Description of problem:
diskdump freeze on Opteron dual core with Athlon SMP kernel.
Version-Release number of selected component (if applicable):
RHEL3 Update 6 (2.4.21-37). 7 (2.4.21-40) and 8 (2.4.21-47)
Arch Athlon SMP kernel
H/W HP DL385 (Opteron 280 dual core x 2)
Scsi card: SmartArray 6i
diskdumputils 1.1.9-3, diskdumputils-1.2.8-1, diskdumputils-1.3.3
Diskdump fail 8-9 times in 10 diskdump test.
Steps to Reproduce:
1. Install OS
2. Setup diskdump
3. ALT+SysRQ+C from keyboard or echo c > /proc/sysrq-trigger
diskdump failed to dump 8-9 times in 10 test.
Display only 3 lines, then system freeze until power cycle.
# echo c > /proc/sysrq-trigger
SysRq: Crashing the kernel by request
Unable to handle kernel NULL pointer dereference at virtual address 00000000
printing eip: c01c8cd0 *pde=46805067
Diskdump dump memory image correctly, all the time.
(1) Same DL385 with i686smp, i686hugemem, athlon-uni-processor dump memory
(2) Same DL385 with Athlon SMP kernel, boot with "notsc" solve this problem.
(3) Same DL385 with Athlon SMP kernel, load crash.o into kernel always panic
the system succesfully. So this symptom only reproducable with
(4) kernel(s) and diskdumputils U6 to U8 are tested. No successful result.
purpose of this Bugzilla are,
(1) Report a bug
(2) Ask for fix (if possible)
(3) Understand why "notsc" solve this problem.
I have a bad news.
Add "notsc" didn't fix the problem.
I have done test with some different memory size configuration such as,
512m, 1000m, 2000m and 4000m. With 512m, it works.
With 1000m, 2000m and 4000m, diskdump still failed intermittently
even if I add "notsc".
Masanari, is this still an issue?
Thanks for reply.
The box has migrated to x86 kernel.
So this is not an issue for me now.