diff options
| author | Mikulas Patocka <mpatocka@redhat.com> | 2014-10-16 14:45:20 -0400 |
|---|---|---|
| committer | Zefan Li <lizefan@huawei.com> | 2015-02-02 17:05:16 +0800 |
| commit | 0e7638d87693175cc479b91a8227a27991363cf0 (patch) | |
| tree | 3c7e96a9785e052bae4d6592838495acb557d53c /drivers | |
| parent | d7b1b1db066e9aa59d25ed13c97353a4dbf6ea6c (diff) | |
dm bufio: change __GFP_IO to __GFP_FS in shrinker callbacks
commit 9d28eb12447ee08bb5d1e8bb3195cf20e1ecd1c0 upstream.
The shrinker uses gfp flags to indicate what kind of operation can the
driver wait for. If __GFP_IO flag is present, the driver can wait for
block I/O operations, if __GFP_FS flag is present, the driver can wait on
operations involving the filesystem.
dm-bufio tested for __GFP_IO. However, dm-bufio can run on a loop block
device that makes calls into the filesystem. If __GFP_IO is present and
__GFP_FS isn't, dm-bufio could still block on filesystem operations if it
runs on a loop block device.
The change from __GFP_IO to __GFP_FS supposedly fixes one observed (though
unreproducible) deadlock involving dm-bufio and loop device.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
[lizf: Backported to 3.4:
- drop changes to dm_bufio_shrink_scan() and dm_bufio_shrink_count()
- change __GFP_IO to __GFP_FS in shrink()]
Signed-off-by: Zefan Li <lizefan@huawei.com>
Diffstat (limited to 'drivers')
| -rw-r--r-- | drivers/md/dm-bufio.c | 10 |
1 files changed, 5 insertions, 5 deletions
diff --git a/drivers/md/dm-bufio.c b/drivers/md/dm-bufio.c index 5abc8d62bb56..535b09cd3cd8 100644 --- a/drivers/md/dm-bufio.c +++ b/drivers/md/dm-bufio.c @@ -1379,9 +1379,9 @@ static void drop_buffers(struct dm_bufio_client *c) /* * Test if the buffer is unused and too old, and commit it. - * At if noio is set, we must not do any I/O because we hold - * dm_bufio_clients_lock and we would risk deadlock if the I/O gets rerouted to - * different bufio client. + * And if GFP_NOFS is used, we must not do any I/O because we hold + * dm_bufio_clients_lock and we would risk deadlock if the I/O gets + * rerouted to different bufio client. */ static int __cleanup_old_buffer(struct dm_buffer *b, gfp_t gfp, unsigned long max_jiffies) @@ -1389,7 +1389,7 @@ static int __cleanup_old_buffer(struct dm_buffer *b, gfp_t gfp, if (jiffies - b->last_accessed < max_jiffies) return 1; - if (!(gfp & __GFP_IO)) { + if (!(gfp & __GFP_FS)) { if (test_bit(B_READING, &b->state) || test_bit(B_WRITING, &b->state) || test_bit(B_DIRTY, &b->state)) @@ -1428,7 +1428,7 @@ static int shrink(struct shrinker *shrinker, struct shrink_control *sc) unsigned long r; unsigned long nr_to_scan = sc->nr_to_scan; - if (sc->gfp_mask & __GFP_IO) + if (sc->gfp_mask & __GFP_FS) dm_bufio_lock(c); else if (!dm_bufio_trylock(c)) return !nr_to_scan ? 0 : -1; |