| Message ID | 20231109082827.2276696-2-ming.lei@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | block: try to make aligned bio in case of big chunk IO | expand |
> if (bio->bi_bdev) { > size_t trim = size & (bdev_logical_block_size(bio->bi_bdev) - 1); > - iov_iter_revert(iter, trim); > - size -= trim; > + > + if (trim) { > + iov_iter_revert(iter, trim); > + size -= trim; > + } We also need to fix this. This isn't supposed to be called without a bdev set, but Kent broke this. But if we don't have a bdev we need to pass the block size manually and still do the revert if needed.
diff --git a/block/bio.c b/block/bio.c index 816d412c06e9..09a5e71a0372 100644 --- a/block/bio.c +++ b/block/bio.c @@ -1262,8 +1262,11 @@ static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) if (bio->bi_bdev) { size_t trim = size & (bdev_logical_block_size(bio->bi_bdev) - 1); - iov_iter_revert(iter, trim); - size -= trim; + + if (trim) { + iov_iter_revert(iter, trim); + size -= trim; + } } if (unlikely(!size)) { @@ -1286,7 +1289,8 @@ static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) offset = 0; } - iov_iter_revert(iter, left); + if (left) + iov_iter_revert(iter, left); out: while (i < nr_pages) bio_release_page(bio, pages[i++]);
Most of times all pages are just added to bio, and nothing is left, so not necessary to call into iov_iter_revert(). Same with block size alignment handling. Signed-off-by: Ming Lei <ming.lei@redhat.com> --- block/bio.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-)