Index: head/UPDATING =================================================================== --- head/UPDATING (revision 327232) +++ head/UPDATING (revision 327233) @@ -1,2041 +1,2041 @@ Updating Information for FreeBSD current users. This file is maintained and copyrighted by M. Warner Losh . See end of file for further details. For commonly done items, please see the COMMON ITEMS: section later in the file. These instructions assume that you basically know what you are doing. If not, then please consult the FreeBSD handbook: https://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/current-stable.html Items affecting the ports and packages system can be found in /usr/ports/UPDATING. Please read that file before running portupgrade. NOTE: FreeBSD has switched from gcc to clang. If you have trouble bootstrapping from older versions of FreeBSD, try WITHOUT_CLANG and WITH_GCC to bootstrap to the tip of head, and then rebuild without this option. The bootstrap process from older version of current across the gcc/clang cutover is a bit fragile. NOTE TO PEOPLE WHO THINK THAT FreeBSD 12.x IS SLOW: FreeBSD 12.x has many debugging features turned on, in both the kernel and userland. These features attempt to detect incorrect use of system primitives, and encourage loud failure through extra sanity checking and fail stop semantics. They also substantially impact system performance. If you want to do performance measurement, benchmarking, and optimization, you'll want to turn them off. This includes various WITNESS- related kernel options, INVARIANTS, malloc debugging flags in userland, and various verbose features in the kernel. Many developers choose to disable these features on build machines to maximize performance. (To completely disable malloc debugging, define MALLOC_PRODUCTION in /etc/make.conf, or to merely disable the most expensive debugging functionality run "ln -s 'abort:false,junk:false' /etc/malloc.conf".) ****************************** SPECIAL WARNING: ****************************** Due to a bug in some versions of clang that's very hard to workaround in the upgrade process, to upgrade to -current you must first upgrade either stable/9 after r286035 or stable/10 after r286033 (including 10.3-RELEASE) or current after r286007 (including stable/11 and 11.0-RELEASE). These revisions post-date the 10.2 and 9.3 releases, so you'll need to take the unusual step of upgrading to the tip of the stable branch before moving to 11 or -current via a source upgrade. stable/11 and 11.0-RELEASE have working newer compiler. This differs from the historical situation where one could upgrade from anywhere on the last couple of stable branches, so be careful. If you're running a hybrid system on 9.x or 10.x with an updated clang compiler or are using an supported external toolchain, the build system will allow the upgrade. Otherwise it will print a reminder. ****************************** SPECIAL WARNING: ****************************** 20171215: r326887 fixes the issue described in the 20171214 UPDATING entry. r326888 flips the switch back to building GELI support always. 20171214: r362593 broke ZFS + GELI support for reasons unknown. However, it also broke ZFS support generally, so GELI has been turned off by default as the lesser evil in r326857. If you boot off ZFS and/or GELI, it might not be a good time to update. 20171125: PowerPC users must update loader(8) by rebuilding world before installing a new kernel, as the protocol connecting them has changed. Without the update, loader metadata will not be passed successfully to the kernel and users will have to enter their root partition at the kernel mountroot prompt to continue booting. Newer versions of loader can boot old kernels without issue. 20171110: The LOADER_FIREWIRE_SUPPORT build variable as been renamed to WITH/OUT_LOADER_FIREWIRE. LOADER_{NO_,}GELI_SUPPORT has been renamed to WITH/OUT_LOADER_GELI. 20171106: The naive and non-compliant support of posix_fallocate(2) in ZFS has been removed as of r325320. The system call now returns EINVAL when used on a ZFS file. Although the new behavior complies with the standard, some consumers are not prepared to cope with it. One known victim is lld prior to r325420. 20171102: Building in a FreeBSD src checkout will automatically create object directories now rather than store files in the current directory if 'make obj' was not ran. Calling 'make obj' is no longer necessary. This feature can be disabled by setting WITHOUT_AUTO_OBJ=yes in /etc/src-env.conf (not /etc/src.conf), or passing the option in the environment. 20171101: The default MAKEOBJDIR has changed from /usr/obj/ for native builds, and /usr/obj// for cross-builds, to a unified /usr/obj//. This behavior can be changed to the old format by setting WITHOUT_UNIFIED_OBJDIR=yes in /etc/src-env.conf, the environment, or with -DWITHOUT_UNIFIED_OBJDIR when building. The UNIFIED_OBJDIR option is a transitional feature that will be removed for 12.0 release; please migrate to the new format for any tools by looking up the OBJDIR used by 'make -V .OBJDIR' means rather than hardcoding paths. 20171028: The native-xtools target no longer installs the files by default to the OBJDIR. Use the native-xtools-install target with a DESTDIR to install to ${DESTDIR}/${NXTP} where NXTP defaults to /nxb-bin. 20171021: As part of the boot loader infrastructure cleanup, LOADER_*_SUPPORT options are changing from controlling the build if defined / undefined to controlling the build with explicit 'yes' or 'no' values. They will shift to WITH/WITHOUT options to match other options in the system. 20171010: libstand has turned into a private library for sys/boot use only. It is no longer supported as a public interface outside of sys/boot. 20171005: The arm port has split armv6 into armv6 and armv7. armv7 is now a valid TARGET_ARCH/MACHINE_ARCH setting. If you have an armv7 system and are running a kernel from before r324363, you will need to add MACHINE_ARCH=armv7 to 'make buildworld' to do a native build. 20171003: When building multiple kernels using KERNCONF, non-existent KERNCONF files will produce an error and buildkernel will fail. Previously missing KERNCONF files silently failed giving no indication as to why, only to subsequently discover during installkernel that the desired kernel was never built in the first place. 20170912: The default serial number format for CTL LUNs has changed. This will affect users who use /dev/diskid/* device nodes, or whose FibreChannel or iSCSI clients care about their LUNs' serial numbers. Users who require serial number stability should hardcode serial numbers in /etc/ctl.conf . 20170912: For 32-bit arm compiled for hard-float support, soft-floating point binaries now always get their shared libraries from LD_SOFT_LIBRARY_PATH (in the past, this was only used if /usr/libsoft also existed). Only users with a hard-float ld.so, but soft-float everything else should be affected. 20170826: The geli password typed at boot is now hidden. To restore the previous behavior, see geli(8) for configuration options. 20170825: Move PMTUD blackhole counters to TCPSTATS and remove them from bare sysctl values. Minor nit, but requires a rebuild of both world/kernel to complete. 20170814: "make check" behavior (made in ^/head@r295380) has been changed to execute from a limited sandbox, as opposed to executing from ${TESTSDIR}. Behavioral changes: - The "beforecheck" and "aftercheck" targets are now specified. - ${CHECKDIR} (added in commit noted above) has been removed. - Legacy behavior can be enabled by setting WITHOUT_MAKE_CHECK_USE_SANDBOX in src.conf(5) or the environment. If the limited sandbox mode is enabled, "make check" will execute "make distribution", then install, execute the tests, and clean up the sandbox if successful. The "make distribution" and "make install" targets are typically run as root to set appropriate permissions and ownership at installation time. The end-user should set "WITH_INSTALL_AS_USER" in src.conf(5) or the environment if executing "make check" with limited sandbox mode using an unprivileged user. 20170808: Since the switch to GPT disk labels, fsck for UFS/FFS has been unable to automatically find alternate superblocks. As of r322297, the information needed to find alternate superblocks has been moved to the end of the area reserved for the boot block. Filesystems created with a newfs of this vintage or later will create the recovery information. If you have a filesystem created prior to this change and wish to have a recovery block created for your filesystem, you can do so by running fsck in forground mode (i.e., do not use the -p or -y options). As it starts, fsck will ask ``SAVE DATA TO FIND ALTERNATE SUPERBLOCKS'' to which you should answer yes. 20170728: As of r321665, an NFSv4 server configuration that services Kerberos mounts or clients that do not support the uid/gid in owner/owner_group string capability, must explicitly enable the nfsuserd daemon by adding nfsuserd_enable="YES" to the machine's /etc/rc.conf file. 20170722: Clang, llvm, lldb, compiler-rt and libc++ have been upgraded to 5.0.0. Please see the 20141231 entry below for information about prerequisites and upgrading, if you are not already using clang 3.5.0 or higher. 20170701: WITHOUT_RCMDS is now the default. Set WITH_RCMDS if you need the r-commands (rlogin, rsh, etc.) to be built with the base system. 20170625: The FreeBSD/powerpc platform now uses a 64-bit type for time_t. This is a very major ABI incompatible change, so users of FreeBSD/powerpc must be careful when performing source upgrades. It is best to run 'make installworld' from an alternate root system, either a live CD/memory stick, or a temporary root partition. Additionally, all ports must be recompiled. powerpc64 is largely unaffected, except in the case of 32-bit compatibility. All 32-bit binaries will be affected. 20170623: Forward compatibility for the "ino64" project have been committed. This will allow most new binaries to run on older kernels in a limited fashion. This prevents many of the common foot-shooting actions in the upgrade as well as the limited ability to roll back the kernel across the ino64 upgrade. Complicated use cases may not work properly, though enough simpler ones work to allow recovery in most situations. 20170620: Switch back to the BSDL dtc (Device Tree Compiler). Set WITH_GPL_DTC if you require the GPL compiler. 20170618: The internal ABI used for communication between the NFS kernel modules was changed by r320085, so __FreeBSD_version was bumped to ensure all the NFS related modules are updated together. 20170617: The ABI of struct event was changed by extending the data member to 64bit and adding ext fields. For upgrade, same precautions as for the entry 20170523 "ino64" must be followed. 20170531: The GNU roff toolchain has been removed from base. To render manpages which are not supported by mandoc(1), man(1) can fallback on GNU roff from ports (and recommends to install it). To render roff(7) documents, consider using GNU roff from ports or the heirloom doctools roff toolchain from ports via pkg install groff or via pkg install heirloom-doctools. 20170524: The ath(4) and ath_hal(4) modules now build piecemeal to allow for smaller runtime footprint builds. This is useful for embedded systems which only require one chipset support. If you load it as a module, make sure this is in /boot/loader.conf: if_ath_load="YES" This will load the HAL, all chip/RF backends and if_ath_pci. If you have if_ath_pci in /boot/loader.conf, ensure it is after if_ath or it will not load any HAL chipset support. If you want to selectively load things (eg on ye cheape ARM/MIPS platforms where RAM is at a premium) you should: * load ath_hal * load the chip modules in question * load ath_rate, ath_dfs * load ath_main * load if_ath_pci and/or if_ath_ahb depending upon your particular bus bind type - this is where probe/attach is done. For further comments/feedback, poke adrian@ . 20170523: The "ino64" 64-bit inode project has been committed, which extends a number of types to 64 bits. Upgrading in place requires care and adherence to the documented upgrade procedure. If using a custom kernel configuration ensure that the COMPAT_FREEBSD11 option is included (as during the upgrade the system will be running the ino64 kernel with the existing world). For the safest in-place upgrade begin by removing previous build artifacts via "rm -rf /usr/obj/*". Then, carefully follow the full procedure documented below under the heading "To rebuild everything and install it on the current system." Specifically, a reboot is required after installing the new kernel before installing world. 20170424: The NATM framework including the en(4), fatm(4), hatm(4), and patm(4) devices has been removed. Consumers should plan a migration before the end-of-life date for FreeBSD 11. 20170420: GNU diff has been replaced by a BSD licensed diff. Some features of GNU diff has not been implemented, if those are needed a newer version of GNU diff is available via the diffutils package under the gdiff name. 20170413: As of r316810 for ipfilter, keep frags is no longer assumed when keep state is specified in a rule. r316810 aligns ipfilter with documentation in man pages separating keep frags from keep state. This allows keep state to be specified without forcing keep frags and allows keep frags to be specified independently of keep state. To maintain previous behaviour, also specify keep frags with keep state (as documented in ipf.conf.5). 20170407: arm64 builds now use the base system LLD 4.0.0 linker by default, instead of requiring that the aarch64-binutils port or package be installed. To continue using aarch64-binutils, set CROSS_BINUTILS_PREFIX=/usr/local/aarch64-freebsd/bin . 20170405: The UDP optimization in entry 20160818 that added the sysctl net.inet.udp.require_l2_bcast has been reverted. L2 broadcast packets will no longer be treated as L3 broadcast packets. 20170331: Binds and sends to the loopback addresses, IPv6 and IPv4, will now use any explicitly assigned loopback address available in the jail instead of using the first assigned address of the jail. 20170329: The ctl.ko module no longer implements the iSCSI target frontend: cfiscsi.ko does instead. If building cfiscsi.ko as a kernel module, the module can be loaded via one of the following methods: - `cfiscsi_load="YES"` in loader.conf(5). - Add `cfiscsi` to `$kld_list` in rc.conf(5). - ctladm(8)/ctld(8), when compiled with iSCSI support (`WITH_ISCSI=yes` in src.conf(5)) Please see cfiscsi(4) for more details. 20170316: The mmcsd.ko module now additionally depends on geom_flashmap.ko. Also, mmc.ko and mmcsd.ko need to be a matching pair built from the same source (previously, the dependency of mmcsd.ko on mmc.ko was missing, but mmcsd.ko now will refuse to load if it is incompatible with mmc.ko). 20170315: The syntax of ipfw(8) named states was changed to avoid ambiguity. If you have used named states in the firewall rules, you need to modify them after installworld and before rebooting. Now named states must be prefixed with colon. 20170311: The old drm (sys/dev/drm/) drivers for i915 and radeon have been removed as the userland we provide cannot use them. The KMS version (sys/dev/drm2) supports the same hardware. 20170302: Clang, llvm, lldb, compiler-rt and libc++ have been upgraded to 4.0.0. Please see the 20141231 entry below for information about prerequisites and upgrading, if you are not already using clang 3.5.0 or higher. 20170221: The code that provides support for ZFS .zfs/ directory functionality has been reimplemented. It's not possible now to create a snapshot by mkdir under .zfs/snapshot/. That should be the only user visible change. 20170216: EISA bus support has been removed. The WITH_EISA option is no longer valid. 20170215: MCA bus support has been removed. 20170127: The WITH_LLD_AS_LD / WITHOUT_LLD_AS_LD build knobs have been renamed WITH_LLD_IS_LD / WITHOUT_LLD_IS_LD, for consistency with CLANG_IS_CC. 20170112: The EM_MULTIQUEUE kernel configuration option is deprecated now that the em(4) driver conforms to iflib specifications. 20170109: The igb(4), em(4) and lem(4) ethernet drivers are now implemented via IFLIB. If you have a custom kernel configuration that excludes em(4) but you use igb(4), you need to re-add em(4) to your custom configuration. 20161217: Clang, llvm, lldb, compiler-rt and libc++ have been upgraded to 3.9.1. Please see the 20141231 entry below for information about prerequisites and upgrading, if you are not already using clang 3.5.0 or higher. 20161124: Clang, llvm, lldb, compiler-rt and libc++ have been upgraded to 3.9.0. Please see the 20141231 entry below for information about prerequisites and upgrading, if you are not already using clang 3.5.0 or higher. 20161119: The layout of the pmap structure has changed for powerpc to put the pmap statistics at the front for all CPU variations. libkvm(3) and all tools that link against it need to be recompiled. 20161030: isl(4) and cyapa(4) drivers now require a new driver, chromebook_platform(4), to work properly on Chromebook-class hardware. On other types of hardware the drivers may need to be configured using device hints. Please see the corresponding manual pages for details. 20161017: The urtwn(4) driver was merged into rtwn(4) and now consists of rtwn(4) main module + rtwn_usb(4) and rtwn_pci(4) bus-specific parts. Also, firmware for RTL8188CE was renamed due to possible name conflict (rtwnrtl8192cU(B) -> rtwnrtl8192cE(B)) 20161015: GNU rcs has been removed from base. It is available as packages: - rcs: Latest GPLv3 GNU rcs version. - rcs57: Copy of the latest version of GNU rcs (GPLv2) before it was removed from base. 20161008: Use of the cc_cdg, cc_chd, cc_hd, or cc_vegas congestion control modules now requires that the kernel configuration contain the TCP_HHOOK option. (This option is included in the GENERIC kernel.) 20161003: The WITHOUT_ELFCOPY_AS_OBJCOPY src.conf(5) knob has been retired. ELF Tool Chain's elfcopy is always installed as /usr/bin/objcopy. 20160924: Relocatable object files with the extension of .So have been renamed to use an extension of .pico instead. The purpose of this change is to avoid a name clash with shared libraries on case-insensitive file systems. On those file systems, foo.So is the same file as foo.so. 20160918: GNU rcs has been turned off by default. It can (temporarily) be built again by adding WITH_RCS knob in src.conf. Otherwise, GNU rcs is available from packages: - rcs: Latest GPLv3 GNU rcs version. - rcs57: Copy of the latest version of GNU rcs (GPLv2) from base. 20160918: The backup_uses_rcs functionality has been removed from rc.subr. 20160908: The queue(3) debugging macro, QUEUE_MACRO_DEBUG, has been split into two separate components, QUEUE_MACRO_DEBUG_TRACE and QUEUE_MACRO_DEBUG_TRASH. Define both for the original QUEUE_MACRO_DEBUG behavior. 20160824: r304787 changed some ioctl interfaces between the iSCSI userspace programs and the kernel. ctladm, ctld, iscsictl, and iscsid must be rebuilt to work with new kernels. __FreeBSD_version has been bumped to 1200005. 20160818: The UDP receive code has been updated to only treat incoming UDP packets that were addressed to an L2 broadcast address as L3 broadcast packets. It is not expected that this will affect any standards-conforming UDP application. The new behaviour can be disabled by setting the sysctl net.inet.udp.require_l2_bcast to 0. 20160818: Remove the openbsd_poll system call. __FreeBSD_version has been bumped because of this. 20160622: The libc stub for the pipe(2) system call has been replaced with a wrapper that calls the pipe2(2) system call and the pipe(2) system call is now only implemented by the kernels that include "options COMPAT_FREEBSD10" in their config file (this is the default). Users should ensure that this option is enabled in their kernel or upgrade userspace to r302092 before upgrading their kernel. 20160527: CAM will now strip leading spaces from SCSI disks' serial numbers. This will affect users who create UFS filesystems on SCSI disks using those disk's diskid device nodes. For example, if /etc/fstab previously contained a line like "/dev/diskid/DISK-%20%20%20%20%20%20%20ABCDEFG0123456", you should change it to "/dev/diskid/DISK-ABCDEFG0123456". Users of geom transforms like gmirror may also be affected. ZFS users should generally be fine. 20160523: The bitstring(3) API has been updated with new functionality and improved performance. But it is binary-incompatible with the old API. Objects built with the new headers may not be linked against objects built with the old headers. 20160520: The brk and sbrk functions have been removed from libc on arm64. Binutils from ports has been updated to not link to these functions and should be updated to the latest version before installing a new libc. 20160517: The armv6 port now defaults to hard float ABI. Limited support for running both hardfloat and soft float on the same system is available using the libraries installed with -DWITH_LIBSOFT. This has only been tested as an upgrade path for installworld and packages may fail or need manual intervention to run. New packages will be needed. To update an existing self-hosted armv6hf system, you must add TARGET_ARCH=armv6 on the make command line for both the build and the install steps. 20160510: Kernel modules compiled outside of a kernel build now default to installing to /boot/modules instead of /boot/kernel. Many kernel modules built this way (such as those in ports) already overrode KMODDIR explicitly to install into /boot/modules. However, manually building and installing a module from /sys/modules will now install to /boot/modules instead of /boot/kernel. 20160414: The CAM I/O scheduler has been committed to the kernel. There should be no user visible impact. This does enable NCQ Trim on ada SSDs. While the list of known rogues that claim support for this but actually corrupt data is believed to be complete, be on the lookout for data corruption. The known rogue list is believed to be complete: o Crucial MX100, M550 drives with MU01 firmware. o Micron M510 and M550 drives with MU01 firmware. o Micron M500 prior to MU07 firmware o Samsung 830, 840, and 850 all firmwares o FCCT M500 all firmwares Crucial has firmware http://www.crucial.com/usa/en/support-ssd-firmware with working NCQ TRIM. For Micron branded drives, see your sales rep for updated firmware. Black listed drives will work correctly because these drives work correctly so long as no NCQ TRIMs are sent to them. Given this list is the same as found in Linux, it's believed there are no other rogues in the market place. All other models from the above vendors work. To be safe, if you are at all concerned, you can quirk each of your drives to prevent NCQ from being sent by setting: kern.cam.ada.X.quirks="0x2" in loader.conf. If the drive requires the 4k sector quirk, set the quirks entry to 0x3. 20160330: The FAST_DEPEND build option has been removed and its functionality is now the one true way. The old mkdep(1) style of 'make depend' has been removed. See 20160311 for further details. 20160317: Resource range types have grown from unsigned long to uintmax_t. All drivers, and anything using libdevinfo, need to be recompiled. 20160311: WITH_FAST_DEPEND is now enabled by default for in-tree and out-of-tree builds. It no longer runs mkdep(1) during 'make depend', and the 'make depend' stage can safely be skipped now as it is auto ran when building 'make all' and will generate all SRCS and DPSRCS before building anything else. Dependencies are gathered at compile time with -MF flags kept in separate .depend files per object file. Users should run 'make cleandepend' once if using -DNO_CLEAN to clean out older stale .depend files. 20160306: On amd64, clang 3.8.0 can now insert sections of type AMD64_UNWIND into kernel modules. Therefore, if you load any kernel modules at boot time, please install the boot loaders after you install the kernel, but before rebooting, e.g.: make buildworld make kernel KERNCONF=YOUR_KERNEL_HERE make -C sys/boot install Then follow the usual steps, described in the General Notes section, below. 20160305: Clang, llvm, lldb and compiler-rt have been upgraded to 3.8.0. Please see the 20141231 entry below for information about prerequisites and upgrading, if you are not already using clang 3.5.0 or higher. 20160301: The AIO subsystem is now a standard part of the kernel. The VFS_AIO kernel option and aio.ko kernel module have been removed. Due to stability concerns, asynchronous I/O requests are only permitted on sockets and raw disks by default. To enable asynchronous I/O requests on all file types, set the vfs.aio.enable_unsafe sysctl to a non-zero value. 20160226: The ELF object manipulation tool objcopy is now provided by the ELF Tool Chain project rather than by GNU binutils. It should be a drop-in replacement, with the addition of arm64 support. The (temporary) src.conf knob WITHOUT_ELFCOPY_AS_OBJCOPY knob may be set to obtain the GNU version if necessary. 20160129: Building ZFS pools on top of zvols is prohibited by default. That feature has never worked safely; it's always been prone to deadlocks. Using a zvol as the backing store for a VM guest's virtual disk will still work, even if the guest is using ZFS. Legacy behavior can be restored by setting vfs.zfs.vol.recursive=1. 20160119: The NONE and HPN patches has been removed from OpenSSH. They are still available in the security/openssh-portable port. 20160113: With the addition of ypldap(8), a new _ypldap user is now required during installworld. "mergemaster -p" can be used to add the user prior to installworld, as documented in the handbook. 20151216: The tftp loader (pxeboot) now uses the option root-path directive. As a consequence it no longer looks for a pxeboot.4th file on the tftp server. Instead it uses the regular /boot infrastructure as with the other loaders. 20151211: The code to start recording plug and play data into the modules has been committed. While the old tools will properly build a new kernel, a number of warnings about "unknown metadata record 4" will be produced for an older kldxref. To avoid such warnings, make sure to rebuild the kernel toolchain (or world). Make sure that you have r292078 or later when trying to build 292077 or later before rebuilding. 20151207: Debug data files are now built by default with 'make buildworld' and installed with 'make installworld'. This facilitates debugging but requires more disk space both during the build and for the installed world. Debug files may be disabled by setting WITHOUT_DEBUG_FILES=yes in src.conf(5). 20151130: r291527 changed the internal interface between the nfsd.ko and nfscommon.ko modules. As such, they must both be upgraded to-gether. __FreeBSD_version has been bumped because of this. 20151108: Add support for unicode collation strings leads to a change of order of files listed by ls(1) for example. To get back to the old behaviour, set LC_COLLATE environment variable to "C". Databases administrators will need to reindex their databases given collation results will be different. Due to a bug in install(1) it is recommended to remove the ancient locales before running make installworld. rm -rf /usr/share/locale/* 20151030: The OpenSSL has been upgraded to 1.0.2d. Any binaries requiring libcrypto.so.7 or libssl.so.7 must be recompiled. 20151020: Qlogic 24xx/25xx firmware images were updated from 5.5.0 to 7.3.0. Kernel modules isp_2400_multi and isp_2500_multi were removed and should be replaced with isp_2400 and isp_2500 modules respectively. 20151017: The build previously allowed using 'make -n' to not recurse into sub-directories while showing what commands would be executed, and 'make -n -n' to recursively show commands. Now 'make -n' will recurse and 'make -N' will not. 20151012: If you specify SENDMAIL_MC or SENDMAIL_CF in make.conf, mergemaster and etcupdate will now use this file. A custom sendmail.cf is now updated via this mechanism rather than via installworld. If you had excluded sendmail.cf in mergemaster.rc or etcupdate.conf, you may want to remove the exclusion or change it to "always install". /etc/mail/sendmail.cf is now managed the same way regardless of whether SENDMAIL_MC/SENDMAIL_CF is used. If you are not using SENDMAIL_MC/SENDMAIL_CF there should be no change in behavior. 20151011: Compatibility shims for legacy ATA device names have been removed. It includes ATA_STATIC_ID kernel option, kern.cam.ada.legacy_aliases and kern.geom.raid.legacy_aliases loader tunables, kern.devalias.* environment variables, /dev/ad* and /dev/ar* symbolic links. 20151006: Clang, llvm, lldb, compiler-rt and libc++ have been upgraded to 3.7.0. Please see the 20141231 entry below for information about prerequisites and upgrading, if you are not already using clang 3.5.0 or higher. 20150924: Kernel debug files have been moved to /usr/lib/debug/boot/kernel/, and renamed from .symbols to .debug. This reduces the size requirements on the boot partition or file system and provides consistency with userland debug files. When using the supported kernel installation method the /usr/lib/debug/boot/kernel directory will be renamed (to kernel.old) as is done with /boot/kernel. Developers wishing to maintain the historical behavior of installing debug files in /boot/kernel/ can set KERN_DEBUGDIR="" in src.conf(5). 20150827: The wireless drivers had undergone changes that remove the 'parent interface' from the ifconfig -l output. The rc.d network scripts used to check presence of a parent interface in the list, so old scripts would fail to start wireless networking. Thus, etcupdate(3) or mergemaster(8) run is required after kernel update, to update your rc.d scripts in /etc. 20150827: pf no longer supports 'scrub fragment crop' or 'scrub fragment drop-ovl' These configurations are now automatically interpreted as 'scrub fragment reassemble'. 20150817: Kernel-loadable modules for the random(4) device are back. To use them, the kernel must have device random options RANDOM_LOADABLE kldload(8) can then be used to load random_fortuna.ko or random_yarrow.ko. Please note that due to the indirect function calls that the loadable modules need to provide, the build-in variants will be slightly more efficient. The random(4) kernel option RANDOM_DUMMY has been retired due to unpopularity. It was not all that useful anyway. 20150813: The WITHOUT_ELFTOOLCHAIN_TOOLS src.conf(5) knob has been retired. Control over building the ELF Tool Chain tools is now provided by the WITHOUT_TOOLCHAIN knob. 20150810: The polarity of Pulse Per Second (PPS) capture events with the uart(4) driver has been corrected. Prior to this change the PPS "assert" event corresponded to the trailing edge of a positive PPS pulse and the "clear" event was the leading edge of the next pulse. As the width of a PPS pulse in a typical GPS receiver is on the order of 1 millisecond, most users will not notice any significant difference with this change. Anyone who has compensated for the historical polarity reversal by configuring a negative offset equal to the pulse width will need to remove that workaround. 20150809: The default group assigned to /dev/dri entries has been changed from 'wheel' to 'video' with the id of '44'. If you want to have access to the dri devices please add yourself to the video group with: # pw groupmod video -m $USER 20150806: The menu.rc and loader.rc files will now be replaced during upgrades. Please migrate local changes to menu.rc.local and loader.rc.local instead. 20150805: GNU Binutils versions of addr2line, c++filt, nm, readelf, size, strings and strip have been removed. The src.conf(5) knob WITHOUT_ELFTOOLCHAIN_TOOLS no longer provides the binutils tools. 20150728: As ZFS requires more kernel stack pages than is the default on some architectures e.g. i386, it now warns if KSTACK_PAGES is less than ZFS_MIN_KSTACK_PAGES (which is 4 at the time of writing). Please consider using 'options KSTACK_PAGES=X' where X is greater than or equal to ZFS_MIN_KSTACK_PAGES i.e. 4 in such configurations. 20150706: sendmail has been updated to 8.15.2. Starting with FreeBSD 11.0 and sendmail 8.15, sendmail uses uncompressed IPv6 addresses by default, i.e., they will not contain "::". For example, instead of ::1, it will be 0:0:0:0:0:0:0:1. This permits a zero subnet to have a more specific match, such as different map entries for IPv6:0:0 vs IPv6:0. This change requires that configuration data (including maps, files, classes, custom ruleset, etc.) must use the same format, so make certain such configuration data is upgrading. As a very simple check search for patterns like 'IPv6:[0-9a-fA-F:]*::' and 'IPv6::'. To return to the old behavior, set the m4 option confUSE_COMPRESSED_IPV6_ADDRESSES or the cf option UseCompressedIPv6Addresses. 20150630: The default kernel entropy-processing algorithm is now Fortuna, replacing Yarrow. Assuming you have 'device random' in your kernel config file, the configurations allow a kernel option to override this default. You may choose *ONE* of: options RANDOM_YARROW # Legacy /dev/random algorithm. options RANDOM_DUMMY # Blocking-only driver. If you have neither, you get Fortuna. For most people, read no further, Fortuna will give a /dev/random that works like it always used to, and the difference will be irrelevant. If you remove 'device random', you get *NO* kernel-processed entropy at all. This may be acceptable to folks building embedded systems, but has complications. Carry on reading, and it is assumed you know what you need. *PLEASE* read random(4) and random(9) if you are in the habit of tweaking kernel configs, and/or if you are a member of the embedded community, wanting specific and not-usual behaviour from your security subsystems. NOTE!! If you use RANDOM_DUMMY and/or have no 'device random', you will NOT have a functioning /dev/random, and many cryptographic features will not work, including SSH. You may also find strange behaviour from the random(3) set of library functions, in particular sranddev(3), srandomdev(3) and arc4random(3). The reason for this is that the KERN_ARND sysctl only returns entropy if it thinks it has some to share, and with RANDOM_DUMMY or no 'device random' this will never happen. 20150623: An additional fix for the issue described in the 20150614 sendmail - entry below has been been committed in revision 284717. + entry below has been committed in revision 284717. 20150616: FreeBSD's old make (fmake) has been removed from the system. It is available as the devel/fmake port or via pkg install fmake. 20150615: The fix for the issue described in the 20150614 sendmail entry - below has been been committed in revision 284436. The work + below has been committed in revision 284436. The work around described in that entry is no longer needed unless the default setting is overridden by a confDH_PARAMETERS configuration setting of '5' or pointing to a 512 bit DH parameter file. 20150614: ALLOW_DEPRECATED_ATF_TOOLS/ATFFILE support has been removed from atf.test.mk (included from bsd.test.mk). Please upgrade devel/atf and devel/kyua to version 0.20+ and adjust any calling code to work with Kyuafile and kyua. 20150614: The import of openssl to address the FreeBSD-SA-15:10.openssl security advisory includes a change which rejects handshakes with DH parameters below 768 bits. sendmail releases prior to 8.15.2 (not yet released), defaulted to a 512 bit DH parameter setting for client connections. To work around this interoperability, sendmail can be configured to use a 2048 bit DH parameter by: 1. Edit /etc/mail/`hostname`.mc 2. If a setting for confDH_PARAMETERS does not exist or exists and is set to a string beginning with '5', replace it with '2'. 3. If a setting for confDH_PARAMETERS exists and is set to a file path, create a new file with: openssl dhparam -out /path/to/file 2048 4. Rebuild the .cf file: cd /etc/mail/; make; make install 5. Restart sendmail: cd /etc/mail/; make restart A sendmail patch is coming, at which time this file will be updated. 20150604: Generation of legacy formatted entries have been disabled by default in pwd_mkdb(8), as all base system consumers of the legacy formatted entries were converted to use the new format by default when the new, machine independent format have been added and supported since FreeBSD 5.x. Please see the pwd_mkdb(8) manual page for further details. 20150525: Clang and llvm have been upgraded to 3.6.1 release. Please see the 20141231 entry below for information about prerequisites and upgrading, if you are not already using 3.5.0 or higher. 20150521: TI platform code switched to using vendor DTS files and this update may break existing systems running on Beaglebone, Beaglebone Black, and Pandaboard: - dtb files should be regenerated/reinstalled. Filenames are the same but content is different now - GPIO addressing was changed, now each GPIO bank (32 pins per bank) has its own /dev/gpiocX device, e.g. pin 121 on /dev/gpioc0 in old addressing scheme is now pin 25 on /dev/gpioc3. - Pandaboard: /etc/ttys should be updated, serial console device is now /dev/ttyu2, not /dev/ttyu0 20150501: soelim(1) from gnu/usr.bin/groff has been replaced by usr.bin/soelim. If you need the GNU extension from groff soelim(1), install groff from package: pkg install groff, or via ports: textproc/groff. 20150423: chmod, chflags, chown and chgrp now affect symlinks in -R mode as defined in symlink(7); previously symlinks were silently ignored. 20150415: The const qualifier has been removed from iconv(3) to comply with POSIX. The ports tree is aware of this from r384038 onwards. 20150416: Libraries specified by LIBADD in Makefiles must have a corresponding DPADD_ variable to ensure correct dependencies. This is now enforced in src.libnames.mk. 20150324: From legacy ata(4) driver was removed support for SATA controllers supported by more functional drivers ahci(4), siis(4) and mvs(4). Kernel modules ataahci and ataadaptec were removed completely, replaced by ahci and mvs modules respectively. 20150315: Clang, llvm and lldb have been upgraded to 3.6.0 release. Please see the 20141231 entry below for information about prerequisites and upgrading, if you are not already using 3.5.0 or higher. 20150307: The 32-bit PowerPC kernel has been changed to a position-independent executable. This can only be booted with a version of loader(8) newer than January 31, 2015, so make sure to update both world and kernel before rebooting. 20150217: If you are running a -CURRENT kernel since r273872 (Oct 30th, 2014), but before r278950, the RNG was not seeded properly. Immediately upgrade the kernel to r278950 or later and regenerate any keys (e.g. ssh keys or openssl keys) that were generated w/ a kernel from that range. This does not affect programs that directly used /dev/random or /dev/urandom. All userland uses of arc4random(3) are affected. 20150210: The autofs(4) ABI was changed in order to restore binary compatibility with 10.1-RELEASE. The automountd(8) daemon needs to be rebuilt to work with the new kernel. 20150131: The powerpc64 kernel has been changed to a position-independent executable. This can only be booted with a new version of loader(8), so make sure to update both world and kernel before rebooting. 20150118: Clang and llvm have been upgraded to 3.5.1 release. This is a bugfix only release, no new features have been added. Please see the 20141231 entry below for information about prerequisites and upgrading, if you are not already using 3.5.0. 20150107: ELF tools addr2line, elfcopy (strip), nm, size, and strings are now taken from the ELF Tool Chain project rather than GNU binutils. They should be drop-in replacements, with the addition of arm64 support. The WITHOUT_ELFTOOLCHAIN_TOOLS= knob may be used to obtain the binutils tools, if necessary. See 20150805 for updated information. 20150105: The default Unbound configuration now enables remote control using a local socket. Users who have already enabled the local_unbound service should regenerate their configuration by running "service local_unbound setup" as root. 20150102: The GNU texinfo and GNU info pages have been removed. To be able to view GNU info pages please install texinfo from ports. 20141231: Clang, llvm and lldb have been upgraded to 3.5.0 release. As of this release, a prerequisite for building clang, llvm and lldb is a C++11 capable compiler and C++11 standard library. This means that to be able to successfully build the cross-tools stage of buildworld, with clang as the bootstrap compiler, your system compiler or cross compiler should either be clang 3.3 or later, or gcc 4.8 or later, and your system C++ library should be libc++, or libdstdc++ from gcc 4.8 or later. On any standard FreeBSD 10.x or 11.x installation, where clang and libc++ are on by default (that is, on x86 or arm), this should work out of the box. On 9.x installations where clang is enabled by default, e.g. on x86 and powerpc, libc++ will not be enabled by default, so libc++ should be built (with clang) and installed first. If both clang and libc++ are missing, build clang first, then use it to build libc++. On 8.x and earlier installations, upgrade to 9.x first, and then follow the instructions for 9.x above. Sparc64 and mips users are unaffected, as they still use gcc 4.2.1 by default, and do not build clang. Many embedded systems are resource constrained, and will not be able to build clang in a reasonable time, or in some cases at all. In those cases, cross building bootable systems on amd64 is a workaround. This new version of clang introduces a number of new warnings, of which the following are most likely to appear: -Wabsolute-value This warns in two cases, for both C and C++: * When the code is trying to take the absolute value of an unsigned quantity, which is effectively a no-op, and almost never what was intended. The code should be fixed, if at all possible. If you are sure that the unsigned quantity can be safely cast to signed, without loss of information or undefined behavior, you can add an explicit cast, or disable the warning. * When the code is trying to take an absolute value, but the called abs() variant is for the wrong type, which can lead to truncation. If you want to disable the warning instead of fixing the code, please make sure that truncation will not occur, or it might lead to unwanted side-effects. -Wtautological-undefined-compare and -Wundefined-bool-conversion These warn when C++ code is trying to compare 'this' against NULL, while 'this' should never be NULL in well-defined C++ code. However, there is some legacy (pre C++11) code out there, which actively abuses this feature, which was less strictly defined in previous C++ versions. Squid and openjdk do this, for example. The warning can be turned off for C++98 and earlier, but compiling the code in C++11 mode might result in unexpected behavior; for example, the parts of the program that are unreachable could be optimized away. 20141222: The old NFS client and server (kernel options NFSCLIENT, NFSSERVER) kernel sources have been removed. The .h files remain, since some utilities include them. This will need to be fixed later. If "mount -t oldnfs ..." is attempted, it will fail. If the "-o" option on mountd(8), nfsd(8) or nfsstat(1) is used, the utilities will report errors. 20141121: The handling of LOCAL_LIB_DIRS has been altered to skip addition of directories to top level SUBDIR variable when their parent directory is included in LOCAL_DIRS. Users with build systems with such hierarchies and without SUBDIR entries in the parent directory Makefiles should add them or add the directories to LOCAL_DIRS. 20141109: faith(4) and faithd(8) have been removed from the base system. Faith has been obsolete for a very long time. 20141104: vt(4), the new console driver, is enabled by default. It brings support for Unicode and double-width characters, as well as support for UEFI and integration with the KMS kernel video drivers. You may need to update your console settings in /etc/rc.conf, most probably the keymap. During boot, /etc/rc.d/syscons will indicate what you need to do. vt(4) still has issues and lacks some features compared to syscons(4). See the wiki for up-to-date information: https://wiki.freebsd.org/Newcons If you want to keep using syscons(4), you can do so by adding the following line to /boot/loader.conf: kern.vty=sc 20141102: pjdfstest has been integrated into kyua as an opt-in test suite. Please see share/doc/pjdfstest/README for more details on how to execute it. 20141009: gperf has been removed from the base system for architectures that use clang. Ports that require gperf will obtain it from the devel/gperf port. 20140923: pjdfstest has been moved from tools/regression/pjdfstest to contrib/pjdfstest . 20140922: At svn r271982, The default linux compat kernel ABI has been adjusted to 2.6.18 in support of the linux-c6 compat ports infrastructure update. If you wish to continue using the linux-f10 compat ports, add compat.linux.osrelease=2.6.16 to your local sysctl.conf. Users are encouraged to update their linux-compat packages to linux-c6 during their next update cycle. 20140729: The ofwfb driver, used to provide a graphics console on PowerPC when using vt(4), no longer allows mmap() of all physical memory. This will prevent Xorg on PowerPC with some ATI graphics cards from initializing properly unless x11-servers/xorg-server is updated to 1.12.4_8 or newer. 20140723: The xdev targets have been converted to using TARGET and TARGET_ARCH instead of XDEV and XDEV_ARCH. 20140719: The default unbound configuration has been modified to address issues with reverse lookups on networks that use private address ranges. If you use the local_unbound service, run "service local_unbound setup" as root to regenerate your configuration, then "service local_unbound reload" to load the new configuration. 20140709: The GNU texinfo and GNU info pages are not built and installed anymore, WITH_INFO knob has been added to allow to built and install them again. UPDATE: see 20150102 entry on texinfo's removal 20140708: The GNU readline library is now an INTERNALLIB - that is, it is statically linked into consumers (GDB and variants) in the base system, and the shared library is no longer installed. The devel/readline port is available for third party software that requires readline. 20140702: The Itanium architecture (ia64) has been removed from the list of known architectures. This is the first step in the removal of the architecture. 20140701: Commit r268115 has added NFSv4.1 server support, merged from projects/nfsv4.1-server. Since this includes changes to the internal interfaces between the NFS related modules, a full build of the kernel and modules will be necessary. __FreeBSD_version has been bumped. 20140629: The WITHOUT_VT_SUPPORT kernel config knob has been renamed WITHOUT_VT. (The other _SUPPORT knobs have a consistent meaning which differs from the behaviour controlled by this knob.) 20140619: Maximal length of the serial number in CTL was increased from 16 to 64 chars, that breaks ABI. All CTL-related tools, such as ctladm and ctld, need to be rebuilt to work with a new kernel. 20140606: The libatf-c and libatf-c++ major versions were downgraded to 0 and 1 respectively to match the upstream numbers. They were out of sync because, when they were originally added to FreeBSD, the upstream versions were not respected. These libraries are private and not yet built by default, so renumbering them should be a non-issue. However, unclean source trees will yield broken test programs once the operator executes "make delete-old-libs" after a "make installworld". Additionally, the atf-sh binary was made private by moving it into /usr/libexec/. Already-built shell test programs will keep the path to the old binary so they will break after "make delete-old" is run. If you are using WITH_TESTS=yes (not the default), wipe the object tree and rebuild from scratch to prevent spurious test failures. This is only needed once: the misnumbered libraries and misplaced binaries have been added to OptionalObsoleteFiles.inc so they will be removed during a clean upgrade. 20140512: Clang and llvm have been upgraded to 3.4.1 release. 20140508: We bogusly installed src.opts.mk in /usr/share/mk. This file should be removed to avoid issues in the future (and has been added to ObsoleteFiles.inc). 20140505: /etc/src.conf now affects only builds of the FreeBSD src tree. In the past, it affected all builds that used the bsd.*.mk files. The old behavior was a bug, but people may have relied upon it. To get this behavior back, you can .include /etc/src.conf from /etc/make.conf (which is still global and isn't changed). This also changes the behavior of incremental builds inside the tree of individual directories. Set MAKESYSPATH to ".../share/mk" to do that. Although this has survived make universe and some upgrade scenarios, other upgrade scenarios may have broken. At least one form of temporary breakage was fixed with MAKESYSPATH settings for buildworld as well... In cases where MAKESYSPATH isn't working with this setting, you'll need to set it to the full path to your tree. One side effect of all this cleaning up is that bsd.compiler.mk is no longer implicitly included by bsd.own.mk. If you wish to use COMPILER_TYPE, you must now explicitly include bsd.compiler.mk as well. 20140430: The lindev device has been removed since /dev/full has been made a standard device. __FreeBSD_version has been bumped. 20140424: The knob WITHOUT_VI was added to the base system, which controls building ex(1), vi(1), etc. Older releases of FreeBSD required ex(1) in order to reorder files share/termcap and didn't build ex(1) as a build tool, so building/installing with WITH_VI is highly advised for build hosts for older releases. This issue has been fixed in stable/9 and stable/10 in r277022 and r276991, respectively. 20140418: The YES_HESIOD knob has been removed. It has been obsolete for a decade. Please move to using WITH_HESIOD instead or your builds will silently lack HESIOD. 20140405: The uart(4) driver has been changed with respect to its handling of the low-level console. Previously the uart(4) driver prevented any process from changing the baudrate or the CLOCAL and HUPCL control flags. By removing the restrictions, operators can make changes to the serial console port without having to reboot. However, when getty(8) is started on the serial device that is associated with the low-level console, a misconfigured terminal line in /etc/ttys will now have a real impact. Before upgrading the kernel, make sure that /etc/ttys has the serial console device configured as 3wire without baudrate to preserve the previous behaviour. E.g: ttyu0 "/usr/libexec/getty 3wire" vt100 on secure 20140306: Support for libwrap (TCP wrappers) in rpcbind was disabled by default to improve performance. To re-enable it, if needed, run rpcbind with command line option -W. 20140226: Switched back to the GPL dtc compiler due to updates in the upstream dts files not being supported by the BSDL dtc compiler. You will need to rebuild your kernel toolchain to pick up the new compiler. Core dumps may result while building dtb files during a kernel build if you fail to do so. Set WITHOUT_GPL_DTC if you require the BSDL compiler. 20140216: Clang and llvm have been upgraded to 3.4 release. 20140216: The nve(4) driver has been removed. Please use the nfe(4) driver for NVIDIA nForce MCP Ethernet adapters instead. 20140212: An ABI incompatibility crept into the libc++ 3.4 import in r261283. This could cause certain C++ applications using shared libraries built against the previous version of libc++ to crash. The incompatibility has now been fixed, but any C++ applications or shared libraries built between r261283 and r261801 should be recompiled. 20140204: OpenSSH will now ignore errors caused by kernel lacking of Capsicum capability mode support. Please note that enabling the feature in kernel is still highly recommended. 20140131: OpenSSH is now built with sandbox support, and will use sandbox as the default privilege separation method. This requires Capsicum capability mode support in kernel. 20140128: The libelf and libdwarf libraries have been updated to newer versions from upstream. Shared library version numbers for these two libraries were bumped. Any ports or binaries requiring these two libraries should be recompiled. __FreeBSD_version is bumped to 1100006. 20140110: If a Makefile in a tests/ directory was auto-generating a Kyuafile instead of providing an explicit one, this would prevent such Makefile from providing its own Kyuafile in the future during NO_CLEAN builds. This has been fixed in the Makefiles but manual intervention is needed to clean an objdir if you use NO_CLEAN: # find /usr/obj -name Kyuafile | xargs rm -f 20131213: The behavior of gss_pseudo_random() for the krb5 mechanism has changed, for applications requesting a longer random string than produced by the underlying enctype's pseudo-random() function. In particular, the random string produced from a session key of enctype aes256-cts-hmac-sha1-96 or aes256-cts-hmac-sha1-96 will be different at the 17th octet and later, after this change. The counter used in the PRF+ construction is now encoded as a big-endian integer in accordance with RFC 4402. __FreeBSD_version is bumped to 1100004. 20131108: The WITHOUT_ATF build knob has been removed and its functionality has been subsumed into the more generic WITHOUT_TESTS. If you were using the former to disable the build of the ATF libraries, you should change your settings to use the latter. 20131025: The default version of mtree is nmtree which is obtained from NetBSD. The output is generally the same, but may vary slightly. If you found you need identical output adding "-F freebsd9" to the command line should do the trick. For the time being, the old mtree is available as fmtree. 20131014: libbsdyml has been renamed to libyaml and moved to /usr/lib/private. This will break ports-mgmt/pkg. Rebuild the port, or upgrade to pkg 1.1.4_8 and verify bsdyml not linked in, before running "make delete-old-libs": # make -C /usr/ports/ports-mgmt/pkg build deinstall install clean or # pkg install pkg; ldd /usr/local/sbin/pkg | grep bsdyml 20131010: The stable/10 branch has been created in subversion from head revision r256279. 20131010: The rc.d/jail script has been updated to support jail(8) configuration file. The "jail__*" rc.conf(5) variables for per-jail configuration are automatically converted to /var/run/jail..conf before the jail(8) utility is invoked. This is transparently backward compatible. See below about some incompatibilities and rc.conf(5) manual page for more details. These variables are now deprecated in favor of jail(8) configuration file. One can use "rc.d/jail config " command to generate a jail(8) configuration file in /var/run/jail..conf without running the jail(8) utility. The default pathname of the configuration file is /etc/jail.conf and can be specified by using $jail_conf or $jail__conf variables. Please note that jail_devfs_ruleset accepts an integer at this moment. Please consider to rewrite the ruleset name with an integer. 20130930: BIND has been removed from the base system. If all you need is a local resolver, simply enable and start the local_unbound service instead. Otherwise, several versions of BIND are available in the ports tree. The dns/bind99 port is one example. With this change, nslookup(1) and dig(1) are no longer in the base system. Users should instead use host(1) and drill(1) which are in the base system. Alternatively, nslookup and dig can be obtained by installing the dns/bind-tools port. 20130916: With the addition of unbound(8), a new unbound user is now required during installworld. "mergemaster -p" can be used to add the user prior to installworld, as documented in the handbook. 20130911: OpenSSH is now built with DNSSEC support, and will by default silently trust signed SSHFP records. This can be controlled with the VerifyHostKeyDNS client configuration setting. DNSSEC support can be disabled entirely with the WITHOUT_LDNS option in src.conf. 20130906: The GNU Compiler Collection and C++ standard library (libstdc++) are no longer built by default on platforms where clang is the system compiler. You can enable them with the WITH_GCC and WITH_GNUCXX options in src.conf. 20130905: The PROCDESC kernel option is now part of the GENERIC kernel configuration and is required for the rwhod(8) to work. If you are using custom kernel configuration, you should include 'options PROCDESC'. 20130905: The API and ABI related to the Capsicum framework was modified in backward incompatible way. The userland libraries and programs have to be recompiled to work with the new kernel. This includes the following libraries and programs, but the whole buildworld is advised: libc, libprocstat, dhclient, tcpdump, hastd, hastctl, kdump, procstat, rwho, rwhod, uniq. 20130903: AES-NI intrinsic support has been added to gcc. The AES-NI module has been updated to use this support. A new gcc is required to build the aesni module on both i386 and amd64. 20130821: The PADLOCK_RNG and RDRAND_RNG kernel options are now devices. Thus "device padlock_rng" and "device rdrand_rng" should be used instead of "options PADLOCK_RNG" & "options RDRAND_RNG". 20130813: WITH_ICONV has been split into two feature sets. WITH_ICONV now enables just the iconv* functionality and is now on by default. WITH_LIBICONV_COMPAT enables the libiconv api and link time compatibility. Set WITHOUT_ICONV to build the old way. If you have been using WITH_ICONV before, you will very likely need to turn on WITH_LIBICONV_COMPAT. 20130806: INVARIANTS option now enables DEBUG for code with OpenSolaris and Illumos origin, including ZFS. If you have INVARIANTS in your kernel configuration, then there is no need to set DEBUG or ZFS_DEBUG explicitly. DEBUG used to enable witness(9) tracking of OpenSolaris (mostly ZFS) locks if WITNESS option was set. Because that generated a lot of witness(9) reports and all of them were believed to be false positives, this is no longer done. New option OPENSOLARIS_WITNESS can be used to achieve the previous behavior. 20130806: Timer values in IPv6 data structures now use time_uptime instead of time_second. Although this is not a user-visible functional change, userland utilities which directly use them---ndp(8), rtadvd(8), and rtsold(8) in the base system---need to be updated to r253970 or later. 20130802: find -delete can now delete the pathnames given as arguments, instead of only files found below them or if the pathname did not contain any slashes. Formerly, the following error message would result: find: -delete: : relative path potentially not safe Deleting the pathnames given as arguments can be prevented without error messages using -mindepth 1 or by changing directory and passing "." as argument to find. This works in the old as well as the new version of find. 20130726: Behavior of devfs rules path matching has been changed. Pattern is now always matched against fully qualified devfs path and slash characters must be explicitly matched by slashes in pattern (FNM_PATHNAME). Rulesets involving devfs subdirectories must be reviewed. 20130716: The default ARM ABI has changed to the ARM EABI. The old ABI is incompatible with the ARM EABI and all programs and modules will need to be rebuilt to work with a new kernel. To keep using the old ABI ensure the WITHOUT_ARM_EABI knob is set. NOTE: Support for the old ABI will be removed in the future and users are advised to upgrade. 20130709: pkg_install has been disconnected from the build if you really need it you should add WITH_PKGTOOLS in your src.conf(5). 20130709: Most of network statistics structures were changed to be able keep 64-bits counters. Thus all tools, that work with networking statistics, must be rebuilt (netstat(1), bsnmpd(1), etc.) 20130618: Fix a bug that allowed a tracing process (e.g. gdb) to write to a memory-mapped file in the traced process's address space even if neither the traced process nor the tracing process had write access to that file. 20130615: CVS has been removed from the base system. An exact copy of the code is available from the devel/cvs port. 20130613: Some people report the following error after the switch to bmake: make: illegal option -- J usage: make [-BPSXeiknpqrstv] [-C directory] [-D variable] ... *** [buildworld] Error code 2 this likely due to an old instance of make in ${MAKEPATH} (${MAKEOBJDIRPREFIX}${.CURDIR}/make.${MACHINE}) which src/Makefile will use that blindly, if it exists, so if you see the above error: rm -rf `make -V MAKEPATH` should resolve it. 20130516: Use bmake by default. Whereas before one could choose to build with bmake via -DWITH_BMAKE one must now use -DWITHOUT_BMAKE to use the old make. The goal is to remove these knobs for 10-RELEASE. It is worth noting that bmake (like gmake) treats the command line as the unit of failure, rather than statements within the command line. Thus '(cd some/where && dosomething)' is safer than 'cd some/where; dosomething'. The '()' allows consistent behavior in parallel build. 20130429: Fix a bug that allows NFS clients to issue READDIR on files. 20130426: The WITHOUT_IDEA option has been removed because the IDEA patent expired. 20130426: The sysctl which controls TRIM support under ZFS has been renamed from vfs.zfs.trim_disable -> vfs.zfs.trim.enabled and has been enabled by default. 20130425: The mergemaster command now uses the default MAKEOBJDIRPREFIX rather than creating it's own in the temporary directory in order allow access to bootstrapped versions of tools such as install and mtree. When upgrading from version of FreeBSD where the install command does not support -l, you will need to install a new mergemaster command if mergemaster -p is required. This can be accomplished with the command (cd src/usr.sbin/mergemaster && make install). 20130404: Legacy ATA stack, disabled and replaced by new CAM-based one since FreeBSD 9.0, completely removed from the sources. Kernel modules atadisk and atapi*, user-level tools atacontrol and burncd are removed. Kernel option `options ATA_CAM` is now permanently enabled and removed. 20130319: SOCK_CLOEXEC and SOCK_NONBLOCK flags have been added to socket(2) and socketpair(2). Software, in particular Kerberos, may automatically detect and use these during building. The resulting binaries will not work on older kernels. 20130308: CTL_DISABLE has also been added to the sparc64 GENERIC (for further information, see the respective 20130304 entry). 20130304: Recent commits to callout(9) changed the size of struct callout, so the KBI is probably heavily disturbed. Also, some functions in callout(9)/sleep(9)/sleepqueue(9)/condvar(9) KPIs were replaced by macros. Every kernel module using it won't load, so rebuild is requested. The ctl device has been re-enabled in GENERIC for i386 and amd64, but does not initialize by default (because of the new CTL_DISABLE option) to save memory. To re-enable it, remove the CTL_DISABLE option from the kernel config file or set kern.cam.ctl.disable=0 in /boot/loader.conf. 20130301: The ctl device has been disabled in GENERIC for i386 and amd64. This was done due to the extra memory being allocated at system initialisation time by the ctl driver which was only used if a CAM target device was created. This makes a FreeBSD system unusable on 128MB or less of RAM. 20130208: A new compression method (lz4) has been merged to -HEAD. Please refer to zpool-features(7) for more information. Please refer to the "ZFS notes" section of this file for information on upgrading boot ZFS pools. 20130129: A BSD-licensed patch(1) variant has been added and is installed as bsdpatch, being the GNU version the default patch. To inverse the logic and use the BSD-licensed one as default, while having the GNU version installed as gnupatch, rebuild and install world with the WITH_BSD_PATCH knob set. 20130121: Due to the use of the new -l option to install(1) during build and install, you must take care not to directly set the INSTALL make variable in your /etc/make.conf, /etc/src.conf, or on the command line. If you wish to use the -C flag for all installs you may be able to add INSTALL+=-C to /etc/make.conf or /etc/src.conf. 20130118: The install(1) option -M has changed meaning and now takes an argument that is a file or path to append logs to. In the unlikely event that -M was the last option on the command line and the command line contained at least two files and a target directory the first file will have logs appended to it. The -M option served little practical purpose in the last decade so its use is expected to be extremely rare. 20121223: After switching to Clang as the default compiler some users of ZFS on i386 systems started to experience stack overflow kernel panics. Please consider using 'options KSTACK_PAGES=4' in such configurations. 20121222: GEOM_LABEL now mangles label names read from file system metadata. Mangling affect labels containing spaces, non-printable characters, '%' or '"'. Device names in /etc/fstab and other places may need to be updated. 20121217: By default, only the 10 most recent kernel dumps will be saved. To restore the previous behaviour (no limit on the number of kernel dumps stored in the dump directory) add the following line to /etc/rc.conf: savecore_flags="" 20121201: With the addition of auditdistd(8), a new auditdistd user is now required during installworld. "mergemaster -p" can be used to add the user prior to installworld, as documented in the handbook. 20121117: The sin6_scope_id member variable in struct sockaddr_in6 is now filled by the kernel before passing the structure to the userland via sysctl or routing socket. This means the KAME-specific embedded scope id in sin6_addr.s6_addr[2] is always cleared in userland application. This behavior can be controlled by net.inet6.ip6.deembed_scopeid. __FreeBSD_version is bumped to 1000025. 20121105: On i386 and amd64 systems WITH_CLANG_IS_CC is now the default. This means that the world and kernel will be compiled with clang and that clang will be installed as /usr/bin/cc, /usr/bin/c++, and /usr/bin/cpp. To disable this behavior and revert to building with gcc, compile with WITHOUT_CLANG_IS_CC. Really old versions of current may need to bootstrap WITHOUT_CLANG first if the clang build fails (its compatibility window doesn't extend to the 9 stable branch point). 20121102: The IPFIREWALL_FORWARD kernel option has been removed. Its functionality now turned on by default. 20121023: The ZERO_COPY_SOCKET kernel option has been removed and split into SOCKET_SEND_COW and SOCKET_RECV_PFLIP. NB: SOCKET_SEND_COW uses the VM page based copy-on-write mechanism which is not safe and may result in kernel crashes. NB: The SOCKET_RECV_PFLIP mechanism is useless as no current driver supports disposeable external page sized mbuf storage. Proper replacements for both zero-copy mechanisms are under consideration and will eventually lead to complete removal of the two kernel options. 20121023: The IPv4 network stack has been converted to network byte order. The following modules need to be recompiled together with kernel: carp(4), divert(4), gif(4), siftr(4), gre(4), pf(4), ipfw(4), ng_ipfw(4), stf(4). 20121022: Support for non-MPSAFE filesystems was removed from VFS. The VFS_VERSION was bumped, all filesystem modules shall be recompiled. 20121018: All the non-MPSAFE filesystems have been disconnected from the build. The full list includes: codafs, hpfs, ntfs, nwfs, portalfs, smbfs, xfs. 20121016: The interface cloning API and ABI has changed. The following modules need to be recompiled together with kernel: ipfw(4), pfsync(4), pflog(4), usb(4), wlan(4), stf(4), vlan(4), disc(4), edsc(4), if_bridge(4), gif(4), tap(4), faith(4), epair(4), enc(4), tun(4), if_lagg(4), gre(4). 20121015: The sdhci driver was split in two parts: sdhci (generic SD Host Controller logic) and sdhci_pci (actual hardware driver). No kernel config modifications are required, but if you load sdhc as a module you must switch to sdhci_pci instead. 20121014: Import the FUSE kernel and userland support into base system. 20121013: The GNU sort(1) program has been removed since the BSD-licensed sort(1) has been the default for quite some time and no serious problems have been reported. The corresponding WITH_GNU_SORT knob has also gone. 20121006: The pfil(9) API/ABI for AF_INET family has been changed. Packet filtering modules: pf(4), ipfw(4), ipfilter(4) need to be recompiled with new kernel. 20121001: The net80211(4) ABI has been changed to allow for improved driver PS-POLL and power-save support. All wireless drivers need to be recompiled to work with the new kernel. 20120913: The random(4) support for the VIA hardware random number generator (`PADLOCK') is no longer enabled unconditionally. Add the padlock_rng device in the custom kernel config if needed. The GENERIC kernels on i386 and amd64 do include the device, so the change only affects the custom kernel configurations. 20120908: The pf(4) packet filter ABI has been changed. pfctl(8) and snmp_pf module need to be recompiled to work with new kernel. 20120828: A new ZFS feature flag "com.delphix:empty_bpobj" has been merged to -HEAD. Pools that have empty_bpobj in active state can not be imported read-write with ZFS implementations that do not support this feature. For more information read the zpool-features(5) manual page. 20120727: The sparc64 ZFS loader has been changed to no longer try to auto- detect ZFS providers based on diskN aliases but now requires these to be explicitly listed in the OFW boot-device environment variable. 20120712: The OpenSSL has been upgraded to 1.0.1c. Any binaries requiring libcrypto.so.6 or libssl.so.6 must be recompiled. Also, there are configuration changes. Make sure to merge /etc/ssl/openssl.cnf. 20120712: The following sysctls and tunables have been renamed for consistency with other variables: kern.cam.da.da_send_ordered -> kern.cam.da.send_ordered kern.cam.ada.ada_send_ordered -> kern.cam.ada.send_ordered 20120628: The sort utility has been replaced with BSD sort. For now, GNU sort is also available as "gnusort" or the default can be set back to GNU sort by setting WITH_GNU_SORT. In this case, BSD sort will be installed as "bsdsort". 20120611: A new version of ZFS (pool version 5000) has been merged to -HEAD. Starting with this version the old system of ZFS pool versioning is superseded by "feature flags". This concept enables forward compatibility against certain future changes in functionality of ZFS pools. The first read-only compatible "feature flag" for ZFS pools is named "com.delphix:async_destroy". For more information read the new zpool-features(5) manual page. Please refer to the "ZFS notes" section of this file for information on upgrading boot ZFS pools. 20120417: The malloc(3) implementation embedded in libc now uses sources imported as contrib/jemalloc. The most disruptive API change is to /etc/malloc.conf. If your system has an old-style /etc/malloc.conf, delete it prior to installworld, and optionally re-create it using the new format after rebooting. See malloc.conf(5) for details (specifically the TUNING section and the "opt.*" entries in the MALLCTL NAMESPACE section). 20120328: Big-endian MIPS TARGET_ARCH values no longer end in "eb". mips64eb is now spelled mips64. mipsn32eb is now spelled mipsn32. mipseb is now spelled mips. This is to aid compatibility with third-party software that expects this naming scheme in uname(3). Little-endian settings are unchanged. If you are updating a big-endian mips64 machine from before this change, you may need to set MACHINE_ARCH=mips64 in your environment before the new build system will recognize your machine. 20120306: Disable by default the option VFS_ALLOW_NONMPSAFE for all supported platforms. 20120229: Now unix domain sockets behave "as expected" on nullfs(5). Previously nullfs(5) did not pass through all behaviours to the underlying layer, as a result if we bound to a socket on the lower layer we could connect only to the lower path; if we bound to the upper layer we could connect only to the upper path. The new behavior is one can connect to both the lower and the upper paths regardless what layer path one binds to. 20120211: The getifaddrs upgrade path broken with 20111215 has been restored. If you have upgraded in between 20111215 and 20120209 you need to recompile libc again with your kernel. You still need to recompile world to be able to configure CARP but this restriction already comes from 20111215. 20120114: The set_rcvar() function has been removed from /etc/rc.subr. All base and ports rc.d scripts have been updated, so if you have a port installed with a script in /usr/local/etc/rc.d you can either hand-edit the rcvar= line, or reinstall the port. An easy way to handle the mass-update of /etc/rc.d: rm /etc/rc.d/* && mergemaster -i 20120109: panic(9) now stops other CPUs in the SMP systems, disables interrupts on the current CPU and prevents other threads from running. This behavior can be reverted using the kern.stop_scheduler_on_panic tunable/sysctl. The new behavior can be incompatible with kern.sync_on_panic. 20111215: The carp(4) facility has been changed significantly. Configuration of the CARP protocol via ifconfig(8) has changed, as well as format of CARP events submitted to devd(8) has changed. See manual pages for more information. The arpbalance feature of carp(4) is currently not supported anymore. Size of struct in_aliasreq, struct in6_aliasreq has changed. User utilities using SIOCAIFADDR, SIOCAIFADDR_IN6, e.g. ifconfig(8), need to be recompiled. 20111122: The acpi_wmi(4) status device /dev/wmistat has been renamed to /dev/wmistat0. 20111108: The option VFS_ALLOW_NONMPSAFE option has been added in order to explicitely support non-MPSAFE filesystems. It is on by default for all supported platform at this present time. 20111101: The broken amd(4) driver has been replaced with esp(4) in the amd64, i386 and pc98 GENERIC kernel configuration files. 20110930: sysinstall has been removed 20110923: The stable/9 branch created in subversion. This corresponds to the RELENG_9 branch in CVS. COMMON ITEMS: General Notes ------------- Avoid using make -j when upgrading. While generally safe, there are sometimes problems using -j to upgrade. If your upgrade fails with -j, please try again without -j. From time to time in the past there have been problems using -j with buildworld and/or installworld. This is especially true when upgrading between "distant" versions (eg one that cross a major release boundary or several minor releases, or when several months have passed on the -current branch). Sometimes, obscure build problems are the result of environment poisoning. This can happen because the make utility reads its environment when searching for values for global variables. To run your build attempts in an "environmental clean room", prefix all make commands with 'env -i '. See the env(1) manual page for more details. When upgrading from one major version to another it is generally best to upgrade to the latest code in the currently installed branch first, then do an upgrade to the new branch. This is the best-tested upgrade path, and has the highest probability of being successful. Please try this approach if you encounter problems with a major version upgrade. Since the stable 4.x branch point, one has generally been able to upgade from anywhere in the most recent stable branch to head / current (or even the last couple of stable branches). See the top of this file when there's an exception. When upgrading a live system, having a root shell around before installing anything can help undo problems. Not having a root shell around can lead to problems if pam has changed too much from your starting point to allow continued authentication after the upgrade. This file should be read as a log of events. When a later event changes information of a prior event, the prior event should not be deleted. Instead, a pointer to the entry with the new information should be placed in the old entry. Readers of this file should also sanity check older entries before relying on them blindly. Authors of new entries should write them with this in mind. ZFS notes --------- When upgrading the boot ZFS pool to a new version, always follow these two steps: 1.) recompile and reinstall the ZFS boot loader and boot block (this is part of "make buildworld" and "make installworld") 2.) update the ZFS boot block on your boot drive The following example updates the ZFS boot block on the first partition (freebsd-boot) of a GPT partitioned drive ada0: "gpart bootcode -p /boot/gptzfsboot -i 1 ada0" Non-boot pools do not need these updates. To build a kernel ----------------- If you are updating from a prior version of FreeBSD (even one just a few days old), you should follow this procedure. It is the most failsafe as it uses a /usr/obj tree with a fresh mini-buildworld, make kernel-toolchain make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE To test a kernel once --------------------- If you just want to boot a kernel once (because you are not sure if it works, or if you want to boot a known bad kernel to provide debugging information) run make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel nextboot -k testkernel To rebuild everything and install it on the current system. ----------------------------------------------------------- # Note: sometimes if you are running current you gotta do more than # is listed here if you are upgrading from a really old current. make buildworld make buildkernel KERNCONF=YOUR_KERNEL_HERE make installkernel KERNCONF=YOUR_KERNEL_HERE [1] [3] mergemaster -Fp [5] make installworld mergemaster -Fi [4] make delete-old [6] To cross-install current onto a separate partition -------------------------------------------------- # In this approach we use a separate partition to hold # current's root, 'usr', and 'var' directories. A partition # holding "/", "/usr" and "/var" should be about 2GB in # size. make buildworld make buildkernel KERNCONF=YOUR_KERNEL_HERE make installworld DESTDIR=${CURRENT_ROOT} -DDB_FROM_SRC make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT} cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd To upgrade in-place from stable to current ---------------------------------------------- make buildworld [9] make kernel KERNCONF=YOUR_KERNEL_HERE [8] [1] [3] mergemaster -Fp [5] make installworld mergemaster -Fi [4] make delete-old [6] Make sure that you've read the UPDATING file to understand the tweaks to various things you need. At this point in the life cycle of current, things change often and you are on your own to cope. The defaults can also change, so please read ALL of the UPDATING entries. Also, if you are tracking -current, you must be subscribed to freebsd-current@freebsd.org. Make sure that before you update your sources that you have read and understood all the recent messages there. If in doubt, please track -stable which has much fewer pitfalls. [1] If you have third party modules, such as vmware, you should disable them at this point so they don't crash your system on reboot. [3] From the bootblocks, boot -s, and then do fsck -p mount -u / mount -a cd src adjkerntz -i # if CMOS is wall time Also, when doing a major release upgrade, it is required that you boot into single user mode to do the installworld. [4] Note: This step is non-optional. Failure to do this step can result in a significant reduction in the functionality of the system. Attempting to do it by hand is not recommended and those that pursue this avenue should read this file carefully, as well as the archives of freebsd-current and freebsd-hackers mailing lists for potential gotchas. The -U option is also useful to consider. See mergemaster(8) for more information. [5] Usually this step is a noop. However, from time to time you may need to do this if you get unknown user in the following step. It never hurts to do it all the time. You may need to install a new mergemaster (cd src/usr.sbin/mergemaster && make install) after the buildworld before this step if you last updated from current before 20130425 or from -stable before 20130430. [6] This only deletes old files and directories. Old libraries can be deleted by "make delete-old-libs", but you have to make sure that no program is using those libraries anymore. [8] In order to have a kernel that can run the 4.x binaries needed to do an installworld, you must include the COMPAT_FREEBSD4 option in your kernel. Failure to do so may leave you with a system that is hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is required to run the 5.x binaries on more recent kernels. And so on for COMPAT_FREEBSD6 and COMPAT_FREEBSD7. Make sure that you merge any new devices from GENERIC since the last time you updated your kernel config file. [9] If CPUTYPE is defined in your /etc/make.conf, make sure to use the "?=" instead of the "=" assignment operator, so that buildworld can override the CPUTYPE if it needs to. MAKEOBJDIRPREFIX must be defined in an environment variable, and not on the command line, or in /etc/make.conf. buildworld will warn if it is improperly defined. FORMAT: This file contains a list, in reverse chronological order, of major breakages in tracking -current. It is not guaranteed to be a complete list of such breakages, and only contains entries since September 23, 2011. If you need to see UPDATING entries from before that date, you will need to fetch an UPDATING file from an older FreeBSD release. Copyright information: Copyright 1998-2009 M. Warner Losh. All Rights Reserved. Redistribution, publication, translation and use, with or without modification, in full or in part, in any form or format of this document are permitted without further permission from the author. THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``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 WARNER LOSH 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. Contact Warner Losh if you have any questions about your use of this document. $FreeBSD$ Index: head/sbin/devfs/rule.c =================================================================== --- head/sbin/devfs/rule.c (revision 327232) +++ head/sbin/devfs/rule.c (revision 327233) @@ -1,464 +1,464 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2002 Dima Dorfman. * 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. * * 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. */ /* * Rule subsystem manipulation. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include "extern.h" static void rulespec_infp(FILE *fp, unsigned long request, devfs_rsnum rsnum); static void rulespec_instr(struct devfs_rule *dr, const char *str, devfs_rsnum rsnum); static void rulespec_intok(struct devfs_rule *dr, int ac, char **av, devfs_rsnum rsnum); static void rulespec_outfp(FILE *fp, struct devfs_rule *dr); static command_t rule_add, rule_apply, rule_applyset; static command_t rule_del, rule_delset, rule_show, rule_showsets; static ctbl_t ctbl_rule = { { "add", rule_add }, { "apply", rule_apply }, { "applyset", rule_applyset }, { "del", rule_del }, { "delset", rule_delset }, { "show", rule_show }, { "showsets", rule_showsets }, { NULL, NULL } }; static struct intstr ist_type[] = { { "disk", D_DISK }, { "mem", D_MEM }, { "tape", D_TAPE }, { "tty", D_TTY }, { NULL, -1 } }; static devfs_rsnum in_rsnum; int rule_main(int ac, char **av) { struct cmd *c; int ch; setprogname("devfs rule"); optreset = optind = 1; while ((ch = getopt(ac, av, "s:")) != -1) switch (ch) { case 's': in_rsnum = eatonum(optarg); break; default: usage(); } ac -= optind; av += optind; if (ac < 1) usage(); for (c = ctbl_rule; c->name != NULL; ++c) if (strcmp(c->name, av[0]) == 0) exit((*c->handler)(ac, av)); errx(1, "unknown command: %s", av[0]); } static int rule_add(int ac, char **av) { struct devfs_rule dr; int rv; if (ac < 2) usage(); if (strcmp(av[1], "-") == 0) rulespec_infp(stdin, DEVFSIO_RADD, in_rsnum); else { rulespec_intok(&dr, ac - 1, av + 1, in_rsnum); rv = ioctl(mpfd, DEVFSIO_RADD, &dr); if (rv == -1) err(1, "ioctl DEVFSIO_RADD"); } return (0); } static int rule_apply(int ac __unused, char **av __unused) { struct devfs_rule dr; devfs_rnum rnum; devfs_rid rid; int rv; if (ac < 2) usage(); if (!atonum(av[1], &rnum)) { if (strcmp(av[1], "-") == 0) rulespec_infp(stdin, DEVFSIO_RAPPLY, in_rsnum); else { rulespec_intok(&dr, ac - 1, av + 1, in_rsnum); rv = ioctl(mpfd, DEVFSIO_RAPPLY, &dr); if (rv == -1) err(1, "ioctl DEVFSIO_RAPPLY"); } } else { rid = mkrid(in_rsnum, rnum); rv = ioctl(mpfd, DEVFSIO_RAPPLYID, &rid); if (rv == -1) err(1, "ioctl DEVFSIO_RAPPLYID"); } return (0); } static int rule_applyset(int ac, char **av __unused) { int rv; if (ac != 1) usage(); rv = ioctl(mpfd, DEVFSIO_SAPPLY, &in_rsnum); if (rv == -1) err(1, "ioctl DEVFSIO_SAPPLY"); return (0); } static int rule_del(int ac __unused, char **av) { devfs_rid rid; int rv; if (av[1] == NULL) usage(); rid = mkrid(in_rsnum, eatoi(av[1])); rv = ioctl(mpfd, DEVFSIO_RDEL, &rid); if (rv == -1) err(1, "ioctl DEVFSIO_RDEL"); return (0); } static int rule_delset(int ac, char **av __unused) { struct devfs_rule dr; int rv; if (ac != 1) usage(); memset(&dr, '\0', sizeof(dr)); dr.dr_magic = DEVFS_MAGIC; dr.dr_id = mkrid(in_rsnum, 0); while (ioctl(mpfd, DEVFSIO_RGETNEXT, &dr) != -1) { rv = ioctl(mpfd, DEVFSIO_RDEL, &dr.dr_id); if (rv == -1) err(1, "ioctl DEVFSIO_RDEL"); } if (errno != ENOENT) err(1, "ioctl DEVFSIO_RGETNEXT"); return (0); } static int rule_show(int ac __unused, char **av) { struct devfs_rule dr; devfs_rnum rnum; int rv; memset(&dr, '\0', sizeof(dr)); dr.dr_magic = DEVFS_MAGIC; if (av[1] != NULL) { rnum = eatoi(av[1]); dr.dr_id = mkrid(in_rsnum, rnum - 1); rv = ioctl(mpfd, DEVFSIO_RGETNEXT, &dr); if (rv == -1) err(1, "ioctl DEVFSIO_RGETNEXT"); if (rid2rn(dr.dr_id) == rnum) rulespec_outfp(stdout, &dr); } else { dr.dr_id = mkrid(in_rsnum, 0); while (ioctl(mpfd, DEVFSIO_RGETNEXT, &dr) != -1) rulespec_outfp(stdout, &dr); if (errno != ENOENT) err(1, "ioctl DEVFSIO_RGETNEXT"); } return (0); } static int rule_showsets(int ac, char **av __unused) { devfs_rsnum rsnum; if (ac != 1) usage(); rsnum = 0; while (ioctl(mpfd, DEVFSIO_SGETNEXT, &rsnum) != -1) printf("%d\n", rsnum); if (errno != ENOENT) err(1, "ioctl DEVFSIO_SGETNEXT"); return (0); } int ruleset_main(int ac, char **av) { devfs_rsnum rsnum; int rv; setprogname("devfs ruleset"); if (ac < 2) usage(); rsnum = eatonum(av[1]); rv = ioctl(mpfd, DEVFSIO_SUSE, &rsnum); if (rv == -1) err(1, "ioctl DEVFSIO_SUSE"); return (0); } /* * Input rules from a file (probably the standard input). This * differs from the other rulespec_in*() routines in that it also * calls ioctl() for the rules, since it is impractical (and not very * useful) to return a list (or array) of rules, just so the caller - * can call call ioctl() for each of them. + * can call ioctl() for each of them. */ static void rulespec_infp(FILE *fp, unsigned long request, devfs_rsnum rsnum) { struct devfs_rule dr; char *line; int rv; assert(fp == stdin); /* XXX: De-hardcode "stdin" from error msg. */ while (efgetln(fp, &line)) { rulespec_instr(&dr, line, rsnum); rv = ioctl(mpfd, request, &dr); if (rv == -1) err(1, "ioctl"); free(line); /* efgetln() always malloc()s. */ } if (ferror(stdin)) err(1, "stdin"); } /* * Construct a /struct devfs_rule/ from a string. */ static void rulespec_instr(struct devfs_rule *dr, const char *str, devfs_rsnum rsnum) { char **av; int ac; tokenize(str, &ac, &av); if (ac == 0) errx(1, "unexpected end of rulespec"); rulespec_intok(dr, ac, av, rsnum); free(av[0]); free(av); } /* * Construct a /struct devfs_rule/ from ac and av. */ static void rulespec_intok(struct devfs_rule *dr, int ac __unused, char **av, devfs_rsnum rsnum) { struct intstr *is; struct passwd *pw; struct group *gr; devfs_rnum rnum; void *set; memset(dr, '\0', sizeof(*dr)); /* * We don't maintain ac hereinafter. */ if (av[0] == NULL) errx(1, "unexpected end of rulespec"); /* If the first argument is an integer, treat it as a rule number. */ if (!atonum(av[0], &rnum)) rnum = 0; /* auto-number */ else ++av; /* * These aren't table-driven since that would result in more * tiny functions than I care to deal with. */ for (;;) { if (av[0] == NULL) break; else if (strcmp(av[0], "type") == 0) { if (av[1] == NULL) errx(1, "expecting argument for type"); for (is = ist_type; is->s != NULL; ++is) if (strcmp(av[1], is->s) == 0) { dr->dr_dswflags |= is->i; break; } if (is->s == NULL) errx(1, "unknown type: %s", av[1]); dr->dr_icond |= DRC_DSWFLAGS; av += 2; } else if (strcmp(av[0], "path") == 0) { if (av[1] == NULL) errx(1, "expecting argument for path"); if (strlcpy(dr->dr_pathptrn, av[1], DEVFS_MAXPTRNLEN) >= DEVFS_MAXPTRNLEN) warnx("pattern specified too long; truncated"); dr->dr_icond |= DRC_PATHPTRN; av += 2; } else break; } while (av[0] != NULL) { if (strcmp(av[0], "hide") == 0) { dr->dr_iacts |= DRA_BACTS; dr->dr_bacts |= DRB_HIDE; ++av; } else if (strcmp(av[0], "unhide") == 0) { dr->dr_iacts |= DRA_BACTS; dr->dr_bacts |= DRB_UNHIDE; ++av; } else if (strcmp(av[0], "user") == 0) { if (av[1] == NULL) errx(1, "expecting argument for user"); dr->dr_iacts |= DRA_UID; pw = getpwnam(av[1]); if (pw != NULL) dr->dr_uid = pw->pw_uid; else dr->dr_uid = eatoi(av[1]); /* XXX overflow */ av += 2; } else if (strcmp(av[0], "group") == 0) { if (av[1] == NULL) errx(1, "expecting argument for group"); dr->dr_iacts |= DRA_GID; gr = getgrnam(av[1]); if (gr != NULL) dr->dr_gid = gr->gr_gid; else dr->dr_gid = eatoi(av[1]); /* XXX overflow */ av += 2; } else if (strcmp(av[0], "mode") == 0) { if (av[1] == NULL) errx(1, "expecting argument for mode"); dr->dr_iacts |= DRA_MODE; set = setmode(av[1]); if (set == NULL) errx(1, "invalid mode: %s", av[1]); dr->dr_mode = getmode(set, 0); av += 2; } else if (strcmp(av[0], "include") == 0) { if (av[1] == NULL) errx(1, "expecting argument for include"); dr->dr_iacts |= DRA_INCSET; dr->dr_incset = eatonum(av[1]); av += 2; } else errx(1, "unknown argument: %s", av[0]); } dr->dr_id = mkrid(rsnum, rnum); dr->dr_magic = DEVFS_MAGIC; } /* * Write a human-readable (and machine-parsable, by rulespec_in*()) * representation of dr to bufp. *bufp should be free(3)'d when the * caller is finished with it. */ static void rulespec_outfp(FILE *fp, struct devfs_rule *dr) { struct intstr *is; struct passwd *pw; struct group *gr; fprintf(fp, "%d", rid2rn(dr->dr_id)); if (dr->dr_icond & DRC_DSWFLAGS) for (is = ist_type; is->s != NULL; ++is) if (dr->dr_dswflags & is->i) fprintf(fp, " type %s", is->s); if (dr->dr_icond & DRC_PATHPTRN) fprintf(fp, " path %s", dr->dr_pathptrn); if (dr->dr_iacts & DRA_BACTS) { if (dr->dr_bacts & DRB_HIDE) fprintf(fp, " hide"); if (dr->dr_bacts & DRB_UNHIDE) fprintf(fp, " unhide"); } if (dr->dr_iacts & DRA_UID) { pw = getpwuid(dr->dr_uid); if (pw == NULL) fprintf(fp, " user %d", dr->dr_uid); else fprintf(fp, " user %s", pw->pw_name); } if (dr->dr_iacts & DRA_GID) { gr = getgrgid(dr->dr_gid); if (gr == NULL) fprintf(fp, " group %d", dr->dr_gid); else fprintf(fp, " group %s", gr->gr_name); } if (dr->dr_iacts & DRA_MODE) fprintf(fp, " mode %o", dr->dr_mode); if (dr->dr_iacts & DRA_INCSET) fprintf(fp, " include %d", dr->dr_incset); fprintf(fp, "\n"); } Index: head/tests/sys/fifo/fifo_io.c =================================================================== --- head/tests/sys/fifo/fifo_io.c (revision 327232) +++ head/tests/sys/fifo/fifo_io.c (revision 327233) @@ -1,1399 +1,1399 @@ /*- * Copyright (c) 2005 Robert N. M. Watson * 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. * * 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. * * $FreeBSD$ */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* * Regression test to exercise POSIX fifo I/O. * * We test a number of aspect of behavior, including: * * - If there's no data to read, then for blocking fifos, we block, and for * non-blocking, we return EAGAIN. * * - If we write ten bytes, ten bytes can be read, and they're the same * bytes, in the same order. * * - If we write two batches of five bytes, we can read the same ten bytes in * one read of ten bytes. * * - If we write ten bytes, we can read the same ten bytes in two reads of * five bytes each. * * - If we over-fill a buffer (by writing 512k, which we take to be a large * number above default buffer sizes), we block if there is no reader. * * - That once 512k (ish) is read from the other end, the blocked writer * wakes up. * * - When a fifo is empty, poll, select, kqueue, and fionread report it is * writable but not readable. * * - When a fifo has data in it, poll, select, and kqueue report that it is * writable. * * - XXX: blocked reader semantics? * * - XXX: event behavior on remote close? * * Although behavior of O_RDWR isn't defined for fifos by POSIX, we expect * "reasonable" behavior, and run some additional tests relating to event * management on O_RDWR fifo descriptors. */ #define KQUEUE_MAX_EVENT 8 /* * All activity occurs within a temporary directory created early in the * test. */ static char temp_dir[PATH_MAX]; static void __unused atexit_temp_dir(void) { rmdir(temp_dir); } static void makefifo(const char *fifoname, const char *testname) { if (mkfifo(fifoname, 0700) < 0) err(-1, "%s: makefifo: mkfifo: %s", testname, fifoname); } static void cleanfifo2(const char *fifoname, int fd1, int fd2) { if (fd1 != -1) close(fd1); if (fd2 != -1) close(fd2); (void)unlink(fifoname); } static void cleanfifo3(const char *fifoname, int fd1, int fd2, int fd3) { if (fd3 != -1) close(fd3); cleanfifo2(fifoname, fd1, fd2); } /* * Open two different file descriptors for a fifo: one read, one write. Do * so using non-blocking opens in order to avoid deadlocking the process. */ static int openfifo(const char *fifoname, int *reader_fdp, int *writer_fdp) { int error, fd1, fd2; fd1 = open(fifoname, O_RDONLY | O_NONBLOCK); if (fd1 < 0) return (-1); fd2 = open(fifoname, O_WRONLY | O_NONBLOCK); if (fd2 < 0) { error = errno; close(fd1); errno = error; return (-1); } *reader_fdp = fd1; *writer_fdp = fd2; return (0); } /* * Open one file descriptor for the fifo, supporting both read and write. */ static int openfifo_rw(const char *fifoname, int *fdp) { int fd; fd = open(fifoname, O_RDWR); if (fd < 0) return (-1); *fdp = fd; return (0); } static int set_nonblocking(int fd, const char *testname) { int flags; flags = fcntl(fd, F_GETFL); if (flags < 0) { warn("%s: fcntl(fd, F_GETFL)", testname); return(-1); } flags |= O_NONBLOCK; if (fcntl(fd, F_SETFL, flags) < 0) { warn("%s: fcntl(fd, 0x%x)", testname, flags); return (-1); } return (0); } static int set_blocking(int fd, const char *testname) { int flags; flags = fcntl(fd, F_GETFL); if (flags < 0) { warn("%s: fcntl(fd, F_GETFL)", testname); return(-1); } flags &= ~O_NONBLOCK; if (fcntl(fd, F_SETFL, flags) < 0) { warn("%s: fcntl(fd, 0x%x)", testname, flags); return (-1); } return (0); } /* * Drain a file descriptor (fifo) of any readable data. Note: resets the * blocking state. */ static int drain_fd(int fd, const char *testname) { ssize_t len; u_char ch; if (set_nonblocking(fd, testname) < 0) return (-1); while ((len = read(fd, &ch, sizeof(ch))) > 0); if (len < 0) { switch (errno) { case EAGAIN: return (0); default: warn("%s: drain_fd: read", testname); return (-1); } } warn("%s: drain_fd: read: returned 0 bytes", testname); return (-1); } /* * Simple I/O test: write ten integers, and make sure we get back the same * integers in the same order. This assumes a minimum fifo buffer > 10 * bytes in order to not block and deadlock. */ static void test_simpleio(void) { int i, reader_fd, writer_fd; u_char buffer[10]; ssize_t len; makefifo("testfifo", __func__); if (openfifo("testfifo", &reader_fd, &writer_fd) < 0) { warn("test_simpleio: openfifo: testfifo"); cleanfifo2("testfifo", -1, -1); exit(-1); } for (i = 0; i < 10; i++) buffer[i] = i; len = write(writer_fd, (char *)buffer, sizeof(buffer)); if (len < 0) { warn("test_simpleio: write"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (len != sizeof(buffer)) { warnx("test_simplio: tried %zu but wrote %zd", sizeof(buffer), len); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } len = read(reader_fd, (char *)buffer, sizeof(buffer)); if (len < 0) { warn("test_simpleio: read"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (len != sizeof(buffer)) { warnx("test_simpleio: tried %zu but read %zd", sizeof(buffer), len); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } for (i = 0; i < 10; i++) { if (buffer[i] == i) continue; warnx("test_simpleio: write byte %d as 0x%02x, but read " "0x%02x", i, i, buffer[i]); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } cleanfifo2("testfifo", reader_fd, writer_fd); } static volatile int alarm_fired; /* * Non-destructive SIGALRM handler. */ static void sigalarm(int signum __unused) { alarm_fired = 1; } /* * Wrapper function for write, which uses a timer to interrupt any blocking. * Because we can't reliably detect EINTR for blocking I/O, we also track * whether or not our timeout fired. */ static int __unused timed_write(int fd, void *data, size_t len, ssize_t *written_lenp, int timeout, int *timedoutp, const char *testname) { struct sigaction act, oact; ssize_t written_len; int error; alarm_fired = 0; bzero(&act, sizeof(oact)); act.sa_handler = sigalarm; if (sigaction(SIGALRM, &act, &oact) < 0) { warn("%s: timed_write: sigaction", testname); return (-1); } alarm(timeout); written_len = write(fd, data, len); error = errno; alarm(0); if (sigaction(SIGALRM, &oact, NULL) < 0) { warn("%s: timed_write: sigaction", testname); return (-1); } if (alarm_fired) *timedoutp = 1; else *timedoutp = 0; errno = error; if (written_len < 0) return (-1); *written_lenp = written_len; return (0); } /* * Wrapper function for read, which uses a timer to interrupt any blocking. * Because we can't reliably detect EINTR for blocking I/O, we also track * whether or not our timeout fired. */ static int timed_read(int fd, void *data, size_t len, ssize_t *read_lenp, int timeout, int *timedoutp, const char *testname) { struct sigaction act, oact; ssize_t read_len; int error; alarm_fired = 0; bzero(&act, sizeof(oact)); act.sa_handler = sigalarm; if (sigaction(SIGALRM, &act, &oact) < 0) { warn("%s: timed_write: sigaction", testname); return (-1); } alarm(timeout); read_len = read(fd, data, len); error = errno; alarm(0); if (sigaction(SIGALRM, &oact, NULL) < 0) { warn("%s: timed_write: sigaction", testname); return (-1); } if (alarm_fired) *timedoutp = 1; else *timedoutp = 0; errno = error; if (read_len < 0) return (-1); *read_lenp = read_len; return (0); } /* * This test operates on blocking and non-blocking fifo file descriptors, in * order to determine whether they block at good moments or not. By good we * mean: don't block for non-blocking sockets, and do block for blocking * ones, assuming there isn't I/O buffer to satisfy the request. * * We use a timeout of 5 seconds, concluding that in 5 seconds either all I/O * that can take place will, and that if we reach the end of the timeout, * then blocking has occurred. * * We assume that the buffer size on a fifo is <512K, and as such, that * writing that much data without an active reader will result in blocking. */ static void test_blocking_read_empty(void) { int reader_fd, ret, timedout, writer_fd; ssize_t len; u_char ch; makefifo("testfifo", __func__); if (openfifo("testfifo", &reader_fd, &writer_fd) < 0) { warn("test_blocking_read_empty: openfifo: testfifo"); cleanfifo2("testfifo", -1, -1); exit(-1); } /* * Read one byte from an empty blocking fifo, block as there is no * data. */ if (set_blocking(reader_fd, __func__) < 0) { cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } ret = timed_read(reader_fd, &ch, sizeof(ch), &len, 5, &timedout, __func__); if (ret != -1) { warnx("test_blocking_read_empty: timed_read: returned " "success"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (errno != EINTR) { warn("test_blocking_read_empty: timed_read"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } /* * Read one byte from an empty non-blocking fifo, return EAGAIN as * there is no data. */ if (set_nonblocking(reader_fd, __func__) < 0) { cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } ret = timed_read(reader_fd, &ch, sizeof(ch), &len, 5, &timedout, __func__); if (ret != -1) { warnx("test_blocking_read_empty: timed_read: returned " "success"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (errno != EAGAIN) { warn("test_blocking_read_empty: timed_read"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } cleanfifo2("testfifo", reader_fd, writer_fd); } /* * Write one byte to an empty fifo, then try to read one byte and make sure * we don't block in either the write or the read. This tests both for * improper blocking in the send and receive code. */ static void test_blocking_one_byte(void) { int reader_fd, ret, timedout, writer_fd; ssize_t len; u_char ch; makefifo("testfifo", __func__); if (openfifo("testfifo", &reader_fd, &writer_fd) < 0) { warn("test_blocking: openfifo: testfifo"); cleanfifo2("testfifo", -1, -1); exit(-1); } if (set_blocking(writer_fd, __func__) < 0) { cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (set_blocking(reader_fd, __func__) < 0) { cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } ch = 0xfe; ret = timed_write(writer_fd, &ch, sizeof(ch), &len, 5, &timedout, __func__); if (ret < 0) { warn("test_blocking_one_byte: timed_write"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (len != sizeof(ch)) { warnx("test_blocking_one_byte: timed_write: tried to write " "%zu, wrote %zd", sizeof(ch), len); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } ch = 0xab; ret = timed_read(reader_fd, &ch, sizeof(ch), &len, 5, &timedout, __func__); if (ret < 0) { warn("test_blocking_one_byte: timed_read"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (len != sizeof(ch)) { warnx("test_blocking_one_byte: timed_read: wanted %zu, " "read %zd", sizeof(ch), len); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (ch != 0xfe) { warnx("test_blocking_one_byte: timed_read: expected to read " "0x%02x, read 0x%02x", 0xfe, ch); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } cleanfifo2("testfifo", reader_fd, writer_fd); } /* * Write one byte to an empty fifo, then try to read one byte and make sure * we don't get back EAGAIN. */ static void test_nonblocking_one_byte(void) { int reader_fd, ret, timedout, writer_fd; ssize_t len; u_char ch; makefifo("testfifo", __func__); if (openfifo("testfifo", &reader_fd, &writer_fd) < 0) { warn("test_nonblocking: openfifo: testfifo"); cleanfifo2("testfifo", -1, -1); exit(-1); } if (set_nonblocking(reader_fd, __func__) < 0) { cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } ch = 0xfe; ret = timed_write(writer_fd, &ch, sizeof(ch), &len, 5, &timedout, __func__); if (ret < 0) { warn("test_nonblocking_one_byte: timed_write"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (len != sizeof(ch)) { warnx("test_nonblocking_one_byte: timed_write: tried to write " "%zu, wrote %zd", sizeof(ch), len); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } ch = 0xab; ret = timed_read(reader_fd, &ch, sizeof(ch), &len, 5, &timedout, __func__); if (ret < 0) { warn("test_nonblocking_one_byte: timed_read"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (len != sizeof(ch)) { warnx("test_nonblocking_one_byte: timed_read: wanted %zu, read " "%zd", sizeof(ch), len); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (ch != 0xfe) { warnx("test_nonblocking_one_byte: timed_read: expected to read " "0x%02x, read 0x%02x", 0xfe, ch); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } cleanfifo2("testfifo", reader_fd, writer_fd); } /* * First of two test cases involving a 512K buffer: write the buffer into a * blocking file descriptor. We'd like to know it blocks, but the closest we * can get is to see if SIGALRM fired during the I/O resulting in a partial * write. */ static void test_blocking_partial_write(void) { int reader_fd, ret, timedout, writer_fd; u_char *buffer; ssize_t len; makefifo("testfifo", __func__); if (openfifo("testfifo", &reader_fd, &writer_fd) < 0) { warn("test_blocking_partial_write: openfifo: testfifo"); cleanfifo2("testfifo", -1, -1); exit(-1); } if (set_blocking(writer_fd, __func__) < 0) { cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } buffer = malloc(512*1024); if (buffer == NULL) { warn("test_blocking_partial_write: malloc"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } bzero(buffer, 512*1024); ret = timed_write(writer_fd, buffer, 512*1024, &len, 5, &timedout, __func__); if (ret < 0) { warn("test_blocking_partial_write: timed_write"); free(buffer); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (!timedout) { warnx("test_blocking_partial_write: timed_write: blocking " "socket didn't time out"); free(buffer); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } free(buffer); if (drain_fd(reader_fd, __func__) < 0) { cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } cleanfifo2("testfifo", reader_fd, writer_fd); } /* * Write a 512K buffer to an empty fifo using a non-blocking file descriptor, * and make sure it doesn't block. */ static void test_nonblocking_partial_write(void) { int reader_fd, ret, timedout, writer_fd; u_char *buffer; ssize_t len; makefifo("testfifo", __func__); if (openfifo("testfifo", &reader_fd, &writer_fd) < 0) { warn("test_blocking_partial_write: openfifo: testfifo"); cleanfifo2("testfifo", -1, -1); exit(-1); } if (set_nonblocking(writer_fd, __func__) < 0) { cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } buffer = malloc(512*1024); if (buffer == NULL) { warn("test_blocking_partial_write: malloc"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } bzero(buffer, 512*1024); ret = timed_write(writer_fd, buffer, 512*1024, &len, 5, &timedout, __func__); if (ret < 0) { warn("test_blocking_partial_write: timed_write"); free(buffer); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (timedout) { warnx("test_blocking_partial_write: timed_write: " "non-blocking socket timed out"); free(buffer); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (len == 0 || len >= 512*1024) { warnx("test_blocking_partial_write: timed_write: requested " "%d, sent %zd", 512*1024, len); free(buffer); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } free(buffer); if (drain_fd(reader_fd, __func__) < 0) { cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } cleanfifo2("testfifo", reader_fd, writer_fd); } /* * test_coalesce_big_read() verifies that data mingles in the fifo across * message boundaries by performing two small writes, then a bigger read * that should return data from both writes. */ static void test_coalesce_big_read(void) { int i, reader_fd, writer_fd; u_char buffer[10]; ssize_t len; makefifo("testfifo", __func__); if (openfifo("testfifo", &reader_fd, &writer_fd) < 0) { warn("test_coalesce_big_read: openfifo: testfifo"); cleanfifo2("testfifo", -1, -1); exit(-1); } /* Write five, write five, read ten. */ for (i = 0; i < 10; i++) buffer[i] = i; len = write(writer_fd, buffer, 5); if (len < 0) { warn("test_coalesce_big_read: write 5"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (len != 5) { warnx("test_coalesce_big_read: write 5 wrote %zd", len); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } len = write(writer_fd, buffer + 5, 5); if (len < 0) { warn("test_coalesce_big_read: write 5"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (len != 5) { warnx("test_coalesce_big_read: write 5 wrote %zd", len); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } len = read(reader_fd, buffer, 10); if (len < 0) { warn("test_coalesce_big_read: read 10"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (len != 10) { warnx("test_coalesce_big_read: read 10 read %zd", len); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } for (i = 0; i < 10; i++) { if (buffer[i] == i) continue; warnx("test_coalesce_big_read: expected to read 0x%02x, " "read 0x%02x", i, buffer[i]); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } cleanfifo2("testfifo", -1, -1); } /* * test_coalesce_big_write() verifies that data mingles in the fifo across * message boundaries by performing one big write, then two smaller reads * that should return sequential elements of data from the write. */ static void test_coalesce_big_write(void) { int i, reader_fd, writer_fd; u_char buffer[10]; ssize_t len; makefifo("testfifo", __func__); if (openfifo("testfifo", &reader_fd, &writer_fd) < 0) { warn("test_coalesce_big_write: openfifo: testfifo"); cleanfifo2("testfifo", -1, -1); exit(-1); } /* Write ten, read five, read five. */ for (i = 0; i < 10; i++) buffer[i] = i; len = write(writer_fd, buffer, 10); if (len < 0) { warn("test_coalesce_big_write: write 10"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (len != 10) { warnx("test_coalesce_big_write: write 10 wrote %zd", len); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } len = read(reader_fd, buffer, 5); if (len < 0) { warn("test_coalesce_big_write: read 5"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (len != 5) { warnx("test_coalesce_big_write: read 5 read %zd", len); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } len = read(reader_fd, buffer + 5, 5); if (len < 0) { warn("test_coalesce_big_write: read 5"); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (len != 5) { warnx("test_coalesce_big_write: read 5 read %zd", len); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } for (i = 0; i < 10; i++) { if (buffer[i] == i) continue; warnx("test_coalesce_big_write: expected to read 0x%02x, " "read 0x%02x", i, buffer[i]); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } cleanfifo2("testfifo", -1, -1); } static int poll_status(int fd, int *readable, int *writable, int *exception, const char *testname) { struct pollfd fds[1]; fds[0].fd = fd; fds[0].events = POLLIN | POLLOUT | POLLERR; fds[0].revents = 0; if (poll(fds, 1, 0) < 0) { warn("%s: poll", testname); return (-1); } *readable = (fds[0].revents & POLLIN) ? 1 : 0; *writable = (fds[0].revents & POLLOUT) ? 1 : 0; *exception = (fds[0].revents & POLLERR) ? 1 : 0; return (0); } static int select_status(int fd, int *readable, int *writable, int *exception, const char *testname) { struct fd_set readfds, writefds, exceptfds; struct timeval timeout; FD_ZERO(&readfds); FD_ZERO(&writefds); FD_ZERO(&exceptfds); FD_SET(fd, &readfds); FD_SET(fd, &writefds); FD_SET(fd, &exceptfds); timeout.tv_sec = 0; timeout.tv_usec = 0; if (select(fd+1, &readfds, &writefds, &exceptfds, &timeout) < 0) { warn("%s: select", testname); return (-1); } *readable = FD_ISSET(fd, &readfds) ? 1 : 0; *writable = FD_ISSET(fd, &writefds) ? 1 : 0; *exception = FD_ISSET(fd, &exceptfds) ? 1 : 0; return (0); } /* * Given an existing kqueue, set up read and write event filters for the * passed file descriptor. Typically called once for the read endpoint, and * once for the write endpoint. */ static int kqueue_setup(int kqueue_fd, int fd, const char *testname) { struct kevent kevent_changelist[2]; struct kevent kevent_eventlist[KQUEUE_MAX_EVENT], *kp; struct timespec timeout; int i, ret; timeout.tv_sec = 0; timeout.tv_nsec = 0; bzero(&kevent_changelist, sizeof(kevent_changelist)); EV_SET(&kevent_changelist[0], fd, EVFILT_READ, EV_ADD, 0, 0, 0); EV_SET(&kevent_changelist[1], fd, EVFILT_WRITE, EV_ADD, 0, 0, 0); bzero(&kevent_eventlist, sizeof(kevent_eventlist)); ret = kevent(kqueue_fd, kevent_changelist, 2, kevent_eventlist, KQUEUE_MAX_EVENT, &timeout); if (ret < 0) { warn("%s:%s: kevent initial register", testname, __func__); return (-1); } /* * Verify that the events registered alright. */ for (i = 0; i < ret; i++) { kp = &kevent_eventlist[i]; if (kp->flags != EV_ERROR) continue; errno = kp->data; warn("%s:%s: kevent register index %d", testname, __func__, i); return (-1); } return (0); } static int kqueue_status(int kqueue_fd, int fd, int *readable, int *writable, int *exception, const char *testname) { struct kevent kevent_eventlist[KQUEUE_MAX_EVENT], *kp; struct timespec timeout; int i, ret; timeout.tv_sec = 0; timeout.tv_nsec = 0; ret = kevent(kqueue_fd, NULL, 0, kevent_eventlist, KQUEUE_MAX_EVENT, &timeout); if (ret < 0) { warn("%s: %s: kevent", testname, __func__); return (-1); } *readable = *writable = *exception = 0; for (i = 0; i < ret; i++) { kp = &kevent_eventlist[i]; if (kp->ident != (u_int)fd) continue; if (kp->filter == EVFILT_READ) *readable = 1; if (kp->filter == EVFILT_WRITE) *writable = 1; } return (0); } static int fionread_status(int fd, int *readable, const char *testname) { int i; if (ioctl(fd, FIONREAD, &i) < 0) { warn("%s: ioctl(FIONREAD)", testname); return (-1); } if (i > 0) *readable = 1; else *readable = 0; return (0); } #define READABLE 1 #define WRITABLE 1 #define EXCEPTION 1 #define NOT_READABLE 0 #define NOT_WRITABLE 0 #define NOT_EXCEPTION 0 static int assert_status(int fd, int kqueue_fd, int assert_readable, int assert_writable, int assert_exception, const char *testname, const char *conditionname, const char *fdname) { int readable, writable, exception; if (poll_status(fd, &readable, &writable, &exception, testname) < 0) return (-1); if (readable != assert_readable || writable != assert_writable || exception != assert_exception) { warnx("%s: %s polls r:%d, w:%d, e:%d on %s", testname, fdname, readable, writable, exception, conditionname); return (-1); } if (select_status(fd, &readable, &writable, &exception, testname) < 0) return (-1); if (readable != assert_readable || writable != assert_writable || exception != assert_exception) { warnx("%s: %s selects r:%d, w:%d, e:%d on %s", testname, fdname, readable, writable, exception, conditionname); return (-1); } if (kqueue_status(kqueue_fd, fd, &readable, &writable, &exception, testname) < 0) return (-1); if (readable != assert_readable || writable != assert_writable || exception != assert_exception) { warnx("%s: %s kevent r:%d, w:%d, e:%d on %s", testname, fdname, readable, writable, exception, conditionname); return (-1); } if (fionread_status(fd, &readable, __func__) < 0) return (-1); if (readable != assert_readable) { warnx("%s: %s fionread r:%d on %s", testname, fdname, readable, conditionname); return (-1); } return (0); } /* * test_events() uses poll(), select(), and kevent() to query the status of * fifo file descriptors and determine whether they match expected state * based on earlier semantic tests: specifically, whether or not poll/select/ * kevent will correctly inform on readable/writable state following I/O. * * It would be nice to also test status changes as a result of closing of one * or another fifo endpoint. */ static void test_events_outofbox(void) { int kqueue_fd, reader_fd, writer_fd; makefifo("testfifo", __func__); if (openfifo("testfifo", &reader_fd, &writer_fd) < 0) { warn("test_events_outofbox: openfifo: testfifo"); cleanfifo2("testfifo", -1, -1); exit(-1); } kqueue_fd = kqueue(); if (kqueue_fd < 0) { warn("%s: kqueue", __func__); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (kqueue_setup(kqueue_fd, reader_fd, __func__) < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } if (kqueue_setup(kqueue_fd, writer_fd, __func__) < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } /* * Make sure that fresh, out-of-the-box fifo file descriptors have * good initial states. The reader_fd should have no active state, * since it will not be readable (no data in pipe), writable (it's * a read-only descriptor), and there's no reason for error yet. */ if (assert_status(reader_fd, kqueue_fd, NOT_READABLE, NOT_WRITABLE, NOT_EXCEPTION, __func__, "create", "reader_fd") < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } /* * Make sure that fresh, out-of-the-box fifo file descriptors have * good initial states. The writer_fd should be ready to write. */ if (assert_status(writer_fd, kqueue_fd, NOT_READABLE, WRITABLE, NOT_EXCEPTION, __func__, "create", "writer_fd") < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); } static void test_events_write_read_byte(void) { int kqueue_fd, reader_fd, writer_fd; ssize_t len; u_char ch; makefifo("testfifo", __func__); if (openfifo("testfifo", &reader_fd, &writer_fd) < 0) { warn("test_events_write_read_byte: openfifo: testfifo"); cleanfifo2("testfifo", -1, -1); exit(-1); } kqueue_fd = kqueue(); if (kqueue_fd < 0) { warn("%s: kqueue", __func__); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (kqueue_setup(kqueue_fd, reader_fd, __func__) < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } if (kqueue_setup(kqueue_fd, writer_fd, __func__) < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } /* * Write a byte to the fifo, and make sure that the read end becomes * readable, and that the write end remains writable (small write). */ ch = 0x00; len = write(writer_fd, &ch, sizeof(ch)); if (len < 0) { warn("%s: write", __func__); cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } if (assert_status(reader_fd, kqueue_fd, READABLE, NOT_WRITABLE, NOT_EXCEPTION, __func__, "write", "reader_fd") < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } /* * the writer_fd should remain writable. */ if (assert_status(writer_fd, kqueue_fd, NOT_READABLE, WRITABLE, NOT_EXCEPTION, __func__, "write", "writer_fd") < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } /* - * Read the byte from the reader_fd, and now confirm that that fifo + * Read the byte from the reader_fd, and now confirm that the fifo * becomes unreadable. */ len = read(reader_fd, &ch, sizeof(ch)); if (len < 0) { warn("%s: read", __func__); cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } if (assert_status(reader_fd, kqueue_fd, NOT_READABLE, NOT_WRITABLE, NOT_EXCEPTION, __func__, "write+read", "reader_fd") < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } /* * The writer_fd should remain writable. */ if (assert_status(writer_fd, kqueue_fd, NOT_READABLE, WRITABLE, NOT_EXCEPTION, __func__, "write+read", "writer_fd") < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); } /* * Write a 512k buffer to the fifo in non-blocking mode, and make sure that * the write end becomes un-writable as a result of a partial write that * fills the fifo buffer. */ static void test_events_partial_write(void) { int kqueue_fd, reader_fd, writer_fd; u_char *buffer; ssize_t len; makefifo("testfifo", __func__); if (openfifo("testfifo", &reader_fd, &writer_fd) < 0) { warn("test_events_partial_write: openfifo: testfifo"); cleanfifo2("testfifo", -1, -1); exit(-1); } kqueue_fd = kqueue(); if (kqueue_fd < 0) { warn("%s: kqueue", __func__); cleanfifo2("testfifo", reader_fd, writer_fd); exit(-1); } if (kqueue_setup(kqueue_fd, reader_fd, __func__) < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } if (kqueue_setup(kqueue_fd, writer_fd, __func__) < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } if (set_nonblocking(writer_fd, "test_events") < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } buffer = malloc(512*1024); if (buffer == NULL) { warn("test_events_partial_write: malloc"); cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } bzero(buffer, 512*1024); len = write(writer_fd, buffer, 512*1024); if (len < 0) { warn("test_events_partial_write: write"); free(buffer); cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } free(buffer); if (assert_status(writer_fd, kqueue_fd, NOT_READABLE, NOT_WRITABLE, NOT_EXCEPTION, __func__, "big write", "writer_fd") < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } if (drain_fd(reader_fd, "test_events") < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } /* * Test that the writer_fd has been restored to writable state after * draining. */ if (assert_status(writer_fd, kqueue_fd, NOT_READABLE, WRITABLE, NOT_EXCEPTION, __func__, "big write + drain", "writer_fd") < 0) { cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); exit(-1); } cleanfifo3("testfifo", reader_fd, writer_fd, kqueue_fd); } /* * We don't comprehensively test O_RDWR file descriptors, but do run a couple * of event tests to make sure that the fifo implementation doesn't mixed up * status checks. In particular, at least one past FreeBSD bug exists in * which the FIONREAD test was performed on the wrong socket implementing the * fifo, resulting in the fifo never returning readable. */ static void test_events_rdwr(void) { int fd, kqueue_fd; ssize_t len; char ch; makefifo("testfifo", __func__); if (openfifo_rw("testfifo", &fd) < 0) { warn("%s: openfifo_rw: testfifo", __func__); cleanfifo2("testfifo", -1, -1); exit(-1); } kqueue_fd = kqueue(); if (kqueue_fd < 0) { warn("%s: kqueue", __func__); cleanfifo2("testifo", fd, -1); exit(-1); } if (kqueue_setup(kqueue_fd, fd, __func__) < 0) { cleanfifo2("testfifo", fd, kqueue_fd); exit(-1); } /* * On first creation, the O_RDWR descriptor should be writable but * not readable. */ if (assert_status(fd, kqueue_fd, NOT_READABLE, WRITABLE, NOT_EXCEPTION, __func__, "create", "fd") < 0) { cleanfifo2("testfifo", fd, kqueue_fd); exit(-1); } /* * Write a byte, which should cause the file descriptor to become * readable and writable. */ ch = 0x00; len = write(fd, &ch, sizeof(ch)); if (len < 0) { warn("%s: write", __func__); cleanfifo2("testfifo", fd, kqueue_fd); exit(-1); } if (assert_status(fd, kqueue_fd, READABLE, WRITABLE, NOT_EXCEPTION, __func__, "write", "fd") < 0) { cleanfifo2("testfifo", fd, kqueue_fd); exit(-1); } /* * Read a byte, which should cause the file descriptor to return to * simply being writable. */ len = read(fd, &ch, sizeof(ch)); if (len < 0) { warn("%s: read", __func__); cleanfifo2("testfifo", fd, kqueue_fd); exit(-1); } if (assert_status(fd, kqueue_fd, NOT_READABLE, WRITABLE, NOT_EXCEPTION, __func__, "write+read", "fd") < 0) { cleanfifo2("testfifo", fd, kqueue_fd); exit(-1); } cleanfifo2("testfifo", fd, kqueue_fd); } int main(void) { strcpy(temp_dir, "fifo_io.XXXXXXXXXXX"); if (mkdtemp(temp_dir) == NULL) err(-1, "mkdtemp"); atexit(atexit_temp_dir); if (chdir(temp_dir) < 0) err(-1, "chdir %s", temp_dir); test_simpleio(); test_blocking_read_empty(); test_blocking_one_byte(); test_nonblocking_one_byte(); test_blocking_partial_write(); test_nonblocking_partial_write(); test_coalesce_big_read(); test_coalesce_big_write(); test_events_outofbox(); test_events_write_read_byte(); test_events_partial_write(); test_events_rdwr(); return (0); } Index: head/tests/sys/kern/ptrace_test.c =================================================================== --- head/tests/sys/kern/ptrace_test.c (revision 327232) +++ head/tests/sys/kern/ptrace_test.c (revision 327233) @@ -1,3696 +1,3696 @@ /*- * Copyright (c) 2015 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. * * 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. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* * A variant of ATF_REQUIRE that is suitable for use in child * processes. This only works if the parent process is tripped up by * the early exit and fails some requirement itself. */ #define CHILD_REQUIRE(exp) do { \ if (!(exp)) \ child_fail_require(__FILE__, __LINE__, \ #exp " not met"); \ } while (0) static __dead2 void child_fail_require(const char *file, int line, const char *str) { char buf[128]; snprintf(buf, sizeof(buf), "%s:%d: %s\n", file, line, str); write(2, buf, strlen(buf)); _exit(32); } static void trace_me(void) { /* Attach the parent process as a tracer of this process. */ CHILD_REQUIRE(ptrace(PT_TRACE_ME, 0, NULL, 0) != -1); /* Trigger a stop. */ raise(SIGSTOP); } static void attach_child(pid_t pid) { pid_t wpid; int status; ATF_REQUIRE(ptrace(PT_ATTACH, pid, NULL, 0) == 0); wpid = waitpid(pid, &status, 0); ATF_REQUIRE(wpid == pid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); } static void wait_for_zombie(pid_t pid) { /* * Wait for a process to exit. This is kind of gross, but * there is not a better way. * * Prior to r325719, the kern.proc.pid. sysctl failed * with ESRCH. After that change, a valid struct kinfo_proc * is returned for zombies with ki_stat set to SZOMB. */ for (;;) { struct kinfo_proc kp; size_t len; int mib[4]; mib[0] = CTL_KERN; mib[1] = KERN_PROC; mib[2] = KERN_PROC_PID; mib[3] = pid; len = sizeof(kp); if (sysctl(mib, nitems(mib), &kp, &len, NULL, 0) == -1) { ATF_REQUIRE(errno == ESRCH); break; } if (kp.ki_stat == SZOMB) break; usleep(5000); } } /* * Verify that a parent debugger process "sees" the exit of a debugged * process exactly once when attached via PT_TRACE_ME. */ ATF_TC_WITHOUT_HEAD(ptrace__parent_wait_after_trace_me); ATF_TC_BODY(ptrace__parent_wait_after_trace_me, tc) { pid_t child, wpid; int status; ATF_REQUIRE((child = fork()) != -1); if (child == 0) { /* Child process. */ trace_me(); _exit(1); } /* Parent process. */ /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, child, (caddr_t)1, 0) != -1); /* The second wait() should report the exit status. */ wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); /* The child should no longer exist. */ wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that a parent debugger process "sees" the exit of a debugged * process exactly once when attached via PT_ATTACH. */ ATF_TC_WITHOUT_HEAD(ptrace__parent_wait_after_attach); ATF_TC_BODY(ptrace__parent_wait_after_attach, tc) { pid_t child, wpid; int cpipe[2], status; char c; ATF_REQUIRE(pipe(cpipe) == 0); ATF_REQUIRE((child = fork()) != -1); if (child == 0) { /* Child process. */ close(cpipe[0]); /* Wait for the parent to attach. */ CHILD_REQUIRE(read(cpipe[1], &c, sizeof(c)) == 0); _exit(1); } close(cpipe[1]); /* Parent process. */ /* Attach to the child process. */ attach_child(child); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, child, (caddr_t)1, 0) != -1); /* Signal the child to exit. */ close(cpipe[0]); /* The second wait() should report the exit status. */ wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); /* The child should no longer exist. */ wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that a parent process "sees" the exit of a debugged process only * after the debugger has seen it. */ ATF_TC_WITHOUT_HEAD(ptrace__parent_sees_exit_after_child_debugger); ATF_TC_BODY(ptrace__parent_sees_exit_after_child_debugger, tc) { pid_t child, debugger, wpid; int cpipe[2], dpipe[2], status; char c; ATF_REQUIRE(pipe(cpipe) == 0); ATF_REQUIRE((child = fork()) != -1); if (child == 0) { /* Child process. */ close(cpipe[0]); /* Wait for parent to be ready. */ CHILD_REQUIRE(read(cpipe[1], &c, sizeof(c)) == sizeof(c)); _exit(1); } close(cpipe[1]); ATF_REQUIRE(pipe(dpipe) == 0); ATF_REQUIRE((debugger = fork()) != -1); if (debugger == 0) { /* Debugger process. */ close(dpipe[0]); CHILD_REQUIRE(ptrace(PT_ATTACH, child, NULL, 0) != -1); wpid = waitpid(child, &status, 0); CHILD_REQUIRE(wpid == child); CHILD_REQUIRE(WIFSTOPPED(status)); CHILD_REQUIRE(WSTOPSIG(status) == SIGSTOP); CHILD_REQUIRE(ptrace(PT_CONTINUE, child, (caddr_t)1, 0) != -1); /* Signal parent that debugger is attached. */ CHILD_REQUIRE(write(dpipe[1], &c, sizeof(c)) == sizeof(c)); /* Wait for parent's failed wait. */ CHILD_REQUIRE(read(dpipe[1], &c, sizeof(c)) == 0); wpid = waitpid(child, &status, 0); CHILD_REQUIRE(wpid == child); CHILD_REQUIRE(WIFEXITED(status)); CHILD_REQUIRE(WEXITSTATUS(status) == 1); _exit(0); } close(dpipe[1]); /* Parent process. */ /* Wait for the debugger to attach to the child. */ ATF_REQUIRE(read(dpipe[0], &c, sizeof(c)) == sizeof(c)); /* Release the child. */ ATF_REQUIRE(write(cpipe[0], &c, sizeof(c)) == sizeof(c)); ATF_REQUIRE(read(cpipe[0], &c, sizeof(c)) == 0); close(cpipe[0]); wait_for_zombie(child); /* * This wait should return a pid of 0 to indicate no status to * report. The parent should see the child as non-exited * until the debugger sees the exit. */ wpid = waitpid(child, &status, WNOHANG); ATF_REQUIRE(wpid == 0); /* Signal the debugger to wait for the child. */ close(dpipe[0]); /* Wait for the debugger. */ wpid = waitpid(debugger, &status, 0); ATF_REQUIRE(wpid == debugger); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 0); /* The child process should now be ready. */ wpid = waitpid(child, &status, WNOHANG); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); } /* * Verify that a parent process "sees" the exit of a debugged process * only after a non-direct-child debugger has seen it. In particular, * various wait() calls in the parent must avoid failing with ESRCH by * checking the parent's orphan list for the debugee. */ ATF_TC_WITHOUT_HEAD(ptrace__parent_sees_exit_after_unrelated_debugger); ATF_TC_BODY(ptrace__parent_sees_exit_after_unrelated_debugger, tc) { pid_t child, debugger, fpid, wpid; int cpipe[2], dpipe[2], status; char c; ATF_REQUIRE(pipe(cpipe) == 0); ATF_REQUIRE((child = fork()) != -1); if (child == 0) { /* Child process. */ close(cpipe[0]); /* Wait for parent to be ready. */ CHILD_REQUIRE(read(cpipe[1], &c, sizeof(c)) == sizeof(c)); _exit(1); } close(cpipe[1]); ATF_REQUIRE(pipe(dpipe) == 0); ATF_REQUIRE((debugger = fork()) != -1); if (debugger == 0) { /* Debugger parent. */ /* * Fork again and drop the debugger parent so that the * debugger is not a child of the main parent. */ CHILD_REQUIRE((fpid = fork()) != -1); if (fpid != 0) _exit(2); /* Debugger process. */ close(dpipe[0]); CHILD_REQUIRE(ptrace(PT_ATTACH, child, NULL, 0) != -1); wpid = waitpid(child, &status, 0); CHILD_REQUIRE(wpid == child); CHILD_REQUIRE(WIFSTOPPED(status)); CHILD_REQUIRE(WSTOPSIG(status) == SIGSTOP); CHILD_REQUIRE(ptrace(PT_CONTINUE, child, (caddr_t)1, 0) != -1); /* Signal parent that debugger is attached. */ CHILD_REQUIRE(write(dpipe[1], &c, sizeof(c)) == sizeof(c)); /* Wait for parent's failed wait. */ CHILD_REQUIRE(read(dpipe[1], &c, sizeof(c)) == sizeof(c)); wpid = waitpid(child, &status, 0); CHILD_REQUIRE(wpid == child); CHILD_REQUIRE(WIFEXITED(status)); CHILD_REQUIRE(WEXITSTATUS(status) == 1); _exit(0); } close(dpipe[1]); /* Parent process. */ /* Wait for the debugger parent process to exit. */ wpid = waitpid(debugger, &status, 0); ATF_REQUIRE(wpid == debugger); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 2); /* A WNOHANG wait here should see the non-exited child. */ wpid = waitpid(child, &status, WNOHANG); ATF_REQUIRE(wpid == 0); /* Wait for the debugger to attach to the child. */ ATF_REQUIRE(read(dpipe[0], &c, sizeof(c)) == sizeof(c)); /* Release the child. */ ATF_REQUIRE(write(cpipe[0], &c, sizeof(c)) == sizeof(c)); ATF_REQUIRE(read(cpipe[0], &c, sizeof(c)) == 0); close(cpipe[0]); wait_for_zombie(child); /* * This wait should return a pid of 0 to indicate no status to * report. The parent should see the child as non-exited * until the debugger sees the exit. */ wpid = waitpid(child, &status, WNOHANG); ATF_REQUIRE(wpid == 0); /* Signal the debugger to wait for the child. */ ATF_REQUIRE(write(dpipe[0], &c, sizeof(c)) == sizeof(c)); /* Wait for the debugger. */ ATF_REQUIRE(read(dpipe[0], &c, sizeof(c)) == 0); close(dpipe[0]); /* The child process should now be ready. */ wpid = waitpid(child, &status, WNOHANG); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); } /* * The parent process should always act the same regardless of how the * debugger is attached to it. */ static __dead2 void follow_fork_parent(bool use_vfork) { pid_t fpid, wpid; int status; if (use_vfork) CHILD_REQUIRE((fpid = vfork()) != -1); else CHILD_REQUIRE((fpid = fork()) != -1); if (fpid == 0) /* Child */ _exit(2); wpid = waitpid(fpid, &status, 0); CHILD_REQUIRE(wpid == fpid); CHILD_REQUIRE(WIFEXITED(status)); CHILD_REQUIRE(WEXITSTATUS(status) == 2); _exit(1); } /* * Helper routine for follow fork tests. This waits for two stops * that report both "sides" of a fork. It returns the pid of the new * child process. */ static pid_t handle_fork_events(pid_t parent, struct ptrace_lwpinfo *ppl) { struct ptrace_lwpinfo pl; bool fork_reported[2]; pid_t child, wpid; int i, status; fork_reported[0] = false; fork_reported[1] = false; child = -1; /* * Each process should report a fork event. The parent should * report a PL_FLAG_FORKED event, and the child should report * a PL_FLAG_CHILD event. */ for (i = 0; i < 2; i++) { wpid = wait(&status); ATF_REQUIRE(wpid > 0); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE((pl.pl_flags & (PL_FLAG_FORKED | PL_FLAG_CHILD)) != 0); ATF_REQUIRE((pl.pl_flags & (PL_FLAG_FORKED | PL_FLAG_CHILD)) != (PL_FLAG_FORKED | PL_FLAG_CHILD)); if (pl.pl_flags & PL_FLAG_CHILD) { ATF_REQUIRE(wpid != parent); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(!fork_reported[1]); if (child == -1) child = wpid; else ATF_REQUIRE(child == wpid); if (ppl != NULL) ppl[1] = pl; fork_reported[1] = true; } else { ATF_REQUIRE(wpid == parent); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(!fork_reported[0]); if (child == -1) child = pl.pl_child_pid; else ATF_REQUIRE(child == pl.pl_child_pid); if (ppl != NULL) ppl[0] = pl; fork_reported[0] = true; } } return (child); } /* * Verify that a new child process is stopped after a followed fork and * that the traced parent sees the exit of the child after the debugger * when both processes remain attached to the debugger. */ ATF_TC_WITHOUT_HEAD(ptrace__follow_fork_both_attached); ATF_TC_BODY(ptrace__follow_fork_both_attached, tc) { pid_t children[2], fpid, wpid; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); follow_fork_parent(false); } /* Parent process. */ children[0] = fpid; /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(children[0], &status, 0); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_FOLLOW_FORK, children[0], NULL, 1) != -1); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); children[1] = handle_fork_events(children[0], NULL); ATF_REQUIRE(children[1] > 0); ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); ATF_REQUIRE(ptrace(PT_CONTINUE, children[1], (caddr_t)1, 0) != -1); /* * The child can't exit until the grandchild reports status, so the * grandchild should report its exit first to the debugger. */ wpid = wait(&status); ATF_REQUIRE(wpid == children[1]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 2); wpid = wait(&status); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that a new child process is stopped after a followed fork * and that the traced parent sees the exit of the child when the new * child process is detached after it reports its fork. */ ATF_TC_WITHOUT_HEAD(ptrace__follow_fork_child_detached); ATF_TC_BODY(ptrace__follow_fork_child_detached, tc) { pid_t children[2], fpid, wpid; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); follow_fork_parent(false); } /* Parent process. */ children[0] = fpid; /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(children[0], &status, 0); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_FOLLOW_FORK, children[0], NULL, 1) != -1); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); children[1] = handle_fork_events(children[0], NULL); ATF_REQUIRE(children[1] > 0); ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); ATF_REQUIRE(ptrace(PT_DETACH, children[1], (caddr_t)1, 0) != -1); /* * Should not see any status from the grandchild now, only the * child. */ wpid = wait(&status); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that a new child process is stopped after a followed fork * and that the traced parent sees the exit of the child when the * traced parent is detached after the fork. */ ATF_TC_WITHOUT_HEAD(ptrace__follow_fork_parent_detached); ATF_TC_BODY(ptrace__follow_fork_parent_detached, tc) { pid_t children[2], fpid, wpid; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); follow_fork_parent(false); } /* Parent process. */ children[0] = fpid; /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(children[0], &status, 0); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_FOLLOW_FORK, children[0], NULL, 1) != -1); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); children[1] = handle_fork_events(children[0], NULL); ATF_REQUIRE(children[1] > 0); ATF_REQUIRE(ptrace(PT_DETACH, children[0], (caddr_t)1, 0) != -1); ATF_REQUIRE(ptrace(PT_CONTINUE, children[1], (caddr_t)1, 0) != -1); /* * The child can't exit until the grandchild reports status, so the * grandchild should report its exit first to the debugger. * * Even though the child process is detached, it is still a * child of the debugger, so it will still report it's exit * after the grandchild. */ wpid = wait(&status); ATF_REQUIRE(wpid == children[1]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 2); wpid = wait(&status); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } static void attach_fork_parent(int cpipe[2]) { pid_t fpid; close(cpipe[0]); /* Double-fork to disassociate from the debugger. */ CHILD_REQUIRE((fpid = fork()) != -1); if (fpid != 0) _exit(3); /* Send the pid of the disassociated child to the debugger. */ fpid = getpid(); CHILD_REQUIRE(write(cpipe[1], &fpid, sizeof(fpid)) == sizeof(fpid)); /* Wait for the debugger to attach. */ CHILD_REQUIRE(read(cpipe[1], &fpid, sizeof(fpid)) == 0); } /* * Verify that a new child process is stopped after a followed fork and * that the traced parent sees the exit of the child after the debugger * when both processes remain attached to the debugger. In this test * the parent that forks is not a direct child of the debugger. */ ATF_TC_WITHOUT_HEAD(ptrace__follow_fork_both_attached_unrelated_debugger); ATF_TC_BODY(ptrace__follow_fork_both_attached_unrelated_debugger, tc) { pid_t children[2], fpid, wpid; int cpipe[2], status; ATF_REQUIRE(pipe(cpipe) == 0); ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { attach_fork_parent(cpipe); follow_fork_parent(false); } /* Parent process. */ close(cpipe[1]); /* Wait for the direct child to exit. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 3); /* Read the pid of the fork parent. */ ATF_REQUIRE(read(cpipe[0], &children[0], sizeof(children[0])) == sizeof(children[0])); /* Attach to the fork parent. */ attach_child(children[0]); ATF_REQUIRE(ptrace(PT_FOLLOW_FORK, children[0], NULL, 1) != -1); /* Continue the fork parent ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); /* Signal the fork parent to continue. */ close(cpipe[0]); children[1] = handle_fork_events(children[0], NULL); ATF_REQUIRE(children[1] > 0); ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); ATF_REQUIRE(ptrace(PT_CONTINUE, children[1], (caddr_t)1, 0) != -1); /* * The fork parent can't exit until the child reports status, * so the child should report its exit first to the debugger. */ wpid = wait(&status); ATF_REQUIRE(wpid == children[1]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 2); wpid = wait(&status); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that a new child process is stopped after a followed fork * and that the traced parent sees the exit of the child when the new * child process is detached after it reports its fork. In this test * the parent that forks is not a direct child of the debugger. */ ATF_TC_WITHOUT_HEAD(ptrace__follow_fork_child_detached_unrelated_debugger); ATF_TC_BODY(ptrace__follow_fork_child_detached_unrelated_debugger, tc) { pid_t children[2], fpid, wpid; int cpipe[2], status; ATF_REQUIRE(pipe(cpipe) == 0); ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { attach_fork_parent(cpipe); follow_fork_parent(false); } /* Parent process. */ close(cpipe[1]); /* Wait for the direct child to exit. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 3); /* Read the pid of the fork parent. */ ATF_REQUIRE(read(cpipe[0], &children[0], sizeof(children[0])) == sizeof(children[0])); /* Attach to the fork parent. */ attach_child(children[0]); ATF_REQUIRE(ptrace(PT_FOLLOW_FORK, children[0], NULL, 1) != -1); /* Continue the fork parent ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); /* Signal the fork parent to continue. */ close(cpipe[0]); children[1] = handle_fork_events(children[0], NULL); ATF_REQUIRE(children[1] > 0); ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); ATF_REQUIRE(ptrace(PT_DETACH, children[1], (caddr_t)1, 0) != -1); /* * Should not see any status from the child now, only the fork * parent. */ wpid = wait(&status); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that a new child process is stopped after a followed fork * and that the traced parent sees the exit of the child when the * traced parent is detached after the fork. In this test the parent * that forks is not a direct child of the debugger. */ ATF_TC_WITHOUT_HEAD(ptrace__follow_fork_parent_detached_unrelated_debugger); ATF_TC_BODY(ptrace__follow_fork_parent_detached_unrelated_debugger, tc) { pid_t children[2], fpid, wpid; int cpipe[2], status; ATF_REQUIRE(pipe(cpipe) == 0); ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { attach_fork_parent(cpipe); follow_fork_parent(false); } /* Parent process. */ close(cpipe[1]); /* Wait for the direct child to exit. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 3); /* Read the pid of the fork parent. */ ATF_REQUIRE(read(cpipe[0], &children[0], sizeof(children[0])) == sizeof(children[0])); /* Attach to the fork parent. */ attach_child(children[0]); ATF_REQUIRE(ptrace(PT_FOLLOW_FORK, children[0], NULL, 1) != -1); /* Continue the fork parent ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); /* Signal the fork parent to continue. */ close(cpipe[0]); children[1] = handle_fork_events(children[0], NULL); ATF_REQUIRE(children[1] > 0); ATF_REQUIRE(ptrace(PT_DETACH, children[0], (caddr_t)1, 0) != -1); ATF_REQUIRE(ptrace(PT_CONTINUE, children[1], (caddr_t)1, 0) != -1); /* * Should not see any status from the fork parent now, only * the child. */ wpid = wait(&status); ATF_REQUIRE(wpid == children[1]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 2); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that a child process does not see an unrelated debugger as its * parent but sees its original parent process. */ ATF_TC_WITHOUT_HEAD(ptrace__getppid); ATF_TC_BODY(ptrace__getppid, tc) { pid_t child, debugger, ppid, wpid; int cpipe[2], dpipe[2], status; char c; ATF_REQUIRE(pipe(cpipe) == 0); ATF_REQUIRE((child = fork()) != -1); if (child == 0) { /* Child process. */ close(cpipe[0]); /* Wait for parent to be ready. */ CHILD_REQUIRE(read(cpipe[1], &c, sizeof(c)) == sizeof(c)); /* Report the parent PID to the parent. */ ppid = getppid(); CHILD_REQUIRE(write(cpipe[1], &ppid, sizeof(ppid)) == sizeof(ppid)); _exit(1); } close(cpipe[1]); ATF_REQUIRE(pipe(dpipe) == 0); ATF_REQUIRE((debugger = fork()) != -1); if (debugger == 0) { /* Debugger process. */ close(dpipe[0]); CHILD_REQUIRE(ptrace(PT_ATTACH, child, NULL, 0) != -1); wpid = waitpid(child, &status, 0); CHILD_REQUIRE(wpid == child); CHILD_REQUIRE(WIFSTOPPED(status)); CHILD_REQUIRE(WSTOPSIG(status) == SIGSTOP); CHILD_REQUIRE(ptrace(PT_CONTINUE, child, (caddr_t)1, 0) != -1); /* Signal parent that debugger is attached. */ CHILD_REQUIRE(write(dpipe[1], &c, sizeof(c)) == sizeof(c)); /* Wait for traced child to exit. */ wpid = waitpid(child, &status, 0); CHILD_REQUIRE(wpid == child); CHILD_REQUIRE(WIFEXITED(status)); CHILD_REQUIRE(WEXITSTATUS(status) == 1); _exit(0); } close(dpipe[1]); /* Parent process. */ /* Wait for the debugger to attach to the child. */ ATF_REQUIRE(read(dpipe[0], &c, sizeof(c)) == sizeof(c)); /* Release the child. */ ATF_REQUIRE(write(cpipe[0], &c, sizeof(c)) == sizeof(c)); /* Read the parent PID from the child. */ ATF_REQUIRE(read(cpipe[0], &ppid, sizeof(ppid)) == sizeof(ppid)); close(cpipe[0]); ATF_REQUIRE(ppid == getpid()); /* Wait for the debugger. */ wpid = waitpid(debugger, &status, 0); ATF_REQUIRE(wpid == debugger); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 0); /* The child process should now be ready. */ wpid = waitpid(child, &status, WNOHANG); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); } /* * Verify that pl_syscall_code in struct ptrace_lwpinfo for a new * child process created via fork() reports the correct value. */ ATF_TC_WITHOUT_HEAD(ptrace__new_child_pl_syscall_code_fork); ATF_TC_BODY(ptrace__new_child_pl_syscall_code_fork, tc) { struct ptrace_lwpinfo pl[2]; pid_t children[2], fpid, wpid; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); follow_fork_parent(false); } /* Parent process. */ children[0] = fpid; /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(children[0], &status, 0); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_FOLLOW_FORK, children[0], NULL, 1) != -1); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); /* Wait for both halves of the fork event to get reported. */ children[1] = handle_fork_events(children[0], pl); ATF_REQUIRE(children[1] > 0); ATF_REQUIRE((pl[0].pl_flags & PL_FLAG_SCX) != 0); ATF_REQUIRE((pl[1].pl_flags & PL_FLAG_SCX) != 0); ATF_REQUIRE(pl[0].pl_syscall_code == SYS_fork); ATF_REQUIRE(pl[0].pl_syscall_code == pl[1].pl_syscall_code); ATF_REQUIRE(pl[0].pl_syscall_narg == pl[1].pl_syscall_narg); ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); ATF_REQUIRE(ptrace(PT_CONTINUE, children[1], (caddr_t)1, 0) != -1); /* * The child can't exit until the grandchild reports status, so the * grandchild should report its exit first to the debugger. */ wpid = wait(&status); ATF_REQUIRE(wpid == children[1]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 2); wpid = wait(&status); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that pl_syscall_code in struct ptrace_lwpinfo for a new * child process created via vfork() reports the correct value. */ ATF_TC_WITHOUT_HEAD(ptrace__new_child_pl_syscall_code_vfork); ATF_TC_BODY(ptrace__new_child_pl_syscall_code_vfork, tc) { struct ptrace_lwpinfo pl[2]; pid_t children[2], fpid, wpid; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); follow_fork_parent(true); } /* Parent process. */ children[0] = fpid; /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(children[0], &status, 0); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_FOLLOW_FORK, children[0], NULL, 1) != -1); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); /* Wait for both halves of the fork event to get reported. */ children[1] = handle_fork_events(children[0], pl); ATF_REQUIRE(children[1] > 0); ATF_REQUIRE((pl[0].pl_flags & PL_FLAG_SCX) != 0); ATF_REQUIRE((pl[1].pl_flags & PL_FLAG_SCX) != 0); ATF_REQUIRE(pl[0].pl_syscall_code == SYS_vfork); ATF_REQUIRE(pl[0].pl_syscall_code == pl[1].pl_syscall_code); ATF_REQUIRE(pl[0].pl_syscall_narg == pl[1].pl_syscall_narg); ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); ATF_REQUIRE(ptrace(PT_CONTINUE, children[1], (caddr_t)1, 0) != -1); /* * The child can't exit until the grandchild reports status, so the * grandchild should report its exit first to the debugger. */ wpid = wait(&status); ATF_REQUIRE(wpid == children[1]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 2); wpid = wait(&status); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } static void * simple_thread(void *arg __unused) { pthread_exit(NULL); } static __dead2 void simple_thread_main(void) { pthread_t thread; CHILD_REQUIRE(pthread_create(&thread, NULL, simple_thread, NULL) == 0); CHILD_REQUIRE(pthread_join(thread, NULL) == 0); exit(1); } /* * Verify that pl_syscall_code in struct ptrace_lwpinfo for a new * thread reports the correct value. */ ATF_TC_WITHOUT_HEAD(ptrace__new_child_pl_syscall_code_thread); ATF_TC_BODY(ptrace__new_child_pl_syscall_code_thread, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; lwpid_t mainlwp; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); simple_thread_main(); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); mainlwp = pl.pl_lwpid; /* * Continue the child ignoring the SIGSTOP and tracing all * system call exits. */ ATF_REQUIRE(ptrace(PT_TO_SCX, fpid, (caddr_t)1, 0) != -1); /* * Wait for the new thread to arrive. pthread_create() might * invoke any number of system calls. For now we just wait * for the new thread to arrive and make sure it reports a * valid system call code. If ptrace grows thread event * reporting then this test can be made more precise. */ for (;;) { wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE((pl.pl_flags & PL_FLAG_SCX) != 0); ATF_REQUIRE(pl.pl_syscall_code != 0); if (pl.pl_lwpid != mainlwp) /* New thread seen. */ break; ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); } /* Wait for the child to exit. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); for (;;) { wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); if (WIFEXITED(status)) break; ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); } ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that the expected LWP events are reported for a child thread. */ ATF_TC_WITHOUT_HEAD(ptrace__lwp_events); ATF_TC_BODY(ptrace__lwp_events, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; lwpid_t lwps[2]; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); simple_thread_main(); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); lwps[0] = pl.pl_lwpid; ATF_REQUIRE(ptrace(PT_LWP_EVENTS, wpid, NULL, 1) == 0); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The first event should be for the child thread's birth. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE((pl.pl_flags & (PL_FLAG_BORN | PL_FLAG_SCX)) == (PL_FLAG_BORN | PL_FLAG_SCX)); ATF_REQUIRE(pl.pl_lwpid != lwps[0]); lwps[1] = pl.pl_lwpid; ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The next event should be for the child thread's death. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE((pl.pl_flags & (PL_FLAG_EXITED | PL_FLAG_SCE)) == (PL_FLAG_EXITED | PL_FLAG_SCE)); ATF_REQUIRE(pl.pl_lwpid == lwps[1]); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The last event should be for the child process's exit. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } static void * exec_thread(void *arg __unused) { execl("/usr/bin/true", "true", NULL); exit(127); } static __dead2 void exec_thread_main(void) { pthread_t thread; CHILD_REQUIRE(pthread_create(&thread, NULL, exec_thread, NULL) == 0); for (;;) sleep(60); exit(1); } /* * Verify that the expected LWP events are reported for a multithreaded * process that calls execve(2). */ ATF_TC_WITHOUT_HEAD(ptrace__lwp_events_exec); ATF_TC_BODY(ptrace__lwp_events_exec, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; lwpid_t lwps[2]; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); exec_thread_main(); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); lwps[0] = pl.pl_lwpid; ATF_REQUIRE(ptrace(PT_LWP_EVENTS, wpid, NULL, 1) == 0); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The first event should be for the child thread's birth. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE((pl.pl_flags & (PL_FLAG_BORN | PL_FLAG_SCX)) == (PL_FLAG_BORN | PL_FLAG_SCX)); ATF_REQUIRE(pl.pl_lwpid != lwps[0]); lwps[1] = pl.pl_lwpid; ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* * The next event should be for the main thread's death due to * single threading from execve(). */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE((pl.pl_flags & (PL_FLAG_EXITED | PL_FLAG_SCE)) == (PL_FLAG_EXITED)); ATF_REQUIRE(pl.pl_lwpid == lwps[0]); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The next event should be for the child process's exec. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE((pl.pl_flags & (PL_FLAG_EXEC | PL_FLAG_SCX)) == (PL_FLAG_EXEC | PL_FLAG_SCX)); ATF_REQUIRE(pl.pl_lwpid == lwps[1]); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The last event should be for the child process's exit. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 0); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } static void handler(int sig __unused) { } static void signal_main(void) { signal(SIGINFO, handler); raise(SIGINFO); exit(0); } /* * Verify that the expected ptrace event is reported for a signal. */ ATF_TC_WITHOUT_HEAD(ptrace__siginfo); ATF_TC_BODY(ptrace__siginfo, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); signal_main(); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The next event should be for the SIGINFO. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGINFO); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_event == PL_EVENT_SIGNAL); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SI); ATF_REQUIRE(pl.pl_siginfo.si_code == SI_LWP); ATF_REQUIRE(pl.pl_siginfo.si_pid == wpid); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The last event should be for the child process's exit. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 0); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that the expected ptrace events are reported for PTRACE_EXEC. */ ATF_TC_WITHOUT_HEAD(ptrace__ptrace_exec_disable); ATF_TC_BODY(ptrace__ptrace_exec_disable, tc) { pid_t fpid, wpid; int events, status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); exec_thread(NULL); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); events = 0; ATF_REQUIRE(ptrace(PT_SET_EVENT_MASK, fpid, (caddr_t)&events, sizeof(events)) == 0); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* Should get one event at exit. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 0); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } ATF_TC_WITHOUT_HEAD(ptrace__ptrace_exec_enable); ATF_TC_BODY(ptrace__ptrace_exec_enable, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; int events, status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); exec_thread(NULL); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); events = PTRACE_EXEC; ATF_REQUIRE(ptrace(PT_SET_EVENT_MASK, fpid, (caddr_t)&events, sizeof(events)) == 0); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The next event should be for the child process's exec. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE((pl.pl_flags & (PL_FLAG_EXEC | PL_FLAG_SCX)) == (PL_FLAG_EXEC | PL_FLAG_SCX)); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The last event should be for the child process's exit. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 0); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } ATF_TC_WITHOUT_HEAD(ptrace__event_mask); ATF_TC_BODY(ptrace__event_mask, tc) { pid_t fpid, wpid; int events, status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); exit(0); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* PT_FOLLOW_FORK should toggle the state of PTRACE_FORK. */ ATF_REQUIRE(ptrace(PT_FOLLOW_FORK, fpid, NULL, 1) != -1); ATF_REQUIRE(ptrace(PT_GET_EVENT_MASK, fpid, (caddr_t)&events, sizeof(events)) == 0); ATF_REQUIRE(events & PTRACE_FORK); ATF_REQUIRE(ptrace(PT_FOLLOW_FORK, fpid, NULL, 0) != -1); ATF_REQUIRE(ptrace(PT_GET_EVENT_MASK, fpid, (caddr_t)&events, sizeof(events)) == 0); ATF_REQUIRE(!(events & PTRACE_FORK)); /* PT_LWP_EVENTS should toggle the state of PTRACE_LWP. */ ATF_REQUIRE(ptrace(PT_LWP_EVENTS, fpid, NULL, 1) != -1); ATF_REQUIRE(ptrace(PT_GET_EVENT_MASK, fpid, (caddr_t)&events, sizeof(events)) == 0); ATF_REQUIRE(events & PTRACE_LWP); ATF_REQUIRE(ptrace(PT_LWP_EVENTS, fpid, NULL, 0) != -1); ATF_REQUIRE(ptrace(PT_GET_EVENT_MASK, fpid, (caddr_t)&events, sizeof(events)) == 0); ATF_REQUIRE(!(events & PTRACE_LWP)); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* Should get one event at exit. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 0); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that the expected ptrace events are reported for PTRACE_VFORK. */ ATF_TC_WITHOUT_HEAD(ptrace__ptrace_vfork); ATF_TC_BODY(ptrace__ptrace_vfork, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; int events, status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); follow_fork_parent(true); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_GET_EVENT_MASK, fpid, (caddr_t)&events, sizeof(events)) == 0); events |= PTRACE_VFORK; ATF_REQUIRE(ptrace(PT_SET_EVENT_MASK, fpid, (caddr_t)&events, sizeof(events)) == 0); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) != -1); /* The next event should report the end of the vfork. */ wpid = wait(&status); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE((pl.pl_flags & PL_FLAG_VFORK_DONE) != 0); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) != -1); wpid = wait(&status); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } ATF_TC_WITHOUT_HEAD(ptrace__ptrace_vfork_follow); ATF_TC_BODY(ptrace__ptrace_vfork_follow, tc) { struct ptrace_lwpinfo pl[2]; pid_t children[2], fpid, wpid; int events, status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); follow_fork_parent(true); } /* Parent process. */ children[0] = fpid; /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(children[0], &status, 0); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_GET_EVENT_MASK, children[0], (caddr_t)&events, sizeof(events)) == 0); events |= PTRACE_FORK | PTRACE_VFORK; ATF_REQUIRE(ptrace(PT_SET_EVENT_MASK, children[0], (caddr_t)&events, sizeof(events)) == 0); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); /* Wait for both halves of the fork event to get reported. */ children[1] = handle_fork_events(children[0], pl); ATF_REQUIRE(children[1] > 0); ATF_REQUIRE((pl[0].pl_flags & PL_FLAG_VFORKED) != 0); ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); ATF_REQUIRE(ptrace(PT_CONTINUE, children[1], (caddr_t)1, 0) != -1); /* * The child can't exit until the grandchild reports status, so the * grandchild should report its exit first to the debugger. */ wpid = waitpid(children[1], &status, 0); ATF_REQUIRE(wpid == children[1]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 2); /* * The child should report it's vfork() completion before it * exits. */ wpid = wait(&status); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl[0], sizeof(pl[0])) != -1); ATF_REQUIRE((pl[0].pl_flags & PL_FLAG_VFORK_DONE) != 0); ATF_REQUIRE(ptrace(PT_CONTINUE, children[0], (caddr_t)1, 0) != -1); wpid = wait(&status); ATF_REQUIRE(wpid == children[0]); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * XXX: There's nothing inherently platform specific about this test, however a * userspace visible breakpoint() is a prerequisite. */ #if defined(__amd64__) || defined(__i386__) || defined(__sparc64__) /* * Verify that no more events are reported after PT_KILL except for the * process exit when stopped due to a breakpoint trap. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_KILL_breakpoint); ATF_TC_BODY(ptrace__PT_KILL_breakpoint, tc) { pid_t fpid, wpid; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); breakpoint(); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The second wait() should report hitting the breakpoint. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); /* Kill the child process. */ ATF_REQUIRE(ptrace(PT_KILL, fpid, 0, 0) == 0); /* The last wait() should report the SIGKILL. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSIGNALED(status)); ATF_REQUIRE(WTERMSIG(status) == SIGKILL); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } #endif /* defined(__amd64__) || defined(__i386__) || defined(__sparc64__) */ /* * Verify that no more events are reported after PT_KILL except for the * process exit when stopped inside of a system call. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_KILL_system_call); ATF_TC_BODY(ptrace__PT_KILL_system_call, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); getpid(); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Continue the child ignoring the SIGSTOP and tracing system calls. */ ATF_REQUIRE(ptrace(PT_SYSCALL, fpid, (caddr_t)1, 0) == 0); /* The second wait() should report a system call entry for getpid(). */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SCE); /* Kill the child process. */ ATF_REQUIRE(ptrace(PT_KILL, fpid, 0, 0) == 0); /* The last wait() should report the SIGKILL. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSIGNALED(status)); ATF_REQUIRE(WTERMSIG(status) == SIGKILL); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that no more events are reported after PT_KILL except for the * process exit when killing a multithreaded process. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_KILL_threads); ATF_TC_BODY(ptrace__PT_KILL_threads, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; lwpid_t main_lwp; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); simple_thread_main(); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); main_lwp = pl.pl_lwpid; ATF_REQUIRE(ptrace(PT_LWP_EVENTS, wpid, NULL, 1) == 0); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The first event should be for the child thread's birth. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE((pl.pl_flags & (PL_FLAG_BORN | PL_FLAG_SCX)) == (PL_FLAG_BORN | PL_FLAG_SCX)); ATF_REQUIRE(pl.pl_lwpid != main_lwp); /* Kill the child process. */ ATF_REQUIRE(ptrace(PT_KILL, fpid, 0, 0) == 0); /* The last wait() should report the SIGKILL. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSIGNALED(status)); ATF_REQUIRE(WTERMSIG(status) == SIGKILL); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } static void * mask_usr1_thread(void *arg) { pthread_barrier_t *pbarrier; sigset_t sigmask; pbarrier = (pthread_barrier_t*)arg; sigemptyset(&sigmask); sigaddset(&sigmask, SIGUSR1); CHILD_REQUIRE(pthread_sigmask(SIG_BLOCK, &sigmask, NULL) == 0); /* Sync up with other thread after sigmask updated. */ pthread_barrier_wait(pbarrier); for (;;) sleep(60); return (NULL); } /* * Verify that the SIGKILL from PT_KILL takes priority over other signals * and prevents spurious stops due to those other signals. */ ATF_TC(ptrace__PT_KILL_competing_signal); ATF_TC_HEAD(ptrace__PT_KILL_competing_signal, tc) { atf_tc_set_md_var(tc, "require.user", "root"); } ATF_TC_BODY(ptrace__PT_KILL_competing_signal, tc) { pid_t fpid, wpid; int status; cpuset_t setmask; pthread_t t; pthread_barrier_t barrier; struct sched_param sched_param; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { /* Bind to one CPU so only one thread at a time will run. */ CPU_ZERO(&setmask); CPU_SET(0, &setmask); cpusetid_t setid; CHILD_REQUIRE(cpuset(&setid) == 0); CHILD_REQUIRE(cpuset_setaffinity(CPU_LEVEL_CPUSET, CPU_WHICH_CPUSET, setid, sizeof(setmask), &setmask) == 0); CHILD_REQUIRE(pthread_barrier_init(&barrier, NULL, 2) == 0); CHILD_REQUIRE(pthread_create(&t, NULL, mask_usr1_thread, (void*)&barrier) == 0); /* * Give the main thread higher priority. The test always * assumes that, if both threads are able to run, the main * thread runs first. */ sched_param.sched_priority = (sched_get_priority_max(SCHED_FIFO) + sched_get_priority_min(SCHED_FIFO)) / 2; CHILD_REQUIRE(pthread_setschedparam(pthread_self(), SCHED_FIFO, &sched_param) == 0); sched_param.sched_priority -= RQ_PPQ; CHILD_REQUIRE(pthread_setschedparam(t, SCHED_FIFO, &sched_param) == 0); sigset_t sigmask; sigemptyset(&sigmask); sigaddset(&sigmask, SIGUSR2); CHILD_REQUIRE(pthread_sigmask(SIG_BLOCK, &sigmask, NULL) == 0); /* Sync up with other thread after sigmask updated. */ pthread_barrier_wait(&barrier); trace_me(); for (;;) sleep(60); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* Send a signal that only the second thread can handle. */ ATF_REQUIRE(kill(fpid, SIGUSR2) == 0); /* The second wait() should report the SIGUSR2. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGUSR2); /* Send a signal that only the first thread can handle. */ ATF_REQUIRE(kill(fpid, SIGUSR1) == 0); /* Replace the SIGUSR2 with a kill. */ ATF_REQUIRE(ptrace(PT_KILL, fpid, 0, 0) == 0); /* The last wait() should report the SIGKILL (not the SIGUSR signal). */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSIGNALED(status)); ATF_REQUIRE(WTERMSIG(status) == SIGKILL); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that the SIGKILL from PT_KILL takes priority over other stop events * and prevents spurious stops caused by those events. */ ATF_TC(ptrace__PT_KILL_competing_stop); ATF_TC_HEAD(ptrace__PT_KILL_competing_stop, tc) { atf_tc_set_md_var(tc, "require.user", "root"); } ATF_TC_BODY(ptrace__PT_KILL_competing_stop, tc) { pid_t fpid, wpid; int status; cpuset_t setmask; pthread_t t; pthread_barrier_t barrier; lwpid_t main_lwp; struct ptrace_lwpinfo pl; struct sched_param sched_param; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); /* Bind to one CPU so only one thread at a time will run. */ CPU_ZERO(&setmask); CPU_SET(0, &setmask); cpusetid_t setid; CHILD_REQUIRE(cpuset(&setid) == 0); CHILD_REQUIRE(cpuset_setaffinity(CPU_LEVEL_CPUSET, CPU_WHICH_CPUSET, setid, sizeof(setmask), &setmask) == 0); CHILD_REQUIRE(pthread_barrier_init(&barrier, NULL, 2) == 0); CHILD_REQUIRE(pthread_create(&t, NULL, mask_usr1_thread, (void*)&barrier) == 0); /* * Give the main thread higher priority. The test always * assumes that, if both threads are able to run, the main * thread runs first. */ sched_param.sched_priority = (sched_get_priority_max(SCHED_FIFO) + sched_get_priority_min(SCHED_FIFO)) / 2; CHILD_REQUIRE(pthread_setschedparam(pthread_self(), SCHED_FIFO, &sched_param) == 0); sched_param.sched_priority -= RQ_PPQ; CHILD_REQUIRE(pthread_setschedparam(t, SCHED_FIFO, &sched_param) == 0); sigset_t sigmask; sigemptyset(&sigmask); sigaddset(&sigmask, SIGUSR2); CHILD_REQUIRE(pthread_sigmask(SIG_BLOCK, &sigmask, NULL) == 0); /* Sync up with other thread after sigmask updated. */ pthread_barrier_wait(&barrier); /* Sync up with the test before doing the getpid(). */ raise(SIGSTOP); getpid(); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); main_lwp = pl.pl_lwpid; /* Continue the child ignoring the SIGSTOP and tracing system calls. */ ATF_REQUIRE(ptrace(PT_SYSCALL, fpid, (caddr_t)1, 0) == 0); /* * Continue until child is done with setup, which is indicated with * SIGSTOP. Ignore system calls in the meantime. */ for (;;) { wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); if (WSTOPSIG(status) == SIGTRAP) { ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & (PL_FLAG_SCE | PL_FLAG_SCX)); } else { ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); break; } ATF_REQUIRE(ptrace(PT_SYSCALL, fpid, (caddr_t)1, 0) == 0); } /* Proceed, allowing main thread to hit syscall entry for getpid(). */ ATF_REQUIRE(ptrace(PT_SYSCALL, fpid, (caddr_t)1, 0) == 0); wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_lwpid == main_lwp); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SCE); /* Prevent the main thread from hitting its syscall exit for now. */ ATF_REQUIRE(ptrace(PT_SUSPEND, main_lwp, 0, 0) == 0); /* * Proceed, allowing second thread to hit syscall exit for * pthread_barrier_wait(). */ ATF_REQUIRE(ptrace(PT_SYSCALL, fpid, (caddr_t)1, 0) == 0); wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_lwpid != main_lwp); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SCX); /* Send a signal that only the second thread can handle. */ ATF_REQUIRE(kill(fpid, SIGUSR2) == 0); ATF_REQUIRE(ptrace(PT_SYSCALL, fpid, (caddr_t)1, 0) == 0); /* The next wait() should report the SIGUSR2. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGUSR2); /* Allow the main thread to try to finish its system call. */ ATF_REQUIRE(ptrace(PT_RESUME, main_lwp, 0, 0) == 0); /* * At this point, the main thread is in the middle of a system call and * has been resumed. The second thread has taken a SIGUSR2 which will * be replaced with a SIGKILL below. The main thread will get to run * first. It should notice the kill request (even though the signal * replacement occurred in the other thread) and exit accordingly. It * should not stop for the system call exit event. */ /* Replace the SIGUSR2 with a kill. */ ATF_REQUIRE(ptrace(PT_KILL, fpid, 0, 0) == 0); /* The last wait() should report the SIGKILL (not a syscall exit). */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSIGNALED(status)); ATF_REQUIRE(WTERMSIG(status) == SIGKILL); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } static void sigusr1_handler(int sig) { CHILD_REQUIRE(sig == SIGUSR1); _exit(2); } /* * Verify that even if the signal queue is full for a child process, * a PT_KILL will kill the process. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_KILL_with_signal_full_sigqueue); ATF_TC_BODY(ptrace__PT_KILL_with_signal_full_sigqueue, tc) { pid_t fpid, wpid; int status; int max_pending_per_proc; size_t len; int i; ATF_REQUIRE(signal(SIGUSR1, sigusr1_handler) != SIG_ERR); ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); len = sizeof(max_pending_per_proc); ATF_REQUIRE(sysctlbyname("kern.sigqueue.max_pending_per_proc", &max_pending_per_proc, &len, NULL, 0) == 0); /* Fill the signal queue. */ for (i = 0; i < max_pending_per_proc; ++i) ATF_REQUIRE(kill(fpid, SIGUSR1) == 0); /* Kill the child process. */ ATF_REQUIRE(ptrace(PT_KILL, fpid, 0, 0) == 0); /* The last wait() should report the SIGKILL. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSIGNALED(status)); ATF_REQUIRE(WTERMSIG(status) == SIGKILL); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that when stopped at a system call entry, a signal can be * requested with PT_CONTINUE which will be delivered once the system * call is complete. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_CONTINUE_with_signal_system_call_entry); ATF_TC_BODY(ptrace__PT_CONTINUE_with_signal_system_call_entry, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; int status; ATF_REQUIRE(signal(SIGUSR1, sigusr1_handler) != SIG_ERR); ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); getpid(); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Continue the child ignoring the SIGSTOP and tracing system calls. */ ATF_REQUIRE(ptrace(PT_SYSCALL, fpid, (caddr_t)1, 0) == 0); /* The second wait() should report a system call entry for getpid(). */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SCE); /* Continue the child process with a signal. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGUSR1) == 0); for (;;) { /* * The last wait() should report exit 2, i.e., a normal _exit * from the signal handler. In the meantime, catch and proceed * past any syscall stops. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); if (WIFSTOPPED(status) && WSTOPSIG(status) == SIGTRAP) { ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & (PL_FLAG_SCE | PL_FLAG_SCX)); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); } else { ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 2); break; } } wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } static void sigusr1_counting_handler(int sig) { static int counter = 0; CHILD_REQUIRE(sig == SIGUSR1); counter++; if (counter == 2) _exit(2); } /* * Verify that, when continuing from a stop at system call entry and exit, * a signal can be requested from both stops, and both will be delivered when * the system call is complete. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_CONTINUE_with_signal_system_call_entry_and_exit); ATF_TC_BODY(ptrace__PT_CONTINUE_with_signal_system_call_entry_and_exit, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; int status; ATF_REQUIRE(signal(SIGUSR1, sigusr1_counting_handler) != SIG_ERR); ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); getpid(); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Continue the child ignoring the SIGSTOP and tracing system calls. */ ATF_REQUIRE(ptrace(PT_SYSCALL, fpid, (caddr_t)1, 0) == 0); /* The second wait() should report a system call entry for getpid(). */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SCE); /* Continue the child process with a signal. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGUSR1) == 0); /* The third wait() should report a system call exit for getpid(). */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SCX); /* Continue the child process with a signal. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGUSR1) == 0); for (;;) { /* * The last wait() should report exit 2, i.e., a normal _exit * from the signal handler. In the meantime, catch and proceed * past any syscall stops. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); if (WIFSTOPPED(status) && WSTOPSIG(status) == SIGTRAP) { ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & (PL_FLAG_SCE | PL_FLAG_SCX)); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); } else { ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 2); break; } } wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that even if the signal queue is full for a child process, * a PT_CONTINUE with a signal will not result in loss of that signal. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_CONTINUE_with_signal_full_sigqueue); ATF_TC_BODY(ptrace__PT_CONTINUE_with_signal_full_sigqueue, tc) { pid_t fpid, wpid; int status; int max_pending_per_proc; size_t len; int i; ATF_REQUIRE(signal(SIGUSR2, handler) != SIG_ERR); ATF_REQUIRE(signal(SIGUSR1, sigusr1_handler) != SIG_ERR); ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); len = sizeof(max_pending_per_proc); ATF_REQUIRE(sysctlbyname("kern.sigqueue.max_pending_per_proc", &max_pending_per_proc, &len, NULL, 0) == 0); /* Fill the signal queue. */ for (i = 0; i < max_pending_per_proc; ++i) ATF_REQUIRE(kill(fpid, SIGUSR2) == 0); /* Continue with signal. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGUSR1) == 0); for (;;) { wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); if (WIFSTOPPED(status)) { ATF_REQUIRE(WSTOPSIG(status) == SIGUSR2); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); } else { /* * The last wait() should report normal _exit from the * SIGUSR1 handler. */ ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 2); break; } } wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } static sem_t sigusr1_sem; static int got_usr1; static void sigusr1_sempost_handler(int sig __unused) { got_usr1++; CHILD_REQUIRE(sem_post(&sigusr1_sem) == 0); } /* * Verify that even if the signal queue is full for a child process, * and the signal is masked, a PT_CONTINUE with a signal will not * result in loss of that signal. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_CONTINUE_with_signal_masked_full_sigqueue); ATF_TC_BODY(ptrace__PT_CONTINUE_with_signal_masked_full_sigqueue, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; int status, err; int max_pending_per_proc; size_t len; int i; sigset_t sigmask; ATF_REQUIRE(signal(SIGUSR2, handler) != SIG_ERR); ATF_REQUIRE(sem_init(&sigusr1_sem, 0, 0) == 0); ATF_REQUIRE(signal(SIGUSR1, sigusr1_sempost_handler) != SIG_ERR); got_usr1 = 0; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { CHILD_REQUIRE(sigemptyset(&sigmask) == 0); CHILD_REQUIRE(sigaddset(&sigmask, SIGUSR1) == 0); CHILD_REQUIRE(sigprocmask(SIG_BLOCK, &sigmask, NULL) == 0); trace_me(); CHILD_REQUIRE(got_usr1 == 0); /* Allow the pending SIGUSR1 in now. */ CHILD_REQUIRE(sigprocmask(SIG_UNBLOCK, &sigmask, NULL) == 0); /* Wait to receive the SIGUSR1. */ do { err = sem_wait(&sigusr1_sem); CHILD_REQUIRE(err == 0 || errno == EINTR); } while (err != 0 && errno == EINTR); CHILD_REQUIRE(got_usr1 == 1); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); len = sizeof(max_pending_per_proc); ATF_REQUIRE(sysctlbyname("kern.sigqueue.max_pending_per_proc", &max_pending_per_proc, &len, NULL, 0) == 0); /* Fill the signal queue. */ for (i = 0; i < max_pending_per_proc; ++i) ATF_REQUIRE(kill(fpid, SIGUSR2) == 0); /* Continue with signal. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGUSR1) == 0); /* Collect and ignore all of the SIGUSR2. */ for (i = 0; i < max_pending_per_proc; ++i) { wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGUSR2); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); } /* Now our PT_CONTINUE'd SIGUSR1 should cause a stop after unmask. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGUSR1); ATF_REQUIRE(ptrace(PT_LWPINFO, fpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_siginfo.si_signo == SIGUSR1); /* Continue the child, ignoring the SIGUSR1. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The last wait() should report exit after receiving SIGUSR1. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that, after stopping due to a signal, that signal can be * replaced with another signal. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_CONTINUE_change_sig); ATF_TC_BODY(ptrace__PT_CONTINUE_change_sig, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); sleep(20); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* Send a signal without ptrace. */ ATF_REQUIRE(kill(fpid, SIGINT) == 0); /* The second wait() should report a SIGINT was received. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGINT); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SI); ATF_REQUIRE(pl.pl_siginfo.si_signo == SIGINT); /* Continue the child process with a different signal. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGTERM) == 0); /* * The last wait() should report having died due to the new * signal, SIGTERM. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSIGNALED(status)); ATF_REQUIRE(WTERMSIG(status) == SIGTERM); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that a signal can be passed through to the child even when there * was no true signal originally. Such cases arise when a SIGTRAP is * invented for e.g, system call stops. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_CONTINUE_with_sigtrap_system_call_entry); ATF_TC_BODY(ptrace__PT_CONTINUE_with_sigtrap_system_call_entry, tc) { struct ptrace_lwpinfo pl; struct rlimit rl; pid_t fpid, wpid; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); /* SIGTRAP expected to cause exit on syscall entry. */ rl.rlim_cur = rl.rlim_max = 0; ATF_REQUIRE(setrlimit(RLIMIT_CORE, &rl) == 0); getpid(); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Continue the child ignoring the SIGSTOP and tracing system calls. */ ATF_REQUIRE(ptrace(PT_SYSCALL, fpid, (caddr_t)1, 0) == 0); /* The second wait() should report a system call entry for getpid(). */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SCE); /* Continue the child process with a SIGTRAP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGTRAP) == 0); for (;;) { /* * The last wait() should report exit due to SIGTRAP. In the * meantime, catch and proceed past any syscall stops. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); if (WIFSTOPPED(status) && WSTOPSIG(status) == SIGTRAP) { ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & (PL_FLAG_SCE | PL_FLAG_SCX)); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); } else { ATF_REQUIRE(WIFSIGNALED(status)); ATF_REQUIRE(WTERMSIG(status) == SIGTRAP); break; } } wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * A mixed bag PT_CONTINUE with signal test. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_CONTINUE_with_signal_mix); ATF_TC_BODY(ptrace__PT_CONTINUE_with_signal_mix, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; int status; ATF_REQUIRE(signal(SIGUSR1, sigusr1_counting_handler) != SIG_ERR); ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); getpid(); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Continue the child ignoring the SIGSTOP and tracing system calls. */ ATF_REQUIRE(ptrace(PT_SYSCALL, fpid, (caddr_t)1, 0) == 0); /* The second wait() should report a system call entry for getpid(). */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SCE); /* Continue with the first SIGUSR1. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGUSR1) == 0); /* The next wait() should report a system call exit for getpid(). */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SCX); /* Send an ABRT without ptrace. */ ATF_REQUIRE(kill(fpid, SIGABRT) == 0); /* Continue normally. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The next wait() should report the SIGABRT. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGABRT); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SI); ATF_REQUIRE(pl.pl_siginfo.si_signo == SIGABRT); /* Continue, replacing the SIGABRT with another SIGUSR1. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGUSR1) == 0); for (;;) { /* * The last wait() should report exit 2, i.e., a normal _exit * from the signal handler. In the meantime, catch and proceed * past any syscall stops. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); if (WIFSTOPPED(status) && WSTOPSIG(status) == SIGTRAP) { ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & (PL_FLAG_SCE | PL_FLAG_SCX)); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); } else { ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 2); break; } } wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify a signal delivered by ptrace is noticed by kevent(2). */ ATF_TC_WITHOUT_HEAD(ptrace__PT_CONTINUE_with_signal_kqueue); ATF_TC_BODY(ptrace__PT_CONTINUE_with_signal_kqueue, tc) { pid_t fpid, wpid; int status, kq, nevents; struct kevent kev; ATF_REQUIRE(signal(SIGUSR1, SIG_IGN) != SIG_ERR); ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { CHILD_REQUIRE((kq = kqueue()) > 0); EV_SET(&kev, SIGUSR1, EVFILT_SIGNAL, EV_ADD, 0, 0, 0); CHILD_REQUIRE(kevent(kq, &kev, 1, NULL, 0, NULL) == 0); trace_me(); for (;;) { nevents = kevent(kq, NULL, 0, &kev, 1, NULL); if (nevents == -1 && errno == EINTR) continue; CHILD_REQUIRE(nevents > 0); CHILD_REQUIRE(kev.filter == EVFILT_SIGNAL); CHILD_REQUIRE(kev.ident == SIGUSR1); break; } exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Continue with the SIGUSR1. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGUSR1) == 0); /* * The last wait() should report normal exit with code 1. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } static void * signal_thread(void *arg) { int err; sigset_t sigmask; pthread_barrier_t *pbarrier = (pthread_barrier_t*)arg; /* Wait for this thread to receive a SIGUSR1. */ do { err = sem_wait(&sigusr1_sem); CHILD_REQUIRE(err == 0 || errno == EINTR); } while (err != 0 && errno == EINTR); /* Free our companion thread from the barrier. */ pthread_barrier_wait(pbarrier); /* * Swap ignore duties; the next SIGUSR1 should go to the * other thread. */ CHILD_REQUIRE(sigemptyset(&sigmask) == 0); CHILD_REQUIRE(sigaddset(&sigmask, SIGUSR1) == 0); CHILD_REQUIRE(pthread_sigmask(SIG_BLOCK, &sigmask, NULL) == 0); /* Sync up threads after swapping signal masks. */ pthread_barrier_wait(pbarrier); /* Wait until our companion has received its SIGUSR1. */ pthread_barrier_wait(pbarrier); return (NULL); } /* * Verify that a traced process with blocked signal received the * signal from kill() once unmasked. */ ATF_TC_WITHOUT_HEAD(ptrace__killed_with_sigmask); ATF_TC_BODY(ptrace__killed_with_sigmask, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; int status, err; sigset_t sigmask; ATF_REQUIRE(sem_init(&sigusr1_sem, 0, 0) == 0); ATF_REQUIRE(signal(SIGUSR1, sigusr1_sempost_handler) != SIG_ERR); got_usr1 = 0; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { CHILD_REQUIRE(sigemptyset(&sigmask) == 0); CHILD_REQUIRE(sigaddset(&sigmask, SIGUSR1) == 0); CHILD_REQUIRE(sigprocmask(SIG_BLOCK, &sigmask, NULL) == 0); trace_me(); CHILD_REQUIRE(got_usr1 == 0); /* Allow the pending SIGUSR1 in now. */ CHILD_REQUIRE(sigprocmask(SIG_UNBLOCK, &sigmask, NULL) == 0); /* Wait to receive a SIGUSR1. */ do { err = sem_wait(&sigusr1_sem); CHILD_REQUIRE(err == 0 || errno == EINTR); } while (err != 0 && errno == EINTR); CHILD_REQUIRE(got_usr1 == 1); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_LWPINFO, fpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_siginfo.si_signo == SIGSTOP); /* Send blocked SIGUSR1 which should cause a stop. */ ATF_REQUIRE(kill(fpid, SIGUSR1) == 0); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The next wait() should report the kill(SIGUSR1) was received. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGUSR1); ATF_REQUIRE(ptrace(PT_LWPINFO, fpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_siginfo.si_signo == SIGUSR1); /* Continue the child, allowing in the SIGUSR1. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGUSR1) == 0); /* The last wait() should report normal exit with code 1. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that a traced process with blocked signal received the * signal from PT_CONTINUE once unmasked. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_CONTINUE_with_sigmask); ATF_TC_BODY(ptrace__PT_CONTINUE_with_sigmask, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; int status, err; sigset_t sigmask; ATF_REQUIRE(sem_init(&sigusr1_sem, 0, 0) == 0); ATF_REQUIRE(signal(SIGUSR1, sigusr1_sempost_handler) != SIG_ERR); got_usr1 = 0; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { CHILD_REQUIRE(sigemptyset(&sigmask) == 0); CHILD_REQUIRE(sigaddset(&sigmask, SIGUSR1) == 0); CHILD_REQUIRE(sigprocmask(SIG_BLOCK, &sigmask, NULL) == 0); trace_me(); CHILD_REQUIRE(got_usr1 == 0); /* Allow the pending SIGUSR1 in now. */ CHILD_REQUIRE(sigprocmask(SIG_UNBLOCK, &sigmask, NULL) == 0); /* Wait to receive a SIGUSR1. */ do { err = sem_wait(&sigusr1_sem); CHILD_REQUIRE(err == 0 || errno == EINTR); } while (err != 0 && errno == EINTR); CHILD_REQUIRE(got_usr1 == 1); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_LWPINFO, fpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_siginfo.si_signo == SIGSTOP); /* Continue the child replacing SIGSTOP with SIGUSR1. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGUSR1) == 0); /* The next wait() should report the SIGUSR1 was received. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGUSR1); ATF_REQUIRE(ptrace(PT_LWPINFO, fpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_siginfo.si_signo == SIGUSR1); /* Continue the child, ignoring the SIGUSR1. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The last wait() should report normal exit with code 1. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that if ptrace stops due to a signal but continues with * a different signal that the new signal is routed to a thread - * that can accept it, and that that thread is awakened by the signal + * that can accept it, and that the thread is awakened by the signal * in a timely manner. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_CONTINUE_with_signal_thread_sigmask); ATF_TC_BODY(ptrace__PT_CONTINUE_with_signal_thread_sigmask, tc) { pid_t fpid, wpid; int status, err; pthread_t t; sigset_t sigmask; pthread_barrier_t barrier; ATF_REQUIRE(pthread_barrier_init(&barrier, NULL, 2) == 0); ATF_REQUIRE(sem_init(&sigusr1_sem, 0, 0) == 0); ATF_REQUIRE(signal(SIGUSR1, sigusr1_sempost_handler) != SIG_ERR); ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { CHILD_REQUIRE(pthread_create(&t, NULL, signal_thread, (void*)&barrier) == 0); /* The other thread should receive the first SIGUSR1. */ CHILD_REQUIRE(sigemptyset(&sigmask) == 0); CHILD_REQUIRE(sigaddset(&sigmask, SIGUSR1) == 0); CHILD_REQUIRE(pthread_sigmask(SIG_BLOCK, &sigmask, NULL) == 0); trace_me(); /* Wait until other thread has received its SIGUSR1. */ pthread_barrier_wait(&barrier); /* * Swap ignore duties; the next SIGUSR1 should go to this * thread. */ CHILD_REQUIRE(pthread_sigmask(SIG_UNBLOCK, &sigmask, NULL) == 0); /* Sync up threads after swapping signal masks. */ pthread_barrier_wait(&barrier); /* * Sync up with test code; we're ready for the next SIGUSR1 * now. */ raise(SIGSTOP); /* Wait for this thread to receive a SIGUSR1. */ do { err = sem_wait(&sigusr1_sem); CHILD_REQUIRE(err == 0 || errno == EINTR); } while (err != 0 && errno == EINTR); /* Free the other thread from the barrier. */ pthread_barrier_wait(&barrier); CHILD_REQUIRE(pthread_join(t, NULL) == 0); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* * Send a signal without ptrace that either thread will accept (USR2, * in this case). */ ATF_REQUIRE(kill(fpid, SIGUSR2) == 0); /* The second wait() should report a SIGUSR2 was received. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGUSR2); /* Continue the child, changing the signal to USR1. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGUSR1) == 0); /* The next wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); ATF_REQUIRE(kill(fpid, SIGUSR2) == 0); /* The next wait() should report a SIGUSR2 was received. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGUSR2); /* Continue the child, changing the signal to USR1. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, SIGUSR1) == 0); /* The last wait() should report normal exit with code 1. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } static void * raise_sigstop_thread(void *arg __unused) { raise(SIGSTOP); return NULL; } static void * sleep_thread(void *arg __unused) { sleep(60); return NULL; } static void terminate_with_pending_sigstop(bool sigstop_from_main_thread) { pid_t fpid, wpid; int status, i; cpuset_t setmask; cpusetid_t setid; pthread_t t; /* * Become the reaper for this process tree. We need to be able to check * that both child and grandchild have died. */ ATF_REQUIRE(procctl(P_PID, getpid(), PROC_REAP_ACQUIRE, NULL) == 0); fpid = fork(); ATF_REQUIRE(fpid >= 0); if (fpid == 0) { fpid = fork(); CHILD_REQUIRE(fpid >= 0); if (fpid == 0) { trace_me(); /* Pin to CPU 0 to serialize thread execution. */ CPU_ZERO(&setmask); CPU_SET(0, &setmask); CHILD_REQUIRE(cpuset(&setid) == 0); CHILD_REQUIRE(cpuset_setaffinity(CPU_LEVEL_CPUSET, CPU_WHICH_CPUSET, setid, sizeof(setmask), &setmask) == 0); if (sigstop_from_main_thread) { /* * We expect the SIGKILL sent when our parent * dies to be delivered to the new thread. * Raise the SIGSTOP in this thread so the * threads compete. */ CHILD_REQUIRE(pthread_create(&t, NULL, sleep_thread, NULL) == 0); raise(SIGSTOP); } else { /* * We expect the SIGKILL to be delivered to * this thread. After creating the new thread, * just get off the CPU so the other thread can * raise the SIGSTOP. */ CHILD_REQUIRE(pthread_create(&t, NULL, raise_sigstop_thread, NULL) == 0); sleep(60); } exit(0); } /* First stop is trace_me() immediately after fork. */ wpid = waitpid(fpid, &status, 0); CHILD_REQUIRE(wpid == fpid); CHILD_REQUIRE(WIFSTOPPED(status)); CHILD_REQUIRE(WSTOPSIG(status) == SIGSTOP); CHILD_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* Second stop is from the raise(SIGSTOP). */ wpid = waitpid(fpid, &status, 0); CHILD_REQUIRE(wpid == fpid); CHILD_REQUIRE(WIFSTOPPED(status)); CHILD_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* * Terminate tracing process without detaching. Our child * should be killed. */ exit(0); } /* * We should get a normal exit from our immediate child and a SIGKILL * exit from our grandchild. The latter case is the interesting one. * Our grandchild should not have stopped due to the SIGSTOP that was * left dangling when its parent died. */ for (i = 0; i < 2; ++i) { wpid = wait(&status); if (wpid == fpid) ATF_REQUIRE(WIFEXITED(status) && WEXITSTATUS(status) == 0); else ATF_REQUIRE(WIFSIGNALED(status) && WTERMSIG(status) == SIGKILL); } } /* * These two tests ensure that if the tracing process exits without detaching * just after the child received a SIGSTOP, the child is cleanly killed and * doesn't go to sleep due to the SIGSTOP. The parent's death will send a * SIGKILL to the child. If the SIGKILL and the SIGSTOP are handled by * different threads, the SIGKILL must win. There are two variants of this * test, designed to catch the case where the SIGKILL is delivered to the * younger thread (the first test) and the case where the SIGKILL is delivered * to the older thread (the second test). This behavior has changed in the * past, so make no assumption. */ ATF_TC(ptrace__parent_terminate_with_pending_sigstop1); ATF_TC_HEAD(ptrace__parent_terminate_with_pending_sigstop1, tc) { atf_tc_set_md_var(tc, "require.user", "root"); } ATF_TC_BODY(ptrace__parent_terminate_with_pending_sigstop1, tc) { terminate_with_pending_sigstop(true); } ATF_TC(ptrace__parent_terminate_with_pending_sigstop2); ATF_TC_HEAD(ptrace__parent_terminate_with_pending_sigstop2, tc) { atf_tc_set_md_var(tc, "require.user", "root"); } ATF_TC_BODY(ptrace__parent_terminate_with_pending_sigstop2, tc) { terminate_with_pending_sigstop(false); } /* * Verify that after ptrace() discards a SIGKILL signal, the event mask * is not modified. */ ATF_TC_WITHOUT_HEAD(ptrace__event_mask_sigkill_discard); ATF_TC_BODY(ptrace__event_mask_sigkill_discard, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; int status, event_mask, new_event_mask; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); raise(SIGSTOP); exit(0); } /* The first wait() should report the stop from trace_me(). */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Set several unobtrusive event bits. */ event_mask = PTRACE_EXEC | PTRACE_FORK | PTRACE_LWP; ATF_REQUIRE(ptrace(PT_SET_EVENT_MASK, wpid, (caddr_t)&event_mask, sizeof(event_mask)) == 0); /* Send a SIGKILL without using ptrace. */ ATF_REQUIRE(kill(fpid, SIGKILL) == 0); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The next stop should be due to the SIGKILL. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGKILL); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SI); ATF_REQUIRE(pl.pl_siginfo.si_signo == SIGKILL); /* Continue the child ignoring the SIGKILL. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The next wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Check the current event mask. It should not have changed. */ new_event_mask = 0; ATF_REQUIRE(ptrace(PT_GET_EVENT_MASK, wpid, (caddr_t)&new_event_mask, sizeof(new_event_mask)) == 0); ATF_REQUIRE(event_mask == new_event_mask); /* Continue the child to let it exit. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The last event should be for the child process's exit. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 0); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } static void * flock_thread(void *arg) { int fd; fd = *(int *)arg; (void)flock(fd, LOCK_EX); (void)flock(fd, LOCK_UN); return (NULL); } /* * Verify that PT_ATTACH will suspend threads sleeping in an SBDRY section. * We rely on the fact that the lockf implementation sets SBDRY before blocking * on a lock. This is a regression test for r318191. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_ATTACH_with_SBDRY_thread); ATF_TC_BODY(ptrace__PT_ATTACH_with_SBDRY_thread, tc) { pthread_barrier_t barrier; pthread_barrierattr_t battr; char tmpfile[64]; pid_t child, wpid; int error, fd, i, status; ATF_REQUIRE(pthread_barrierattr_init(&battr) == 0); ATF_REQUIRE(pthread_barrierattr_setpshared(&battr, PTHREAD_PROCESS_SHARED) == 0); ATF_REQUIRE(pthread_barrier_init(&barrier, &battr, 2) == 0); (void)snprintf(tmpfile, sizeof(tmpfile), "./ptrace.XXXXXX"); fd = mkstemp(tmpfile); ATF_REQUIRE(fd >= 0); ATF_REQUIRE((child = fork()) != -1); if (child == 0) { pthread_t t[2]; int cfd; error = pthread_barrier_wait(&barrier); if (error != 0 && error != PTHREAD_BARRIER_SERIAL_THREAD) _exit(1); cfd = open(tmpfile, O_RDONLY); if (cfd < 0) _exit(1); /* * We want at least two threads blocked on the file lock since * the SIGSTOP from PT_ATTACH may kick one of them out of * sleep. */ if (pthread_create(&t[0], NULL, flock_thread, &cfd) != 0) _exit(1); if (pthread_create(&t[1], NULL, flock_thread, &cfd) != 0) _exit(1); if (pthread_join(t[0], NULL) != 0) _exit(1); if (pthread_join(t[1], NULL) != 0) _exit(1); _exit(0); } ATF_REQUIRE(flock(fd, LOCK_EX) == 0); error = pthread_barrier_wait(&barrier); ATF_REQUIRE(error == 0 || error == PTHREAD_BARRIER_SERIAL_THREAD); /* * Give the child some time to block. Is there a better way to do this? */ sleep(1); /* * Attach and give the child 3 seconds to stop. */ ATF_REQUIRE(ptrace(PT_ATTACH, child, NULL, 0) == 0); for (i = 0; i < 3; i++) { wpid = waitpid(child, &status, WNOHANG); if (wpid == child && WIFSTOPPED(status) && WSTOPSIG(status) == SIGSTOP) break; sleep(1); } ATF_REQUIRE_MSG(i < 3, "failed to stop child process after PT_ATTACH"); ATF_REQUIRE(ptrace(PT_DETACH, child, NULL, 0) == 0); ATF_REQUIRE(flock(fd, LOCK_UN) == 0); ATF_REQUIRE(unlink(tmpfile) == 0); ATF_REQUIRE(close(fd) == 0); } static void sigusr1_step_handler(int sig) { CHILD_REQUIRE(sig == SIGUSR1); raise(SIGABRT); } /* * Verify that PT_STEP with a signal invokes the signal before * stepping the next instruction (and that the next instruction is * stepped correctly). */ ATF_TC_WITHOUT_HEAD(ptrace__PT_STEP_with_signal); ATF_TC_BODY(ptrace__PT_STEP_with_signal, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; int status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); signal(SIGUSR1, sigusr1_step_handler); raise(SIGABRT); exit(1); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The next stop should report the SIGABRT in the child body. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGABRT); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SI); ATF_REQUIRE(pl.pl_siginfo.si_signo == SIGABRT); /* Step the child process inserting SIGUSR1. */ ATF_REQUIRE(ptrace(PT_STEP, fpid, (caddr_t)1, SIGUSR1) == 0); /* The next stop should report the SIGABRT in the signal handler. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGABRT); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SI); ATF_REQUIRE(pl.pl_siginfo.si_signo == SIGABRT); /* Continue the child process discarding the signal. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The next stop should report a trace trap from PT_STEP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(pl.pl_flags & PL_FLAG_SI); ATF_REQUIRE(pl.pl_siginfo.si_signo == SIGTRAP); ATF_REQUIRE(pl.pl_siginfo.si_code == TRAP_TRACE); /* Continue the child to let it exit. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* The last event should be for the child process's exit. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } #if defined(__amd64__) || defined(__i386__) /* * Only x86 both define breakpoint() and have a PC after breakpoint so * that restarting doesn't retrigger the breakpoint. */ static void * continue_thread(void *arg) { breakpoint(); return (NULL); } static __dead2 void continue_thread_main(void) { pthread_t threads[2]; CHILD_REQUIRE(pthread_create(&threads[0], NULL, continue_thread, NULL) == 0); CHILD_REQUIRE(pthread_create(&threads[1], NULL, continue_thread, NULL) == 0); CHILD_REQUIRE(pthread_join(threads[0], NULL) == 0); CHILD_REQUIRE(pthread_join(threads[1], NULL) == 0); exit(1); } /* * Ensure that PT_CONTINUE clears the status of the thread that * triggered the stop even if a different thread's LWP was passed to * PT_CONTINUE. */ ATF_TC_WITHOUT_HEAD(ptrace__PT_CONTINUE_different_thread); ATF_TC_BODY(ptrace__PT_CONTINUE_different_thread, tc) { struct ptrace_lwpinfo pl; pid_t fpid, wpid; lwpid_t lwps[2]; bool hit_break[2]; int i, j, status; ATF_REQUIRE((fpid = fork()) != -1); if (fpid == 0) { trace_me(); continue_thread_main(); } /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE(ptrace(PT_LWP_EVENTS, wpid, NULL, 1) == 0); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* One of the new threads should report it's birth. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE((pl.pl_flags & (PL_FLAG_BORN | PL_FLAG_SCX)) == (PL_FLAG_BORN | PL_FLAG_SCX)); lwps[0] = pl.pl_lwpid; /* * Suspend this thread to ensure both threads are alive before * hitting the breakpoint. */ ATF_REQUIRE(ptrace(PT_SUSPEND, lwps[0], NULL, 0) != -1); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* Second thread should report it's birth. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE((pl.pl_flags & (PL_FLAG_BORN | PL_FLAG_SCX)) == (PL_FLAG_BORN | PL_FLAG_SCX)); ATF_REQUIRE(pl.pl_lwpid != lwps[0]); lwps[1] = pl.pl_lwpid; /* Resume both threads waiting for breakpoint events. */ hit_break[0] = hit_break[1] = false; ATF_REQUIRE(ptrace(PT_RESUME, lwps[0], NULL, 0) != -1); ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); /* One thread should report a breakpoint. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); ATF_REQUIRE((pl.pl_flags & PL_FLAG_SI) != 0); ATF_REQUIRE(pl.pl_siginfo.si_signo == SIGTRAP && pl.pl_siginfo.si_code == TRAP_BRKPT); if (pl.pl_lwpid == lwps[0]) i = 0; else i = 1; hit_break[i] = true; /* * Resume both threads but pass the other thread's LWPID to * PT_CONTINUE. */ ATF_REQUIRE(ptrace(PT_CONTINUE, lwps[i ^ 1], (caddr_t)1, 0) == 0); /* * Will now get two thread exit events and one more breakpoint * event. */ for (j = 0; j < 3; j++) { wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(wpid == fpid); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGTRAP); ATF_REQUIRE(ptrace(PT_LWPINFO, wpid, (caddr_t)&pl, sizeof(pl)) != -1); if (pl.pl_lwpid == lwps[0]) i = 0; else i = 1; ATF_REQUIRE_MSG(lwps[i] != 0, "event for exited thread"); if (pl.pl_flags & PL_FLAG_EXITED) { ATF_REQUIRE_MSG(hit_break[i], "exited thread did not report breakpoint"); lwps[i] = 0; } else { ATF_REQUIRE((pl.pl_flags & PL_FLAG_SI) != 0); ATF_REQUIRE(pl.pl_siginfo.si_signo == SIGTRAP && pl.pl_siginfo.si_code == TRAP_BRKPT); ATF_REQUIRE_MSG(!hit_break[i], "double breakpoint event"); hit_break[i] = true; } ATF_REQUIRE(ptrace(PT_CONTINUE, fpid, (caddr_t)1, 0) == 0); } /* Both threads should have exited. */ ATF_REQUIRE(lwps[0] == 0); ATF_REQUIRE(lwps[1] == 0); /* The last event should be for the child process's exit. */ wpid = waitpid(fpid, &status, 0); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); wpid = wait(&status); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } #endif ATF_TP_ADD_TCS(tp) { ATF_TP_ADD_TC(tp, ptrace__parent_wait_after_trace_me); ATF_TP_ADD_TC(tp, ptrace__parent_wait_after_attach); ATF_TP_ADD_TC(tp, ptrace__parent_sees_exit_after_child_debugger); ATF_TP_ADD_TC(tp, ptrace__parent_sees_exit_after_unrelated_debugger); ATF_TP_ADD_TC(tp, ptrace__follow_fork_both_attached); ATF_TP_ADD_TC(tp, ptrace__follow_fork_child_detached); ATF_TP_ADD_TC(tp, ptrace__follow_fork_parent_detached); ATF_TP_ADD_TC(tp, ptrace__follow_fork_both_attached_unrelated_debugger); ATF_TP_ADD_TC(tp, ptrace__follow_fork_child_detached_unrelated_debugger); ATF_TP_ADD_TC(tp, ptrace__follow_fork_parent_detached_unrelated_debugger); ATF_TP_ADD_TC(tp, ptrace__getppid); ATF_TP_ADD_TC(tp, ptrace__new_child_pl_syscall_code_fork); ATF_TP_ADD_TC(tp, ptrace__new_child_pl_syscall_code_vfork); ATF_TP_ADD_TC(tp, ptrace__new_child_pl_syscall_code_thread); ATF_TP_ADD_TC(tp, ptrace__lwp_events); ATF_TP_ADD_TC(tp, ptrace__lwp_events_exec); ATF_TP_ADD_TC(tp, ptrace__siginfo); ATF_TP_ADD_TC(tp, ptrace__ptrace_exec_disable); ATF_TP_ADD_TC(tp, ptrace__ptrace_exec_enable); ATF_TP_ADD_TC(tp, ptrace__event_mask); ATF_TP_ADD_TC(tp, ptrace__ptrace_vfork); ATF_TP_ADD_TC(tp, ptrace__ptrace_vfork_follow); #if defined(__amd64__) || defined(__i386__) || defined(__sparc64__) ATF_TP_ADD_TC(tp, ptrace__PT_KILL_breakpoint); #endif ATF_TP_ADD_TC(tp, ptrace__PT_KILL_system_call); ATF_TP_ADD_TC(tp, ptrace__PT_KILL_threads); ATF_TP_ADD_TC(tp, ptrace__PT_KILL_competing_signal); ATF_TP_ADD_TC(tp, ptrace__PT_KILL_competing_stop); ATF_TP_ADD_TC(tp, ptrace__PT_KILL_with_signal_full_sigqueue); ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_with_signal_system_call_entry); ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_with_signal_system_call_entry_and_exit); ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_with_signal_full_sigqueue); ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_with_signal_masked_full_sigqueue); ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_change_sig); ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_with_sigtrap_system_call_entry); ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_with_signal_mix); ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_with_signal_kqueue); ATF_TP_ADD_TC(tp, ptrace__killed_with_sigmask); ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_with_sigmask); ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_with_signal_thread_sigmask); ATF_TP_ADD_TC(tp, ptrace__parent_terminate_with_pending_sigstop1); ATF_TP_ADD_TC(tp, ptrace__parent_terminate_with_pending_sigstop2); ATF_TP_ADD_TC(tp, ptrace__event_mask_sigkill_discard); ATF_TP_ADD_TC(tp, ptrace__PT_ATTACH_with_SBDRY_thread); ATF_TP_ADD_TC(tp, ptrace__PT_STEP_with_signal); #if defined(__amd64__) || defined(__i386__) ATF_TP_ADD_TC(tp, ptrace__PT_CONTINUE_different_thread); #endif return (atf_no_error()); } Index: head/tools/regression/bpf/bpf_filter/bpf_test.c =================================================================== --- head/tools/regression/bpf/bpf_filter/bpf_test.c (revision 327232) +++ head/tools/regression/bpf/bpf_filter/bpf_test.c (revision 327233) @@ -1,281 +1,281 @@ /*- * Copyright (C) 2008-2009 Jung-uk Kim . 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. * * THIS SOFTWARE IS PROVIDED BY 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 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. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include BPF_TEST_H #define PASSED 0 #define FAILED 1 #define FATAL -1 #ifndef LOG_LEVEL #define LOG_LEVEL 1 #endif #ifdef BPF_BENCHMARK #define BPF_NRUNS 10000000 #else #define BPF_NRUNS 1 #endif static void sig_handler(int); static int nins = sizeof(pc) / sizeof(pc[0]); static int verbose = LOG_LEVEL; #ifdef BPF_JIT_COMPILER #include #include static u_int bpf_compile_and_filter(void) { bpf_jit_filter *filter; u_int i, ret; /* Compile the BPF filter program and generate native code. */ if ((filter = bpf_jitter(pc, nins)) == NULL) { if (verbose > 1) printf("Failed to allocate memory:\t"); if (verbose > 0) printf("FATAL\n"); exit(FATAL); } if (verbose > 2) { printf("\n"); hexdump(filter->func, filter->size, NULL, HD_OMIT_CHARS); } for (i = 0; i < BPF_NRUNS; i++) ret = (*(filter->func))(pkt, wirelen, buflen); bpf_destroy_jit_filter(filter); return (ret); } #else u_int bpf_filter(const struct bpf_insn *, u_char *, u_int, u_int); #endif #ifdef BPF_VALIDATE static const u_short bpf_code_map[] = { 0x10ff, /* 0x00-0x0f: 1111111100001000 */ 0x3070, /* 0x10-0x1f: 0000111000001100 */ 0x3131, /* 0x20-0x2f: 1000110010001100 */ 0x3031, /* 0x30-0x3f: 1000110000001100 */ 0x3131, /* 0x40-0x4f: 1000110010001100 */ 0x1011, /* 0x50-0x5f: 1000100000001000 */ 0x1013, /* 0x60-0x6f: 1100100000001000 */ 0x1010, /* 0x70-0x7f: 0000100000001000 */ 0x0093, /* 0x80-0x8f: 1100100100000000 */ 0x1010, /* 0x90-0x9f: 0000100000001000 */ 0x1010, /* 0xa0-0xaf: 0000100000001000 */ 0x0002, /* 0xb0-0xbf: 0100000000000000 */ 0x0000, /* 0xc0-0xcf: 0000000000000000 */ 0x0000, /* 0xd0-0xdf: 0000000000000000 */ 0x0000, /* 0xe0-0xef: 0000000000000000 */ 0x0000 /* 0xf0-0xff: 0000000000000000 */ }; #define BPF_VALIDATE_CODE(c) \ ((c) <= 0xff && (bpf_code_map[(c) >> 4] & (1 << ((c) & 0xf))) != 0) /* * XXX Copied from sys/net/bpf_filter.c and modified. * * Return true if the 'fcode' is a valid filter program. * The constraints are that each jump be forward and to a valid * code. The code must terminate with either an accept or reject. * * The kernel needs to be able to verify an application's filter code. * Otherwise, a bogus program could easily crash the system. */ static int bpf_validate(const struct bpf_insn *f, int len) { register int i; register const struct bpf_insn *p; /* Do not accept negative length filter. */ if (len < 0) return (0); /* An empty filter means accept all. */ if (len == 0) return (1); for (i = 0; i < len; ++i) { p = &f[i]; /* * Check that the code is valid. */ if (!BPF_VALIDATE_CODE(p->code)) return (0); /* - * Check that that jumps are forward, and within + * Check that jumps are forward, and within * the code block. */ if (BPF_CLASS(p->code) == BPF_JMP) { register u_int offset; if (p->code == (BPF_JMP|BPF_JA)) offset = p->k; else offset = p->jt > p->jf ? p->jt : p->jf; if (offset >= (u_int)(len - i) - 1) return (0); continue; } /* * Check that memory operations use valid addresses. */ if (p->code == BPF_ST || p->code == BPF_STX || p->code == (BPF_LD|BPF_MEM) || p->code == (BPF_LDX|BPF_MEM)) { if (p->k >= BPF_MEMWORDS) return (0); continue; } /* * Check for constant division by 0. */ if ((p->code == (BPF_ALU|BPF_DIV|BPF_K) || p->code == (BPF_ALU|BPF_MOD|BPF_K)) && p->k == 0) return (0); } return (BPF_CLASS(f[len - 1].code) == BPF_RET); } #endif int main(void) { #ifndef BPF_JIT_COMPILER u_int i; #endif u_int ret; int sig; #ifdef BPF_VALIDATE int valid; #endif /* Try to catch all signals */ for (sig = SIGHUP; sig <= SIGUSR2; sig++) signal(sig, sig_handler); #ifdef BPF_VALIDATE valid = bpf_validate(pc, nins); if (valid != 0 && invalid != 0) { if (verbose > 1) printf("Validated invalid instruction(s):\t"); if (verbose > 0) printf("FAILED\n"); return (FAILED); } else if (valid == 0 && invalid == 0) { if (verbose > 1) printf("Invalidated valid instruction(s):\t"); if (verbose > 0) printf("FAILED\n"); return (FAILED); } else if (invalid != 0) { if (verbose > 1) printf("Expected and invalidated:\t"); if (verbose > 0) printf("PASSED\n"); return (PASSED); } #endif #ifdef BPF_JIT_COMPILER ret = bpf_compile_and_filter(); #else for (i = 0; i < BPF_NRUNS; i++) ret = bpf_filter(nins != 0 ? pc : NULL, pkt, wirelen, buflen); #endif if (expect_signal != 0) { if (verbose > 1) printf("Expected signal %d but got none:\t", expect_signal); if (verbose > 0) printf("FAILED\n"); return (FAILED); } if (ret != expect) { if (verbose > 1) printf("Expected 0x%x but got 0x%x:\t", expect, ret); if (verbose > 0) printf("FAILED\n"); return (FAILED); } if (verbose > 1) printf("Expected and got 0x%x:\t", ret); if (verbose > 0) printf("PASSED\n"); return (PASSED); } static void sig_handler(int sig) { if (expect_signal == 0) { if (verbose > 1) printf("Received unexpected signal %d:\t", sig); if (verbose > 0) printf("FATAL\n"); exit(FATAL); } if (expect_signal != sig) { if (verbose > 1) printf("Expected signal %d but got %d:\t", expect_signal, sig); if (verbose > 0) printf("FAILED\n"); exit(FAILED); } if (verbose > 1) printf("Expected and got signal %d:\t", sig); if (verbose > 0) printf("PASSED\n"); exit(PASSED); }