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

Re: [Xen-devel] [PATCH RFC] VT-d: honor firmware-first mode in XSA-59 workaround code



Jan Beulich wrote on 2014-05-23:
>>>> On 23.05.14 at 08:46, <yang.z.zhang@xxxxxxxxx> wrote:
>> Jan Beulich wrote on 2014-05-23:
>>>>>> On 23.05.14 at 04:32, <yang.z.zhang@xxxxxxxxx> wrote:
>>>> Jan Beulich wrote on 2014-05-22:
>>>>> @@ -447,18 +448,26 @@ void pci_vtd_quirk(const struct pci_dev
>>>>>          }
>>>>>          
>>>>>          val = pci_conf_read32(seg, bus, dev, func, pos +
>>>>> PCI_ERR_UNCOR_MASK); -        pci_conf_write32(seg, bus, dev, func,
>>>>> pos + PCI_ERR_UNCOR_MASK, -                         val |
>>>>> PCI_ERR_UNC_UNSUP); -        val = pci_conf_read32(seg, bus, dev,
>>>>> func, pos + PCI_ERR_COR_MASK); -        pci_conf_write32(seg, bus,
>>>>> dev, func, pos + PCI_ERR_COR_MASK, -                         val |
>>>>> PCI_ERR_COR_ADV_NFAT); +        val2 = pci_conf_read32(seg, bus,
>>>>> dev, func, pos + PCI_ERR_COR_MASK); +        if ( (val &
>>>>> PCI_ERR_UNC_UNSUP) && (val2 & PCI_ERR_COR_ADV_NFAT) ) +           
>>>>> action = "Found masked";
>>>> 
>>>> What happened if dom0 unmasked it later?
>>> 
>>> That question you should have raised on the first of the XSA-59
>>> related patch sets, but the answer is we expect Dom0 to be well
>>> behaved. Of course, if you have a non-intrusive suggestion on how
>>> to enforce
>> this in Xen, I'm all ears.
>> 
>> So you mean we need to stare at Linux upstream to make sure it
>> doesn't do any un-friendly modification to Xen.
> 
> Not sure what you mean by "stare", but relying on Dom0 to not do bad
> things is a fundamental model with Xen - if the Dom0 kernel wanted,
> it'd have other ways of doing bad things.

From dom0's point, hypervisor is similar to a firmware especially in this case. 
So if hypervisor want to disable the AER, it can act as a firmware to set the 
firmware first mode. Then it is safe to mask it and dom0 will never try to 
touch it. This is just an idea and may introduce side effect. Need more 
thinking.

> 
> Jan


Best regards,
Yang



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


 


Rackspace

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