Mystery Segfaults Solved

After much painful debugging, I finally found the cause of the segfaults I was seeing on EABI images, particularly with glibc 2.4/2.5. It turned out to be a cache flushing problem affecting StrongARM, XScale and ARMv6 processors. The bug must have been in the kernel for years and only triggered by the changed usage patterns in glibc 2.4/2.5.

I submitted a patch for the problem which has been accepted into the mainline kernel and will be in 2.6.20.
(, I’d recommend anyone using the affected processors to apply this patch.

Comments are closed.