]> git.kernelconcepts.de Git - karo-tx-linux.git/commit
writeback: plug writeback in wb_writeback() and writeback_inodes_wb()
authorLinus Torvalds <torvalds@linux-foundation.org>
Fri, 11 Sep 2015 20:37:19 +0000 (13:37 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 12 Sep 2015 18:13:07 +0000 (11:13 -0700)
commit505a666ee3fc611518e85df203eb8c707995ceaa
treec4df72f1504a245d027e7c3c7617e3d08c0c7385
parent0ba13fd19d39b7cb672bcec052bc813389c079a4
writeback: plug writeback in wb_writeback() and writeback_inodes_wb()

We had to revert the pluggin in writeback_sb_inodes() because the
wb->list_lock is held, but we could easily plug at a higher level before
taking that lock, and unplug after releasing it.  This does that.

Chris will run performance numbers, just to verify that this approach is
comparable to the alternative (we could just drop and re-take the lock
around the blk_finish_plug() rather than these two commits.

I'd have preferred waiting for actual performance numbers before picking
one approach over the other, but I don't want to release rc1 with the
known "sleeping function called from invalid context" issue, so I'll
pick this cleanup version for now.  But if the numbers show that we
really want to plug just at the writeback_sb_inodes() level, and we
should just play ugly games with the spinlock, we'll switch to that.

Cc: Chris Mason <clm@fb.com>
Cc: Josef Bacik <jbacik@fb.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Neil Brown <neilb@suse.de>
Cc: Jan Kara <jack@suse.cz>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/fs-writeback.c