-
Notifications
You must be signed in to change notification settings - Fork 5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
GPU is reset on Raspberry Pi 3B+ #5780
Comments
I am investigating on similar issue but observed on RPi2 I am able to reproduce while playing with settings of kodi Some inputs :
|
I'm seeing the same after doing an update this afternoon.
|
The changes added in that version: https://release.debian.org/proposed-updates/bullseye_diffs/xorg-server_1.20.11-1+deb11u11.debdiff |
Same issue here with Raspbian GNU/Linux 12 (bookworm) armv7l on a Pi2B Rev1.1 with fresh installed RetroPie as well as with kodi |
I'm here running into the same issue but then with a image build by buildroot running webkit on a RPI3b. I updated the kernel from 6.1.61 to 6.1.77 due to this issue #5674. my config:
This is the register dump when the message appears. # cat /sys/kernel/debug/dri/0/v3d_regs
V3D_IDENT0 = 0x02443356
V3D_IDENT1 = 0xc1102431
V3D_IDENT2 = 0x00000121
V3D_SCRATCH = 0x00000000
V3D_L2CACTL = 0x00000001
V3D_SLCACTL = 0x00000000
V3D_INTCTL = 0x00000000
V3D_INTENA = 0x00000007
V3D_INTDIS = 0x00000007
V3D_CT0CS = 0x00004020
V3D_CT1CS = 0x00000120
V3D_CT0EA = 0xd7d330ba
V3D_CT1EA = 0xd7d33a5b
V3D_CT0CA = 0xd7d3208c
V3D_CT1CA = 0xd85085a9
V3D_CT00RA0 = 0x00000000
V3D_CT01RA0 = 0xd7dccac1
V3D_CT0LC = 0x00000000
V3D_CT1LC = 0x00000000
V3D_CT0PC = 0x00000000
V3D_CT1PC = 0x00000000
V3D_PCS = 0x00000005
V3D_BFC = 0x000000b8
V3D_RFC = 0x000000f2
V3D_BPCA = 0xd8506e00
V3D_BPCS = 0x00079000
V3D_BPOA = 0xd8300000
V3D_BPOS = 0x00000000
V3D_BXCF = 0x00000000
V3D_SQRSV0 = 0x00000000
V3D_SQRSV1 = 0x00000000
V3D_SQCNTL = 0x00000000
V3D_SRQPC = 0x00000000
V3D_SRQUA = 0x00000000
V3D_SRQUL = 0x00000000
V3D_SRQCS = 0x00000000
V3D_VPACNTL = 0x00000000
V3D_VPMBASE = 0x00000000
V3D_PCTRC = 0x00000000
V3D_PCTRE = 0x00000000
V3D_PCTR(0) = 0x00000000
V3D_PCTRS(0) = 0x00000000
V3D_PCTR(1) = 0x00000000
V3D_PCTRS(1) = 0x00000000
V3D_PCTR(2) = 0x00000000
V3D_PCTRS(2) = 0x00000000
V3D_PCTR(3) = 0x00000000
V3D_PCTRS(3) = 0x00000000
V3D_PCTR(4) = 0x00000000
V3D_PCTRS(4) = 0x00000000
V3D_PCTR(5) = 0x00000000
V3D_PCTRS(5) = 0x00000000
V3D_PCTR(6) = 0x00000000
V3D_PCTRS(6) = 0x00000000
V3D_PCTR(7) = 0x00000000
V3D_PCTRS(7) = 0x00000000
V3D_PCTR(8) = 0x00000000
V3D_PCTRS(8) = 0x00000000
V3D_PCTR(9) = 0x00000000
V3D_PCTRS(9) = 0x00000000
V3D_PCTR(10) = 0x00000000
V3D_PCTRS(10) = 0x00000000
V3D_PCTR(11) = 0x00000000
V3D_PCTRS(11) = 0x00000000
V3D_PCTR(12) = 0x00000000
V3D_PCTRS(12) = 0x00000000
V3D_PCTR(13) = 0x00000000
V3D_PCTRS(13) = 0x00000000
V3D_PCTR(14) = 0x00000000
V3D_PCTRS(14) = 0x00000000
V3D_PCTR(15) = 0x00000000
V3D_PCTRS(15) = 0x00000000
V3D_DBGE = 0x00000000
V3D_FDBGO = 0x00000006
V3D_FDBGB = 0x1c00000d
V3D_FDBGR = 0x595d68fd
V3D_FDBGS = 0x1246d1ef
V3D_ERRSTAT = 0x00007000 # cat /sys/kernel/debug/dri/0/v3d_ident
Revision: 1
Slices: 3
TMUs: 6
QPUs: 12
Semaphores: 16 # cat /sys/kernel/debug/dri/0/bo_stats
V3D: 67368kb BOs (77)
V3D shader: 96kb BOs (24)
dumb: 4052kb BOs (1)
binner: 16384kb BOs (1)
RCL: 48kb BOs (8)
BCL: 32kb BOs (8)
userspace BO cache: 1028kb BOs (2)
@6by9 is there anything I can provide to get a more clear picture of what is going on? |
Sorry, I know very little of the 3D pipeline. @popcornmix possibly, or @melissawen for more knowledge on 3D. |
I can confirm this issue on latest RaspiOS + kernel on pretty much anything below raspberry pi4 using full v3d KMS. |
@mairacanal is following the 3D work more closely than I am. |
I've started to investigate this issue and any report explaining how to reproduce the error, kernel version, Mesa version, and attaching the debugfs files are really helpful. Thanks for all the reports here (and thanks @melissawen for pointing me to this issue)! |
i have the same issue with the Raspberry Pi 0 2w I use Recalbox 9.2 |
Hi @mairacanal. First thank you for the investigation. |
@mairacanal I'd be happy to give you the link for a bare minimum IMG for a Raspberry Pi of your choice that you can simply burn and reproduce the issue if you let me know what Pi you'd prefer, Pi0/1, Pi3, Pi Zero2W ect |
Hi, Repeats several times during the week Raspberry Pi 3 Model B Plus Rev 1.3, Raspbian GNU/Linux 11 (bullseye), 6.6.21-v7+ # lsb_release -a
No LSB modules are available.
Distributor ID: Raspbian
Description: Raspbian GNU/Linux 11 (bullseye)
Release: 11
Codename: bullseye # vcgencmd version
Feb 29 2024 12:26:47
Copyright (c) 2012 Broadcom
version f4e2138c2adc8f3a92a3a65939e458f11d7298ba (clean) (release) (start) # cat /proc/version
Linux version 6.6.21-v7+ (dom@buildbot) (arm-linux-gnueabihf-gcc-8 (Ubuntu/Linaro 8.4.0-3ubuntu1) 8.4.0, GNU ld (GNU Binutils for Ubuntu) 2.34) #1743 SMP Thu Mar 14 11:35:08 GMT 2024 # cat /proc/device-tree/model
Raspberry Pi 3 Model B Plus Rev 1.3 # cat /etc/os-release
PRETTY_NAME="Raspbian GNU/Linux 11 (bullseye)"
NAME="Raspbian GNU/Linux"
VERSION_ID="11"
VERSION="11 (bullseye)"
VERSION_CODENAME=bullseye
ID=raspbian
ID_LIKE=debian
HOME_URL="http://www.raspbian.org/"
SUPPORT_URL="http://www.raspbian.org/RaspbianForums"
BUG_REPORT_URL="http://www.raspbian.org/RaspbianBugs" # vcdbg log msg
Failed to allocate -201253077 bytes for message buffer # cat /sys/kernel/debug/dri/0/v3d_regs
V3D_IDENT0 = 0x02443356
V3D_IDENT1 = 0xc1102431
V3D_IDENT2 = 0x00000121
V3D_SCRATCH = 0x00000000
V3D_L2CACTL = 0x00000001
V3D_SLCACTL = 0x00000000
V3D_INTCTL = 0x00000000
V3D_INTENA = 0x00000007
V3D_INTDIS = 0x00000007
V3D_CT0CS = 0x00007020
V3D_CT1CS = 0x00000020
V3D_CT0EA = 0xdeec70d1
V3D_CT1EA = 0xdeec602d
V3D_CT0CA = 0xdeec70d0
V3D_CT1CA = 0xded1551a
V3D_CT00RA0 = 0x00000000
V3D_CT01RA0 = 0x00000000
V3D_CT0LC = 0x00000000
V3D_CT1LC = 0x00000000
V3D_CT0PC = 0x00000000
V3D_CT1PC = 0x00fce537
V3D_PCS = 0x00000005
V3D_BFC = 0x00000078
V3D_RFC = 0x000000a0
V3D_BPCA = 0xe0b85300
V3D_BPCS = 0x0007a000
V3D_BPOA = 0xe0b00000
V3D_BPOS = 0x00000000
V3D_BXCF = 0x00000000
V3D_SQRSV0 = 0x00000000
V3D_SQRSV1 = 0x00000000
V3D_SQCNTL = 0x00000000
V3D_SRQPC = 0x00000000
V3D_SRQUA = 0x00000000
V3D_SRQUL = 0x00000000
V3D_SRQCS = 0x00000000
V3D_VPACNTL = 0x00000000
V3D_VPMBASE = 0x00000000
V3D_PCTRC = 0x00000000
V3D_PCTRE = 0x00000000
V3D_PCTR(0) = 0x00000000
V3D_PCTRS(0) = 0x00000000
V3D_PCTR(1) = 0x00000000
V3D_PCTRS(1) = 0x00000000
V3D_PCTR(2) = 0x00000000
V3D_PCTRS(2) = 0x00000000
V3D_PCTR(3) = 0x00000000
V3D_PCTRS(3) = 0x00000000
V3D_PCTR(4) = 0x00000000
V3D_PCTRS(4) = 0x00000000
V3D_PCTR(5) = 0x00000000
V3D_PCTRS(5) = 0x00000000
V3D_PCTR(6) = 0x00000000
V3D_PCTRS(6) = 0x00000000
V3D_PCTR(7) = 0x00000000
V3D_PCTRS(7) = 0x00000000
V3D_PCTR(8) = 0x00000000
V3D_PCTRS(8) = 0x00000000
V3D_PCTR(9) = 0x00000000
V3D_PCTRS(9) = 0x00000000
V3D_PCTR(10) = 0x00000000
V3D_PCTRS(10) = 0x00000000
V3D_PCTR(11) = 0x00000000
V3D_PCTRS(11) = 0x00000000
V3D_PCTR(12) = 0x00000000
V3D_PCTRS(12) = 0x00000000
V3D_PCTR(13) = 0x00000000
V3D_PCTRS(13) = 0x00000000
V3D_PCTR(14) = 0x00000000
V3D_PCTRS(14) = 0x00000000
V3D_PCTR(15) = 0x00000000
V3D_PCTRS(15) = 0x00000000
V3D_DBGE = 0x00000000
V3D_FDBGO = 0x00000006
V3D_FDBGB = 0x10000000
V3D_FDBGR = 0x10006800
V3D_FDBGS = 0x00001147
V3D_ERRSTAT = 0x00007000 # cat /sys/kernel/debug/dri/0/v3d_ident
Revision: 1
Slices: 3
TMUs: 6
QPUs: 12
Semaphores: 16 # cat /sys/kernel/debug/dri/0/bo_stats
V3D: 35240kb BOs (27)
V3D shader: 144kb BOs (36)
dumb: 12168kb BOs (3)
binner: 16384kb BOs (1)
RCL: 44kb BOs (9)
BCL: 40kb BOs (9)
userspace BO cache: 3504kb BOs (13) # cat /var/log/syslog
May 14 08:21:33 <name> kernel: [85463.717960] [drm] Resetting GPU.
May 14 08:21:34 <name> kernel: [85464.757986] [drm] Resetting GPU.
May 14 08:21:35 <name> kernel: [85465.717977] [drm] Resetting GPU.
May 14 08:21:36 <name> kernel: [85466.758010] [drm] Resetting GPU.
May 14 08:21:37 <name> kernel: [85467.717930] [drm] Resetting GPU.
May 14 08:21:38 <name> kernel: [85468.758008] [drm] Resetting GPU.
May 14 08:21:39 <name> kernel: [85469.717938] [drm] Resetting GPU.
May 14 08:21:40 <name> kernel: [85470.758021] [drm] Resetting GPU.
May 14 08:21:41 <name> kernel: [85471.717918] [drm] Resetting GPU. |
Hello, I'm the original author of the report #3221. In 2019 I could only solve the issue by not using KMS and keep using the old legacy stack on the rpi3, which worked well. I recently tried to update the system, but what I found is that the issue is still there and makes the system crash in minutes. May I ask if someone is actually working on this issue? Are there the resources to fix it? Cause if it is unlikely to be fixed in rpi3, the other option is to go back to the legacy stack once again and remain there. That would require a considerable amount of hours for porting software and for the maintenance of two different systems, so before I do, will the legacy stack remain supported in rpi3 in newer kernels? Should we consider that the reasonable solution to this long-standing issue? Thanks. |
Hi! I was pointed out that the same bug is reproducible very easy on Recalbox 9.2.1 with Gpi case 2w. |
Hello, I invested some time on this issue. This patch #6239 prevents rendering issues after the GPU is reset. So something still causes the GPU to reset, I did not investigate that, but when it happens, proper behavior is restored. Result is just a short interruption of the rendering. May be tolerable in some cases. |
I can confirm that workaround you provide works as expected. It does not prevent the GPU reset but there is no more screen distortion. |
Hey guys, I was out for a while due to some health issues, but I'm back to this issue once more. I only have an RPi 3B+ available here and what I did to reproduce the error was to install Recalbox 9.2.3-Pulstar and try to play some games. Unfortunately, I didn't manage to reproduce the error while playing the games. Any chance you guys could provide me with more information about how to reproduce the error? For example, which game were you playing when you hit the error? How much time were you playing (5 minutes, 15 minutes, an hour)? I worked on a solution for this issue, but I can't confirm it fixes the issue if I can reproduce the error at first. Thank you guys for the feedback! |
Hi Maira! Hope you are doing well! Recalbox 9.2.3 contains the workaround pointed by @carlonluca here #5780 (comment). That's why you won't reproduce the bug (however, the GPU error still spawns in /var/log/messages). As said, this is only a workaround and if you can track the real issue, that would be great :) I'm currently looking for a way to trig the bug. |
Yesterday I made a new image with Raspberry Pi Imager.
|
What were you running? If you provide reproducible instructions we can investigate. |
Raspberry Pi uses chrome in kiosk mode. Unfortunately, the page loads from the local server and displays Chart.js charts. |
Describe the bug
The same bug as #3221. Request for reopening.
Steps to reproduce the behaviour
Occurs occasionally near 1 time at 1-2 days.
Device (s)
Raspberry Pi 3 Mod. B+
System
LibreELEC (official): 11.0.3, Kernel 6.1.38.
Logs
No response
Additional context
No response
The text was updated successfully, but these errors were encountered: