On Mon, Dec 8, 2008 at 5:39 PM, John David Anglin <dave@xxxxxxxxxxxxxxxxxx> wrote: >> Otherwise nothing in glibc should make futex calls. >> >> Broken futex calls shouldn't crash the kernel. > > The kernel doesn't crash since Helge's stack backtrace fix. We are > just getting a lot of segfaults in application code. Excellent. That is certainly progress. > I'm fairly certain libc6 in debian testing is still built with linuxthreads. > I'm not sure libc is actually making futex calls, but one is compiled > into _dl_fixup. I suppose I could add a printk in the kernel to see > if futex's are being used. Yes, your assembly dump shows a futex syscall in _dl_fixup, and AFAIK there should not be one there. I used printk *very* effectively in the kernel/futex.c code to debug futex issues with the NPTL bringup. Remember that the futex is essentially the address of the userspace object, so it's rather easy to trace. > I'm currently building debian libc6. In a quick attempt, I wasn't > successful in building the glibc cvs source. It pushes hard to get > you to configure with nptl. My build with linuxthreads failed with > a type size problem. > > I'll try the testsuite with this build on my current kernel. If that > works, I'll flip to 2.6.28. Thanks, please keep me posted. Cheers, Carlos. -- To unsubscribe from this list: send the line "unsubscribe linux-parisc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
