cleanup the bio handling in iomap v2

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
> 

[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 7898c1c47370e6..d31e0d3b50c683 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

[Darrick J. Wong]

On Thu, Jul 22, 2021 at 07:42:56AM +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.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>

Pretty straightforward.

Reviewed-by: Darrick J. Wong <djwong@kernel.org>

--D

> ---
>  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 7898c1c47370e6..d31e0d3b50c683 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
> 

[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

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>