| Message ID | 20241025115818.54976-1-hch@lst.de (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | block: fix queue limits checks in blk_rq_map_user_bvec for real | expand |
On Fri, Oct 25, 2024 at 01:58:11PM +0200, Christoph Hellwig wrote: > blk_rq_map_user_bvec currently only has ad-hoc checks for queue limits, > and the last fix to it enabled valid NVMe I/O to pass, but also allowed > invalid one for drivers that set a max_segment_size or seg_boundary > limit. > > Fix it once for all by using the bio_split_rw_at helper from the I/O > path that indicates if and where a bio would be have to be split to > adhere to the queue limits, and it it returns a positive value, turn > that into -EREMOTEIO to retry using the copy path. > > Fixes: 2ff949441802 ("block: fix sanity checks in blk_rq_map_user_bvec") > Signed-off-by: Christoph Hellwig <hch@lst.de> This passes my test for NVMe passthrough I/O using a strict subset of a preregistered buffer (see 2ff949441802 for details). Tested-by: Uday Shankar <ushankar@purestorage.com>
On Fri, Oct 25, 2024 at 01:58:11PM +0200, Christoph Hellwig wrote: > blk_rq_map_user_bvec currently only has ad-hoc checks for queue limits, > and the last fix to it enabled valid NVMe I/O to pass, but also allowed > invalid one for drivers that set a max_segment_size or seg_boundary > limit. > > Fix it once for all by using the bio_split_rw_at helper from the I/O > path that indicates if and where a bio would be have to be split to > adhere to the queue limits, and it it returns a positive value, turn > that into -EREMOTEIO to retry using the copy path. Nice cleanup. Reviewed-by: Keith Busch <kbusch@kernel.org>
On 25/10/2024 12:58, Christoph Hellwig wrote: > > + /* check that the data layout matches the hardware restrictions */ > + ret = bio_split_rw_at(bio, lim, &nsegs, lim->max_hw_sectors); eh, but doesn't bio_split_rw_at() accept bytes (and not a value in sectors) for max size?
On Fri, Oct 25, 2024 at 09:43:04PM +0100, John Garry wrote: > On 25/10/2024 12:58, Christoph Hellwig wrote: >> + /* check that the data layout matches the hardware restrictions */ >> + ret = bio_split_rw_at(bio, lim, &nsegs, lim->max_hw_sectors); > > eh, but doesn't bio_split_rw_at() accept bytes (and not a value in sectors) > for max size? Yes. Thanks for the careful review, I've sent out a new version. (and this really helped me finding a bug I had been debugging in another user of this helper :))
diff --git a/block/blk-map.c b/block/blk-map.c index 6ef2ec1f7d78..d1dce1470054 100644 --- a/block/blk-map.c +++ b/block/blk-map.c @@ -561,55 +561,35 @@ EXPORT_SYMBOL(blk_rq_append_bio); /* Prepare bio for passthrough IO given ITER_BVEC iter */ static int blk_rq_map_user_bvec(struct request *rq, const struct iov_iter *iter) { - struct request_queue *q = rq->q; - size_t nr_iter = iov_iter_count(iter); - size_t nr_segs = iter->nr_segs; - struct bio_vec *bvecs, *bvprvp = NULL; - const struct queue_limits *lim = &q->limits; - unsigned int nsegs = 0, bytes = 0; + const struct queue_limits *lim = &rq->q->limits; + unsigned int nsegs; struct bio *bio; - size_t i; + int ret; - if (!nr_iter || (nr_iter >> SECTOR_SHIFT) > queue_max_hw_sectors(q)) - return -EINVAL; - if (nr_segs > queue_max_segments(q)) + if (!iov_iter_count(iter) || + (iov_iter_count(iter) >> SECTOR_SHIFT) > lim->max_hw_sectors) return -EINVAL; - /* no iovecs to alloc, as we already have a BVEC iterator */ + /* reuse the bvecs from the iterator instead of allocating new ones */ bio = blk_rq_map_bio_alloc(rq, 0, GFP_KERNEL); - if (bio == NULL) + if (!bio) return -ENOMEM; - bio_iov_bvec_set(bio, (struct iov_iter *)iter); - blk_rq_bio_prep(rq, bio, nr_segs); - - /* loop to perform a bunch of sanity checks */ - bvecs = (struct bio_vec *)iter->bvec; - for (i = 0; i < nr_segs; i++) { - struct bio_vec *bv = &bvecs[i]; + /* check that the data layout matches the hardware restrictions */ + ret = bio_split_rw_at(bio, lim, &nsegs, lim->max_hw_sectors); + if (ret) { /* - * If the queue doesn't support SG gaps and adding this - * offset would create a gap, fallback to copy. + * If we would have to split the bio, try to copy. */ - if (bvprvp && bvec_gap_to_prev(lim, bvprvp, bv->bv_offset)) { - blk_mq_map_bio_put(bio); - return -EREMOTEIO; - } - /* check full condition */ - if (nsegs >= nr_segs || bytes > UINT_MAX - bv->bv_len) - goto put_bio; - if (bytes + bv->bv_len > nr_iter) - break; - - nsegs++; - bytes += bv->bv_len; - bvprvp = bv; + if (ret > 0) + ret = -EREMOTEIO; + blk_mq_map_bio_put(bio); + return ret; } + + blk_rq_bio_prep(rq, bio, nsegs); return 0; -put_bio: - blk_mq_map_bio_put(bio); - return -EINVAL; } /**
blk_rq_map_user_bvec currently only has ad-hoc checks for queue limits, and the last fix to it enabled valid NVMe I/O to pass, but also allowed invalid one for drivers that set a max_segment_size or seg_boundary limit. Fix it once for all by using the bio_split_rw_at helper from the I/O path that indicates if and where a bio would be have to be split to adhere to the queue limits, and it it returns a positive value, turn that into -EREMOTEIO to retry using the copy path. Fixes: 2ff949441802 ("block: fix sanity checks in blk_rq_map_user_bvec") Signed-off-by: Christoph Hellwig <hch@lst.de> --- block/blk-map.c | 54 ++++++++++++++++--------------------------------- 1 file changed, 17 insertions(+), 37 deletions(-)