Uploaded image for project: 'Lustre'
  1. Lustre
  2. LU-20587

lnetctl peer show -v 2 causes kernel oops

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Unresolved
    • Icon: Medium Medium
    • None
    • None
    • None
    • 3
    • 9223372036854775807

      Command triggering the issue: 
      lnetctl peer show -v 2 
      Summary: Net peer generic-netlink dump NULL dereference with large verbose peer output
      Environment:

      • Lustre client: upstream 2.15.90 also 2.17.x . Doesn't exist in 2.15.8 as the issue LU-9680 is not include there. Where ever the patch LU-9680  exist. 
      • Kernel: 6.8.0-1051-nvidia-64k or  6.8.0-1028-nvidia-64k but it is there in x86 as well.
      • Reproducer: ARM64 client, kernel 6.8.0-1028-nvidia-64k
      • Trigger: lnetctl peer show -v 2
      • Scale: approximately 800 LNet peers. Also tried with 2000
      The stack trace and other details in the dmesg are as following. If the kernel parameter kernel.panic_on_oopsis set, it will panic the system. 

      CPU: 141 PID: 3688398 Comm: lnetctl
      pc : lnet_peer_ni_show_dump+0xa20/0xf10 [lnet]
      x22: 0000000000000000
      Code: 2a1a03f4 b940be61 8b010000 cb160000 (790002c0)
      Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000 

       

            simmonsja James A Simmons
            abooback Vadakkumuri Valappil Abooback
            Votes:
            0 Vote for this issue
            Watchers:
            7 Start watching this issue

              Created:
              Updated: