Re: vfio-pci: dvb-s2 pcie card stopped working after a short time

[Wolfgang Nothdurft <Wolfgang.Nothdurft@linogate.de>]

Am 09.04.2015 um 17:40 schrieb Alex Williamson:
> On Thu, 2015-04-09 at 15:13 +0200, Wolfgang Nothdurft wrote:
>> Hi,
>>
>> I'm using kvm + libvirt + pci pass-through (vfio-pci) for virtualizing
>> my mythtv server.
>>
>>       <hostdev mode='subsystem' type='pci' managed='yes'>
>>         <driver name='vfio'/>
>>         <source>
>>           <address domain='0x0000' bus='0x01' slot='0x00' function='0x0'/>
>>         </source>
>>         <address type='pci' domain='0x0000' bus='0x02' slot='0x03'
>> function='0x0'/>
>>       </hostdev>
>>
>>
>> I had no problems with my previous dvb cards at least no big known problems.
>>
>> With my new dvb-s2 card (DVBSky S952 V3), the card stopped working after
>> recording 1 or 2 hour.
>>
>> After than the dvbstream from the card is broken and I have to reload
>> the driver.
>>
>> MythTV Log shows:
>>
>> Apr  7 19:19:10 mythtv mythlogserver: mythbackend[2087]: E DVBRead
>> recorders/dtvrecorder.cpp:855 (FindH264Keyframes) DTVRec[1]: PES start
>> code not found in TS packet with PUSI set
>> Apr  7 19:19:12 mythtv mythlogserver: mythbackend[2087]: E DVBRead
>> mpeg/mpegstreamdata.cpp:364 (AssemblePSIP)
>> MPEGStream[1](0x7f9bd412b898): Error: offset>181, pes length & current
>> cannot be queried
>>
>> On the kvm host the card works properly. With my old cards I got these
>> problems sporadically, so this was not a big deal for me.
>> But now it is unusable.
>>
>> /proc/interrupts and lspci -vv from the kvmhost:
>>
>>    29:    1061275    1065204    1060158    1066001   PCI-MSI-edge
>> SMI_PCIE
>>
>> 01:00.0 Multimedia video controller: Spin Master Ltd. Device 3038 (rev 01)
>>           Subsystem: DVBSky Device 0552
>>           Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop-
>> ParErr- Stepping- SERR- FastB2B- DisINTx+
>>           Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort-
>> <TAbort- <MAbort- >SERR- <PERR- INTx-
>>           Latency: 0, Cache Line Size: 32 bytes
>>           Interrupt: pin A routed to IRQ 29
>>           Region 0: Memory at d0100000 (32-bit, non-prefetchable) [size=4K]
>>           Capabilities: [40] Power Management version 3
>>                   Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA
>> PME(D0-,D1-,D2-,D3hot-,D3cold-)
>>                   Status: D0 NoSoftRst- PME-Enable- DSel=0 DScale=0 PME-
>>           Capabilities: [50] MSI: Enable+ Count=1/16 Maskable- 64bit+
>>                   Address: 00000000fee0f00c  Data: 4127
>>           Capabilities: [70] Express (v1) Endpoint, MSI 00
>>                   DevCap: MaxPayload 256 bytes, PhantFunc 0, Latency L0s
>> <64ns, L1 <1us
>>                           ExtTag- AttnBtn- AttnInd- PwrInd- RBE+ FLReset-
>>                   DevCtl: Report errors: Correctable- Non-Fatal- Fatal-
>> Unsupported-
>>                           RlxdOrd+ ExtTag- PhantFunc- AuxPwr- NoSnoop-
>>                           MaxPayload 128 bytes, MaxReadReq 512 bytes
>>                   DevSta: CorrErr- UncorrErr- FatalErr- UnsuppReq-
>> AuxPwr- TransPend-
>>                   LnkCap: Port #0, Speed 2.5GT/s, Width x1, ASPM L0s,
>> Latency L0 unlimited, L1 unlimited
>>                           ClockPM- Surprise- LLActRep- BwNot-
>>                   LnkCtl: ASPM Disabled; RCB 128 bytes Disabled- Retrain-
>> CommClk+
>>                           ExtSynch- ClockPM- AutWidDis- BWInt- AutBWInt-
>>                   LnkSta: Speed 2.5GT/s, Width x1, TrErr- Train- SlotClk+
>> DLActive- BWMgmt- ABWMgmt-
>>           Kernel driver in use: SMI PCIe driver
>>           Kernel modules: smipcie
>>
>>
>> /proc/interrupts from kvmhost when the vm is started:
>>
>>    29:          0          0          0          0   PCI-MSI-edge
>> vfio-msi[0](0000:01:00.0)
>>
>> /proc/interrupts and lspci -vv from the vm
>>
>>    25:          0          0   PCI-MSI-edge      SMI_PCIE
>
>
> Are you suggesting with these zero counts that interrupts aren't
> working, or do they increment normally when the card is in use?
>
>
>> 02:03.0 Multimedia video controller: Spin Master Ltd. Device 3038 (rev 01)
>>           Subsystem: DVBSky Device 0552
>>           Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop-
>> ParErr- Stepping- SERR+ FastB2B- DisINTx+
>>           Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort-
>> <TAbort- <MAbort- >SERR- <PERR- INTx-
>>           Latency: 0, Cache Line Size: 32 bytes
>>           Interrupt: pin A routed to IRQ 25
>>           Region 0: Memory at fe660000 (32-bit, non-prefetchable) [size=4K]
>>           Capabilities: [40] Power Management version 3
>>                   Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA
>> PME(D0-,D1-,D2-,D3hot-,D3cold-)
>>                   Status: D0 NoSoftRst- PME-Enable- DSel=0 DScale=0 PME-
>>           Capabilities: [50] MSI: Enable+ Count=1/16 Maskable- 64bit+
>>                   Address: 00000000fee0300c  Data: 41a1
>>           Capabilities: [70] Express (v1) Endpoint, MSI 00
>>                   DevCap: MaxPayload 256 bytes, PhantFunc 0, Latency L0s
>> <64ns, L1 <1us
>>                           ExtTag- AttnBtn- AttnInd- PwrInd- RBE+ FLReset-
>>                   DevCtl: Report errors: Correctable- Non-Fatal- Fatal-
>> Unsupported-
>>                           RlxdOrd+ ExtTag- PhantFunc- AuxPwr- NoSnoop-
>>                           MaxPayload 128 bytes, MaxReadReq 512 bytes
>>                   DevSta: CorrErr- UncorrErr- FatalErr- UnsuppReq-
>> AuxPwr- TransPend-
>>                   LnkCap: Port #0, Speed 2.5GT/s, Width x1, ASPM L0s,
>> Latency L0 unlimited, L1 unlimited
>>                           ClockPM- Surprise- LLActRep- BwNot-
>>                   LnkCtl: ASPM Disabled; RCB 128 bytes Disabled- Retrain-
>> CommClk+
>>                           ExtSynch- ClockPM- AutWidDis- BWInt- AutBWInt-
>>                   LnkSta: Speed 2.5GT/s, Width x1, TrErr- Train- SlotClk+
>> DLActive- BWMgmt- ABWMgmt-
>>           Kernel driver in use: SMI PCIe driver
>>           Kernel modules: smipcie
>>
>> I have tried several things, all with the same result.
>>
>> * Swapping PCIe Port
>> * using kvm-based-passthrough instead of vfio-pci
>> * starting qemu direct and using bus=pcie.0 ( -device
>> vfio-pci,host=01:00.0,id=hostdev0,bus=pcie.0,addr=0x3)
>>
>>
>> Board: Supermicro C2SBC-Q
>> dvb-s2: DVBSky S952 V3
>> System: Gentoo
>> Kernel: 3.19.0-gentoo (on both host and vm)
>> qemu: 2.1.2
>> libvirt: 1.2.10
>>
>> Is there anything I can do to find and solve this problem?
>> Or is this simply a bad combination of hardware components and I will
>> stuck here?
>
> If neither vfio-pci nor legacy kvm based assignment work reliably, then
> it sounds like there's some quirkiness to this card's operation that
> hasn't been figured out.  Does the old card work in the same
> configuration?  Do the old and new cards use the same driver?  Is there
> any relevant dmesg output in either the host or the guest?  If you can
> kick the new card to work again by reloading the driver in the guest,
> that would seem to indicate some lack of robustness in the driver
> itself.  Thanks,

so it's not hopeless :)

The old card works in exact this configuration, but uses an other 
driver. The driver for the new card is newish and was first introduced 
in 3.19. I tested the media-build driver from dvbsky with the same result.
But I was wondering, because the card was running stable with the same 
driver on the bare host for days, so I thought it would be due to 
virtualization.

The only relevant dmesg message is

kvm: zapping shadow pages for mmio generation wraparound

on the host when I start the vm.

One thing to mention is, that I need allow_unsafe_interrupts to get the 
passthrough to work.

Thanks
Wolfgang