xfs: increase the default parallelism levels of pwork clients
authorDarrick J. Wong <djwong@kernel.org>
Sat, 23 Jan 2021 00:48:41 +0000 (16:48 -0800)
committerDarrick J. Wong <djwong@kernel.org>
Wed, 3 Feb 2021 17:18:49 +0000 (09:18 -0800)
commitf83d436aef5def77b318effc14809fdc57092588
treee1fc94c9c3e2dcf9ebf881d30716d7c33dd78897
parenta1a7d05a05765eec042942a5c360e909c0dd0131
xfs: increase the default parallelism levels of pwork clients

Increase the parallelism level for pwork clients to the workqueue
defaults so that we can take advantage of computers with a lot of CPUs
and a lot of hardware.  On fast systems this will speed up quotacheck by
a large factor, and the following posteof/cowblocks cleanup series will
use the functionality presented in this patch to run garbage collection
as quickly as possible.

We do this by switching the pwork workqueue to unbounded, since the
current user (quotacheck) runs lengthy scans for each work item and we
don't care about dispatching the work on a warm cpu cache or anything
like that.  Also set WQ_SYSFS so that we can monitor where the wq is
running.

Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Documentation/admin-guide/xfs.rst
fs/xfs/xfs_iwalk.c
fs/xfs/xfs_pwork.c
fs/xfs/xfs_pwork.h