On Jan 7, 2006, at 11:49 AM, Ulrich Drepper wrote:
Robust futexes will have -EDEADLK returned to them since there is no
POSIX specification for
robust mutexes, yet, and returning -EDEADLK is more in the spirit of
robustness.
I disagree. Robustness is only an additional mutex attribute on top of
the mutex type. I.e., there can be robust
normal/error-checking/recursive mutexes. What kind it is can be told to
the kernel at the time the robust mutex is registered with the kernel.
You are correct. And I believe the kernel will have to have
these attributes reflected on the rt_mutex as well. (Guess I better
get started on a new patch).
This is the way I'll write up the proposal for robust mutexes for
inclusion in the next POSIX revision.
I'm hoping, however, that the POSIX spec will not state that the app must
hang if it deadlocks itself on a robust mutex. I prefer that the kernel, when it can, return
-EDEADLK, so the app can at least try and recover. That was the original
goal of using DEBUG_DEADLOCKS, so apps could be debugged and corrected
without just hanging.
David
--
➧ Ulrich Drepper ➧ Red Hat, Inc. ➧ 444 Castro St ➧ Mountain View, CA ❖
_______________________________________________ robustmutexes mailing list [email protected] https://lists.osdl.org/mailman/listinfo/robustmutexes
