Red Hat Bugzilla – Bug 101040
Massive oops & filesystem corruption on AMD Athlon
Last modified: 2007-11-30 17:06:57 EST
Description of problem: After installing Taroon Beta 1 on my AMD Athlon Compaq Presario laptop, I booted it up and found a lot of oopses during boot. The kernel is oopsing at "do_generic_file_read" at a very fast rate which overflows the "dmesg" kernel ring buffer. I have attached to thig bug report the output of "dmesg", plus "lspci" and a copy of "/proc/ioports" and "/proc/interrupts". Version-Release number of selected component (if applicable): Stock Red Hat Linux Taroon Beta 1 kernel. How reproducible: Always. Steps to Reproduce: 1. Install Taroon AS Beta 1 on a Compaq Presario 706EA, or a similar computer, since I don't know if this problem is related to hardware configuration. 2. Choose to install "Everything" 3. After installation and upon rebooting, the kernel starts oopsing. The result of the "dmesg" command after booting into runlevel 1 is attached in the bug report. Actual results: The system oopses continuously and ends up corrupting the filesystem, requiring a full system reinstall. Expected results: The system should be stable with the selected hardware configuration. At least, Red Hat Linux Beta is completely stable on the same machine. Additional info: The problem is reproducible with either ext2 or ext3 filesystems. The machine used to install RHL AS is a Compaq Presario 706EA.
Created attachment 93190 [details] output of the "dmesg" command The kernel oopses at a very fast rate, and thus, I've been unable to capture the full kernel ring as it overflows.
Created attachment 93191 [details] output of the "lspci -vvv" command
Created attachment 93192 [details] output of "cat /proc/interrupts"
Created attachment 93193 [details] output of "cat /proc/ioports"
Thank you for your detailed bug report. Red Hat is aware of this problem, and several options for fixing it are currently under review. The original bug report is #99507. *** This bug has been marked as a duplicate of 99507 ***
*** Bug 101481 has been marked as a duplicate of this bug. ***
I can't see bug 99507, but I'm getting the same problem on an untainted NForce2 based platform.
Nitin, I didn't want to add your "outside" e-mail address to the cc: list of bug #99507 (which is restricted to beta partners), so here is a copy of my final comment for 99507 posted on 2003-07-31: This problem is believed to be caused by a flaw in the Athlon XP cpu, which under unusual circumstances, allows a "prefetch" machine instruction with certain alignments and/or cache conflicts to cause a page fault on address 0 (in violation of the instruction's functional description). A work-around has been committed to Taroon today to the Athlon-specific prefetch routine to test for NULL addresses. The change is applied by linux-2.4.18-smallpatches.patch to include/asm-i386/processor.h, and will be incorporated into any kernel version 2.4.21-1.1931.2.364 or later (and will thus be part of Taroon B2).
Changed to 'CLOSED' state since 'RESOLVED' has been deprecated.