From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 5.5; Windows NT 5.0)
Description of problem:
When using a large number of Linux software RAID arrays I receive a kernel
OOPS. With some help from users on the kernel mailing list and the linux-raid
list we have determined that /proc/mdstat is overflowing it's 4k page and
overwriting some other part of kernel memory causing the OOPS.
This problems appears to mainfest itself around 24-27 RAID1 arrays.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Boot with 2.4.18-26.7.x
2. Create a large number of software RAID1 arrays.
3. Watch the kernel go bye-bye
Actual Results: Kernel OOPS in do_try_to_free_pages
Expected Results: No OOPS
Created attachment 90582 [details]
Patch to convert /proc/mdstat to use seq_file
The attached patch seems to resolve the OOPS I was seeing. I currently have 52
SCSI disks in 26 RAID1 arrays and one RAID0 stripe across those. It has been
running successfully on test hardware under load for approximately 24 hours
As an FYI, the 2.4.20-x series of errata kernels seem to have switched
to seq_file for md.