diff options
| author | Darren Hart <dvhart@linux.intel.com> | 2015-12-18 22:36:37 +0100 |
|---|---|---|
| committer | Thomas Gleixner <tglx@linutronix.de> | 2015-12-20 12:43:25 +0100 |
| commit | 337f13046ff03717a9e99675284a817527440a49 (patch) | |
| tree | f60d2ac3de33d48447601c1bae638091864de0bb /kernel/locking | |
| parent | futex: Cleanup the goto confusion in requeue_pi() (diff) | |
| download | linux-337f13046ff03717a9e99675284a817527440a49.tar.xz linux-337f13046ff03717a9e99675284a817527440a49.zip | |
futex: Allow FUTEX_CLOCK_REALTIME with FUTEX_WAIT op
While reviewing Michael Kerrisk's recent futex manpage update, I noticed
that we allow the FUTEX_CLOCK_REALTIME flag for FUTEX_WAIT_BITSET but
not for FUTEX_WAIT.
FUTEX_WAIT is treated as a simple version for FUTEX_WAIT_BITSET
internally (with a bitmask of FUTEX_BITSET_MATCH_ANY). As such, I cannot
come up with a reason for this exclusion for FUTEX_WAIT.
This change does modify the behavior of the futex syscall, changing a
call with FUTEX_WAIT | FUTEX_CLOCK_REALTIME from returning -ENOSYS, to be
equivalent to FUTEX_WAIT_BITSET | FUTEX_CLOCK_REALTIME with a bitset of
FUTEX_BITSET_MATCH_ANY.
Reported-by: Michael Kerrisk <mtk.manpages@gmail.com>
Signed-off-by: Darren Hart <dvhart@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Link: http://lkml.kernel.org/r/9f3bdc116d79d23f5ee72ceb9a2a857f5ff8fa29.1450474525.git.dvhart@linux.intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Diffstat (limited to 'kernel/locking')
0 files changed, 0 insertions, 0 deletions