Skip to contentSkip to conversation

nfsd: use system_unbound_wq for nfsd_file_gc_worker()

PATCH In review Youzhong Yang 2 participants 2 messages Jul 11, 2024

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 nfsd: use system_unbound_wq for nfsd_file_gc_worker()

After many rounds of changes in filecache.c, the fix by commit ce7df055(NFSD: Make the file_delayed_close workqueue UNBOUND) is gone, now we are getting syslog messages like these:

[ 1618.186688] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 4 times, consider switching to WQ_UNBOUND
[ 1638.661616] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 8 times, consider switching to WQ_UNBOUND
[ 1665.284542] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 16 times, consider switching to WQ_UNBOUND
[ 1759.491342] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 32 times, consider switching to WQ_UNBOUND
[ 3013.012308] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 64 times, consider switching to WQ_UNBOUND
[ 3154.172827] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 128 times, consider switching to WQ_UNBOUND
[ 3422.461924] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 256 times, consider switching to WQ_UNBOUND
[ 3963.152054] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 512 times, consider switching to WQ_UNBOUND

Consider use system_unbound_wq instead of system_wq for nfsd_file_gc_worker().

---
 fs/nfsd/filecache.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/nfsd/filecache.c +1 -1
index ad9083ca144b..e7faa373d45e 100644
@@ -111,7 +111,7 @@ static void
 nfsd_file_schedule_laundrette(void)
 {
 	if (test_bit(NFSD_FILE_CACHE_UP, &nfsd_file_flags))
-		queue_delayed_work(system_wq, &nfsd_filecache_laundrette,
+		queue_delayed_work(system_unbound_wq, &nfsd_filecache_laundrette,
 				   NFSD_LAUNDRETTE_DELAY);
 }
 
Signature · 1 line
2.45.2
#1 lore raw
Quoted 13 lines · Youzhong Yang

After many rounds of changes in filecache.c, the fix by commit ce7df055(NFSD: Make the file_delayed_close workqueue UNBOUND) is gone, now we are getting syslog messages like these:

[ 1618.186688] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 4 times, consider switching to WQ_UNBOUND
[ 1638.661616] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 8 times, consider switching to WQ_UNBOUND
[ 1665.284542] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 16 times, consider switching to WQ_UNBOUND
[ 1759.491342] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 32 times, consider switching to WQ_UNBOUND
[ 3013.012308] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 64 times, consider switching to WQ_UNBOUND
[ 3154.172827] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 128 times, consider switching to WQ_UNBOUND
[ 3422.461924] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 256 times, consider switching to WQ_UNBOUND
[ 3963.152054] workqueue: nfsd_file_gc_worker [nfsd] hogged CPU for >13333us 512 times, consider switching to WQ_UNBOUND
Quoted 2 lines

Consider use system_unbound_wq instead of system_wq for nfsd_file_gc_worker().

Seems sensible. I will let Neil have the last word.

Quoted 21 lines
---
 fs/nfsd/filecache.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/nfsd/filecache.c +1 -1
index ad9083ca144b..e7faa373d45e 100644
@@ -111,7 +111,7 @@ static void
 nfsd_file_schedule_laundrette(void)
 {
 	if (test_bit(NFSD_FILE_CACHE_UP, &nfsd_file_flags))
-		queue_delayed_work(system_wq, &nfsd_filecache_laundrette,
+		queue_delayed_work(system_unbound_wq, &nfsd_filecache_laundrette,
 				   NFSD_LAUNDRETTE_DELAY);
 }
 
Signature · 1 line
2.45.2
Signature · 1 line
Chuck Lever