d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
c82c5ba499
freebsd-skq/sys/kern
History
rwatson 9487c057e2 Correct a number of serious and closely related bugs in the UNIX domain 
socket file descriptor garbage collection code, which is intended to
detect and clear cycles of orphaned file descriptors that are "in-flight"
in a socket when that socket is closed before they are received.  The
algorithm present was both run at poor times (resulting in recursion and
reentrance), and also buggy in the presence of parallelism.  In order to
fix these problems, make the following changes:

- When there are in-flight sockets and a UNIX domain socket is destroyed,
  asynchronously schedule the garbage collector, rather than running it
  synchronously in the current context.  This avoids lock order issues
  when the garbage collection code reenters the UNIX domain socket code,
  avoiding lock order reversals, deadlocks, etc.  Run the code
  asynchronously in a task queue.

- In the garbage collector, when skipping file descriptors that have
  entered a closing state (i.e., have f_count == 0), re-test the FDEFER
  flag, and decrement unp_defer.  As file descriptors can now transition
  to a closed state, while the garbage collector is running, it is no
  longer the case that unp_defer will remain an accurate count of
  deferred sockets in the mark portion of the GC algorithm.  Otherwise,
  the garbage collector will loop waiting waiting for unp_defer to reach
  zero, which it will never do as it is skipping file descriptors that
  were marked in an earlier pass, but now closed.

- Acquire the UNIX domain socket subsystem lock in unp_discard() when
  modifying the unp_rights counter, or a read/write race is risked with
  other threads also manipulating the counter.

While here:

- Remove #if 0'd code regarding acquiring the socket buffer sleep lock in
  the garbage collector, this is not required as we are able to use the
  socket buffer receive lock to protect scanning the receive buffer for
  in-flight file descriptors on the socket buffer.

- Annotate that the description of the garbage collector implementation
  is increasingly inaccurate and needs to be updated.

- Add counters of the number of deferred garbage collections and recycled
  file descriptors.  This will be removed and is here temporarily for
  debugging purposes.

With these changes in place, the unp_passfd regression test now appears
to be passed consistently on UP and SMP systems for extended runs,
whereas before it hung quickly or panicked, depending on which bug was
triggered.

Reported by:	Philip Kizer <pckizer at nostrum dot com>
MFC after:	2 weeks
2005-11-10 16:06:04 +00:00
..
bus_if.m
clock_if.m
cpufreq_if.m
device_if.m
genassym.sh
imgact_aout.c
imgact_elf32.c
imgact_elf64.c
imgact_elf.c Back out alpha/alpha/trap.c:1.124, osf1_ioctl.c:1.14, osf1_misc.c:1.57, 2005-09-28 07:03:03 +00:00
imgact_gzip.c
imgact_shell.c Fix a panic which could occur parsing #!-lines in a shell-script. If the 2005-06-19 02:21:03 +00:00
inflate.c Normalize a significant number of kernel malloc type names: 2005-10-31 15:41:29 +00:00
init_main.c Remove mac_create_root_mount() and mpo_create_root_mount(), which 2005-09-19 13:59:57 +00:00
init_sysent.c Regen. 2005-10-30 02:14:37 +00:00
kern_acct.c Pass 'curthread' into VFS_STATFS() from acctwatch(), rather than passing 2005-09-21 15:28:07 +00:00
kern_acl.c Implement new world order in VFS locking for ACLs. This will remove the 2005-09-17 22:01:14 +00:00
kern_alq.c
kern_clock.c In watchdog_config enable the software watchdog iff the WD_ACTIVE flag is 2005-10-27 17:22:47 +00:00
kern_condvar.c
kern_conf.c Move the initialization of the devmtx into the mutex_init() function 2005-10-18 18:27:44 +00:00
kern_context.c
kern_cpu.c make saved cpu level stackable. 2005-10-03 06:57:29 +00:00
kern_descrip.c Add the f_msgcount field to the set of struct file fields printed in show 2005-11-10 13:26:29 +00:00
kern_environment.c Reorder statements to avoid accessing unknown memory. 2005-09-26 14:14:55 +00:00
kern_event.c Add in kqueue support to LIO event notification and fix how it handled 2005-10-12 17:51:31 +00:00
kern_exec.c Cleanup some signal interfaces. Now the tdsignal function accepts 2005-11-03 04:49:16 +00:00
kern_exit.c Giant clean up for exit(2) 2005-11-08 17:11:03 +00:00
kern_fork.c Fix the recent panics/LORs/hangs created by my kqueue commit by: 2005-07-01 16:28:32 +00:00
kern_idle.c
kern_intr.c Add a swi_remove() function to teardown software interrupt handlers. For 2005-10-26 15:51:05 +00:00
kern_jail.c Push Giant down in jails. Pass the MPSAFE flag to NDINIT, and keep track 2005-09-28 00:30:56 +00:00
kern_kse.c Cleanup some signal interfaces. Now the tdsignal function accepts 2005-11-03 04:49:16 +00:00
kern_kthread.c
kern_ktr.c - Add curthread to the state that ktr is saving. The extra information is 2005-06-10 23:21:29 +00:00
kern_ktrace.c Reuse ktr_unused field in ktr_header structure as ktr_tid; populate 2005-11-01 14:46:37 +00:00
kern_linker.c Fix a bug in the kernel module runtime linker that made it impossible 2005-10-19 20:40:30 +00:00
kern_lock.c Include kdb.h so that kdb_active is declared regardless of KDB being 2005-10-02 10:03:51 +00:00
kern_lockf.c
kern_mac.c Standard Giant push down operations for the Mandatory Access Control (MAC) 2005-10-04 14:32:58 +00:00
kern_malloc.c Detect memory leaks when memory type is being destroyed. 2005-11-03 13:48:59 +00:00
kern_mbuf.c Fix panic string in last revision. 2005-11-06 16:47:59 +00:00
kern_mib.c mp_ncpus is always (properly) initialized, even on UP kernels, so just use it. 2005-08-21 18:03:31 +00:00
kern_module.c
kern_mtxpool.c
kern_mutex.c Move the initialization of the devmtx into the mutex_init() function 2005-10-18 18:27:44 +00:00
kern_ntptime.c Explicitly acquire Giant around the ntp_gettime() and assert it in the 2005-05-28 14:34:41 +00:00
kern_physio.c
kern_pmc.c Fail the module loading process if the currently executing kernel 2005-07-30 09:02:42 +00:00
kern_poll.c - Don't pollute opt_global.h with DEVICE_POLLING and introduce 2005-10-05 10:09:17 +00:00
kern_proc.c Add support for queueing SIGCHLD same as other UNIX systems did. 2005-11-08 09:09:26 +00:00
kern_prot.c Use the refcount API to manage the reference count for user credentials 2005-09-27 18:09:42 +00:00
kern_resource.c Calling setrlimit from 32bit apps could potentially increase certain 2005-11-02 21:18:07 +00:00
kern_sema.c
kern_shutdown.c Don't pretend to be thread0 when calling sync(). 2005-09-22 15:34:15 +00:00
kern_sig.c WIFxxx macros requires an int type but p_xstat is short, convert it 2005-11-09 07:58:16 +00:00
kern_subr.c
kern_switch.c In adjustrunqueue(), add code to handle thread migrating case for 2005-08-03 01:23:45 +00:00
kern_sx.c
kern_synch.c Use low level constructs borrowed from interrupt threads to wait for 2005-05-23 23:01:53 +00:00
kern_syscalls.c
kern_sysctl.c Drop in a WITNESS_WARN into SYSCTL_IN to make sure that we are 2005-08-08 21:06:42 +00:00
kern_tc.c Start time_uptime with 1 instead of 0. 2005-09-19 22:16:31 +00:00
kern_thr.c Cleanup some signal interfaces. Now the tdsignal function accepts 2005-11-03 04:49:16 +00:00
kern_thread.c Add support for queueing SIGCHLD same as other UNIX systems did. 2005-11-08 09:09:26 +00:00
kern_time.c Fix name compatible problem with POSIX standard. the sigval_ptr and 2005-11-04 09:41:00 +00:00
kern_timeout.c Oops, missed adding the required include. 2005-09-15 20:20:36 +00:00
kern_umtx.c do umtx_wake at userland thread exit address, so that others userland 2005-10-26 06:55:46 +00:00
kern_uuid.c Add parse_uuid() that creates a binary representation of an UUID from 2005-10-07 13:37:10 +00:00
kern_xxx.c
ksched.c
link_elf_obj.c Handle vm_map_wire()'s failure. 2005-08-28 05:38:40 +00:00
link_elf.c Check to see if the hash table is present in link_elf_lookup_symbol() 2005-10-31 19:17:32 +00:00
linker_if.m
Make.tags.inc
Makefile Add a 'sysent' target that depends on the various files built from 2005-07-13 20:50:17 +00:00
makesyscalls.sh Introduce a new field in the syscalls.master file format to hold the 2005-05-30 15:09:18 +00:00
md4c.c
md5c.c
p1003_1b.c
posix4_mib.c
sched_4bsd.c Move HWPMC_HOOKS into its own opt_hwpmc_hooks.h file. It doesn't merit 2005-06-24 00:16:57 +00:00
sched_ule.c Temporarily disable nice threshold detection code, as it can starve 2005-09-22 01:19:37 +00:00
subr_acl_posix1e.c Implement new world order in VFS locking for ACLs. This will remove the 2005-09-17 22:01:14 +00:00
subr_autoconf.c
subr_blist.c
subr_bus.c When data passed into devctl_notify is NULL, don't print (null). Instead 2005-10-04 22:25:14 +00:00
subr_clist.c
subr_clock.c
subr_devstat.c - Remove two mtx_asserts that can incorrectly trigger if 2005-05-03 10:58:05 +00:00
subr_disk.c - Fix insertions of bios which represent data earlier than anything else 2005-06-15 23:32:07 +00:00
subr_eventhandler.c
subr_hints.c The resource_xxx routines in subr_hints.c are called before and after the 2005-07-31 10:46:55 +00:00
subr_kdb.c Add stoppcbs[] arrays on Alpha and sparc64 and have each CPU save its 2005-11-03 21:08:20 +00:00
subr_kobj.c Export a routine, kobj_machdep_init(), that allows platforms 2005-08-07 02:20:35 +00:00
subr_log.c
subr_mbpool.c
subr_mchain.c Change API of mb_copy_t in libmchain so that netsmb can handle 2005-07-29 13:22:37 +00:00
subr_module.c
subr_msgbuf.c
subr_param.c Partially revert revision 1.66, which contained a change that did not 2005-10-14 19:15:10 +00:00
subr_pcpu.c Fix 'show allpcpu' ddb command on non-x86. CPU IDs are in the range 0 .. 2005-11-03 21:06:29 +00:00
subr_power.c General consensus is that it would be even better to run this in a 2005-11-09 16:22:56 +00:00
subr_prf.c Acquire Giant in uprintf() and tprintf() rather than asserting it. In 2005-09-26 08:02:24 +00:00
subr_prof.c
subr_rman.c Eliminate __RMAN_RESOURCE_VISIBLE hack entirely by moving the struct 2005-10-06 21:49:31 +00:00
subr_rtc.c
subr_sbuf.c
subr_scanf.c
subr_sleepqueue.c Clarify panic message, I parsed the old one 'trying to sleep while sleeping' 2005-11-09 07:28:52 +00:00
subr_smp.c Rename the KDB_STOP_NMI kernel option to STOP_NMI and make it apply to all 2005-10-24 21:04:19 +00:00
subr_stack.c Add 'depth' argument to CTRSTACK() macro, which allows to reduce number 2005-08-29 11:34:08 +00:00
subr_taskqueue.c Use shorter names for the Giant and fast taskqueues so that their names 2005-10-25 19:29:02 +00:00
subr_trap.c 1. Change prototype of trapsignal and sendsig to use ksiginfo_t *, most 2005-10-14 12:43:47 +00:00
subr_turnstile.c Trim a couple of unneeded includes. 2005-09-29 19:13:52 +00:00
subr_unit.c
subr_witness.c Reorganize the interrupt handling code a bit to make a few things cleaner 2005-10-25 19:48:48 +00:00
sys_generic.c - Add two new system calls: preadv() and pwritev() which are like readv() 2005-07-07 18:17:55 +00:00
sys_pipe.c Fix the recent panics/LORs/hangs created by my kqueue commit by: 2005-07-01 16:28:32 +00:00
sys_process.c Make sure pending SIGCHLD is removed from previous parent when process 2005-11-08 23:28:12 +00:00
sys_socket.c
syscalls.c Regen. 2005-10-30 02:14:37 +00:00
syscalls.master Fix sigevent's POSIX incompatible problem by adding member fields 2005-10-30 02:12:49 +00:00
sysv_ipc.c
sysv_msg.c
sysv_sem.c Gratuitous renaming of four System V Semaphore MAC Framework entry 2005-06-07 05:03:28 +00:00
sysv_shm.c Re-order MAC and DAC checks in shmget() in order to give precedence to 2005-10-04 16:40:20 +00:00
tty_compat.c Make ttsetcompat() static 2005-10-16 20:40:40 +00:00
tty_conf.c
tty_cons.c
tty_pty.c Merge the dev_clone and dev_clone_cred event handlers into a single 2005-08-08 19:55:32 +00:00
tty_subr.c
tty_tty.c Merge the dev_clone and dev_clone_cred event handlers into a single 2005-08-08 19:55:32 +00:00
tty.c Revert most of revision 1.235 and fix the problem a different way. We 2005-10-27 16:47:28 +00:00
uipc_accf.c o setsockopt(2) cannot remove accept filter. [1] 2005-06-11 11:59:48 +00:00
uipc_cow.c Previously, nothing prevented the page that was returned by pmap_extract() 2005-10-23 07:41:56 +00:00
uipc_domain.c
uipc_mbuf2.c Define four constants, MBUF_{,MEM,CLUSTER,PACKET,TAG}_MEM_NAME, which 2005-07-17 14:04:03 +00:00
uipc_mbuf.c Free only those mbuf+clusters back to the packet zone that were allocated 2005-11-05 19:43:55 +00:00
uipc_proto.c Use sparse initializers for "struct domain" and "struct protosw", 2005-11-09 13:29:16 +00:00
uipc_sem.c Const-qualify ksem_timedwait's parameter abstime as it's only passed in. 2005-10-18 11:46:24 +00:00
uipc_sockbuf.c Retire MT_HEADER mbuf type and change its users to use MT_DATA. 2005-11-02 13:46:32 +00:00
uipc_socket2.c Retire MT_HEADER mbuf type and change its users to use MT_DATA. 2005-11-02 13:46:32 +00:00
uipc_socket.c Retire MT_HEADER mbuf type and change its users to use MT_DATA. 2005-11-02 13:46:32 +00:00
uipc_syscalls.c Reformat socket control messages on input/output for 32bit compatibility 2005-10-31 21:09:56 +00:00
uipc_usrreq.c Correct a number of serious and closely related bugs in the UNIX domain 2005-11-10 16:06:04 +00:00
vfs_acl.c Implement new world order in VFS locking for ACLs. This will remove the 2005-09-17 22:01:14 +00:00
vfs_aio.c In aio_waitcomplete, do not return EAGAIN if no other threads 2005-11-08 23:48:32 +00:00
vfs_bio.c Normalize a significant number of kernel malloc type names: 2005-10-31 15:41:29 +00:00
vfs_cache.c - Fix a leaked reference to a vnode via v_dd. We rely on cache_purge() and 2005-06-17 01:05:13 +00:00
vfs_cluster.c Normalize a significant number of kernel malloc type names: 2005-10-31 15:41:29 +00:00
vfs_default.c In vop_stdpathconf(ap) also default for _PC_NAME_MAX and _PC_PATH_MAX. 2005-08-17 06:59:23 +00:00
vfs_export.c Normalize a significant number of kernel malloc type names: 2005-10-31 15:41:29 +00:00
vfs_extattr.c Implement new world order in VFS locking for extended attributes. This will 2005-09-24 23:47:04 +00:00
vfs_hash.c Normalize a significant number of kernel malloc type names: 2005-10-31 15:41:29 +00:00
vfs_init.c
vfs_lookup.c Use correct VFS locking rather than unconditionally grabbing Giant around 2005-09-21 19:49:42 +00:00
vfs_mount.c For nmount(), allow a text string error message to be propagated back 2005-11-09 02:26:38 +00:00
vfs_subr.c This is a workaround for a complicated issue involving VFS cookies and devfs. 2005-11-09 22:03:50 +00:00
vfs_syscalls.c Implement new world order in VFS locking for extended attributes. This will 2005-09-24 23:47:04 +00:00
vfs_vnops.c - Replace the series of DEBUG_LOCKS hacks which tried to save the vn_lock 2005-08-03 04:48:22 +00:00
vnode_if.src Allow EVFILT_VNODE events to work on every filesystem type, not just 2005-06-09 20:20:31 +00:00