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

Re: [Xen-devel] Regression PCI passthrough from 4.5.5 to 4.6.0-rc1

>>> On 10.07.17 at 18:56, <ml-ak@xxxxxxxxx> wrote:
> Hello Jan, Pasi, all
>>> I noticed that PCI passthrough for an LSI SAS HBA 9211 did not longer work 
> (at least under Windows) when using Xen 4.8.1.
>>> I then bisected through various released versions and finally I narrowed it 
> down to
>>> 4.5.5 (with qemu from Xen 4.6.5) -> working
>>> 4.6.0-rc1 (with qemu from Xen 4.6.5) -> no longer working
>> So can you please bisect which exact commit between Xen 4.5 and 4.6 causes 
> the problem?
> Initially I did not bisect through unstable code as I expected to hit broken 
> intermediate snapshots. But finally it worked.
> I found that commit 
> http://xenbits.xen.org/gitweb/?p=xen.git;a=commit;h=ad28e42bd1d28d746988ed716 
> 54e8aa670629753 is causing the problem (x86/MSI: track host and guest masking 
> separately) which appeared on the mailing list already:
> https://lists.xenproject.org/archives/html/xen-devel/2015-06/msg03923.html 
> https://lists.xenproject.org/archives/html/xen-devel/2016-04/msg00028.html 
> Commit 
> http://xenbits.xen.org/gitweb/?p=xen.git;a=commit;h=84d6add5593d865736831d150 
> da7c38588f669f6 does not fix it.

So it's possibly a different issue from at least what Sander had seen,
as that change has been tested by him successfully.

> Attached is some first debug info.

What I first of all notice is that there's no interrupt at all which is
bound to either Dom1 or Dom2 in the debug key output. qemu
logs may therefore also be necessary, likely even with verbosity
first increased in the sources (by #define-ing
XEN_PT_LOGGING_ENABLED in xen_pt.h prior to its first use).

It further looks like it's Windows you're having the problem with.
Unless Windows provides you with some kind of diagnostics,
could you check with Linux (allowing us to also see the kernel

Finally, it would help if you could do all debugging on at least 4.9,
preferably even master. Among other benefits that'll exclude
there potentially being some other change missing from the older

> Jan, I still have access to the hardware so perhaps we can finally solve 
> this problem.

Feel free to go ahead; I'll be on vacation for the next three weeks.


Xen-devel mailing list



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