diff options
| author | Adhemerval Zanella <adhemerval.zanella@linaro.org> | 2024-08-05 11:27:35 -0300 |
|---|---|---|
| committer | Adhemerval Zanella <adhemerval.zanella@linaro.org> | 2024-08-05 17:07:57 -0300 |
| commit | c6af8a9a3ce137a9704825d173be22a2b2d9cb49 (patch) | |
| tree | daa57dfeb5ae6e1d1c61b7b414400b3bb28d3546 /stdlib/exit.c | |
| parent | 5097cd344fd243fb8deb6dec96e8073753f962f9 (diff) | |
| download | glibc-c6af8a9a3ce137a9704825d173be22a2b2d9cb49.tar.xz glibc-c6af8a9a3ce137a9704825d173be22a2b2d9cb49.zip | |
stdlib: Allow concurrent quick_exit (BZ 31997)
As for exit, also allows concurrent quick_exit to avoid race
conditions when it is called concurrently. Since it uses the same
internal function as exit, the __exit_lock lock is moved to
__run_exit_handlers. It also solved a potential concurrent when
calling exit and quick_exit concurrently.
The test case 'expected' is expanded to a value larger than the
minimum required by C/POSIX (32 entries) so at_quick_exit() will
require libc to allocate a new block. This makes the test mre likely to
trigger concurrent issues (through free() at __run_exit_handlers)
if quick_exit() interacts with the at_quick_exit list concurrently.
This is also the latest interpretation of the Austin Ticket [1].
Checked on x86_64-linux-gnu.
[1] https://austingroupbugs.net/view.php?id=1845
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Diffstat (limited to 'stdlib/exit.c')
| -rw-r--r-- | stdlib/exit.c | 18 |
1 files changed, 10 insertions, 8 deletions
diff --git a/stdlib/exit.c b/stdlib/exit.c index bbaf138806..8d7e2e53d0 100644 --- a/stdlib/exit.c +++ b/stdlib/exit.c @@ -28,6 +28,13 @@ __exit_funcs_lock is declared. */ bool __exit_funcs_done = false; +/* The lock handles concurrent exit() and quick_exit(), even though the + C/POSIX standard states that calling exit() more than once is UB. The + recursive lock allows atexit() handlers or destructors to call exit() + itself. In this case, the handler list execution will resume at the + point of the current handler. */ +__libc_lock_define_initialized_recursive (static, __exit_lock) + /* Call all functions registered with `atexit' and `on_exit', in the reverse of the order in which they were registered perform stdio cleanup, and terminate program execution with STATUS. */ @@ -36,6 +43,9 @@ attribute_hidden __run_exit_handlers (int status, struct exit_function_list **listp, bool run_list_atexit, bool run_dtors) { + /* The exit should never return, so there is no need to unlock it. */ + __libc_lock_lock_recursive (__exit_lock); + /* First, call the TLS destructors. */ if (run_dtors) call_function_static_weak (__call_tls_dtors); @@ -132,17 +142,9 @@ __run_exit_handlers (int status, struct exit_function_list **listp, } -/* The lock handles concurrent exit(), even though the C/POSIX standard states - that calling exit() more than once is UB. The recursive lock allows - atexit() handlers or destructors to call exit() itself. In this case, the - handler list execution will resume at the point of the current handler. */ -__libc_lock_define_initialized_recursive (static, __exit_lock) - void exit (int status) { - /* The exit should never return, so there is no need to unlock it. */ - __libc_lock_lock_recursive (__exit_lock); __run_exit_handlers (status, &__exit_funcs, true, true); } libc_hidden_def (exit) |
