Re: [PATCH] mm/tracing: Add check for kmem_cache

From: Vlastimil Babka
Date: Wed Nov 23 2022 - 08:33:25 EST


On 11/23/22 12:10, Jiasheng Jiang wrote:
> In __kmem_cache_alloc_lru() of mm/slab.c, cachep may be NULL when error
> occurs in slab_alloc().

Are you sure? AFAICS internally in slab_alloc_node(), slab_pre_alloc_hook()
can return NULL which is assigned to cachep local variable, but only NULL
propagated back to __kmem_cache_alloc_lru() is the object, not cachep.
So trace_kmem_cache_alloc() should only see NULL s, if it was already NULL
when passed to kmem_cache_alloc(), at which point the NULL is dereferenced
and crashing (as expected) earlier than reaching the tracepoint.

> However, trace_kmem_cache_alloc() will still use it as the parameter,
> which may cause NULL poineter dereference.
> Therefore, it should be better to add the check and handle the error
> in the trace_kmem_cache_alloc().
>
> Fixes: 36555751c675 ("kmemtrace: SLAB hooks.")
> Signed-off-by: Jiasheng Jiang <jiasheng@xxxxxxxxxxx>
> ---
> include/trace/events/kmem.h | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/include/trace/events/kmem.h b/include/trace/events/kmem.h
> index 243073cfc29d..d99507d32ef5 100644
> --- a/include/trace/events/kmem.h
> +++ b/include/trace/events/kmem.h
> @@ -32,8 +32,8 @@ TRACE_EVENT(kmem_cache_alloc,
> TP_fast_assign(
> __entry->call_site = call_site;
> __entry->ptr = ptr;
> - __entry->bytes_req = s->object_size;
> - __entry->bytes_alloc = s->size;
> + __entry->bytes_req = s ? s->object_size : 0;
> + __entry->bytes_alloc = s ? s->size : 0;
> __entry->gfp_flags = (__force unsigned long)gfp_flags;
> __entry->node = node;
> __entry->accounted = IS_ENABLED(CONFIG_MEMCG_KMEM) ?