kexec boot fails with a kernel panic

  • 7023983
  • 01-Jul-2019
  • 02-Jul-2019

Environment

SUSE Linux Enterprise Server 15 Service Pack 1

Situation

With kexec boot the system could fail to boot further with an error similar to the below:

Kernel panic - not syncing: ERROR: Failed to allocate 0x0000000000004000 bytes below 0x0000000020000000.

CPU: 0 PID: 0 Comm: swapper Not tainted 4.12.14-195.gbb021ec-default #1 SLE15-SP1
Call Trace:
[c00000000122fcf0] [c000000000a19368] dump_stack+0xb0/0xf0 (unreliable)
[c00000000122fd30] [c000000000a15efc] panic+0x144/0x31c
[c00000000122fdc0] [c000000000d3c8e4] memblock_alloc_base+0x44/0x58


This issue is seen on large configuration systems when the "crashkernel=" parameter is passed to the kernel.

Resolution

Working on a fix to ensure the memory available in the first memory block (other than what is used for kdump) is larger than 128MB. This change can have an impact on older configurations. While a fix that works for all cases is arrived at, use an offset greater than 128MB for crashkernel during kexec boot. For example:

kexec -l --append="`cat /proc/cmdline` crashkernel=X@160M" --initrd=/boot/initrd-`uname -r` /boot/vmlinux-`uname -r`

Where X is the memory to be reserved for kdump.

Cause

System resources need memory to be reserved in the first memory block. While the size of the first memory block is usually 512MB or greater, memory for crashkernel is reserved at an offset of 128MB by default. This leaves less than 128MB memory to reserve for system resources that need memory in the first memory block. On large configuration systems this memory may not be sufficient, especially with kexec boot where initrd is also placed below 128MB adding to these memory constraints and eventually leading to an allocation failure.

Feedback service temporarily unavailable. For content questions or problems, please contact Support.