Re: What's become with PPC patches?

* Sergei Shtylyov ([email protected]) wrote:
> Hello.
> 
> Mathieu Desnoyers wrote:
> 
> >>>>Since we were about to start working on LTTng in arch/powerpc/ tree (at
> >>>>last), I've looked thru the current 2.6.18 patches, and was surprised 
> >>>>in a bad
> >>>>way about PPC patches which only looked like a shadow of what they were
> >>>>several months ago, largely reminding the result of a failed rebase 
> >>>>with only
> >>>>some added lines and braces serving no particular purpose. Looks like 
> >>>>that
> >>>>disaster (I think it's the right word) happened after 
> >>>>2.6.17-lttng-0.5.108.
> >>>>Mattieau, could you clarify what happened? The patch has lost a lot of 
> >>>>trace points (assmingly borrowed from my last year's LTT work), so it's 
> >>>>going to be quite a setback for us, and completely unexpected... :-(
> 
> >>>Can you point to some examples of the trace points that went away ? I 
> >>>will
> >>>probably be able to explain where they are now : most likely in 
> >>>architecture
> >>>independant locations.
> 
> >>  Look at arch/powerpc/mm/fault.c in your patchset, for example...
> 
> >Look at ltt/probes/ltt-probe-ppc.c: probe_mm_handle_fault_entry/exit
> >And also at 
> 
> >mm/memory.c:       MARK(mm_handle_fault_entry, "%lu %ld", address, 
> >                    KSTK_EIP(current));
> >[...]
> >mm/memory.c:       MARK(mm_handle_fault_exit, MARK_NOARGS);
> 
> >Note that there should also be a call to trace_kernel_trap_entry/exit in
> >probe_mm_handle_fault_entry/exit in ltt/probes/ltt-probe-ppc.c (see
> >ltt-probe-i386.c for an example). I will fix in the next releases.
> 
>    Hm, that's all good... but where's the syscall entry/exit hook gone from 
> arch/powerpc/kernel/entry_32.S?
> 

arch/powerpc/kernel$ grep MARK *

process.c:      MARK(kernel_thread_create, "%ld %p", retval, fn);
process.c:#define REGS_MARKER   0x7265677368657265ul
process.c:#define FRAME_MARKER  12
process.c:#define REGS_MARKER   0x72656773ul
process.c:#define FRAME_MARKER  2
process.c:                  && stack[FRAME_MARKER] == REGS_MARKER) {
ptrace.c:       MARK(kernel_syscall_entry, "%ld struct pt_regs %p",
ptrace.c:       MARK(kernel_syscall_exit, MARK_NOARGS);
syscalls.c:     MARK(ipc_call, "%u %d", call, first);
time.c: MARK(kernel_trap_entry, "%ld struct pt_regs %p", regs->trap, regs);
time.c: MARK(kernel_trap_exit, MARK_NOARGS);
traps.c:        MARK(kernel_trap_entry, "%ld struct pt_regs %p", regs->trap,
regs);
traps.c:        MARK(kernel_trap_exit, MARK_NOARGS);
traps.c:        MARK(kernel_trap_entry, "%ld struct pt_regs %p", regs->trap,
regs);
traps.c:        MARK(kernel_trap_exit, MARK_NOARGS);
traps.c:        MARK(kernel_trap_entry, "%ld struct pt_regs %p", regs->trap,
regs);
traps.c:        MARK(kernel_trap_exit, MARK_NOARGS);


So look at ptrace.c,

Mathieu


OpenPGP public key:              http://krystal.dyndns.org:8080/key/compudj.gpg
Key fingerprint:     8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68