Messages in this thread |  | | Date | Wed, 3 Sep 2025 12:37:20 +0200 | | From | Jan Kara <> | | Subject | Re: [PATCH] fs: quota: create dedicated workqueue for quota_release_work |
| |
On Wed 03-09-25 14:01:46, Shashank A P wrote: > On 9/2/2025 7:17 PM, Jan Kara wrote: > > On Mon 01-09-25 14:59:00, Shashank A P wrote: > >> There is a kernel panic due to WARN_ONCE when panic_on_warn is set. > >> > >> This issue occurs when writeback is triggered due to sync call for an > >> opened file(ie, writeback reason is WB_REASON_SYNC). When f2fs balance > >> is needed at sync path, flush for quota_release_work is triggered. > >> By default quota_release_work is queued to "events_unbound" queue which > >> does not have WQ_MEM_RECLAIM flag. During f2fs balance "writeback" > >> workqueue tries to flush quota_release_work causing kernel panic due to > >> MEM_RECLAIM flag mismatch errors. > >> > >> This patch creates dedicated workqueue with WQ_MEM_RECLAIM flag > >> for work quota_release_work. > >> > >> ------------[ cut here ]------------ > >> WARNING: CPU: 4 PID: 14867 at kernel/workqueue.c:3721 check_flush_dependency+0x13c/0x148 > >> Call trace: > >> check_flush_dependency+0x13c/0x148 > >> __flush_work+0xd0/0x398 > >> flush_delayed_work+0x44/0x5c > >> dquot_writeback_dquots+0x54/0x318 > >> f2fs_do_quota_sync+0xb8/0x1a8 > >> f2fs_write_checkpoint+0x3cc/0x99c > >> f2fs_gc+0x190/0x750 > >> f2fs_balance_fs+0x110/0x168 > >> f2fs_write_single_data_page+0x474/0x7dc > >> f2fs_write_data_pages+0x7d0/0xd0c > >> do_writepages+0xe0/0x2f4 > >> __writeback_single_inode+0x44/0x4ac > >> writeback_sb_inodes+0x30c/0x538 > >> wb_writeback+0xf4/0x440 > >> wb_workfn+0x128/0x5d4 > >> process_scheduled_works+0x1c4/0x45c > >> worker_thread+0x32c/0x3e8 > >> kthread+0x11c/0x1b0 > >> ret_from_fork+0x10/0x20 > >> Kernel panic - not syncing: kernel: panic_on_warn set ... > >> > >> Fixes: ac6f420291b3 ("quota: flush quota_release_work upon quota writeback") > >> CC: stable@vger.kernel.org > >> Signed-off-by: Shashank A P <shashank.ap@samsung.com> > > Thanks. It seems a bit unfortunate that we have to create a separate > > workqueue just for this but I don't see a different easy solution. So I've > > added your patch to my tree. > > > > Honza > > Hi Jan Kara, > Thanks for your comments. > > We've got a below kernel warning from kernel test > robot(lkp@intel.com)that related to the changes, and we need your > suggestions to resolve this warning as its merged already in your tree. > > Should we post a new patch version to fix this warning, or can you > please incorporate the fix while applying it to linux-next? > > > sparse warnings: (new ones prefixed by >>) > >> fs/quota/dquot.c:166:25: sparse: sparse: symbol 'quota_unbound_wq' > was not declared. Should it be static? > > https://lore.kernel.org/all/202509031153.0ACADDn6-lkp@intel.com/
Thanks for noticing. Somehow I didn't get this notification. Anyway, I've fixed up the commit in my tree to declare quota_unbound_wq as static.
Honza
> >> --- > >> fs/quota/dquot.c | 10 +++++++++- > >> 1 file changed, 9 insertions(+), 1 deletion(-) > >> > >> diff --git a/fs/quota/dquot.c b/fs/quota/dquot.c > >> index df4a9b348769..d0f83a0c42df 100644 > >> --- a/fs/quota/dquot.c > >> +++ b/fs/quota/dquot.c > >> @@ -162,6 +162,9 @@ static struct quota_module_name module_names[] = INIT_QUOTA_MODULE_NAMES; > >> /* SLAB cache for dquot structures */ > >> static struct kmem_cache *dquot_cachep; > >> > >> +/* workqueue for work quota_release_work*/ > >> +struct workqueue_struct *quota_unbound_wq; > >> + > >> void register_quota_format(struct quota_format_type *fmt) > >> { > >> spin_lock(&dq_list_lock); > >> @@ -881,7 +884,7 @@ void dqput(struct dquot *dquot) > >> put_releasing_dquots(dquot); > >> atomic_dec(&dquot->dq_count); > >> spin_unlock(&dq_list_lock); > >> - queue_delayed_work(system_unbound_wq, "a_release_work, 1); > >> + queue_delayed_work(quota_unbound_wq, "a_release_work, 1); > >> } > >> EXPORT_SYMBOL(dqput); > >> > >> @@ -3041,6 +3044,11 @@ static int __init dquot_init(void) > >> > >> shrinker_register(dqcache_shrinker); > >> > >> + quota_unbound_wq = alloc_workqueue("quota_events_unbound", > >> + WQ_UNBOUND | WQ_MEM_RECLAIM, WQ_MAX_ACTIVE); > >> + if (!quota_unbound_wq) > >> + panic("Cannot create quota_unbound_wq\n"); > >> + > >> return 0; > >> } > >> fs_initcall(dquot_init); > >> -- > >> 2.34.1 > >> -- Jan Kara <jack@suse.com> SUSE Labs, CR
|  |