Re: 2.6.12-rc4-mm1

["Artem B. Bityuckiy" <dedekind@infradead.org>]

[-- Attachment #1: Type: text/plain, Size: 1040 bytes --]

Steve,

> okay, reproduced after a couple hours of running (nothing intensive), 
> then doing a grep blah -r /*
I can't reproduce your problem using 2.6.12-rc4 + the 2 patches.
I created 50GB Reiserfs partition, copied a lot of data there (built
linux sources in many exemplars) and issued 'grep -r blah *'. I also
tried it with parallel writing to the partition - no help.

I wonder, could you please try 2.6.12-rc4 + the 2 patches and see if the
warning still there ?

The above referred patches are (and they are also attached):
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.12-
rc4/2.6.12-rc4-mm1/broken-out/vfs-bugfix-two-read_inode-calles-
without.patch 
ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.12-rc4/2.6.12-rc4-mm1/broken-out/__wait_on_freeing_inode-fix.patch

P.S. I wanted to try 2.6.12-rc4-mm2, but it OOPSes during the kernel
load in my system. Then I switched to 2.6.12-rc4 in order not to fight
with -mm2's problems.

-- 
Best Regards,
Artem B. Bityuckiy,
St.-Petersburg, Russia.

[-- Attachment #2: vfs-bugfix-two-read_inode-calles-without.patch --]
[-- Type: text/x-patch, Size: 1184 bytes --]

diff -puN fs/inode.c~vfs-bugfix-two-read_inode-calles-without fs/inode.c
--- 25/fs/inode.c~vfs-bugfix-two-read_inode-calles-without	Fri May  6 15:12:47 2005
+++ 25-akpm/fs/inode.c	Fri May  6 15:12:47 2005
@@ -282,6 +282,13 @@ static void dispose_list(struct list_hea
 		if (inode->i_data.nrpages)
 			truncate_inode_pages(&inode->i_data, 0);
 		clear_inode(inode);
+
+		spin_lock(&inode_lock);
+		hlist_del_init(&inode->i_hash);
+		list_del_init(&inode->i_sb_list);
+		spin_unlock(&inode_lock);
+
+		wake_up_inode(inode);
 		destroy_inode(inode);
 		nr_disposed++;
 	}
@@ -317,8 +324,6 @@ static int invalidate_list(struct list_h
 		inode = list_entry(tmp, struct inode, i_sb_list);
 		invalidate_inode_buffers(inode);
 		if (!atomic_read(&inode->i_count)) {
-			hlist_del_init(&inode->i_hash);
-			list_del(&inode->i_sb_list);
 			list_move(&inode->i_list, dispose);
 			inode->i_state |= I_FREEING;
 			count++;
@@ -439,8 +444,6 @@ static void prune_icache(int nr_to_scan)
 			if (!can_unuse(inode))
 				continue;
 		}
-		hlist_del_init(&inode->i_hash);
-		list_del_init(&inode->i_sb_list);
 		list_move(&inode->i_list, &freeable);
 		inode->i_state |= I_FREEING;
 		nr_pruned++;


[-- Attachment #3: __wait_on_freeing_inode-fix.patch --]
[-- Type: text/x-patch, Size: 1655 bytes --]

diff -puN fs/inode.c~__wait_on_freeing_inode-fix fs/inode.c
--- 25/fs/inode.c~__wait_on_freeing_inode-fix	2005-05-09 20:09:33.000000000 -0700
+++ 25-akpm/fs/inode.c	2005-05-09 20:09:33.000000000 -0700
@@ -1241,29 +1241,21 @@ int inode_wait(void *word)
 }
 
 /*
- * If we try to find an inode in the inode hash while it is being deleted, we
- * have to wait until the filesystem completes its deletion before reporting
- * that it isn't found.  This is because iget will immediately call
- * ->read_inode, and we want to be sure that evidence of the deletion is found
- * by ->read_inode.
+ * If we try to find an inode in the inode hash while it is being
+ * deleted, we have to wait until the filesystem completes its
+ * deletion before reporting that it isn't found.  This function waits
+ * until the deletion _might_ have completed.  Callers are responsible
+ * to recheck inode state.
+ *
+ * It doesn't matter if I_LOCK is not set initially, a call to
+ * wake_up_inode() after removing from the hash list will DTRT.
+ *
  * This is called with inode_lock held.
  */
 static void __wait_on_freeing_inode(struct inode *inode)
 {
 	wait_queue_head_t *wq;
 	DEFINE_WAIT_BIT(wait, &inode->i_state, __I_LOCK);
-
-	/*
-	 * I_FREEING and I_CLEAR are cleared in process context under
-	 * inode_lock, so we have to give the tasks who would clear them
-	 * a chance to run and acquire inode_lock.
-	 */
-	if (!(inode->i_state & I_LOCK)) {
-		spin_unlock(&inode_lock);
-		yield();
-		spin_lock(&inode_lock);
-		return;
-	}
 	wq = bit_waitqueue(&inode->i_state, __I_LOCK);
 	prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
 	spin_unlock(&inode_lock);