L4T 32.4.2 - GPU error on boot

I am receiving a persistent GPU error on boot, and subsequent intermittent errors following boot (specifically when loading GPU-intensive apps such as hello AI world). I am running L4T 32.4.2 - any ideas what could be causing this?

[  398.409779] nvgpu: 57000000.gpu   nvgpu_set_error_notifier_locked:137  [ERR]  error notifier set to 8 for ch 502
[  398.419964] nvgpu: 57000000.gpu     gk20a_fifo_handle_sched_error:2531 [ERR]  fifo sched ctxsw timeout error: engine=0, tsg=6, ms=3100
[  398.432169] ---- mlocks ----

[  398.436556] ---- syncpts ----
[  398.439532] id 1 (disp0_a) min 4 max 4 refs 1 (previous client : )
[  398.445763] id 2 (disp0_b) min 3 max 3 refs 1 (previous client : )
[  398.451960] id 3 (disp0_c) min 3 max 3 refs 1 (previous client : )
[  398.458160] id 7 (gm20b_507) min 27914 max 27914 refs 1 (previous client : )
[  398.465226] id 8 (gm20b_506) min 17022 max 17022 refs 1 (previous client : )
[  398.472287] id 9 (gm20b_505) min 2 max 2 refs 1 (previous client : )
[  398.478651] id 11 (gm20b_504) min 2 max 2 refs 1 (previous client : )
[  398.485189] id 12 (gm20b_503) min 54 max 54 refs 1 (previous client : )
[  398.491812] id 13 (gm20b_502) min 6902 max 6908 refs 1 (previous client : )
[  398.498781] id 14 (gm20b_501) min 4 max 4 refs 1 (previous client : )
[  398.505226] id 26 (vblank0) min 717 max -6 refs 1 (previous client : )

[  398.513320] ---- channels ----
[  398.516379]
               channel 0 - 54340000.vic

[  398.522910] 0-54340000.vic (0):
[  398.525966] inactive

[  398.529633]
               ---- host general irq ----

[  398.536333] sync_hintmask_ext = 0xc0000000
[  398.540440] sync_hintmask = 0x80000000
[  398.544191] sync_intc0mask = 0x00000001
[  398.548029] sync_intmask = 0x00000011
[  398.551692]
               ---- host syncpt irq mask ----

[  398.558739] syncpt_thresh_int_mask(0) = 0x04000001
[  398.563528] syncpt_thresh_int_mask(1) = 0x00000000
[  398.568319] syncpt_thresh_int_mask(2) = 0x00000000
[  398.573123] syncpt_thresh_int_mask(3) = 0x00000000
[  398.577914] syncpt_thresh_int_mask(4) = 0x00000000
[  398.582704] syncpt_thresh_int_mask(5) = 0x00000000
[  398.587495] syncpt_thresh_int_mask(6) = 0x00000000
[  398.592284] syncpt_thresh_int_mask(7) = 0x00000000
[  398.597071] syncpt_thresh_int_mask(8) = 0x00000000
[  398.601860] syncpt_thresh_int_mask(9) = 0x00000000
[  398.606653] syncpt_thresh_int_mask(10) = 0x00000000
[  398.611529] syncpt_thresh_int_mask(11) = 0x00000000
[  398.616407]
               ---- host syncpt irq status ----

[  398.623627] syncpt_thresh_cpu0_int_status(0) = 0x00000000
[  398.629030] syncpt_thresh_cpu0_int_status(1) = 0x00000000
[  398.634427] syncpt_thresh_cpu0_int_status(2) = 0x00000000
[  398.639826] syncpt_thresh_cpu0_int_status(3) = 0x00000000
[  398.645223] syncpt_thresh_cpu0_int_status(4) = 0x00000000
[  398.650620] syncpt_thresh_cpu0_int_status(5) = 0x00000000
[  398.656014]
               ---- host syncpt thresh ----

[  398.662889] syncpt_int_thresh_thresh_0(0) = 1
[  398.667250] syncpt_int_thresh_thresh_0(13) = 6904
[  398.672027] gm20b pbdma 0:
[  398.674639] id: 6 (tsg), next_id: 6 (tsg) chan status: valid
[  398.680303] PBDMA_PUT: 0000001e00008678 PBDMA_GET: 0000001e00008600 GP_PUT: 00000074 GP_GET: 00000072 FETCH: 00000074 HEADER: 60030100
               HDR: 60116040 SHADOW0: 000085e4 SHADOW1: 0000941e

[  398.698113] gm20b eng 0:
[  398.700563] id: 0 (channel), next_id: 6 (tsg), ctx status: load
[  398.706567] busy

[  398.708508] gm20b eng 1:
[  398.710946] id: 0 (channel), next_id: 0 (channel), ctx status: invalid


[  398.720729] 501-gm20b, pid 11802, refs 2:
[  398.724659] channel status:  in use idle not busy
[  398.729364] RAMFC : TOP: 8000001f0000c030 PUT: 0000001f0000c030 GET: 0000001f0000c030 FETCH: 0000001f0000c030
               HEADER: 60400000 COUNT: 80000000
               SYNCPOINT 00000000 00000e01 SEMAPHORE 00000000 00000000 00000000 00000000

[  398.752820] 502-gm20b, pid 11802, refs 8:
[  398.756746] channel status:  in use on_pbdma_and_eng busy
[  398.762147] RAMFC : TOP: 8000001f00006f60 PUT: 0000001f00006f60 GET: 0000001f00006f60 FETCH: 0000001f00006f60
               HEADER: 60400000 COUNT: 80000000
               SYNCPOINT 00000000 00000d01 SEMAPHORE 0000000d fc001000 05e82179 01100002

[  398.785590] 503-gm20b, pid 9069, refs 2:
[  398.789425] channel status:  in use idle not busy
[  398.794131] RAMFC : TOP: 8000001f00280288 PUT: 0000001f00280288 GET: 0000001f00280288 FETCH: 0000001f00280288
               HEADER: 60400000 COUNT: 80000000
               SYNCPOINT 00000000 00000c01 SEMAPHORE 00000000 00000000 00000000 00000000

[  398.817570] 504-gm20b, pid 9069, refs 2:
[  398.821406] channel status:  in use idle not busy
[  398.826109] RAMFC : TOP: 8000001f00240018 PUT: 0000001f00240018 GET: 0000001f00240018 FETCH: 0000001f00240018
               HEADER: 60400000 COUNT: 80000000
               SYNCPOINT 00000000 00000b01 SEMAPHORE 00000000 00000000 00000000 00000000

[  398.849564] 505-gm20b, pid 9069, refs 2:
[  398.853400] channel status:  in use idle not busy
[  398.858106] RAMFC : TOP: 8000001f00200018 PUT: 0000001f00200018 GET: 0000001f00200018 FETCH: 0000001f00200018
               HEADER: 60400000 COUNT: 80000000
               SYNCPOINT 00000000 00000901 SEMAPHORE 00000000 00000000 00000000 00000000

[  398.881544] 506-gm20b, pid 9069, refs 2:
[  398.885382] channel status:  in use idle not busy
[  398.890086] RAMFC : TOP: 8000001f00171de8 PUT: 0000001f00171de8 GET: 0000001f00171de8 FETCH: 0000001f00171de8
               HEADER: 60400000 COUNT: 80000000
               SYNCPOINT 00000000 00000801 SEMAPHORE 00000001 0002fff0 0000367d 00000004

[  398.913527] 507-gm20b, pid 9069, refs 2:
[  398.917360] channel status:  in use idle not busy
[  398.922065] RAMFC : TOP: 8000001f00051c88 PUT: 0000001f00051c88 GET: 0000001f00051c88 FETCH: 0000001f00051c88
               HEADER: 60400000 COUNT: 80000000
               SYNCPOINT 00000000 00000701 SEMAPHORE 00000001 0002ffb0 00000003 00001004

[  398.945506] 508-gm20b, pid 3989, refs 2:
[  398.949341] channel status:  in use idle not busy
[  398.954044] RAMFC : TOP: 0000000000000000 PUT: 0000000000000000 GET: 0000000000000000 FETCH: 0000000000000000
               HEADER: 60400000 COUNT: 00000000
               SYNCPOINT 00000000 00000000 SEMAPHORE 00000000 00000000 00000000 00000000

[  398.977496] 509-gm20b, pid 3989, refs 2:
[  398.981331] channel status:  in use idle not busy
[  398.986035] RAMFC : TOP: 0000000000000000 PUT: 0000000000000000 GET: 0000000000000000 FETCH: 0000000000000000
               HEADER: 60400000 COUNT: 00000000
               SYNCPOINT 00000000 00000000 SEMAPHORE 00000000 00000000 00000000 00000000

[  399.009476] 510-gm20b, pid 3989, refs 2:
[  399.013313] channel status:  in use idle not busy
[  399.018017] RAMFC : TOP: 0000000000000000 PUT: 0000000000000000 GET: 0000000000000000 FETCH: 0000000000000000
               HEADER: 60400000 COUNT: 00000000
               SYNCPOINT 00000000 00000000 SEMAPHORE 00000000 00000000 00000000 00000000

[  399.041462] 511-gm20b, pid 3989, refs 2:
[  399.045297] channel status:  in use idle not busy
[  399.050001] RAMFC : TOP: 0000000000000000 PUT: 0000000000000000 GET: 0000000000000000 FETCH: 0000000000000000
               HEADER: 60400000 COUNT: 00000000
               SYNCPOINT 00000000 00000000 SEMAPHORE 00000000 00000000 00000000 00000000

[  399.073687] nvgpu: 57000000.gpu gk20a_fifo_handle_mmu_fault_locked:1721 [ERR]  fake mmu fault on engine 0, engine subid 0 (gpc), client 0 (l1 0), addr 0x547b119000, type 4 (unbound inst), access_type 0x00000001,inst_ptr 0x20fc8000
[  399.094101] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:129  [ERR]  gr_fecs_os_r : 0
[  399.102708] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:131  [ERR]  gr_fecs_cpuctl_r : 0x40
[  399.111930] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:133  [ERR]  gr_fecs_idlestate_r : 0x1
[  399.121320] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:135  [ERR]  gr_fecs_mailbox0_r : 0x0
[  399.130624] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:137  [ERR]  gr_fecs_mailbox1_r : 0x0
[  399.139921] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:139  [ERR]  gr_fecs_irqstat_r : 0x0
[  399.149133] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:141  [ERR]  gr_fecs_irqmode_r : 0x4
[  399.158343] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:143  [ERR]  gr_fecs_irqmask_r : 0x8704
[  399.167817] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:145  [ERR]  gr_fecs_irqdest_r : 0x0
[  399.177026] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:147  [ERR]  gr_fecs_debug1_r : 0x40
[  399.186239] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:149  [ERR]  gr_fecs_debuginfo_r : 0x0
[  399.195624] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:151  [ERR]  gr_fecs_ctxsw_status_1_r : 0x304
[  399.205615] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(0) : 0x1
[  399.215605] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(1) : 0x0
[  399.225599] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(2) : 0x90009
[  399.235944] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(3) : 0x14000
[  399.246287] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(4) : 0x1ffda0
[  399.256710] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(5) : 0x0
[  399.266702] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(6) : 0x0
[  399.276691] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(7) : 0x0
[  399.286684] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(8) : 0x0
[  399.296673] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(9) : 0x0
[  399.306667] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(10) : 0x0
[  399.316745] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(11) : 0x0
[  399.326826] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(12) : 0x0
[  399.336902] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(13) : 0x0
[  399.346981] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(14) : 0x0
[  399.357058] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:155  [ERR]  gr_fecs_ctxsw_mailbox_r(15) : 0x0
[  399.367136] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:159  [ERR]  gr_fecs_engctl_r : 0x0
[  399.376269] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:161  [ERR]  gr_fecs_curctx_r : 0x0
[  399.385396] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:163  [ERR]  gr_fecs_nxtctx_r : 0x0
[  399.394519] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:169  [ERR]  FECS_FALCON_REG_IMB : 0xbadfbadf
[  399.404521] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:175  [ERR]  FECS_FALCON_REG_DMB : 0xbadfbadf
[  399.414519] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:181  [ERR]  FECS_FALCON_REG_CSW : 0xbadfbadf
[  399.424517] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:187  [ERR]  FECS_FALCON_REG_CTX : 0xbadfbadf
[  399.434511] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:193  [ERR]  FECS_FALCON_REG_EXCI : 0xbadfbadf
[  399.444597] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:200  [ERR]  FECS_FALCON_REG_PC : 0xbadfbadf
[  399.454506] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:206  [ERR]  FECS_FALCON_REG_SP : 0xbadfbadf
[  399.464414] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:200  [ERR]  FECS_FALCON_REG_PC : 0xbadfbadf
[  399.474321] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:206  [ERR]  FECS_FALCON_REG_SP : 0xbadfbadf
[  399.484230] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:200  [ERR]  FECS_FALCON_REG_PC : 0xbadfbadf
[  399.494138] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:206  [ERR]  FECS_FALCON_REG_SP : 0xbadfbadf
[  399.504046] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:200  [ERR]  FECS_FALCON_REG_PC : 0xbadfbadf
[  399.513952] nvgpu: 57000000.gpu      gk20a_fecs_dump_falcon_stats:206  [ERR]  FECS_FALCON_REG_SP : 0xbadfbadf
[  399.523862] nvgpu: 57000000.gpu gk20a_fifo_handle_mmu_fault_locked:1726 [ERR]  gr_status_r : 0x81
[  399.533597] nvgpu: 57000000.gpu                    fifo_error_isr:2605 [ERR]  channel reset initiated from fifo_error_isr; intr=0x00000100

Hi,

We have seen this issue before when some forum users are running gpu stress.

Could you share what application you are running now? Also, could you move to rel-32.4.3 and try again?
I guess the issue is still but still worth trying.

And…I notice you say “on boot”. Why do you mention that? Do you hit this issue right after you boot up?

Isn’t this issue happened after you run the application?

Hi, i have runnung gpu stress by running about 15 oceanFFT and i have the same error dmesg info ,could you help ?dmesg_log_file.txt (253.8 KB)

You don’t have enough RAM:

[225745.639156] Out of memory: Kill process 10126 (oceanFFT) score 27 or sacrifice child
[225745.698914] Killed process 10126 (oceanFFT) total-vm:8613140kB, anon-rss:0kB, file-rss:31652kB, shmem-rss:0kB
[225745.807352] oom_reaper: reaped process 10126 (oceanFFT), now anon-rss:0kB, file-rss:31644kB, shmem-rss:0kB

Note that a GPU requires physical RAM, and swap will not help (indirectly it could help if other processes are swapped out and more physical RAM becomes available, but you’ve reach the memory limits).

Ok,when this case happens ,it makes my hdmi monitor without any display ,is there any way can i recovery it with showing display on the monitor?

Could you share the exact steps to reproduce this issue? Just running oceanFFT for 15 hours without any other configuration?

1.first step i connect the hdmi monitor , mouse and keyboard on board powering on .
2. second i log in using nvidia account ,then i run oceanFFT programs about 15 numbers typing “ps -e” you can find 15 processes such as the picture shows 图片

  1. third after i keep running serveral hours or days it happens this case . the case happens may several hours or several days which has no fixed time.

Are you saying that after this failed the system no longer boots correctly? When you run out of RAM it can force instant reboot without completing writes to the filesystem, and thus might corrupt if it exceeds the journal capacity. If you are just trying to get it to run all of those programs and not fail, then running out of RAM will always occur and do this.

It seems to happen intermittently on boot. Sometimes I see it within a few minutes of booting (around when the GPU / CUDA drivers are loaded, but not running any inference or other intensive apps) and other times it happens intermittently during running of GPU apps. I realize this is not a lot to go on, but it’s what I’m dealing with… I did notice some unusual items in the startup logs relating to the imx219 camera, curious if this could be a cause of the issue?

[    3.802857] i2c /dev entries driver
[    3.803991] i2c i2c-6: Added multiplexed i2c bus 7
[    3.804258] i2c i2c-6: Added multiplexed i2c bus 8
[    3.804263] i2c-mux-gpio cam_i2cmux: 2 port mux on Tegra I2C adapter adapter
[    3.804740] imx219 7-0010: tegracam sensor driver:imx219_v2.0.6
[    3.858118] usb 1-2: new high-speed USB device number 2 using tegra-xusb
[    3.867096] imx219 8-0010: tegracam sensor driver:imx219_v2.0.6
[    3.894697] tegra-vii2c 546c0000.i2c: no acknowledge from address 0x10
[    3.894771] imx219 8-0010: imx219_board_setup: error during i2c read probe (-121)
[    3.894799] imx219 8-0010: board setup failed
[    3.894862] imx219: probe of 8-0010 failed with error -121

Not yet , i want to recovery it no reboot ,is there any way ? i unpluggin the hdmi monitor when the board runs several hours ,then i connect the hdmi monitor again ,it shows nothing but black screen.

To drcnyc,

That imx219 log is not related. I see it every time. This issue is definitely triggered by GPU. Actually, as I know there are two gpu issues

  1. On boot, you shall see something from gpu driver related to PMU. This is not fatal.

  2. When running intensive GPU log, a timeout error will be printed as your log. This one is fatal.

If you read the log carefully, these two error should have different error logs.

To wangzhaofeng

I think you could try to rmmod nvpug and modprobe it again.

@WayneWWW one more thing I just noticed that could be a clue, the error seems to happen on startup whenever there was a failure during GPU processing that caused a hard reboot. This is probably a long shot, but the error happens right after initializing the GPU and the message references the GPU being “locked” – I wonder if this has to do with the fact that it hard reboot during a running inference session?

This still doesn’t explain why it’s rebooting during hard usage, but it could explain the startup issue? And if so would there be any way to “unlock” the GPU?

Hi,

I think we should clarify how many errors you have here.

It seems you have one error that happens right after device boot up. -> What is the error log of this one?
Same as below?

gk20a_fifo_handle_sched_error:2531 [ERR] fifo sched ctxsw timeout error: engine=0, tsg=6, ms=3100
[ 398.432169] ---- mlocks ----

i have already try it ,but it failed
rmmod: ERROR: Module nvgpu is in use
and i try rmmod -f nvgpu
rmmod: ERROR: …/libkmod/libkmod-module.c:793 kmod_module_remove_module() could not remove ‘nvgpu’: Resource temporarily unavailable
rmmod: ERROR: could not remove module nvgpu: Resource temporarily unavailable

cause the oceanFFT use gpu module but i cannot kill the oceanFFT success by kill -9 pid

wangzhaofeng,

How about using “modprobe -r” to remove it?

modprobe -r nvgpu
modprobe: FATAL: Module nvgpu is in use.

the same error.

rmmod -f to force remove it?

1 Like

@WayneWWW Yes - the error above is the one that occurs on boot. I actually don’t have the dmesg output of the error that occurs during GPU usage because it is an instant shutoff - no logs or records of what is causing it. It’s as if the power is shut off instantly. Could this be an out of memory issue? You would think it would be a softer shutdown if so, with errors etc…