inotify: use GFP_NOFS under potential memory pressure
inotify can have a watchs removed under filesystem reclaim.
=================================
[ INFO: inconsistent lock state ]
2.6.31-rc2 #16
---------------------------------
inconsistent {IN-RECLAIM_FS-W} -> {RECLAIM_FS-ON-W} usage.
khubd/217 [HC0[0]:SC0[0]:HE1:SE1] takes:
(iprune_mutex){+.+.?.}, at: [<c10ba899>] invalidate_inodes+0x20/0xe3
{IN-RECLAIM_FS-W} state was registered at:
[<c10536ab>] __lock_acquire+0x2c9/0xac4
[<c1053f45>] lock_acquire+0x9f/0xc2
[<c1308872>] __mutex_lock_common+0x2d/0x323
[<c1308c00>] mutex_lock_nested+0x2e/0x36
[<c10ba6ff>] shrink_icache_memory+0x38/0x1b2
[<c108bfb6>] shrink_slab+0xe2/0x13c
[<c108c3e1>] kswapd+0x3d1/0x55d
[<c10449b5>] kthread+0x66/0x6b
[<c1003fdf>] kernel_thread_helper+0x7/0x10
[<ffffffff>] 0xffffffff
Two things are needed to fix this. First we need a method to tell
fsnotify_create_event() to use GFP_NOFS and second we need to stop using
one global IN_IGNORED event and allocate them one at a time. This solves
current issues with multiple IN_IGNORED on a queue having tail drop
problems and simplifies the allocations since we don't have to worry about
two tasks opperating on the IGNORED event concurrently.
Signed-off-by: Eric Paris <eparis@redhat.com>
diff --git a/fs/notify/fsnotify.c b/fs/notify/fsnotify.c
index ec2f7bd..037e878 100644
--- a/fs/notify/fsnotify.c
+++ b/fs/notify/fsnotify.c
@@ -159,7 +159,9 @@
if (!group->ops->should_send_event(group, to_tell, mask))
continue;
if (!event) {
- event = fsnotify_create_event(to_tell, mask, data, data_is, file_name, cookie);
+ event = fsnotify_create_event(to_tell, mask, data,
+ data_is, file_name, cookie,
+ GFP_KERNEL);
/* shit, we OOM'd and now we can't tell, maybe
* someday someone else will want to do something
* here */
diff --git a/fs/notify/inotify/inotify_user.c b/fs/notify/inotify/inotify_user.c
index 726118a..f30d9bb 100644
--- a/fs/notify/inotify/inotify_user.c
+++ b/fs/notify/inotify/inotify_user.c
@@ -57,7 +57,6 @@
static struct kmem_cache *inotify_inode_mark_cachep __read_mostly;
struct kmem_cache *event_priv_cachep __read_mostly;
-static struct fsnotify_event *inotify_ignored_event;
/*
* When inotify registers a new group it increments this and uses that
@@ -384,12 +383,19 @@
struct fsnotify_group *group)
{
struct inotify_inode_mark_entry *ientry;
+ struct fsnotify_event *ignored_event;
struct inotify_event_private_data *event_priv;
struct fsnotify_event_private_data *fsn_event_priv;
+ ignored_event = fsnotify_create_event(NULL, FS_IN_IGNORED, NULL,
+ FSNOTIFY_EVENT_NONE, NULL, 0,
+ GFP_NOFS);
+ if (!ignored_event)
+ return;
+
ientry = container_of(entry, struct inotify_inode_mark_entry, fsn_entry);
- event_priv = kmem_cache_alloc(event_priv_cachep, GFP_KERNEL);
+ event_priv = kmem_cache_alloc(event_priv_cachep, GFP_NOFS);
if (unlikely(!event_priv))
goto skip_send_ignore;
@@ -398,7 +404,7 @@
fsn_event_priv->group = group;
event_priv->wd = ientry->wd;
- fsnotify_add_notify_event(group, inotify_ignored_event, fsn_event_priv);
+ fsnotify_add_notify_event(group, ignored_event, fsn_event_priv);
/* did the private data get added? */
if (list_empty(&fsn_event_priv->event_list))
@@ -406,6 +412,9 @@
skip_send_ignore:
+ /* matches the reference taken when the event was created */
+ fsnotify_put_event(ignored_event);
+
/* remove this entry from the idr */
inotify_remove_from_idr(group, ientry);
@@ -748,9 +757,6 @@
inotify_inode_mark_cachep = KMEM_CACHE(inotify_inode_mark_entry, SLAB_PANIC);
event_priv_cachep = KMEM_CACHE(inotify_event_private_data, SLAB_PANIC);
- inotify_ignored_event = fsnotify_create_event(NULL, FS_IN_IGNORED, NULL, FSNOTIFY_EVENT_NONE, NULL, 0);
- if (!inotify_ignored_event)
- panic("unable to allocate the inotify ignored event\n");
inotify_max_queued_events = 16384;
inotify_max_user_instances = 128;
diff --git a/fs/notify/notification.c b/fs/notify/notification.c
index 2b20feaf..5213685 100644
--- a/fs/notify/notification.c
+++ b/fs/notify/notification.c
@@ -153,7 +153,7 @@
return true;
break;
case (FSNOTIFY_EVENT_NONE):
- return true;
+ return false;
};
}
return false;
@@ -345,18 +345,19 @@
* @name the filename, if available
*/
struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32 mask, void *data,
- int data_type, const char *name, u32 cookie)
+ int data_type, const char *name, u32 cookie,
+ gfp_t gfp)
{
struct fsnotify_event *event;
- event = kmem_cache_alloc(fsnotify_event_cachep, GFP_KERNEL);
+ event = kmem_cache_alloc(fsnotify_event_cachep, gfp);
if (!event)
return NULL;
initialize_event(event);
if (name) {
- event->file_name = kstrdup(name, GFP_KERNEL);
+ event->file_name = kstrdup(name, gfp);
if (!event->file_name) {
kmem_cache_free(fsnotify_event_cachep, event);
return NULL;
diff --git a/include/linux/fsnotify_backend.h b/include/linux/fsnotify_backend.h
index 6c3de99..4d6f47b 100644
--- a/include/linux/fsnotify_backend.h
+++ b/include/linux/fsnotify_backend.h
@@ -352,7 +352,7 @@
/* put here because inotify does some weird stuff when destroying watches */
extern struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32 mask,
void *data, int data_is, const char *name,
- u32 cookie);
+ u32 cookie, gfp_t gfp);
#else