I've encountered a loopback bug on kernel-2.4.3+international_patch. I
haven't been able to pin down what the bug is exactly, but these are the
conditions under which it manifests:
1. connect file to loopback device
2. build fs on loopback device
3. mount fs and use it
4. read the file
5. lockup, about 1 out 5 times.
The lockup occurs randomly and seems to have random effects. Sometimes
user-space scheduling is lost and sometimes there is a total lockup. This
is all I could extract from /var/log/messages in several crashes:
Apr 9 04:04:48 gs176 kernel: kernel BUG at dcache.c:658!
Apr 9 04:04:48 gs176 kernel: invalid operand: 0000
Apr 9 04:04:48 gs176 kernel: CPU: 0
-John
Stats:
Linux version 2.4.3 + patch-int-2.4.3.1
compiler - redhat's hacked 2.96 (I'll try gcc->kgcc next.)
Also appeared with 2.4.2 + patch-2.4.3-pre4 + patch-int-2.4.0.3
compiler egcs-1.1.2