Buss Reset Problem

Buss Reset Problem

[CACook]

CPU: Intel(R) Xeon(R) CPU E3-1265L v3 @ 2.50GHz
KVM: qemu-kvm-1.5.3-86.el7_1.2.x86_64
Kernel: 4.1.1-1.el7.elrepo.x86_64, and kernel-3.10.0-229.7.2.el7.x86_64
Host & Guest: CentOS 7.1
Using virt-manager-1.1.0-12.el7.noarch to create, configure, and start guest
I am trying to do a PCI passthrough of a QCA988X wifi card. It's a Doodle Labs military-grade 802.11ac miniPCI card. This card configures nicely on the host, and seems to pass through to the KVM guest, but early in the boot of the guest it says "Unknown header type" at the wifi's bus address. And sure enough, lspci -vv on the host then shows:
        !!! Unknown header type 7f
        Kernel driver in use: vfio-pci
It seems to be the buss reset problem, as outlined here:
http://www.gossamer-threads.com/lists/linux/kernel/2054846
 
... Except my card locks up on boot of the KVM guest.
 
When I logged a bug at QEMU (https://bugs.launchpad.net/qemu/+bug/1471583), Alex suggested I add to the patch:
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATHEROS, 0x003c, quirk_no_bus_reset);
 
... So I am in the process of trying that now.



_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: Buss Reset Problem

[CACook]

On Monday 06 July 2015 8:04:39 CACook@quantum-sci.com wrote:
> 
> CPU: Intel(R) Xeon(R) CPU E3-1265L v3 @ 2.50GHz
> KVM: qemu-kvm-1.5.3-86.el7_1.2.x86_64
> Kernel: 4.1.1-1.el7.elrepo.x86_64, and kernel-3.10.0-229.7.2.el7.x86_64
> Host & Guest: CentOS 7.1
> Using virt-manager-1.1.0-12.el7.noarch to create, configure, and start guest
> I am trying to do a PCI passthrough of a QCA988X wifi card. It's a Doodle Labs military-grade 802.11ac miniPCI card. This card configures nicely on the host, and seems to pass through to the KVM guest, but early in the boot of the guest it says "Unknown header type" at the wifi's bus address. And sure enough, lspci -vv on the host then shows:
>         !!! Unknown header type 7f
>         Kernel driver in use: vfio-pci
> It seems to be the buss reset problem, as outlined here:
> http://www.gossamer-threads.com/lists/linux/kernel/2054846
>  
> ... Except my card locks up on boot of the KVM guest.
>  
> When I logged a bug at QEMU (https://bugs.launchpad.net/qemu/+bug/1471583), Alex suggested I add to the patch:
> DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATHEROS, 0x003c, quirk_no_bus_reset);
>  
> ... So I am in the process of trying that now.

So that's it for the Doodle Labs ACE-DB-3.  It went back yesterday -- doesn't work with Linux.

It's not the buss reset problem as I patched quirks.c.





_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: Buss Reset Problem

[Michal Kazior]

On 9 July 2015 at 15:33,  <CACook@quantum-sci.com> wrote:
> On Monday 06 July 2015 8:04:39 CACook@quantum-sci.com wrote:
>>
>> CPU: Intel(R) Xeon(R) CPU E3-1265L v3 @ 2.50GHz
>> KVM: qemu-kvm-1.5.3-86.el7_1.2.x86_64
>> Kernel: 4.1.1-1.el7.elrepo.x86_64, and kernel-3.10.0-229.7.2.el7.x86_64
>> Host & Guest: CentOS 7.1
>> Using virt-manager-1.1.0-12.el7.noarch to create, configure, and start guest
>> I am trying to do a PCI passthrough of a QCA988X wifi card. It's a Doodle Labs military-grade 802.11ac miniPCI card. This card configures nicely on the host, and seems to pass through to the KVM guest, but early in the boot of the guest it says "Unknown header type" at the wifi's bus address. And sure enough, lspci -vv on the host then shows:
>>         !!! Unknown header type 7f
>>         Kernel driver in use: vfio-pci
>> It seems to be the buss reset problem, as outlined here:
>> http://www.gossamer-threads.com/lists/linux/kernel/2054846
>>
>> ... Except my card locks up on boot of the KVM guest.
>>
>> When I logged a bug at QEMU (https://bugs.launchpad.net/qemu/+bug/1471583), Alex suggested I add to the patch:
>> DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATHEROS, 0x003c, quirk_no_bus_reset);
>>
>> ... So I am in the process of trying that now.
>
> So that's it for the Doodle Labs ACE-DB-3.  It went back yesterday -- doesn't work with Linux.

I think someone complained they couldn't get an XB design to work on
some of their PCI-E hosts (some Atom laptop even refused to POST with
the card if I recall correctly). Apparently there's an electrical
quirk - or at least that was to conclusion back then.


> It's not the buss reset problem as I patched quirks.c.

I wonder how much related this could be to Data Bus Error ath10k sees
when doing cold resets sometimes. Maybe it's good to have the quirk
fixup nonetheless? But I'm no PCI expert so what do I know.


Michał

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: Buss Reset Problem

[CACook]

On Thursday 09 July 2015 15:46:55 Michal Kazior wrote:
> > So that's it for the Doodle Labs ACE-DB-3.  It went back yesterday -- doesn't work with Linux.
> 
> I think someone complained they couldn't get an XB design to work on
> some of their PCI-E hosts (some Atom laptop even refused to POST with
> the card if I recall correctly). Apparently there's an electrical
> quirk - or at least that was to conclusion back then.
> 
> 
> > It's not the buss reset problem as I patched quirks.c.
> 
> I wonder how much related this could be to Data Bus Error ath10k sees
> when doing cold resets sometimes. Maybe it's good to have the quirk
> fixup nonetheless? But I'm no PCI expert so what do I know.

Yes the problem is very early in boot.  In KVM guest almost right after Grub it says that PCIe address does not exist (or is not accessible), and sure enough I check it on the host and it's flipped to "Invalid header".  I have to reboot the _host_ to get the card back then.

The card has good specs but Doodle doesn't care about this problem, and I just don't have time to uselessly fiddle with it.





_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

Re: Buss Reset Problem

[Michal Kazior]

On 9 July 2015 at 16:11,  <CACook@quantum-sci.com> wrote:
> On Thursday 09 July 2015 15:46:55 Michal Kazior wrote:
>> > So that's it for the Doodle Labs ACE-DB-3.  It went back yesterday -- doesn't work with Linux.
>>
>> I think someone complained they couldn't get an XB design to work on
>> some of their PCI-E hosts (some Atom laptop even refused to POST with
>> the card if I recall correctly). Apparently there's an electrical
>> quirk - or at least that was to conclusion back then.
>>
>>
>> > It's not the buss reset problem as I patched quirks.c.
>>
>> I wonder how much related this could be to Data Bus Error ath10k sees
>> when doing cold resets sometimes. Maybe it's good to have the quirk
>> fixup nonetheless? But I'm no PCI expert so what do I know.
>
> Yes the problem is very early in boot.  In KVM guest almost right after Grub it says that PCIe address does not exist (or is not accessible), and sure enough I check it on the host and it's flipped to "Invalid header".  I have to reboot the _host_ to get the card back then.

I wonder how much this is fault of the card and the IOMMU and/or vfio itself.

I've been using pci-stub and qemu's pci-assign to virtualize ath10k
for quite some time now. It does herp-derp on me occasionally but
never* did it require me to reboot the host.

I briefly tried to use vfio but it wasn't working well for me (can't
remember the details).

* The cold reset problem can hard lock-up the host sometimes. If you
load ath10k_pci with reset_mode=1 both on host and guest then it
doesn't require reboots of any sorts.


>
> The card has good specs but Doodle doesn't care about this problem, and I just don't have time to uselessly fiddle with it.

Sorry to hear that.


Michał

_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k