Commit Graph

144 Commits

Author SHA1 Message Date
Jason Evans
98a1f447bb Add a wrapper for the sendfile() system call.
PR:		bin/17366
2000-04-27 00:59:44 +00:00
Jason Evans
7e5e179982 Explicitly include sys/cdefs.h to get the definition of __strong_reference(),
rather than getting lucky due to header dependencies.
2000-03-18 22:36:46 +00:00
Daniel Eischen
1d013a86ed Fix pthread_suspend_np/pthread_resume_np. For the record, suspending a
thread waiting on an event (I/O, condvar, etc) will, when resumed using
pthread_resume_np, return with EINTR.  For example, suspending and resuming
a thread blocked on read() will not requeue the thread for the read, but
will return -1 with errno = EINTR.  If the suspended thread is in a critical
region, the thread is suspended as soon as it leaves the critical region.

Fix a bogon in pthread_kill() where a signal was being delivered twice
to threads waiting in sigwait().

Reported by (suspend/resume bug):	jdp
Reviewed by:	jasone
2000-03-15 13:59:27 +00:00
Jason Evans
8a668b9049 For errors, return -1 and set errno to indicate the error type, rather than
returning the error directly.

For sem_post(), make sure that the correct thread is woken up.  This has
unfortunate performance implications, but is necessary for POSIX compliance.

Approved by:	jkh
2000-02-16 19:34:53 +00:00
Jason Evans
d77639f25f Use __strong_reference() instead of __weak_reference() to assure that the
weak symbols of the same name are not used.
2000-01-29 22:53:55 +00:00
Jason Evans
4ec6f67d10 Use _fcntl() (not fcntl()) inside of fcntl().
Reported by:	green
2000-01-28 22:10:27 +00:00
Jason Evans
9233c4d942 Simplify sytem call renaming. Instead of _foo() <-- _libc_foo <-- foo(),
just use _foo() <-- foo().  In the case of a libpthread that doesn't do
call conversion (such as linuxthreads and our upcoming libpthread), this
is adequate.  In the case of libc_r, we still need three names, which are
now _thread_sys_foo() <-- _foo() <-- foo().

Convert all internal libc usage of: aio_suspend(), close(), fsync(), msync(),
nanosleep(), open(), fcntl(), read(), and write() to _foo() instead of foo().

Remove all internal libc usage of: creat(), pause(), sleep(), system(),
tcdrain(), wait(), and waitpid().

Make thread cancellation fully POSIX-compliant.

Suggested by:	deischen
2000-01-27 23:07:25 +00:00
Jason Evans
0c5d1a3361 Fix millisecond to nanosecond conversion.
PR:		misc/16245
2000-01-22 09:15:15 +00:00
Jason Evans
beab1ec9b5 Minor *jmp() cleanups. 2000-01-20 21:53:59 +00:00
Jason Evans
120bfc9ded Add sem_*() functions. Named semaphores and process-shared semaphores
are not supported by this implementation, and the error return values
from sem_init(), sem_open(), sem_close(), and sem_unlink() reflect this.

Approved by:	jkh
2000-01-20 07:54:49 +00:00
Jason Evans
adbd6ee028 Do signal deferral for pthread_kill() as it was done in the old days.
Submitted by:	deischen
2000-01-20 04:46:52 +00:00
Jason Evans
0a3fa43c7e Implement continuations to correctly handle [sig|_]longjmp() inside of a
signal handler.  Explicitly check for jumps to anywhere other than the
current stack, since such jumps are undefined according to POSIX.

While we're at it, convert thread cancellation to use continuations, since
it's cleaner than the original cancellation code.

Avoid delivering a signal to a thread twice.  This was a pre-existing bug,
but was likely unexposed until these other changes were made.

Defer signals generated by pthread_kill() so that they can be delivered on
the appropriate stack.  deischen claims that this is unnecessary, which is
likely true, but without this change, pthread_kill() can cause undefined
priority queue states and/or PANICs in [sig|_]longjmp(), so I'm leaving
this in for now.  To compile this code out and exercise the bug, define
the _NO_UNDISPATCH cpp macro.  Defining _PTHREADS_INVARIANTS as well will
cause earlier crashes.

PR:			kern/14685
Collaboration with:	deischen
2000-01-19 07:04:50 +00:00
Daniel Eischen
7e416d8e49 Properly initialize the last active time of the initial thread. This fixes
the case that a CPU hungry main thread is prevented from being preempted
due to a negative calculation of its time slice.

Reported by:	Alexander Litvin <archer@lucky.net>
2000-01-18 11:35:25 +00:00
Jason Evans
f560c4e709 Track libc's three-tier symbol naming. libc_r must currently implement
the _libc_*() entry points and add *() weak aliases.  This will all
change for the better when libc_r becomes libpthread.
2000-01-12 09:28:58 +00:00
Daniel Eischen
3eadb69d1f Make sched_param parameter a const to comply with POSIX and SUSv2 specs.
This doesn't need to be applied to stable, because somehow -stable seems
to have gotten it right.

Reviewed by:	jasone
2000-01-10 04:14:08 +00:00
Jason Evans
91e92a2d3d Don't explicitly mmap() red zones at the bottom of thread stacks (except
the initial thread).  Instead, just leave an unmapped gap between thread
stacks and make sure that the thread stacks won't grow into these gaps,
simply by limiting the size of the stacks with the 'len' argument to
mmap().  This (if I understand correctly) reduces VM overhead
considerably.

Reviewed by:	deischen
1999-12-29 15:44:59 +00:00
Daniel Eischen
1752e66d49 -Wall and minor style(9) cleanups. 1999-12-28 18:13:04 +00:00
Daniel Eischen
69186ed701 Change stack allocation algorithm to make better use of memory
(it was leaving an unused block).  Also protect the global stack
pointer from context changes while fiddling with it.
1999-12-28 18:12:07 +00:00
Daniel Eischen
8d048bba15 Don't wakeup threads when there is a process signal and no installed
handler.  Thread-to-thread signals (pthread_signal) are treated differently
than process signals; a pthread_signal can wakeup a blocked thread if
a signal handler is not installed for that signal.

Found by:	ACE tests
1999-12-28 18:08:09 +00:00
Jason Evans
386812d440 Fix some minor POSIX/SUSv2 compliance nits.
PR:		kern/11982
1999-12-18 01:00:42 +00:00
Daniel Eischen
fc8f3f5bfe Fix problems with cancellation while in critical regions.
o Cancellation flags were not getting properly set/cleared.
  o Loops waiting for internal locks were not being exited
    correctly by a cancelled thread.
  o Minor spelling (cancelation -> cancellation) and formatting
    corrections (missing tab).

Found by:	tg
Reviewed by:	jasone
1999-12-17 00:57:54 +00:00
Daniel Eischen
3dc268f4e7 Fixes for signal handling:
o Don't call signal handlers with the signal handler access lock
    held.
  o Remove pending signals before calling signal handlers.  If
    pending signals were not removed prior to handling them,
    invocation of the handler could cause the handler to be
    called more than once for the same signal.  Found by: JB
  o When SIGCHLD arrives, wake up all threads in PS_WAIT_WAIT
    (wait4).

PR:		bin/15328
Reviewed by:	jasone
1999-12-17 00:56:36 +00:00
Jason Evans
e6a5e33c6b Avoid an infinite loop if the last element of the iov array passed to
writev() has an iov_len of 0.

PR:		bin/8281
1999-12-16 22:35:40 +00:00
Daniel Eischen
4fc937ef8f Change signal handling to conform to POSIX specified semantics.
Before this change, a signal was delivered to each thread that
didn't have the signal masked.  Signals also improperly woke up
threads waiting on I/O.  With this change, signals are now
handled in the following way:

  o If a thread is waiting in a sigwait for the signal,
    then the thread is woken up.

  o If no threads are sigwait'ing on the signal and a
    thread is in a sigsuspend waiting for the signal,
    then the thread is woken up.

  o In the case that no threads are waiting or suspended
    on the signal, then the signal is delivered to the
    first thread we find that has the signal unmasked.

  o If no threads are waiting or suspended on the signal,
    and no threads have the signal unmasked, then the signal
    is added to the process wide pending signal set.  The
    signal will be delivered to the first thread that unmasks
    the signal.

If there is an installed signal handler, it is only invoked
if the chosen thread was not in a sigwait.

In the case that multiple threads are waiting or suspended
on a signal, or multiple threads have the signal unmasked,
we wake up/deliver the signal to the first thread we find.
The above rules still apply.

Reported by:	Scott Hess <scott@avantgo.com>
Reviewed by:	jb, jasone
1999-12-04 22:55:59 +00:00
Doug Rabson
3c085f72bb * Fix the stack allocation code so that it works for alpha. Change it
to use mmap(..., MAP_STACK, ...) on alpha too since that should work
  now.
* Add hooks to allow GDB to access the internals of pthreads without
  having to know the exact layout of struct pthread.

Reviewed by: deischen
1999-11-28 19:47:43 +00:00
Alfred Perlstein
7285bccf1a add pthread_cancel, obtained from OpenBSD.
eischen (Daniel Eischen) added wrappers to protect against cancled
threads orphaning internal resources.

the cancelability code is still a bit fuzzy but works for test
programs of my own, OpenBSD's and some examples from ORA's books.

add readdir_r to both libc and libc_r

add some 'const' attributes to function parameters

Reviewed by: eischen, jasone
1999-11-28 05:38:13 +00:00
Marcel Moolenaar
74562f1a2a Add the proper headers so that the SIGNOTEMPTY and SIGSETAND
macros are defined.
1999-09-30 14:51:31 +00:00
Marcel Moolenaar
3cf3c5d9dd sigset_t change (part 5 of 5)
-----------------------------

Most of the userland changes are in libc. For both the alpha
and the i386 setjmp has been changed to accomodate for the
new sigset_t. Internally, libc is mostly rewritten to use the
new syscalls. The exception is in compat-43/sigcompat.c

The POSIX thread library has also been rewritten to use the
new sigset_t. Except, that it currently only handles NSIG
signals instead of the maximum _SIG_MAXSIG. This should not
be a problem because current applications don't use any
signals higher than NSIG.

There are version bumps for the following libraries:
  libdialog
  libreadline
  libc
  libc_r
  libedit
  libftpio
  libss

These libraries either a) have one of the modified structures
visible in the interface, or b) use sigset_t internally and
may cause breakage if new binaries are used against libraries
that don't have the sigset_t change. This not an immediate
issue, but will be as soon as applications start using the
new range to its fullest.

NOTE: libncurses already had an version bump and has not been
      given one now.

NOTE: doscmd is a real casualty and has been disconnected for
      the moment. Reconnection will eventually happen after
      doscmd has been fixed. I'm aware that being the last one
      to touch it, I'm automaticly promoted to being maintainer.
      According to good taste this means that I will receive a
      badge which either will be glued or mechanically stapled,
      drilled or otherwise violently forced onto me :-)

NOTE: pcvt/vttest cannot be compiled with -traditional. The
      change cause sys/types to be included along the way which
      contains the const and volatile modifiers. I don't consider
      this a solution, but more a workaround.
1999-09-29 15:18:46 +00:00
Dmitrij Tejblum
ffd73a0043 Fix a memory leak: free the thread-specific poll_data, used in the select()
implementation.

PR:		13368
Submitted by:	Steve Bernacki, Jr. <steve@copacetic.net>
1999-08-30 15:45:42 +00:00
Daniel Eischen
3e12058d25 When checking for valid timevals in the wrapped select() and poll()
routines, don't return EINVAL but set errno to EINVAL and return -1.
Added a check in pthread_cond_timedwait for a null timespec pointer.
1999-08-30 00:02:08 +00:00
Peter Wemm
7f3dea244c $Id$ -> $FreeBSD$ 1999-08-28 00:22:10 +00:00
Alfred Perlstein
89f6acf96a move sanity check of timeval struct so as not to segfault when passed
a null pointer.

Pointed out by: Thomas Gellekum <tg@ihf.rwth-aachen.de>
Reviewed by:    eivind
1999-08-20 21:06:20 +00:00
John Birrell
e4065e8294 When checking if there is a stack to free, observe the fact that it
might have been mmapped, and if so, passing the pointer to free() is
really not a good idea.

[ In the next millenium, when I've taken over the world, I'm going
  to ban 8 character tabs. You've been warned. ]
1999-08-20 12:17:09 +00:00
Alfred Perlstein
4c263f605c time_to_sleep->tv_nsec > 1000000000
-to-
time_to_sleep->tv_nsec >= 1000000000
1999-08-19 23:18:07 +00:00
Alfred Perlstein
91518882e0 Sanity check time structures passed in, return EINVAL like the system
calls do to avoid corrupting the thread library's concept of wakeup
time.

PR:		kern/12141
Reviewed by:	deischen, eivind
1999-08-19 23:06:11 +00:00
Alfred Perlstein
c81c6baf1d handle under/overflow of time values in a more robust manner,
there may be an overflow that need to be adjusted more than once.

Pointed out by: Fabian Thylmann <fthylmann@stats.net>

Reviewed by:	eivind, jb
1999-08-19 16:49:53 +00:00
Daniel Eischen
936aa6b443 Add check for runnable threads before polling file descriptors.
Submitted by:	tegge
1999-08-12 19:34:39 +00:00
Daniel Eischen
fa7c4d5575 Add RCS IDs to those files without them.
Fix copyrights (s/REGENTS/AUTHOR).

Suggested by:	tg
Approved by:	jb
1999-08-05 12:15:30 +00:00
Daniel Eischen
8b5d18ec76 Fix thread initialization to allow for the case where stdio file
descriptors are not opened.

PR:		bin/12853
Reviewed by:	jb
1999-08-05 12:08:10 +00:00
Dmitrij Tejblum
6ff19e189d Use USRSTACK (defined in <machine/vmparam.h>) to get top of the initial stack.
PTHREAD_STACK_TOP was wrong for all supported architectures.
1999-07-12 16:09:30 +00:00
Jason Evans
66da783384 Fix a couple more coding style nits. 1999-07-11 06:06:52 +00:00
Jason Evans
34582929f1 Modify previous changes to conform better to libc_r's coding style.
Always use mmap() for default-size stack allocation.  Use MAP_ANON instead
of MAP_STACK on the alpha architecture.

Reduce the amount of code executed while owning _gc_mutex during stack
allocation.
1999-07-11 05:56:37 +00:00
Jason Evans
876cc3dae0 Always use growable thread stacks on the i386. The VM_STACK kernel option
must be made default for the alpha before growable thread stacks are
enabled for the alpha.
1999-07-06 00:25:38 +00:00
Jason Evans
ecaa6e8c9e Use growable stacks for thread stacks that are the default stack size.
Cache discarded default thread stacks for use in subsequent thread creations.

Create a red zone at the end of each stack (including the initial thread
stack), with the hope of causing a segfault if a stack overflows.

To activate these modifications, add -D_PTHREAD_GSTACK to CFLAGS in
src/lib/libc_r/Makefile.  Since the modifications depend on the VM_STACK
kernel option, I'm not sure how to safely use growable stacks by default.

Testing, as well as algorithmic and stylistic comments are welcome.
1999-07-05 00:35:19 +00:00
Dmitrij Tejblum
db06cf96d5 Defer signals, so we will not wait for SIGCHLD after it was delivered. 1999-06-29 19:57:07 +00:00
Dmitrij Tejblum
2467b7977b Fix few warnings on the alpha. 1999-06-23 15:01:22 +00:00
John Birrell
02292f131a In the words of the author:
o The polling mechanism for I/O readiness was changed from
    select() to poll().  In additon, a wrapped version of poll()
    is now provided.

  o The wrapped select routine now converts each fd_set to a
    poll array so that the thread scheduler doesn't have to
    perform a bitwise search for selected fds each time file
    descriptors are polled for I/O readiness.

  o The thread scheduler was modified to use a new queue (_workq)
    for threads that need work.  Threads waiting for I/O readiness
    and spinblocks are added to the work queue in addition to the
    waiting queue.  This reduces the time spent forming/searching
    the array of file descriptors being polled.

  o The waiting queue (_waitingq) is now maintained in order of
    thread wakeup time.  This allows the thread scheduler to
    find the nearest wakeup time by looking at the first thread
    in the queue instead of searching the entire queue.

  o Removed file descriptor locking for select/poll routines.  An
    application should not rely on the threads library for providing
    this locking; if necessary, the application should use mutexes
    to protect selecting/polling of file descriptors.

  o Retrieve and use the kernel clock rate/resolution at startup
    instead of hardcoding the clock resolution to 10 msec (tested
    with kernel running at 1000 HZ).

  o All queues have been changed to use queue.h macros.  These
    include the queues of all threads, dead threads, and threads
    waiting for file descriptor locks.

  o Added reinitialization of the GC mutex and condition variable
    after a fork.  Also prevented reallocation of the ready queue
    after a fork.

  o Prevented the wrapped close routine from closing the thread
    kernel pipes.

  o Initialized file descriptor table for stdio entries at thread
    init.

  o Provided additional flags to indicate to what queues threads
    belong.

  o Moved TAILQ initialization for statically allocated mutex and
    condition variables to after the spinlock.

  o Added dispatching of signals to pthread_kill.  Removing the
    dispatching of signals from thread activation broke sigsuspend
    when pthread_kill was used to send a signal to a thread.

  o Temporarily set the state of a thread to PS_SUSPENDED when it
    is first created and placed in the list of threads so that it
    will not be accidentally scheduled before becoming a member
    of one of the scheduling queues.

  o Change the signal handler to queue signals to the thread kernel
    pipe if the scheduling queues are protected.  When scheduling
    queues are unprotected, signals are then dequeued and handled.

  o Ensured that all installed signal handlers block the scheduling
    signal and that the scheduling signal handler blocks all
    other signals.  This ensures that the signal handler is only
    interruptible for and by non-scheduling signals.  An atomic
    lock is used to decide which instance of the signal handler
    will handle pending signals.

  o Removed _lock_thread_list and _unlock_thread_list as they are
    no longer used to protect the thread list.

  o Added missing RCS IDs to modified files.

  o Added checks for appropriate queue membership and activity when
    adding, removing, and searching the scheduling queues.  These
    checks add very little overhead and are enabled when compiled
    with _PTHREADS_INVARIANTS defined.  Suggested and implemented
    by Tor Egge with some modification by me.

  o Close a race condition in uthread_close.  (Tor Egge)

  o Protect the scheduling queues while modifying them in
    pthread_cond_signal and _thread_fd_unlock.  (Tor Egge)

  o Ensure that when a thread gets a mutex, the mutex is on that
    threads list of owned mutexes.  (Tor Egge)

  o Set the kernel-in-scheduler flag in _thread_kern_sched_state
    and _thread_kern_sched_state_unlock to prevent a scheduling
    signal from calling the scheduler again.  (Tor Egge)

  o Don't use TAILQ_FOREACH macro while searching the waiting
    queue for threads in a sigwait state, because a change of
    state destroys the TAILQ link.  It is actually safe to do
    so, though, because once a sigwaiting thread is found, the
    loop ends and the function returns.  (Tor Egge)

  o When dispatching signals to threads, make the thread inherit
    the signal deferral flag of the currently running thread.
    (Tor Egge)

Submitted by: Daniel Eischen <eischen@vigrid.com> and
              Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
John Birrell
994d9b67a5 Added a missing comma to the static condition variable initialisation
definition.

Submitted by: David Leonard <David.Leonard@csee.uq.edu.au>, an OpenBSD guy.
1999-05-24 07:22:55 +00:00
John Birrell
eb9dc34d8b Fix a problem with static initialisation of mutexes and condition
variables.

Submitted by: Dan Eischen <eischen@vigrid.com>
1999-05-23 10:55:33 +00:00
John Birrell
65d33b21e6 Ensure that an existing thread gets it's state set to PS_DEAD to
prevent being rescheduled.

Submitted by: Dan Eischen <eischen@vigrid.com>
1999-05-16 05:25:37 +00:00