[PATCH] erofs: get rid of raw bi_end_io() usage

[PATCH] erofs: get rid of raw bi_end_io() usage

[Gao Xiang]

These BIOs are actually harmless in practice, as they are all pseudo
BIOs and do not use advanced features like chaining.  Using the BIO
interface is a more friendly and unified approach for both bdev and
and file-backed I/Os.

Let's use bio_endio() instead.

Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
 fs/erofs/fileio.c  | 2 +-
 fs/erofs/fscache.c | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/fs/erofs/fileio.c b/fs/erofs/fileio.c
index b7b3432a9882..a65f86dc79b6 100644
--- a/fs/erofs/fileio.c
+++ b/fs/erofs/fileio.c
@@ -34,7 +34,7 @@ static void erofs_fileio_ki_complete(struct kiocb *iocb, long ret)
 	if (rq->bio.bi_end_io) {
 		if (ret < 0 && !rq->bio.bi_status)
 			rq->bio.bi_status = errno_to_blk_status(ret);
-		rq->bio.bi_end_io(&rq->bio);
+		bio_endio(&rq->bio);
 	} else {
 		bio_for_each_folio_all(fi, &rq->bio) {
 			DBG_BUGON(folio_test_uptodate(fi.folio));
diff --git a/fs/erofs/fscache.c b/fs/erofs/fscache.c
index 362acf828279..7a346e20f7b7 100644
--- a/fs/erofs/fscache.c
+++ b/fs/erofs/fscache.c
@@ -185,7 +185,7 @@ static void erofs_fscache_bio_endio(void *priv, ssize_t transferred_or_error)
 
 	if (IS_ERR_VALUE(transferred_or_error))
 		io->bio.bi_status = errno_to_blk_status(transferred_or_error);
-	io->bio.bi_end_io(&io->bio);
+	bio_endio(&io->bio);
 	BUILD_BUG_ON(offsetof(struct erofs_fscache_bio, io) != 0);
 	erofs_fscache_io_put(&io->io);
 }
@@ -216,7 +216,7 @@ void erofs_fscache_submit_bio(struct bio *bio)
 	if (!ret)
 		return;
 	bio->bi_status = errno_to_blk_status(ret);
-	bio->bi_end_io(bio);
+	bio_endio(bio);
 }
 
 static int erofs_fscache_meta_read_folio(struct file *data, struct folio *folio)
-- 
2.43.5

[PATCH v2] erofs: get rid of raw bi_end_io() usage

[Gao Xiang]

These BIOs are actually harmless in practice, as they are all pseudo
BIOs and do not use advanced features like chaining.  Using the BIO
interface is a more friendly and unified approach for both bdev and
and file-backed I/Os.

Let's use bio_endio() instead.

Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
v2:
 - call bio_endio() unconditionally in erofs_fileio_ki_complete().

 fs/erofs/fileio.c  | 2 +-
 fs/erofs/fscache.c | 4 ++--
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/fs/erofs/fileio.c b/fs/erofs/fileio.c
index b7b3432a9882..301647b84ca9 100644
--- a/fs/erofs/fileio.c
+++ b/fs/erofs/fileio.c
@@ -34,13 +34,13 @@ static void erofs_fileio_ki_complete(struct kiocb *iocb, long ret)
 	if (rq->bio.bi_end_io) {
 		if (ret < 0 && !rq->bio.bi_status)
 			rq->bio.bi_status = errno_to_blk_status(ret);
-		rq->bio.bi_end_io(&rq->bio);
 	} else {
 		bio_for_each_folio_all(fi, &rq->bio) {
 			DBG_BUGON(folio_test_uptodate(fi.folio));
 			erofs_onlinefolio_end(fi.folio, ret, false);
 		}
 	}
+	bio_endio(&rq->bio);
 	bio_uninit(&rq->bio);
 	kfree(rq);
 }
diff --git a/fs/erofs/fscache.c b/fs/erofs/fscache.c
index 362acf828279..7a346e20f7b7 100644
--- a/fs/erofs/fscache.c
+++ b/fs/erofs/fscache.c
@@ -185,7 +185,7 @@ static void erofs_fscache_bio_endio(void *priv, ssize_t transferred_or_error)
 
 	if (IS_ERR_VALUE(transferred_or_error))
 		io->bio.bi_status = errno_to_blk_status(transferred_or_error);
-	io->bio.bi_end_io(&io->bio);
+	bio_endio(&io->bio);
 	BUILD_BUG_ON(offsetof(struct erofs_fscache_bio, io) != 0);
 	erofs_fscache_io_put(&io->io);
 }
@@ -216,7 +216,7 @@ void erofs_fscache_submit_bio(struct bio *bio)
 	if (!ret)
 		return;
 	bio->bi_status = errno_to_blk_status(ret);
-	bio->bi_end_io(bio);
+	bio_endio(bio);
 }
 
 static int erofs_fscache_meta_read_folio(struct file *data, struct folio *folio)
-- 
2.43.5

Re: [PATCH v2] erofs: get rid of raw bi_end_io() usage

[Christoph Hellwig]

On Thu, Nov 27, 2025 at 04:07:56PM +0800, Gao Xiang wrote:
> These BIOs are actually harmless in practice, as they are all pseudo
> BIOs and do not use advanced features like chaining.  Using the BIO
> interface is a more friendly and unified approach for both bdev and
> and file-backed I/Os.
> 
> Let's use bio_endio() instead.

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH v2] erofs: get rid of raw bi_end_io() usage

[Ming Lei]

On Thu, Nov 27, 2025 at 04:07:56PM +0800, Gao Xiang wrote:
> These BIOs are actually harmless in practice, as they are all pseudo
> BIOs and do not use advanced features like chaining.  Using the BIO
> interface is a more friendly and unified approach for both bdev and
> and file-backed I/Os.
> 
> Let's use bio_endio() instead.
> 
> Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
> ---
> v2:
>  - call bio_endio() unconditionally in erofs_fileio_ki_complete().

bio_endio() can cover bio not submitted yet, so:

Reviewed-by: Ming Lei <ming.lei@redhat.com>


Thanks,
Ming

Re: [PATCH v2] erofs: get rid of raw bi_end_io() usage

[Gao Xiang]

On 2025/11/28 08:58, Ming Lei wrote:
> On Thu, Nov 27, 2025 at 04:07:56PM +0800, Gao Xiang wrote:
>> These BIOs are actually harmless in practice, as they are all pseudo
>> BIOs and do not use advanced features like chaining.  Using the BIO
>> interface is a more friendly and unified approach for both bdev and
>> and file-backed I/Os.
>>
>> Let's use bio_endio() instead.
>>
>> Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
>> ---
>> v2:
>>   - call bio_endio() unconditionally in erofs_fileio_ki_complete().
> 
> bio_endio() can cover bio not submitted yet, so:
> 
> Reviewed-by: Ming Lei <ming.lei@redhat.com>

Thank you and Christoph for review!

Thanks,
Gao Xiang

Re: [PATCH v2] erofs: get rid of raw bi_end_io() usage

[Gao Xiang]

On 2025/11/28 09:40, Gao Xiang wrote:
> 
> 
> On 2025/11/28 08:58, Ming Lei wrote:
>> On Thu, Nov 27, 2025 at 04:07:56PM +0800, Gao Xiang wrote:
>>> These BIOs are actually harmless in practice, as they are all pseudo
>>> BIOs and do not use advanced features like chaining.  Using the BIO
>>> interface is a more friendly and unified approach for both bdev and
>>> and file-backed I/Os.
>>>
>>> Let's use bio_endio() instead.
>>>
>>> Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
>>> ---
>>> v2:
>>>   - call bio_endio() unconditionally in erofs_fileio_ki_complete().
>>
>> bio_endio() can cover bio not submitted yet, so:

( btw, nevertheless, it's not submitted, just with or without
   a given bio callback. )

I believe that is ok anyway.

>>
>> Reviewed-by: Ming Lei <ming.lei@redhat.com>
> 
> Thank you and Christoph for review!
> 
> Thanks,
> Gao Xiang

Re: [PATCH v2] erofs: get rid of raw bi_end_io() usage

[Chao Yu]

On 11/27/25 16:07, Gao Xiang wrote:
> These BIOs are actually harmless in practice, as they are all pseudo
> BIOs and do not use advanced features like chaining.  Using the BIO
> interface is a more friendly and unified approach for both bdev and
> and file-backed I/Os.
> 
> Let's use bio_endio() instead.
> 
> Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>

Reviewed-by: Chao Yu <chao@kernel.org>

Thanks,