[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [Xen-devel] MSI causing softpanics in guest



Hi, Keir,

This is an RFC patch, which is not tested now (It will cost some time to setup 
my test environments). I want to ensure I am on the same page with you. So 
could you pleas have a look and point to me which part I may be missing?
Thanks!

Shan Haitao

-----Original Message-----
From: Keir Fraser [mailto:keir.fraser@xxxxxxxxxxxxx]
Sent: 2008年9月24日 14:05
To: Shan, Haitao; Anish Bhatt
Cc: Jan Beulich; xen-devel@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Xen-devel] MSI causing softpanics in guest

So, which of us is going to patch pcifront? :-)

 -- Keir

On 24/9/08 02:24, "Shan, Haitao" <haitao.shan@xxxxxxxxx> wrote:

> It is supposed to work on dom0. This is a bug.
> Please refer to Keir's reply in this mail thread.
>
> Shan Haitao
>
> -----Original Message-----
> From: Anish Bhatt [mailto:anish@xxxxxxxxxxxxx]
> Sent: 2008年9月24日 8:08
> To: Shan, Haitao
> Cc: Jan Beulich; xen-devel@xxxxxxxxxxxxxxxxxxx; Keir Fraser
> Subject: Re: [Xen-devel] MSI causing softpanics in guest
>
> No, the bug_on is not triggered in dom0. I also saw the following error
> message in xm dmesg, :
> /physdev.c:90:d0 remap pirq fa vector 49 while not unmap/
> Its part of the code added for MSI, I wonder if its of any significance.
>
> -Anish
>
> Shan, Haitao wrote:
>> Just some thoughts on Anish's problem. It seems he reveals a bug in current
>> code, I think.
>>
>> Currently, evnchn_map_pirq is hooked on msi_map_vector
>> (pci_enable_msi->msi_capability_init->msi_map_vector). Here the pirq type is
>> set to IRQT_PIRQ. Then request_irq should succeed since pirq type is already
>> set.
>>
>> However, this path only works on the assumption that msi_map_vector and
>> request_irq are executed in the same domain, to be more specific, dom0.
>> For PV domU guest, there is no msi_map_vector in its pci_enable_msi ( PV domU
>> actually asks dom0 to do the msi map, instead. This is a decision made by
>> Keir and Yunhong long ago. I think it is base on security considerations). So
>> its irq type is not set correctly (What's more, it incorrectly sets the
>> dom0's irq type). Then request_irq can trigger this bug_on().
>>
>> Anish, if you use the device in dom0 itself, does it also have this bug_on?
>>
>> Keir and Jan,
>> Do you think this explanation is reasonable for issues Anish met?
>>
>> Best Regards
>> Shan Haitao
>>
>> -----Original Message-----
>> From: Jan Beulich [mailto:jbeulich@xxxxxxxxxx]
>> Sent: 2008年9月23日 15:24
>> To: Anish Bhatt
>> Cc: Keir Fraser; Shan, Haitao; xen-devel@xxxxxxxxxxxxxxxxxxx
>> Subject: Re: [Xen-devel] MSI causing softpanics in guest
>>
>> So the question is what irq you pass in to request_irq() and where you got
>> it from. Jan
>>
>>
>>>>> Anish Bhatt <anish@xxxxxxxxxx> 23.09.08 07:04 >>>
>>>>>
>> I put a printk before BUG_ON, Its expecting IRQT_PIRQ, but getting
>> IRQT_UNBOUND instead.
>> -Anish
>>
>> Jan Beulich wrote:
>>
>>>>>> Keir Fraser <keir.fraser@xxxxxxxxxxxxx> 20.09.08 10:15 >>>
>>>>>>
>>>>>>
>>>> On 3.3.0 you need to specify msi=1 on Xen's command line to enable MSI. In
>>>> xen-unstable, MSI support is always enabled.
>>>>
>>>> I changed a check in evtchn_get_xen_pirq() into a BUG_ON() as it looks to
>>>> me
>>>> like it should never trigger. Jan Beulich was the original authour of that
>>>> function -- cc'ed in case he can indicate whether I've actually broken the
>>>> function. :-)
>>>>
>>>>
>>> No, I think that BUG_ON() you added is valid. It definitely is in the
>>> context
>>> of startup_pirq(). I'd be curious to know what the type of that irq really
>>> is,
>>> but that cannot be determined from the backtrace alone. Anish, could you
>>> perhaps add a simple printk() before the BUG_ON()? Unfortunately the
>>> driver in question does not appear to be part of the tree, so we can't even
>>> check what it does prior to calling request_irq()...
>>>
>>>  -- Keir
>>>
>>> On 19/9/08 20:53, "Anish Bhatt" <anish@xxxxxxxxxxxxx> wrote:
>>>
>>>
>>>
>>>> lspci shows MSI enabled for PCI device. PCI passthrough works fine.
>>>> However, as soon as the MSI driver for card is insmodded, kernel panics.
>>>> This is on xen-unstable.  Tried the same with xen-3.3.0 which is
>>>> supposed to have MSI passthrough, but the same guest shows MSI as disabled.
>>>> Any else seen this bug, or know of a workaround ?
>>>>
>>>> Trace is as follows :
>>>>
>>>> ------------[ cut here ]------------
>>>> kernel BUG at
>>>>
>>>>
>>>>
>>> /usr/src/xen/xen-unstable.hg/linux-2.6.18-xen.hg/drivers/xen/core/evtchn.c:8
>>> 09>
>>> !
>>>
>>>
>>>> invalid opcode: 0000 [#1]
>>>> SMP
>>>> Modules linked in: nfemsg nfdvnet ipv6 binfmt_misc dm_mod nfe usbcore
>>>> ext3 jbd processor fuse
>>>> CPU:    0
>>>> EIP:    0061:[<c02487f5>]    Tainted: GF     VLI
>>>> EFLAGS: 00210097   (2.6.18.8-xen #2)
>>>> EIP is at evtchn_get_xen_pirq+0x35/0x40
>>>> eax: ffffffff   ebx: 00000002   ecx: c0372e60   edx: 00000000
>>>> esi: c2103560   edi: c03d3080   ebp: 000004f9   esp: ed385dac
>>>> ds: 007b   es: 007b   ss: 0069
>>>> Process modprobe (pid: 2590, ti=ed384000 task=ed7b1100 task.ti=ed384000)
>>>> Stack: c0248aef 00000000 00000000 00000000 00000000 00000000 00000000
>>>> 00000000
>>>>        00000000 00000000 00000000 00000000 00000000 00000000 00000000
>>>> 00000000
>>>>        00000000 00000000 00000000 00000000 00000000 00000000 00000000
>>>> 00000000
>>>> Call Trace:
>>>>  [<c0248aef>] startup_pirq+0x3f/0x250
>>>>  [<c0150b50>] setup_irq+0x160/0x1b0
>>>>  [<ee0dd270>] nfe_interrupt_handler+0x0/0x30 [nfemsg]
>>>>  [<c0150c43>] request_irq+0xa3/0xc0
>>>>  [<ee02c8ad>] nfemsg_module_init+0x8ad/0x133e [nfemsg]
>>>>  [<c030531b>] cond_resched+0x2b/0x40
>>>>  [<c0305369>] wait_for_completion+0x19/0xf0
>>>>  [<c0142818>] sys_init_module+0x148/0x1b50
>>>>  [<c010595f>] syscall_call+0x7/0xb
>>>> Code: 00 00 89 d1 d3 e0 85 05 c4 2c 42 c0 74 1a 8b 14 95 c0 c0 44 c0 89
>>>> d0 c1 e8 1c 83 e8 01 75 0c c1 ea 0c 81 e2 ff ff 00 00 89 d0 c3 <0f> 0b
>>>> 29 03 a4 f1 32 c0 eb ea 90 83 ec 08 89 74 24 04 89 c6 a1
>>>> EIP: [<c02487f5>] evtchn_get_xen_pirq+0x35/0x40 SS:ESP 0069:ed385dac
>>>>  ### card [0] start: host progs ###
>>>> -bash-3.2#
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: ------------[ cut here ]------------
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: kernel BUG at
>>>>
>>>>
>>>>
>>> /usr/src/xen/xen-unstable.hg/linux-2.6.18-xen.hg/drivers/xen/core/evtchn.c:8
>>> 09>
>>> !
>>>
>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: invalid opcode: 0000 [#1]
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: SMP
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: CPU:    0
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: EIP is at evtchn_get_xen_pirq+0x35/0x40
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: eax: ffffffff   ebx: 00000002   ecx: c0372e60   edx: 00000000
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: esi: c2103560   edi: c03d3080   ebp: 000004f9   esp: ed385dac
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: ds: 007b   es: 007b   ss: 0069
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: Process modprobe (pid: 2590, ti=ed384000 task=ed7b1100
>>>> task.ti=ed384000)
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: Stack: c0248aef 00000000 00000000 00000000 00000000 00000000
>>>> 00000000 00000000
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel:        00000000 00000000 00000000 00000000 00000000 00000000
>>>> 00000000 00000000
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel:        00000000 00000000 00000000 00000000 00000000 00000000
>>>> 00000000 00000000
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: Call Trace:
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: Code: 00 00 89 d1 d3 e0 85 05 c4 2c 42 c0 74 1a 8b 14 95 c0 c0
>>>> 44 c0 89 d0 c1 e8 1c 83 e8 01 75 0c c1 ea 0c 81 e2 ff ff 00 00 89 d0 c3
>>>> <0f> 0b 29 03 a4 f1 32 c0 eb ea 90 83 ec 08 89 74 24 04 89 c6 a1
>>>>
>>>> Message from syslogd@drake at Sep 19 15:36:44 ...
>>>>  kernel: EIP: [<c02487f5>] evtchn_get_xen_pirq+0x35/0x40 SS:ESP
>>>> 0069:ed385dac
>>>>
>>>>
>>>
>>>
>>
>>
>> --
>> As long as the music's loud enough, we won't hear the world falling apart.
>>
>>
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@xxxxxxxxxxxxxxxxxxx
>> http://lists.xensource.com/xen-devel
>>
>
>
> --
> As long as the music's loud enough, we won't hear the world falling apart.
>


Attachment: fixup_translate_xen_pirqs.patch
Description: fixup_translate_xen_pirqs.patch

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.