summary |
shortlog |
log |
commit | commitdiff |
tree
raw |
patch |
inline | side by side (from parent 1:
e26ade5)
Gunnar Beutner did all the hard work on this one by correctly
identifying that this issue is a race between dmu_sync() and
dbuf_dirty().
Now in all cases the caller is responsible for preventing this
race by making sure the zfs_range_lock() is held when dirtying
a buffer which may be referenced in a log record. The mmap
case which relies on zfs_putpage() was not taking the range
lock. This code was accidentally dropped when the function
was rewritten for the Linux VFS.
This patch adds the required range locking to zfs_putpage().
It also adds the missing ZFS_ENTER()/ZFS_EXIT() macros which
aren't strictly required due to the VFS holding a reference.
However, this makes the code more consistent with the upsteam
code and there's no harm in being extra careful here.
Original-patch-by: Gunnar Beutner <gunnar@beutner.name>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #541
zfs_sb_t *zsb = ITOZSB(ip);
loff_t offset;
loff_t pgoff;
zfs_sb_t *zsb = ITOZSB(ip);
loff_t offset;
loff_t pgoff;
+ unsigned int pglen;
+ rl_t *rl;
dmu_tx_t *tx;
caddr_t va;
int err = 0;
dmu_tx_t *tx;
caddr_t va;
int err = 0;
+ ZFS_ENTER(zsb);
+ ZFS_VERIFY_ZP(zp);
/* Page is beyond end of file */
if (pgoff >= offset) {
unlock_page(pp);
/* Page is beyond end of file */
if (pgoff >= offset) {
unlock_page(pp);
set_page_writeback(pp);
unlock_page(pp);
set_page_writeback(pp);
unlock_page(pp);
+ rl = zfs_range_lock(zp, pgoff, pglen, RL_WRITER);
tx = dmu_tx_create(zsb->z_os);
sync = ((zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) ||
tx = dmu_tx_create(zsb->z_os);
sync = ((zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) ||
if (sync)
zfs_putpage_commit_cb(pp, ECANCELED);
if (sync)
zfs_putpage_commit_cb(pp, ECANCELED);
+ zfs_range_unlock(rl);
+ ZFS_EXIT(zsb);
zfs_log_write(zsb->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0);
dmu_tx_commit(tx);
zfs_log_write(zsb->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0);
dmu_tx_commit(tx);
ASSERT3S(err, ==, 0);
if (sync) {
ASSERT3S(err, ==, 0);
if (sync) {
zfs_putpage_commit_cb(pp, err);
}
zfs_putpage_commit_cb(pp, err);
}