[PATCH 0/5] Remove zero_user()

[PATCH 0/5] Remove zero_user()

[Matthew Wilcox (Oracle)]

The zero_user() API is almost unused these days.  Finish the job of
removing it.

Matthew Wilcox (Oracle) (5):
  bio: Use memzero_page() in bio_truncate()
  null_blk: Use memzero_page()
  direct-io: Use memzero_page()
  ceph: Convert ceph_zero_partial_page() to use a folio
  mm: Remove zero_user()

 block/bio.c                   |  4 ++--
 drivers/block/null_blk/main.c |  2 +-
 fs/ceph/file.c                | 21 ++++++++++-----------
 fs/direct-io.c                |  2 +-
 include/linux/highmem.h       |  6 ------
 5 files changed, 14 insertions(+), 21 deletions(-)

-- 
2.47.2

[PATCH 1/5] bio: Use memzero_page() in bio_truncate()

[Matthew Wilcox (Oracle)]

memzero_page() is the new name for zero_user().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
---
 block/bio.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/block/bio.c b/block/bio.c
index 3c0a558c90f5..ce16c34ec6de 100644
--- a/block/bio.c
+++ b/block/bio.c
@@ -653,13 +653,13 @@ static void bio_truncate(struct bio *bio, unsigned new_size)
 
 	bio_for_each_segment(bv, bio, iter) {
 		if (done + bv.bv_len > new_size) {
-			unsigned offset;
+			size_t offset;
 
 			if (!truncated)
 				offset = new_size - done;
 			else
 				offset = 0;
-			zero_user(bv.bv_page, bv.bv_offset + offset,
+			memzero_page(bv.bv_page, bv.bv_offset + offset,
 				  bv.bv_len - offset);
 			truncated = true;
 		}
-- 
2.47.2

[PATCH 2/5] null_blk: Use memzero_page()

[Matthew Wilcox (Oracle)]

memzero_page() is the new name for zero_user().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
---
 drivers/block/null_blk/main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/block/null_blk/main.c b/drivers/block/null_blk/main.c
index aa163ae9b2aa..91642c9a3b29 100644
--- a/drivers/block/null_blk/main.c
+++ b/drivers/block/null_blk/main.c
@@ -1179,7 +1179,7 @@ static int copy_from_nullb(struct nullb *nullb, struct page *dest,
 			memcpy_page(dest, off + count, t_page->page, offset,
 				    temp);
 		else
-			zero_user(dest, off + count, temp);
+			memzero_page(dest, off + count, temp);
 
 		count += temp;
 		sector += temp >> SECTOR_SHIFT;
-- 
2.47.2

[PATCH 3/5] direct-io: Use memzero_page()

[Matthew Wilcox (Oracle)]

memzero_page() is the new name for zero_user().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
---
 fs/direct-io.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/direct-io.c b/fs/direct-io.c
index bbd05f1a2145..111958634def 100644
--- a/fs/direct-io.c
+++ b/fs/direct-io.c
@@ -996,7 +996,7 @@ static int do_direct_IO(struct dio *dio, struct dio_submit *sdio,
 					dio_unpin_page(dio, page);
 					goto out;
 				}
-				zero_user(page, from, 1 << blkbits);
+				memzero_page(page, from, 1 << blkbits);
 				sdio->block_in_file++;
 				from += 1 << blkbits;
 				dio->result += 1 << blkbits;
-- 
2.47.2

[PATCH 4/5] ceph: Convert ceph_zero_partial_page() to use a folio

[Matthew Wilcox (Oracle)]

Retrieve a folio from the pagecache instead of a page and operate on it.
Removes several hidden calls to compound_head() along with calls to
deprecated functions like wait_on_page_writeback() and find_lock_page().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
---
 fs/ceph/file.c | 21 ++++++++++-----------
 1 file changed, 10 insertions(+), 11 deletions(-)

diff --git a/fs/ceph/file.c b/fs/ceph/file.c
index a7254cab44cc..d5c674d2ba8a 100644
--- a/fs/ceph/file.c
+++ b/fs/ceph/file.c
@@ -2530,18 +2530,17 @@ static loff_t ceph_llseek(struct file *file, loff_t offset, int whence)
 	return generic_file_llseek(file, offset, whence);
 }
 
-static inline void ceph_zero_partial_page(
-	struct inode *inode, loff_t offset, unsigned size)
+static inline void ceph_zero_partial_page(struct inode *inode,
+		loff_t offset, size_t size)
 {
-	struct page *page;
-	pgoff_t index = offset >> PAGE_SHIFT;
-
-	page = find_lock_page(inode->i_mapping, index);
-	if (page) {
-		wait_on_page_writeback(page);
-		zero_user(page, offset & (PAGE_SIZE - 1), size);
-		unlock_page(page);
-		put_page(page);
+	struct folio *folio;
+
+	folio = filemap_lock_folio(inode->i_mapping, offset >> PAGE_SHIFT);
+	if (folio) {
+		folio_wait_writeback(folio);
+		folio_zero_range(folio, offset_in_folio(folio, offset), size);
+		folio_unlock(folio);
+		folio_put(folio);
 	}
 }
 
-- 
2.47.2

[PATCH 5/5] mm: Remove zero_user()

[Matthew Wilcox (Oracle)]

All users have now been converted to either memzero_page() or
folio_zero_range().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
---
 include/linux/highmem.h | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/include/linux/highmem.h b/include/linux/highmem.h
index e48d7f27b0b9..a30526cc53a7 100644
--- a/include/linux/highmem.h
+++ b/include/linux/highmem.h
@@ -292,12 +292,6 @@ static inline void zero_user_segment(struct page *page,
 	zero_user_segments(page, start, end, 0, 0);
 }
 
-static inline void zero_user(struct page *page,
-	unsigned start, unsigned size)
-{
-	zero_user_segments(page, start, start + size, 0, 0);
-}
-
 #ifndef __HAVE_ARCH_COPY_USER_HIGHPAGE
 
 static inline void copy_user_highpage(struct page *to, struct page *from,
-- 
2.47.2

Re: [PATCH 4/5] ceph: Convert ceph_zero_partial_page() to use a folio

[Viacheslav Dubeyko]

On Thu, 2025-06-12 at 15:34 +0100, Matthew Wilcox (Oracle) wrote:
> Retrieve a folio from the pagecache instead of a page and operate on it.
> Removes several hidden calls to compound_head() along with calls to
> deprecated functions like wait_on_page_writeback() and find_lock_page().
> 
> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> ---
>  fs/ceph/file.c | 21 ++++++++++-----------
>  1 file changed, 10 insertions(+), 11 deletions(-)
> 
> diff --git a/fs/ceph/file.c b/fs/ceph/file.c
> index a7254cab44cc..d5c674d2ba8a 100644
> --- a/fs/ceph/file.c
> +++ b/fs/ceph/file.c
> @@ -2530,18 +2530,17 @@ static loff_t ceph_llseek(struct file *file, loff_t offset, int whence)
>  	return generic_file_llseek(file, offset, whence);
>  }
>  
> -static inline void ceph_zero_partial_page(
> -	struct inode *inode, loff_t offset, unsigned size)
> +static inline void ceph_zero_partial_page(struct inode *inode,
> +		loff_t offset, size_t size)
>  {
> -	struct page *page;
> -	pgoff_t index = offset >> PAGE_SHIFT;
> -
> -	page = find_lock_page(inode->i_mapping, index);
> -	if (page) {
> -		wait_on_page_writeback(page);
> -		zero_user(page, offset & (PAGE_SIZE - 1), size);
> -		unlock_page(page);
> -		put_page(page);
> +	struct folio *folio;
> +
> +	folio = filemap_lock_folio(inode->i_mapping, offset >> PAGE_SHIFT);
> +	if (folio) {
> +		folio_wait_writeback(folio);
> +		folio_zero_range(folio, offset_in_folio(folio, offset), size);
> +		folio_unlock(folio);
> +		folio_put(folio);
>  	}
>  }
>  

Looks really good. And filemap_lock_folio() is more efficient than
find_lock_page() now.

Reviewed-by: Viacheslav Dubeyko <Slava.Dubeyko@ibm.com>

Thanks,
Slava.

Re: [PATCH 0/5] Remove zero_user()

[Christoph Hellwig]

On Thu, Jun 12, 2025 at 03:34:36PM +0100, Matthew Wilcox (Oracle) wrote:
> The zero_user() API is almost unused these days.  Finish the job of
> removing it.

Both the block layer users really should use bvec based helpers.
I was planning to get to that this merge window.  Can we queue up
just the other two removals for and remove zero_user after -rc1
to reduce conflicts?

Re: [PATCH 0/5] Remove zero_user()

[Matthew Wilcox]

On Fri, Jun 13, 2025 at 07:24:32AM +0200, Christoph Hellwig wrote:
> On Thu, Jun 12, 2025 at 03:34:36PM +0100, Matthew Wilcox (Oracle) wrote:
> > The zero_user() API is almost unused these days.  Finish the job of
> > removing it.
> 
> Both the block layer users really should use bvec based helpers.
> I was planning to get to that this merge window.  Can we queue up
> just the other two removals for and remove zero_user after -rc1
> to reduce conflicts?

If I'd known you were doing that, I wouldn't've bothered.  However,
Andrew's taken the patches now, so I'm inclined to leave them in.
No matter which tree it gets merged through, this is a relatively easy
conflict to resolve (ie just take your version).  I have some more
patches which build on the removal of zero_user() so it'd be nice to
not hold them up.

Re: [PATCH 0/5] Remove zero_user()

[Andrew Morton]

On Fri, 13 Jun 2025 20:51:06 +0100 Matthew Wilcox <willy@infradead.org> wrote:

> On Fri, Jun 13, 2025 at 07:24:32AM +0200, Christoph Hellwig wrote:
> > On Thu, Jun 12, 2025 at 03:34:36PM +0100, Matthew Wilcox (Oracle) wrote:
> > > The zero_user() API is almost unused these days.  Finish the job of
> > > removing it.
> > 
> > Both the block layer users really should use bvec based helpers.
> > I was planning to get to that this merge window.  Can we queue up
> > just the other two removals for and remove zero_user after -rc1
> > to reduce conflicts?
> 
> If I'd known you were doing that, I wouldn't've bothered.  However,
> Andrew's taken the patches now, so I'm inclined to leave them in.
> No matter which tree it gets merged through, this is a relatively easy
> conflict to resolve (ie just take your version).  I have some more
> patches which build on the removal of zero_user() so it'd be nice to
> not hold them up.

Sure, Christoph, please just proceed with the block changes and we can
see what the conflicts look like when Stephen hits them.  If Matthew's
series needs modification then so be it.

Re: [PATCH 3/5] direct-io: Use memzero_page()

[Alex Markuze]

Good cleanup.

Reviewed-by: Alex Markuze amarkuze@redhat.com

On Thu, Jun 12, 2025 at 5:36 PM Matthew Wilcox (Oracle)
<willy@infradead.org> wrote:
>
> memzero_page() is the new name for zero_user().
>
> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> ---
>  fs/direct-io.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/direct-io.c b/fs/direct-io.c
> index bbd05f1a2145..111958634def 100644
> --- a/fs/direct-io.c
> +++ b/fs/direct-io.c
> @@ -996,7 +996,7 @@ static int do_direct_IO(struct dio *dio, struct dio_submit *sdio,
>                                         dio_unpin_page(dio, page);
>                                         goto out;
>                                 }
> -                               zero_user(page, from, 1 << blkbits);
> +                               memzero_page(page, from, 1 << blkbits);
>                                 sdio->block_in_file++;
>                                 from += 1 << blkbits;
>                                 dio->result += 1 << blkbits;
> --
> 2.47.2
>
>

Re: [PATCH 1/5] bio: Use memzero_page() in bio_truncate()

[Alex Markuze]

Good cleanup.

Reviewed-by: Alex Markuze amarkuze@redhat.com

On Thu, Jun 12, 2025 at 5:35 PM Matthew Wilcox (Oracle)
<willy@infradead.org> wrote:
>
> memzero_page() is the new name for zero_user().
>
> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> ---
>  block/bio.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/block/bio.c b/block/bio.c
> index 3c0a558c90f5..ce16c34ec6de 100644
> --- a/block/bio.c
> +++ b/block/bio.c
> @@ -653,13 +653,13 @@ static void bio_truncate(struct bio *bio, unsigned new_size)
>
>         bio_for_each_segment(bv, bio, iter) {
>                 if (done + bv.bv_len > new_size) {
> -                       unsigned offset;
> +                       size_t offset;
>
>                         if (!truncated)
>                                 offset = new_size - done;
>                         else
>                                 offset = 0;
> -                       zero_user(bv.bv_page, bv.bv_offset + offset,
> +                       memzero_page(bv.bv_page, bv.bv_offset + offset,
>                                   bv.bv_len - offset);
>                         truncated = true;
>                 }
> --
> 2.47.2
>
>

Re: [PATCH 2/5] null_blk: Use memzero_page()

[Alex Markuze]

Good cleanup.

Reviewed-by: Alex Markuze amarkuze@redhat.com

On Thu, Jun 12, 2025 at 5:35 PM Matthew Wilcox (Oracle)
<willy@infradead.org> wrote:
>
> memzero_page() is the new name for zero_user().
>
> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> ---
>  drivers/block/null_blk/main.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/block/null_blk/main.c b/drivers/block/null_blk/main.c
> index aa163ae9b2aa..91642c9a3b29 100644
> --- a/drivers/block/null_blk/main.c
> +++ b/drivers/block/null_blk/main.c
> @@ -1179,7 +1179,7 @@ static int copy_from_nullb(struct nullb *nullb, struct page *dest,
>                         memcpy_page(dest, off + count, t_page->page, offset,
>                                     temp);
>                 else
> -                       zero_user(dest, off + count, temp);
> +                       memzero_page(dest, off + count, temp);
>
>                 count += temp;
>                 sector += temp >> SECTOR_SHIFT;
> --
> 2.47.2
>
>

Re: [PATCH 5/5] mm: Remove zero_user()

[Alex Markuze]

Good cleanup.

Reviewed-by: Alex Markuze amarkuze@redhat.com

On Thu, Jun 12, 2025 at 5:35 PM Matthew Wilcox (Oracle)
<willy@infradead.org> wrote:
>
> All users have now been converted to either memzero_page() or
> folio_zero_range().
>
> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> ---
>  include/linux/highmem.h | 6 ------
>  1 file changed, 6 deletions(-)
>
> diff --git a/include/linux/highmem.h b/include/linux/highmem.h
> index e48d7f27b0b9..a30526cc53a7 100644
> --- a/include/linux/highmem.h
> +++ b/include/linux/highmem.h
> @@ -292,12 +292,6 @@ static inline void zero_user_segment(struct page *page,
>         zero_user_segments(page, start, end, 0, 0);
>  }
>
> -static inline void zero_user(struct page *page,
> -       unsigned start, unsigned size)
> -{
> -       zero_user_segments(page, start, start + size, 0, 0);
> -}
> -
>  #ifndef __HAVE_ARCH_COPY_USER_HIGHPAGE
>
>  static inline void copy_user_highpage(struct page *to, struct page *from,
> --
> 2.47.2
>
>