Kernel level memory leak 36.3 nvarguscamerasrc

Robotics & Edge Computing Jetson Systems Jetson Orin Nano
, ,
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) Jetson Orin Nano
• DeepStream Version 7
• JetPack Version (valid for Jetson only) 6.0
• TensorRT Version 8.6.2.3 cuda 12.2
• Issue Type( questions, new requirements, bugs) bug

After applying the patch recommended in the last post:

Now there’s a leak with kmalloc-256 when running a simple pipeline:

gst-launch-1.0 nvarguscamerasrc ! fakesink

Applying kmemleak to the kernel shows the issue as:

unreferenced object 0xffff0001688f4d00 (size 256):
  comm "EglStrmComm*111", pid 2129, jiffies 4294917016 (age 969.964s)
  hex dump (first 32 bytes):
    00 03 a6 cd 00 00 ff ff 18 fe 58 af 99 d8 ff ff  ..........X.....
    10 4d 8f 68 01 00 ff ff 10 4d 8f 68 01 00 ff ff  .M.h.....M.h....
  backtrace:
    [<00000000cc7e2574>] kmem_cache_alloc_trace+0x2bc/0x3d0
    [<0000000067d879be>] host1x_fence_create+0x44/0xe0 [host1x]
    [<0000000056bfb32b>] dev_file_ioctl+0xb4/0x858 [host1x_fence]
    [<00000000d41d542f>] __arm64_sys_ioctl+0xb4/0x100
    [<00000000085bc6eb>] invoke_syscall+0x5c/0x130
    [<0000000063137684>] el0_svc_common.constprop.0+0x64/0x110
    [<00000000d5712572>] do_el0_svc+0x74/0xa0
    [<0000000008ecc488>] el0_svc+0x28/0x80
    [<00000000432b8db2>] el0t_64_sync_handler+0xa4/0x130
    [<00000000a7bb9996>] el0t_64_sync+0x1a4/0x1a8

We can’t upgrade to JP 6.1 yet because it doesn’t support Deepstream 7, and there’s no updated deepstream available.

Is there a workaround for the additional leak on kmalloc-256?
Is this fixed in the next JP?

5

It’s know issue we are checking it.

Thanks