musl
10 years agoadd netpacket/packet.h
Rich Felker [Fri, 10 Jun 2011 01:47:24 +0000 (21:47 -0400)]
add netpacket/packet.h

looks like busybox is going to want it, and apparently some other
low-level network software does too...

10 years agoit's called getgrgid_r, not getgruid_r...
Rich Felker [Wed, 8 Jun 2011 20:47:08 +0000 (16:47 -0400)]
it's called getgrgid_r, not getgruid_r...

10 years agoimplement psignal and psiginfo
Rich Felker [Wed, 8 Jun 2011 20:41:44 +0000 (16:41 -0400)]
implement psignal and psiginfo

10 years agoisgreater etc. relation macros for math.h
Rich Felker [Wed, 8 Jun 2011 20:33:04 +0000 (16:33 -0400)]
isgreater etc. relation macros for math.h

10 years agomq names without leading / have impl-def behavior; allowing them is easier
Rich Felker [Tue, 7 Jun 2011 19:07:54 +0000 (15:07 -0400)]
mq names without leading / have impl-def behavior; allowing them is easier

10 years agodefine MQ_PRIO_MAX
Rich Felker [Tue, 7 Jun 2011 19:05:04 +0000 (15:05 -0400)]
define MQ_PRIO_MAX

10 years agouse __WCHAR_TYPE__ on i386 if it is defined
Rich Felker [Tue, 7 Jun 2011 15:26:42 +0000 (11:26 -0400)]
use __WCHAR_TYPE__ on i386 if it is defined

unfortunately traditional i386 practice was to use "long" rather than
"int" for wchar_t, despite the latter being much more natural and
logical. we followed this practice, but it seems some compilers (clang
and maybe certain gcc builds or others too..?) have switched to using
int, resulting in spurious pointer type mismatches when L"..." wide
strings are used. the best solution I could find is to use the
compiler's definition of wchar_t if it exists, and otherwise fallback
to the traditional definition.

there's no point in duplicating this approach on 64-bit archs, as
their only 32-bit type is int.

10 years agomq send/recv functions are cancellation points
Rich Felker [Tue, 7 Jun 2011 15:14:39 +0000 (11:14 -0400)]
mq send/recv functions are cancellation points

10 years agoimplement mq_notify
Rich Felker [Tue, 7 Jun 2011 06:42:55 +0000 (02:42 -0400)]
implement mq_notify

10 years agoadd support for POSIX message queues, except mq_notify
Rich Felker [Tue, 7 Jun 2011 05:52:27 +0000 (01:52 -0400)]
add support for POSIX message queues, except mq_notify

10 years agouse __attribute__((const)) for errno and pthread_self if __GNUC__ is defined
Rich Felker [Tue, 7 Jun 2011 00:12:42 +0000 (20:12 -0400)]
use __attribute__((const)) for errno and pthread_self if __GNUC__ is defined

this is not too ugly and should result in significant code size and
performance improvements for many programs.

10 years agouse volatile pointers for intentional-crash code.
Rich Felker [Mon, 6 Jun 2011 22:10:43 +0000 (18:10 -0400)]
use volatile pointers for intentional-crash code.

10 years agofix handling of d_name in struct dirent
Rich Felker [Mon, 6 Jun 2011 22:04:28 +0000 (18:04 -0400)]
fix handling of d_name in struct dirent

basically there are 3 choices for how to implement this variable-size
string member:
1. C99 flexible array member: breaks using dirent.h with pre-C99 compiler.
2. old way: length-1 string: generates array bounds warnings in caller.
3. new way: length-NAME_MAX string. no problems, simplifies all code.

of course the usable part in the pointer returned by readdir might be
shorter than NAME_MAX+1 bytes, but that is allowed by the standard and
doesn't hurt anything.

10 years agosafety fix for glob's vla usage: disallow patterns longer than PATH_MAX
Rich Felker [Sun, 5 Jun 2011 23:29:52 +0000 (19:29 -0400)]
safety fix for glob's vla usage: disallow patterns longer than PATH_MAX

this actually inadvertently disallows some valid patterns with
redundant / or * characters, but it's better than allowing unbounded
vla allocation.

eventually i'll write code to move the pattern to the stack and
eliminate redundancy to ensure that it fits in PATH_MAX at the
beginning of glob. this would also allow it to be modified in place
for passing to fnmatch rather than copied at each level of recursion.

10 years agoeliminate (harmless in this case) vla usage in fnmatch.c
Rich Felker [Sun, 5 Jun 2011 17:30:56 +0000 (13:30 -0400)]
eliminate (harmless in this case) vla usage in fnmatch.c

10 years agomissing prototypes for mbsnrtowcs and wcsnrtombs
Rich Felker [Mon, 30 May 2011 17:32:40 +0000 (13:32 -0400)]
missing prototypes for mbsnrtowcs and wcsnrtombs

10 years agoimplement pthread_[sg]etconcurrency.
Rich Felker [Mon, 30 May 2011 15:31:07 +0000 (11:31 -0400)]
implement pthread_[sg]etconcurrency.

there is a resource limit of 0 bits to store the concurrency level
requested. thus any positive level exceeds a resource limit, resulting
in EAGAIN. :-)

10 years agoimplement uselocale function (minimal)
Rich Felker [Mon, 30 May 2011 05:41:23 +0000 (01:41 -0400)]
implement uselocale function (minimal)

10 years agoadd useless, obsolescent function ulimit
Rich Felker [Sun, 29 May 2011 18:09:03 +0000 (14:09 -0400)]
add useless, obsolescent function ulimit

10 years agofix backwards posix_spawn file action order
Rich Felker [Sun, 29 May 2011 16:58:02 +0000 (12:58 -0400)]
fix backwards posix_spawn file action order

10 years agoadd accidentally-omitted file needed for posix_spawn file actions
Rich Felker [Sun, 29 May 2011 03:31:11 +0000 (23:31 -0400)]
add accidentally-omitted file needed for posix_spawn file actions

10 years agoadd file actions support to posix_spawn
Rich Felker [Sun, 29 May 2011 03:30:47 +0000 (23:30 -0400)]
add file actions support to posix_spawn

10 years agoposix_spawn: honor POSIX_SPAWN_SETSIGDEF flag
Rich Felker [Sat, 28 May 2011 22:39:43 +0000 (18:39 -0400)]
posix_spawn: honor POSIX_SPAWN_SETSIGDEF flag

10 years agoinitial implementation of posix_spawn
Rich Felker [Sat, 28 May 2011 22:36:30 +0000 (18:36 -0400)]
initial implementation of posix_spawn

file actions are not yet implemented, but everything else should be
mostly complete and roughly correct.

10 years agomodernize coding style in sjlj asm
Rich Felker [Fri, 27 May 2011 00:59:02 +0000 (20:59 -0400)]
modernize coding style in sjlj asm

10 years agofix strncat and wcsncat (double null termination)
Rich Felker [Mon, 23 May 2011 01:58:43 +0000 (21:58 -0400)]
fix strncat and wcsncat (double null termination)

also modify wcsncpy to use the same loop logic

10 years agofix wcsncpy writing past end of buffer
Rich Felker [Mon, 23 May 2011 01:54:42 +0000 (21:54 -0400)]
fix wcsncpy writing past end of buffer

10 years agofix brk/sbrk behavior to match the real legacy functions
Rich Felker [Sun, 22 May 2011 16:34:33 +0000 (12:34 -0400)]
fix brk/sbrk behavior to match the real legacy functions

11 years agorelease notes for 0.7.10
Rich Felker [Tue, 17 May 2011 18:13:06 +0000 (14:13 -0400)]
release notes for 0.7.10

11 years agofix the last known rounding bug in floating point printing
Rich Felker [Wed, 11 May 2011 23:58:03 +0000 (19:58 -0400)]
fix the last known rounding bug in floating point printing

the observed symptom was that the code was incorrectly rounding up
1.0625 to 1.063 despite the rounding mode being round-to-nearest with
ties broken by rounding to even last place. however, the code was just
not right in many respects, and i'm surprised it worked as well as it
did. this time i tested the values that end up in the variables round,
small, and the expression round+small, and all look good.

11 years agoeventfd syscall wrapper and read/write wrappers
Rich Felker [Sun, 8 May 2011 04:03:00 +0000 (00:03 -0400)]
eventfd syscall wrapper and read/write wrappers

11 years agooptimize out useless default-attribute object in pthread_create
Rich Felker [Sun, 8 May 2011 03:39:48 +0000 (23:39 -0400)]
optimize out useless default-attribute object in pthread_create

11 years agooptimize compound-literal sigset_t's not to contain useless hurd bits
Rich Felker [Sun, 8 May 2011 03:37:10 +0000 (23:37 -0400)]
optimize compound-literal sigset_t's not to contain useless hurd bits

11 years agooverhaul implementation-internal signal protections
Rich Felker [Sun, 8 May 2011 03:23:58 +0000 (23:23 -0400)]
overhaul implementation-internal signal protections

the new approach relies on the fact that the only ways to create
sigset_t objects without invoking UB are to use the sig*set()
functions, or from the masks returned by sigprocmask, sigaction, etc.
or in the ucontext_t argument to a signal handler. thus, as long as
sigfillset and sigaddset avoid adding the "protected" signals, there
is no way the application will ever obtain a sigset_t including these
bits, and thus no need to add the overhead of checking/clearing them
when sigprocmask or sigaction is called.

note that the old code actually *failed* to remove the bits from
sa_mask when sigaction was called.

the new implementations are also significantly smaller, simpler, and
faster due to ignoring the useless "GNU HURD signals" 65-1024, which
are not used and, if there's any sanity in the world, never will be
used.

11 years agoreduce some ridiculously large spin counts
Rich Felker [Sat, 7 May 2011 01:45:48 +0000 (21:45 -0400)]
reduce some ridiculously large spin counts

these should be tweaked according to testing. offhand i know 1000 is
too low and 5000 is likely to be sufficiently high. consider trying to
add futexes to file locking, too...

11 years agoremove debug code that was missed in barrier commit
Rich Felker [Sat, 7 May 2011 00:27:45 +0000 (20:27 -0400)]
remove debug code that was missed in barrier commit

11 years agocompletely new barrier implementation, addressing major correctness issues
Rich Felker [Sat, 7 May 2011 00:00:59 +0000 (20:00 -0400)]
completely new barrier implementation, addressing major correctness issues

the previous implementation had at least 2 problems:

1. the case where additional threads reached the barrier before the
first wave was finished leaving the barrier was untested and seemed
not to be working.

2. threads leaving the barrier continued to access memory within the
barrier object after other threads had successfully returned from
pthread_barrier_wait. this could lead to memory corruption or crashes
if the barrier object had automatic storage in one of the waiting
threads and went out of scope before all threads finished returning,
or if one thread unmapped the memory in which the barrier object
lived.

the new implementation avoids both problems by making the barrier
state essentially local to the first thread which enters the barrier
wait, and forces that thread to be the last to return.

11 years agoadd SA_NOMASK alias for SA_NODEFER with _GNU_SOURCE
Rich Felker [Thu, 5 May 2011 21:44:06 +0000 (17:44 -0400)]
add SA_NOMASK alias for SA_NODEFER with _GNU_SOURCE

11 years agofix fclose return status logic, again
Rich Felker [Mon, 2 May 2011 13:18:03 +0000 (09:18 -0400)]
fix fclose return status logic, again

the previous fix was incorrect, as it would prevent f->close(f) from
being called if fflush(f) failed. i believe this was the original
motivation for using | rather than ||. so now let's just use a second
statement to constrain the order of function calls, and to back to
using |.

11 years agofix undefined call order in fclose, possible lost output depending on compiler
Rich Felker [Mon, 2 May 2011 02:59:14 +0000 (22:59 -0400)]
fix undefined call order in fclose, possible lost output depending on compiler

pcc turned up this bug by calling f->close(f) before fflush(f),
resulting in lost output and error on flush.

11 years agoworkaround for preprocessor bug in pcc
Rich Felker [Mon, 2 May 2011 02:16:04 +0000 (22:16 -0400)]
workaround for preprocessor bug in pcc

with this patch, musl compiles and mostly works with pcc 1.0.0. a few
tests are still failing and i'm uncertain whether they are due to
portability problems in musl, or bugs in pcc, but i suspect the
latter.

11 years agominor optimization in puts: use inline putc_unlocked macro for newline
Rich Felker [Mon, 2 May 2011 00:12:51 +0000 (20:12 -0400)]
minor optimization in puts: use inline putc_unlocked macro for newline

11 years agoavoid crashing when nel==0 is passed to qsort
Rich Felker [Fri, 29 Apr 2011 15:14:55 +0000 (11:14 -0400)]
avoid crashing when nel==0 is passed to qsort

11 years agouse compiler builtins for variadic macros when available
Rich Felker [Thu, 28 Apr 2011 03:41:48 +0000 (23:41 -0400)]
use compiler builtins for variadic macros when available

this slightly cuts down on the degree musl "fights with" gcc, but more
importantly, it fixes a critical bug when gcc inlines a variadic
function and optimizes out the variadic arguments due to noticing that
they were "not used" (by __builtin_va_arg).

we leave the old code in place if __GNUC__ >= 3 is false; it seems
like it might be necessary at least for tinycc support and perhaps if
anyone ever gets around to fixing gcc 2.95.3 enough to make it work..

11 years agocorrect variadic prototypes for execl* family
Rich Felker [Wed, 27 Apr 2011 20:06:33 +0000 (16:06 -0400)]
correct variadic prototypes for execl* family

the old versions worked, but conflicted with programs which declared
their own prototypes and generated warnings with some versions of gcc.

11 years agoreplace heap sort with smoothsort implementation by Valentin Ochs
Rich Felker [Wed, 27 Apr 2011 17:27:04 +0000 (13:27 -0400)]
replace heap sort with smoothsort implementation by Valentin Ochs

Smoothsort is an adaptive variant of heapsort. This version was
written by Valentin Ochs (apo) specifically for inclusion in musl. I
worked with him to get it working in O(1) memory usage even with giant
array element widths, and to optimize it heavily for size and speed.
It's still roughly 4 times as large as the old heap sort
implementation, but roughly 20 times faster given an almost-sorted
array of 1M elements (20 being the base-2 log of 1M), i.e. it really
does reduce O(n log n) to O(n) in the mostly-sorted case. It's still
somewhat slower than glibc's Introsort for random input, but now
considerably faster than glibc when the input is already sorted, or
mostly sorted.

11 years agoadd word-sized ctz function to atomic.h
Rich Felker [Wed, 27 Apr 2011 16:19:49 +0000 (12:19 -0400)]
add word-sized ctz function to atomic.h

strictly speaking this and a few other ops should be factored into
asm.h or the file should just be renamed to asm.h, but whatever. clean
it up someday.

11 years agofunction signature fix: add const qualifier to mempcpy src arg
Rich Felker [Tue, 26 Apr 2011 16:28:41 +0000 (12:28 -0400)]
function signature fix: add const qualifier to mempcpy src arg

11 years agotypo in prototype for mempcpy
Rich Felker [Tue, 26 Apr 2011 12:42:55 +0000 (08:42 -0400)]
typo in prototype for mempcpy

11 years agoprototype for mempcpy
Rich Felker [Tue, 26 Apr 2011 12:41:54 +0000 (08:41 -0400)]
prototype for mempcpy

11 years agofix bug in ipv6 parsing that prevented parsing a lone "::"
Rich Felker [Mon, 25 Apr 2011 21:49:21 +0000 (17:49 -0400)]
fix bug in ipv6 parsing that prevented parsing a lone "::"

11 years agoipv6 parsing code (formerly dummied-out)
Rich Felker [Mon, 25 Apr 2011 21:04:40 +0000 (17:04 -0400)]
ipv6 parsing code (formerly dummied-out)

11 years agofix 2 eof-related bugs in scanf
Rich Felker [Mon, 25 Apr 2011 14:40:25 +0000 (10:40 -0400)]
fix 2 eof-related bugs in scanf

1. failed match of literal chars from the format string would always
return matching failure rather than input failure at eof, leading to
infinite loops in some programs.

2. unread of eof would wrongly adjust the character counts reported by
%n, yielding an off-by-one error.

11 years agoupdate release notes to new format, add more
Rich Felker [Sat, 23 Apr 2011 00:14:04 +0000 (20:14 -0400)]
update release notes to new format, add more

11 years agofix initial stack alignment in new threads on x86_64
Rich Felker [Fri, 22 Apr 2011 22:48:57 +0000 (18:48 -0400)]
fix initial stack alignment in new threads on x86_64

11 years agoupdate release notes in prep for 0.7.9 release
Rich Felker [Fri, 22 Apr 2011 03:33:39 +0000 (23:33 -0400)]
update release notes in prep for 0.7.9 release

11 years agoomit errno update path for syscalls that cannot fail
Rich Felker [Fri, 22 Apr 2011 01:52:41 +0000 (21:52 -0400)]
omit errno update path for syscalls that cannot fail

11 years agofix bogus return values for inet_pton
Rich Felker [Thu, 21 Apr 2011 20:57:00 +0000 (16:57 -0400)]
fix bogus return values for inet_pton

11 years agomove wait.h macros out of bits. they do not vary.
Rich Felker [Thu, 21 Apr 2011 18:27:28 +0000 (14:27 -0400)]
move wait.h macros out of bits. they do not vary.

11 years agoinclude signal.h to avoid thorny __sigcontext/sigcontext issues
Rich Felker [Thu, 21 Apr 2011 18:21:57 +0000 (14:21 -0400)]
include signal.h to avoid thorny __sigcontext/sigcontext issues

this is explicitly allowed by POSIX

11 years agofix minor bugs due to incorrect threaded-predicate semantics
Rich Felker [Thu, 21 Apr 2011 01:41:45 +0000 (21:41 -0400)]
fix minor bugs due to incorrect threaded-predicate semantics

some functions that should have been testing whether pthread_self()
had been called and initialized the thread pointer were instead
testing whether pthread_create() had been called and actually made the
program "threaded". while it's unlikely any mismatch would occur in
real-world problems, this could have introduced subtle bugs. now, we
store the address of the main thread's thread descriptor in the libc
structure and use its presence as a flag that the thread register is
initialized. note that after fork, the calling thread (not necessarily
the original main thread) is the new main thread.

11 years agoworkaround bug in linux dup2
Rich Felker [Thu, 21 Apr 2011 01:05:10 +0000 (21:05 -0400)]
workaround bug in linux dup2

the linux documentation for dup2 says it can fail with EBUSY due to a
race condition with open and dup in the kernel. shield applications
(and the rest of libc) from this nonsense by looping until it succeeds

11 years agoproperly create new session/controlling terminal in forkpty
Rich Felker [Thu, 21 Apr 2011 01:01:42 +0000 (21:01 -0400)]
properly create new session/controlling terminal in forkpty

11 years agoimplement (nonstandard) forkpty
Rich Felker [Thu, 21 Apr 2011 00:55:13 +0000 (20:55 -0400)]
implement (nonstandard) forkpty

11 years agodisallow blank strings as service or host name
Rich Felker [Thu, 21 Apr 2011 00:02:35 +0000 (20:02 -0400)]
disallow blank strings as service or host name

11 years agofix bugs in ipv4 parsing
Rich Felker [Thu, 21 Apr 2011 00:00:59 +0000 (20:00 -0400)]
fix bugs in ipv4 parsing

11 years agofix initgroups (uninitialized count passed to getgrouplist)
Rich Felker [Wed, 20 Apr 2011 23:15:27 +0000 (19:15 -0400)]
fix initgroups (uninitialized count passed to getgrouplist)

11 years agoshadow password fixes: empty fields should read as -1 not 0
Rich Felker [Wed, 20 Apr 2011 23:05:59 +0000 (19:05 -0400)]
shadow password fixes: empty fields should read as -1 not 0

11 years agonamespace fixes for sys/mman.h
Rich Felker [Wed, 20 Apr 2011 19:55:58 +0000 (15:55 -0400)]
namespace fixes for sys/mman.h

11 years agofix missing include in posix_madvise.c (compile error)
Rich Felker [Wed, 20 Apr 2011 19:36:15 +0000 (15:36 -0400)]
fix missing include in posix_madvise.c (compile error)

11 years agosupport posix_madvise (previous a stub)
Rich Felker [Wed, 20 Apr 2011 19:25:28 +0000 (15:25 -0400)]
support posix_madvise (previous a stub)

the check against MADV_DONTNEED to because linux MADV_DONTNEED
semantics conflict dangerously with the POSIX semantics

11 years agoadd syscall wrappers for posix_fadvise, posix_fallocate
Rich Felker [Wed, 20 Apr 2011 19:20:22 +0000 (15:20 -0400)]
add syscall wrappers for posix_fadvise, posix_fallocate

11 years agoblock cancellation in wordexp, handle more errors
Rich Felker [Wed, 20 Apr 2011 03:42:56 +0000 (23:42 -0400)]
block cancellation in wordexp, handle more errors

11 years agoavoid malloc of potentially-large string in wordexp
Rich Felker [Wed, 20 Apr 2011 03:37:57 +0000 (23:37 -0400)]
avoid malloc of potentially-large string in wordexp

11 years agomove some more code out of pthread_create.c
Rich Felker [Wed, 20 Apr 2011 03:09:14 +0000 (23:09 -0400)]
move some more code out of pthread_create.c

this also de-uglifies the dummy function aliasing a bit.

11 years agofix uninitialized waiters field in semaphores
Rich Felker [Tue, 19 Apr 2011 17:16:59 +0000 (13:16 -0400)]
fix uninitialized waiters field in semaphores

11 years agodns lookups: protect against cancellation and fix incorrect error codes
Rich Felker [Tue, 19 Apr 2011 01:35:14 +0000 (21:35 -0400)]
dns lookups: protect against cancellation and fix incorrect error codes

11 years agoavoid fd leak if opendir is cancelled when calloc has failed
Rich Felker [Tue, 19 Apr 2011 01:22:14 +0000 (21:22 -0400)]
avoid fd leak if opendir is cancelled when calloc has failed

11 years agoprotect ftw and nftw against cancellation
Rich Felker [Tue, 19 Apr 2011 01:17:03 +0000 (21:17 -0400)]
protect ftw and nftw against cancellation

11 years agoprotect syslog against cancellation
Rich Felker [Tue, 19 Apr 2011 01:11:23 +0000 (21:11 -0400)]
protect syslog against cancellation

these functions are allowed to be cancellation points, but then we
would have to install cleanup handlers to avoid termination with locks
held.

11 years agorecheck cancellation disabled flag after syscall returns EINTR
Rich Felker [Tue, 19 Apr 2011 00:50:37 +0000 (20:50 -0400)]
recheck cancellation disabled flag after syscall returns EINTR

we already checked before making the syscall, but it's possible that a
signal handler interrupted the blocking syscall and disabled
cancellation, and that this is the cause of EINTR. in this case, the
old behavior was testably wrong.

11 years agoremove bogus extra logic for close cancellability
Rich Felker [Mon, 18 Apr 2011 22:42:34 +0000 (18:42 -0400)]
remove bogus extra logic for close cancellability

like all other syscalls, close should return to the caller if and only
if it successfully performed its action. it is necessary that the
application be able to determine whether the close succeeded.

11 years agofix typo in x86_64 cancellable syscall asm
Rich Felker [Sun, 17 Apr 2011 23:25:17 +0000 (19:25 -0400)]
fix typo in x86_64 cancellable syscall asm

11 years agominimal realpath implementation using /proc
Rich Felker [Sun, 17 Apr 2011 21:32:36 +0000 (17:32 -0400)]
minimal realpath implementation using /proc

clean and simple, but fails when the caller does not have permissions
to open the file for reading or when /proc is not available. i may
replace this with a full implementation later, possibly leaving this
version as an optimization to use when it works.

11 years agopthread_exit is not supposed to affect cancellability
Rich Felker [Sun, 17 Apr 2011 21:09:41 +0000 (17:09 -0400)]
pthread_exit is not supposed to affect cancellability

if the exit was caused by cancellation, __cancel has already set these
flags anyway.

11 years agofix pthread_exit from cancellation handler
Rich Felker [Sun, 17 Apr 2011 21:06:05 +0000 (17:06 -0400)]
fix pthread_exit from cancellation handler

cancellation frames were not correctly popped, so this usage would not
only loop, but also reuse discarded and invalid parts of the stack.

11 years agoclean up handling of thread/nothread mode, locking
Rich Felker [Sun, 17 Apr 2011 20:53:54 +0000 (16:53 -0400)]
clean up handling of thread/nothread mode, locking

11 years agodebloat: use __syscall instead of syscall where possible
Rich Felker [Sun, 17 Apr 2011 20:32:15 +0000 (16:32 -0400)]
debloat: use __syscall instead of syscall where possible

don't waste time (and significant code size due to function call
overhead!) setting errno when the result of a syscall does not matter
or when it can't fail.

11 years agofix bugs in cancellable syscall asm
Rich Felker [Sun, 17 Apr 2011 19:30:08 +0000 (15:30 -0400)]
fix bugs in cancellable syscall asm

x86_64 was just plain wrong in the cancel-flag-already-set path, and
crashing.

the more subtle error was not clearing the saved stack pointer before
returning to c code. this could result in the signal handler
misidentifying c code as the pre-syscall part of the asm, and acting
on cancellation at the wrong time, and thus resource leak race
conditions.

also, now __cancel (in the c code) is responsible for clearing the
saved sp in the already-cancelled branch. this means we have to use
call rather than jmp to ensure the stack pointer in the c will never
match what the asm saved.

11 years agooptimize cancellation enable/disable code
Rich Felker [Sun, 17 Apr 2011 17:21:13 +0000 (13:21 -0400)]
optimize cancellation enable/disable code

the goal is to be able to use pthread_setcancelstate internally in
the implementation, whenever a function might want to use functions
which are cancellation points but avoid becoming a cancellation point
itself. i could have just used a separate internal function for
temporarily inhibiting cancellation, but the solution in this commit
is better because (1) it's one less implementation-specific detail in
functions that need to use it, and (2) application code can also get
the same benefit.

previously, pthread_setcancelstate dependend on pthread_self, which
would pull in unwanted thread setup overhead for non-threaded
programs. now, it temporarily stores the state in the global libc
struct if threads have not been initialized, and later moves it if
needed. this way we can instead use __pthread_self, which has no
dependencies and assumes that the thread register is already valid.

11 years agodon't use pthread_once when there is no danger in race
Rich Felker [Sun, 17 Apr 2011 16:15:55 +0000 (12:15 -0400)]
don't use pthread_once when there is no danger in race

11 years agofix some minor issues in cancellation handling patch
Rich Felker [Sun, 17 Apr 2011 16:09:47 +0000 (12:09 -0400)]
fix some minor issues in cancellation handling patch

signals were wrongly left masked, and cancellability state was not
switched to disabled, during the execution of cleanup handlers.

11 years agooverhaul pthread cancellation
Rich Felker [Sun, 17 Apr 2011 15:43:03 +0000 (11:43 -0400)]
overhaul pthread cancellation

this patch improves the correctness, simplicity, and size of
cancellation-related code. modulo any small errors, it should now be
completely conformant, safe, and resource-leak free.

the notion of entering and exiting cancellation-point context has been
completely eliminated and replaced with alternative syscall assembly
code for cancellable syscalls. the assembly is responsible for setting
up execution context information (stack pointer and address of the
syscall instruction) which the cancellation signal handler can use to
determine whether the interrupted code was in a cancellable state.

these changes eliminate race conditions in the previous generation of
cancellation handling code (whereby a cancellation request received
just prior to the syscall would not be processed, leaving the syscall
to block, potentially indefinitely), and remedy an issue where
non-cancellable syscalls made from signal handlers became cancellable
if the signal handler interrupted a cancellation point.

x86_64 asm is untested and may need a second try to get it right.

11 years agoremove stupid debug code in wordexp
Rich Felker [Fri, 15 Apr 2011 16:07:26 +0000 (12:07 -0400)]
remove stupid debug code in wordexp

11 years agoimplement wordexp. first try, may be buggy. intended to be safe.
Rich Felker [Fri, 15 Apr 2011 16:06:34 +0000 (12:06 -0400)]
implement wordexp. first try, may be buggy. intended to be safe.

11 years agoavoid setting errno when checking for tty
Rich Felker [Fri, 15 Apr 2011 16:04:13 +0000 (12:04 -0400)]
avoid setting errno when checking for tty

setting errno here is completely valid, but some programs, notably
busybox printf, assume that errno will not be set during output and
treat this as an error condition. in any case, skipping it slightly
reduces code size and saves time.

11 years agodocument some changes in the upcoming release
Rich Felker [Fri, 15 Apr 2011 03:33:46 +0000 (23:33 -0400)]
document some changes in the upcoming release

11 years agofix O_SYNC definition, cleanup fcntl.h
Rich Felker [Fri, 15 Apr 2011 02:06:30 +0000 (22:06 -0400)]
fix O_SYNC definition, cleanup fcntl.h

11 years agofix FAPPEND typo on x86_64 (previously only fixed on i386)
Rich Felker [Fri, 15 Apr 2011 01:50:07 +0000 (21:50 -0400)]
fix FAPPEND typo on x86_64 (previously only fixed on i386)

11 years agofcntl.h: move macros that do not vary between archs out of bits
Rich Felker [Fri, 15 Apr 2011 01:49:22 +0000 (21:49 -0400)]
fcntl.h: move macros that do not vary between archs out of bits