Re: [Xen-devel] [PATCH-for-4.13] x86/mm: don't needlessly veto migration

[Jan Beulich <jbeulich@xxxxxxxx>]

On 01.10.2019 13:08, Andrew Cooper wrote:
> On 01/10/2019 09:46, Jan Beulich wrote:
>> On 01.10.2019 10:28, Paul Durrant wrote:
>>> Now that xl.cfg has an option to explicitly enable IOMMU mappings for a
>>> domain, migration may be needlessly vetoed due to the check of
>>> is_iommu_enabled() in paging_log_dirty_enable().
>>> There is actually no need to prevent logdirty from being enabled unless
>>> devices are assigned to a domain and that domain is sharing HAP mappings
>>> with the IOMMU (in which case disabling write permissions in the P2M may
>>> cause DMA faults).
>> But that's taking into account only half of the reason of the
>> exclusion. The other half is that assigned devices may cause pages
>> to be dirtied behind the back of the log-dirty logic.
> 
> No.  Very specifically not.
> 
> Xen cannot correctly evaluate whether a domain is migrateable or not,
> and therefore is unable to make a correct veto decision.

Then it should be put in the position, shouldn't it? Like with the
suggested reporting of what devices have log-dirty handling taken care
of (which, to take into account a remark by Paul, I think should go
without saying for emulated devices).

> The existing DOMCTL_disable_migrate is bogus and I'm in the process of
> purging it from Xen.

That domctl is pretty much orthogonal I think.

> Also, I seem to recall nacking this change originally, specifically on
> the basis that it breaks live migration with GPUs, so I'm rather
> confused to find the code present...

I've gone and checked - there's no response by you that I could find in
the list archives on any of the v2 through v4 submissions by Roger (in
the April-June 2014 time frame; v1 looks to not have had this change yet).
Hence it became commit 37201c6203.

Jan

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