On Sun, Mar 28, 2010 at 4:22 PM, John David Anglin <dave@xxxxxxxxxxxxxxxxxx> wrote: >> Does this mean that we need to either (a) fix the kernel to stop >> clobbering r19 and r21 or fix userspace? > > r21 is an syscall argument register and it looks like CLOB_ARGS_* > handles it. So, I doubt we need to worry about about r21. Right. >> I'm inclined to think the kernel is wrong, and lazy here, and pt_regs >> should have another PT_GR2_CHILD and PT_GR30_CHILD entries for the >> case above. > > Fixing the kernel will fix broken userspace code. Fixing glibc to > save and restore r19 in non-PIC code will allow new code to work > with broken kernels. Thus, I tend to think both should be fixed. Agreed, I'll get a fix for this into glibc. > Instead of using PT_GR19 and PT_GR21, can't we use PT_SR4 and PT_SR5 > to store r2 and r30 for the child. We don't seem to ever actually > save sr4 or sr5 in the syscall path and we restore fixed values. > This is a hack, but avoids increasing the size of the task structure. I don't like hacks like this. They are confusing and hard to maintain. 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
