| Message ID | 20191105000617.26923-1-hch@lst.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | block: skip the split micro-optimization for devices with chunk size | expand |
On 11/4/19 5:06 PM, Christoph Hellwig wrote: > If the devices sets a chunk size we might have to split I/O that is > smaller than a page size if it crosses the chunk boundary. Skip the > micro-optimization for small I/Os in that case. I'm just going to fold this in with the previous, which is still head of tree.
On 11/4/19 4:06 PM, Christoph Hellwig wrote: > If the devices sets a chunk size we might have to split I/O that is > smaller than a page size if it crosses the chunk boundary. Skip the > micro-optimization for small I/Os in that case. > > Fixes: b072e20f0084 ("block: merge invalidate_partitions into rescan_partitions") > Signed-off-by: Christoph Hellwig <hch@lst.de> > --- > block/blk-merge.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/block/blk-merge.c b/block/blk-merge.c > index 06eb38357b41..f22cb6251d06 100644 > --- a/block/blk-merge.c > +++ b/block/blk-merge.c > @@ -317,7 +317,8 @@ void __blk_queue_split(struct request_queue *q, struct bio **bio, > * are cloned, but compared to the performance impact of cloned > * bios themselves the loop below doesn't matter anyway. ^^^^ Did you perhaps mean "test" instead of "loop"? Thanks, Bart. > */ > - if ((*bio)->bi_vcnt == 1 && > + if (!q->limits.chunk_sectors && > + (*bio)->bi_vcnt == 1 && > (*bio)->bi_io_vec[0].bv_len <= PAGE_SIZE) { > *nr_segs = 1; > break; >
On Mon, Nov 04, 2019 at 04:14:10PM -0800, Bart Van Assche wrote: > On 11/4/19 4:06 PM, Christoph Hellwig wrote: >> If the devices sets a chunk size we might have to split I/O that is >> smaller than a page size if it crosses the chunk boundary. Skip the >> micro-optimization for small I/Os in that case. >> >> Fixes: b072e20f0084 ("block: merge invalidate_partitions into rescan_partitions") >> Signed-off-by: Christoph Hellwig <hch@lst.de> >> --- >> block/blk-merge.c | 3 ++- >> 1 file changed, 2 insertions(+), 1 deletion(-) >> >> diff --git a/block/blk-merge.c b/block/blk-merge.c >> index 06eb38357b41..f22cb6251d06 100644 >> --- a/block/blk-merge.c >> +++ b/block/blk-merge.c >> @@ -317,7 +317,8 @@ void __blk_queue_split(struct request_queue *q, struct bio **bio, >> * are cloned, but compared to the performance impact of cloned >> * bios themselves the loop below doesn't matter anyway. > ^^^^ > Did you perhaps mean "test" instead of "loop"? No. I did mean the loop in blk_bio_segment_split.
diff --git a/block/blk-merge.c b/block/blk-merge.c index 06eb38357b41..f22cb6251d06 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -317,7 +317,8 @@ void __blk_queue_split(struct request_queue *q, struct bio **bio, * are cloned, but compared to the performance impact of cloned * bios themselves the loop below doesn't matter anyway. */ - if ((*bio)->bi_vcnt == 1 && + if (!q->limits.chunk_sectors && + (*bio)->bi_vcnt == 1 && (*bio)->bi_io_vec[0].bv_len <= PAGE_SIZE) { *nr_segs = 1; break;
If the devices sets a chunk size we might have to split I/O that is smaller than a page size if it crosses the chunk boundary. Skip the micro-optimization for small I/Os in that case. Fixes: b072e20f0084 ("block: merge invalidate_partitions into rescan_partitions") Signed-off-by: Christoph Hellwig <hch@lst.de> --- block/blk-merge.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)