Re: [PATCH BUGFIX] block: add missing group association in bio_split

[Paolo <paolo.valente@linaro.org>]

Il 10/05/2016 18:12, Jeff Moyer ha scritto:
> Paolo Valente <paolo.valente@linaro.org> writes:
>
>> When a bio is split, the newly created bio must be associated with the
>> same blkcg as the original bio (if BLK_CGROUP is enabled). If this
>> operation is not performed, then the new bio is not associated with
>> any group, and the group of the current task is returned when the
>> group of the bio is requested.
>>
>> Depending on the frequency of splits, this may cause a large
>> percentage of the bios belonging to a given group to be treated as if
>> belonging to other groups (in most cases as if belonging to the root
>> group). The expected group isolation may thereby be then broken.
>>
>> This commit adds the missing association in bio_split.
>>
>> Signed-off-by: Paolo Valente <paolo.valente@linaro.org>
>> ---
>>   block/bio.c         | 15 +++++++++++++++
>>   include/linux/bio.h |  2 ++
>>   2 files changed, 17 insertions(+)
>>
>> diff --git a/block/bio.c b/block/bio.c
>> index 807d25e..20795eb 100644
>> --- a/block/bio.c
>> +++ b/block/bio.c
>> @@ -1811,6 +1811,8 @@ struct bio *bio_split(struct bio *bio, int sectors,
>>
>>   	bio_advance(bio, split->bi_iter.bi_size);
>>
>> +	bio_clone_blkcg_association(split, bio);
>> +
>
> First, I think this should be done inside of bio_clone_fast and
> bio_clone_bioset instead of in bio_split.  Otherwise you miss other
> places where bios are cloned, and I'm guessing that's bad.  You could
> also then get rid of this code in btrfs/extent_io.c:
>
> #ifdef CONFIG_BLK_CGROUP
>                  /* FIXME, put this into bio_clone_bioset */
>                  if (bio->bi_css)
>                          bio_associate_blkcg(new, bio->bi_css);
> #endif
>
> Next, you went to the trouble of propagating the return value from
> bio_associate_blkcg, and then it goes unchecked here in the only caller
> of your new function.  Since bio_associate_blkcg should not fail when
> cloning (right?), I'd change bio_clone_blkcg_association to return void
> and to WARN_ON a failure return from bio_associate_blkcg.
>

Changing as suggested.

Thanks,
Paolo

> Cheers,
> Jeff
>
>>   	return split;
>>   }
>>   EXPORT_SYMBOL(bio_split);
>> @@ -2016,6 +2018,19 @@ void bio_disassociate_task(struct bio *bio)
>>   	}
>>   }
>>
>> +/**
>> + * bio_clone_blkcg_association - clone blkcg association from src to dst bio
>> + * @dst: destination bio
>> + * @src: source bio
>> + */
>> +int bio_clone_blkcg_association(struct bio *dst, struct bio *src)
>> +{
>> +	if (src->bi_css)
>> +		return bio_associate_blkcg(dst, src->bi_css);
>> +
>> +	return 0;
>> +}
>> +
>>   #endif /* CONFIG_BLK_CGROUP */
>>
>>   static void __init biovec_init_slabs(void)
>> diff --git a/include/linux/bio.h b/include/linux/bio.h
>> index 6b7481f..8535f81 100644
>> --- a/include/linux/bio.h
>> +++ b/include/linux/bio.h
>> @@ -527,11 +527,13 @@ extern unsigned int bvec_nr_vecs(unsigned short idx);
>>   int bio_associate_blkcg(struct bio *bio, struct cgroup_subsys_state *blkcg_css);
>>   int bio_associate_current(struct bio *bio);
>>   void bio_disassociate_task(struct bio *bio);
>> +int bio_clone_blkcg_association(struct bio *dst, struct bio *src);
>>   #else	/* CONFIG_BLK_CGROUP */
>>   static inline int bio_associate_blkcg(struct bio *bio,
>>   			struct cgroup_subsys_state *blkcg_css) { return 0; }
>>   static inline int bio_associate_current(struct bio *bio) { return -ENOENT; }
>>   static inline void bio_disassociate_task(struct bio *bio) { }
>> +int bio_clone_blkcg_association(struct bio *dst, struct bio *src) { return 0; }
>>   #endif	/* CONFIG_BLK_CGROUP */
>>
>>   #ifdef CONFIG_HIGHMEM