lkml.org 
[lkml]   [2014]   [Oct]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] mm/kmemleak: Do not skip stack frames
On Tue, Oct 21, 2014 at 01:19:29PM +0100, Thierry Reding wrote:
> From: Thierry Reding <treding@nvidia.com>
>
> Trying to chase down memory leaks is much easier when the complete stack
> trace is available.
>
> Signed-off-by: Thierry Reding <treding@nvidia.com>
> ---
> It seems like this was initially set to 1 when merged in commit
> 3c7b4e6b8be4 (kmemleak: Add the base support) and later increased to 2
> in commit fd6789675ebf (kmemleak: Save the stack trace for early
> allocations). Perhaps there was a reason to skip the first few frames,
> but I've certainly found it difficult to find leaks when the stack trace
> doesn't point at the proper location.
> ---
> mm/kmemleak.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/mm/kmemleak.c b/mm/kmemleak.c
> index 3cda50c1e394..55d9ad0f40d4 100644
> --- a/mm/kmemleak.c
> +++ b/mm/kmemleak.c
> @@ -503,7 +503,7 @@ static int __save_stack_trace(unsigned long *trace)
> stack_trace.max_entries = MAX_TRACE;
> stack_trace.nr_entries = 0;
> stack_trace.entries = trace;
> - stack_trace.skip = 2;
> + stack_trace.skip = 0;

The reason for this was to avoid listing some of the kmemleak internals
(kmemleak_alloc -> create_object -> __save_stack_trace). I can see how
inlining of __save_stack_trace() would cause some of the last frames to
be missed. I would still prefer to keep it at 1 rather than 0?

Which architecture are you testing on? What's the additional trace you
get with this patch?

--
Catalin


\
 
 \ /
  Last update: 2014-10-21 18:41    [W:0.044 / U:0.092 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site