[Bug libstdc++/104442] atomic::wait incorrectly loops in case of spurious notification when __waiter is shared
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104442 --- Comment #9 from Marc Poulhiès --- FWIW, I confirm that the vxworks issue is fixed in latest gcc-11 branch ! Thanks
[Bug libstdc++/104442] atomic::wait incorrectly loops in case of spurious notification when __waiter is shared
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104442 Thomas Rodgers changed: What|Removed |Added Status|ASSIGNED|RESOLVED Resolution|--- |FIXED --- Comment #8 from Thomas Rodgers --- (In reply to Marc Poulhiès from comment #7) > Thanks ! Since the patch committed is effectively the same as your proposed patch, I am going to assume this resolves the issue on vxworks, but I have no way to independently confirm that.
[Bug libstdc++/104442] atomic::wait incorrectly loops in case of spurious notification when __waiter is shared
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104442 --- Comment #7 from Marc Poulhiès --- Thanks !
[Bug libstdc++/104442] atomic::wait incorrectly loops in case of spurious notification when __waiter is shared
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104442 --- Comment #6 from CVS Commits --- The releases/gcc-11 branch has been updated by Thomas Rodgers : https://gcc.gnu.org/g:5669a60e2fcc90afaa16fe20391cdcf18ab398d7 commit r11-9547-g5669a60e2fcc90afaa16fe20391cdcf18ab398d7 Author: Thomas Rodgers Date: Wed Feb 9 12:29:19 2022 -0800 libstdc++: Fix deadlock in atomic wait [PR104442] This issue was observed as a deadlock in 29_atomics/atomic/wait_notify/100334.cc on vxworks. When a wait is "laundered" (e.g. type T* does not suffice as a waitable address for the platform's native waiting primitive), the address waited is that of the _M_ver member of __waiter_pool_base, so several threads may wait on the same address for unrelated atomic objects. As noted in the PR, the implementation correctly exits the wait for the thread whose data changed, but not for any other threads waiting on the same address. As noted in the PR the __waiter::_M_do_wait_v member was correctly exiting but the other waiters were not reloading the value of _M_ver before re-entering the wait. Moving the spin call inside the loop accomplishes this, and is consistent with the predicate accepting version of __waiter::_M_do_wait. libstdc++-v3/ChangeLog: PR libstdc++/104442 * include/bits/atomic_wait.h (__waiter::_M_do_wait_v): Move spin loop inside do loop so that threads failing the wait, reload _M_ver. (cherry picked from commit 4cf3c339815cdfa636b25a512f91b63d7c313fd6)
[Bug libstdc++/104442] atomic::wait incorrectly loops in case of spurious notification when __waiter is shared
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104442 --- Comment #5 from CVS Commits --- The master branch has been updated by Thomas Rodgers : https://gcc.gnu.org/g:4cf3c339815cdfa636b25a512f91b63d7c313fd6 commit r12-7151-g4cf3c339815cdfa636b25a512f91b63d7c313fd6 Author: Thomas Rodgers Date: Wed Feb 9 12:29:19 2022 -0800 libstdc++: Fix deadlock in atomic wait [PR104442] This issue was observed as a deadlock in 29_atomics/atomic/wait_notify/100334.cc on vxworks. When a wait is "laundered" (e.g. type T* does not suffice as a waitable address for the platform's native waiting primitive), the address waited is that of the _M_ver member of __waiter_pool_base, so several threads may wait on the same address for unrelated atomic objects. As noted in the PR, the implementation correctly exits the wait for the thread whose data changed, but not for any other threads waiting on the same address. As noted in the PR the __waiter::_M_do_wait_v member was correctly exiting but the other waiters were not reloading the value of _M_ver before re-entering the wait. Moving the spin call inside the loop accomplishes this, and is consistent with the predicate accepting version of __waiter::_M_do_wait. libstdc++-v3/ChangeLog: PR libstdc++/104442 * include/bits/atomic_wait.h (__waiter::_M_do_wait_v): Move spin loop inside do loop so that threads failing the wait, reload _M_ver.
[Bug libstdc++/104442] atomic::wait incorrectly loops in case of spurious notification when __waiter is shared
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104442 --- Comment #4 from Thomas Rodgers --- (In reply to Jonathan Wakely from comment #3) > Tom submitted > https://gcc.gnu.org/pipermail/gcc-patches/2022-February/590050.html already, > which is slightly different. The primary difference was changing the memory order of the load in _M_do_spin. I may revert that part of the change and submit it separately, as I've noticed one other case where I'd like to change it from RELAXED to SEQ_CST.
[Bug libstdc++/104442] atomic::wait incorrectly loops in case of spurious notification when __waiter is shared
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104442 --- Comment #3 from Jonathan Wakely --- Tom submitted https://gcc.gnu.org/pipermail/gcc-patches/2022-February/590050.html already, which is slightly different.
[Bug libstdc++/104442] atomic::wait incorrectly loops in case of spurious notification when __waiter is shared
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104442 --- Comment #2 from Marc Poulhiès --- Ok thanks for looking into this. I'll submit the patch on the list then !
[Bug libstdc++/104442] atomic::wait incorrectly loops in case of spurious notification when __waiter is shared
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104442 Thomas Rodgers changed: What|Removed |Added Ever confirmed|0 |1 Last reconfirmed||2022-02-09 CC||rodgertq at gcc dot gnu.org Status|UNCONFIRMED |ASSIGNED --- Comment #1 from Thomas Rodgers --- I took a look at whether or not it made sense to do a different refactoring with this version of the implementation and I don't believe that it does (the implementation detail here is likely to change substantially when we commit to an ABI stable version) and indeed, the predicate version does exactly what this patch proposes.
