Re: ENOSPC error returned too late

["Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>]

On Mon, Aug 18, 2008 at 03:00:26PM +0530, Aneesh Kumar K.V wrote:
> On Sun, Aug 17, 2008 at 08:53:11AM -0400, Theodore Ts'o wrote:
> > 
> > While trying to track down kernel BZ #11341 (which is really an
> > e2fsprogs bug, not a 2.6.26 regression) I created a fresh ext4
> > filesystem which was 5GB in length, and then filled it to ENOSPC via:
> > 
> > dd if=/dev/zero of=/mnt/bigfile bs=1024 count=10485760
> > 
> > Unfortunately, it sent the ENOSPC few blocks too late, so the following
> > appeared in dmesg, and an attempt to umount the filesystem hung in D
> > wait.  :-(
> > 
> > 						- Ted
> > 
> > 
> > [63518.930337] mpage_da_map_blocks block allocation failed for inode 12 at logical offset 1254547 with max blocks 2 with error -28
> > [63518.930337] This should not happen.!! Data will be lost
> 
> I will look into this. Are we really out of free space or is it that
> mballoc is not able to find free space ?. I am not able to reproduce
> this here. Will try more.
> 
> 

How about 

commit 08085a642ac6e7c3bfae0b6daf538e64a99695b2
Author: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Date:   Mon Aug 18 16:20:39 2008 +0530

    ext4: make sure ext4_has_free_blocks return 0 for ENOSPC
    
    Fix ext4_has_free_blocks to return 0 when we don't have
    enought space.
    
    Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

diff --git a/fs/ext4/balloc.c b/fs/ext4/balloc.c
index 27323cd..dfe2d4f 100644
--- a/fs/ext4/balloc.c
+++ b/fs/ext4/balloc.c
@@ -1628,6 +1628,9 @@ ext4_fsblk_t ext4_has_free_blocks(struct ext4_sb_info *sbi,
 		free_blocks =
 			percpu_counter_sum_and_set(&sbi->s_freeblocks_counter);
 #endif
+	if (free_blocks <= root_blocks)
+		/* we don't have free space */
+		return 0;
 	if (free_blocks - root_blocks < nblocks)
 		return free_blocks - root_blocks;
 	return nblocks;