Oracle Linux: Sched/Fair: "Prevent a Division by 0 in Scale_rt_Capacity() " or "Avoid Divide by Zero When Rebalancing Domains" (Doc ID 2678431.1)

Last updated on JUNE 16, 2020

Applies to:

Symptoms

Server crashed on below code.

KERNEL: /share/linuxrpm/vmlinux_repo/64/4.14.35-1902.5.1.4.el7uek.x86_64/vmlinux
DUMPFILE: vmcore [PARTIAL DUMP]
CPUS: 384
DATE: Fri Nov 29 21:51:45 2019
UPTIME: 7 days, 02:10:11
LOAD AVERAGE: 68.54, 61.76, 51.40
TASKS: 12239
NODENAME: hostname
RELEASE: 4.14.35-1902.5.1.4.el7uek.x86_64
VERSION: #2 SMP Wed Oct 9 19:29:16 PDT 2019
MACHINE: x86_64 (2700 Mhz)
MEMORY: 6142.6 GB
PANIC: "kernel BUG at mm/mmap.c:636!"
PID: 336404
COMMAND: "java"
TASK: ffff996c1e3b3d80 [THREAD_INFO: ffff996c1e3b3d80]
CPU: 260
STATE: TASK_RUNNING (PANIC)

from further review kernel/sched/fair.c: had this change.

- used = div_u64(avg, total);
+ used = div64_u64(avg, total);

Cause

My Oracle Support provides customers with access to over a million knowledge articles and a vibrant support community of peers and Oracle experts.