On 8/13/16 5:45 AM, Warner Losh wrote:
> On Fri, Aug 12, 2016 at 3:06 PM, Ngie Cooper <yaneurab...@gmail.com> wrote:
>> On Fri, Aug 12, 2016 at 8:23 AM, Bryan Drewery <bdrew...@freebsd.org> wrote:
>>
>> ...
>>
>>> No, I was missing r303637.  Hard to say if it is related... Andrey says
>>> it's not.  I haven't dived into it yet and it's so far only happened
>>> once (out of a few tests).  We do have various customizations but I'm
>>> inclined to think it's the stock code having problems.
>>
>> I hit it again on our internal vendor tree (what bdrewery was noting
>> previously). I can provide some details about the panic if need be.
> 
> Ideally, you'd reproduce it on an unmodified -current. In a vendor tree 
> there's
> many chances that something else is odd or off that would make it hard for
> the community to help.
> 
> Warner
> 

#4  0xffffffff80477b82 in g_resize_provider_event
(arg=0xfffff80007f8d3a0, flag=<optimized out>) at
/b/mnt/src/sys/geom/geom_subr.c:668
668             KASSERT(!(pp->flags & G_PF_WITHER),
(gdb) p pp->mediasize
$2 = 0

It seems that r303637 really does cover the panic we're hitting since
the g_resize_provider_event was posted with mediasize==0 which r303637
avoids.

-- 
Regards,
Bryan Drewery

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to