From: Linus Torvalds > Sent: 28 November 2017 17:33 > > On Mon, Nov 27, 2017 at 10:26 PM, Eric W. Biederman > <ebiederm@xxxxxxxxxxxx> wrote: > >> > >> Oh well, I just did /proc/<pid>/stack by making it just print 0 > >> unconditionally rather than the hex number. > > > > Patch? > > Oh, apparently I never pushed out yesterday. > > The patch literally just affects the (useless) hex number. So: > > cat /proc/self/stack > > now prints out > > [<0>] proc_pid_stack+0xaa/0x100 > [<0>] proc_single_show+0x48/0x80 > [<0>] seq_read+0xd2/0x410 > ... > > instead of putting some randomized kernel address there. Not sure I've done it on Linux - getting a hexdump of the stack is hard. But I know I've used the absolute return addresses to help hand-decode the stack. Usually needed to work out which stack frame is which - especially when the stack decode doesn't actually (obviously) contain the addresses of each frame. I don't know how these new stack traceback methods work, but the best one I've seen in the past disassembled forwards remembering the stack offset and unprocessed branch targets until it found a return address. It only had to track %sp and %bp. David
