Re: [PATCH] ext4: bio_alloc never fails

Re: [PATCH] ext4: bio_alloc never fails

[Ritesh Harjani]

On 10/30/19 9:56 AM, Gao Xiang wrote:
> Similar to [1] [2], it seems a trivial cleanup since
> bio_alloc can handle memory allocation as mentioned in
> fs/direct-io.c (also see fs/block_dev.c, fs/buffer.c, ..)
> 

AFAIU, the reason is that, bio_alloc with __GFP_DIRECT_RECLAIM
flags guarantees bio allocation under some given restrictions,
as stated in fs/direct-io.c
So here it is ok to not check for NULL value from bio_alloc.

I think we can update above info too in your commit msg.

> [1] https://lore.kernel.org/r/20191030035518.65477-1-gaoxiang25@huawei.com
> [2] https://lore.kernel.org/r/20190830162812.GA10694@infradead.org
> Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
> ---
>   fs/ext4/page-io.c  | 11 +++--------
>   fs/ext4/readpage.c |  2 --
>   2 files changed, 3 insertions(+), 10 deletions(-)
> 
> diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c
> index 12ceadef32c5..f1f7b6601780 100644
> --- a/fs/ext4/page-io.c
> +++ b/fs/ext4/page-io.c
> @@ -358,14 +358,12 @@ void ext4_io_submit_init(struct ext4_io_submit *io,
>   	io->io_end = NULL;
>   }
> 
> -static int io_submit_init_bio(struct ext4_io_submit *io,
> -			      struct buffer_head *bh)
> +static void io_submit_init_bio(struct ext4_io_submit *io,
> +			       struct buffer_head *bh)
>   {
>   	struct bio *bio;
> 
>   	bio = bio_alloc(GFP_NOIO, BIO_MAX_PAGES);
> -	if (!bio)
> -		return -ENOMEM;
>   	bio->bi_iter.bi_sector = bh->b_blocknr * (bh->b_size >> 9);
>   	bio_set_dev(bio, bh->b_bdev);
>   	bio->bi_end_io = ext4_end_bio;
> @@ -373,7 +371,6 @@ static int io_submit_init_bio(struct ext4_io_submit *io,
>   	io->io_bio = bio;
>   	io->io_next_block = bh->b_blocknr;
>   	wbc_init_bio(io->io_wbc, bio);
> -	return 0;
>   }
> 
>   static int io_submit_add_bh(struct ext4_io_submit *io,
> @@ -388,9 +385,7 @@ static int io_submit_add_bh(struct ext4_io_submit *io,
>   		ext4_io_submit(io);
>   	}
>   	if (io->io_bio == NULL) {
> -		ret = io_submit_init_bio(io, bh);
> -		if (ret)
> -			return ret;
> +		io_submit_init_bio(io, bh);
>   		io->io_bio->bi_write_hint = inode->i_write_hint;
>   	}
>   	ret = bio_add_page(io->io_bio, page, bh->b_size, bh_offset(bh));


Also we can further simplify it like below. Please check.

diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c
index f1f7b6601780..a3a2edeb3bbf 100644
--- a/fs/ext4/page-io.c
+++ b/fs/ext4/page-io.c
@@ -373,7 +373,7 @@ static void io_submit_init_bio(struct ext4_io_submit 
*io,
  	wbc_init_bio(io->io_wbc, bio);
  }

-static int io_submit_add_bh(struct ext4_io_submit *io,
+static void io_submit_add_bh(struct ext4_io_submit *io,
  			    struct inode *inode,
  			    struct page *page,
  			    struct buffer_head *bh)
@@ -393,7 +393,6 @@ static int io_submit_add_bh(struct ext4_io_submit *io,
  		goto submit_and_retry;
  	wbc_account_cgroup_owner(io->io_wbc, page, bh->b_size);
  	io->io_next_block++;
-	return 0;
  }

  int ext4_bio_write_page(struct ext4_io_submit *io,
@@ -495,30 +494,23 @@ int ext4_bio_write_page(struct ext4_io_submit *io,
  	do {
  		if (!buffer_async_write(bh))
  			continue;
-		ret = io_submit_add_bh(io, inode, bounce_page ?: page, bh);
-		if (ret) {
-			/*
-			 * We only get here on ENOMEM.  Not much else
-			 * we can do but mark the page as dirty, and
-			 * better luck next time.
-			 */
-			break;
-		}
+		io_submit_add_bh(io, inode, bounce_page ?: page, bh);
  		nr_submitted++;
  		clear_buffer_dirty(bh);
  	} while ((bh = bh->b_this_page) != head);

-	/* Error stopped previous loop? Clean up buffers... */
-	if (ret) {
-	out:
-		fscrypt_free_bounce_page(bounce_page);
-		printk_ratelimited(KERN_ERR "%s: ret = %d\n", __func__, ret);
-		redirty_page_for_writepage(wbc, page);
-		do {
-			clear_buffer_async_write(bh);
-			bh = bh->b_this_page;
-		} while (bh != head);
-	}
+	goto unlock;
+
+out:
+	fscrypt_free_bounce_page(bounce_page);
+	printk_ratelimited(KERN_ERR "%s: ret = %d\n", __func__, ret);
+	redirty_page_for_writepage(wbc, page);
+	do {
+		clear_buffer_async_write(bh);
+		bh = bh->b_this_page;
+	} while (bh != head);
+
+unlock:
  	unlock_page(page);
  	/* Nothing submitted - we have to end page writeback */
  	if (!nr_submitted)


-ritesh

Re: [PATCH] ext4: bio_alloc never fails

[Gao Xiang]

Hi Ritech,

On Wed, Oct 30, 2019 at 03:43:10PM +0530, Ritesh Harjani wrote:
> 
> 
> On 10/30/19 9:56 AM, Gao Xiang wrote:
> > Similar to [1] [2], it seems a trivial cleanup since
> > bio_alloc can handle memory allocation as mentioned in
> > fs/direct-io.c (also see fs/block_dev.c, fs/buffer.c, ..)
> > 
> 
> AFAIU, the reason is that, bio_alloc with __GFP_DIRECT_RECLAIM
> flags guarantees bio allocation under some given restrictions,
> as stated in fs/direct-io.c
> So here it is ok to not check for NULL value from bio_alloc.
> 
> I think we can update above info too in your commit msg.

Ok, I will update commit msg as you suggested.

> 
> > [1] https://lore.kernel.org/r/20191030035518.65477-1-gaoxiang25@huawei.com
> > [2] https://lore.kernel.org/r/20190830162812.GA10694@infradead.org
> > Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
> > ---
> >   fs/ext4/page-io.c  | 11 +++--------
> >   fs/ext4/readpage.c |  2 --
> >   2 files changed, 3 insertions(+), 10 deletions(-)
> > 
> > diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c
> > index 12ceadef32c5..f1f7b6601780 100644
> > --- a/fs/ext4/page-io.c
> > +++ b/fs/ext4/page-io.c
> > @@ -358,14 +358,12 @@ void ext4_io_submit_init(struct ext4_io_submit *io,
> >   	io->io_end = NULL;
> >   }
> > 
> > -static int io_submit_init_bio(struct ext4_io_submit *io,
> > -			      struct buffer_head *bh)
> > +static void io_submit_init_bio(struct ext4_io_submit *io,
> > +			       struct buffer_head *bh)
> >   {
> >   	struct bio *bio;
> > 
> >   	bio = bio_alloc(GFP_NOIO, BIO_MAX_PAGES);
> > -	if (!bio)
> > -		return -ENOMEM;
> >   	bio->bi_iter.bi_sector = bh->b_blocknr * (bh->b_size >> 9);
> >   	bio_set_dev(bio, bh->b_bdev);
> >   	bio->bi_end_io = ext4_end_bio;
> > @@ -373,7 +371,6 @@ static int io_submit_init_bio(struct ext4_io_submit *io,
> >   	io->io_bio = bio;
> >   	io->io_next_block = bh->b_blocknr;
> >   	wbc_init_bio(io->io_wbc, bio);
> > -	return 0;
> >   }
> > 
> >   static int io_submit_add_bh(struct ext4_io_submit *io,
> > @@ -388,9 +385,7 @@ static int io_submit_add_bh(struct ext4_io_submit *io,
> >   		ext4_io_submit(io);
> >   	}
> >   	if (io->io_bio == NULL) {
> > -		ret = io_submit_init_bio(io, bh);
> > -		if (ret)
> > -			return ret;
> > +		io_submit_init_bio(io, bh);
> >   		io->io_bio->bi_write_hint = inode->i_write_hint;
> >   	}
> >   	ret = bio_add_page(io->io_bio, page, bh->b_size, bh_offset(bh));
> 
> 
> Also we can further simplify it like below. Please check.

Got it, let me update the patch later. :-)
Thanks for your suggestion. I will wait for a while and
see if other opinions raise up...

Thanks,
Gao Xiang

> 
> diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c
> index f1f7b6601780..a3a2edeb3bbf 100644
> --- a/fs/ext4/page-io.c
> +++ b/fs/ext4/page-io.c
> @@ -373,7 +373,7 @@ static void io_submit_init_bio(struct ext4_io_submit
> *io,
>  	wbc_init_bio(io->io_wbc, bio);
>  }
> 
> -static int io_submit_add_bh(struct ext4_io_submit *io,
> +static void io_submit_add_bh(struct ext4_io_submit *io,
>  			    struct inode *inode,
>  			    struct page *page,
>  			    struct buffer_head *bh)
> @@ -393,7 +393,6 @@ static int io_submit_add_bh(struct ext4_io_submit *io,
>  		goto submit_and_retry;
>  	wbc_account_cgroup_owner(io->io_wbc, page, bh->b_size);
>  	io->io_next_block++;
> -	return 0;
>  }
> 
>  int ext4_bio_write_page(struct ext4_io_submit *io,
> @@ -495,30 +494,23 @@ int ext4_bio_write_page(struct ext4_io_submit *io,
>  	do {
>  		if (!buffer_async_write(bh))
>  			continue;
> -		ret = io_submit_add_bh(io, inode, bounce_page ?: page, bh);
> -		if (ret) {
> -			/*
> -			 * We only get here on ENOMEM.  Not much else
> -			 * we can do but mark the page as dirty, and
> -			 * better luck next time.
> -			 */
> -			break;
> -		}
> +		io_submit_add_bh(io, inode, bounce_page ?: page, bh);
>  		nr_submitted++;
>  		clear_buffer_dirty(bh);
>  	} while ((bh = bh->b_this_page) != head);
> 
> -	/* Error stopped previous loop? Clean up buffers... */
> -	if (ret) {
> -	out:
> -		fscrypt_free_bounce_page(bounce_page);
> -		printk_ratelimited(KERN_ERR "%s: ret = %d\n", __func__, ret);
> -		redirty_page_for_writepage(wbc, page);
> -		do {
> -			clear_buffer_async_write(bh);
> -			bh = bh->b_this_page;
> -		} while (bh != head);
> -	}
> +	goto unlock;
> +
> +out:
> +	fscrypt_free_bounce_page(bounce_page);
> +	printk_ratelimited(KERN_ERR "%s: ret = %d\n", __func__, ret);
> +	redirty_page_for_writepage(wbc, page);
> +	do {
> +		clear_buffer_async_write(bh);
> +		bh = bh->b_this_page;
> +	} while (bh != head);
> +
> +unlock:
>  	unlock_page(page);
>  	/* Nothing submitted - we have to end page writeback */
>  	if (!nr_submitted)
> 
> 
> -ritesh
> 

Re: [PATCH] ext4: bio_alloc never fails

[Theodore Y. Ts'o]

On Wed, Oct 30, 2019 at 03:43:10PM +0530, Ritesh Harjani wrote:
> 
> 
> On 10/30/19 9:56 AM, Gao Xiang wrote:
> > Similar to [1] [2], it seems a trivial cleanup since
> > bio_alloc can handle memory allocation as mentioned in
> > fs/direct-io.c (also see fs/block_dev.c, fs/buffer.c, ..)
> > 
> 
> AFAIU, the reason is that, bio_alloc with __GFP_DIRECT_RECLAIM
> flags guarantees bio allocation under some given restrictions,
> as stated in fs/direct-io.c
> So here it is ok to not check for NULL value from bio_alloc.
> 
> I think we can update above info too in your commit msg.

Please also add a short comment in the code itself, so it's clear why
it's OK to skip the error check, and reference the comments for
bio_alloc_bioset().  This is the fairly subtle bit which makes this
change not obvious:

 *   When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will
 *   always be able to allocate a bio. This is due to the mempool guarantees.
 *   To make this work, callers must never allocate more than 1 bio at a time
 *   from this pool. Callers that need to allocate more than 1 bio must always
 *   submit the previously allocated bio for IO before attempting to allocate
 *   a new one. Failure to do so can cause deadlocks under memory pressure.
 *
 *   Note that when running under generic_make_request() (i.e. any block
 *   driver), bios are not submitted until after you return - see the code in
 *   generic_make_request() that converts recursion into iteration, to prevent
 *   stack overflows.
 *
 *   This would normally mean allocating multiple bios under
 *   generic_make_request() would be susceptible to deadlocks, but we have
 *   deadlock avoidance code that resubmits any blocked bios from a rescuer
 *   thread.

Otherwise, someone else may not understand why it's safe to not check
the error return then submit cleanup patch to add the error checking
back.  :-)

					- Ted
					

Re: [PATCH] ext4: bio_alloc never fails

[Gao Xiang]

Hi Ted,

On Wed, Oct 30, 2019 at 11:04:37AM -0400, Theodore Y. Ts'o wrote:
> On Wed, Oct 30, 2019 at 03:43:10PM +0530, Ritesh Harjani wrote:
> > 
> > 
> > On 10/30/19 9:56 AM, Gao Xiang wrote:
> > > Similar to [1] [2], it seems a trivial cleanup since
> > > bio_alloc can handle memory allocation as mentioned in
> > > fs/direct-io.c (also see fs/block_dev.c, fs/buffer.c, ..)
> > > 
> > 
> > AFAIU, the reason is that, bio_alloc with __GFP_DIRECT_RECLAIM
> > flags guarantees bio allocation under some given restrictions,
> > as stated in fs/direct-io.c
> > So here it is ok to not check for NULL value from bio_alloc.
> > 
> > I think we can update above info too in your commit msg.
> 
> Please also add a short comment in the code itself, so it's clear why
> it's OK to skip the error check, and reference the comments for
> bio_alloc_bioset().  This is the fairly subtle bit which makes this
> change not obvious:

OK, I will add short comments in code then, and tidy up later since
it's not urgent (but I'm surprised that so many in-kernel code handles
that, those also makes me misleaded before, but I think mempool back
maybe better since the total efficient path is shorter compared with
error handling path)... and I'd like to know the f2fs side as well :-)

> 
>  *   When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will
>  *   always be able to allocate a bio. This is due to the mempool guarantees.
>  *   To make this work, callers must never allocate more than 1 bio at a time
>  *   from this pool. Callers that need to allocate more than 1 bio must always
>  *   submit the previously allocated bio for IO before attempting to allocate
>  *   a new one. Failure to do so can cause deadlocks under memory pressure.
>  *
>  *   Note that when running under generic_make_request() (i.e. any block
>  *   driver), bios are not submitted until after you return - see the code in
>  *   generic_make_request() that converts recursion into iteration, to prevent
>  *   stack overflows.
>  *
>  *   This would normally mean allocating multiple bios under
>  *   generic_make_request() would be susceptible to deadlocks, but we have
>  *   deadlock avoidance code that resubmits any blocked bios from a rescuer
>  *   thread.
> 
> Otherwise, someone else may not understand why it's safe to not check
> the error return then submit cleanup patch to add the error checking
> back.  :-)

Got it.

Thanks,
Gao Xiang

> 
> 					- Ted
>