On Tue, 2 Apr 2013 00:06:24 -0400 (EDT), CAI Qian <caiqian@redhat.com> wrote:
> Bisect indicated this is the culprit,
> 
> 0e401101db49959f5783f6ee9e676124b5a183ac
> ext4: fix memory leakage in mext_check_coverage
Strange... 
It changes a bug in move_extent.c (e4defrag functionality)
ASAIU you just previously stopped your bisecting process here. Right?
Is this indeed a first bad commit?
> 
> This following with Dmitry's debug patch applied,
> 
> CAI Qian
> 
> Ý  101.408610¨ ES cache assertation failed for inode: 753 es_cached ex Ý56/5/744
> 81/20¨ != found ex Ý56/5/3396400/0¨ retval 0 flags 5
> Ý  209.858899¨ ES cache assertation failed for inode: 384 es_cached ex Ý57/7/332
> 82/20¨ != found ex Ý57/7/3396400/0¨ retval 0 flags 5
> Ý  209.860656¨ ES cache assertation failed for inode: 384 es_cached ex Ý25/1/332
> 50/20¨ != found ex Ý25/1/0/0¨ retval 0 flags 0
> Ý  209.893587¨ ES cache assertation failed for inode: 384 es_cached ex Ý22/1/332
> 47/20¨ != found ex Ý22/1/34838/1000¨ retval 1 flags 0
> Ý  209.913482¨ ES cache assertation failed for inode: 384 es_cached ex Ý27/1/329
> 40/20¨ != found ex Ý27/1/0/0¨ retval 0 flags 0
> Ý  209.919950¨ ES cache assertation failed for inode: 384 es_cached ex Ý59/5/338
> 48/20¨ != found ex Ý59/5/3396400/0¨ retval 0 flags 5
> Ý  209.931856¨ ES cache assertation failed for inode: 384 es_cached ex Ý7/1/3292
> 0/20¨ != found ex Ý7/1/35879/20¨ retval 1 flags 43
> Ý  209.969282¨ ES cache assertation failed for inode: 384 es_cached ex Ý35/1/361
> 97/20¨ != found ex Ý35/1/36197/1000¨ retval 1 flags 0
> Ý  209.969290¨ ES cache assertation failed for inode: 384 es_cached ex Ý48/1/362
> 10/20¨ != found ex Ý48/1/0/0¨ retval 0 flags 0
> Ý  209.980724¨ ES cache assertation failed for inode: 384 es_cached ex Ý13/4/334
> 89/20¨ != found ex Ý13/4/2161372/0¨ retval 0 flags 5
> Ý  209.980744¨ ES cache assertation failed for inode: 384 es_cached ex Ý61/3/335
> 37/20¨ != found ex Ý61/3/3396400/0¨ retval 0 flags 5
> Ý  209.983848¨ ES cache assertation failed for inode: 384 es_cached ex Ý44/2/335
> 20/20¨ != found ex Ý44/2/36216/20¨ retval 2 flags 43
> Ý  210.020041¨ ES cache assertation failed for inode: 384 es_cached ex Ý61/3/341
> 91/20¨ != found ex Ý61/3/3396400/0¨ retval 0 flags 5
> Ý  210.050100¨ ES cache assertation failed for inode: 384 es_cached ex Ý22/11/34
> 565/20¨ != found ex Ý22/11/3396400/0¨ retval 0 flags 5
> Ý  210.053271¨ ES cache assertation failed for inode: 384 es_cached ex Ý15/1/334
> 90/20¨ != found ex Ý15/1/33579/1000¨ retval 1 flags 1
It does not looks like bigendian issue, actually I cant find any logical
system in the log. The only thing I see is that es_cache is
horribly out of sync with extent_tree. 
Please try this patch: