Looking at the errors in comment #11 from Natty, we see the following:
[ 134.061517] WARNING: at .../ubuntu/aufs/plink.c:457 au_plink_put+0x4c/0xa0 [aufs]()
[ 134.061519] Hardware name: 5129CTO
[ 134.061520] pseudo-link is not flushed
[...]
[ 134.061641] [<ffffffff81182a00>] sys_umount+0x60/0xd0
This is being triggered out of an unmount of the union mount. Looking at the
code this seems to be a benign warning, note how the code goes on to flush out the links and clean up:
void au_plink_put(struct super_block *sb, int verbose)
[...] WARN(verbose && !list_empty(plink_list), "pseudo-link is not flushed"); list_for_each_entry_safe(plink, tmp, plink_list, list) do_put_plink(plink, 0); INIT_LIST_HEAD(plink_list);
[...]
I would not expect this to trigger any instability. I think this warning likely should be squashed in general on the unmount path.
Looking at the Maverick errors as originally reported we see the following:
WARNING: at .../ubuntu/aufs/plink.c:332 au_plink_append+0x1f1/0x2e0 [aufs]()
Hardware name: 6459CTO
unexpectedly many pseudo links, 101
[...]
[<ffffffff8115fe85>] sys_linkat+0x165/0x170
This is a completly different error. However this also does not appear to be anything other than a diagnostic. It implies we have a large number of pseudo links, but processing goes on as normal:
As this list is maintained via RCU it is possible for this list to grow quite large under high load. This also looks like a candidate for removal as a benign message.
@All -- can you confirm that there are stability issues on Maverick associated with aufs use here? If so what are the symptoms. Just seeing these messages are not in of themselves indications of anything other than use of aufs itself.
@Jamie -- could you confirm whether you are seeing any bad effects with the Natty kernel in the face of the error there. I believe that is a different and very likely benign message.
Looking at the errors in comment #11 from Natty, we see the following:
[ 134.061517] WARNING: at .../ubuntu/ aufs/plink. c:457 au_plink_ put+0x4c/ 0xa0 [aufs]() a00>] sys_umount+ 0x60/0xd0
[ 134.061519] Hardware name: 5129CTO
[ 134.061520] pseudo-link is not flushed
[...]
[ 134.061641] [<ffffffff81182
This is being triggered out of an unmount of the union mount. Looking at the
code this seems to be a benign warning, note how the code goes on to flush out the links and clean up:
void au_plink_put(struct super_block *sb, int verbose)
WARN(verbose && !list_empty( plink_list) , "pseudo-link is not flushed");
list_for_ each_entry_ safe(plink, tmp, plink_list, list)
do_put_ plink(plink, 0);
INIT_LIST_ HEAD(plink_ list);
[...]
[...]
I would not expect this to trigger any instability. I think this warning likely should be squashed in general on the unmount path.
Looking at the Maverick errors as originally reported we see the following:
WARNING: at .../ubuntu/ aufs/plink. c:332 au_plink_ append+ 0x1f1/0x2e0 [aufs]() ffffffff8115fe8 5>] sys_linkat+ 0x165/0x170
Hardware name: 6459CTO
unexpectedly many pseudo links, 101
[...]
[<
This is a completly different error. However this also does not appear to be anything other than a diagnostic. It implies we have a large number of pseudo links, but processing goes on as normal:
void au_plink_ append( struct inode *inode, aufs_bindex_t bindex,
struct dentry *h_dentry)
cnt+ +;
WARN_ ONCE(cnt > AUFS_PLINK_WARN,
"unexpectedly many pseudo links, %d\n", cnt);
{
[...]
if (!found) {
[...]
As this list is maintained via RCU it is possible for this list to grow quite large under high load. This also looks like a candidate for removal as a benign message.
@All -- can you confirm that there are stability issues on Maverick associated with aufs use here? If so what are the symptoms. Just seeing these messages are not in of themselves indications of anything other than use of aufs itself.
@Jamie -- could you confirm whether you are seeing any bad effects with the Natty kernel in the face of the error there. I believe that is a different and very likely benign message.