Re: [Xen-devel] [PATCH v2 08/11] IOMMU: propagate IOMMU Device-TLB flush error up to iommu_iotlb_flush{, _all} (leaf ones).

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 26.04.16 at 13:50, <quan.xu@xxxxxxxxx> wrote:
> On April 25, 2016 7:40 PM, Jan Beulich <JBeulich@xxxxxxxx> wrote:
>> >>> On 18.04.16 at 16:00, <quan.xu@xxxxxxxxx> wrote:
>> > While IOMMU Device-TLB flush timed out, xen calls panic() at present.
>> > However the existing panic() is going to be eliminated, so we must
>> > propagate the IOMMU Device-TLB flush error up to the
>> > iommu_iotlb_flush{,_all}.
>> >
>> > If IOMMU mapping and unmapping failed, the domain (with the exception
>> > of the hardware domain) is crashed, treated as a fatal error. Rollback
>> > can be lighter weight.
>> >
>> > This patch fixes the leaf ones.
>> >
>> > Signed-off-by: Quan Xu <quan.xu@xxxxxxxxx>
>> 
>> Looks okay except for the again missing uses of __must_check on function
>> declarations or, where needed, definitions.
> 
> This patch modifies the common part  'iommu_ops', but not to
> fix the ARM/AMD callbacks ( it seems the callbacks are not initialized for 
> AMD code, but ARM does).
> I need to fix ARM callbacks as well.
> 
> 
>  I wonder whether it is good to use  __must_check on these callbacks or not.

For callbacks it's indeed questionable.

>  IMO I'd better use __must_check on functions, i.e. iommu_iotlb_flush_all(), 
>  invoking these callbacks,  instead of these callbacks. 

At least the two functions in xen/drivers/passthrough/iommu.c that
this patch alters need to gain __must_check annotations (if they
don't already in an earlier patch - that's simply not possible to tell on
this version, which fails to add any such annotations).

Jan


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