Skip to contentSkip to conversation

inotify: IN_DELETE events missing in -mm

PATCH In review Amy Griffis 4 participants 4 messages Mar 29, 2006 – Oct 16, 2013 VFS

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 inotify: IN_DELETE events missing in -mm

In recent -mm kernels (e.g. 2.6.16-mm1), IN_DELETE events are no longer generated for the removal of a file from a watched directory.

This seems to be a result of clearing DCACHE_INOTIFY_PARENT_WATCHED in d_delete() directly before calling fsnotify_nameremove().

Assuming the flag doesn't need to be cleared before dentry_iput(), this should do the trick.

fs/dcache.c +3 -3
index 363cd4b..344ce91 100644
@@ -1198,11 +1198,11 @@ void d_delete(struct dentry * dentry)
 	spin_lock(&dentry->d_lock);
 	isdir = S_ISDIR(dentry->d_inode->i_mode);
 	if (atomic_read(&dentry->d_count) == 1) {
-		/* remove this and other inotify debug checks after 2.6.18 */
-		dentry->d_flags &= ~DCACHE_INOTIFY_PARENT_WATCHED;
-
 		dentry_iput(dentry);
 		fsnotify_nameremove(dentry, isdir);
+
+		/* remove this and other inotify debug checks after 2.6.18 */
+		dentry->d_flags &= ~DCACHE_INOTIFY_PARENT_WATCHED;
 		return;
 	}
 
#1 lore raw
Quoted 8 lines · Amy Griffis

In recent -mm kernels (e.g. 2.6.16-mm1), IN_DELETE events are no longer generated for the removal of a file from a watched directory.

This seems to be a result of clearing DCACHE_INOTIFY_PARENT_WATCHED in d_delete() directly before calling fsnotify_nameremove().

Assuming the flag doesn't need to be cleared before dentry_iput(), this should do the trick.

I took a quick look at nick's patch again, and it doesn't seem to require the flag being cleared before dentry_iput. This looks good to me. Nick?

Signature · 1 line
John McCutchan <[email protected]>
#2 lore raw
In reply to Amy Griffis
Quoted 9 lines · Amy Griffis

In recent -mm kernels (e.g. 2.6.16-mm1), IN_DELETE events are no longer generated for the removal of a file from a watched directory.

This seems to be a result of clearing DCACHE_INOTIFY_PARENT_WATCHED in d_delete() directly before calling fsnotify_nameremove().

Assuming the flag doesn't need to be cleared before dentry_iput(), this should do the trick.

Thanks Amy. This should go to the next -stable kernel too.

Quoted 11 lines
fs/dcache.c +0 -1
index 363cd4b..344ce91 100644
@@ -1198,11 +1198,11 @@ void d_delete(struct dentry * dentry)
 	spin_lock(&dentry->d_lock);
 	isdir = S_ISDIR(dentry->d_inode->i_mode);
 	if (atomic_read(&dentry->d_count) == 1) {
-		/* remove this and other inotify debug checks after 2.6.18 */

fs/dcache.c void d_delete(struct dentry * dentry) · line 1208

Quoted 10 lines
fs/dcache.c +3 -2
-		dentry->d_flags &= ~DCACHE_INOTIFY_PARENT_WATCHED;
-
 		dentry_iput(dentry);
 		fsnotify_nameremove(dentry, isdir);
+
+		/* remove this and other inotify debug checks after 2.6.18 */
+		dentry->d_flags &= ~DCACHE_INOTIFY_PARENT_WATCHED;
 		return;
 	}
 

Send instant messages to your online friends http://au.messenger.yahoo.com

#3 lore raw
In reply to Amy Griffis

You cannot move this line of code after dentry_iput(dentry); without considerations for locking.

Another thread can get this dentry and be changing dentry->d_flags at the same time!
We have hit this in testing. The code must be changed to have locks (if this field needs to be changed or ...)

In our tests, the flag 0x10 in __d_drop goes missing because this line of code executed at the same time on another processor.

--
View this message in context: http://linux-kernel.2935.n7.nabble.com/PATCH-inotify-IN-DELETE-events-missing-in-mm-tp105372p738511.html
Sent from the Linux Kernel mailing list archive at Nabble.com.