kernel memory not released

NVIDIA Open GPU Kernel Modules Version

535.161.07

Please confirm this issue does not happen with the proprietary driver (of the same version). This issue tracker is only for bugs specific to the open kernel driver.

• I confirm that this does not happen with the proprietary driver package.

Operating System and Version

Debian GNU/Linux 12 (bookworm)

Kernel Release

Linux kv-research-1 6.1.76 #8 SMP PREEMPT_DYNAMIC Fri Mar 15 01:20:32 PDT 2024 x86_64 GNU/Linux

Please confirm you are running a stable release kernel (e.g. not a -rc). We do not accept bug reports for unreleased kernels.

• I am running on a stable kernel release.

Hardware: GPU

NVIDIA GeForce RTX 4090

Describe the bug

leak.zip
please see the reproduce code attached.

I built the kernel myself with CONFIG_DEBUG_KMEMLEAK=y

the reproduce code is simply cudaHostAlloc a large area of memory and copy some random data into GPU memory.

after the program finishes, do
echo 'scan' > /sys/kernel/debug/kmemleak && cat /sys/kernel/debug/kmemleak

and you will got some leak info. please see attached file.

sometimes, maybe need some time. e.g 2-3min to get the kmemleak info after program finished which I don't know why.
kmemleak.log

To Reproduce

leak.zip

please see the reproduce code

Bug Incidence

Always

nvidia-bug-report.log.gz

nvidia-bug-report.log.gz

More Info

No response

Start with the reproduction code in leak.zip and the attached kmemleak.log; review the reported allocations after cudaHostAlloc and GPU data transfer. Reproduce on Debian 12 with kernel 6.1.76 using CONFIG_DEBUG_KMEMLEAK and the documented scan command. Done means the reported kernel memory is released after the program exits and no corresponding leak remains.