[PATCH 1/2] iomap: simplify iomap_readpage_actor

[PATCH 1/2] iomap: simplify iomap_readpage_actor

[Christoph Hellwig]

Now that the outstanding reads are counted in bytes, there is no need
to use the low-level __bio_try_merge_page API, we can switch back to
always using bio_add_page and simply iomap_readpage_actor again.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/iomap/buffered-io.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)

diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
index 87ccb3438becd9..4eaadbd265fcfa 100644
--- a/fs/iomap/buffered-io.c
+++ b/fs/iomap/buffered-io.c
@@ -241,7 +241,6 @@ iomap_readpage_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
 	struct iomap_readpage_ctx *ctx = data;
 	struct page *page = ctx->cur_page;
 	struct iomap_page *iop;
-	bool same_page = false, is_contig = false;
 	loff_t orig_pos = pos;
 	unsigned poff, plen;
 	sector_t sector;
@@ -268,16 +267,10 @@ iomap_readpage_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
 	if (iop)
 		atomic_add(plen, &iop->read_bytes_pending);
 
-	/* Try to merge into a previous segment if we can */
 	sector = iomap_sector(iomap, pos);
-	if (ctx->bio && bio_end_sector(ctx->bio) == sector) {
-		if (__bio_try_merge_page(ctx->bio, page, plen, poff,
-				&same_page))
-			goto done;
-		is_contig = true;
-	}
-
-	if (!is_contig || bio_full(ctx->bio, plen)) {
+	if (!ctx->bio ||
+	    bio_full(ctx->bio, plen) ||
+	    bio_end_sector(ctx->bio) != sector) {
 		gfp_t gfp = mapping_gfp_constraint(page->mapping, GFP_KERNEL);
 		gfp_t orig_gfp = gfp;
 		unsigned int nr_vecs = DIV_ROUND_UP(length, PAGE_SIZE);
-- 
2.30.2

[PATCH 2/2] iomap: simplify iomap_add_to_ioend

[Christoph Hellwig]

Now that the outstanding writes are counted in bytes, there is no need
to use the low-level __bio_try_merge_page API, we can switch back to
always using bio_add_page and simply iomap_add_to_ioend again.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/iomap/buffered-io.c | 17 +++++------------
 1 file changed, 5 insertions(+), 12 deletions(-)

diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
index 4eaadbd265fcfa..b5f355f3f8ee51 100644
--- a/fs/iomap/buffered-io.c
+++ b/fs/iomap/buffered-io.c
@@ -1252,7 +1252,6 @@ iomap_add_to_ioend(struct inode *inode, loff_t offset, struct page *page,
 	sector_t sector = iomap_sector(&wpc->iomap, offset);
 	unsigned len = i_blocksize(inode);
 	unsigned poff = offset & (PAGE_SIZE - 1);
-	bool merged, same_page = false;
 
 	if (!wpc->ioend || !iomap_can_add_to_ioend(wpc, offset, sector)) {
 		if (wpc->ioend)
@@ -1260,19 +1259,13 @@ iomap_add_to_ioend(struct inode *inode, loff_t offset, struct page *page,
 		wpc->ioend = iomap_alloc_ioend(inode, wpc, offset, sector, wbc);
 	}
 
-	merged = __bio_try_merge_page(wpc->ioend->io_bio, page, len, poff,
-			&same_page);
-	if (iop)
-		atomic_add(len, &iop->write_bytes_pending);
-
-	if (!merged) {
-		if (bio_full(wpc->ioend->io_bio, len)) {
-			wpc->ioend->io_bio =
-				iomap_chain_bio(wpc->ioend->io_bio);
-		}
-		bio_add_page(wpc->ioend->io_bio, page, len, poff);
+	if (bio_add_page(wpc->ioend->io_bio, page, len, poff) != len) {
+		wpc->ioend->io_bio = iomap_chain_bio(wpc->ioend->io_bio);
+		__bio_add_page(wpc->ioend->io_bio, page, len, poff);
 	}
 
+	if (iop)
+		atomic_add(len, &iop->write_bytes_pending);
 	wpc->ioend->io_size += len;
 	wbc_account_cgroup_owner(wbc, page, len);
 }
-- 
2.30.2

Re: [PATCH 2/2] iomap: simplify iomap_add_to_ioend

[Matthew Wilcox]

On Tue, Jul 20, 2021 at 10:43:20AM +0200, Christoph Hellwig wrote:
> Now that the outstanding writes are counted in bytes, there is no need
> to use the low-level __bio_try_merge_page API, we can switch back to
> always using bio_add_page and simply iomap_add_to_ioend again.

These two callers were the only external users of __bio_try_merge_page(),
so it can now be made static to block/bio.c.

From f3599f22b9e107c7a6cb0ace813ca505843316ce Mon Sep 17 00:00:00 2001
From: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Date: Tue, 20 Jul 2021 08:04:29 -0400
Subject: [PATCH] block: Make __bio_try_merge_page static

iomap was the only external user of __bio_try_merge_page().  Make it
static (and move it above a caller so it doesn't need to be declared
in advance).

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
---
 block/bio.c         | 77 ++++++++++++++++++++++-----------------------
 include/linux/bio.h |  2 --
 2 files changed, 38 insertions(+), 41 deletions(-)

diff --git a/block/bio.c b/block/bio.c
index 1fab762e079b..d36a4d3d3cd5 100644
--- a/block/bio.c
+++ b/block/bio.c
@@ -709,6 +709,44 @@ static inline bool page_is_mergeable(const struct bio_vec *bv,
 	return (bv->bv_page + bv_end / PAGE_SIZE) == (page + off / PAGE_SIZE);
 }
 
+/**
+ * __bio_try_merge_page - try appending data to an existing bvec.
+ * @bio: destination bio
+ * @page: start page to add
+ * @len: length of the data to add
+ * @off: offset of the data relative to @page
+ * @same_page: return if the segment has been merged inside the same page
+ *
+ * Try to add the data at @page + @off to the last bvec of @bio.  This is a
+ * useful optimisation for file systems with a block size smaller than the
+ * page size.
+ *
+ * Warn if (@len, @off) crosses pages in case that @same_page is true.
+ *
+ * Return %true on success or %false on failure.
+ */
+static bool __bio_try_merge_page(struct bio *bio, struct page *page,
+		unsigned int len, unsigned int off, bool *same_page)
+{
+	if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED)))
+		return false;
+
+	if (bio->bi_vcnt > 0) {
+		struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt - 1];
+
+		if (page_is_mergeable(bv, page, len, off, same_page)) {
+			if (bio->bi_iter.bi_size > UINT_MAX - len) {
+				*same_page = false;
+				return false;
+			}
+			bv->bv_len += len;
+			bio->bi_iter.bi_size += len;
+			return true;
+		}
+	}
+	return false;
+}
+
 /*
  * Try to merge a page into a segment, while obeying the hardware segment
  * size limit.  This is not for normal read/write bios, but for passthrough
@@ -840,45 +878,6 @@ int bio_add_zone_append_page(struct bio *bio, struct page *page,
 }
 EXPORT_SYMBOL_GPL(bio_add_zone_append_page);
 
-/**
- * __bio_try_merge_page - try appending data to an existing bvec.
- * @bio: destination bio
- * @page: start page to add
- * @len: length of the data to add
- * @off: offset of the data relative to @page
- * @same_page: return if the segment has been merged inside the same page
- *
- * Try to add the data at @page + @off to the last bvec of @bio.  This is a
- * useful optimisation for file systems with a block size smaller than the
- * page size.
- *
- * Warn if (@len, @off) crosses pages in case that @same_page is true.
- *
- * Return %true on success or %false on failure.
- */
-bool __bio_try_merge_page(struct bio *bio, struct page *page,
-		unsigned int len, unsigned int off, bool *same_page)
-{
-	if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED)))
-		return false;
-
-	if (bio->bi_vcnt > 0) {
-		struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt - 1];
-
-		if (page_is_mergeable(bv, page, len, off, same_page)) {
-			if (bio->bi_iter.bi_size > UINT_MAX - len) {
-				*same_page = false;
-				return false;
-			}
-			bv->bv_len += len;
-			bio->bi_iter.bi_size += len;
-			return true;
-		}
-	}
-	return false;
-}
-EXPORT_SYMBOL_GPL(__bio_try_merge_page);
-
 /**
  * __bio_add_page - add page(s) to a bio in a new segment
  * @bio: destination bio
diff --git a/include/linux/bio.h b/include/linux/bio.h
index 2203b686e1f0..c779e0e3140a 100644
--- a/include/linux/bio.h
+++ b/include/linux/bio.h
@@ -467,8 +467,6 @@ extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
 			   unsigned int, unsigned int);
 int bio_add_zone_append_page(struct bio *bio, struct page *page,
 			     unsigned int len, unsigned int offset);
-bool __bio_try_merge_page(struct bio *bio, struct page *page,
-		unsigned int len, unsigned int off, bool *same_page);
 void __bio_add_page(struct bio *bio, struct page *page,
 		unsigned int len, unsigned int off);
 int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter);
-- 
2.30.2

Re: [PATCH 2/2] iomap: simplify iomap_add_to_ioend

[Christoph Hellwig]

On Tue, Jul 20, 2021 at 01:10:04PM +0100, Matthew Wilcox wrote:
> On Tue, Jul 20, 2021 at 10:43:20AM +0200, Christoph Hellwig wrote:
> > Now that the outstanding writes are counted in bytes, there is no need
> > to use the low-level __bio_try_merge_page API, we can switch back to
> > always using bio_add_page and simply iomap_add_to_ioend again.
> 
> These two callers were the only external users of __bio_try_merge_page(),
> so it can now be made static to block/bio.c.

Yeah, but doing so in the same merge window will create an annoying
cross dependency.  I have a series to cleanup a whole lot of the
bio_add_page related code including starting to clean up the return
value for the next merge window.

Re: [PATCH 1/2] iomap: simplify iomap_readpage_actor

[Matthew Wilcox]

On Tue, Jul 20, 2021 at 10:43:19AM +0200, Christoph Hellwig wrote:
> Now that the outstanding reads are counted in bytes, there is no need
> to use the low-level __bio_try_merge_page API, we can switch back to
> always using bio_add_page and simply iomap_readpage_actor again.

I don't think this quite works.  You need to check the return value
from bio_add_page(), otherwise you can be in a situation where you try
to add a page to the last bvec and it's not contiguous, so it fails.

I was imagining something more like this:

-       bool same_page = false, is_contig = false;
+       bool is_contig = false;
...
        /* Try to merge into a previous segment if we can */
        sector = iomap_sector(iomap, pos);
-       if (ctx->bio && bio_end_sector(ctx->bio) == sector) {
-               if (__bio_try_merge_page(ctx->bio, page, plen, poff,
-                               &same_page))
-                       goto done;
-               is_contig = true;
-       }
+       if (ctx->bio && bio_end_sector(ctx->bio) == sector)
+               is_contig = bio_add_page(ctx->bio, page, plen, poff) > 0;

-       if (!is_contig || bio_full(ctx->bio, plen)) {
+       if (!is_contig) {
                gfp_t gfp = mapping_gfp_constraint(page->mapping, GFP_KERNEL);
                gfp_t orig_gfp = gfp;
...
                bio_set_dev(ctx->bio, iomap->bdev);
                ctx->bio->bi_end_io = iomap_read_end_io;
+               bio_add_page(ctx->bio, page, plen, poff);
        }

-       bio_add_page(ctx->bio, page, plen, poff);
 done:
        /*

Re: [PATCH 1/2] iomap: simplify iomap_readpage_actor

[Christoph Hellwig]

On Tue, Jul 20, 2021 at 01:27:24PM +0100, Matthew Wilcox wrote:
> On Tue, Jul 20, 2021 at 10:43:19AM +0200, Christoph Hellwig wrote:
> > Now that the outstanding reads are counted in bytes, there is no need
> > to use the low-level __bio_try_merge_page API, we can switch back to
> > always using bio_add_page and simply iomap_readpage_actor again.
> 
> I don't think this quite works.  You need to check the return value
> from bio_add_page(), otherwise you can be in a situation where you try
> to add a page to the last bvec and it's not contiguous, so it fails.

Indeed.  While bio_full covers the number of vectors, if we run out of
bytes in bi_size this won't work.

I think we can do this version, but I haven't tested it yet:

---
From 4198cd5805d45f83c9029743ad5d0ce774a4e0f8 Mon Sep 17 00:00:00 2001
From: Christoph Hellwig <hch@lst.de>
Date: Mon, 12 Jul 2021 11:00:58 +0200
Subject: iomap: simplify iomap_readpage_actor

Now that the outstanding reads are counted in bytes, there is no need
to use the low-level __bio_try_merge_page API, we can switch back to
always using bio_add_page and simply iomap_readpage_actor again.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/iomap/buffered-io.c | 16 ++++------------
 1 file changed, 4 insertions(+), 12 deletions(-)

diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
index 87ccb3438becd9..712b6513a0c449 100644
--- a/fs/iomap/buffered-io.c
+++ b/fs/iomap/buffered-io.c
@@ -241,7 +241,6 @@ iomap_readpage_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
 	struct iomap_readpage_ctx *ctx = data;
 	struct page *page = ctx->cur_page;
 	struct iomap_page *iop;
-	bool same_page = false, is_contig = false;
 	loff_t orig_pos = pos;
 	unsigned poff, plen;
 	sector_t sector;
@@ -268,16 +267,10 @@ iomap_readpage_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
 	if (iop)
 		atomic_add(plen, &iop->read_bytes_pending);
 
-	/* Try to merge into a previous segment if we can */
 	sector = iomap_sector(iomap, pos);
-	if (ctx->bio && bio_end_sector(ctx->bio) == sector) {
-		if (__bio_try_merge_page(ctx->bio, page, plen, poff,
-				&same_page))
-			goto done;
-		is_contig = true;
-	}
-
-	if (!is_contig || bio_full(ctx->bio, plen)) {
+	if (!ctx->bio ||
+	    bio_end_sector(ctx->bio) != sector ||
+	    bio_add_page(ctx->bio, page, plen, poff) != plen) {
 		gfp_t gfp = mapping_gfp_constraint(page->mapping, GFP_KERNEL);
 		gfp_t orig_gfp = gfp;
 		unsigned int nr_vecs = DIV_ROUND_UP(length, PAGE_SIZE);
@@ -301,9 +294,8 @@ iomap_readpage_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
 		ctx->bio->bi_iter.bi_sector = sector;
 		bio_set_dev(ctx->bio, iomap->bdev);
 		ctx->bio->bi_end_io = iomap_read_end_io;
+		__bio_add_page(ctx->bio, page, plen, poff);
 	}
-
-	bio_add_page(ctx->bio, page, plen, poff);
 done:
 	/*
 	 * Move the caller beyond our range so that it keeps making progress.
-- 
2.30.2

Re: [PATCH 1/2] iomap: simplify iomap_readpage_actor

[Matthew Wilcox]

On Tue, Jul 20, 2021 at 03:26:44PM +0200, Christoph Hellwig wrote:
> I think we can do this version, but I haven't tested it yet:

I see no problems with this version.

Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>

cleanup the bio handling in iomap v2

[Christoph Hellwig]

Hi all,

this series cleans up the bio handling in the iomap buffered I/O code
by dropping use of the very lowlevel helpers.  We did only need them
before Matthew converted the inflight counters to be byte based

Changes sine v1:
 - fix readpage to properly handle bios that hit their byte size limit

[PATCH 1/2] iomap: simplify iomap_readpage_actor

[Christoph Hellwig]

Now that the outstanding reads are counted in bytes, there is no need
to use the low-level __bio_try_merge_page API, we can switch back to
always using bio_add_page and simply iomap_readpage_actor again.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
---
 fs/iomap/buffered-io.c | 16 ++++------------
 1 file changed, 4 insertions(+), 12 deletions(-)

diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
index 0597f5c186a33f..7898c1c47370e6 100644
--- a/fs/iomap/buffered-io.c
+++ b/fs/iomap/buffered-io.c
@@ -245,7 +245,6 @@ iomap_readpage_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
 	struct iomap_readpage_ctx *ctx = data;
 	struct page *page = ctx->cur_page;
 	struct iomap_page *iop;
-	bool same_page = false, is_contig = false;
 	loff_t orig_pos = pos;
 	unsigned poff, plen;
 	sector_t sector;
@@ -269,16 +268,10 @@ iomap_readpage_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
 	if (iop)
 		atomic_add(plen, &iop->read_bytes_pending);
 
-	/* Try to merge into a previous segment if we can */
 	sector = iomap_sector(iomap, pos);
-	if (ctx->bio && bio_end_sector(ctx->bio) == sector) {
-		if (__bio_try_merge_page(ctx->bio, page, plen, poff,
-				&same_page))
-			goto done;
-		is_contig = true;
-	}
-
-	if (!is_contig || bio_full(ctx->bio, plen)) {
+	if (!ctx->bio ||
+	    bio_end_sector(ctx->bio) != sector ||
+	    bio_add_page(ctx->bio, page, plen, poff) != plen) {
 		gfp_t gfp = mapping_gfp_constraint(page->mapping, GFP_KERNEL);
 		gfp_t orig_gfp = gfp;
 		unsigned int nr_vecs = DIV_ROUND_UP(length, PAGE_SIZE);
@@ -302,9 +295,8 @@ iomap_readpage_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
 		ctx->bio->bi_iter.bi_sector = sector;
 		bio_set_dev(ctx->bio, iomap->bdev);
 		ctx->bio->bi_end_io = iomap_read_end_io;
+		__bio_add_page(ctx->bio, page, plen, poff);
 	}
-
-	bio_add_page(ctx->bio, page, plen, poff);
 done:
 	/*
 	 * Move the caller beyond our range so that it keeps making progress.
-- 
2.30.2

Re: [PATCH 1/2] iomap: simplify iomap_readpage_actor

[Darrick J. Wong]

On Thu, Jul 22, 2021 at 07:42:55AM +0200, Christoph Hellwig wrote:
> Now that the outstanding reads are counted in bytes, there is no need
> to use the low-level __bio_try_merge_page API, we can switch back to
> always using bio_add_page and simply iomap_readpage_actor again.

s/simply/simplify/ ?

> Signed-off-by: Christoph Hellwig <hch@lst.de>
> Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>

With that corrected,
Reviewed-by: Darrick J. Wong <djwong@kernel.org>

--D

> ---
>  fs/iomap/buffered-io.c | 16 ++++------------
>  1 file changed, 4 insertions(+), 12 deletions(-)
> 
> diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
> index 0597f5c186a33f..7898c1c47370e6 100644
> --- a/fs/iomap/buffered-io.c
> +++ b/fs/iomap/buffered-io.c
> @@ -245,7 +245,6 @@ iomap_readpage_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
>  	struct iomap_readpage_ctx *ctx = data;
>  	struct page *page = ctx->cur_page;
>  	struct iomap_page *iop;
> -	bool same_page = false, is_contig = false;
>  	loff_t orig_pos = pos;
>  	unsigned poff, plen;
>  	sector_t sector;
> @@ -269,16 +268,10 @@ iomap_readpage_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
>  	if (iop)
>  		atomic_add(plen, &iop->read_bytes_pending);
>  
> -	/* Try to merge into a previous segment if we can */
>  	sector = iomap_sector(iomap, pos);
> -	if (ctx->bio && bio_end_sector(ctx->bio) == sector) {
> -		if (__bio_try_merge_page(ctx->bio, page, plen, poff,
> -				&same_page))
> -			goto done;
> -		is_contig = true;
> -	}
> -
> -	if (!is_contig || bio_full(ctx->bio, plen)) {
> +	if (!ctx->bio ||
> +	    bio_end_sector(ctx->bio) != sector ||
> +	    bio_add_page(ctx->bio, page, plen, poff) != plen) {
>  		gfp_t gfp = mapping_gfp_constraint(page->mapping, GFP_KERNEL);
>  		gfp_t orig_gfp = gfp;
>  		unsigned int nr_vecs = DIV_ROUND_UP(length, PAGE_SIZE);
> @@ -302,9 +295,8 @@ iomap_readpage_actor(struct inode *inode, loff_t pos, loff_t length, void *data,
>  		ctx->bio->bi_iter.bi_sector = sector;
>  		bio_set_dev(ctx->bio, iomap->bdev);
>  		ctx->bio->bi_end_io = iomap_read_end_io;
> +		__bio_add_page(ctx->bio, page, plen, poff);
>  	}
> -
> -	bio_add_page(ctx->bio, page, plen, poff);
>  done:
>  	/*
>  	 * Move the caller beyond our range so that it keeps making progress.
> -- 
> 2.30.2
>