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

Re: [Xen-devel] SMMU, Unhandled context fault



Hi Julien,

On Thu, Apr 28, 2016 at 02:14:58PM +0100, Julien Grall wrote:
>Hello,
>
>On 28/04/16 13:56, Peng Fan wrote:
>>On Thu, Apr 28, 2016 at 11:27:22AM +0100, Julien Grall wrote:
>>>
>>>
>>>On 28/04/16 07:39, Peng Fan wrote:
>>>>Hi Julien,
>>>
>>>Hello Peng,
>>>
>>>>On Thu, Apr 28, 2016 at 10:37:54AM +0800, Peng Fan wrote:
>>>>>Hi Julien,
>>>>>On Wed, Apr 27, 2016 at 10:58:28AM +0100, Julien Grall wrote:
>>>>>>Hello Peng,
>>>>>>
>>>>>>On 27/04/2016 03:02, Peng Fan wrote:
>>>>>>>On Tue, Apr 26, 2016 at 04:30:03PM +0200, Edgar E. Iglesias wrote:
>>>>>>>>On Tue, Apr 26, 2016 at 09:56:33PM +0800, Peng Fan wrote:
>>>>>>>>>You mean the PNU bit(Privileged Not Unprivileged) is 1?
>>>>>>>>>I did not met Unhandled context fault each time.
>>>>>>>>>Actually during my serveral boot test, I only met two times.
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>I meant the NSSTATE and NSATTR bits in FSYNR are set to zero. I get the
>>>>>>>>impression that the TrustZone state for the SD controller may be
>>>>>>>
>>>>>>>oh. The NSATTR bit is 0. I did not find NSSTATE in my Issue D SMMU spec.
>>>>>>>If without xen, only one linux boots up, sd controller can access memory 
>>>>>>>using
>>>>>>>DMA without issue.
>>>>>>
>>>>>>IIRC, by default Linux baremetal does not protect the devices with the 
>>>>>>SMMU.
>>>>>>
>>>>>>I would recommend you to check whether the SMMUs are in-used and 
>>>>>>configured
>>>>>>to generate a fault (disable_bypass = 1).
>>>>>
>>>>>Ok. I'll set S2CRn to generate fault in xen smmu driver to see whether 
>>>>>SMMUs in-used or not
>>>
>>>I meant in Linux.
>>
>>My bad. Do you mean enabling SMMU driver in Linux with KVM support?
>
>Yes.
>
>[...]
>
>>Is there any big difference between XEN SMMU driver and linux SMMU driver?
>>I know that XEN only support Stage 2. But the initliaization flow is almost 
>>the same.
>
>The SMMU driver for Xen is a port from Linux 3.19-rc0. Since then the Linux
>driver has been reworked and it might be possible that we have missed some
>bug fix.
>
>Aside that, for Xen, the page tables are always shared between the SMMU and
>the processor.

Thanks. I shared two picture that dumped using TRACE32.

https://drive.google.com/file/d/0B9ruJqJLIGp7cHhhSFNSNC00MHc/view?usp=sharing
https://drive.google.com/file/d/0B9ruJqJLIGp7dmlqVllXYTIxajQ/view?usp=sharing

Would you please help check?

The block that marked red, seems not correct. I am also adding debug info
in xen memory allocation part to see what happends.

Thanks,
Peng.

>
>Regards,
>
>-- 
>Julien Grall

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

 


Rackspace

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