| Message ID | 20190416153847.8173-1-ming.lei@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | block: make sure that bvec length can't be overflowed | expand |
s/overflowed/overflow/ in the subject.
Otherwise this looks good to me:
Reviewed-by: Christoph Hellwig <hch@lst.de>
Although this will create yet another conflict between Linus' tree
and the 5.2 block tree.
Although maybe Jens still reset the tree and move the merge past this..
On 4/16/19 10:46 AM, Christoph Hellwig wrote: > s/overflowed/overflow/ in the subject. > > Otherwise this looks good to me: > > Reviewed-by: Christoph Hellwig <hch@lst.de> > > Although this will create yet another conflict between Linus' tree > and the 5.2 block tree. > > Although maybe Jens still reset the tree and move the merge past this.. That was my main worry too... But I'll probably just deal with it and redo the merge, I don't want to do two, and I don't want to have to deal with answering for why we get repeated conflicts in this area. For the patch, I do generally prefer doing an AND with PAGE_MASK rather than a modulo with PAGE_SIZE. I know the compiler will take care of it, but still, it's more prudent imho.
On Tue, Apr 16, 2019 at 11:03:00AM -0600, Jens Axboe wrote: > On 4/16/19 10:46 AM, Christoph Hellwig wrote: > > s/overflowed/overflow/ in the subject. > > > > Otherwise this looks good to me: > > > > Reviewed-by: Christoph Hellwig <hch@lst.de> > > > > Although this will create yet another conflict between Linus' tree > > and the 5.2 block tree. > > > > Although maybe Jens still reset the tree and move the merge past this.. > > That was my main worry too... But I'll probably just deal with it and > redo the merge, I don't want to do two, and I don't want to have to > deal with answering for why we get repeated conflicts in this area. This one is a fix which should affect on loop and nvme-loop only in theory. That is typical conflict between fix and feature(improvement). Anyway, sorry for the a bit late fix. > > For the patch, I do generally prefer doing an AND with PAGE_MASK > rather than a modulo with PAGE_SIZE. I know the compiler will take > care of it, but still, it's more prudent imho. OK, do it in V2. Thanks, Ming
On 4/16/19 5:38 PM, Ming Lei wrote: > bvec->bv_offset may be bigger than PAGE_SIZE sometimes, such as, > when one bio is splitted in the middle of one bvec via bio_split(), > and bi_iter.bi_bvec_done is used to build offset of the 1st bvec of > remained bio. > > So we have to make sure that every bvec's offset is less than > PAGE_SIZE from bio_for_each_segment(). > > This patch fixes this issue reported by Zhang Yi When running nvme/011. > > Cc: Christoph Hellwig <hch@lst.de> > Cc: Yi Zhang <yi.zhang@redhat.com> > Reported-by: Yi Zhang <yi.zhang@redhat.com> > Fixes: 6dc4f100c175 ("block: allow bio_for_each_segment_all() to iterate over multi-page bvec") > Signed-off-by: Ming Lei <ming.lei@redhat.com> > --- > include/linux/bvec.h | 5 +++-- > 1 file changed, 3 insertions(+), 2 deletions(-) > > diff --git a/include/linux/bvec.h b/include/linux/bvec.h > index 3bc91879e1e2..f179b370066f 100644 > --- a/include/linux/bvec.h > +++ b/include/linux/bvec.h > @@ -160,8 +160,9 @@ static inline void bvec_advance(const struct bio_vec *bvec, > bv->bv_page = nth_page(bv->bv_page, 1); > bv->bv_offset = 0; > } else { > - bv->bv_page = bvec->bv_page; > - bv->bv_offset = bvec->bv_offset; > + bv->bv_page = bvec_nth_page(bvec->bv_page, bvec->bv_offset / > + PAGE_SIZE); > + bv->bv_offset = bvec->bv_offset % PAGE_SIZE; > } > bv->bv_len = min_t(unsigned int, PAGE_SIZE - bv->bv_offset, > bvec->bv_len - iter_all->done); > Looks okay. Reviewed-by: Hannes Reinecke <hare@suse.com> Cheers, Hannes
diff --git a/include/linux/bvec.h b/include/linux/bvec.h index 3bc91879e1e2..f179b370066f 100644 --- a/include/linux/bvec.h +++ b/include/linux/bvec.h @@ -160,8 +160,9 @@ static inline void bvec_advance(const struct bio_vec *bvec, bv->bv_page = nth_page(bv->bv_page, 1); bv->bv_offset = 0; } else { - bv->bv_page = bvec->bv_page; - bv->bv_offset = bvec->bv_offset; + bv->bv_page = bvec_nth_page(bvec->bv_page, bvec->bv_offset / + PAGE_SIZE); + bv->bv_offset = bvec->bv_offset % PAGE_SIZE; } bv->bv_len = min_t(unsigned int, PAGE_SIZE - bv->bv_offset, bvec->bv_len - iter_all->done);
bvec->bv_offset may be bigger than PAGE_SIZE sometimes, such as, when one bio is splitted in the middle of one bvec via bio_split(), and bi_iter.bi_bvec_done is used to build offset of the 1st bvec of remained bio. So we have to make sure that every bvec's offset is less than PAGE_SIZE from bio_for_each_segment(). This patch fixes this issue reported by Zhang Yi When running nvme/011. Cc: Christoph Hellwig <hch@lst.de> Cc: Yi Zhang <yi.zhang@redhat.com> Reported-by: Yi Zhang <yi.zhang@redhat.com> Fixes: 6dc4f100c175 ("block: allow bio_for_each_segment_all() to iterate over multi-page bvec") Signed-off-by: Ming Lei <ming.lei@redhat.com> --- include/linux/bvec.h | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)