staging: lustre: race condition for check/use cfs_fail_val
authorFan Yong <fan.yong@intel.com>
Wed, 4 Nov 2015 18:40:04 +0000 (13:40 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Mon, 16 Nov 2015 04:02:47 +0000 (20:02 -0800)
commit78368d578e9eab8dc9d42b9e9a2be78a98f4e9d9
treead585a2fba1d691f3419afe92909b69dd0661f05
parenta0455471582117d31659618c02146804df527ff4
staging: lustre: race condition for check/use cfs_fail_val

There are some race conditions when check/use cfs_fail_val.
For example: when inject failure stub for LFSCK test as following:

764   if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY2) &&
765       cfs_fail_val > 0) {
766           struct l_wait_info lwi;
767
768           lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
769                             NULL, NULL);
770           l_wait_event(thread->t_ctl_waitq,
771                        !thread_is_running(thread),
772                        &lwi);
773
774           if (unlikely(!thread_is_running(thread))) {
775                   CDEBUG(D_LFSCK, "%s: scan dir exit for engine "
776                          "stop, parent "DFID", cookie "LPX64"n",
777                          lfsck_lfsck2name(lfsck),
778                          PFID(lfsck_dto2fid(dir)),
779                          lfsck->li_cookie_dir);
780                   RETURN(0);
781           }
782   }

The "cfs_fail_val" may be changed as zero by others after the check
at the line 765 but before using it at the line 768. Then the LFSCK
engine will fall into "wait" until someone run "lfsck_stop".

Signed-off-by: Fan Yong <fan.yong@intel.com>
Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-6146
Reviewed-on: http://review.whamcloud.com/13481
Reviewed-by: Lai Siyao <lai.siyao@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/staging/lustre/lustre/libcfs/fail.c