This is a new implementation for condition variables, required
after http://austingroupbugs.net/view.php?id=609 to fix bug 13165. In
essence, we need to be stricter in which waiters a signal or broadcast
is required to wake up; this couldn't be solved using the old algorithm.
ISO C++ made a similar clarification, so this also fixes a bug in
current libstdc++, for example.
We can't use the old algorithm anymore because futexes do not guarantee
to wake in FIFO order. Thus, when we wake, we can't simply let any
waiter grab a signal, but we need to ensure that one of the waiters
happening before the signal is woken up. This is something the previous
algorithm violated (see bug 13165).
There's another issue specific to condvars: ABA issues on the underlying
futexes. Unlike mutexes that have just three states, or semaphores that
have no tokens or a limited number of them, the state of a condvar is
the *order* of the waiters. A waiter on a semaphore can grab a token
whenever one is available; a condvar waiter must only consume a signal
if it is eligible to do so as determined by the relative order of the
waiter and the signal.
Therefore, this new algorithm maintains two groups of waiters: Those
eligible to consume signals (G1), and those that have to wait until
previous waiters have consumed signals (G2). Once G1 is empty, G2
becomes the new G1. 64b counters are used to avoid ABA issues.
This condvar doesn't yet use a requeue optimization (ie, on a broadcast,
waking just one thread and requeueing all others on the futex of the
mutex supplied by the program). I don't think doing the requeue is
necessarily the right approach (but I haven't done real measurements
yet):
* If a program expects to wake many threads at the same time and make
that scalable, a condvar isn't great anyway because of how it requires
waiters to operate mutually exclusive (due to the mutex usage). Thus, a
thundering herd problem is a scalability problem with or without the
optimization. Using something like a semaphore might be more
appropriate in such a case.
* The scalability problem is actually at the mutex side; the condvar
could help (and it tries to with the requeue optimization), but it
should be the mutex who decides how that is done, and whether it is done
at all.
* Forcing all but one waiter into the kernel-side wait queue of the
mutex prevents/avoids the use of lock elision on the mutex. Thus, it
prevents the only cure against the underlying scalability problem
inherent to condvars.
* If condvars use short critical sections (ie, hold the mutex just to
check a binary flag or such), which they should do ideally, then forcing
all those waiter to proceed serially with kernel-based hand-off (ie,
futex ops in the mutex' contended state, via the futex wait queues) will
be less efficient than just letting a scalable mutex implementation take
care of it. Our current mutex impl doesn't employ spinning at all, but
if critical sections are short, spinning can be much better.
* Doing the requeue stuff requires all waiters to always drive the mutex
into the contended state. This leads to each waiter having to call
futex_wake after lock release, even if this wouldn't be necessary.
[BZ #13165]
* nptl/pthread_cond_broadcast.c (__pthread_cond_broadcast): Rewrite to
use new algorithm.
* nptl/pthread_cond_destroy.c (__pthread_cond_destroy): Likewise.
* nptl/pthread_cond_init.c (__pthread_cond_init): Likewise.
* nptl/pthread_cond_signal.c (__pthread_cond_signal): Likewise.
* nptl/pthread_cond_wait.c (__pthread_cond_wait): Likewise.
(__pthread_cond_timedwait): Move here from pthread_cond_timedwait.c.
(__condvar_confirm_wakeup, __condvar_cancel_waiting,
__condvar_cleanup_waiting, __condvar_dec_grefs,
__pthread_cond_wait_common): New.
(__condvar_cleanup): Remove.
* npt/pthread_condattr_getclock.c (pthread_condattr_getclock): Adapt.
* npt/pthread_condattr_setclock.c (pthread_condattr_setclock):
Likewise.
* npt/pthread_condattr_getpshared.c (pthread_condattr_getpshared):
Likewise.
* npt/pthread_condattr_init.c (pthread_condattr_init): Likewise.
* nptl/tst-cond1.c: Add comment.
* nptl/tst-cond20.c (do_test): Adapt.
* nptl/tst-cond22.c (do_test): Likewise.
* sysdeps/aarch64/nptl/bits/pthreadtypes.h (pthread_cond_t): Adapt
structure.
* sysdeps/arm/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise.
* sysdeps/ia64/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise.
* sysdeps/m68k/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise.
* sysdeps/microblaze/nptl/bits/pthreadtypes.h (pthread_cond_t):
Likewise.
* sysdeps/mips/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise.
* sysdeps/nios2/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise.
* sysdeps/s390/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise.
* sysdeps/sh/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise.
* sysdeps/tile/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise.
* sysdeps/unix/sysv/linux/alpha/bits/pthreadtypes.h (pthread_cond_t):
Likewise.
* sysdeps/unix/sysv/linux/powerpc/bits/pthreadtypes.h (pthread_cond_t):
Likewise.
* sysdeps/x86/bits/pthreadtypes.h (pthread_cond_t): Likewise.
* sysdeps/nptl/internaltypes.h (COND_NWAITERS_SHIFT): Remove.
(COND_CLOCK_BITS): Adapt.
* sysdeps/nptl/pthread.h (PTHREAD_COND_INITIALIZER): Adapt.
* nptl/pthreadP.h (__PTHREAD_COND_CLOCK_MONOTONIC_MASK,
__PTHREAD_COND_SHARED_MASK): New.
* nptl/nptl-printers.py (CLOCK_IDS): Remove.
(ConditionVariablePrinter, ConditionVariableAttributesPrinter): Adapt.
* nptl/nptl_lock_constants.pysym: Adapt.
* nptl/test-cond-printers.py: Adapt.
* sysdeps/unix/sysv/linux/hppa/internaltypes.h (cond_compat_clear,
cond_compat_check_and_clear): Adapt.
* sysdeps/unix/sysv/linux/hppa/pthread_cond_timedwait.c: Remove file ...
* sysdeps/unix/sysv/linux/hppa/pthread_cond_wait.c
(__pthread_cond_timedwait): ... and move here.
* nptl/DESIGN-condvar.txt: Remove file.
* nptl/lowlevelcond.sym: Likewise.
* nptl/pthread_cond_timedwait.c: Likewise.
* sysdeps/unix/sysv/linux/i386/i486/pthread_cond_broadcast.S: Likewise.
* sysdeps/unix/sysv/linux/i386/i486/pthread_cond_signal.S: Likewise.
* sysdeps/unix/sysv/linux/i386/i486/pthread_cond_timedwait.S: Likewise.
* sysdeps/unix/sysv/linux/i386/i486/pthread_cond_wait.S: Likewise.
* sysdeps/unix/sysv/linux/i386/i586/pthread_cond_broadcast.S: Likewise.
* sysdeps/unix/sysv/linux/i386/i586/pthread_cond_signal.S: Likewise.
* sysdeps/unix/sysv/linux/i386/i586/pthread_cond_timedwait.S: Likewise.
* sysdeps/unix/sysv/linux/i386/i586/pthread_cond_wait.S: Likewise.
* sysdeps/unix/sysv/linux/i386/i686/pthread_cond_broadcast.S: Likewise.
* sysdeps/unix/sysv/linux/i386/i686/pthread_cond_signal.S: Likewise.
* sysdeps/unix/sysv/linux/i386/i686/pthread_cond_timedwait.S: Likewise.
* sysdeps/unix/sysv/linux/i386/i686/pthread_cond_wait.S: Likewise.
* sysdeps/unix/sysv/linux/x86_64/pthread_cond_broadcast.S: Likewise.
* sysdeps/unix/sysv/linux/x86_64/pthread_cond_signal.S: Likewise.
* sysdeps/unix/sysv/linux/x86_64/pthread_cond_timedwait.S: Likewise.
* sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S: Likewise.
This patch remove the PID cache and usage in current GLIBC code. Current
usage is mainly used a performance optimization to avoid the syscall,
however it adds some issues:
- The exposed clone syscall will try to set pid/tid to make the new
thread somewhat compatible with current GLIBC assumptions. This cause
a set of issue with new workloads and usecases (such as BZ#17214 and
[1]) as well for new internal usage of clone to optimize other algorithms
(such as clone plus CLONE_VM for posix_spawn, BZ#19957).
- The caching complexity also added some bugs in the past [2] [3] and
requires more effort of each port to handle such requirements (for
both clone and vfork implementation).
- Caching performance gain in mainly on getpid and some specific
code paths. The getpid performance leverage is questionable [4],
either by the idea of getpid being a hotspot as for the getpid
implementation itself (if it is indeed a justifiable hotspot a
vDSO symbol could let to a much more simpler solution).
Other usage is mainly for non usual code paths, such as pthread
cancellation signal and handling.
For thread creation (on stack allocation) the code simplification in fact
adds some performance gain due the no need of transverse the stack cache
and invalidate each element pid.
Other thread usages will require a direct getpid syscall, such as
cancellation/setxid signal, thread cancellation, thread fail path (at
create_thread), and thread signal (pthread_kill and pthread_sigqueue).
However these are hardly usual hotspots and I think adding a syscall is
justifiable.
It also simplifies both the clone and vfork arch-specific implementation.
And by review each fork implementation there are some discrepancies that
this patch also solves:
- microblaze clone/vfork does not set/reset the pid/tid field
- hppa uses the default vfork implementation that fallback to fork.
Since vfork is deprecated I do not think we should bother with it.
The patch also removes the TID caching in clone. My understanding for
such semantic is try provide some pthread usage after a user program
issue clone directly (as done by thread creation with CLONE_PARENT_SETTID
and pthread tid member). However, as stated before in multiple discussions
threads, GLIBC provides clone syscalls without further supporting all this
semantics.
I ran a full make check on x86_64, x32, i686, armhf, aarch64, and powerpc64le.
For sparc32, sparc64, and mips I ran the basic fork and vfork tests from
posix/ folder (on a qemu system). So it would require further testing
on alpha, hppa, ia64, m68k, nios2, s390, sh, and tile (I excluded microblaze
because it is already implementing the patch semantic regarding clone/vfork).
[1] https://codereview.chromium.org/800183004/
[2] https://sourceware.org/ml/libc-alpha/2006-07/msg00123.html
[3] https://sourceware.org/bugzilla/show_bug.cgi?id=15368
[4] http://yarchive.net/comp/linux/getpid_caching.html
* sysdeps/nptl/fork.c (__libc_fork): Remove pid cache setting.
* nptl/allocatestack.c (allocate_stack): Likewise.
(__reclaim_stacks): Likewise.
(setxid_signal_thread): Obtain pid through syscall.
* nptl/nptl-init.c (sigcancel_handler): Likewise.
(sighandle_setxid): Likewise.
* nptl/pthread_cancel.c (pthread_cancel): Likewise.
* sysdeps/unix/sysv/linux/pthread_kill.c (__pthread_kill): Likewise.
* sysdeps/unix/sysv/linux/pthread_sigqueue.c (pthread_sigqueue):
Likewise.
* sysdeps/unix/sysv/linux/createthread.c (create_thread): Likewise.
* sysdeps/unix/sysv/linux/getpid.c: Remove file.
* nptl/descr.h (struct pthread): Change comment about pid value.
* nptl/pthread_getattr_np.c (pthread_getattr_np): Remove thread
pid assert.
* sysdeps/unix/sysv/linux/pthread-pids.h (__pthread_initialize_pids):
Do not set pid value.
* nptl_db/td_ta_thr_iter.c (iterate_thread_list): Remove thread
pid cache check.
* nptl_db/td_thr_validate.c (td_thr_validate): Likewise.
* sysdeps/aarch64/nptl/tcb-offsets.sym: Remove pid offset.
* sysdeps/alpha/nptl/tcb-offsets.sym: Likewise.
* sysdeps/arm/nptl/tcb-offsets.sym: Likewise.
* sysdeps/hppa/nptl/tcb-offsets.sym: Likewise.
* sysdeps/i386/nptl/tcb-offsets.sym: Likewise.
* sysdeps/ia64/nptl/tcb-offsets.sym: Likewise.
* sysdeps/m68k/nptl/tcb-offsets.sym: Likewise.
* sysdeps/microblaze/nptl/tcb-offsets.sym: Likewise.
* sysdeps/mips/nptl/tcb-offsets.sym: Likewise.
* sysdeps/nios2/nptl/tcb-offsets.sym: Likewise.
* sysdeps/powerpc/nptl/tcb-offsets.sym: Likewise.
* sysdeps/s390/nptl/tcb-offsets.sym: Likewise.
* sysdeps/sh/nptl/tcb-offsets.sym: Likewise.
* sysdeps/sparc/nptl/tcb-offsets.sym: Likewise.
* sysdeps/tile/nptl/tcb-offsets.sym: Likewise.
* sysdeps/x86_64/nptl/tcb-offsets.sym: Likewise.
* sysdeps/unix/sysv/linux/aarch64/clone.S: Remove pid and tid caching.
* sysdeps/unix/sysv/linux/alpha/clone.S: Likewise.
* sysdeps/unix/sysv/linux/arm/clone.S: Likewise.
* sysdeps/unix/sysv/linux/hppa/clone.S: Likewise.
* sysdeps/unix/sysv/linux/i386/clone.S: Likewise.
* sysdeps/unix/sysv/linux/ia64/clone2.S: Likewise.
* sysdeps/unix/sysv/linux/mips/clone.S: Likewise.
* sysdeps/unix/sysv/linux/nios2/clone.S: Likewise.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/clone.S: Likewise.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/clone.S: Likewise.
* sysdeps/unix/sysv/linux/s390/s390-32/clone.S: Likewise.
* sysdeps/unix/sysv/linux/s390/s390-64/clone.S: Likewise.
* sysdeps/unix/sysv/linux/sh/clone.S: Likewise.
* sysdeps/unix/sysv/linux/sparc/sparc32/clone.S: Likewise.
* sysdeps/unix/sysv/linux/sparc/sparc64/clone.S: Likewise.
* sysdeps/unix/sysv/linux/tile/clone.S: Likewise.
* sysdeps/unix/sysv/linux/x86_64/clone.S: Likewise.
* sysdeps/unix/sysv/linux/aarch64/vfork.S: Remove pid set and reset.
* sysdeps/unix/sysv/linux/alpha/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/arm/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/i386/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/ia64/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/m68k/clone.S: Likewise.
* sysdeps/unix/sysv/linux/m68k/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/mips/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/nios2/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/s390/s390-32/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/s390/s390-64/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/sh/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/sparc/sparc32/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/sparc/sparc64/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/tile/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/x86_64/vfork.S: Likewise.
* sysdeps/unix/sysv/linux/tst-clone2.c (f): Remove direct pthread
struct access.
(clone_test): Remove function.
(do_test): Rewrite to take in consideration pid is not cached anymore.
This patch relies on the C version of the rwlocks posted earlier.
With C rwlocks it is very straight forward to do adaptive elision
using TSX. It is based on the infrastructure added earlier
for mutexes, but uses its own elision macros. The macros
are fairly general purpose and could be used for other
elision purposes too.
This version is much cleaner than the earlier assembler based
version, and in particular implements adaptation which makes
it safer.
I changed the behavior slightly to not require any changes
in the test suite and fully conform to all expected
behaviors (generally at the cost of not eliding in
various situations). In particular this means the timedlock
variants are not elided. Nested trylock aborts.
I've moved the ARM port from ports to the main sysdeps hierarchy.
Beyond the README update, the move of the files was simply
git mv ports/sysdeps/arm sysdeps/arm
git mv ports/sysdeps/unix/arm sysdeps/unix/arm
git mv ports/sysdeps/unix/sysv/linux/arm sysdeps/unix/sysv/linux/arm
and in addition to the ChangeLog entries here, I put a note at the top
of ports/ChangeLog.arm similar to that at the top of
ChangeLog.powerpc. There is deliberately no NEWS change, as I think
it makes the most sense to put in a general note above all ports
having moved if we can achieve that for 2.20.
Tested that disassembly of installed shared libraries for arm is the
same before and after this patch, except for data (not instructions)
in ld.so (there are assertions in sysdeps/arm/dl-machine.h, and the
path by which that file is found, and so by which it appears in the
assertion message, changes as a result of the move).
* sysdeps/arm: Move directory from ports/sysdeps/arm.
* sysdeps/unix/arm: Move directory from ports/sysdeps/unix/arm.
* sysdeps/unix/sysv/linux/arm: Move directory from
ports/sysdeps/unix/sysv/linux/arm.
* README: Update listing for arm-*-linux-gnueabi.
ports/ChangeLog.arm:
* sysdeps/arm: Move directory to ../sysdeps/arm.
* sysdeps/unix/arm: Move directory to ../sysdeps.arm.
* sysdeps/unix/sysv/linux/arm: Move directory to
../sysdeps/unix/sysv/linux/arm.
