Re: [PATCH][GIT PULL] ftrace: Rename config option HAVE_C_MCOUNT_RECORD to HAVE_C_RECORDMCOUNT

[Steven Rostedt <rostedt@goodmis.org>]

On Fri, 2010-10-15 at 09:38 +0200, Ingo Molnar wrote:
> FYI, a recent tracing change causes this new warning on latest -tip:
> 
> Freeing SMP alternatives: 20k freed
> ftrace: allocating 14134 entries in 28 pages
> ------------[ cut here ]------------
> WARNING: at kernel/trace/ftrace.c:1006 ftrace_bug+0x94/0x1f9()
> Hardware name: System Product Name
> Modules linked in:
> Pid: 0, comm: swapper Not tainted 2.6.36-rc8-tip+ #49604
> Call Trace:
>  [<c102c224>] warn_slowpath_common+0x6a/0x7f
>  [<c10629f3>] ? ftrace_bug+0x94/0x1f9
>  [<c10633d8>] ? ftrace_process_locs+0x152/0x225
>  [<c102c24d>] warn_slowpath_null+0x14/0x18
>  [<c10629f3>] ftrace_bug+0x94/0x1f9
>  [<c1063404>] ftrace_process_locs+0x17e/0x225
>  [<c10633d8>] ? ftrace_process_locs+0x152/0x225
>  [<c13fd7b4>] ftrace_init+0x112/0x12a
>  [<c13ed7ca>] start_kernel+0x287/0x291
>  [<c13ed079>] i386_start_kernel+0x79/0x7b
> ---[ end trace 4eaa2a86a8e2da22 ]---
> ftrace failed to modify [<c10633d8>] ftrace_process_locs+0x152/0x225

I know exactly what happened here, and it has to be because of the C
recordmcount code, because the perl version will not process a file
called ftrace.o

What happened is that ftrace_process_locs() references the mcount call,
and since we are adding relocations of mcount to the __mcount_loc
section, we just added a place that references the call but does not
actually call mcount. So we fail the check when we try to modify a
caller to mcount.

I'll see if I can add a fix to that in the code.

-- Steve

>  actual: b9:44:2b:00:c1
> Enabling APIC mode:  Flat.  Using 1 I/O APICs
> enabled ExtINT on CPU#0
> 
> (it booted up fine after this)
> 
> Candidates would be:
> 
> 0fdf136: Merge branch 'tip/perf/recordmcount-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-2.6-trace into perf/core
> cf4db25: ftrace: Rename config option HAVE_C_MCOUNT_RECORD to HAVE_C_RECORDMCOUNT
> d9d572a: Merge branch 'perf/core' of git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing into perf/core
> c28d507: ftrace: Remove duplicate code for 64 and 32 bit in recordmcount.c
> 72441cb: ftrace/x86: Add support for C version of recordmcount
> 81d3858: ftrace: Add C version of recordmcount compile time code
> ebc8827: x86: Barf when vmalloc and kmemcheck faults happen in NMI
> 3cba11d: kconfig/x86: Add HAVE_TEXT_POKE_SMP config for stop_machine dependency
> 3caa375: x86: Use __stop_machine() in text_poke_smp()
> 087a4eb: stopmachine: Define __stop_machine when CONFIG_STOP_MACHINE=n
> fd02e6f: kprobes: Fix selftest to clear flags field for reusing probes
> 0f55a2f: kprobes: Update document about irq disabled state in kprobe handler
> 
> Config attached.
> 
> Gut feeling suggests that it's the new in-C recordmcount code causing 
> this bug.
> 
> Thanks,
> 
> 	Ingo