Google Coral TPU PCIe Passthrough Woes

exime

@olivierlambert said in Google Coral TPU PCIe Passthrough Woes:

Can you try an on older kernel in your VM just to be sure? (Eg Debian 10 guest with default bundled kernel)

I'm just now getting back to this.

Might the problem be related to this issue?

"Unfortunately the device in question violates PCI specification by mapping PBA, MSI-X vector table, and other registers into same 4KB page (PBA is at 0x46068, VT at 0x46800, but there is a bunch of other registers in 0x46XXX range)."

https://github.com/google-coral/edgetpu/issues/343#issuecomment-1287251821

dakota created this issue in google-coral/edgetpu

open Apex failing with error -110 (No /dev/apex_0) #343

andSmv

Hello @exime,

Here we have the EPT violation (write access to the r-x page) at 0x3fff2046. This address is tagged as an MMIO address, so very probably belongs to the device you're trying to passthrough.

Normally this has nothing to do with the 0x46xxx range (where MSI-X caps are pointing)? But the fact that there's some hacking in there make me think that Google engs also did some hacking all the place around.

Can you please while starting a native guest (or in dom0 before the passthrough) give a PCI dump for your device

lspci -vvv -s $YOUR_DEV_BDF

YOUR_DEV_BDF is your device PCI id (ex: 00:1:0)

jjgg

@andSmv I'm pretty sure I'm having the same problem as @exime except I'm getting the VM just cold turning off.

Output from your command on the host:

[16:09 xenhost04 ~]# lspci -vvv -s 41:00.0
41:00.0 Non-VGA unclassified device: Global Unichip Corp. Coral Edge TPU (prog-if ff)
        Subsystem: Global Unichip Corp. Coral Edge TPU
        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: 64 bytes
        Interrupt: pin A routed to IRQ 66
        Region 0: Memory at d01fc000 (64-bit, prefetchable) [size=16K]
        Region 2: Memory at d0000000 (64-bit, prefetchable) [size=1M]
        Capabilities: [80] Express (v2) Endpoint, MSI 00
                DevCap: MaxPayload 256 bytes, PhantFunc 0, Latency L0s <64ns, L1 <1us
                        ExtTag+ AttnBtn- AttnInd- PwrInd- RBE+ FLReset- SlotPowerLimit 25.000W
                DevCtl: Report errors: Correctable- Non-Fatal+ Fatal+ Unsupported+
                        RlxdOrd+ ExtTag+ PhantFunc- AuxPwr- NoSnoop+
                        MaxPayload 256 bytes, MaxReadReq 512 bytes
                DevSta: CorrErr- UncorrErr- FatalErr- UnsuppReq- AuxPwr- TransPend-
                LnkCap: Port #1, Speed 5GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <64ns, L1 <1us
                        ClockPM+ Surprise- LLActRep- BwNot- ASPMOptComp+
                LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- CommClk-
                        ExtSynch- ClockPM- AutWidDis- BWInt- AutBWInt-
                LnkSta: Speed 5GT/s, Width x1, TrErr- Train- SlotClk- DLActive- BWMgmt- ABWMgmt-
                DevCap2: Completion Timeout: Range ABCD, TimeoutDis+, LTR+, OBFF Not Supported
                DevCtl2: Completion Timeout: 65ms to 210ms, TimeoutDis-, LTR-, OBFF Disabled
                LnkCtl2: Target Link Speed: 5GT/s, EnterCompliance- SpeedDis-
                         Transmit Margin: Normal Operating Range, EnterModifiedCompliance- ComplianceSOS-
                         Compliance De-emphasis: -6dB
                LnkSta2: Current De-emphasis Level: -6dB, EqualizationComplete-, EqualizationPhase1-
                         EqualizationPhase2-, EqualizationPhase3-, LinkEqualizationRequest-
        Capabilities: [d0] MSI-X: Enable- Count=128 Masked-
                Vector table: BAR=2 offset=00046800
                PBA: BAR=2 offset=00046068
        Capabilities: [e0] MSI: Enable- Count=1/32 Maskable- 64bit+
                Address: 0000000000000000  Data: 0000
        Capabilities: [f8] 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: [100 v1] Vendor Specific Information: ID=1556 Rev=1 Len=008 <?>
        Capabilities: [108 v1] Latency Tolerance Reporting
                Max snoop latency: 0ns
                Max no snoop latency: 0ns
        Capabilities: [110 v1] L1 PM Substates
                L1SubCap: PCI-PM_L1.2+ PCI-PM_L1.1+ ASPM_L1.2+ ASPM_L1.1+ L1_PM_Substates+
                          PortCommonModeRestoreTime=10us PortTPowerOnTime=10us
        Capabilities: [200 v2] Advanced Error Reporting
                UESta:  DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt- UnxCmplt- RxOF- MalfTLP- ECRC- UnsupReq- ACSViol-
                UEMsk:  DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt+ UnxCmplt+ RxOF- MalfTLP- ECRC- UnsupReq- ACSViol-
                UESvrt: DLP+ SDES- TLP+ FCP+ CmpltTO+ CmpltAbrt- UnxCmplt- RxOF- MalfTLP+ ECRC+ UnsupReq- ACSViol-
                CESta:  RxErr- BadTLP- BadDLLP- Rollover- Timeout- NonFatalErr-
                CEMsk:  RxErr+ BadTLP+ BadDLLP+ Rollover+ Timeout+ NonFatalErr+
                AERCap: First Error Pointer: 00, GenCap+ CGenEn- ChkCap+ ChkEn-
        Kernel driver in use: pciback

andSmv

@jjgg Can you please also post XEN traces after the VM is stopped.
(either in hypervisor.log or just type xl dmesg (under root account in your dom0)

jjgg

@andSmv

xl dmesg from the host in question. Note the last set prefixed with 739760 appeared just after I attempted to install the drivers, the depmod process was just running, it turns off before dpkg finishes. I've just included the whole set as I suspect this is from several different attempts.

Additional note, I actually have two Coral PCI devices installed. For the final test I only passed one through to the VM.

(XEN) [523317.173290] d19v4 EPT violation 0x1aa (-w-/r-x) gpa 0x000000f1846800 mfn 0xd0246 type 5
(XEN) [523317.173297] d19v4 Walking EPT tables for GFN f1846:
(XEN) [523317.173302] d19v4  epte 9c00005519618007
(XEN) [523317.173306] d19v4  epte 9c00002af3b9b007
(XEN) [523317.173309] d19v4  epte 9c0000244004d007
(XEN) [523317.173313] d19v4  epte 9c500000d0246845
(XEN) [523317.173318] d19v4  --- GLA 0xffffb4c880209800
(XEN) [523317.173333] domain_crash called from vmx_vmexit_handler+0xea6/0x1b00
(XEN) [523317.173338] Domain 19 (vcpu#4) crashed on cpu#23:
(XEN) [523317.173348] ----[ Xen-4.13.5-9.30  x86_64  debug=n   Not tainted ]----
(XEN) [523317.173351] CPU:    23
(XEN) [523317.173355] RIP:    0010:[<ffffffffad54987f>]
(XEN) [523317.173359] RFLAGS: 0000000000010202   CONTEXT: hvm guest (d19v4)
(XEN) [523317.173369] rax: ffffb4c880209800   rbx: ffffb4c8806a3aac   rcx: 0000000000000000
(XEN) [523317.173373] rdx: 00000000fee97000   rsi: ffffb4c8806a3aac   rdi: 0000000000000001
(XEN) [523317.173381] rbp: ffff904f409e6380   rsp: ffffb4c8806a3a60   r8:  000000000000000d
(XEN) [523317.173385] r9:  ffff904f413d52f8   r10: ffff904f4162a700   r11: 00000000000393e0
(XEN) [523317.173389] r12: 0000000000000097   r13: 0000000000000011   r14: 000000000000000d
(XEN) [523317.173393] r15: ffff904f413d52f8   cr0: 0000000080050033   cr4: 00000000001706e0
(XEN) [523317.173398] cr3: 0000000103736006   cr2: 00007f23d938f6b7
(XEN) [523317.173402] fsb: 00007f23d9e298c0   gsb: ffff90563e700000   gss: 0000000000000000
(XEN) [523317.173407] ds: 0000   es: 0000   fs: 0000   gs: 0000   ss: 0018   cs: 0010
(XEN) [524620.959493] d23v1 EPT violation 0x1aa (-w-/r-x) gpa 0x000000f1846800 mfn 0xd0246 type 5
(XEN) [524620.959498] d23v1 Walking EPT tables for GFN f1846:
(XEN) [524620.959501] d23v1  epte 9c00005730d5f007
(XEN) [524620.959504] d23v1  epte 9c0000245bedd007
(XEN) [524620.959506] d23v1  epte 9c00002af3b93007
(XEN) [524620.959509] d23v1  epte 9c500000d0246845
(XEN) [524620.959512] d23v1  --- GLA 0xffffb82103411800
(XEN) [524620.959520] domain_crash called from vmx_vmexit_handler+0xea6/0x1b00
(XEN) [524620.959523] Domain 23 (vcpu#1) crashed on cpu#10:
(XEN) [524620.959529] ----[ Xen-4.13.5-9.30  x86_64  debug=n   Not tainted ]----
(XEN) [524620.959531] CPU:    10
(XEN) [524620.959534] RIP:    0010:[<ffffffff95027075>]
(XEN) [524620.959537] RFLAGS: 0000000000010202   CONTEXT: hvm guest (d23v1)
(XEN) [524620.959542] rax: 00000000fee97000   rbx: ffff9213b9164480   rcx: 0000000000000000
(XEN) [524620.959545] rdx: ffffb82103411800   rsi: 0000000000000000   rdi: ffffb82103411800
(XEN) [524620.959549] rbp: ffffb8210664f9ac   rsp: ffffb8210664f960   r8:  0000000000000001
(XEN) [524620.959552] r9:  ffff9213be007600   r10: 0000000000001001   r11: 0000000000001001
(XEN) [524620.959556] r12: 0000000000000011   r13: ffff9213ba74f2a8   r14: ffff9213ba74f0a8
(XEN) [524620.959558] r15: 0000000000000097   cr0: 0000000080050033   cr4: 00000000001606e0
(XEN) [524620.959561] cr3: 00000007c9230004   cr2: 000055d28bf759a8
(XEN) [524620.959564] fsb: 00007f9f132264c0   gsb: ffff9213be440000   gss: 0000000000000000
(XEN) [524620.959567] ds: 0000   es: 0000   fs: 0000   gs: 0000   ss: 0018   cs: 0010
(XEN) [524887.457380] d24v0 EPT violation 0x1aa (-w-/r-x) gpa 0x000000f1846800 mfn 0xd0246 type 5
(XEN) [524887.457385] d24v0 Walking EPT tables for GFN f1846:
(XEN) [524887.457389] d24v0  epte 9c000057150c0007
(XEN) [524887.457392] d24v0  epte 9c000022442f9007
(XEN) [524887.457395] d24v0  epte 9c00002af3b93007
(XEN) [524887.457398] d24v0  epte 9c500000d0246845
(XEN) [524887.457401] d24v0  --- GLA 0xffffa20b43411800
(XEN) [524887.457411] domain_crash called from vmx_vmexit_handler+0xea6/0x1b00
(XEN) [524887.457415] Domain 24 (vcpu#0) crashed on cpu#27:
(XEN) [524887.457422] ----[ Xen-4.13.5-9.30  x86_64  debug=n   Not tainted ]----
(XEN) [524887.457424] CPU:    27
(XEN) [524887.457427] RIP:    0010:[<ffffffff97027075>]
(XEN) [524887.457430] RFLAGS: 0000000000010202   CONTEXT: hvm guest (d24v0)
(XEN) [524887.457437] rax: 00000000fee97000   rbx: ffff96aab4ed67e0   rcx: 0000000000000000
(XEN) [524887.457440] rdx: ffffa20b43411800   rsi: 0000000000000000   rdi: ffffa20b43411800
(XEN) [524887.457445] rbp: ffffa20b4ca4f9ac   rsp: ffffa20b4ca4f960   r8:  0000000000000001
(XEN) [524887.457448] r9:  ffff96aabe007600   r10: 0000000000001001   r11: 0000000000001001
(XEN) [524887.457452] r12: 0000000000000011   r13: ffff96aaba7482a8   r14: ffff96aaba7480a8
(XEN) [524887.457456] r15: 0000000000000097   cr0: 0000000080050033   cr4: 00000000001606f0
(XEN) [524887.457459] cr3: 00000007eb0ea003   cr2: 00007ff2ba681fa7
(XEN) [524887.457463] fsb: 00007ff2ba6844c0   gsb: ffff96aabe400000   gss: 0000000000000000
(XEN) [524887.457466] ds: 0000   es: 0000   fs: 0000   gs: 0000   ss: 0018   cs: 0010
(XEN) [525662.181596] d29v1 EPT violation 0x1aa (-w-/r-x) gpa 0x000000f1846800 mfn 0xd0046 type 5
(XEN) [525662.181599] d29v1 Walking EPT tables for GFN f1846:
(XEN) [525662.181602] d29v1  epte 9c00005519559007
(XEN) [525662.181604] d29v1  epte 9c000022442f9007
(XEN) [525662.181605] d29v1  epte 9c0000284bed3007
(XEN) [525662.181607] d29v1  epte 9c500000d0046845
(XEN) [525662.181609] d29v1  --- GLA 0xffffb41dc3409800
(XEN) [525662.181615] domain_crash called from vmx_vmexit_handler+0xea6/0x1b00
(XEN) [525662.181617] Domain 29 (vcpu#1) crashed on cpu#0:
(XEN) [525662.181621] ----[ Xen-4.13.5-9.30  x86_64  debug=n   Not tainted ]----
(XEN) [525662.181623] CPU:    0
(XEN) [525662.181625] RIP:    0010:[<ffffffffb8c27075>]
(XEN) [525662.181626] RFLAGS: 0000000000010202   CONTEXT: hvm guest (d29v1)
(XEN) [525662.181630] rax: 00000000fee97000   rbx: ffff8abaf4d72060   rcx: 0000000000000000
(XEN) [525662.181632] rdx: ffffb41dc3409800   rsi: 0000000000000000   rdi: ffffb41dc3409800
(XEN) [525662.181635] rbp: ffffb41dc67b79ac   rsp: ffffb41dc67b7960   r8:  0000000000000001
(XEN) [525662.181637] r9:  ffff8abafe007600   r10: 0000000000001001   r11: 0000000000001001
(XEN) [525662.181639] r12: 0000000000000011   r13: ffff8abafa7492a8   r14: ffff8abafa7490a8
(XEN) [525662.181641] r15: 0000000000000097   cr0: 0000000080050033   cr4: 00000000001606e0
(XEN) [525662.181643] cr3: 00000007f6df2006   cr2: 000055f23e274c78
(XEN) [525662.181645] fsb: 00007f3828fce4c0   gsb: ffff8abafe440000   gss: 0000000000000000
(XEN) [525662.181647] ds: 0000   es: 0000   fs: 0000   gs: 0000   ss: 0018   cs: 0010
(XEN) [739760.414382] d30v3 EPT violation 0x1aa (-w-/r-x) gpa 0x000000f1846800 mfn 0xd0046 type 5
(XEN) [739760.414387] d30v3 Walking EPT tables for GFN f1846:
(XEN) [739760.414391] d30v3  epte 9c00005730d5f007
(XEN) [739760.414394] d30v3  epte 9c000022442f8007
(XEN) [739760.414396] d30v3  epte 9c0000224437a007
(XEN) [739760.414399] d30v3  epte 9c500000d0046845
(XEN) [739760.414402] d30v3  --- GLA 0xffffae9003411800
(XEN) [739760.414411] domain_crash called from vmx_vmexit_handler+0xea6/0x1b00
(XEN) [739760.414414] Domain 30 (vcpu#3) crashed on cpu#9:
(XEN) [739760.414420] ----[ Xen-4.13.5-9.30  x86_64  debug=n   Not tainted ]----
(XEN) [739760.414422] CPU:    9
(XEN) [739760.414425] RIP:    0010:[<ffffffff97427075>]
(XEN) [739760.414428] RFLAGS: 0000000000010202   CONTEXT: hvm guest (d30v3)
(XEN) [739760.414433] rax: 00000000fee97000   rbx: ffff96a3f95d7cc0   rcx: 0000000000000000
(XEN) [739760.414437] rdx: ffffae9003411800   rsi: 0000000000000000   rdi: ffffae9003411800
(XEN) [739760.414441] rbp: ffffae900d0779ac   rsp: ffffae900d077960   r8:  0000000000000001
(XEN) [739760.414444] r9:  ffff96a3fe007600   r10: 0000000000001001   r11: 0000000000001001
(XEN) [739760.414448] r12: 0000000000000011   r13: ffff96a3fa7482a8   r14: ffff96a3fa7480a8
(XEN) [739760.414451] r15: 0000000000000097   cr0: 0000000080050033   cr4: 00000000001606e0
(XEN) [739760.414454] cr3: 00000007f7266001   cr2: 00007f22e42b42b0
(XEN) [739760.414457] fsb: 00007f22e412f4c0   gsb: ffff96a3fe4c0000   gss: 0000000000000000
(XEN) [739760.414460] ds: 0000   es: 0000   fs: 0000   gs: 0000   ss: 0018   cs: 0010

andSmv

@jjgg Thank you. Yes the same problem - ept violation.. Look, I'll try to figure out what we can do here. There's a patch that comes from Qubes OS guys that normally shold fix the MSI-x PBA issue (not sure that this is the good fix, but still... worth trying) This patch applies on recent Xen and wasn't accepted yet. I will take a look if it can be easily backported to XCP-ng Xen and come back to you.

exime

@andSmv thanks!

@jjgg glad you're providing the info, sorry for abandoning the thread

jjgg

@andSmv thanks. Completely understand this appears more of a hardware issue here, but happy to test anything. The host that these cards are installed in isn't running critical infrastructure and can be rebooted relatively easily.

@exime your initial post was a top Google result, was still helpful! Thanks.

andSmv

@jjgg Here's the link to xen.gz.

You need to put it in your /boot folder (backup your existent file!) and make sure your grub.cfg is pointing to it.

But first: Backup all you want to backup! The patch is totally untested and doesn't apply as is (so I needed to adapt it). Normally not such a big deal and should not do no harm, but... you never know.

I'm also not sure that the issue would be fixed. We unfortunatelly do not have Coral TPU device at Vates, so we can't do the more deep analysis on this. The guy who wrote this patch tried to fix other device.

@exime - this is 4.13.5 XCP-ng patched xen, so there's chances it wouldn't work for you (from what I saw you're running 4.13.4 xen)

Anyway, if we have good news, we'll find the way to fix it for everybody.

exime

@andSmv ack - I'll wait and see if it works out for @jjgg since my Xen server is in active use

jjgg

Ok so testing setup.

Downloaded xen.gz, renamed to xenept.gz and put into /boot:

Updated grub.cfg to point to that file:

Rebooted host.

Got to the grub screen, let it load as normal, as soon as it disappeared (so it had made the default selection) the server rebooted.

I ended up just placing xen.gz in /boot and removing the symbolic link to the original and attempting, no difference.

Of note, messing with kernels / grub is not something I've got experience with. I may have made a mistake / need things explained in a bit more detail if I'm potentially misunderstood some instructions above.

jjgg

Booted into fallback and put things back the way they were. Happy to keep testing if there's additional bits to test.

jmccoy555

Ah, just found these things exist..... Then just found this issue exists too ️

olivierlambert

If only they could have done PCI hardware that follow the PCI specifications

jmccoy555

Maybe this one will come to life again https://xcp-ng.org/forum/topic/7066/coral-tpu-pci-passthrough/14

Don't really want to buy one knowing its not working!!

jjgg

Definitely frustrating and no fault of xcp-ng - I have a lot of spare cpu cycles so it isn't majorly impacting me that I know of. I'm still available to test fixes though.

Looks like most of the Proxmox users have got this working in an LXC container by installing the drivers on the host itself and passing through the actual Apex devices. Not a route that's applicable to us but just a datapoint.

jmccoy555

@jjgg it would be great if we could get this working. My CPU utilisation is fine too, but when I shut down my Zoneminder VM things go a lot quieter (fans) so I'm sure there would be a benefit CPU and power wise.

Nornode

@jmccoy555 // @jjgg

Did anyone of you get your Coral USB TPU working and passthrough to a VM?

jjgg

@Nornode hey, nope I did not. I ended up moving my infrastructure to Proxmox.

Honestly this is no fault of XCP-ng and XCP-ng suits my hardware / setup a lot better, but it was either that or I had two servers that needed to be bare metal.

olivierlambert

PCI passthrough might cause problems with this device, but USB could work.