one space after the comment character. # To comment out a negative option that disables code and thus should not be # enabled for LINT builds, precede 'options' with "#!". # # # This is the ``identification'' of the kernel. Usually this should # be the same as the name of your kernel. # ident LINT # # The `maxusers' parameter controls the static sizing of a number of # internal system tables by a formula defined in subr_param.c. # Omitting this parameter or setting it to 0 will cause the system to # auto-size based on physical memory. # maxusers 10 # # The `makeoptions' parameter allows variables to be passed to the # generated Makefile in the build area. # # CONF_CFLAGS gives some extra compiler flags that are added to ${CFLAGS} # after most other flags. Here we use it to inhibit use of non-optimal # gcc builtin functions (e.g., memcmp). # # DEBUG happens to be magic. # The following is equivalent to 'config -g KERNELNAME' and creates # 'kernel.debug' compiled with -g debugging as well as a normal # 'kernel'. Use 'make install.debug' to install the debug kernel # but that isn't normally necessary as the debug symbols are not loaded # by the kernel and are not useful there anyway. # # KERNEL can be overridden so that you can change the default name of your # kernel. # # MODULES_OVERRIDE can be used to limit modules built to a specific list. # makeoptions CONF_CFLAGS=-fno-builtin #Don't allow use of memcmp, etc. #makeoptions DEBUG=-g #Build kernel with gdb(1) debug symbols #makeoptions KERNEL=foo #Build kernel "foo" and install "/foo" # Only build Linux API modules and plus those parts of the sound system I need. #makeoptions MODULES_OVERRIDE="linux sound/snd sound/pcm sound/driver/maestro3" makeoptions DESTDIR=/tmp # # Certain applications can grow to be larger than the 512M limit # that FreeBSD initially imposes. Below are some options to # allow that limit to grow to 1GB, and can be increased further # with changing the parameters. MAXDSIZ is the maximum that the # limit can be set to, and the DFLDSIZ is the default value for # the limit. MAXSSIZ is the maximum that the stack limit can be # set to. You might want to set the default lower than the max, # and explicitly set the maximum with a shell command for processes # that regularly exceed the limit like INND. # options MAXDSIZ=(1024UL*1024*1024) options MAXSSIZ=(128UL*1024*1024) options DFLDSIZ=(1024UL*1024*1024) # # BLKDEV_IOSIZE sets the default block size used in user block # device I/O. Note that this value will be overridden by the label # when specifying a block device from a label with a non-0 # partition blocksize. The default is PAGE_SIZE. # options BLKDEV_IOSIZE=8192 # Options for the VM subsystem # L2 cache size (in KB) can be specified in PQ_CACHESIZE options PQ_CACHESIZE=512 # color for 512k cache # Deprecated options supported for backwards compatibility #options PQ_NOOPT # No coloring #options PQ_LARGECACHE # color for 512k cache #options PQ_HUGECACHE # color for 1024k cache #options PQ_MEDIUMCACHE # color for 256k cache #options PQ_NORMALCACHE # color for 64k cache # This allows you to actually store this configuration file into # the kernel binary itself, where it may be later read by saying: # strings -n 3 /boot/kernel/kernel | sed -n 's/^___//p' > MYKERNEL # options INCLUDE_CONFIG_FILE # Include this file in kernel options GEOM_AES # Don't use, use GEOM_BDE options GEOM_APPLE # Apple partitioning options GEOM_BDE # Disk encryption. options GEOM_BSD # BSD disklabels options GEOM_CONCAT # Disk concatenation. options GEOM_FOX # Redundant path mitigation options GEOM_GATE # Userland services. options GEOM_GPT # GPT partitioning options GEOM_MBR # DOS/MBR partitioning options GEOM_NOP # Test class. options GEOM_PC98 # NEC PC9800 partitioning options GEOM_STRIPE # Disk striping. options GEOM_SUNLABEL # Sun/Solaris partitioning options GEOM_VOL # Volume names from UFS superblock # # The root device and filesystem type can be compiled in; # this provides a fallback option if the root device cannot # be correctly guessed by the bootstrap code, or an override if # the RB_DFLTROOT flag (-r) is specified when booting the kernel. # options ROOTDEVNAME=\"ufs:da0s2e\" ##################################################################### # Scheduler options: # # Specifying one of SCHED_4BSD or SCHED_ULE is mandatory. These options # select which scheduler is compiled in. # # SCHED_4BSD is the historical, proven, BSD scheduler. It has a global run # queue and no cpu affinity which makes it suboptimal for SMP. It has very # good interactivity and priority selection. # # SCHED_ULE is a new scheduler that has been designed for SMP and has some # advantages for UP as well. It is intended to replace the 4BSD scheduler # over time. # options SCHED_4BSD #options SCHED_ULE ##################################################################### # SMP OPTIONS: # # SMP enables building of a Symmetric MultiProcessor Kernel. # Mandatory: options SMP # Symmetric MultiProcessor Kernel # ADAPTIVE_MUTEXES changes the behavior of blocking mutexes to spin # if the thread that currently owns the mutex is executing on another # CPU. options ADAPTIVE_MUTEXES # MUTEX_NOINLINE forces mutex operations to call functions to perform each # operation rather than inlining the simple cases. This can be used to # shrink the size of the kernel text segment. Note that this behavior is # already implied by the INVARIANT_SUPPORT, INVARIANTS, MUTEX_PROFILING, # and WITNESS options. options MUTEX_NOINLINE # SMP Debugging Options: # # MUTEX_DEBUG enables various extra assertions in the mutex code. +# SLEEPQUEUE_PROFILING enables rudimentary profiling of the hash table +# used to hold active sleep queues. +# TURNSTILE_PROFILING enables rudimentary profiling of the hash table +# used to hold active lock queues. # WITNESS enables the witness code which detects deadlocks and cycles # during locking operations. # WITNESS_DDB causes the witness code to drop into the kernel debugger if # a lock hierarchy violation occurs or if locks are held when going to # sleep. # WITNESS_SKIPSPIN disables the witness checks on spin mutexes. options MUTEX_DEBUG options WITNESS options WITNESS_DDB options WITNESS_SKIPSPIN # MUTEX_PROFILING - Profiling mutual exclusion locks (mutexes). See # MUTEX_PROFILING(9) for details. options MUTEX_PROFILING + +# Profiling for internal hash tables. +options SLEEPQUEUE_PROFILING +options TURNSTILE_PROFILING ##################################################################### # COMPATIBILITY OPTIONS # # Implement system calls compatible with 4.3BSD and older versions of # FreeBSD. You probably do NOT want to remove this as much current code # still relies on the 4.3 emulation. Note that some architectures that # are supported by FreeBSD do not include support for certain important # aspects of this compatibility option, namely those related to the # signal delivery mechanism. # options COMPAT_43 # Enable FreeBSD4 compatibility syscalls options COMPAT_FREEBSD4 # # These three options provide support for System V Interface # Definition-style interprocess communication, in the form of shared # memory, semaphores, and message queues, respectively. # options SYSVSHM options SYSVSEM options SYSVMSG ##################################################################### # DEBUGGING OPTIONS # # Enable the kernel debugger. # options DDB # # Use direct symbol lookup routines for ddb instead of the kernel linker # ones, so that symbols (mostly) work before the kernel linker has been # initialized. This is not the default because it breaks ddb's lookup of # symbols in loaded modules. # #!options DDB_NOKLDSYM # # Print the numerical value of symbols in addition to the symbolic # representation. # options DDB_NUMSYM # # Print a stack trace of the current thread out on the console for a panic. # options DDB_TRACE # # Don't drop into DDB for a panic. Intended for unattended operation # where you may want to drop to DDB from the console, but still want # the machine to recover from a panic # options DDB_UNATTENDED # # If using GDB remote mode to debug the kernel, there's a non-standard # extension to the remote protocol that can be used to use the serial # port as both the debugging port and the system console. It's non- # standard and you're on your own if you enable it. See also the # "remotechat" variables in the FreeBSD specific version of gdb. # options GDB_REMOTE_CHAT # # KTRACE enables the system-call tracing facility ktrace(2). To be more # SMP-friendly, KTRACE uses a worker thread to process most trace events # asynchronously to the thread generating the event. This requires a # pre-allocated store of objects representing trace events. The # KTRACE_REQUEST_POOL option specifies the initial size of this store. # The size of the pool can be adjusted both at boottime and runtime via # the kern.ktrace_request_pool tunable and sysctl. # options KTRACE #kernel tracing options KTRACE_REQUEST_POOL=101 # # KTR is a kernel tracing mechanism imported from BSD/OS. Currently it # has no userland interface aside from a few sysctl's. It is enabled with # the KTR option. KTR_ENTRIES defines the number of entries in the circular # trace buffer. KTR_COMPILE defines the mask of events to compile into the # kernel as defined by the KTR_* constants in . KTR_MASK defines the # initial value of the ktr_mask variable which determines at runtime what # events to trace. KTR_CPUMASK determines which CPU's log events, with # bit X corresponding to cpu X. KTR_VERBOSE enables dumping of KTR events # to the console by default. This functionality can be toggled via the # debug.ktr_verbose sysctl and defaults to off if KTR_VERBOSE is not defined. # options KTR options KTR_ENTRIES=1024 options KTR_COMPILE=(KTR_INTR|KTR_PROC) options KTR_MASK=KTR_INTR options KTR_CPUMASK=0x3 options KTR_VERBOSE # # The INVARIANTS option is used in a number of source files to enable # extra sanity checking of internal structures. This support is not # enabled by default because of the extra time it would take to check # for these conditions, which can only occur as a result of # programming errors. # options INVARIANTS # # The INVARIANT_SUPPORT option makes us compile in support for # verifying some of the internal structures. It is a prerequisite for # 'INVARIANTS', as enabling 'INVARIANTS' will make these functions be # called. The intent is that you can set 'INVARIANTS' for single # source files (by changing the source file or specifying it on the # command line) if you have 'INVARIANT_SUPPORT' enabled. Also, if you # wish to build a kernel module with 'INVARIANTS', then adding # 'INVARIANT_SUPPORT' to your kernel will provide all the necessary # infrastructure without the added overhead. # options INVARIANT_SUPPORT # # The DIAGNOSTIC option is used to enable extra debugging information # from some parts of the kernel. As this makes everything more noisy, # it is disabled by default. # options DIAGNOSTIC # # REGRESSION causes optional kernel interfaces necessary only for regression # testing to be enabled. These interfaces may constitute security risks # when enabled, as they permit processes to easily modify aspects of the # run-time environment to reproduce unlikely or unusual (possibly normally # impossible) scenarios. # options REGRESSION # # RESTARTABLE_PANICS allows one to continue from a panic as if it were # a call to the debugger via the Debugger() function instead. It is only # useful if a kernel debugger is present. To restart from a panic, reset # the panicstr variable to NULL and continue execution. This option is # for development use only and should NOT be used in production systems # to "workaround" a panic. # #options RESTARTABLE_PANICS # # This option let some drivers co-exist that can't co-exist in a running # system. This is used to be able to compile all kernel code in one go for # quality assurance purposes (like this file, which the option takes it name # from.) # options COMPILING_LINT ##################################################################### # NETWORKING OPTIONS # # Protocol families: # Only the INET (Internet) family is officially supported in FreeBSD. # options INET #Internet communications protocols options INET6 #IPv6 communications protocols options IPSEC #IP security options IPSEC_ESP #IP security (crypto; define w/ IPSEC) options IPSEC_DEBUG #debug for IP security # # Set IPSEC_FILTERGIF to force packets coming through a gif tunnel # to be processed by any configured packet filtering (ipfw, ipf). # The default is that packets coming from a tunnel are _not_ processed; # they are assumed trusted. # # IPSEC history is preserved for such packets, and can be filtered # using ipfw(8)'s 'ipsec' keyword, when this option is enabled. # #options IPSEC_FILTERGIF #filter ipsec packets from a tunnel #options FAST_IPSEC #new IPsec (cannot define w/ IPSEC) options IPX #IPX/SPX communications protocols options IPXIP #IPX in IP encapsulation (not available) #options NCP #NetWare Core protocol options NETATALK #Appletalk communications protocols options NETATALKDEBUG #Appletalk debugging # # SMB/CIFS requester # NETSMB enables support for SMB protocol, it requires LIBMCHAIN and LIBICONV # options. # NETSMBCRYPTO enables support for encrypted passwords. options NETSMB #SMB/CIFS requester options NETSMBCRYPTO #encrypted password support for SMB # mchain library. It can be either loaded as KLD or compiled into kernel options LIBMCHAIN # altq(9). Enable the base part of the hooks with the ALTQ option. # Individual disciplines must be built into the base system and can not be # loaded as modules at this point. In order to build a SMP kernel you must # also have the ALTQ_NOPCC option. options ALTQ options ALTQ_CBQ # Class Bases Queueing options ALTQ_RED # Random Early Drop options ALTQ_RIO # RED In/Out options ALTQ_HFSC # Hierarchical Packet Scheduler options ALTQ_CDNR # Traffic conditioner options ALTQ_PRIQ # Priority Queueing options ALTQ_NOPCC # Required for SMP build options ALTQ_DEBUG # netgraph(4). Enable the base netgraph code with the NETGRAPH option. # Individual node types can be enabled with the corresponding option # listed below; however, this is not strictly necessary as netgraph # will automatically load the corresponding KLD module if the node type # is not already compiled into the kernel. Each type below has a # corresponding man page, e.g., ng_async(8). options NETGRAPH #netgraph(4) system options NETGRAPH_ASYNC options NETGRAPH_ATMLLC options NETGRAPH_ATM_ATMPIF options NETGRAPH_BLUETOOTH # ng_bluetooth(4) options NETGRAPH_BLUETOOTH_BT3C # ng_bt3c(4) options NETGRAPH_BLUETOOTH_H4 # ng_h4(4) options NETGRAPH_BLUETOOTH_HCI # ng_hci(4) options NETGRAPH_BLUETOOTH_L2CAP # ng_l2cap(4) options NETGRAPH_BLUETOOTH_SOCKET # ng_btsocket(4) options NETGRAPH_BLUETOOTH_UBT # ng_ubt(4) options NETGRAPH_BLUETOOTH_UBTBCMFW # ubtbcmfw(4) options NETGRAPH_BPF options NETGRAPH_BRIDGE options NETGRAPH_CISCO options NETGRAPH_ECHO options NETGRAPH_EIFACE options NETGRAPH_ETHER options NETGRAPH_FEC options NETGRAPH_FRAME_RELAY options NETGRAPH_GIF options NETGRAPH_GIF_DEMUX options NETGRAPH_HOLE options NETGRAPH_IFACE options NETGRAPH_IP_INPUT options NETGRAPH_KSOCKET options NETGRAPH_L2TP options NETGRAPH_LMI # MPPC compression requires proprietary files (not included) #options NETGRAPH_MPPC_COMPRESSION options NETGRAPH_MPPC_ENCRYPTION options NETGRAPH_ONE2MANY options NETGRAPH_PPP options NETGRAPH_PPPOE options NETGRAPH_PPTPGRE options NETGRAPH_RFC1490 options NETGRAPH_SOCKET options NETGRAPH_SPLIT options NETGRAPH_SPPP options NETGRAPH_TEE options NETGRAPH_TTY options NETGRAPH_UI options NETGRAPH_VJC # NgATM - Netgraph ATM options NGATM_ATM options NGATM_ATMBASE options NGATM_SSCOP options NGATM_SSCFU options NGATM_UNI device mn # Munich32x/Falc54 Nx64kbit/sec cards. device musycc # LMC/SBE LMC1504 quad T1/E1 # # Network interfaces: # The `loop' device is MANDATORY when networking is enabled. # The `ether' device provides generic code to handle # Ethernets; it is MANDATORY when an Ethernet device driver is # configured or token-ring is enabled. # The `wlan' device provides generic code to support 802.11 # drivers, including host AP mode; it is MANDATORY for the wi # driver and will eventually be required by all 802.11 drivers. # The `fddi' device provides generic code to support FDDI. # The `arcnet' device provides generic code to support Arcnet. # The `sppp' device serves a similar role for certain types # of synchronous PPP links (like `cx', `ar'). # The `sl' device implements the Serial Line IP (SLIP) service. # The `ppp' device implements the Point-to-Point Protocol. # The `bpf' device enables the Berkeley Packet Filter. Be # aware of the legal and administrative consequences of enabling this # option. The number of devices determines the maximum number of # simultaneous BPF clients programs runnable. # The `disc' device implements a minimal network interface, # which throws away all packets sent and never receives any. It is # included for testing purposes. This shows up as the `ds' interface. # The `tap' device is a pty-like virtual Ethernet interface # The `tun' device implements (user-)ppp and nos-tun # The `gif' device implements IPv6 over IP4 tunneling, # IPv4 over IPv6 tunneling, IPv4 over IPv4 tunneling and # IPv6 over IPv6 tunneling. # The `gre' device implements two types of IP4 over IP4 tunneling: # GRE and MOBILE, as specified in the RFC1701 and RFC2004. # The XBONEHACK option allows the same pair of addresses to be configured on # multiple gif interfaces. # The `faith' device captures packets sent to it and diverts them # to the IPv4/IPv6 translation daemon. # The `stf' device implements 6to4 encapsulation. # The `ef' device provides support for multiple ethernet frame types # specified via ETHER_* options. See ef(4) for details. # # The pf packet filter consists of three devices: # The `pf' device provides /dev/pf and the firewall code itself. # The `pflog' device provides the pflog0 interface which logs packets. # The `pfsync' device provides the pfsync0 interface used for # synchronization of firewall state tables (over the net). # Requires option PFIL_HOOKS and (when used as a module) option RANDOM_IP_ID # # The PPP_BSDCOMP option enables support for compress(1) style entire # packet compression, the PPP_DEFLATE is for zlib/gzip style compression. # PPP_FILTER enables code for filtering the ppp data stream and selecting # events for resetting the demand dial activity timer - requires bpf. # See pppd(8) for more details. # device ether #Generic Ethernet device vlan #VLAN support device wlan #802.11 support device token #Generic TokenRing device fddi #Generic FDDI device arcnet #Generic Arcnet device sppp #Generic Synchronous PPP device loop #Network loopback device device bpf #Berkeley packet filter device disc #Discard device (ds0, ds1, etc) device tap #Virtual Ethernet driver device tun #Tunnel driver (ppp(8), nos-tun(8)) device sl #Serial Line IP device gre #IP over IP tunneling device pf #PF OpenBSD packet-filter firewall device pflog #logging support interface for PF device pfsync #synchronization interface for PF device ppp #Point-to-point protocol options PPP_BSDCOMP #PPP BSD-compress support options PPP_DEFLATE #PPP zlib/deflate/gzip support options PPP_FILTER #enable bpf filtering (needs bpf) device ef # Multiple ethernet frames support options ETHER_II # enable Ethernet_II frame options ETHER_8023 # enable Ethernet_802.3 (Novell) frame options ETHER_8022 # enable Ethernet_802.2 frame options ETHER_SNAP # enable Ethernet_802.2/SNAP frame # for IPv6 device gif #IPv6 and IPv4 tunneling options XBONEHACK device faith #for IPv6 and IPv4 translation device stf #6to4 IPv6 over IPv4 encapsulation # # Internet family options: # # MROUTING enables the kernel multicast packet forwarder, which works # with mrouted(8). # # PIM enables Protocol Independent Multicast in the kernel. # Requires MROUTING enabled. # # IPFIREWALL enables support for IP firewall construction, in # conjunction with the `ipfw' program. IPFIREWALL_VERBOSE sends # logged packets to the system logger. IPFIREWALL_VERBOSE_LIMIT # limits the number of times a matching entry can be logged. # # WARNING: IPFIREWALL defaults to a policy of "deny ip from any to any" # and if you do not add other rules during startup to allow access, # YOU WILL LOCK YOURSELF OUT. It is suggested that you set firewall_type=open # in /etc/rc.conf when first enabling this feature, then refining the # firewall rules in /etc/rc.firewall after you've tested that the new kernel # feature works properly. # # IPFIREWALL_DEFAULT_TO_ACCEPT causes the default rule (at boot) to # allow everything. Use with care, if a cracker can crash your # firewall machine, they can get to your protected machines. However, # if you are using it as an as-needed filter for specific problems as # they arise, then this may be for you. Changing the default to 'allow' # means that you won't get stuck if the kernel and /sbin/ipfw binary get # out of sync. # # IPDIVERT enables the divert IP sockets, used by ``ipfw divert'' # # IPSTEALTH enables code to support stealth forwarding (i.e., forwarding # packets without touching the ttl). This can be useful to hide firewalls # from traceroute and similar tools. # # PFIL_HOOKS enables an abstraction layer which is meant to be used in # network code where filtering is required. See pfil(9). This option is # required by the IPFILTER option and the PF device. # # TCPDEBUG enables code which keeps traces of the TCP state machine # for sockets with the SO_DEBUG option set, which can then be examined # using the trpt(8) utility. # options MROUTING # Multicast routing options PIM # Protocol Independent Multicast options IPFIREWALL #firewall options IPFIREWALL_VERBOSE #enable logging to syslogd(8) options IPFIREWALL_VERBOSE_LIMIT=100 #limit verbosity options IPFIREWALL_DEFAULT_TO_ACCEPT #allow everything by default options IPV6FIREWALL #firewall for IPv6 options IPV6FIREWALL_VERBOSE options IPV6FIREWALL_VERBOSE_LIMIT=100 options IPV6FIREWALL_DEFAULT_TO_ACCEPT options IPDIVERT #divert sockets options IPFILTER #ipfilter support options IPFILTER_LOG #ipfilter logging options IPFILTER_DEFAULT_BLOCK #block all packets by default options IPSTEALTH #support for stealth forwarding options PFIL_HOOKS #required by IPFILTER options TCPDEBUG # The MBUF_STRESS_TEST option enables options which create # various random failures / extreme cases related to mbuf # functions. See mbuf(9) for a list of available test cases. options MBUF_STRESS_TEST # RANDOM_IP_ID causes the ID field in IP packets to be randomized # instead of incremented by 1 with each packet generated. This # option closes a minor information leak which allows remote # observers to determine the rate of packet generation on the # machine by watching the counter. options RANDOM_IP_ID # Statically Link in accept filters options ACCEPT_FILTER_DATA options ACCEPT_FILTER_HTTP # TCP_DROP_SYNFIN adds support for ignoring TCP packets with SYN+FIN. This # prevents nmap et al. from identifying the TCP/IP stack, but breaks support # for RFC1644 extensions and is not recommended for web servers. # options TCP_DROP_SYNFIN #drop TCP packets with SYN+FIN # TCP_SIGNATURE adds support for RFC 2385 (TCP-MD5) digests. These are # carried in TCP option 19. This option is commonly used to protect # TCP sessions (e.g. BGP) where IPSEC is not available nor desirable. # This is enabled on a per-socket basis using the TCP_MD5SIG socket option. # This requires the use of 'device crypto', 'options FAST_IPSEC', and # 'device cryptodev' as it depends on the non-KAME IPSEC SADB code. #options TCP_SIGNATURE #include support for RFC 2385 # DUMMYNET enables the "dummynet" bandwidth limiter. You need IPFIREWALL # as well. See dummynet(4) and ipfw(8) for more info. When you run # DUMMYNET it is advisable to also have "options HZ=1000" to achieve a # smoother scheduling of the traffic. # # BRIDGE enables bridging between ethernet cards -- see bridge(4). # You can use IPFIREWALL and DUMMYNET together with bridging. # options DUMMYNET options BRIDGE # Zero copy sockets support. This enables "zero copy" for sending and # receiving data via a socket. The send side works for any type of NIC, # the receive side only works for NICs that support MTUs greater than the # page size of your architecture and that support header splitting. See # zero_copy(9) for more details. options ZERO_COPY_SOCKETS # # ATM (HARP version) options # # ATM_CORE includes the base ATM functionality code. This must be included # for ATM support. # # ATM_IP includes support for running IP over ATM. # # At least one (and usually only one) of the following signalling managers # must be included (note that all signalling managers include PVC support): # ATM_SIGPVC includes support for the PVC-only signalling manager `sigpvc'. # ATM_SPANS includes support for the `spans' signalling manager, which runs # the FORE Systems's proprietary SPANS signalling protocol. # ATM_UNI includes support for the `uni30' and `uni31' signalling managers, # which run the ATM Forum UNI 3.x signalling protocols. # # The `hfa' driver provides support for the FORE Systems, Inc. # PCA-200E ATM PCI Adapter. # # The `harp' pseudo-driver makes all NATM interface drivers available to HARP. # options ATM_CORE #core ATM protocol family options ATM_IP #IP over ATM support options ATM_SIGPVC #SIGPVC signalling manager options ATM_SPANS #SPANS signalling manager options ATM_UNI #UNI signalling manager device hfa #FORE PCA-200E ATM PCI device harp #Pseudo-interface for NATM ##################################################################### # FILESYSTEM OPTIONS # # Only the root, /usr, and /tmp filesystems need be statically # compiled; everything else will be automatically loaded at mount # time. (Exception: the UFS family--- FFS --- cannot # currently be demand-loaded.) Some people still prefer to statically # compile other filesystems as well. # # NB: The NULL, PORTAL, UMAP and UNION filesystems are known to be # buggy, and WILL panic your system if you attempt to do anything with # them. They are included here as an incentive for some enterprising # soul to sit down and fix them. # # One of these is mandatory: options FFS #Fast filesystem options NFSCLIENT #Network File System client # The rest are optional: options CD9660 #ISO 9660 filesystem options FDESCFS #File descriptor filesystem options HPFS #OS/2 File system options MSDOSFS #MS DOS File System (FAT, FAT32) options NFSSERVER #Network File System server options NTFS #NT File System options NULLFS #NULL filesystem # Broken (depends on NCP): #options NWFS #NetWare filesystem options PORTALFS #Portal filesystem options PROCFS #Process filesystem (requires PSEUDOFS) options PSEUDOFS #Pseudo-filesystem framework options SMBFS #SMB/CIFS filesystem options UDF #Universal Disk Format # Broken (seriously (functionally) broken): #options UMAPFS #UID map filesystem options UNIONFS #Union filesystem # The xFS_ROOT options REQUIRE the associated ``options xFS'' options NFS_ROOT #NFS usable as root device # Soft updates is a technique for improving filesystem speed and # making abrupt shutdown less risky. # options SOFTUPDATES # Extended attributes allow additional data to be associated with files, # and is used for ACLs, Capabilities, and MAC labels. # See src/sys/ufs/ufs/README.extattr for more information. options UFS_EXTATTR options UFS_EXTATTR_AUTOSTART # Access Control List support for UFS filesystems. The current ACL # implementation requires extended attribute support, UFS_EXTATTR, # for the underlying filesystem. # See src/sys/ufs/ufs/README.acls for more information. options UFS_ACL # Directory hashing improves the speed of operations on very large # directories at the expense of some memory. options UFS_DIRHASH # Make space in the kernel for a root filesystem on a md device. # Define to the number of kilobytes to reserve for the filesystem. options MD_ROOT_SIZE=10 # Make the md device a potential root device, either with preloaded # images of type mfs_root or md_root. options MD_ROOT # Disk quotas are supported when this option is enabled. options QUOTA #enable disk quotas # If you are running a machine just as a fileserver for PC and MAC # users, using SAMBA or Netatalk, you may consider setting this option # and keeping all those users' directories on a filesystem that is # mounted with the suiddir option. This gives new files the same # ownership as the directory (similar to group). It's a security hole # if you let these users run programs, so confine it to file-servers # (but it'll save you lots of headaches in those cases). Root owned # directories are exempt and X bits are cleared. The suid bit must be # set on the directory as well; see chmod(1) PC owners can't see/set # ownerships so they keep getting their toes trodden on. This saves # you all the support calls as the filesystem it's used on will act as # they expect: "It's my dir so it must be my file". # options SUIDDIR # NFS options: options NFS_MINATTRTIMO=3 # VREG attrib cache timeout in sec options NFS_MAXATTRTIMO=60 options NFS_MINDIRATTRTIMO=30 # VDIR attrib cache timeout in sec options NFS_MAXDIRATTRTIMO=60 options NFS_GATHERDELAY=10 # Default write gather delay (msec) options NFS_WDELAYHASHSIZ=16 # and with this options NFS_DEBUG # Enable NFS Debugging # Coda stuff: options CODA #CODA filesystem. device vcoda 4 #coda minicache <-> venus comm. # Use the old Coda 5.x venus<->kernel interface instead of the new # realms-aware 6.x protocol. #options CODA_COMPAT_5 # # Add support for the EXT2FS filesystem of Linux fame. Be a bit # careful with this - the ext2fs code has a tendency to lag behind # changes and not be exercised very much, so mounting read/write could # be dangerous (and even mounting read only could result in panics.) # options EXT2FS # Use real implementations of the aio_* system calls. There are numerous # stability and security issues in the current aio code that make it # unsuitable for inclusion on machines with untrusted local users. options VFS_AIO # Cryptographically secure random number generator; /dev/[u]random device random # Optional character code conversion support with LIBICONV. # Each option requires their base file system and LIBICONV. options CD9660_ICONV options MSDOSFS_ICONV options NTFS_ICONV options UDF_ICONV ##################################################################### # POSIX P1003.1B # Real time extensions added in the 1993 Posix # _KPOSIX_PRIORITY_SCHEDULING: Build in _POSIX_PRIORITY_SCHEDULING options _KPOSIX_PRIORITY_SCHEDULING # p1003_1b_semaphores are very experimental, # user should be ready to assist in debugging if problems arise. options P1003_1B_SEMAPHORES ##################################################################### # SECURITY POLICY PARAMETERS # Support for Mandatory Access Control (MAC): options MAC options MAC_BIBA options MAC_BSDEXTENDED options MAC_DEBUG options MAC_IFOFF options MAC_LOMAC options MAC_MLS options MAC_NONE options MAC_PARTITION options MAC_PORTACL options MAC_SEEOTHERUIDS options MAC_STUB options MAC_TEST ##################################################################### # CLOCK OPTIONS # The granularity of operation is controlled by the kernel option HZ whose # default value (100) means a granularity of 10ms (1s/HZ). # Some subsystems, such as DUMMYNET, might benefit from a smaller # granularity such as 1ms or less, for a smoother scheduling of packets. # Consider, however, that reducing the granularity too much might # cause excessive overhead in clock interrupt processing, # potentially causing ticks to be missed and thus actually reducing # the accuracy of operation. options HZ=100 # Enable support for the kernel PLL to use an external PPS signal, # under supervision of [x]ntpd(8) # More info in ntpd documentation: http://www.eecis.udel.edu/~ntp options PPS_SYNC ##################################################################### # SCSI DEVICES # SCSI DEVICE CONFIGURATION # The SCSI subsystem consists of the `base' SCSI code, a number of # high-level SCSI device `type' drivers, and the low-level host-adapter # device drivers. The host adapters are listed in the ISA and PCI # device configuration sections below. # # It is possible to wire down your SCSI devices so that a given bus, # target, and LUN always come on line as the same device unit. In # earlier versions the unit numbers were assigned in the order that # the devices were probed on the SCSI bus. This means that if you # removed a disk drive, you may have had to rewrite your /etc/fstab # file, and also that you had to be careful when adding a new disk # as it may have been probed earlier and moved your device configuration # around. (See also option GEOM_VOL for a different solution to this # problem.) # This old behavior is maintained as the default behavior. The unit # assignment begins with the first non-wired down unit for a device # type. For example, if you wire a disk as "da3" then the first # non-wired disk will be assigned da4. # The syntax for wiring down devices is: hint.scbus.0.at="ahc0" hint.scbus.1.at="ahc1" hint.scbus.1.bus="0" hint.scbus.3.at="ahc2" hint.scbus.3.bus="0" hint.scbus.2.at="ahc2" hint.scbus.2.bus="1" hint.da.0.at="scbus0" hint.da.0.target="0" hint.da.0.unit="0" hint.da.1.at="scbus3" hint.da.1.target="1" hint.da.2.at="scbus2" hint.da.2.target="3" hint.sa.1.at="scbus1" hint.sa.1.target="6" # "units" (SCSI logical unit number) that are not specified are # treated as if specified as LUN 0. # All SCSI devices allocate as many units as are required. # The ch driver drives SCSI Media Changer ("jukebox") devices. # # The da driver drives SCSI Direct Access ("disk") and Optical Media # ("WORM") devices. # # The sa driver drives SCSI Sequential Access ("tape") devices. # # The cd driver drives SCSI Read Only Direct Access ("cd") devices. # # The ses driver drives SCSI Environment Services ("ses") and # SAF-TE ("SCSI Accessible Fault-Tolerant Enclosure") devices. # # The pt driver drives SCSI Processor devices. # # # Target Mode support is provided here but also requires that a SIM # (SCSI Host Adapter Driver) provide support as well. # # The targ driver provides target mode support as a Processor type device. # It exists to give the minimal context necessary to respond to Inquiry # commands. There is a sample user application that shows how the rest # of the command support might be done in /usr/share/examples/scsi_target. # # The targbh driver provides target mode support and exists to respond # to incoming commands that do not otherwise have a logical unit assigned # to them. # # The "unknown" device (uk? in pre-2.0.5) is now part of the base SCSI # configuration as the "pass" driver. device scbus #base SCSI code device ch #SCSI media changers device da #SCSI direct access devices (aka disks) device sa #SCSI tapes device cd #SCSI CD-ROMs device ses #SCSI Environmental Services (and SAF-TE) device pt #SCSI processor device targ #SCSI Target Mode Code device targbh #SCSI Target Mode Blackhole Device device pass #CAM passthrough driver # CAM OPTIONS: # debugging options: # -- NOTE -- If you specify one of the bus/target/lun options, you must # specify them all! # CAMDEBUG: When defined enables debugging macros # CAM_DEBUG_BUS: Debug the given bus. Use -1 to debug all busses. # CAM_DEBUG_TARGET: Debug the given target. Use -1 to debug all targets. # CAM_DEBUG_LUN: Debug the given lun. Use -1 to debug all luns. # CAM_DEBUG_FLAGS: OR together CAM_DEBUG_INFO, CAM_DEBUG_TRACE, # CAM_DEBUG_SUBTRACE, and CAM_DEBUG_CDB # # CAM_MAX_HIGHPOWER: Maximum number of concurrent high power (start unit) cmds # CAM_NEW_TRAN_CODE: this is the new transport layer code that will be switched # to soon # SCSI_NO_SENSE_STRINGS: When defined disables sense descriptions # SCSI_NO_OP_STRINGS: When defined disables opcode descriptions # SCSI_DELAY: The number of MILLISECONDS to freeze the SIM (scsi adapter) # queue after a bus reset, and the number of milliseconds to # freeze the device queue after a bus device reset. This # can be changed at boot and runtime with the # kern.cam.scsi_delay tunable/sysctl. options CAMDEBUG options CAM_DEBUG_BUS=-1 options CAM_DEBUG_TARGET=-1 options CAM_DEBUG_LUN=-1 options CAM_DEBUG_FLAGS=(CAM_DEBUG_INFO|CAM_DEBUG_TRACE|CAM_DEBUG_CDB) options CAM_MAX_HIGHPOWER=4 options SCSI_NO_SENSE_STRINGS options SCSI_NO_OP_STRINGS options SCSI_DELAY=8000 # Be pessimistic about Joe SCSI device # Options for the CAM CDROM driver: # CHANGER_MIN_BUSY_SECONDS: Guaranteed minimum time quantum for a changer LUN # CHANGER_MAX_BUSY_SECONDS: Maximum time quantum per changer LUN, only # enforced if there is I/O waiting for another LUN # The compiled in defaults for these variables are 2 and 10 seconds, # respectively. # # These can also be changed on the fly with the following sysctl variables: # kern.cam.cd.changer.min_busy_seconds # kern.cam.cd.changer.max_busy_seconds # options CHANGER_MIN_BUSY_SECONDS=2 options CHANGER_MAX_BUSY_SECONDS=10 # Options for the CAM sequential access driver: # SA_IO_TIMEOUT: Timeout for read/write/wfm operations, in minutes # SA_SPACE_TIMEOUT: Timeout for space operations, in minutes # SA_REWIND_TIMEOUT: Timeout for rewind operations, in minutes # SA_ERASE_TIMEOUT: Timeout for erase operations, in minutes # SA_1FM_AT_EOD: Default to model which only has a default one filemark at EOT. options SA_IO_TIMEOUT=4 options SA_SPACE_TIMEOUT=60 options SA_REWIND_TIMEOUT=(2*60) options SA_ERASE_TIMEOUT=(4*60) options SA_1FM_AT_EOD # Optional timeout for the CAM processor target (pt) device # This is specified in seconds. The default is 60 seconds. options SCSI_PT_DEFAULT_TIMEOUT=60 # Optional enable of doing SES passthrough on other devices (e.g., disks) # # Normally disabled because a lot of newer SCSI disks report themselves # as having SES capabilities, but this can then clot up attempts to build # build a topology with the SES device that's on the box these drives # are in.... options SES_ENABLE_PASSTHROUGH ##################################################################### # MISCELLANEOUS DEVICES AND OPTIONS # The `pty' device usually turns out to be ``effectively mandatory'', # as it is required for `telnetd', `rlogind', `screen', `emacs', and # `xterm', among others. device pty #Pseudo ttys device nmdm #back-to-back tty devices device md #Memory/malloc disk device snp #Snoop device - to look at pty/vty/etc.. device ccd #Concatenated disk driver # Configuring Vinum into the kernel is not necessary, since the kld # module gets started automatically when vinum(8) starts. This # device is also untested. Use at your own risk. # # The option VINUMDEBUG must match the value set in CFLAGS # in src/sbin/vinum/Makefile. Failure to do so will result in # the following message from vinum(8): # # Can't get vinum config: Invalid argument # # see vinum(4) for more reasons not to use these options. device vinum #Vinum concat/mirror/raid driver options VINUMDEBUG #enable Vinum debugging hooks # Kernel side iconv library options LIBICONV # Size of the kernel message buffer. Should be N * pagesize. options MSGBUF_SIZE=40960 # Maximum size of a tty or pty input buffer. options TTYHOG=8193 ##################################################################### # HARDWARE DEVICE CONFIGURATION # For ISA the required hints are listed. # EISA, MCA, PCI and pccard are self identifying buses, so no hints # are needed. # # Mandatory devices: # # The keyboard controller; it controls the keyboard and the PS/2 mouse. device atkbdc hint.atkbdc.0.at="isa" hint.atkbdc.0.port="0x060" # The AT keyboard device atkbd hint.atkbd.0.at="atkbdc" hint.atkbd.0.irq="1" # Options for atkbd: options ATKBD_DFLT_KEYMAP # specify the built-in keymap makeoptions ATKBD_DFLT_KEYMAP=jp.106 # These options are valid for other keyboard drivers as well. options KBD_DISABLE_KEYMAP_LOAD # refuse to load a keymap options KBD_INSTALL_CDEV # install a CDEV entry in /dev # `flags' for atkbd: # 0x01 Force detection of keyboard, else we always assume a keyboard # 0x02 Don't reset keyboard, useful for some newer ThinkPads # 0x03 Force detection and avoid reset, might help with certain # dockingstations # 0x04 Old-style (XT) keyboard support, useful for older ThinkPads # PS/2 mouse device psm hint.psm.0.at="atkbdc" hint.psm.0.irq="12" # Options for psm: options PSM_HOOKRESUME #hook the system resume event, useful #for some laptops options PSM_RESETAFTERSUSPEND #reset the device at the resume event # Video card driver for VGA adapters. device vga hint.vga.0.at="isa" # Options for vga: # Try the following option if the mouse pointer is not drawn correctly # or font does not seem to be loaded properly. May cause flicker on # some systems. options VGA_ALT_SEQACCESS # If you can dispense with some vga driver features, you may want to # use the following options to save some memory. #options VGA_NO_FONT_LOADING # don't save/load font #options VGA_NO_MODE_CHANGE # don't change video modes # Older video cards may require this option for proper operation. options VGA_SLOW_IOACCESS # do byte-wide i/o's to TS and GDC regs # The following option probably won't work with the LCD displays. options VGA_WIDTH90 # support 90 column modes options FB_DEBUG # Frame buffer debugging device splash # Splash screen and screen saver support # Various screen savers. device blank_saver device daemon_saver device fade_saver device fire_saver device green_saver device logo_saver device rain_saver device star_saver device warp_saver # The syscons console driver (sco color console compatible). device sc hint.sc.0.at="isa" options MAXCONS=16 # number of virtual consoles options SC_ALT_MOUSE_IMAGE # simplified mouse cursor in text mode options SC_DFLT_FONT # compile font in makeoptions SC_DFLT_FONT=cp850 options SC_DISABLE_DDBKEY # disable `debug' key options SC_DISABLE_REBOOT # disable reboot key sequence options SC_HISTORY_SIZE=200 # number of history buffer lines options SC_MOUSE_CHAR=0x3 # char code for text mode mouse cursor options SC_PIXEL_MODE # add support for the raster text mode # The following options will let you change the default colors of syscons. options SC_NORM_ATTR=(FG_GREEN|BG_BLACK) options SC_NORM_REV_ATTR=(FG_YELLOW|BG_GREEN) options SC_KERNEL_CONS_ATTR=(FG_RED|BG_BLACK) options SC_KERNEL_CONS_REV_ATTR=(FG_BLACK|BG_RED) # The following options will let you change the default behaviour of # cut-n-paste feature options SC_CUT_SPACES2TABS # convert leading spaces into tabs options SC_CUT_SEPCHARS=\"x09\" # set of characters that delimit words # (default is single space - \"x20\") # If you have a two button mouse, you may want to add the following option # to use the right button of the mouse to paste text. options SC_TWOBUTTON_MOUSE # You can selectively disable features in syscons. options SC_NO_CUTPASTE options SC_NO_FONT_LOADING options SC_NO_HISTORY options SC_NO_SYSMOUSE options SC_NO_SUSPEND_VTYSWITCH # `flags' for sc # 0x80 Put the video card in the VESA 800x600 dots, 16 color mode # 0x100 Probe for a keyboard device periodically if one is not present # # Optional devices: # # # SCSI host adapters: # # adv: All Narrow SCSI bus AdvanSys controllers. # adw: Second Generation AdvanSys controllers including the ADV940UW. # aha: Adaptec 154x/1535/1640 # ahb: Adaptec 174x EISA controllers # ahc: Adaptec 274x/284x/2910/293x/294x/394x/3950x/3960x/398X/4944/ # 19160x/29160x, aic7770/aic78xx # ahd: Adaptec 29320/39320 Controllers. # aic: Adaptec 6260/6360, APA-1460 (PC Card), NEC PC9801-100 (C-BUS) # amd: Support for the AMD 53C974 SCSI host adapter chip as found on devices # such as the Tekram DC-390(T). # bt: Most Buslogic controllers: including BT-445, BT-54x, BT-64x, BT-74x, # BT-75x, BT-946, BT-948, BT-956, BT-958, SDC3211B, SDC3211F, SDC3222F # esp: NCR53c9x. Only for SBUS hardware right now. # isp: Qlogic ISP 1020, 1040 and 1040B PCI SCSI host adapters, # ISP 1240 Dual Ultra SCSI, ISP 1080 and 1280 (Dual) Ultra2, # ISP 12160 Ultra3 SCSI, # Qlogic ISP 2100 and ISP 2200 1Gb Fibre Channel host adapters. # Qlogic ISP 2300 and ISP 2312 2Gb Fibre Channel host adapters. # ispfw: Firmware module for Qlogic host adapters # mpt: LSI-Logic MPT/Fusion 53c1020 or 53c1030 Ultra4 # or FC9x9 Fibre Channel host adapters. # ncr: NCR 53C810, 53C825 self-contained SCSI host adapters. # sym: Symbios/Logic 53C8XX family of PCI-SCSI I/O processors: # 53C810, 53C810A, 53C815, 53C825, 53C825A, 53C860, 53C875, # 53C876, 53C885, 53C895, 53C895A, 53C896, 53C897, 53C1510D, # 53C1010-33, 53C1010-66. # trm: Tekram DC395U/UW/F DC315U adapters. # wds: WD7000 # # Note that the order is important in order for Buslogic ISA/EISA cards to be # probed correctly. # device bt hint.bt.0.at="isa" hint.bt.0.port="0x330" device adv hint.adv.0.at="isa" device adw device aha hint.aha.0.at="isa" device aic hint.aic.0.at="isa" device ahb device ahc device ahd device amd device esp device isp hint.isp.0.disable="1" hint.isp.0.role="3" hint.isp.0.prefer_iomap="1" hint.isp.0.prefer_memmap="1" hint.isp.0.fwload_disable="1" hint.isp.0.ignore_nvram="1" hint.isp.0.fullduplex="1" hint.isp.0.topology="lport" hint.isp.0.topology="nport" hint.isp.0.topology="lport-only" hint.isp.0.topology="nport-only" # we can't get u_int64_t types, nor can we get strings if it's got # a leading 0x, hence this silly dodge. hint.isp.0.portwnn="w50000000aaaa0000" hint.isp.0.nodewnn="w50000000aaaa0001" device ispfw device mpt device ncr device sym device trm device wds hint.wds.0.at="isa" hint.wds.0.port="0x350" hint.wds.0.irq="11" hint.wds.0.drq="6" # The aic7xxx driver will attempt to use memory mapped I/O for all PCI # controllers that have it configured only if this option is set. Unfortunately, # this doesn't work on some motherboards, which prevents it from being the # default. options AHC_ALLOW_MEMIO # Dump the contents of the ahc controller configuration PROM. options AHC_DUMP_EEPROM # Bitmap of units to enable targetmode operations. options AHC_TMODE_ENABLE # Compile in Aic7xxx Debugging code. options AHC_DEBUG # Aic7xxx driver debugging options. See sys/dev/aic7xxx/aic7xxx.h options AHC_DEBUG_OPTS # Print register bitfields in debug output. Adds ~128k to driver # See ahc(4). options AHC_REG_PRETTY_PRINT # Compile in aic79xx debugging code. options AHD_DEBUG # Aic79xx driver debugging options. Adds ~215k to driver. See ahd(4). options AHD_DEBUG_OPTS=0xFFFFFFFF # Print human-readable register definitions when debugging options AHD_REG_PRETTY_PRINT # Bitmap of units to enable targetmode operations. options AHD_TMODE_ENABLE # The adw driver will attempt to use memory mapped I/O for all PCI # controllers that have it configured only if this option is set. options ADW_ALLOW_MEMIO # Options used in dev/isp/ (Qlogic SCSI/FC driver). # # ISP_TARGET_MODE - enable target mode operation # options ISP_TARGET_MODE=1 # Options used in dev/sym/ (Symbios SCSI driver). #options SYM_SETUP_LP_PROBE_MAP #-Low Priority Probe Map (bits) # Allows the ncr to take precedence # 1 (1<<0) -> 810a, 860 # 2 (1<<1) -> 825a, 875, 885, 895 # 4 (1<<2) -> 895a, 896, 1510d #options SYM_SETUP_SCSI_DIFF #-HVD support for 825a, 875, 885 # disabled:0 (default), enabled:1 #options SYM_SETUP_PCI_PARITY #-PCI parity checking # disabled:0, enabled:1 (default) #options SYM_SETUP_MAX_LUN #-Number of LUNs supported # default:8, range:[1..64] # The 'asr' driver provides support for current DPT/Adaptec SCSI RAID # controllers (SmartRAID V and VI and later). # These controllers require the CAM infrastructure. # device asr # The 'dpt' driver provides support for old DPT controllers (http://www.dpt.com/). # These have hardware RAID-{0,1,5} support, and do multi-initiator I/O. # The DPT controllers are commonly re-licensed under other brand-names - # some controllers by Olivetti, Dec, HP, AT&T, SNI, AST, Alphatronic, NEC and # Compaq are actually DPT controllers. # # See src/sys/dev/dpt for debugging and other subtle options. # DPT_MEASURE_PERFORMANCE Enables a set of (semi)invasive metrics. Various # instruments are enabled. The tools in # /usr/sbin/dpt_* assume these to be enabled. # DPT_HANDLE_TIMEOUTS Normally device timeouts are handled by the DPT. # If you ant the driver to handle timeouts, enable # this option. If your system is very busy, this # option will create more trouble than solve. # DPT_TIMEOUT_FACTOR Used to compute the excessive amount of time to # wait when timing out with the above option. # DPT_DEBUG_xxxx These are controllable from sys/dev/dpt/dpt.h # DPT_LOST_IRQ When enabled, will try, once per second, to catch # any interrupt that got lost. Seems to help in some # DPT-firmware/Motherboard combinations. Minimal # cost, great benefit. # DPT_RESET_HBA Make "reset" actually reset the controller # instead of fudging it. Only enable this if you # are 100% certain you need it. device dpt # DPT options #!CAM# options DPT_MEASURE_PERFORMANCE #!CAM# options DPT_HANDLE_TIMEOUTS options DPT_TIMEOUT_FACTOR=4 options DPT_LOST_IRQ options DPT_RESET_HBA # # Compaq "CISS" RAID controllers (SmartRAID 5* series) # These controllers have a SCSI-like interface, and require the # CAM infrastructure. # device ciss # # Intel Integrated RAID controllers. # This driver was developed and is maintained by Intel. Contacts # at Intel for this driver are # "Kannanthanam, Boji T" and # "Leubner, Achim" . # device iir # # Mylex AcceleRAID and eXtremeRAID controllers with v6 and later # firmware. These controllers have a SCSI-like interface, and require # the CAM infrastructure. # device mly # # Compaq Smart RAID, Mylex DAC960 and AMI MegaRAID controllers. Only # one entry is needed; the code will find and configure all supported # controllers. # device ida # Compaq Smart RAID device mlx # Mylex DAC960 device amr # AMI MegaRAID # # 3ware ATA RAID # device twe # 3ware ATA RAID # # The 'ATA' driver supports all ATA and ATAPI devices, including PC Card # devices. You only need one "device ata" for it to find all # PCI and PC Card ATA/ATAPI devices on modern machines. device ata device atadisk # ATA disk drives device ataraid # ATA RAID drives device atapicd # ATAPI CDROM drives device atapifd # ATAPI floppy drives device atapist # ATAPI tape drives device atapicam # emulate ATAPI devices as SCSI ditto via CAM # needs CAM to be present (scbus & pass) # # For older non-PCI, non-PnPBIOS systems, these are the hints lines to add: hint.ata.0.at="isa" hint.ata.0.port="0x1f0" hint.ata.0.irq="14" hint.ata.1.at="isa" hint.ata.1.port="0x170" hint.ata.1.irq="15" # # The following options are valid on the ATA driver: # # ATA_STATIC_ID: controller numbering is static ie depends on location # else the device numbers are dynamically allocated. options ATA_STATIC_ID # # Standard floppy disk controllers and floppy tapes, supports # the Y-E DATA External FDD (PC Card) # device fdc hint.fdc.0.at="isa" hint.fdc.0.port="0x3F0" hint.fdc.0.irq="6" hint.fdc.0.drq="2" # # FDC_DEBUG enables floppy debugging. Since the debug output is huge, you # gotta turn it actually on by setting the variable fd_debug with DDB, # however. options FDC_DEBUG # # Activate this line if you happen to have an Insight floppy tape. # Probing them proved to be dangerous for people with floppy disks only, # so it's "hidden" behind a flag: #hint.fdc.0.flags="1" # Specify floppy devices hint.fd.0.at="fdc0" hint.fd.0.drive="0" hint.fd.1.at="fdc0" hint.fd.1.drive="1" # # sio: serial ports (see sio(4)), including support for various # PC Card devices, such as Modem and NICs (see etc/defaults/pccard.conf) # device sio hint.sio.0.at="isa" hint.sio.0.port="0x3F8" hint.sio.0.flags="0x10" hint.sio.0.irq="4" # Options for sio: options COM_ESP # Code for Hayes ESP. options COM_MULTIPORT # Code for some cards with shared IRQs. options CONSPEED=115200 # Speed for serial console # (default 9600). # `flags' specific to sio(4). See below for flags used by both sio(4) and # uart(4). # 0x20 force this unit to be the console (unless there is another # higher priority console). This replaces the COMCONSOLE option. # 0x40 reserve this unit for low level console operations. Do not # access the device in any normal way. # PnP `flags' # 0x1 disable probing of this device. Used to prevent your modem # from being attached as a PnP modem. # Other flags for sio that aren't documented in the man page. # 0x20000 enable hardware RTS/CTS and larger FIFOs. Only works for # ST16650A-compatible UARTs. # # uart: newbusified driver for serial interfaces. It consolidates the sio(4), # sab(4) and zs(4) drivers. # device uart # Options for uart(4) options UART_PPS_ON_CTS # Do time pulse capturing using CTS # instead of DCD. # The following hint should only be used for pure ISA devices. It is not # needed otherwise. Use of hints is strongly discouraged. hint.uart.0.at="isa" # The following 3 hints are used when the UART is a system device (i.e., a # console or debug port), but only on platforms that don't have any other # means to pass the information to the kernel. The unit number of the hint # is only used to bundle the hints together. There is no relation to the # unit number of the probed UART. hint.uart.0.port="0x3f8" hint.uart.0.flags="0x10" hint.uart.0.baud="115200" # `flags' for serial drivers that support consoles like sio(4) and uart(4): # 0x10 enable console support for this unit. Other console flags # (if applicable) are ignored unless this is set. Enabling # console support does not make the unit the preferred console. # Boot with -h or set boot_serial=YES in the loader. For sio(4) # specifically, the 0x20 flag can also be set (see above). # Currently, at most one unit can have console support; the # first one (in config file order) with this flag set is # preferred. Setting this flag for sio0 gives the old behaviour. # 0x80 use this port for serial line gdb support in ddb. Also known # as debug port. # # Options for serial drivers that support consoles: options BREAK_TO_DEBUGGER # A BREAK on a serial console goes to # ddb, if available. # Solaris implements a new BREAK which is initiated by a character # sequence CR ~ ^b which is similar to a familiar pattern used on # Sun servers by the Remote Console. options ALT_BREAK_TO_DEBUGGER # PCI Universal Communications driver # Supports various single and multi port PCI serial cards. Maybe later # also the parallel ports on combination serial/parallel cards. New cards # can be added in src/sys/dev/puc/pucdata.c. # # If the PUC_FASTINTR option is used the driver will try to use fast # interrupts. The card must then be the only user of that interrupt. # Interrupts cannot be shared when using PUC_FASTINTR. device puc options PUC_FASTINTR # # Network interfaces: # # MII bus support is required for some PCI 10/100 ethernet NICs, # namely those which use MII-compliant transceivers or implement # transceiver control interfaces that operate like an MII. Adding # "device miibus0" to the kernel config pulls in support for # the generic miibus API and all of the PHY drivers, including a # generic one for PHYs that aren't specifically handled by an # individual driver. device miibus # an: Aironet 4500/4800 802.11 wireless adapters. Supports the PCMCIA, # PCI and ISA varieties. # awi: Support for IEEE 802.11 PC Card devices using the AMD Am79C930 and # Harris (Intersil) Chipset with PCnetMobile firmware by AMD. # bge: Support for gigabit ethernet adapters based on the Broadcom # BCM570x family of controllers, including the 3Com 3c996-T, # the Netgear GA302T, the SysKonnect SK-9D21 and SK-9D41, and # the embedded gigE NICs on Dell PowerEdge 2550 servers. # cm: Arcnet SMC COM90c26 / SMC COM90c56 # (and SMC COM90c66 in '56 compatibility mode) adapters. # cnw: Xircom CNW/Netware Airsurfer PC Card adapter # cs: IBM Etherjet and other Crystal Semi CS89x0-based adapters # dc: Support for PCI fast ethernet adapters based on the DEC/Intel 21143 # and various workalikes including: # the ADMtek AL981 Comet and AN985 Centaur, the ASIX Electronics # AX88140A and AX88141, the Davicom DM9100 and DM9102, the Lite-On # 82c168 and 82c169 PNIC, the Lite-On/Macronix LC82C115 PNIC II # and the Macronix 98713/98713A/98715/98715A/98725 PMAC. This driver # replaces the old al, ax, dm, pn and mx drivers. List of brands: # Digital DE500-BA, Kingston KNE100TX, D-Link DFE-570TX, SOHOware SFA110, # SVEC PN102-TX, CNet Pro110B, 120A, and 120B, Compex RL100-TX, # LinkSys LNE100TX, LNE100TX V2.0, Jaton XpressNet, Alfa Inc GFC2204, # KNE110TX. # de: Digital Equipment DC21040 # em: Intel Pro/1000 Gigabit Ethernet 82542, 82543, 82544 based adapters. # ep: 3Com 3C509, 3C529, 3C556, 3C562D, 3C563D, 3C572, 3C574X, 3C579, 3C589 # and PC Card devices using these chipsets. # ex: Intel EtherExpress Pro/10 and other i82595-based adapters, # Olicom Ethernet PC Card devices. # fe: Fujitsu MB86960A/MB86965A Ethernet # fea: DEC DEFEA EISA FDDI adapter # fpa: Support for the Digital DEFPA PCI FDDI. `device fddi' is also needed. # fxp: Intel EtherExpress Pro/100B # (hint of prefer_iomap can be done to prefer I/O instead of Mem mapping) # gx: Intel Pro/1000 Gigabit Ethernet (82542, 82543-F, 82543-T) # lge: Support for PCI gigabit ethernet adapters based on the Level 1 # LXT1001 NetCellerator chipset. This includes the D-Link DGE-500SX, # SMC TigerCard 1000 (SMC9462SX), and some Addtron cards. # my: Myson Fast Ethernet (MTD80X, MTD89X) # nge: Support for PCI gigabit ethernet adapters based on the National # Semiconductor DP83820 and DP83821 chipset. This includes the # SMC EZ Card 1000 (SMC9462TX), D-Link DGE-500T, Asante FriendlyNet # GigaNIX 1000TA and 1000TPC, the Addtron AEG320T, the LinkSys # EG1032 and EG1064, the Surecom EP-320G-TX and the Netgear GA622T. # pcn: Support for PCI fast ethernet adapters based on the AMD Am79c97x # chipsets, including the PCnet/FAST, PCnet/FAST+, PCnet/PRO and # PCnet/Home. These were previously handled by the lnc driver (and # still will be if you leave this driver out of the kernel). # rl: Support for PCI fast ethernet adapters based on the RealTek 8129/8139 # chipset. Note that the RealTek driver defaults to using programmed # I/O to do register accesses because memory mapped mode seems to cause # severe lockups on SMP hardware. This driver also supports the # Accton EN1207D `Cheetah' adapter, which uses a chip called # the MPX 5030/5038, which is either a RealTek in disguise or a # RealTek workalike. Note that the D-Link DFE-530TX+ uses the RealTek # chipset and is supported by this driver, not the 'vr' driver. # sf: Support for Adaptec Duralink PCI fast ethernet adapters based on the # Adaptec AIC-6915 "starfire" controller. # This includes dual and quad port cards, as well as one 100baseFX card. # Most of these are 64-bit PCI devices, except for one single port # card which is 32-bit. # sis: Support for NICs based on the Silicon Integrated Systems SiS 900, # SiS 7016 and NS DP83815 PCI fast ethernet controller chips. # sbsh: Support for Granch SBNI16 SHDSL modem PCI adapters # sk: Support for the SysKonnect SK-984x series PCI gigabit ethernet NICs. # This includes the SK-9841 and SK-9842 single port cards (single mode # and multimode fiber) and the SK-9843 and SK-9844 dual port cards # (also single mode and multimode). # The driver will autodetect the number of ports on the card and # attach each one as a separate network interface. # sn: Support for ISA and PC Card Ethernet devices using the # SMC91C90/92/94/95 chips. # ste: Sundance Technologies ST201 PCI fast ethernet controller, includes # the D-Link DFE-550TX. # ti: Support for PCI gigabit ethernet NICs based on the Alteon Networks # Tigon 1 and Tigon 2 chipsets. This includes the Alteon AceNIC, the # 3Com 3c985, the Netgear GA620 and various others. Note that you will # probably want to bump up NMBCLUSTERS a lot to use this driver. # tl: Support for the Texas Instruments TNETE100 series 'ThunderLAN' # cards and integrated ethernet controllers. This includes several # Compaq Netelligent 10/100 cards and the built-in ethernet controllers # in several Compaq Prosignia, Proliant and Deskpro systems. It also # supports several Olicom 10Mbps and 10/100 boards. # tx: SMC 9432 TX, BTX and FTX cards. (SMC EtherPower II series) # txp: Support for 3Com 3cR990 cards with the "Typhoon" chipset # vr: Support for various fast ethernet adapters based on the VIA # Technologies VT3043 `Rhine I' and VT86C100A `Rhine II' chips, # including the D-Link DFE530TX (see 'rl' for DFE530TX+), the Hawking # Technologies PN102TX, and the AOpen/Acer ALN-320. # vx: 3Com 3C590 and 3C595 # wb: Support for fast ethernet adapters based on the Winbond W89C840F chip. # Note: this is not the same as the Winbond W89C940F, which is a # NE2000 clone. # wi: Lucent WaveLAN/IEEE 802.11 PCMCIA adapters. Note: this supports both # the PCMCIA and ISA cards: the ISA card is really a PCMCIA to ISA # bridge with a PCMCIA adapter plugged into it. # xe: Xircom/Intel EtherExpress Pro100/16 PC Card ethernet controller, # Accton Fast EtherCard-16, Compaq Netelligent 10/100 PC Card, # Toshiba 10/100 Ethernet PC Card, Xircom 16-bit Ethernet + Modem 56 # xl: Support for the 3Com 3c900, 3c905, 3c905B and 3c905C (Fast) # Etherlink XL cards and integrated controllers. This includes the # integrated 3c905B-TX chips in certain Dell Optiplex and Dell # Precision desktop machines and the integrated 3c905-TX chips # in Dell Latitude laptop docking stations. # Also supported: 3Com 3c980(C)-TX, 3Com 3cSOHO100-TX, 3Com 3c450-TX # Order for ISA/EISA devices is important here device cm hint.cm.0.at="isa" hint.cm.0.port="0x2e0" hint.cm.0.irq="9" hint.cm.0.maddr="0xdc000" device cs hint.cs.0.at="isa" hint.cs.0.port="0x300" device ep device ex device fe hint.fe.0.at="isa" hint.fe.0.port="0x300" device fea device sn hint.sn.0.at="isa" hint.sn.0.port="0x300" hint.sn.0.irq="10" device an device awi device cnw device wi device xe # PCI Ethernet NICs that use the common MII bus controller code. device dc # DEC/Intel 21143 and various workalikes device fxp # Intel EtherExpress PRO/100B (82557, 82558) hint.fxp.0.prefer_iomap="0" device my # Myson Fast Ethernet (MTD80X, MTD89X) device rl # RealTek 8129/8139 device pcn # AMD Am79C97x PCI 10/100 NICs device sf # Adaptec AIC-6915 (``Starfire'') device sbsh # Granch SBNI16 SHDSL modem device sis # Silicon Integrated Systems SiS 900/SiS 7016 device ste # Sundance ST201 (D-Link DFE-550TX) device tl # Texas Instruments ThunderLAN device tx # SMC EtherPower II (83c170 ``EPIC'') device vr # VIA Rhine, Rhine II device wb # Winbond W89C840F device xl # 3Com 3c90x (``Boomerang'', ``Cyclone'') # PCI Ethernet NICs. device de # DEC/Intel DC21x4x (``Tulip'') device txp # 3Com 3cR990 (``Typhoon'') device vx # 3Com 3c590, 3c595 (``Vortex'') # PCI Gigabit & FDDI NICs. device bge device gx device lge device nge device sk device ti device fpa # Use "private" jumbo buffers allocated exclusively for the ti(4) driver. # This option is incompatible with the TI_JUMBO_HDRSPLIT option below. #options TI_PRIVATE_JUMBOS # Turn on the header splitting option for the ti(4) driver firmware. This # only works for Tigon II chips, and has no effect for Tigon I chips. options TI_JUMBO_HDRSPLIT # These two options allow manipulating the mbuf cluster size and mbuf size, # respectively. Be very careful with NIC driver modules when changing # these from their default values, because that can potentially cause a # mismatch between the mbuf size assumed by the kernel and the mbuf size # assumed by a module. The only driver that currently has the ability to # detect a mismatch is ti(4). options MCLSHIFT=12 # mbuf cluster shift in bits, 12 == 4KB options MSIZE=512 # mbuf size in bytes # # ATM related options (Cranor version) # (note: this driver cannot be used with the HARP ATM stack) # # The `en' device provides support for Efficient Networks (ENI) # ENI-155 PCI midway cards, and the Adaptec 155Mbps PCI ATM cards (ANA-59x0). # # The `hatm' device provides support for Fore/Marconi HE155 and HE622 # ATM PCI cards. # # The `fatm' device provides support for Fore PCA200E ATM PCI cards. # # The `patm' device provides support for IDT77252 based cards like # ProSum's ProATM-155 and ProATM-25 and IDT's evaluation boards. # # atm device provides generic atm functions and is required for # atm devices. # NATM enables the netnatm protocol family that can be used to # bypass TCP/IP. # # utopia provides the access to the ATM PHY chips and is required for en, # hatm and fatm. # # the current driver supports only PVC operations (no atm-arp, no multicast). # for more details, please read the original documents at # http://www.ccrc.wustl.edu/pub/chuck/tech/bsdatm/bsdatm.html # device atm device en device fatm #Fore PCA200E device hatm #Fore/Marconi HE155/622 device patm #IDT77252 cards (ProATM and IDT) device utopia #ATM PHY driver options NATM #native ATM options LIBMBPOOL #needed by patm, iatm # # Audio drivers: `pcm', `sbc', `gusc' # # pcm: PCM audio through various sound cards. # # This has support for a large number of new audio cards, based on # CS423x, OPTi931, Yamaha OPL-SAx, and also for SB16, GusPnP. # For more information about this driver and supported cards, see pcm(4). # # The flags of the device tells the device a bit more info about the # device that normally is obtained through the PnP interface. # bit 2..0 secondary DMA channel; # bit 4 set if the board uses two dma channels; # bit 15..8 board type, overrides autodetection; leave it # zero if don't know what to put in (and you don't, # since this is unsupported at the moment...). # # Supported cards include: # Creative SoundBlaster ISA PnP/non-PnP # Supports ESS and Avance ISA chips as well. # Gravis UltraSound ISA PnP/non-PnP # Crystal Semiconductor CS461x/428x PCI # Neomagic 256AV (ac97) # Most of the more common ISA/PnP sb/mss/ess compatible cards. device pcm # For non-pnp sound cards with no bridge drivers only: hint.pcm.0.at="isa" hint.pcm.0.irq="10" hint.pcm.0.drq="1" hint.pcm.0.flags="0x0" # The bridge drivers for sound cards. These can be separately configured # for providing services to the likes of new-midi. # When used with 'device pcm' they also provide pcm sound services. # # sbc: Creative SoundBlaster ISA PnP/non-PnP # Supports ESS and Avance ISA chips as well. # gusc: Gravis UltraSound ISA PnP/non-PnP # csa: Crystal Semiconductor CS461x/428x PCI # For non-PnP cards: device sbc hint.sbc.0.at="isa" hint.sbc.0.port="0x220" hint.sbc.0.irq="5" hint.sbc.0.drq="1" hint.sbc.0.flags="0x15" device gusc hint.gusc.0.at="isa" hint.gusc.0.port="0x220" hint.gusc.0.irq="5" hint.gusc.0.drq="1" hint.gusc.0.flags="0x13" # # Miscellaneous hardware: # # scd: Sony CD-ROM using proprietary (non-ATAPI) interface # mcd: Mitsumi CD-ROM using proprietary (non-ATAPI) interface # bktr: Brooktree bt848/848a/849a/878/879 video capture and TV Tuner board # cy: Cyclades serial driver # joy: joystick (including IO DATA PCJOY PC Card joystick) # rc: RISCom/8 multiport card # rp: Comtrol Rocketport(ISA/PCI) - single card # si: Specialix SI/XIO 4-32 port terminal multiplexor # nmdm: nullmodem terminal driver (see nmdm(4)) # Notes on the Comtrol Rocketport driver: # # The exact values used for rp0 depend on how many boards you have # in the system. The manufacturer's sample configs are listed as: # # device rp # core driver support # # Comtrol Rocketport ISA single card # hint.rp.0.at="isa" # hint.rp.0.port="0x280" # # If instead you have two ISA cards, one installed at 0x100 and the # second installed at 0x180, then you should add the following to # your kernel probe hints: # hint.rp.0.at="isa" # hint.rp.0.port="0x100" # hint.rp.1.at="isa" # hint.rp.1.port="0x180" # # For 4 ISA cards, it might be something like this: # hint.rp.0.at="isa" # hint.rp.0.port="0x180" # hint.rp.1.at="isa" # hint.rp.1.port="0x100" # hint.rp.2.at="isa" # hint.rp.2.port="0x340" # hint.rp.3.at="isa" # hint.rp.3.port="0x240" # # For PCI cards, you need no hints. # Mitsumi CD-ROM device mcd hint.mcd.0.at="isa" hint.mcd.0.port="0x300" # for the Sony CDU31/33A CDROM device scd hint.scd.0.at="isa" hint.scd.0.port="0x230" device joy # PnP aware, hints for nonpnp only hint.joy.0.at="isa" hint.joy.0.port="0x201" device rc hint.rc.0.at="isa" hint.rc.0.port="0x220" hint.rc.0.irq="12" device rp hint.rp.0.at="isa" hint.rp.0.port="0x280" device si options SI_DEBUG hint.si.0.at="isa" hint.si.0.maddr="0xd0000" hint.si.0.irq="12" device nmdm # # The 'bktr' device is a PCI video capture device using the Brooktree # bt848/bt848a/bt849a/bt878/bt879 chipset. When used with a TV Tuner it forms a # TV card, eg Miro PC/TV, Hauppauge WinCast/TV WinTV, VideoLogic Captivator, # Intel Smart Video III, AverMedia, IMS Turbo, FlyVideo. # # options OVERRIDE_CARD=xxx # options OVERRIDE_TUNER=xxx # options OVERRIDE_MSP=1 # options OVERRIDE_DBX=1 # These options can be used to override the auto detection # The current values for xxx are found in src/sys/dev/bktr/bktr_card.h # Using sysctl(8) run-time overrides on a per-card basis can be made # # options BROOKTREE_SYSTEM_DEFAULT=BROOKTREE_PAL # or # options BROOKTREE_SYSTEM_DEFAULT=BROOKTREE_NTSC # Specifies the default video capture mode. # This is required for Dual Crystal (28&35Mhz) boards where PAL is used # to prevent hangs during initialisation. eg VideoLogic Captivator PCI. # # options BKTR_USE_PLL # PAL or SECAM users who have a 28Mhz crystal (and no 35Mhz crystal) # must enable PLL mode with this option. eg some new Bt878 cards. # # options BKTR_GPIO_ACCESS # This enable IOCTLs which give user level access to the GPIO port. # # options BKTR_NO_MSP_RESET # Prevents the MSP34xx reset. Good if you initialise the MSP in another OS first # # options BKTR_430_FX_MODE # Switch Bt878/879 cards into Intel 430FX chipset compatibility mode. # # options BKTR_SIS_VIA_MODE # Switch Bt878/879 cards into SIS/VIA chipset compatibility mode which is # needed for some old SiS and VIA chipset motherboards. # This also allows Bt878/879 chips to work on old OPTi (<1997) chipset # motherboards and motherboards with bad or incomplete PCI 2.1 support. # As a rough guess, old = before 1998 # # options BKTR_NEW_MSP34XX_DRIVER # Use new, more complete initialization scheme for the msp34* soundchip. # Should fix stereo autodetection if the old driver does only output # mono sound. # # options BKTR_USE_FREEBSD_SMBUS # Compile with FreeBSD SMBus implementation # # Brooktree driver has been ported to the new I2C framework. Thus, # you'll need to have the following 3 lines in the kernel config. # device smbus # device iicbus # device iicbb # device iicsmb # The iic and smb devices are only needed if you want to control other # I2C slaves connected to the external connector of some cards. # device bktr # # PC Card/PCMCIA # (OLDCARD) # # card: pccard slots # pcic: isa/pccard bridge #device pcic #hint.pcic.0.at="isa" #hint.pcic.1.at="isa" #device card 1 # # PC Card/PCMCIA and Cardbus # (NEWCARD) # # Note that NEWCARD and OLDCARD are incompatible. Do not use both at the same # time. # # pccbb: pci/cardbus bridge implementing YENTA interface # pccard: pccard slots # cardbus: cardbus slots device cbb device pccard device cardbus #device pcic ISA attachment currently busted #hint.pcic.0.at="isa" #hint.pcic.1.at="isa" # # SMB bus # # System Management Bus support is provided by the 'smbus' device. # Access to the SMBus device is via the 'smb' device (/dev/smb*), # which is a child of the 'smbus' device. # # Supported devices: # smb standard io through /dev/smb* # # Supported SMB interfaces: # iicsmb I2C to SMB bridge with any iicbus interface # bktr brooktree848 I2C hardware interface # intpm Intel PIIX4 (82371AB, 82443MX) Power Management Unit # alpm Acer Aladdin-IV/V/Pro2 Power Management Unit # ichsmb Intel ICH SMBus controller chips (82801AA, 82801AB, 82801BA) # viapm VIA VT82C586B/596B/686A and VT8233 Power Management Unit # amdpm AMD 756 Power Management Unit # nfpm NVIDIA nForce Power Management Unit # device smbus # Bus support, required for smb below. device intpm device alpm device ichsmb device viapm device amdpm device nfpm device smb # # I2C Bus # # Philips i2c bus support is provided by the `iicbus' device. # # Supported devices: # ic i2c network interface # iic i2c standard io # iicsmb i2c to smb bridge. Allow i2c i/o with smb commands. # # Supported interfaces: # bktr brooktree848 I2C software interface # # Other: # iicbb generic I2C bit-banging code (needed by lpbb, bktr) # device iicbus # Bus support, required for ic/iic/iicsmb below. device iicbb device ic device iic device iicsmb # smb over i2c bridge # Parallel-Port Bus # # Parallel port bus support is provided by the `ppbus' device. # Multiple devices may be attached to the parallel port, devices # are automatically probed and attached when found. # # Supported devices: # vpo Iomega Zip Drive # Requires SCSI disk support ('scbus' and 'da'), best # performance is achieved with ports in EPP 1.9 mode. # lpt Parallel Printer # plip Parallel network interface # ppi General-purpose I/O ("Geek Port") + IEEE1284 I/O # pps Pulse per second Timing Interface # lpbb Philips official parallel port I2C bit-banging interface # # Supported interfaces: # ppc ISA-bus parallel port interfaces. # options PPC_PROBE_CHIPSET # Enable chipset specific detection # (see flags in ppc(4)) options DEBUG_1284 # IEEE1284 signaling protocol debug options PERIPH_1284 # Makes your computer act as an IEEE1284 # compliant peripheral options DONTPROBE_1284 # Avoid boot detection of PnP parallel devices options VP0_DEBUG # ZIP/ZIP+ debug options LPT_DEBUG # Printer driver debug options PPC_DEBUG # Parallel chipset level debug options PLIP_DEBUG # Parallel network IP interface debug options PCFCLOCK_VERBOSE # Verbose pcfclock driver options PCFCLOCK_MAX_RETRIES=5 # Maximum read tries (default 10) device ppc hint.ppc.0.at="isa" hint.ppc.0.irq="7" device ppbus device vpo device lpt device plip device ppi device pps device lpbb device pcfclock # Kernel BOOTP support options BOOTP # Use BOOTP to obtain IP address/hostname # Requires NFSCLIENT and NFS_ROOT options BOOTP_NFSROOT # NFS mount root filesystem using BOOTP info options BOOTP_NFSV3 # Use NFS v3 to NFS mount root options BOOTP_COMPAT # Workaround for broken bootp daemons. options BOOTP_WIRED_TO=fxp0 # Use interface fxp0 for BOOTP # # Add tie-ins for a hardware watchdog. This only enables the hooks; # the user must still supply the actual driver. # options HW_WDOG # # Add software watchdog routines. # options SW_WATCHDOG # # Disable swapping of upages and stack pages. This option removes all # code which actually performs swapping, so it's not possible to turn # it back on at run-time. # # This is sometimes usable for systems which don't have any swap space # (see also sysctls "vm.defer_swapspace_pageouts" and # "vm.disable_swapspace_pageouts") # #options NO_SWAPPING # Set the number of sf_bufs to allocate. sf_bufs are virtual buffers # for sendfile(2) that are used to map file VM pages, and normally # default to a quantity that is roughly 16*MAXUSERS+512. You would # typically want about 4 of these for each simultaneous file send. # options NSFBUFS=1024 # # Enable extra debugging code for locks. This stores the filename and # line of whatever acquired the lock in the lock itself, and change a # number of function calls to pass around the relevant data. This is # not at all useful unless you are debugging lock code. Also note # that it is likely to break e.g. fstat(1) unless you recompile your # userland with -DDEBUG_LOCKS as well. # options DEBUG_LOCKS ##################################################################### # USB support # UHCI controller device uhci # OHCI controller device ohci # EHCI controller device ehci # General USB code (mandatory for USB) device usb # # USB Double Bulk Pipe devices device udbp # USB Fm Radio device ufm # Generic USB device driver device ugen # Human Interface Device (anything with buttons and dials) device uhid # USB keyboard device ukbd # USB printer device ulpt # USB Iomega Zip 100 Drive (Requires scbus and da) device umass # USB support for Belkin F5U109 and Magic Control Technology serial adapters device umct # USB modem support device umodem # USB mouse device ums # Diamond Rio 500 Mp3 player device urio # USB scanners device uscanner # # USB serial support device ucom # USB support for Belkin F5U103 and compatible serial adapters device ubsa # USB support for BWCT console serial adapters device ubser # USB support for serial adapters based on the FT8U100AX and FT8U232AM device uftdi # USB support for Prolific PL-2303 serial adapters device uplcom # USB Visor and Palm devices device uvisor # USB serial support for DDI pocket's PHS device uvscom # # ADMtek USB ethernet. Supports the LinkSys USB100TX, # the Billionton USB100, the Melco LU-ATX, the D-Link DSB-650TX # and the SMC 2202USB. Also works with the ADMtek AN986 Pegasus # eval board. device aue # # CATC USB-EL1201A USB ethernet. Supports the CATC Netmate # and Netmate II, and the Belkin F5U111. device cue # # Kawasaki LSI ethernet. Supports the LinkSys USB10T, # Entrega USB-NET-E45, Peracom Ethernet Adapter, the # 3Com 3c19250, the ADS Technologies USB-10BT, the ATen UC10T, # the Netgear EA101, the D-Link DSB-650, the SMC 2102USB # and 2104USB, and the Corega USB-T. device kue # # RealTek RTL8150 USB to fast ethernet. Supports the Melco LUA-KTX # and the GREEN HOUSE GH-USB100B. device rue # # Davicom DM9601E USB to fast ethernet. Supports the Corega FEther USB-TXC. device udav # debugging options for the USB subsystem # options USB_DEBUG # options for ukbd: options UKBD_DFLT_KEYMAP # specify the built-in keymap makeoptions UKBD_DFLT_KEYMAP=it.iso # options for uplcom: options UPLCOM_INTR_INTERVAL=100 # interrupt pipe interval # in milliseconds # options for uvscom: options UVSCOM_DEFAULT_OPKTSIZE=8 # default output packet size options UVSCOM_INTR_INTERVAL=100 # interrupt pipe interval # in milliseconds ##################################################################### # FireWire support device firewire # FireWire bus code device sbp # SCSI over Firewire (Requires scbus and da) device sbp_targ # SBP-2 Target mode (Requires scbus and targ) device fwe # Ethernet over FireWire (non-standard!) device fwip # IP over FireWire (rfc2734 and rfc3146) ##################################################################### # dcons support (Dumb Console Device) device dcons # dumb console driver device dcons_crom # FireWire attachment options DCONS_BUF_SIZE=16384 # buffer size options DCONS_POLL_HZ=100 # polling rate options DCONS_FORCE_CONSOLE=0 # force to be the primary console options DCONS_FORCE_GDB=1 # force to be the gdb device ##################################################################### # crypto subsystem # # This is a port of the openbsd crypto framework. Include this when # configuring FAST_IPSEC and when you have a h/w crypto device to accelerate # user applications that link to openssl. # # Drivers are ports from openbsd with some simple enhancements that have # been fed back to openbsd. device crypto # core crypto support device cryptodev # /dev/crypto for access to h/w device rndtest # FIPS 140-2 entropy tester device hifn # Hifn 7951, 7781, etc. options HIFN_DEBUG # enable debugging support: hw.hifn.debug options HIFN_RNDTEST # enable rndtest support device ubsec # Broadcom 5501, 5601, 58xx options UBSEC_DEBUG # enable debugging support: hw.ubsec.debug options UBSEC_RNDTEST # enable rndtest support ##################################################################### # # Embedded system options: # # An embedded system might want to run something other than init. options INIT_PATH=/sbin/init:/stand/sysinstall # Debug options options BUS_DEBUG # enable newbus debugging options DEBUG_VFS_LOCKS # enable vfs lock debugging options SOCKBUF_DEBUG # enable sockbuf last record/mb tail checking ##################################################################### # SYSV IPC KERNEL PARAMETERS # # Maximum number of entries in a semaphore map. options SEMMAP=31 # Maximum number of System V semaphores that can be used on the system at # one time. options SEMMNI=11 # Total number of semaphores system wide options SEMMNS=61 # Total number of undo structures in system options SEMMNU=31 # Maximum number of System V semaphores that can be used by a single process # at one time. options SEMMSL=61 # Maximum number of operations that can be outstanding on a single System V # semaphore at one time. options SEMOPM=101 # Maximum number of undo operations that can be outstanding on a single # System V semaphore at one time. options SEMUME=11 # Maximum number of shared memory pages system wide. options SHMALL=1025 # Maximum size, in bytes, of a single System V shared memory region. options SHMMAX=(SHMMAXPGS*PAGE_SIZE+1) options SHMMAXPGS=1025 # Minimum size, in bytes, of a single System V shared memory region. options SHMMIN=2 # Maximum number of shared memory regions that can be used on the system # at one time. options SHMMNI=33 # Maximum number of System V shared memory regions that can be attached to # a single process at one time. options SHMSEG=9 # Set the amount of time (in seconds) the system will wait before # rebooting automatically when a kernel panic occurs. If set to (-1), # the system will wait indefinitely until a key is pressed on the # console. options PANIC_REBOOT_WAIT_TIME=16 # Attempt to bypass the buffer cache and put data directly into the # userland buffer for read operation when O_DIRECT flag is set on the # file. Both offset and length of the read operation must be # multiples of the physical media sector size. # #options DIRECTIO # Specify a lower limit for the number of swap I/O buffers. They are # (among other things) used when bypassing the buffer cache due to # DIRECTIO kernel option enabled and O_DIRECT flag set on file. # #options NSWBUF_MIN=120 ##################################################################### # More undocumented options for linting. # Note that documenting these are not considered an affront. options CAM_DEBUG_DELAY # VFS cluster debugging. options CLUSTERDEBUG options DEBUG # Kernel filelock debugging. options LOCKF_DEBUG # System V compatible message queues # Please note that the values provided here are used to test kernel # building. The defaults in the sources provide almost the same numbers. # MSGSSZ must be a power of 2 between 8 and 1024. options MSGMNB=2049 # Max number of chars in queue options MSGMNI=41 # Max number of message queue identifiers options MSGSEG=2049 # Max number of message segments options MSGSSZ=16 # Size of a message segment options MSGTQL=41 # Max number of messages in system options NBUF=512 # Number of buffer headers options NMBCLUSTERS=1024 # Number of mbuf clusters options SCSI_NCR_DEBUG options SCSI_NCR_MAX_SYNC=10000 options SCSI_NCR_MAX_WIDE=1 options SCSI_NCR_MYADDR=7 options SC_DEBUG_LEVEL=5 # Syscons debug level options SC_RENDER_DEBUG # syscons rendering debugging options SHOW_BUSYBUFS # List buffers that prevent root unmount options SLIP_IFF_OPTS options VFS_BIO_DEBUG # VFS buffer I/O debugging options KSTACK_MAX_PAGES=32 # Maximum pages to give the kernel stack # Adaptec Array Controller driver options options AAC_DEBUG # Debugging levels: # 0 - quiet, only emit warnings # 1 - noisy, emit major function # points and things done # 2 - extremely noisy, emit trace # items in loops, etc. # Yet more undocumented options for linting. # BKTR_ALLOC_PAGES has no effect except to cause warnings, and # BROOKTREE_ALLOC_PAGES hasn't actually been superseded by it, since the # driver still mostly spells this option BROOKTREE_ALLOC_PAGES. ##options BKTR_ALLOC_PAGES=(217*4+1) options BROOKTREE_ALLOC_PAGES=(217*4+1) options MAXFILES=999 options NDEVFSINO=1025 options NDEVFSOVERFLOW=32769 # Yet more undocumented options for linting. options VGA_DEBUG Index: head/sys/conf/options =================================================================== --- head/sys/conf/options (revision 131258) +++ head/sys/conf/options (revision 131259) @@ -1,666 +1,668 @@ # $FreeBSD$ # # On the handling of kernel options # # All kernel options should be listed in NOTES, with suitable # descriptions. Negative options (options that make some code not # compile) should be commented out; LINT (generated from NOTES) should # compile as much code as possible. Try to structure option-using # code so that a single option only switch code on, or only switch # code off, to make it possible to have a full compile-test. If # necessary, you can check for COMPILING_LINT to get maximum code # coverage. # # All new options shall also be listed in either "conf/options" or # "conf/options.". Options that affect a single source-file # .[c|s] should be directed into "opt_.h", while options # that affect multiple files should either go in "opt_global.h" if # this is a kernel-wide option (used just about everywhere), or in # "opt_.h" if it affect only some files. # Note that the effect of listing only an option without a # header-file-name in conf/options (and cousins) is that the last # convention is followed. # # This handling scheme is not yet fully implemented. # # # Format of this file: # Option name filename # # If filename is missing, the default is # opt_.h AAC_DEBUG opt_aac.h AHC_ALLOW_MEMIO opt_aic7xxx.h AHC_TMODE_ENABLE opt_aic7xxx.h AHC_DUMP_EEPROM opt_aic7xxx.h AHC_DEBUG opt_aic7xxx.h AHC_DEBUG_OPTS opt_aic7xxx.h AHC_REG_PRETTY_PRINT opt_aic7xxx.h AHD_DEBUG opt_aic79xx.h AHD_DEBUG_OPTS opt_aic79xx.h AHD_TMODE_ENABLE opt_aic79xx.h AHD_REG_PRETTY_PRINT opt_aic79xx.h ADW_ALLOW_MEMIO opt_adw.h TWA_DEBUG opt_twa.h TWA_FLASH_FIRMWARE opt_twa.h # Miscellaneous options. ADAPTIVE_MUTEXES ALQ CODA_COMPAT_5 opt_coda.h COMPAT_43 opt_compat.h COMPAT_FREEBSD4 opt_compat.h COMPILING_LINT opt_global.h CONSPEED opt_comconsole.h CY_PCI_FASTINTR DDB DDB_NOKLDSYM opt_ddb.h DDB_NUMSYM opt_ddb.h DDB_TRACE DDB_UNATTENDED DIRECTIO opt_directio.h GDB_REMOTE_CHAT opt_ddb.h GDBSPEED opt_ddb.h GEOM_AES opt_geom.h GEOM_APPLE opt_geom.h GEOM_BDE opt_geom.h GEOM_BSD opt_geom.h GEOM_CONCAT opt_geom.h GEOM_FOX opt_geom.h GEOM_GATE opt_geom.h GEOM_GPT opt_geom.h GEOM_MBR opt_geom.h GEOM_MIRROR opt_geom.h GEOM_NOP opt_geom.h GEOM_PC98 opt_geom.h GEOM_STRIPE opt_geom.h GEOM_SUNLABEL opt_geom.h GEOM_VOL opt_geom.h HW_WDOG KSTACK_MAX_PAGES KSTACK_PAGES KTRACE KTRACE_REQUEST_POOL opt_ktrace.h LIBICONV MAC MAC_ALWAYS_LABEL_MBUF opt_mac.h MAC_BIBA opt_dontuse.h MAC_BSDEXTENDED opt_dontuse.h MAC_DEBUG opt_mac.h MAC_IFOFF opt_dontuse.h MAC_LOMAC opt_dontuse.h MAC_MLS opt_dontuse.h MAC_NONE opt_dontuse.h MAC_PARTITION opt_dontuse.h MAC_PORTACL opt_dontuse.h MAC_SEEOTHERUIDS opt_dontuse.h MAC_STATIC opt_mac.h MAC_STUB opt_dontuse.h MAC_TEST opt_dontuse.h MD_ROOT opt_md.h MD_ROOT_SIZE opt_md.h MUTEX_WAKE_ALL NSWBUF_MIN opt_swap.h PANIC_REBOOT_WAIT_TIME opt_panic.h PPS_SYNC opt_ntp.h PUC_FASTINTR opt_puc.h QUOTA SCHED_4BSD opt_sched.h SCHED_ULE opt_sched.h SHOW_BUSYBUFS +SLEEPQUEUE_PROFILING SPX_HACK SUIDDIR opt_suiddir.h MSGMNB opt_sysvipc.h MSGMNI opt_sysvipc.h MSGSEG opt_sysvipc.h MSGSSZ opt_sysvipc.h MSGTQL opt_sysvipc.h SEMMAP opt_sysvipc.h SEMMNI opt_sysvipc.h SEMMNS opt_sysvipc.h SEMMNU opt_sysvipc.h SEMMSL opt_sysvipc.h SEMOPM opt_sysvipc.h SEMUME opt_sysvipc.h SHMALL opt_sysvipc.h SHMMAX opt_sysvipc.h SHMMAXPGS opt_sysvipc.h SHMMIN opt_sysvipc.h SHMMNI opt_sysvipc.h SHMSEG opt_sysvipc.h SYSVMSG opt_sysvipc.h SYSVSEM opt_sysvipc.h SYSVSHM opt_sysvipc.h SW_WATCHDOG opt_watchdog.h +TURNSTILE_PROFILING TTYHOG opt_tty.h VFS_AIO WLCACHE opt_wavelan.h WLDEBUG opt_wavelan.h # POSIX kernel options P1003_1B_SEMAPHORES opt_posix.h _KPOSIX_PRIORITY_SCHEDULING opt_posix.h # Do we want the config file compiled into the kernel? INCLUDE_CONFIG_FILE opt_config.h # Options for static filesystems. These should only be used at config # time, since the corresponding lkms cannot work if there are any static # dependencies. Unusability is enforced by hiding the defines for the # options in a never-included header. CD9660 opt_dontuse.h CODA opt_dontuse.h EXT2FS opt_dontuse.h FDESCFS opt_dontuse.h HPFS opt_dontuse.h LINPROCFS opt_dontuse.h MSDOSFS opt_dontuse.h NTFS opt_dontuse.h NULLFS opt_dontuse.h NWFS opt_dontuse.h PORTALFS opt_dontuse.h PROCFS opt_dontuse.h PSEUDOFS opt_dontuse.h SMBFS opt_dontuse.h UDF opt_dontuse.h UMAPFS opt_dontuse.h UNIONFS opt_dontuse.h # Broken - ffs_snapshot() dependency from ufs_lookup() :-( FFS opt_ffs_broken_fixme.h # These static filesystems have one slightly bogus static dependency in # sys/i386/i386/autoconf.c. If any of these filesystems are # statically compiled into the kernel, code for mounting them as root # filesystems will be enabled - but look below. NFSCLIENT opt_nfs.h NFSSERVER opt_nfs.h # filesystems and libiconv bridge CD9660_ICONV opt_dontuse.h MSDOSFS_ICONV opt_dontuse.h NTFS_ICONV opt_dontuse.h UDF_ICONV opt_dontuse.h # If you are following the conditions in the copyright, # you can enable soft-updates which will speed up a lot of thigs # and make the system safer from crashes at the same time. # otherwise a STUB module will be compiled in. SOFTUPDATES opt_ffs.h # Enabling this option turns on support for Access Control Lists in UFS, # which can be used to support high security configurations. Depends on # UFS_EXTATTR. UFS_ACL opt_ufs.h # Enabling this option turns on support for extended attributes in UFS-based # filesystems, which can be used to support high security configurations # as well as new filesystem features. UFS_EXTATTR opt_ufs.h UFS_EXTATTR_AUTOSTART opt_ufs.h # Enable fast hash lookups for large directories on UFS-based filesystems. UFS_DIRHASH opt_ufs.h # The below sentence is not in English, and neither is this one. # We plan to remove the static dependences above, with a # _ROOT option to control if it usable as root. This list # allows these options to be present in config files already (though # they won't make any difference yet). NFS_ROOT opt_nfsroot.h # SMB/CIFS requester NETSMB opt_netsmb.h NETSMBCRYPTO opt_netsmb.h # Options used only in subr_param.c. HZ opt_param.h MAXFILES opt_param.h NBUF opt_param.h NMBCLUSTERS opt_param.h NSFBUFS opt_param.h VM_BCACHE_SIZE_MAX opt_param.h VM_SWZONE_SIZE_MAX opt_param.h MAXUSERS DFLDSIZ opt_param.h MAXDSIZ opt_param.h MAXSSIZ opt_param.h # Generic SCSI options. CAM_MAX_HIGHPOWER opt_cam.h CAMDEBUG opt_cam.h CAM_DEBUG_DELAY opt_cam.h CAM_DEBUG_BUS opt_cam.h CAM_DEBUG_TARGET opt_cam.h CAM_DEBUG_LUN opt_cam.h CAM_DEBUG_FLAGS opt_cam.h CAM_NEW_TRAN_CODE opt_cam.h SCSI_DELAY opt_scsi.h SCSI_NO_SENSE_STRINGS opt_scsi.h SCSI_NO_OP_STRINGS opt_scsi.h # Options used only in cam/scsi/scsi_cd.c CHANGER_MIN_BUSY_SECONDS opt_cd.h CHANGER_MAX_BUSY_SECONDS opt_cd.h # Options used only in cam/scsi/scsi_sa.c. SA_IO_TIMEOUT opt_sa.h SA_SPACE_TIMEOUT opt_sa.h SA_REWIND_TIMEOUT opt_sa.h SA_ERASE_TIMEOUT opt_sa.h SA_1FM_AT_EOD opt_sa.h # Options used only in cam/scsi/scsi_pt.c SCSI_PT_DEFAULT_TIMEOUT opt_pt.h # Options used only in cam/scsi/scsi_ses.c SES_ENABLE_PASSTHROUGH opt_ses.h # Options used in dev/sym/ (Symbios SCSI driver). SYM_SETUP_LP_PROBE_MAP opt_sym.h #-Low Priority Probe Map (bits) # Allows the ncr to take precedence # 1 (1<<0) -> 810a, 860 # 2 (1<<1) -> 825a, 875, 885, 895 # 4 (1<<2) -> 895a, 896, 1510d SYM_SETUP_SCSI_DIFF opt_sym.h #-HVD support for 825a, 875, 885 # disabled:0 (default), enabled:1 SYM_SETUP_PCI_PARITY opt_sym.h #-PCI parity checking # disabled:0, enabled:1 (default) SYM_SETUP_MAX_LUN opt_sym.h #-Number of LUNs supported # default:8, range:[1..64] # Options used only in pci/ncr.c SCSI_NCR_DEBUG opt_ncr.h SCSI_NCR_MAX_SYNC opt_ncr.h SCSI_NCR_MAX_WIDE opt_ncr.h SCSI_NCR_MYADDR opt_ncr.h # Options used only in dev/isp/* ISP_TARGET_MODE opt_isp.h ISP_FW_CRASH_DUMP opt_isp.h # Options used in the 'ata' ATA/ATAPI driver ATA_STATIC_ID opt_ata.h ATA_NOPCI opt_ata.h DEV_ATADISK opt_ata.h DEV_ATAPICD opt_ata.h DEV_ATAPIST opt_ata.h DEV_ATAPIFD opt_ata.h DEV_ATAPICAM opt_ata.h DEV_ATARAID opt_ata.h # Net stuff. ACCEPT_FILTER_DATA ACCEPT_FILTER_HTTP ALTQ opt_global.h ALTQ_CBQ opt_altq.h ALTQ_RED opt_altq.h ALTQ_RIO opt_altq.h ALTQ_HFSC opt_altq.h ALTQ_CDNR opt_altq.h ALTQ_PRIQ opt_altq.h ALTQ_NOPCC opt_altq.h ALTQ_DEBUG opt_altq.h BOOTP opt_bootp.h BOOTP_COMPAT opt_bootp.h BOOTP_NFSROOT opt_bootp.h BOOTP_NFSV3 opt_bootp.h BOOTP_WIRED_TO opt_bootp.h BRIDGE opt_bdg.h DEV_PF opt_pf.h DEV_PFLOG opt_pf.h DEV_PFSYNC opt_pf.h ETHER_II opt_ef.h ETHER_8023 opt_ef.h ETHER_8022 opt_ef.h ETHER_SNAP opt_ef.h MROUTING opt_mrouting.h PIM opt_mrouting.h INET opt_inet.h INET6 opt_inet6.h IPSEC opt_ipsec.h IPSEC_ESP opt_ipsec.h IPSEC_DEBUG opt_ipsec.h IPSEC_FILTERGIF opt_ipsec.h FAST_IPSEC opt_ipsec.h IPDIVERT DUMMYNET opt_ipdn.h IPFILTER opt_ipfilter.h IPFILTER_LOG opt_ipfilter.h IPFILTER_DEFAULT_BLOCK opt_ipfilter.h PFIL_HOOKS opt_pfil_hooks.h IPFIREWALL opt_ipfw.h IPFIREWALL_VERBOSE opt_ipfw.h IPFIREWALL_VERBOSE_LIMIT opt_ipfw.h IPFIREWALL_DEFAULT_TO_ACCEPT opt_ipfw.h IPV6FIREWALL opt_ip6fw.h IPV6FIREWALL_VERBOSE opt_ip6fw.h IPV6FIREWALL_VERBOSE_LIMIT opt_ip6fw.h IPV6FIREWALL_DEFAULT_TO_ACCEPT opt_ip6fw.h IPSTEALTH IPX IPXIP opt_ipx.h LIBMBPOOL LIBMCHAIN MBUF_STRESS_TEST opt_mbuf_stress_test.h NCP NETATALK opt_atalk.h PPP_BSDCOMP opt_ppp.h PPP_DEFLATE opt_ppp.h PPP_FILTER opt_ppp.h RANDOM_IP_ID SLIP_IFF_OPTS opt_slip.h TCPDEBUG TCP_SIGNATURE opt_inet.h TCP_SACK_DEBUG opt_tcp_sack.h TCP_DROP_SYNFIN opt_tcp_input.h XBONEHACK # Netgraph(4). Use option NETGRAPH to enable the base netgraph code. # Each netgraph node type can be either be compiled into the kernel # or loaded dynamically. To get the former, include the corresponding # option below. Each type has its own man page, e.g. ng_async(4). NETGRAPH NETGRAPH_ASYNC opt_netgraph.h NETGRAPH_ATMLLC opt_netgraph.h NETGRAPH_ATM_ATMPIF opt_netgraph.h NETGRAPH_BLUETOOTH opt_netgraph.h NETGRAPH_BLUETOOTH_BT3C opt_netgraph.h NETGRAPH_BLUETOOTH_H4 opt_netgraph.h NETGRAPH_BLUETOOTH_HCI opt_netgraph.h NETGRAPH_BLUETOOTH_L2CAP opt_netgraph.h NETGRAPH_BLUETOOTH_SOCKET opt_netgraph.h NETGRAPH_BLUETOOTH_UBT opt_netgraph.h NETGRAPH_BLUETOOTH_UBTBCMFW opt_netgraph.h NETGRAPH_BPF opt_netgraph.h NETGRAPH_BRIDGE opt_netgraph.h NETGRAPH_CISCO opt_netgraph.h NETGRAPH_ECHO opt_netgraph.h NETGRAPH_EIFACE opt_netgraph.h NETGRAPH_ETHER opt_netgraph.h NETGRAPH_FEC opt_netgraph.h NETGRAPH_FRAME_RELAY opt_netgraph.h NETGRAPH_GIF opt_netgraph.h NETGRAPH_GIF_DEMUX opt_netgraph.h NETGRAPH_HOLE opt_netgraph.h NETGRAPH_IFACE opt_netgraph.h NETGRAPH_IP_INPUT opt_netgraph.h NETGRAPH_KSOCKET opt_netgraph.h NETGRAPH_L2TP opt_netgraph.h NETGRAPH_LMI opt_netgraph.h # MPPC compression requires proprietary files (not included) NETGRAPH_MPPC_COMPRESSION opt_netgraph.h NETGRAPH_MPPC_ENCRYPTION opt_netgraph.h NETGRAPH_ONE2MANY opt_netgraph.h NETGRAPH_PPP opt_netgraph.h NETGRAPH_PPPOE opt_netgraph.h NETGRAPH_PPTPGRE opt_netgraph.h NETGRAPH_RFC1490 opt_netgraph.h NETGRAPH_SOCKET opt_netgraph.h NETGRAPH_SPLIT opt_netgraph.h NETGRAPH_SPPP opt_netgraph.h NETGRAPH_TEE opt_netgraph.h NETGRAPH_TTY opt_netgraph.h NETGRAPH_UI opt_netgraph.h NETGRAPH_VJC opt_netgraph.h # NgATM options NGATM_ATM opt_netgraph.h NGATM_ATMBASE opt_netgraph.h NGATM_SSCOP opt_netgraph.h NGATM_SSCFU opt_netgraph.h NGATM_UNI opt_netgraph.h # DRM options DRM_DEBUG opt_drm.h ZERO_COPY_SOCKETS opt_zero.h TI_PRIVATE_JUMBOS opt_ti.h TI_JUMBO_HDRSPLIT opt_ti.h # ATM (HARP version) ATM_CORE opt_atm.h ATM_IP opt_atm.h ATM_SIGPVC opt_atm.h ATM_SPANS opt_atm.h ATM_UNI opt_atm.h # XXX Conflict: # of devices vs network protocol (Native ATM). # This makes "atm.h" unusable. NATM # DPT driver debug flags DPT_MEASURE_PERFORMANCE opt_dpt.h DPT_HANDLE_TIMEOUTS opt_dpt.h DPT_TIMEOUT_FACTOR opt_dpt.h DPT_LOST_IRQ opt_dpt.h DPT_RESET_HBA opt_dpt.h # Misc debug flags. Most of these should probably be replaced with # 'DEBUG', and then let people recompile just the interesting modules # with 'make CC="cc -DDEBUG"'. CLUSTERDEBUG opt_debug_cluster.h DEBUG_1284 opt_ppb_1284.h VP0_DEBUG opt_vpo.h LPT_DEBUG opt_lpt.h PLIP_DEBUG opt_plip.h LOCKF_DEBUG opt_debug_lockf.h NPX_DEBUG opt_debug_npx.h NETATALKDEBUG opt_atalk.h SI_DEBUG opt_debug_si.h SX_DEBUG opt_debug_sx.h # Fb options FB_DEBUG opt_fb.h FB_INSTALL_CDEV opt_fb.h # ppbus related options PERIPH_1284 opt_ppb_1284.h DONTPROBE_1284 opt_ppb_1284.h # smbus related options ENABLE_ALART opt_intpm.h # These cause changes all over the kernel BLKDEV_IOSIZE opt_global.h BURN_BRIDGES opt_global.h DEBUG opt_global.h DEBUG_LOCKS opt_global.h DEBUG_VFS_LOCKS opt_global.h DIAGNOSTIC opt_global.h INVARIANT_SUPPORT opt_global.h INVARIANTS opt_global.h LOOKUP_SHARED opt_global.h MCLSHIFT opt_global.h MUTEX_DEBUG opt_global.h MUTEX_NOINLINE opt_global.h MUTEX_PROFILING opt_global.h MSIZE opt_global.h REGRESSION opt_global.h RESTARTABLE_PANICS opt_global.h VFS_BIO_DEBUG opt_global.h # These are VM related options VM_KMEM_SIZE opt_vm.h VM_KMEM_SIZE_SCALE opt_vm.h VM_KMEM_SIZE_MAX opt_vm.h NO_SWAPPING opt_vm.h MALLOC_MAKE_FAILURES opt_vm.h MALLOC_PROFILE opt_vm.h PQ_NOOPT opt_vmpage.h PQ_NORMALCACHE opt_vmpage.h PQ_MEDIUMCACHE opt_vmpage.h PQ_LARGECACHE opt_vmpage.h PQ_HUGECACHE opt_vmpage.h PQ_CACHESIZE opt_vmpage.h # Standard SMP options SMP opt_global.h # Size of the kernel message buffer MSGBUF_SIZE opt_msgbuf.h # NFS options NFS_MINATTRTIMO opt_nfs.h NFS_MAXATTRTIMO opt_nfs.h NFS_MINDIRATTRTIMO opt_nfs.h NFS_MAXDIRATTRTIMO opt_nfs.h NFS_GATHERDELAY opt_nfs.h NFS_WDELAYHASHSIZ opt_nfs.h NFS_DEBUG opt_nfs.h # For the Bt848/Bt848A/Bt849/Bt878/Bt879 driver OVERRIDE_CARD opt_bktr.h OVERRIDE_TUNER opt_bktr.h OVERRIDE_DBX opt_bktr.h OVERRIDE_MSP opt_bktr.h BROOKTREE_SYSTEM_DEFAULT opt_bktr.h BROOKTREE_ALLOC_PAGES opt_bktr.h BKTR_OVERRIDE_CARD opt_bktr.h BKTR_OVERRIDE_TUNER opt_bktr.h BKTR_OVERRIDE_DBX opt_bktr.h BKTR_OVERRIDE_MSP opt_bktr.h BKTR_SYSTEM_DEFAULT opt_bktr.h BKTR_ALLOC_PAGES opt_bktr.h BKTR_USE_PLL opt_bktr.h BKTR_GPIO_ACCESS opt_bktr.h BKTR_NO_MSP_RESET opt_bktr.h BKTR_430_FX_MODE opt_bktr.h BKTR_SIS_VIA_MODE opt_bktr.h BKTR_USE_FREEBSD_SMBUS opt_bktr.h BKTR_NEW_MSP34XX_DRIVER opt_bktr.h # options for serial support COM_ESP opt_sio.h COM_MULTIPORT opt_sio.h BREAK_TO_DEBUGGER opt_comconsole.h ALT_BREAK_TO_DEBUGGER opt_comconsole.h # Options to support PPS UART_PPS_ON_CTS opt_uart.h # options for bus/device framework BUS_DEBUG opt_bus.h # options for USB support USB_DEBUG opt_usb.h USBVERBOSE opt_usb.h UKBD_DFLT_KEYMAP opt_ukbd.h UPLCOM_INTR_INTERVAL opt_uplcom.h UVSCOM_DEFAULT_OPKTSIZE opt_uvscom.h UVSCOM_INTR_INTERVAL opt_uvscom.h # Vinum options VINUMDEBUG opt_vinum.h # Embedded system options INIT_PATH opt_init_path.h ROOTDEVNAME opt_rootdevname.h FDC_DEBUG opt_fdc.h PCFCLOCK_VERBOSE opt_pcfclock.h PCFCLOCK_MAX_RETRIES opt_pcfclock.h TDFX_LINUX opt_tdfx.h KTR opt_global.h KTR_ALQ opt_ktr.h KTR_MASK opt_ktr.h KTR_CPUMASK opt_ktr.h KTR_COMPILE opt_global.h KTR_ENTRIES opt_global.h KTR_VERBOSE opt_ktr.h WITNESS opt_global.h WITNESS_DDB opt_witness.h WITNESS_SKIPSPIN opt_witness.h # options for ACPI support ACPI_DEBUG opt_acpi.h ACPI_MAX_THREADS opt_acpi.h ACPI_NO_SEMAPHORES opt_acpi.h ACPICA_PEDANTIC opt_acpi.h # options for DEVFS, see sys/fs/devfs/devfs.h NDEVFSINO opt_devfs.h NDEVFSOVERFLOW opt_devfs.h # various 'device presence' options. DEV_BPF opt_bpf.h DEV_ISA opt_isa.h DEV_MCA opt_mca.h DEV_SPLASH opt_splash.h EISA_SLOTS opt_eisa.h # ed driver ED_NO_MIIBUS opt_ed.h # wi driver WI_SYMBOL_FIRMWARE opt_wi.h # XXX bogusly global. DEVICE_POLLING opt_global.h SOCKBUF_DEBUG opt_global.h # options for ubsec driver UBSEC_DEBUG opt_ubsec.h UBSEC_RNDTEST opt_ubsec.h UBSEC_NO_RNG opt_ubsec.h # options for hifn driver HIFN_DEBUG opt_hifn.h HIFN_RNDTEST opt_hifn.h # options for safenet driver SAFE_DEBUG opt_safe.h SAFE_NO_RNG opt_safe.h SAFE_RNDTEST opt_safe.h # syscons options MAXCONS opt_syscons.h SC_ALT_MOUSE_IMAGE opt_syscons.h SC_CUT_SPACES2TABS opt_syscons.h SC_CUT_SEPCHARS opt_syscons.h SC_DEBUG_LEVEL opt_syscons.h SC_DFLT_FONT opt_syscons.h SC_DISABLE_DDBKEY opt_syscons.h SC_DISABLE_REBOOT opt_syscons.h SC_HISTORY_SIZE opt_syscons.h SC_KERNEL_CONS_ATTR opt_syscons.h SC_KERNEL_CONS_REV_ATTR opt_syscons.h SC_MOUSE_CHAR opt_syscons.h SC_NO_CUTPASTE opt_syscons.h SC_NO_FONT_LOADING opt_syscons.h SC_NO_HISTORY opt_syscons.h SC_NO_SUSPEND_VTYSWITCH opt_syscons.h SC_NO_SYSMOUSE opt_syscons.h SC_NORM_ATTR opt_syscons.h SC_NORM_REV_ATTR opt_syscons.h SC_PIXEL_MODE opt_syscons.h SC_RENDER_DEBUG opt_syscons.h SC_TWOBUTTON_MOUSE opt_syscons.h # kbd options KBD_DISABLE_KEYMAP_LOAD opt_kbd.h KBD_INSTALL_CDEV opt_kbd.h KBD_MAXRETRY opt_kbd.h KBD_MAXWAIT opt_kbd.h KBD_RESETDELAY opt_kbd.h KBDIO_DEBUG opt_kbd.h # options for the Atheros HAL (only useful with source code) AH_SUPPORT_AR5210 opt_ah.h AH_SUPPORT_AR5211 opt_ah.h AH_SUPPORT_AR5212 opt_ah.h AH_DEBUG opt_ah.h AH_DEBUG_ALQ opt_ah.h AH_ASSERT opt_ah.h # dcons options DCONS_BUF_SIZE opt_dcons.h DCONS_POLL_HZ opt_dcons.h DCONS_FORCE_CONSOLE opt_dcons.h DCONS_FORCE_GDB opt_dcons.h Index: head/sys/kern/subr_sleepqueue.c =================================================================== --- head/sys/kern/subr_sleepqueue.c (revision 131258) +++ head/sys/kern/subr_sleepqueue.c (revision 131259) @@ -1,795 +1,836 @@ /* * Copyright (c) 2004 John Baldwin * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the author nor the names of any co-contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ /* * Implementation of sleep queues used to hold queue of threads blocked on * a wait channel. Sleep queues different from turnstiles in that wait * channels are not owned by anyone, so there is no priority propagation. * Sleep queues can also provide a timeout and can also be interrupted by * signals. That said, there are several similarities between the turnstile * and sleep queue implementations. (Note: turnstiles were implemented * first.) For example, both use a hash table of the same size where each * bucket is referred to as a "chain" that contains both a spin lock and * a linked list of queues. An individual queue is located by using a hash * to pick a chain, locking the chain, and then walking the chain searching * for the queue. This means that a wait channel object does not need to * embed it's queue head just as locks do not embed their turnstile queue * head. Threads also carry around a sleep queue that they lend to the * wait channel when blocking. Just as in turnstiles, the queue includes * a free list of the sleep queues of other threads blocked on the same * wait channel in the case of multiple waiters. * * Some additional functionality provided by sleep queues include the * ability to set a timeout. The timeout is managed using a per-thread * callout that resumes a thread if it is asleep. A thread may also * catch signals while it is asleep (aka an interruptible sleep). The * signal code uses sleepq_abort() to interrupt a sleeping thread. Finally, * sleep queues also provide some extra assertions. One is not allowed to * mix the sleep/wakeup and cv APIs for a given wait channel. Also, one * must consistently use the same lock to synchronize with a wait channel, * though this check is currently only a warning for sleep/wakeup due to * pre-existing abuse of that API. The same lock must also be held when * awakening threads, though that is currently only enforced for condition * variables. */ +#include "opt_sleepqueue_profiling.h" + #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include +#include /* * Constants for the hash table of sleep queue chains. These constants are * the same ones that 4BSD (and possibly earlier versions of BSD) used. * Basically, we ignore the lower 8 bits of the address since most wait * channel pointers are aligned and only look at the next 7 bits for the * hash. SC_TABLESIZE must be a power of two for SC_MASK to work properly. */ #define SC_TABLESIZE 128 /* Must be power of 2. */ #define SC_MASK (SC_TABLESIZE - 1) #define SC_SHIFT 8 #define SC_HASH(wc) (((uintptr_t)(wc) >> SC_SHIFT) & SC_MASK) #define SC_LOOKUP(wc) &sleepq_chains[SC_HASH(wc)] /* * There two different lists of sleep queues. Both lists are connected * via the sq_hash entries. The first list is the sleep queue chain list * that a sleep queue is on when it is attached to a wait channel. The * second list is the free list hung off of a sleep queue that is attached * to a wait channel. * * Each sleep queue also contains the wait channel it is attached to, the * list of threads blocked on that wait channel, flags specific to the * wait channel, and the lock used to synchronize with a wait channel. * The flags are used to catch mismatches between the various consumers * of the sleep queue API (e.g. sleep/wakeup and condition variables). * The lock pointer is only used when invariants are enabled for various * debugging checks. * * Locking key: * c - sleep queue chain lock */ struct sleepqueue { TAILQ_HEAD(, thread) sq_blocked; /* (c) Blocked threads. */ LIST_ENTRY(sleepqueue) sq_hash; /* (c) Chain and free list. */ LIST_HEAD(, sleepqueue) sq_free; /* (c) Free queues. */ void *sq_wchan; /* (c) Wait channel. */ int sq_flags; /* (c) Flags. */ #ifdef INVARIANTS struct mtx *sq_lock; /* (c) Associated lock. */ #endif }; struct sleepqueue_chain { LIST_HEAD(, sleepqueue) sc_queues; /* List of sleep queues. */ struct mtx sc_lock; /* Spin lock for this chain. */ +#ifdef SLEEPQUEUE_PROFILING + u_int sc_depth; /* Length of sc_queues. */ + u_int sc_max_depth; /* Max length of sc_queues. */ +#endif }; +#ifdef SLEEPQUEUE_PROFILING +u_int sleepq_max_depth; +SYSCTL_NODE(_debug, OID_AUTO, sleepq, CTLFLAG_RD, 0, "sleepq profiling"); +SYSCTL_NODE(_debug_sleepq, OID_AUTO, chains, CTLFLAG_RD, 0, + "sleepq chain stats"); +SYSCTL_UINT(_debug_sleepq, OID_AUTO, max_depth, CTLFLAG_RD, &sleepq_max_depth, + 0, "maxmimum depth achieved of a single chain"); +#endif static struct sleepqueue_chain sleepq_chains[SC_TABLESIZE]; MALLOC_DEFINE(M_SLEEPQUEUE, "sleep queues", "sleep queues"); /* * Prototypes for non-exported routines. */ static int sleepq_check_timeout(void); static void sleepq_switch(void *wchan); static void sleepq_timeout(void *arg); static void sleepq_remove_thread(struct sleepqueue *sq, struct thread *td); static void sleepq_resume_thread(struct thread *td, int pri); /* * Early initialization of sleep queues that is called from the sleepinit() * SYSINIT. */ void init_sleepqueues(void) { +#ifdef SLEEPQUEUE_PROFILING + struct sysctl_oid *chain_oid; + char chain_name[10]; +#endif int i; for (i = 0; i < SC_TABLESIZE; i++) { LIST_INIT(&sleepq_chains[i].sc_queues); mtx_init(&sleepq_chains[i].sc_lock, "sleepq chain", NULL, MTX_SPIN); +#ifdef SLEEPQUEUE_PROFILING + snprintf(chain_name, sizeof(chain_name), "%d", i); + chain_oid = SYSCTL_ADD_NODE(NULL, + SYSCTL_STATIC_CHILDREN(_debug_sleepq_chains), OID_AUTO, + chain_name, CTLFLAG_RD, NULL, "sleepq chain stats"); + SYSCTL_ADD_UINT(NULL, SYSCTL_CHILDREN(chain_oid), OID_AUTO, + "depth", CTLFLAG_RD, &sleepq_chains[i].sc_depth, 0, NULL); + SYSCTL_ADD_UINT(NULL, SYSCTL_CHILDREN(chain_oid), OID_AUTO, + "max_depth", CTLFLAG_RD, &sleepq_chains[i].sc_max_depth, 0, + NULL); +#endif } thread0.td_sleepqueue = sleepq_alloc(); } /* * Malloc and initialize a new sleep queue for a new thread. */ struct sleepqueue * sleepq_alloc(void) { struct sleepqueue *sq; sq = malloc(sizeof(struct sleepqueue), M_SLEEPQUEUE, M_WAITOK | M_ZERO); TAILQ_INIT(&sq->sq_blocked); LIST_INIT(&sq->sq_free); return (sq); } /* * Free a sleep queue when a thread is destroyed. */ void sleepq_free(struct sleepqueue *sq) { MPASS(sq != NULL); MPASS(TAILQ_EMPTY(&sq->sq_blocked)); free(sq, M_SLEEPQUEUE); } /* * Look up the sleep queue associated with a given wait channel in the hash * table locking the associated sleep queue chain. Return holdind the sleep * queue chain lock. If no queue is found in the table, NULL is returned. */ struct sleepqueue * sleepq_lookup(void *wchan) { struct sleepqueue_chain *sc; struct sleepqueue *sq; KASSERT(wchan != NULL, ("%s: invalid NULL wait channel", __func__)); sc = SC_LOOKUP(wchan); mtx_lock_spin(&sc->sc_lock); LIST_FOREACH(sq, &sc->sc_queues, sq_hash) if (sq->sq_wchan == wchan) return (sq); return (NULL); } /* * Unlock the sleep queue chain associated with a given wait channel. */ void sleepq_release(void *wchan) { struct sleepqueue_chain *sc; sc = SC_LOOKUP(wchan); mtx_unlock_spin(&sc->sc_lock); } /* * Places the current thread on the sleepqueue for the specified wait * channel. If INVARIANTS is enabled, then it associates the passed in * lock with the sleepq to make sure it is held when that sleep queue is * woken up. */ void sleepq_add(struct sleepqueue *sq, void *wchan, struct mtx *lock, const char *wmesg, int flags) { struct sleepqueue_chain *sc; struct thread *td, *td1; td = curthread; sc = SC_LOOKUP(wchan); mtx_assert(&sc->sc_lock, MA_OWNED); MPASS(td->td_sleepqueue != NULL); MPASS(wchan != NULL); /* If the passed in sleep queue is NULL, use this thread's queue. */ if (sq == NULL) { +#ifdef SLEEPQUEUE_PROFILING + sc->sc_depth++; + if (sc->sc_depth > sc->sc_max_depth) { + sc->sc_max_depth = sc->sc_depth; + if (sc->sc_max_depth > sleepq_max_depth) + sleepq_max_depth = sc->sc_max_depth; + } +#endif sq = td->td_sleepqueue; LIST_INSERT_HEAD(&sc->sc_queues, sq, sq_hash); KASSERT(TAILQ_EMPTY(&sq->sq_blocked), ("thread's sleep queue has a non-empty queue")); KASSERT(LIST_EMPTY(&sq->sq_free), ("thread's sleep queue has a non-empty free list")); KASSERT(sq->sq_wchan == NULL, ("stale sq_wchan pointer")); sq->sq_wchan = wchan; #ifdef INVARIANTS sq->sq_lock = lock; #endif sq->sq_flags = flags; TAILQ_INSERT_TAIL(&sq->sq_blocked, td, td_slpq); } else { MPASS(wchan == sq->sq_wchan); MPASS(lock == sq->sq_lock); TAILQ_FOREACH(td1, &sq->sq_blocked, td_slpq) if (td1->td_priority > td->td_priority) break; if (td1 != NULL) TAILQ_INSERT_BEFORE(td1, td, td_slpq); else TAILQ_INSERT_TAIL(&sq->sq_blocked, td, td_slpq); LIST_INSERT_HEAD(&sq->sq_free, td->td_sleepqueue, sq_hash); } td->td_sleepqueue = NULL; mtx_lock_spin(&sched_lock); td->td_wchan = wchan; td->td_wmesg = wmesg; mtx_unlock_spin(&sched_lock); } /* * Sets a timeout that will remove the current thread from the specified * sleep queue after timo ticks if the thread has not already been awakened. */ void sleepq_set_timeout(void *wchan, int timo) { struct sleepqueue_chain *sc; struct thread *td; td = curthread; sc = SC_LOOKUP(wchan); mtx_assert(&sc->sc_lock, MA_OWNED); MPASS(TD_ON_SLEEPQ(td)); MPASS(td->td_sleepqueue == NULL); MPASS(wchan != NULL); callout_reset(&td->td_slpcallout, timo, sleepq_timeout, td); } /* * Marks the pending sleep of the current thread as interruptible and * makes an initial check for pending signals before putting a thread * to sleep. */ int sleepq_catch_signals(void *wchan) { struct sleepqueue_chain *sc; struct sleepqueue *sq; struct thread *td; struct proc *p; int do_upcall; int sig; do_upcall = 0; td = curthread; p = td->td_proc; sc = SC_LOOKUP(wchan); mtx_assert(&sc->sc_lock, MA_OWNED); MPASS(td->td_sleepqueue == NULL); MPASS(wchan != NULL); CTR3(KTR_PROC, "sleepq catching signals: thread %p (pid %ld, %s)", (void *)td, (long)p->p_pid, p->p_comm); /* Mark thread as being in an interruptible sleep. */ mtx_lock_spin(&sched_lock); MPASS(TD_ON_SLEEPQ(td)); td->td_flags |= TDF_SINTR; mtx_unlock_spin(&sched_lock); sleepq_release(wchan); /* See if there are any pending signals for this thread. */ PROC_LOCK(p); mtx_lock(&p->p_sigacts->ps_mtx); sig = cursig(td); mtx_unlock(&p->p_sigacts->ps_mtx); if (sig == 0 && thread_suspend_check(1)) sig = SIGSTOP; else do_upcall = thread_upcall_check(td); PROC_UNLOCK(p); /* * If there were pending signals and this thread is still on * the sleep queue, remove it from the sleep queue. */ sq = sleepq_lookup(wchan); mtx_lock_spin(&sched_lock); if (TD_ON_SLEEPQ(td) && (sig != 0 || do_upcall != 0)) { mtx_unlock_spin(&sched_lock); sleepq_remove_thread(sq, td); } else mtx_unlock_spin(&sched_lock); return (sig); } /* * Switches to another thread if we are still asleep on a sleep queue and * drop the lock on the sleepqueue chain. Returns with sched_lock held. */ static void sleepq_switch(void *wchan) { struct sleepqueue_chain *sc; struct thread *td; td = curthread; sc = SC_LOOKUP(wchan); mtx_assert(&sc->sc_lock, MA_OWNED); /* * If we have a sleep queue, then we've already been woken up, so * just return. */ if (td->td_sleepqueue != NULL) { MPASS(!TD_ON_SLEEPQ(td)); mtx_unlock_spin(&sc->sc_lock); mtx_lock_spin(&sched_lock); return; } /* * Otherwise, actually go to sleep. */ mtx_lock_spin(&sched_lock); mtx_unlock_spin(&sc->sc_lock); sched_sleep(td); TD_SET_SLEEPING(td); mi_switch(SW_VOL); KASSERT(TD_IS_RUNNING(td), ("running but not TDS_RUNNING")); CTR3(KTR_PROC, "sleepq resume: thread %p (pid %ld, %s)", (void *)td, (long)td->td_proc->p_pid, (void *)td->td_proc->p_comm); } /* * Check to see if we timed out. */ static int sleepq_check_timeout(void) { struct thread *td; mtx_assert(&sched_lock, MA_OWNED); td = curthread; /* * If TDF_TIMEOUT is set, we timed out. */ if (td->td_flags & TDF_TIMEOUT) { td->td_flags &= ~TDF_TIMEOUT; return (EWOULDBLOCK); } /* * If TDF_TIMOFAIL is set, the timeout ran after we had * already been woken up. */ if (td->td_flags & TDF_TIMOFAIL) td->td_flags &= ~TDF_TIMOFAIL; /* * If callout_stop() fails, then the timeout is running on * another CPU, so synchronize with it to avoid having it * accidentally wake up a subsequent sleep. */ else if (callout_stop(&td->td_slpcallout) == 0) { td->td_flags |= TDF_TIMEOUT; TD_SET_SLEEPING(td); mi_switch(SW_INVOL); } return (0); } /* * Check to see if we were awoken by a signal. */ static int sleepq_check_signals(void) { struct thread *td; mtx_assert(&sched_lock, MA_OWNED); td = curthread; /* We are no longer in an interruptible sleep. */ td->td_flags &= ~TDF_SINTR; if (td->td_flags & TDF_INTERRUPT) return (td->td_intrval); return (0); } /* * If we were in an interruptible sleep and we weren't interrupted and * didn't timeout, check to see if there are any pending signals and * which return value we should use if so. The return value from an * earlier call to sleepq_catch_signals() should be passed in as the * argument. */ int sleepq_calc_signal_retval(int sig) { struct thread *td; struct proc *p; int rval; td = curthread; p = td->td_proc; PROC_LOCK(p); mtx_lock(&p->p_sigacts->ps_mtx); /* XXX: Should we always be calling cursig()? */ if (sig == 0) sig = cursig(td); if (sig != 0) { if (SIGISMEMBER(p->p_sigacts->ps_sigintr, sig)) rval = EINTR; else rval = ERESTART; } else rval = 0; mtx_unlock(&p->p_sigacts->ps_mtx); PROC_UNLOCK(p); return (rval); } /* * Block the current thread until it is awakened from its sleep queue. */ void sleepq_wait(void *wchan) { sleepq_switch(wchan); mtx_unlock_spin(&sched_lock); } /* * Block the current thread until it is awakened from its sleep queue * or it is interrupted by a signal. */ int sleepq_wait_sig(void *wchan) { int rval; sleepq_switch(wchan); rval = sleepq_check_signals(); mtx_unlock_spin(&sched_lock); return (rval); } /* * Block the current thread until it is awakened from its sleep queue * or it times out while waiting. */ int sleepq_timedwait(void *wchan) { int rval; sleepq_switch(wchan); rval = sleepq_check_timeout(); mtx_unlock_spin(&sched_lock); return (rval); } /* * Block the current thread until it is awakened from its sleep queue, * it is interrupted by a signal, or it times out waiting to be awakened. */ int sleepq_timedwait_sig(void *wchan, int signal_caught) { int rvalt, rvals; sleepq_switch(wchan); rvalt = sleepq_check_timeout(); rvals = sleepq_check_signals(); mtx_unlock_spin(&sched_lock); if (signal_caught || rvalt == 0) return (rvals); else return (rvalt); } /* * Removes a thread from a sleep queue. */ static void sleepq_remove_thread(struct sleepqueue *sq, struct thread *td) { struct sleepqueue_chain *sc; MPASS(td != NULL); MPASS(sq->sq_wchan != NULL); MPASS(td->td_wchan == sq->sq_wchan); sc = SC_LOOKUP(sq->sq_wchan); mtx_assert(&sc->sc_lock, MA_OWNED); /* Remove the thread from the queue. */ TAILQ_REMOVE(&sq->sq_blocked, td, td_slpq); /* * Get a sleep queue for this thread. If this is the last waiter, * use the queue itself and take it out of the chain, otherwise, * remove a queue from the free list. */ if (LIST_EMPTY(&sq->sq_free)) { td->td_sleepqueue = sq; #ifdef INVARIANTS sq->sq_wchan = NULL; +#endif +#ifdef SLEEPQUEUE_PROFILING + sc->sc_depth--; #endif } else td->td_sleepqueue = LIST_FIRST(&sq->sq_free); LIST_REMOVE(td->td_sleepqueue, sq_hash); mtx_lock_spin(&sched_lock); td->td_wmesg = NULL; td->td_wchan = NULL; mtx_unlock_spin(&sched_lock); } /* * Resumes a thread that was asleep on a queue. */ static void sleepq_resume_thread(struct thread *td, int pri) { /* * Note that thread td might not be sleeping if it is running * sleepq_catch_signals() on another CPU or is blocked on * its proc lock to check signals. It doesn't hurt to clear * the sleeping flag if it isn't set though, so we just always * do it. However, we can't assert that it is set. */ mtx_lock_spin(&sched_lock); CTR3(KTR_PROC, "sleepq_wakeup: thread %p (pid %ld, %s)", (void *)td, (long)td->td_proc->p_pid, td->td_proc->p_comm); TD_CLR_SLEEPING(td); /* Adjust priority if requested. */ MPASS(pri == -1 || (pri >= PRI_MIN && pri <= PRI_MAX)); if (pri != -1 && td->td_priority > pri) td->td_priority = pri; setrunnable(td); mtx_unlock_spin(&sched_lock); } /* * Find the highest priority thread sleeping on a wait channel and resume it. */ void sleepq_signal(void *wchan, int flags, int pri) { struct sleepqueue *sq; struct thread *td; CTR2(KTR_PROC, "sleepq_signal(%p, %d)", wchan, flags); KASSERT(wchan != NULL, ("%s: invalid NULL wait channel", __func__)); sq = sleepq_lookup(wchan); if (sq == NULL) { sleepq_release(wchan); return; } KASSERT(sq->sq_flags == flags, ("%s: mismatch between sleep/wakeup and cv_*", __func__)); /* XXX: Do for all sleep queues eventually. */ if (flags & SLEEPQ_CONDVAR) mtx_assert(sq->sq_lock, MA_OWNED); /* Remove first thread from queue and awaken it. */ td = TAILQ_FIRST(&sq->sq_blocked); sleepq_remove_thread(sq, td); sleepq_release(wchan); sleepq_resume_thread(td, pri); } /* * Resume all threads sleeping on a specified wait channel. */ void sleepq_broadcast(void *wchan, int flags, int pri) { TAILQ_HEAD(, thread) list; struct sleepqueue *sq; struct thread *td; CTR2(KTR_PROC, "sleepq_broadcast(%p, %d)", wchan, flags); KASSERT(wchan != NULL, ("%s: invalid NULL wait channel", __func__)); sq = sleepq_lookup(wchan); if (sq == NULL) { sleepq_release(wchan); return; } KASSERT(sq->sq_flags == flags, ("%s: mismatch between sleep/wakeup and cv_*", __func__)); /* XXX: Do for all sleep queues eventually. */ if (flags & SLEEPQ_CONDVAR) mtx_assert(sq->sq_lock, MA_OWNED); /* Move blocked threads from the sleep queue to a temporary list. */ TAILQ_INIT(&list); while (!TAILQ_EMPTY(&sq->sq_blocked)) { td = TAILQ_FIRST(&sq->sq_blocked); sleepq_remove_thread(sq, td); TAILQ_INSERT_TAIL(&list, td, td_slpq); } sleepq_release(wchan); /* Resume all the threads on the temporary list. */ while (!TAILQ_EMPTY(&list)) { td = TAILQ_FIRST(&list); TAILQ_REMOVE(&list, td, td_slpq); sleepq_resume_thread(td, pri); } } /* * Time sleeping threads out. When the timeout expires, the thread is * removed from the sleep queue and made runnable if it is still asleep. */ static void sleepq_timeout(void *arg) { struct sleepqueue *sq; struct thread *td; void *wchan; td = arg; CTR3(KTR_PROC, "sleepq_timeout: thread %p (pid %ld, %s)", (void *)td, (long)td->td_proc->p_pid, (void *)td->td_proc->p_comm); /* * First, see if the thread is asleep and get the wait channel if * it is. */ mtx_lock_spin(&sched_lock); if (TD_ON_SLEEPQ(td)) { wchan = td->td_wchan; mtx_unlock_spin(&sched_lock); sq = sleepq_lookup(wchan); mtx_lock_spin(&sched_lock); } else { wchan = NULL; sq = NULL; } /* * At this point, if the thread is still on the sleep queue, * we have that sleep queue locked as it cannot migrate sleep * queues while we dropped sched_lock. If it had resumed and * was on another CPU while the lock was dropped, it would have * seen that TDF_TIMEOUT and TDF_TIMOFAIL are clear and the * call to callout_stop() to stop this routine would have failed * meaning that it would have already set TDF_TIMEOUT to * synchronize with this function. */ if (TD_ON_SLEEPQ(td)) { MPASS(td->td_wchan == wchan); MPASS(sq != NULL); td->td_flags |= TDF_TIMEOUT; mtx_unlock_spin(&sched_lock); sleepq_remove_thread(sq, td); sleepq_release(wchan); sleepq_resume_thread(td, -1); return; } else if (wchan != NULL) sleepq_release(wchan); /* * Now check for the edge cases. First, if TDF_TIMEOUT is set, * then the other thread has already yielded to us, so clear * the flag and resume it. If TDF_TIMEOUT is not set, then the * we know that the other thread is not on a sleep queue, but it * hasn't resumed execution yet. In that case, set TDF_TIMOFAIL * to let it know that the timeout has already run and doesn't * need to be canceled. */ if (td->td_flags & TDF_TIMEOUT) { MPASS(TD_IS_SLEEPING(td)); td->td_flags &= ~TDF_TIMEOUT; TD_CLR_SLEEPING(td); setrunnable(td); } else td->td_flags |= TDF_TIMOFAIL; mtx_unlock_spin(&sched_lock); } /* * Resumes a specific thread from the sleep queue associated with a specific * wait channel if it is on that queue. */ void sleepq_remove(struct thread *td, void *wchan) { struct sleepqueue *sq; /* * Look up the sleep queue for this wait channel, then re-check * that the thread is asleep on that channel, if it is not, then * bail. */ MPASS(wchan != NULL); sq = sleepq_lookup(wchan); mtx_lock_spin(&sched_lock); if (!TD_ON_SLEEPQ(td) || td->td_wchan != wchan) { mtx_unlock_spin(&sched_lock); sleepq_release(wchan); return; } mtx_unlock_spin(&sched_lock); MPASS(sq != NULL); /* Thread is asleep on sleep queue sq, so wake it up. */ sleepq_remove_thread(sq, td); sleepq_release(wchan); sleepq_resume_thread(td, -1); } /* * Abort a thread as if an interrupt had occurred. Only abort * interruptible waits (unfortunately it isn't safe to abort others). * * XXX: What in the world does the comment below mean? * Also, whatever the signal code does... */ void sleepq_abort(struct thread *td) { void *wchan; mtx_assert(&sched_lock, MA_OWNED); MPASS(TD_ON_SLEEPQ(td)); MPASS(td->td_flags & TDF_SINTR); /* * If the TDF_TIMEOUT flag is set, just leave. A * timeout is scheduled anyhow. */ if (td->td_flags & TDF_TIMEOUT) return; CTR3(KTR_PROC, "sleepq_abort: thread %p (pid %ld, %s)", (void *)td, (long)td->td_proc->p_pid, (void *)td->td_proc->p_comm); wchan = td->td_wchan; mtx_unlock_spin(&sched_lock); sleepq_remove(td, wchan); mtx_lock_spin(&sched_lock); } Index: head/sys/kern/subr_turnstile.c =================================================================== --- head/sys/kern/subr_turnstile.c (revision 131258) +++ head/sys/kern/subr_turnstile.c (revision 131259) @@ -1,721 +1,766 @@ /*- * Copyright (c) 1998 Berkeley Software Design, Inc. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Berkeley Software Design Inc's name may not be used to endorse or * promote products derived from this software without specific prior * written permission. * * THIS SOFTWARE IS PROVIDED BY BERKELEY SOFTWARE DESIGN INC ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL BERKELEY SOFTWARE DESIGN INC BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from BSDI $Id: mutex_witness.c,v 1.1.2.20 2000/04/27 03:10:27 cp Exp $ * and BSDI $Id: synch_machdep.c,v 2.3.2.39 2000/04/27 03:10:25 cp Exp $ */ /* * Implementation of turnstiles used to hold queue of threads blocked on * non-sleepable locks. Sleepable locks use condition variables to * implement their queues. Turnstiles differ from a sleep queue in that * turnstile queue's are assigned to a lock held by an owning thread. Thus, * when one thread is enqueued onto a turnstile, it can lend its priority * to the owning thread. * * We wish to avoid bloating locks with an embedded turnstile and we do not * want to use back-pointers in the locks for the same reason. Thus, we * use a similar approach to that of Solaris 7 as described in Solaris * Internals by Jim Mauro and Richard McDougall. Turnstiles are looked up * in a hash table based on the address of the lock. Each entry in the * hash table is a linked-lists of turnstiles and is called a turnstile * chain. Each chain contains a spin mutex that protects all of the * turnstiles in the chain. * * Each time a thread is created, a turnstile is malloc'd and attached to * that thread. When a thread blocks on a lock, if it is the first thread * to block, it lends its turnstile to the lock. If the lock already has * a turnstile, then it gives its turnstile to the lock's turnstile's free * list. When a thread is woken up, it takes a turnstile from the free list * if there are any other waiters. If it is the only thread blocked on the * lock, then it reclaims the turnstile associated with the lock and removes * it from the hash table. */ +#include "opt_turnstile_profiling.h" + #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include -#include #include +#include +#include /* * Constants for the hash table of turnstile chains. TC_SHIFT is a magic * number chosen because the sleep queue's use the same value for the * shift. Basically, we ignore the lower 8 bits of the address. * TC_TABLESIZE must be a power of two for TC_MASK to work properly. */ #define TC_TABLESIZE 128 /* Must be power of 2. */ #define TC_MASK (TC_TABLESIZE - 1) #define TC_SHIFT 8 #define TC_HASH(lock) (((uintptr_t)(lock) >> TC_SHIFT) & TC_MASK) #define TC_LOOKUP(lock) &turnstile_chains[TC_HASH(lock)] /* * There are three different lists of turnstiles as follows. The list * connected by ts_link entries is a per-thread list of all the turnstiles * attached to locks that we own. This is used to fixup our priority when * a lock is released. The other two lists use the ts_hash entries. The * first of these two is the turnstile chain list that a turnstile is on * when it is attached to a lock. The second list to use ts_hash is the * free list hung off of a turnstile that is attached to a lock. * * Each turnstile contains two lists of threads. The ts_blocked list is * a linked list of threads blocked on the turnstile's lock. The * ts_pending list is a linked list of threads previously awakened by * turnstile_signal() or turnstile_wait() that are waiting to be put on * the run queue. * * Locking key: * c - turnstile chain lock * q - td_contested lock */ struct turnstile { TAILQ_HEAD(, thread) ts_blocked; /* (c + q) Blocked threads. */ TAILQ_HEAD(, thread) ts_pending; /* (c) Pending threads. */ LIST_ENTRY(turnstile) ts_hash; /* (c) Chain and free list. */ LIST_ENTRY(turnstile) ts_link; /* (q) Contested locks. */ LIST_HEAD(, turnstile) ts_free; /* (c) Free turnstiles. */ struct lock_object *ts_lockobj; /* (c) Lock we reference. */ struct thread *ts_owner; /* (c + q) Who owns the lock. */ }; struct turnstile_chain { LIST_HEAD(, turnstile) tc_turnstiles; /* List of turnstiles. */ struct mtx tc_lock; /* Spin lock for this chain. */ +#ifdef TURNSTILE_PROFILING + u_int tc_depth; /* Length of tc_queues. */ + u_int tc_max_depth; /* Max length of tc_queues. */ +#endif }; +#ifdef TURNSTILE_PROFILING +u_int turnstile_max_depth; +SYSCTL_NODE(_debug, OID_AUTO, turnstile, CTLFLAG_RD, 0, "turnstile profiling"); +SYSCTL_NODE(_debug_turnstile, OID_AUTO, chains, CTLFLAG_RD, 0, + "turnstile chain stats"); +SYSCTL_UINT(_debug_turnstile, OID_AUTO, max_depth, CTLFLAG_RD, + &turnstile_max_depth, 0, "maxmimum depth achieved of a single chain"); +#endif static struct mtx td_contested_lock; static struct turnstile_chain turnstile_chains[TC_TABLESIZE]; MALLOC_DEFINE(M_TURNSTILE, "turnstiles", "turnstiles"); /* * Prototypes for non-exported routines. */ static void init_turnstile0(void *dummy); static void propagate_priority(struct thread *); static void turnstile_setowner(struct turnstile *ts, struct thread *owner); /* * Walks the chain of turnstiles and their owners to propagate the priority * of the thread being blocked to all the threads holding locks that have to * release their locks before this thread can run again. */ static void propagate_priority(struct thread *td) { struct turnstile_chain *tc; struct turnstile *ts; struct thread *td1; int pri; mtx_assert(&sched_lock, MA_OWNED); pri = td->td_priority; ts = td->td_blocked; for (;;) { td = ts->ts_owner; if (td == NULL) { /* * This really isn't quite right. Really * ought to bump priority of thread that * next acquires the lock. */ return; } MPASS(td->td_proc != NULL); MPASS(td->td_proc->p_magic == P_MAGIC); /* * XXX: The owner of a turnstile can be stale if it is the * first thread to grab a slock of a sx lock. In that case * it is possible for us to be at SSLEEP or some other * weird state. We should probably just return if the state * isn't SRUN or SLOCK. */ KASSERT(!TD_IS_SLEEPING(td), ("sleeping thread (pid %d) owns a non-sleepable lock", td->td_proc->p_pid)); /* * If this thread already has higher priority than the * thread that is being blocked, we are finished. */ if (td->td_priority <= pri) return; /* * If lock holder is actually running, just bump priority. */ if (TD_IS_RUNNING(td)) { td->td_priority = pri; return; } #ifndef SMP /* * For UP, we check to see if td is curthread (this shouldn't * ever happen however as it would mean we are in a deadlock.) */ KASSERT(td != curthread, ("Deadlock detected")); #endif /* * If on run queue move to new run queue, and quit. * XXXKSE this gets a lot more complicated under threads * but try anyhow. */ if (TD_ON_RUNQ(td)) { MPASS(td->td_blocked == NULL); sched_prio(td, pri); return; } /* * Bump this thread's priority. */ td->td_priority = pri; /* * If we aren't blocked on a lock, we should be. */ KASSERT(TD_ON_LOCK(td), ( "process %d(%s):%d holds %s but isn't blocked on a lock\n", td->td_proc->p_pid, td->td_proc->p_comm, td->td_state, ts->ts_lockobj->lo_name)); /* * Pick up the lock that td is blocked on. */ ts = td->td_blocked; MPASS(ts != NULL); tc = TC_LOOKUP(ts->ts_lockobj); mtx_lock_spin(&tc->tc_lock); /* * This thread may not be blocked on this turnstile anymore * but instead might already be woken up on another CPU * that is waiting on sched_lock in turnstile_unpend() to * finish waking this thread up. We can detect this case * by checking to see if this thread has been given a * turnstile by either turnstile_signal() or * turnstile_broadcast(). In this case, treat the thread as * if it was already running. */ if (td->td_turnstile != NULL) { mtx_unlock_spin(&tc->tc_lock); return; } /* * Check if the thread needs to be moved up on * the blocked chain. It doesn't need to be moved * if it is already at the head of the list or if * the item in front of it still has a higher priority. */ if (td == TAILQ_FIRST(&ts->ts_blocked)) { mtx_unlock_spin(&tc->tc_lock); continue; } td1 = TAILQ_PREV(td, threadqueue, td_lockq); if (td1->td_priority <= pri) { mtx_unlock_spin(&tc->tc_lock); continue; } /* * Remove thread from blocked chain and determine where * it should be moved up to. Since we know that td1 has * a lower priority than td, we know that at least one * thread in the chain has a lower priority and that * td1 will thus not be NULL after the loop. */ mtx_lock_spin(&td_contested_lock); TAILQ_REMOVE(&ts->ts_blocked, td, td_lockq); TAILQ_FOREACH(td1, &ts->ts_blocked, td_lockq) { MPASS(td1->td_proc->p_magic == P_MAGIC); if (td1->td_priority > pri) break; } MPASS(td1 != NULL); TAILQ_INSERT_BEFORE(td1, td, td_lockq); mtx_unlock_spin(&td_contested_lock); CTR4(KTR_LOCK, "propagate_priority: td %p moved before %p on [%p] %s", td, td1, ts->ts_lockobj, ts->ts_lockobj->lo_name); mtx_unlock_spin(&tc->tc_lock); } } /* * Early initialization of turnstiles. This is not done via a SYSINIT() * since this needs to be initialized very early when mutexes are first * initialized. */ void init_turnstiles(void) { +#ifdef TURNSTILE_PROFILING + struct sysctl_oid *chain_oid; + char chain_name[10]; +#endif int i; for (i = 0; i < TC_TABLESIZE; i++) { LIST_INIT(&turnstile_chains[i].tc_turnstiles); mtx_init(&turnstile_chains[i].tc_lock, "turnstile chain", NULL, MTX_SPIN); +#ifdef TURNSTILE_PROFILING + snprintf(chain_name, sizeof(chain_name), "%d", i); + chain_oid = SYSCTL_ADD_NODE(NULL, + SYSCTL_STATIC_CHILDREN(_debug_turnstile_chains), OID_AUTO, + chain_name, CTLFLAG_RD, NULL, "turnstile chain stats"); + SYSCTL_ADD_UINT(NULL, SYSCTL_CHILDREN(chain_oid), OID_AUTO, + "depth", CTLFLAG_RD, &turnstile_chains[i].tc_depth, 0, + NULL); + SYSCTL_ADD_UINT(NULL, SYSCTL_CHILDREN(chain_oid), OID_AUTO, + "max_depth", CTLFLAG_RD, &turnstile_chains[i].tc_max_depth, + 0, NULL); +#endif } mtx_init(&td_contested_lock, "td_contested", NULL, MTX_SPIN); thread0.td_turnstile = NULL; } static void init_turnstile0(void *dummy) { thread0.td_turnstile = turnstile_alloc(); } SYSINIT(turnstile0, SI_SUB_LOCK, SI_ORDER_ANY, init_turnstile0, NULL); /* * Set the owner of the lock this turnstile is attached to. */ static void turnstile_setowner(struct turnstile *ts, struct thread *owner) { mtx_assert(&td_contested_lock, MA_OWNED); MPASS(owner->td_proc->p_magic == P_MAGIC); MPASS(ts->ts_owner == NULL); ts->ts_owner = owner; LIST_INSERT_HEAD(&owner->td_contested, ts, ts_link); } /* * Malloc a turnstile for a new thread, initialize it and return it. */ struct turnstile * turnstile_alloc(void) { struct turnstile *ts; ts = malloc(sizeof(struct turnstile), M_TURNSTILE, M_WAITOK | M_ZERO); TAILQ_INIT(&ts->ts_blocked); TAILQ_INIT(&ts->ts_pending); LIST_INIT(&ts->ts_free); return (ts); } /* * Free a turnstile when a thread is destroyed. */ void turnstile_free(struct turnstile *ts) { MPASS(ts != NULL); MPASS(TAILQ_EMPTY(&ts->ts_blocked)); MPASS(TAILQ_EMPTY(&ts->ts_pending)); free(ts, M_TURNSTILE); } /* * Look up the turnstile for a lock in the hash table locking the associated * turnstile chain along the way. Return with the turnstile chain locked. * If no turnstile is found in the hash table, NULL is returned. */ struct turnstile * turnstile_lookup(struct lock_object *lock) { struct turnstile_chain *tc; struct turnstile *ts; tc = TC_LOOKUP(lock); mtx_lock_spin(&tc->tc_lock); LIST_FOREACH(ts, &tc->tc_turnstiles, ts_hash) if (ts->ts_lockobj == lock) return (ts); return (NULL); } /* * Unlock the turnstile chain associated with a given lock. */ void turnstile_release(struct lock_object *lock) { struct turnstile_chain *tc; tc = TC_LOOKUP(lock); mtx_unlock_spin(&tc->tc_lock); } /* * Take ownership of a turnstile and adjust the priority of the new * owner appropriately. */ void turnstile_claim(struct turnstile *ts) { struct turnstile_chain *tc; struct thread *td, *owner; tc = TC_LOOKUP(ts->ts_lockobj); mtx_assert(&tc->tc_lock, MA_OWNED); owner = curthread; mtx_lock_spin(&td_contested_lock); turnstile_setowner(ts, owner); mtx_unlock_spin(&td_contested_lock); td = TAILQ_FIRST(&ts->ts_blocked); MPASS(td != NULL); MPASS(td->td_proc->p_magic == P_MAGIC); mtx_unlock_spin(&tc->tc_lock); /* * Update the priority of the new owner if needed. */ mtx_lock_spin(&sched_lock); if (td->td_priority < owner->td_priority) owner->td_priority = td->td_priority; mtx_unlock_spin(&sched_lock); } /* * Block the current thread on the turnstile ts. This function will context * switch and not return until this thread has been woken back up. This * function must be called with the appropriate turnstile chain locked and * will return with it unlocked. */ void turnstile_wait(struct turnstile *ts, struct lock_object *lock, struct thread *owner) { struct turnstile_chain *tc; struct thread *td, *td1; td = curthread; tc = TC_LOOKUP(lock); mtx_assert(&tc->tc_lock, MA_OWNED); MPASS(td->td_turnstile != NULL); MPASS(owner != NULL); MPASS(owner->td_proc->p_magic == P_MAGIC); /* If the passed in turnstile is NULL, use this thread's turnstile. */ if (ts == NULL) { +#ifdef TURNSTILE_PROFILING + tc->tc_depth++; + if (tc->tc_depth > tc->tc_max_depth) { + tc->tc_max_depth = tc->tc_depth; + if (tc->tc_max_depth > turnstile_max_depth) + turnstile_max_depth = tc->tc_max_depth; + } +#endif ts = td->td_turnstile; LIST_INSERT_HEAD(&tc->tc_turnstiles, ts, ts_hash); KASSERT(TAILQ_EMPTY(&ts->ts_pending), ("thread's turnstile has pending threads")); KASSERT(TAILQ_EMPTY(&ts->ts_blocked), ("thread's turnstile has a non-empty queue")); KASSERT(LIST_EMPTY(&ts->ts_free), ("thread's turnstile has a non-empty free list")); KASSERT(ts->ts_lockobj == NULL, ("stale ts_lockobj pointer")); ts->ts_lockobj = lock; mtx_lock_spin(&td_contested_lock); TAILQ_INSERT_TAIL(&ts->ts_blocked, td, td_lockq); turnstile_setowner(ts, owner); mtx_unlock_spin(&td_contested_lock); } else { TAILQ_FOREACH(td1, &ts->ts_blocked, td_lockq) if (td1->td_priority > td->td_priority) break; mtx_lock_spin(&td_contested_lock); if (td1 != NULL) TAILQ_INSERT_BEFORE(td1, td, td_lockq); else TAILQ_INSERT_TAIL(&ts->ts_blocked, td, td_lockq); mtx_unlock_spin(&td_contested_lock); MPASS(td->td_turnstile != NULL); LIST_INSERT_HEAD(&ts->ts_free, td->td_turnstile, ts_hash); MPASS(owner == ts->ts_owner); } td->td_turnstile = NULL; mtx_unlock_spin(&tc->tc_lock); mtx_lock_spin(&sched_lock); /* * Handle race condition where a thread on another CPU that owns * lock 'lock' could have woken us in between us dropping the * turnstile chain lock and acquiring the sched_lock. */ if (td->td_flags & TDF_TSNOBLOCK) { td->td_flags &= ~TDF_TSNOBLOCK; mtx_unlock_spin(&sched_lock); return; } #ifdef notyet /* * If we're borrowing an interrupted thread's VM context, we * must clean up before going to sleep. */ if (td->td_ithd != NULL) { struct ithd *it = td->td_ithd; if (it->it_interrupted) { if (LOCK_LOG_TEST(lock, 0)) CTR3(KTR_LOCK, "%s: %p interrupted %p", __func__, it, it->it_interrupted); intr_thd_fixup(it); } } #endif /* Save who we are blocked on and switch. */ td->td_blocked = ts; td->td_lockname = lock->lo_name; TD_SET_LOCK(td); propagate_priority(td); if (LOCK_LOG_TEST(lock, 0)) CTR4(KTR_LOCK, "%s: td %p blocked on [%p] %s", __func__, td, lock, lock->lo_name); mi_switch(SW_VOL); if (LOCK_LOG_TEST(lock, 0)) CTR4(KTR_LOCK, "%s: td %p free from blocked on [%p] %s", __func__, td, lock, lock->lo_name); mtx_unlock_spin(&sched_lock); } /* * Pick the highest priority thread on this turnstile and put it on the * pending list. This must be called with the turnstile chain locked. */ int turnstile_signal(struct turnstile *ts) { struct turnstile_chain *tc; struct thread *td; int empty; MPASS(ts != NULL); MPASS(curthread->td_proc->p_magic == P_MAGIC); MPASS(ts->ts_owner == curthread); tc = TC_LOOKUP(ts->ts_lockobj); mtx_assert(&tc->tc_lock, MA_OWNED); /* * Pick the highest priority thread blocked on this lock and * move it to the pending list. */ td = TAILQ_FIRST(&ts->ts_blocked); MPASS(td->td_proc->p_magic == P_MAGIC); mtx_lock_spin(&td_contested_lock); TAILQ_REMOVE(&ts->ts_blocked, td, td_lockq); mtx_unlock_spin(&td_contested_lock); TAILQ_INSERT_TAIL(&ts->ts_pending, td, td_lockq); /* * If the turnstile is now empty, remove it from its chain and * give it to the about-to-be-woken thread. Otherwise take a * turnstile from the free list and give it to the thread. */ empty = TAILQ_EMPTY(&ts->ts_blocked); - if (empty) + if (empty) { MPASS(LIST_EMPTY(&ts->ts_free)); - else +#ifdef TURNSTILE_PROFILING + tc->tc_depth--; +#endif + } else ts = LIST_FIRST(&ts->ts_free); MPASS(ts != NULL); LIST_REMOVE(ts, ts_hash); td->td_turnstile = ts; return (empty); } /* * Put all blocked threads on the pending list. This must be called with * the turnstile chain locked. */ void turnstile_broadcast(struct turnstile *ts) { struct turnstile_chain *tc; struct turnstile *ts1; struct thread *td; MPASS(ts != NULL); MPASS(curthread->td_proc->p_magic == P_MAGIC); MPASS(ts->ts_owner == curthread); tc = TC_LOOKUP(ts->ts_lockobj); mtx_assert(&tc->tc_lock, MA_OWNED); /* * Transfer the blocked list to the pending list. */ mtx_lock_spin(&td_contested_lock); TAILQ_CONCAT(&ts->ts_pending, &ts->ts_blocked, td_lockq); mtx_unlock_spin(&td_contested_lock); /* * Give a turnstile to each thread. The last thread gets * this turnstile. */ TAILQ_FOREACH(td, &ts->ts_pending, td_lockq) { if (LIST_EMPTY(&ts->ts_free)) { MPASS(TAILQ_NEXT(td, td_lockq) == NULL); ts1 = ts; +#ifdef TURNSTILE_PROFILING + tc->tc_depth--; +#endif } else ts1 = LIST_FIRST(&ts->ts_free); MPASS(ts1 != NULL); LIST_REMOVE(ts1, ts_hash); td->td_turnstile = ts1; } } /* * Wakeup all threads on the pending list and adjust the priority of the * current thread appropriately. This must be called with the turnstile * chain locked. */ void turnstile_unpend(struct turnstile *ts) { TAILQ_HEAD( ,thread) pending_threads; struct turnstile_chain *tc; struct thread *td; int cp, pri; MPASS(ts != NULL); MPASS(ts->ts_owner == curthread); tc = TC_LOOKUP(ts->ts_lockobj); mtx_assert(&tc->tc_lock, MA_OWNED); MPASS(!TAILQ_EMPTY(&ts->ts_pending)); /* * Move the list of pending threads out of the turnstile and * into a local variable. */ TAILQ_INIT(&pending_threads); TAILQ_CONCAT(&pending_threads, &ts->ts_pending, td_lockq); #ifdef INVARIANTS if (TAILQ_EMPTY(&ts->ts_blocked)) ts->ts_lockobj = NULL; #endif /* * Remove the turnstile from this thread's list of contested locks * since this thread doesn't own it anymore. New threads will * not be blocking on the turnstile until it is claimed by a new * owner. */ mtx_lock_spin(&td_contested_lock); ts->ts_owner = NULL; LIST_REMOVE(ts, ts_link); mtx_unlock_spin(&td_contested_lock); mtx_unlock_spin(&tc->tc_lock); /* * Adjust the priority of curthread based on other contested * locks it owns. Don't lower the priority below the base * priority however. */ td = curthread; pri = PRI_MAX; mtx_lock_spin(&sched_lock); mtx_lock_spin(&td_contested_lock); LIST_FOREACH(ts, &td->td_contested, ts_link) { cp = TAILQ_FIRST(&ts->ts_blocked)->td_priority; if (cp < pri) pri = cp; } mtx_unlock_spin(&td_contested_lock); if (pri > td->td_base_pri) pri = td->td_base_pri; td->td_priority = pri; /* * Wake up all the pending threads. If a thread is not blocked * on a lock, then it is currently executing on another CPU in * turnstile_wait() or sitting on a run queue waiting to resume * in turnstile_wait(). Set a flag to force it to try to acquire * the lock again instead of blocking. */ while (!TAILQ_EMPTY(&pending_threads)) { td = TAILQ_FIRST(&pending_threads); TAILQ_REMOVE(&pending_threads, td, td_lockq); MPASS(td->td_proc->p_magic == P_MAGIC); if (TD_ON_LOCK(td)) { td->td_blocked = NULL; td->td_lockname = NULL; TD_CLR_LOCK(td); MPASS(TD_CAN_RUN(td)); setrunqueue(td); } else { td->td_flags |= TDF_TSNOBLOCK; MPASS(TD_IS_RUNNING(td) || TD_ON_RUNQ(td)); } } mtx_unlock_spin(&sched_lock); } /* * Return the first thread in a turnstile. */ struct thread * turnstile_head(struct turnstile *ts) { #ifdef INVARIANTS struct turnstile_chain *tc; MPASS(ts != NULL); tc = TC_LOOKUP(ts->ts_lockobj); mtx_assert(&tc->tc_lock, MA_OWNED); #endif return (TAILQ_FIRST(&ts->ts_blocked)); } /* * Returns true if a turnstile is empty. */ int turnstile_empty(struct turnstile *ts) { #ifdef INVARIANTS struct turnstile_chain *tc; MPASS(ts != NULL); tc = TC_LOOKUP(ts->ts_lockobj); mtx_assert(&tc->tc_lock, MA_OWNED); #endif return (TAILQ_EMPTY(&ts->ts_blocked)); }