Le mardi 25 mars 2014 à 15:24 -0500, Alex Elder a écrit :
> On 03/25/2014 03:21 PM, Olivier Bonvalet wrote:
> > Le mardi 25 mars 2014 à 22:18 +0200, Ilya Dryomov a écrit :
> >> On Tue, Mar 25, 2014 at 9:03 PM, Alex Elder <elder@xxxxxxxx> wrote:
> >>> On 03/25/2014 01:53 PM, Olivier Bonvalet wrote:
> >>>> Le mardi 25 mars 2014 à 12:43 -0500, Alex Elder a écrit :
> >>>>> Please try applying this, on top of the previous patch.
> >>>>> If you can then reproduce the problem we'll have a bunch
> >>>>> of new information about the particular request that's
> >>>>> leading to the failure. That might tell us what more we
> >>>>> can do to find the root cause. Thank you.
> >>>>>
> >>>>> -Alex
> >>>>>
> >>>>> PS I hope my mailer doesn't botch the long lines. It might.
> >>>>>
> >>>>
> >>>> Here the execution will continue, no more kernel panic after this
> >>>> debugging display. Is it wanted ?
> >>>
> >>>
> >>> I guess it should panic. I'm glad you mentioned this.
> >>
> >> Just in case, if you haven't done it already: stick rbd_assert(0);
> >> after the last printk in that if statement, so it looks like this:
> >>
> >> if (which != img_request->next_completion) {
> >> printk("%s: bad image object request information:\n", __func__);
> >> printk("obj_request %p\n", obj_request);
> >> printk(" ->object_name <%s>\n", obj_request->object_name);
> >> ...
> >>
> >> printk("img_request %p\n", img_request);
> >> printk(" ->snap 0x%016llx\n", img_request->snap_id);
> >> ...
> >> printk(" ->result %d\n", img_request->result);
> >>
> >> rbd_assert(0);
> >> }
> >>
> >> Thanks,
> >>
> >> Ilya
> >>
> >
> > Without the rbd_assert(0), I add this hang :
> >
> >
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255933] rbd_img_obj_callback: bad image object request information:
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255938] obj_request ffff88025a2b3c48
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255940] ->object_name <rb.0.1536881.238e1f29.000000000439>
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255941] ->offset 0
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255943] ->length 28672
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255944] ->type 0x1
> BIO request
>
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255945] ->flags 0x3
> IMG_DATA, KNOWN
>
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255946] ->which 1
> Second object in the request
>
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255948] ->xferred 28672
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255949] ->result 0
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255950] img_request ffff8802536c4a60
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255952] ->snap 0xffff880257f85ec0
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255953] ->offset 4534026240
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255954] ->length 45056
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255955] ->flags 0x1
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255957] ->obj_request_count 1
> !!! There is only one request... (?)
>
> So obj_request_count might be getting computed incorrectly.
>
> -Alex
>
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255958] ->next_completion 2
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255959] ->xferred 45056
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255960] ->result 0
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255962]
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255962] Assertion failure in rbd_img_obj_callback() at line 2162:
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255962]
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255962] rbd_assert(which < img_request->obj_request_count);
> > Mar 25 21:17:58 murmillia kernel: [ 2205.255962]
> > Mar 25 21:17:58 murmillia kernel: [ 2205.256141] ------------[ cut here ]------------
> > Mar 25 21:17:58 murmillia kernel: [ 2205.256178] kernel BUG at drivers/block/rbd.c:2162!
> >
> >
>
> --
> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
Here I was migrating (xen live migrate) ~20 VM from one host to
"murmillia", just after booting "murmillia" on the new kernel.
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
