Re: [Xen-devel] [PATCH v3] IOMMU: make DMA containment of quarantined devices optional

[Jan Beulich <jbeulich@suse.com>]

On 10.03.2020 13:30, Paul Durrant wrote:
>> -----Original Message-----
>> From: Jan Beulich <jbeulich@suse.com>
>> Sent: 10 March 2020 10:27
>> To: Tian, Kevin <kevin.tian@intel.com>; Paul Durrant <paul@xen.org>
>> Cc: xen-devel@lists.xenproject.org; Andrew Cooper <andrew.cooper3@citrix.com>
>> Subject: Re: [PATCH v3] IOMMU: make DMA containment of quarantined devices optional
>>
>> On 10.03.2020 04:43, Tian, Kevin wrote:
>>>> From: Jan Beulich <jbeulich@suse.com>
>>>> Sent: Monday, March 9, 2020 7:09 PM
>>>>
>>>> I'm happy to take better suggestions to replace the "full" command line
>>>> option and Kconfig prompt tokens. I don't think though that "fault" and
>>>> "write-fault" are really suitable there.
>>>
>>> I think we may just allow both r/w access to scratch page for such bogus
>>> device, which may make 'full' more reasonable since we now fully
>>> contain in-fly DMAs. I'm not sure about the value of keeping write-fault
>>> alone for such devices (just because one observed his specific device only
>>> has problem with read-fault).
>>
>> Well, a fundamental problem I have here is that I still don't know
>> the _exact_ conditions for the observed hangs. I consider it unlikely
>> for IOMMU read faults to cause hangs, but for write faults to be
>> "fine".
> 
> AFAIK it's because the writes are posted and so any faults are just ignored, whereas a read fault being synchronous causes the device's state machine to lock up. It really is observed behaviour.
> 
>> It would seem more likely to me that e.g. a non-present
>> context entry might cause issues. If that was the case, we wouldn't
>> need to handle reads and writes differently; we could instead install
>> an all zero top level page table. And we'd still get all faults that
>> are supposed to surface. But perhaps Paul did try this back then, and
>> it turned out to not be an option.
>>
> 
> The only info I had was that faults on DMA reads had to avoided
> completely. I did not have access to the h/w in question at the
> time. I may be able to get it now.

I see. The implication then is, as Kevin said, that we mustn't run
guests with _any_ IOMMU PTEs having their "read" bits clear.
Anything that's "not present" now would need directing to a scratch
page. I then further wonder what effect reads to addresses beyond
AGAW would have. It may be impossible to arrange for sufficiently
secure pass-through with such a device, at which point - again as
said by Kevin - there may be little point in the scratch page
based quarantining.

Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel