c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2025-02-05 06:43:31 +03:00
Code Issues Packages Projects Releases Wiki Activity

Fix random ztest_deadman_thread failures

Browse Source 
The zloop test has been failing in buildbot for the last few weeks
with various failures in ztest_deadman_thread(). This is due to the
fact that this thread is not stopped when performing pool import /
export tests as it should be. This patch simply corrects this.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Serapheim Dimitropoulos <serapheim.dimitro@delphix.com>
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #8010
This commit is contained in:
Tom Caputi 2018-10-10 16:48:33 -04:00 committed by Brian Behlendorf
parent e871a8f058
commit 9410257800
2 changed files with 26 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

36
cmd/ztest/ztest.c
View File

@ -6495,13 +6495,20 @@ ztest_deadman_thread(void *arg)
{ {
ztest_shared_t *zs = arg; ztest_shared_t *zs = arg;
spa_t *spa = ztest_spa; spa_t *spa = ztest_spa;
hrtime_t delta, overdue, total = 0; hrtime_t delay, overdue, last_run = gethrtime();
for (;;) { delay = (zs->zs_thread_stop - zs->zs_thread_start) +
delta = zs->zs_thread_stop - zs->zs_thread_start + MSEC2NSEC(zfs_deadman_synctime_ms);
MSEC2NSEC(zfs_deadman_synctime_ms);
(void) poll(NULL, 0, (int)NSEC2MSEC(delta)); while (!ztest_exiting) {
/*
* Wait for the delay timer while checking occasionally
* if we should stop.
*/
if (gethrtime() < last_run + delay) {
(void) poll(NULL, 0, 1000);
continue;
}
/* /*
* If the pool is suspended then fail immediately. Otherwise, * If the pool is suspended then fail immediately. Otherwise,
@ -6522,15 +6529,20 @@ ztest_deadman_thread(void *arg)
* then it may be hung and is terminated. * then it may be hung and is terminated.
*/ */
overdue = zs->zs_proc_stop + MSEC2NSEC(zfs_deadman_synctime_ms); overdue = zs->zs_proc_stop + MSEC2NSEC(zfs_deadman_synctime_ms);
total += zfs_deadman_synctime_ms / 1000;
if (gethrtime() > overdue) { if (gethrtime() > overdue) {
fatal(0, "aborting test after %llu seconds because " fatal(0, "aborting test after %llu seconds because "
"the process is overdue for termination.", total); "the process is overdue for termination.",
(gethrtime() - zs->zs_proc_start) / NANOSEC);
} }
(void) printf("ztest has been running for %lld seconds\n", (void) printf("ztest has been running for %lld seconds\n",
total); (gethrtime() - zs->zs_proc_start) / NANOSEC);
last_run = gethrtime();
delay = MSEC2NSEC(zfs_deadman_checktime_ms);
} }
thread_exit();
} }
static void static void
@ -6724,7 +6736,7 @@ ztest_run(ztest_shared_t *zs)
{ {
spa_t *spa; spa_t *spa;
objset_t *os; objset_t *os;
kthread_t *resume_thread; kthread_t *resume_thread, *deadman_thread;
kthread_t **run_threads; kthread_t **run_threads;
uint64_t object; uint64_t object;
int error; int error;
@ -6782,7 +6794,7 @@ ztest_run(ztest_shared_t *zs)
/* /*
* Create a deadman thread and set to panic if we hang. * Create a deadman thread and set to panic if we hang.
*/ */
(void) thread_create(NULL, 0, ztest_deadman_thread, deadman_thread = thread_create(NULL, 0, ztest_deadman_thread,
zs, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); zs, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri);
spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC; spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC;
@ -6849,9 +6861,10 @@ ztest_run(ztest_shared_t *zs)
umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *)); umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *));
/* Kill the resume thread */ /* Kill the resume and deadman threads */
ztest_exiting = B_TRUE; ztest_exiting = B_TRUE;
VERIFY0(thread_join(resume_thread)); VERIFY0(thread_join(resume_thread));
VERIFY0(thread_join(deadman_thread));
ztest_resume(spa); ztest_resume(spa);
/* /*
@ -7351,6 +7364,7 @@ main(int argc, char **argv)
dprintf_setup(&argc, argv); dprintf_setup(&argc, argv);
zfs_deadman_synctime_ms = 300000; zfs_deadman_synctime_ms = 300000;
zfs_deadman_checktime_ms = 30000;
/* /*
* As two-word space map entries may not come up often (especially * As two-word space map entries may not come up often (especially
* if pool and vdev sizes are small) we want to force at least some * if pool and vdev sizes are small) we want to force at least some

2
module/zfs/spa_misc.c
View File

@ -312,7 +312,7 @@ unsigned long zfs_deadman_ziotime_ms = 300000ULL;
* Check time in milliseconds. This defines the frequency at which we check * Check time in milliseconds. This defines the frequency at which we check
* for hung I/O. * for hung I/O.
*/ */
unsigned long zfs_deadman_checktime_ms = 60000ULL; unsigned long zfs_deadman_checktime_ms = 60000ULL;
/* /*
* By default the deadman is enabled. * By default the deadman is enabled.