Re: [Xen-devel] Issue about domU missing interrupt

[Sander Eikelenboom <linux@xxxxxxxxxxxxxx>]

Wednesday, December 5, 2012, 1:07:07 PM, you wrote:

> On Wed, 5 Dec 2012, Sander Eikelenboom wrote:
>> Wednesday, December 5, 2012, 12:51:13 PM, you wrote:
>> 
>> > On Tue, 4 Dec 2012, Sander Eikelenboom wrote:
>> >> Tuesday, December 4, 2012, 12:01:05 PM, you wrote:
>> >> 
>> >> > On Tue, Dec 04, 2012 at 10:15:58AM +0000, Jan Beulich wrote:
>> >> >> >>> On 04.12.12 at 04:07, "G.R." <firemeteor@xxxxxxxxxxxxxxxxxxxxx> 
>> >> >> >>> wrote:
>> >> >> >>  I had a quick look, and it doesn't look that hard to backport that 
>> >> >> >> patch.
>> >> >> > 
>> >> >> > Thanks, Mat.
>> >> >> > I'm glad to report that the patch do fix my problem.
>> >> >> > 
>> >> >> > And yest it is really easy to port since the code did not change 
>> >> >> > across the
>> >> >> > two release.
>> >> >> > The only change would be line numbers (3841 vs 3803) and one extra 
>> >> >> > comments
>> >> >> > before this line:
>> >> >> > 
>> >> >> >  static int pt_msix_update_one(struct pt_dev *dev, int entry_nr)
>> >> >> > 
>> >> >> > I'm not sure if you are going to release another maintenance version 
>> >> >> > that
>> >> >> > include this patch,
>> >> >> > but I'll report this to Debain maintainer since it's about to freeze 
>> >> >> > for
>> >> >> > v7.0 release and v4.2.0 will not make it.
>> >> >> 
>> >> >> Any thoughts on backporting this to 4.1.x, now or after 4.1.4 is
>> >> >> out?
>> >> >> 
>> >> 
>> >> > It's already in Xen 4.2, it's confirmed to fix a bug in Xen 4.1,
>> >> > so I'd say it should be a candidate for Xen 4.1.4.
>> >> 
>> >> Just tried switching the device model to "qemu-xen", seems this one isn't 
>> >> upstream either.
>> >> (XEN) [2012-12-04 22:49:25] vmsi.c:108:d32767 Unsupported delivery mode 3
>> >> 
>> >> Running xen-unstable as of today, with device-model "qemu-xen" for this 
>> >> hvm guest.
>> >> 
>> 
>> > The patch is supposed to fix a bug affecting msi_translate but in
>> > upstream QEMU we haven't implemented msi_translate at all! So in this
>> > case the cause of the bug (and as a consequence the fix) must be a
>> > different one.

Using msi_translate=0 didn't change a thing, still got "unsupported delivery 
mode" and a non working passthrough device.
lspci in the HVM shows MSI-X enabled for the device, but it doesn't function 
properly.

00:05.0 USB controller: NEC Corporation uPD720200 USB 3.0 Host Controller (rev 
03) (prog-if 30 [XHCI])
        Subsystem: Micro-Star International Co., Ltd. Device 4257
        Physical Slot: 5
        Flags: bus master, fast devsel, latency 0, IRQ 36
        Memory at f3250000 (64-bit, non-prefetchable) [size=8K]
        Capabilities: [50] Power Management version 3
        Capabilities: [70] MSI: Enable- Count=1/1 Maskable- 64bit+
        Capabilities: [90] MSI-X: Enable+ Count=8 Masked-
        Capabilities: [a0] Express Endpoint, MSI 00
        Capabilities: [100] #1033
        Kernel driver in use: xhci_hcd


Disabling MSI completly for the HVM (by using pci=nomsi for the HVM kernel) 
makes passthrough device work ok with normal interrupts.

Looking into qemu-xen-dir/hw i do see xen-msi.c, so there seems to be work done 
in that area ?

>>
>> Ok will see if i can find out what is going on. Probably have to force 
>> msi_translate=0.
>> 
>> I noticed "qemu-xen" doesn't make a log file in /var/log/xen as "qemu-dm" 
>> does, is this expected ?

> No, it is not. You should get the usual /var/log/xen/qemu-dm-domainname.log 
> file.

You were right, all though it lacks a bit in verbosity i guess ... the only 2 
lines i'm getting are:

char device redirected to /dev/pts/17
Issued domain 25 poweroff

Instead of the 99 lines of output when using qemu-xen-traditional.


>> >> Sander
>> >> 
>> >> > -- Pasi
>> >> 
>> >> >> Jan
>> >> >> 
>> >> >> > On Mon, Dec 3, 2012 at 9:58 PM, Mats Petersson 
>> >> >> > <mats.petersson@xxxxxxxxxx>wrote:
>> >> >> > 
>> >> >> >> On 03/12/12 13:19, Mats Petersson wrote:
>> >> >> >>
>> >> >> >>> On 03/12/12 13:14, G.R. wrote:
>> >> >> >>>
>> >> >> >>>> On Mon, Dec 3, 2012 at 6:15 PM, Mats Petersson
>> >> >> >>>> <mats.petersson@xxxxxxxxxx 
>> >> >> > <mailto:mats.petersson@citrix.**com<mats.petersson@xxxxxxxxxx>>>
>> >> >> >>>> wrote:
>> >> >> >>>>
>> >> >> >>>>      On 03/12/12 03:47, G.R. wrote:
>> >> >> >>>>
>> >> >> >>>>          Hi developers,
>> >> >> >>>>          I met some domU issues and the log suggests missing 
>> >> >> >>>> interrupt.
>> >> >> >>>>          Details from here:
>> >> >> >>>>          
>> >> >> >>>> http://www.gossamer-threads.**com/lists/xen/users/263938#** 
>> >> >> >>>> 263938 
>> >> >> >>>> <http://www.gossamer-threads.com/lists/xen/users/263938#263938>
>> >> >> >>>>          In summary, this is the suspicious log:
>> >> >> >>>>
>> >> >> >>>>          (XEN) vmsi.c:122:d32767 Unsupported delivery mode 3
>> >> >> >>>>
>> >> >> >>>>          I've checked the code in question and found that mode 3 
>> >> >> >>>> is an
>> >> >> >>>>          'reserved_1' mode.
>> >> >> >>>>          I want to trace down the source of this mode setting to
>> >> >> >>>>          root-cause the issue.
>> >> >> >>>>          But I'm not an xen developer, and am even a newbie as a 
>> >> >> >>>> xen
>> >> >> >>>> user.
>> >> >> >>>>          Could anybody give me instructions about how to enable
>> >> >> >>>>          detailed debug log?
>> >> >> >>>>          It could be better if I can get advice about experiments 
>> >> >> >>>> to
>> >> >> >>>>          perform / switches to try out etc.
>> >> >> >>>>
>> >> >> >>>>          My SW config:
>> >> >> >>>>          dom0: Debian wheezy (3.6 experimental kernel, xen 
>> >> >> >>>> 4.1.3-4)
>> >> >> >>>>          domU: Debian wheezy 3.2.x stock kernel.
>> >> >> >>>>
>> >> >> >>>>          Thanks,
>> >> >> >>>>          Timothy
>> >> >> >>>>
>> >> >> >>>>      Are you passing hardware (PCI Passthrough) to the HVM guest?
>> >> >> >>>>      What are the exact messages in the DomU?
>> >> >> >>>>
>> >> >> >>>>
>> >> >> >>>> Yes, I'm doing PCI passthrough (the IGD, audio && USB controller).
>> >> >> >>>> But this is actually a PVHVM guest since debian stock kernel has 
>> >> >> >>>> PVOP
>> >> >> >>>> enabled.
>> >> >> >>>> And when I tried another PVOP disabled linux distro (openelec 
>> >> >> >>>> v2.0), I
>> >> >> >>>> did not see such msi related error message.
>> >> >> >>>> Actually, with that domU I do not see anything obvious wrong from 
>> >> >> >>>> the
>> >> >> >>>> log, but I also see nothing from panel (panel receive no signal 
>> >> >> >>>> and go
>> >> >> >>>> power-saving) :-(
>> >> >> >>>>
>> >> >> >>>>
>> >> >> >>>> Back to the issue I was reporting, the domU log looks like this:
>> >> >> >>>>
>> >> >> >>>> Dec 2 21:52:44 debvm kernel: [ 1085.604071]
>> >> >> >>>> [drm:i915_hangcheck_ring_idle]
>> >> >> >>>> *ERROR* Hangcheck timer elapsed... blt ring idle [waiting on 
>> >> >> >>>> 3354, at
>> >> >> >>>> 3354], missed IRQ?
>> >> >> >>>> Dec 2 21:56:50 debvm kernel: [ 1332.076071]
>> >> >> >>>> [drm:i915_hangcheck_ring_idle]
>> >> >> >>>> *ERROR* Hangcheck timer elapsed... render ring idle [waiting on 
>> >> >> >>>> 11297, at
>> >> >> >>>> 11297], missed IRQ?
>> >> >> >>>> Dec 2 22:32:48 debvm kernel: [ 7.220073] hda-intel: 
>> >> >> >>>> azx_get_response
>> >> >> >>>> timeout, switching to polling mode: last cmd=0x000f0000
>> >> >> >>>> Dec 2 22:45:32 debvm kernel: [ 776.392084] hda-intel: No response 
>> >> >> >>>> from
>> >> >> >>>> codec, disabling MSI: last cmd=0x002f0600
>> >> >> >>>> Dec 2 22:45:33 debvm kernel: [ 777.400075] hda_intel: 
>> >> >> >>>> azx_get_response
>> >> >> >>>> timeout, switching to single_cmd mode: last cmd=0x002f0600
>> >> >> >>>>
>> >> >> >>>>
>> >> >> >>>> Thanks,
>> >> >> >>>> Timothy
>> >> >> >>>>
>> >> >> >>> It does sound like there is a fix in 4.2.0, as indicated by Jan, 
>> >> >> >>> that
>> >> >> >>> fixes this. I'm not fully clued up to what the policy for 
>> >> >> >>> backporting
>> >> >> >>> fixes are, and I haven't looked at the complexity of the fix 
>> >> >> >>> itself, but
>> >> >> >>> either updating to the 4.2.0 or a (personal) backport sounds like 
>> >> >> >>> the
>> >> >> >>> right solution here.
>> >> >> >>>
>> >> >> >> I had a quick look, and it doesn't look that hard to backport that 
>> >> >> >> patch.
>> >> >> >>
>> >> >> >> --
>> >> >> >> Mats
>> >> >> >>
>> >> >> >>
>> >> >> >>> Unfortunately, I hadn't seen Jan's reply by the time I wrote my 
>> >> >> >>> response
>> >> >> >>> to your original email.
>> >> >> >>>
>> >> >> >>> --
>> >> >> >>> Mats
>> >> >> >>>
>> >> >> >>> ______________________________**_________________
>> >> >> >>> Xen-devel mailing list
>> >> >> >>> Xen-devel@xxxxxxxxxxxxx 
>> >> >> >>> http://lists.xen.org/xen-devel 
>> >> >> >>>
>> >> >> >>>
>> >> >> >>>
>> >> >> >>
>> >> >> >> ______________________________**_________________
>> >> >> >> Xen-devel mailing list
>> >> >> >> Xen-devel@xxxxxxxxxxxxx 
>> >> >> >> http://lists.xen.org/xen-devel 
>> >> >> >>
>> >> >> 
>> >> >> 
>> >> >> 
>> >> >> _______________________________________________
>> >> >> Xen-devel mailing list
>> >> >> Xen-devel@xxxxxxxxxxxxx
>> >> >> http://lists.xen.org/xen-devel
>> >> 
>> >> 
>> >> 
>> 
>> 



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel