Hi: David Miller wrote: [Thu Sep 27 2012, 10:15:54PM EDT] > From: Bob Picco <bpicco@xxxxxxxxxx> > Date: Thu, 27 Sep 2012 11:36:45 -0400 > > > do_sigaltstack expects the stack we are returning to. An issue will manifest > > with an alternate stack. That is, a sigaltstack followed by a sigaction > > (aka. rt_sigaction) with SA_ONSTACK for flags will sigsegv. do_sigaltstack > > returns -EPERM having found on_sig_stack to be true and a sigsegv results. > > > > Let's teach do_rt_sigreturn to provide do_sigaltstack with our target stack. > > "sf" is our target stack frame in the stack frame pointed to by > UREG_FP. > > You really haven't described the real reason why this doesn't work > as-is, and therefore your commit message is inadequate and made me do > a lot of work I shouldn't have to do. > > Your commit message implies that we are not passing in something > related to the stack frame, but we actually are. > > And if this location is wrong, why didn't you update all of the other > do_sigaltstack() callers in the sparc port. By your description they > should be wrong too. > > But those cases don't matter like this one does. This case here is > special, but you haven't described what is so unique about this case > which causes only it to need special handling. > > Do you know the real reason why it fails without your change? Sorry about the deficient change log. Won't happen again! It was a couple months ago when I had full context and was perhaps wrong. It was a combination of glibc and kernel investigation/debug and very subtle. I'll review and have a thorough explanation in the change log soon. thanx, bob -- To unsubscribe from this list: send the line "unsubscribe sparclinux" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
