Re: [Xen-devel] [PATCH] iommu: leave IOMMU enabled by default during kexec crash transition

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 20.02.19 at 20:05, <igor.druzhinin@xxxxxxxxxx> wrote:
> On 20/02/2019 08:55, Jan Beulich wrote:
>> 
>> Everything, absolutely everything is possible as a cause for a crash.
>> I don't see why device isolation would matter here at all. Page table
>> corruption (be it IOMMU or CPU one) can be caused by
>> malfunctioning kernel code, entirely unrelated to what devices do.
>> 
> 
> Yes, I blindly assumed that stray DMA is the most common source of
> kernel data corruption. Unfortunately, we don't really have a choice now
> - almost all platforms these days raise some form of an error in the
> event of unhanded translation fault (at least most of those we have now
> in our labs). So it's either slight chance of entering the crash kernel
> with corrupted IOMMU pages or not entering it at all in most of the cases.

And I'm fine with this, as long as the new (even if smaller) risk is
mentioned in the description.

Jan



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