On Mon, Sep 07, 2015 at 05:11:29PM +0300, Andrey Ryabinin wrote: > > > On 09/07/2015 04:56 PM, Vladimir Davydov wrote: > > On Fri, Sep 04, 2015 at 05:06:14PM +0300, Andrey Ryabinin wrote: > > ... > >> @@ -457,6 +457,8 @@ static void get_map(struct kmem_cache *s, struct page > >> *page, unsigned long *map) > >> */ > >> #ifdef CONFIG_SLUB_DEBUG_ON > >> static int slub_debug = DEBUG_DEFAULT_FLAGS; > >> +#elif defined (CONFIG_KASAN) > >> +static int slub_debug = SLAB_STORE_USER; > > > > Why not just enable SLUB_DEBUG_ON for debug kernels? > > We could, but I think that this patch make sense anyway. > > When user enables CONFIG_KASAN=y, he also have to change boot cmdline, > So the problem here is that just replacing kernel is not enough, you also > need to change boot cmdline. > And once user is going to switch to CONFIG_KASAN=n back, he will need to > change boot cmdline again. > > Of course we also could 'select SLUB_DEBUG_ON' under CONFIG_KASAN, > but this will enable useless (under CONFIG_KASAN=y) red zoning and poisoning.
OK, I see. Since turning off LOCKDEP on detecting kasan error doesn't really matter, Reviewed-by: Vladimir Davydov <vdavy...@parallels.com> for the whole set. _______________________________________________ Devel mailing list Devel@openvz.org https://lists.openvz.org/mailman/listinfo/devel