Re: [PATCH v10 06/17] mm: slub: introduce metadata_access_enable()/metadata_access_disable()

[Andrey Ryabinin]

2015-01-31 2:16 GMT+03:00 Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>:
> On Sat, 31 Jan 2015 03:11:55 +0400 Andrey Ryabinin <ryabinin.a.a@xxxxxxxxx> wrote:
>
>> >> > kasan_disable_local/kasan_enable_local are also undocumented doesn't
>> >> > help.
>> >> >
>> >>
>> >> Ok, How about this?
>> >>
>> >> /*
>> >>  * This hooks separate payload access from metadata access.
>> >>  * Useful for memory checkers that have to know when slub
>> >>  * accesses metadata.
>> >>  */
>> >
>> > "These hooks".
>> >
>> > I still don't understand :( Maybe I'm having a more-stupid-than-usual
>> > day.
>>
>> I think it's me being stupid today ;) I'll try to explain better.
>>
>> > How can a function "separate access"?  What does this mean?  More
>> > details, please.  I think I've only once seen a comment which had too
>> > much info!
>> >
>>
>> slub could access memory marked by kasan as inaccessible (object's metadata).
>> Kasan shouldn't print report in that case because this access is valid.
>> Disabling instrumentation of slub.c code is not enough to achieve this
>> because slub passes pointer to object's metadata into memchr_inv().
>>
>> We can't disable instrumentation for memchr_inv() because this is quite
>> generic function.
>>
>> So metadata_access_enable/metadata_access_disable wrap some
>> places in slub.c where access to object's metadata starts/end.
>> And kasan_disable_local/kasan_enable_local just disable/enable
>> error reporting in this places.
>
> ooh, I see.  Something like this?
>

Yes! Thank you, this looks much better.

> /*
>  * slub is about to manipulate internal object metadata.  This memory lies
>  * outside the range of the allocated object, so accessing it would normally
>  * be reported by kasan as a bounds error.  metadata_access_enable() is used
>  * to tell kasan that these accesses are OK.
>  */
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/