Re: reiserfs oops [2.5.65]

[Andrew Morton <akpm@digeo.com>]

Dave Jones <davej@codemonkey.org.uk> wrote:
>
> There's lots of "slab error in cache_alloc_debugcheck_after()"
> warnings. cache reiser_inode_cache memory after object was overwritten
> 
> Some call traces.
>  check_poison_obj <- kmem_cache_alloc <- reierfs_alloc_inode <-
>  reiserfs_alloc_inode <- alloc_inode <- get_new_inode <-
>  reiserfs_init_locked <- reiserfs_find_actor <- reiserfs_iget <-
>  reiserfs_find_actor <- reiserfs_init_locked_inode <- reiserfs_lookup <-
>  real_lookup <- do_lookup <- link_path_walk <- kmem_cache_alloc <-
>  __user_walk <- vfs_lstat <- sys_lstat64 <- syscall_call
> 
> Slab corruption: start=c70c7044, expend=c70c7213 problemat=c70c7044
> Last user: [<c0280dcb>](reiserfs_alloc_inode+0x1b/0x30)
> Data: (lots of hex)

Alas, the "(lots of hex)" is important - it lets us determine which member of
struct reiserfs_inode was actually altered.

> I'll give that box a run of memtest to rule out memory corruption
> problems. I'll also hook up a serial terminal tonight to catch tomorrow
> nights 'activity' in full 8-)

Good, thanks.

It would be nice if we had a more robust way of capturing all this info,
especially the oops-while-running-X lossage.  Dump-to-floppy or something.