Hibernate/Resume failure with galcore (gpu-viv) driver

[ripplejr]

I'm working with an i.MX8QXP processor and am having inconsistent crashing during resume from hibernation (suspend to disk). I've narrowed it down to the galcore module causing the issue since it the crash happens only when the galcore module (gpu-viv driver) is loaded. The system will save the image fine but during resume it will fail to decompress the image in every 1 in 4 resumes. The only consistent work fixes that I've found are either using the resume=nocompress option so lzo copmression is never used (causes a large slow down with saving or reading the image) or by setting acpi=off in the kernel options.

This problem has shown up on both the 5.4 and 5.10 kernels.

This problem looks similar to this issue but the patch provided is for a different system.

https://community.nxp.com/t5/i-MX-Processors/GPU-hang-after-multiple-suspend-resume-cycles/m-p/96209...

Steps to reproduce:

1. echo reboot > /sys/power/disk; pm-hibernate

Error Log during boot:

[ 1.809743] PM: hibernation: resume from hibernation
[ 1.815829] Freezing user space processes ... (elapsed 0.000 seconds) done.
[ 1.822933] OOM killer disabled.
[ 1.826197] Freezing remaining freezable tasks ...
[ 1.828292] mmc1: SDHCI controller on 5b020000.mmc [5b020000.mmc] using ADMA
[ 1.903968] (elapsed 0.077 seconds) done.
[ 1.951290] PM: Using 3 thread(s) for decompression
[ 1.956288] PM: Loading and decompressing image data (201761 pages)...
[ 1.963110] hibernate: Hibernated on CPU 0 [mpidr:0x0]
[ 2.103104] PM: Image loading progress: 0%
[ 2.799921] imx6q-pcie 5f010000.pcie: Phy link never came up
[ 2.805756] imx6q-pcie 5f010000.pcie: failed to initialize host
[ 2.811730] imx6q-pcie 5f010000.pcie: unable to add pcie port.
[ 3.033214] PM: Image loading progress: 10%
[ 3.518651] PM: Image loading progress: 20%
[ 3.764275] PM: Image loading progress: 30%
[ 4.013727] PM: Image loading progress: 40%
[ 4.027386] PM: LZO decompression failed
[ 4.031422] PM: hibernation: Read 807044 kbytes in 2.06 seconds (391.76 MB/s)
[ 4.046593] PM: hibernation: Failed to load image, recovering.
[ 4.319290] OOM killer enabled.
[ 4.322518] Restarting tasks ... done.
[ 4.327082] PM: hibernation: resume failed (-4)

[Bio_TICFSL]

Hello ripplejr,

Please do you use the latest BSP? Can you please state the detailed duplicated steps?

Can you please provide the detailed failure log for the failure? It hangs before suspend or after resume?

Do this happens only for Android?

Regards

[ripplejr]

I'm currently using the 5.10.72 release.

Steps to reproduce.

1) echo disk > /sys/power/state

This is the only step to do, if hibernation is enabled in the kernel config and there is a swap space for the image then running the above command will cause the issue. This only happens in 1 in 4 hibernations and only on the resume. When the kernel is loading the image during resume, LZO decompression fails. This is not on Android but a custom Linux build.

[Bio_TICFSL]

ok, it looks like and issue, don't enable at the moment the kernel hypervisor, will check with the developers.

Thanks for the catch!

[ripplejr]

I tried turning off the hypervisor but it doesn't seem to do anything. The LZO compression error still shows up.

[Bio_TICFSL]

Hi,

The i.MX8QXP do not support hypervisor, it support jailhouse but in a teoricall manner, the Hypervisor is not part of the BSP for MX8QXP.

Regards

[ripplejr]

Hi,

Thanks for the help. I wanted to ask if there was any update on a potential fix?