Re: [Qemu-devel] [PULL 00/47] virtio, vhost, pc, pci: tests, documentation, fixes and cleanups

[John Fastabend]

On 16-11-10 03:44 PM, Alex Williamson wrote:
> On Fri, 11 Nov 2016 01:09:05 +0200
> "Michael S. Tsirkin" <address@hidden> wrote:
> 
>> On Thu, Nov 10, 2016 at 03:48:28PM -0700, Alex Williamson wrote:
>>> On Fri, 11 Nov 2016 00:33:17 +0200
>>> "Michael S. Tsirkin" <address@hidden> wrote:
>>>   
>>>> On Thu, Nov 10, 2016 at 03:29:21PM -0700, Alex Williamson wrote:  
>>>>> On Thu, 10 Nov 2016 18:12:20 +0200
>>>>> "Michael S. Tsirkin" <address@hidden> wrote:
>>>>>     
>>>>>> The following changes since commit 
>>>>>> 6bbcb76301a72dc80c8d29af13d40bb9a759c9c6:
>>>>>>
>>>>>>   MAINTAINERS: Remove obsolete stable branches (2016-11-10 15:29:59 
>>>>>> +0000)
>>>>>>
>>>>>> are available in the git repository at:
>>>>>>
>>>>>>   git://git.kernel.org/pub/scm/virt/kvm/mst/qemu.git tags/for_upstream
>>>>>>
>>>>>> for you to fetch changes up to 8038753b86f4cb1e79d4225a799395c4dae96b17:
>>>>>>
>>>>>>   docs: add PCIe devices placement guidelines (2016-11-10 18:08:06 +0200)
>>>>>>
>>>>>> ----------------------------------------------------------------
>>>>>> virtio, vhost, pc, pci: tests, documentation, fixes and cleanups
>>>>>>
>>>>>> Lots of fixes all over the place.  I allowed some cleanups in even 
>>>>>> though they
>>>>>> are not strictly bugfixes, they might prevent bugs and seem very safe.
>>>>>>
>>>>>> Most importantly, this fixes a regression with vhost introduced
>>>>>> by the last pull.    
>>>>>
>>>>> I think I'm hitting this previous regression, I see this in my libvirt
>>>>> log:
>>>>>
>>>>> kvm_mem_ioeventfd_add: error adding ioeventfd: File exists
>>>>>
>>>>> And a backtrace as seen here
>>>>> https://paste.fedoraproject.org/477562/88144131/
>>>>>
>>>>> So I merge this tag to try to resolve it, now I get qemu segfaulting
>>>>> bisected to:
>>>>>
>>>>> commit a6d8372bc6764ee279b473d13ff4ecc8acb7a978
>>>>> Author: Cao jin <address@hidden>
>>>>> Date:   Sat Nov 5 10:07:21 2016 +0800
>>>>>
>>>>>     hcd-xhci: check & correct param before using it  
>>>
>>> Here's the backtrace from this one:
>>>
>>> #0  0x0000555555a99d90 in xhci_running (xhci=0x0) at hw/usb/hcd-xhci.c:824
>>> #1  0x0000555555a9f092 in xhci_port_notify (port=0x7fffbfb39330, 
>>> bits=131072)
>>>     at hw/usb/hcd-xhci.c:2870
>>> #2  0x0000555555a9f221 in xhci_port_update (port=0x7fffbfb39330, 
>>> is_detach=0)
>>>     at hw/usb/hcd-xhci.c:2905
>>> #3  0x0000555555a9f435 in xhci_reset (dev=0x7fffbfb38010)
>>>     at hw/usb/hcd-xhci.c:2961
>>> #4  0x0000555555984a2e in device_reset (dev=0x7fffbfb38010)
>>>     at hw/core/qdev.c:1145
>>> #5  0x00005555559828db in qdev_reset_one (dev=0x7fffbfb38010, opaque=0x0)
>>>     at hw/core/qdev.c:295
>>> #6  0x00005555559834df in qdev_walk_children (dev=0x7fffbfb38010, 
>>> pre_devfn=0x0, pre_busfn=0x0, post_devfn=0x5555559828bf <qdev_reset_one>, 
>>> post_busfn=0x5555559828e2 <qbus_reset_one>, opaque=0x0) at 
>>> hw/core/qdev.c:610
>>> #7  0x000055555598763b in qbus_walk_children (bus=0x555556a40f30, 
>>> pre_devfn=0x0, pre_busfn=0x0, post_devfn=0x5555559828bf <qdev_reset_one>, 
>>> post_busfn=0x5555559828e2 <qbus_reset_one>, opaque=0x0) at hw/core/bus.c:59
>>> #8  0x00005555559834a3 in qdev_walk_children (dev=0x555556a3f220, 
>>> pre_devfn=0x0, pre_busfn=0x0, post_devfn=0x5555559828bf <qdev_reset_one>, 
>>> post_busfn=0x5555559828e2 <qbus_reset_one>, opaque=0x0) at 
>>> hw/core/qdev.c:602
>>> #9  0x000055555598763b in qbus_walk_children (bus=0x5555568296c0, 
>>> pre_devfn=0x0, pre_busfn=0x0, post_devfn=0x5555559828bf <qdev_reset_one>, 
>>> post_busfn=0x5555559828e2 <qbus_reset_one>, opaque=0x0) at hw/core/bus.c:59
>>> #10 0x00005555559829f5 in qbus_reset_all (bus=0x5555568296c0)
>>>     at hw/core/qdev.c:321
>>> #11 0x0000555555982a18 in qbus_reset_all_fn (opaque=0x5555568296c0)
>>>     at hw/core/qdev.c:327
>>> #12 0x00005555558e76f0 in qemu_devices_reset () at vl.c:1765
>>> #13 0x00005555558367ce in pc_machine_reset ()
>>>     at /net/gimli/home/alwillia/Work/qemu.git/hw/i386/pc.c:2181
>>> #14 0x00005555558e778d in qemu_system_reset (report=false) at vl.c:1778
>>> #15 0x00005555558ef44b in main (argc=50, argv=0x7fffffffdf48, 
>>> envp=0x7fffffffe0e0) at vl.c:4656
>>>
>>> Here's a commandline sufficient to trigger it:
>>>
>>> qemu-system-x86_64 -m 1G -nodefaults -no-user-config -display none -monitor 
>>> stdio -device nec-usb-xhci
>>>
>>> So apparently that never got tested or something got mangled in the
>>> commit.  Thanks,
>>>
>>> Alex  
>>
>> On the original tree (without this mangled pull), does one of
>> [PATCH] vhost-scsi: Update 'ioeventfd_started' with host notifiers
>> and
>> [PATCH v2] vhost: Update 'ioeventfd_started' with host notifiers
> 
> The latter I already have from your tag as f99d0d394758, adding the
> former on top of your tag does not resolve the assert (I simply removed
> the xhci device from the vm config to avoid the segfault).  Thanks,
> 
> Alex
> 

hmm with those patches it doesn't even get past boot for me.

qemu-system-x86_64: /home/john/git/qemu/memory.c:2012:
memory_region_del_eventfd: Assertion `i != mr->ioeventfd_nb' failed.
Aborted (core dumped)

I need to run out for a bit but can look at it more tomorrow if needed.

.John