[Bug 212168] [panic] [UFS] use-after-free panic (0xdeadc0dedeadc0de)

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=212168

--- Comment #13 from Glen Barber <gjb at FreeBSD.org> ---
(In reply to Kirk McKusick from comment #10)
> It appears that there is an item on the worklist that has been prematurely
> freed. This freeing should be impossible as anything on a worklist has its
> ONWORKLIST flag set and the worklist freeing routine will panic if the
> ONWORKLIST flag is set. However I have written the following check to see if
> we are finding freed items on the worklist in softdep_disk_io_initiation():
> 
> Index: sys/ufs/ffs/ffs_softdep.c
> ===================================================================
> --- sys/ufs/ffs/ffs_softdep.c   (revision 305057)
> +++ sys/ufs/ffs/ffs_softdep.c   (working copy)
> @@ -9949,6 +9949,13 @@
>  
>         if ((wk = LIST_FIRST(&bp->b_dep)) == NULL)
>                 return;
> +       if (wk->wk_type > D_LAST + 1) {
> +               printf("dead worklist: %p next %p prev %p type %s state
> 0x%X\n",
> +                   wk, wk->wk_list.le_next, wk->wk_list.le_prev,
> +                   TYPENAME(wk->wk_type), wk->wk_state);
> +               LIST_INIT(&bp->b_dep);
> +               return;
> +       }
>         ump = VFSTOUFS(wk->wk_mp);
>  
>         marker.wk_type = D_LAST + 1;    /* Not a normal workitem */
> 
> I would appreciate your running with this patch and reporting if you get the
> "dead worklist" error message.

I do not know what information you want from ddb, and I never thought I'd ever
say this, but "yay, the machine finally crashed again."

I will leave the machine in its current state unless instructed otherwise.

dead worklist: 0xfffffd080edd5400 next 0xdeadc0dedeadc0de prev
0xdeadc0dedeadc0de type ??? state 0xDEADC0
panic: Bad link elm 0xfffffd00b7e0ba00 prev->next != elm

Please let me know what additional information you want.

-- 
You are receiving this mail because:
You are the assignee for the bug.