Kernel warning during Debian boot


#1

I flashed latest Debian (release 260) to Dragonboard from SD card. It boots and works, but kernel has a crash at boot time. The crash happens also in releases 246 and 252. What am I doing wrong?
The info is below. I uploaded full log to https://pastebin.com/tTUQ7WsL

[ 9.870937] msm_cci_init_gpio_params:1383 gpio count 0
[ 9.880711] iommu: Adding device 1d00000.video-codec to group 0
[ 9.895679] 1d00000.video-codec:firmware: assigned reserved memory node venus@89900000
[ 10.018393] iommu: Adding device 1b0ac00.camss to group 1
[ 10.025690] ------------[ cut here ]------------
[ 10.032663] WARNING: CPU: 2 PID: 1577 at …/arch/arm64/mm/dma-mapping.c:843 arch_teardown_dma_ops+0x28/0x48
[ 10.032689] Modules linked in: qcom_camss(+) msm_rng rng_core venus_core videobuf2_dma_sg mdt_loader videobuf2_memops v4l2_mem2mem videobuf2_v4l2 qcom_cci videobuf2_core ip_tables x_tables

[ 10.032697] CPU: 2 PID: 1577 Comm: systemd-udevd Not tainted 4.9.39-linaro-lt-qcom #1
[ 10.032698] Hardware name: Qualcomm Technologies, Inc. APQ 8016 SBC (DT)
[ 10.032700] task: ffff800036675b00 task.stack: ffff800035f60000
[ 10.032704] PC is at arch_teardown_dma_ops+0x28/0x48
[ 10.032707] LR is at arch_teardown_dma_ops+0x14/0x48


#2

Thanks for reporting, Actually this is not a crash but a warning which is then correctly managed by the kernel.
Seems the issue comes from a probing issue in the qcom_camss_driver which returns an error.
The attached dma is then ‘deconfigured’ (arch_teardown_dma_ops) whereas an iommu domain is still attached, this should be fixable.


#3

Hi everyone,

i have the exact same issue as the poster. i know its not a kernel error. But i am trying to load the qcom_camss_driver, this is stopping me to get it work.

Loic mentioned its dma issue, is there a solution for this?

Thanks so much in advance for your help
Jay