Re: xen-unstabel + linux 4.2: MMIO emulation failed: d23v0 64bit @ 0010:ffffffff814e2b1c -> 66 89 02 48 8d 55 c0 48 89 5d c0 44 89 65 c8 e8

[Sander Eikelenboom <linux@eikelenboom.it>]

Wednesday, July 29, 2015, 11:06:37 AM, you wrote:

>> -----Original Message-----
>> From: linux@eikelenboom.it [mailto:linux@eikelenboom.it]
>> Sent: 28 July 2015 22:16
>> To: Andrew Cooper
>> Cc: xen-devel@lists.xen.org; Paul Durrant
>> Subject: Re: [Xen-devel] xen-unstabel + linux 4.2: MMIO emulation failed:
>> d23v0 64bit @ 0010:ffffffff814e2b1c -> 66 89 02 48 8d 55 c0 48 89 5d c0 44 89
>> 65 c8 e8
>> 
>> On 2015-07-27 16:07, Andrew Cooper wrote:
>> > On 24/07/15 19:56, linux@eikelenboom.it wrote:
>> >> Hi All,
>> >>
>> >> On my AMD system running xen-unstable (last commit: ),
>> >>  after a few restarts of a HVM guest with pci-passthrough i got these
>> >> on shutdown of the guest:
>> >> (never seen this before, so it should be something triggered by a
>> >> recent commit)
>> >>
>> >> --
>> >> Sander
>> >>
>> >>
>> >> .... (probably lost before but that's lost)
>> >>
>> >> (XEN) [2015-07-24 18:46:53.732] domain_crash called from io.c:166
>> >> (XEN) [2015-07-24 18:46:53.732] io.c:165:d23v0 Weird HVM ioemulation
>> >> status 1.
>> >> (XEN) [2015-07-24 18:46:53.732] domain_crash called from io.c:166
>> >> (XEN) [2015-07-24 18:46:53.732] io.c:165:d23v0 Weird HVM ioemulation
>> >> status 1.
>> >> (XEN) [2015-07-24 18:46:53.732] domain_crash called from io.c:166
>> >> (XEN) [2015-07-24 18:46:53.732] io.c:165:d23v0 Weird HVM ioemulation
>> >> status 1.
>> >
>> > Paul: this is very likely an issue your emulation series.
>> >
>> > 66 89 02 is mov %ax,(%rdx), but has ended up in handle_pio() which
>> > seems
>> > wrong.
>> >
>> > Sander: Please can you rerun with the following debug
>> >
>> > diff --git a/xen/arch/x86/hvm/io.c b/xen/arch/x86/hvm/io.c
>> > index d3b9cae..7560d08 100644
>> > --- a/xen/arch/x86/hvm/io.c
>> > +++ b/xen/arch/x86/hvm/io.c
>> > @@ -163,7 +163,9 @@ int handle_pio(uint16_t port, unsigned int size,
>> > int
>> > dir)
>> >          break;
>> >      default:
>> >          gdprintk(XENLOG_ERR, "Weird HVM ioemulation status %d.\n",
>> > rc);
>> > -        domain_crash(curr->domain);
>> > +        show_execution_state(curr);
>> > +        dump_execution_state();
>> > +        domain_crash_synchronous(curr->domain);
>> >          break;
>> >      }
>> >
>> > ~Andrew
>> 
>> 
>> Hi Andrew,
>> 
>> I tried your debug patch, however that doesn't print anything new to xl
>> dmesg.
>> It seems to be too busy trying to crash .. (although that actually seems
>> to fail)
>> 
>> This was testing on a tree with a619657e2d5725a61cb72f65901db42e98f2a896
>> as last commit and it fails at the first shutdown of the guest.
>> 
>> After that i tried it with the current staging tree with
>> 44313ab77f3e3c5b566ea4f23b0e32bfd5eafa29 as last commit
>> and i don't to get those failures anymore .. i even done a start
>> shutdown loop of 100 iterations .. everything was fine.
>> 

> Good. I can't see anything particularly relevant in the range between those two commits, but it doesn't sound like it's anything to do with my emulation series.

>   Paul

Well one thing i still find weird is why didn't the domain crash properly ?


>> --
>> Sander