c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2024-12-25 18:59:33 +03:00
Code Issues Packages Projects Releases Wiki Activity

Fix zfs_putpage() lock inversion (again)

Browse Source 
This is a follow up commit to 74328ee which correctly resolved a lock
inversion between zfs_putpage() and zfs_free_range().  Unfortunately,
in the process it accidentally introduced another inversion between
zfs_putpage() and zfs_read().  The page must be unlocked before taking
the range lock.  This patch corrects that issue.

In addition, because the locking rules here are subtle a block comment
has been added clearly explaining why the ordering here is critical.

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Ned Bass <bass6@llnl.gov>
Issue #2976
This commit is contained in:
Brian Behlendorf 2015-01-06 16:54:57 -08:00
parent 33b6dbbc51
commit d958324f97
1 changed files with 16 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
module/zfs/zfs_vnops.c
View File

@ -3899,15 +3899,28 @@ zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc)
}
#endif
rl = zfs_range_lock(zp, pgoff, pglen, RL_WRITER);
set_page_writeback(pp);
/*
* The ordering here is critical and must adhere to the following
* rules in order to avoid deadlocking in either zfs_read() or
* zfs_free_range() due to a lock inversion.
*
* 1) The page must be unlocked prior to acquiring the range lock.
* This is critical because zfs_read() calls find_lock_page()
* which may block on the page lock while holding the range lock.
*
* 2) Before setting or clearing write back on a page the range lock
* must be held in order to prevent a lock inversion with the
* zfs_free_range() function.
*/
unlock_page(pp);
rl = zfs_range_lock(zp, pgoff, pglen, RL_WRITER);
set_page_writeback(pp);
tx = dmu_tx_create(zsb->z_os);
dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, zp);
err = dmu_tx_assign(tx, TXG_NOWAIT);
if (err != 0) {
if (err == ERESTART)