]> Git Repo - linux.git/commit - fs/fs-writeback.c
fs: remove WB_SYNC_HOLD
authorNick Piggin <[email protected]>
Tue, 6 Jan 2009 22:40:25 +0000 (14:40 -0800)
committerLinus Torvalds <[email protected]>
Tue, 6 Jan 2009 23:59:09 +0000 (15:59 -0800)
commit4f5a99d64c17470a784a6c68064207d82e3e74a5
tree2a3e0f0c3990bb8dbda2cdaa506a64180e5cbff2
parente8ea1759138d4279869f52bfb7dca8f02f8ccfe5
fs: remove WB_SYNC_HOLD

Remove WB_SYNC_HOLD.  The primary motiviation is the design of my
anti-starvation code for fsync.  It requires taking an inode lock over the
sync operation, so we could run into lock ordering problems with multiple
inodes.  It is possible to take a single global lock to solve the ordering
problem, but then that would prevent a future nice implementation of "sync
multiple inodes" based on lock order via inode address.

Seems like a backward step to remove this, but actually it is busted
anyway: we can't use the inode lists for data integrity wait: an inode can
be taken off the dirty lists but still be under writeback.  In order to
satisfy data integrity semantics, we should wait for it to finish
writeback, but if we only search the dirty lists, we'll miss it.

It would be possible to have a "writeback" list, for sys_sync, I suppose.
But why complicate things by prematurely optimise?  For unmounting, we
could avoid the "livelock avoidance" code, which would be easier, but
again premature IMO.

Fixing the existing data integrity problem will come next.

Signed-off-by: Nick Piggin <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
fs/fs-writeback.c
include/linux/writeback.h
This page took 0.055027 seconds and 4 git commands to generate.