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

Re: [Xen-devel] [PATCH 0/4] x86/IOMMU: multi-vector MSI prerequisites

>>> On 11.04.13 at 17:40, suravee suthikulpanit <suravee.suthikulpanit@xxxxxxx> 
>>> wrote:
> I have attached the xl dmesg output from the machine which run the 
> non-patched Xen (w/ iommu=debug). It is showing that there is no IVHD 
> entry setting up for the bdf 0x2 (the IOMMU).  Also, there is a line 
> "AMD-Vi: Fail to find iommu for MSI device id = 0x2) which can be traced 
> to xen/drivers/passthrough/amd/iommu_intr.c: 
> amd_iommu_msi_msg_update_ire() which is called from xen/arch/x86/msi.c: 
> write_msi_msg().

Which confirms that adding a check as suggested in an earlier
reply should b e all that's missing.

> When comparing with other AMD systems in the past (i.e. with SR56XX 
> chipset), the IVHD contains IVHD entries such as below:
> (XEN) AMD-Vi: IVHD Device Entry: type 0x3 id 0 flags 0
> (XEN) AMD-Vi:  Dev_Id Range: 0 -> 0x2
> which will generate the mapping entry for bfd 0x0, 0x1, 0x2.

Now that is confusing me: I understand this to mean that on such
a system, the IOMMU's MSI would get remapped, whereas on the
system you saw the crash on it would remain untouched. Can that
really be correct?

> PS: I could not get the serial console out from the trouble target 
> system (Trinity based), so I could not get the crash message.  But from 
> the prior email I sent to first report the issue, you can see that it 
> failed during "enable_iommu".

Yeah, sure, that's understood and fine.

Thanks for all the help so far - I'll get an updated patch set out


Xen-devel mailing list



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