diff mbox series

[v2,10/10] ext4: factor out a common helper to lock and flush data before fallocate

Message ID 20240904062925.716856-11-yi.zhang@huaweicloud.com (mailing list archive)
State New
Headers show
Series ext4: clean up and refactor fallocate | expand

Commit Message

Zhang Yi Sept. 4, 2024, 6:29 a.m. UTC 
From: Zhang Yi <yi.zhang@huawei.com>

Now the beginning of the first four functions in ext4_fallocate() (punch
hole, zero range, insert range and collapse range) are almost the same,
they need to wait for the dio to finish, get filemap invalidate lock,
write back dirty data and finally drop page cache. Factor out a common
helper to do these work can reduce a lot of the redundant code.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
---
 fs/ext4/ext4.h    |   2 +
 fs/ext4/extents.c | 125 ++++++++++++++++++++--------------------------
 fs/ext4/inode.c   |  25 +---------
 3 files changed, 57 insertions(+), 95 deletions(-)

Comments

Jan Kara Sept. 23, 2024, 8:54 a.m. UTC | #1 
On Wed 04-09-24 14:29:25, Zhang Yi wrote:
> From: Zhang Yi <yi.zhang@huawei.com>
> 
> Now the beginning of the first four functions in ext4_fallocate() (punch
> hole, zero range, insert range and collapse range) are almost the same,
> they need to wait for the dio to finish, get filemap invalidate lock,
> write back dirty data and finally drop page cache. Factor out a common
> helper to do these work can reduce a lot of the redundant code.
> 
> Signed-off-by: Zhang Yi <yi.zhang@huawei.com>

I like that we factor out this functionality in a common helper. But see
below:

> @@ -4731,6 +4707,52 @@ static int ext4_fallocate_check(struct inode *inode, int mode,
>  	return 0;
>  }
>  
> +int ext4_prepare_falloc(struct file *file, loff_t start, loff_t end, int mode)
> +{
> +	struct inode *inode = file_inode(file);
> +	struct address_space *mapping = inode->i_mapping;
> +	int ret;
> +
> +	/* Wait all existing dio workers, newcomers will block on i_rwsem */
> +	inode_dio_wait(inode);
> +	ret = file_modified(file);
> +	if (ret)
> +		return ret;
> +
> +	/*
> +	 * Prevent page faults from reinstantiating pages we have released
> +	 * from page cache.
> +	 */
> +	filemap_invalidate_lock(mapping);
> +
> +	ret = ext4_break_layouts(inode);
> +	if (ret)
> +		goto failed;
> +
> +	/*
> +	 * Write data that will be zeroed to preserve them when successfully
> +	 * discarding page cache below but fail to convert extents.
> +	 */
> +	ret = filemap_write_and_wait_range(mapping, start, end);

The comment is somewhat outdated now. Also the range is wrong for collapse
and insert range. There we need to writeout data upto the EOF because we
truncate it below.

								Honza

> +	if (ret)
> +		goto failed;
> +
> +	/*
> +	 * For insert range and collapse range, COWed private pages should
> +	 * be removed since the file's logical offset will be changed, but
> +	 * punch hole and zero range doesn't.
> +	 */
> +	if (mode & (FALLOC_FL_INSERT_RANGE | FALLOC_FL_COLLAPSE_RANGE))
> +		truncate_pagecache(inode, start);
> +	else
> +		truncate_pagecache_range(inode, start, end);
> +
> +	return 0;
> +failed:
> +	filemap_invalidate_unlock(mapping);
> +	return ret;
> +}
Zhang Yi Sept. 24, 2024, 8:11 a.m. UTC | #2 
On 2024/9/23 16:54, Jan Kara wrote:
> On Wed 04-09-24 14:29:25, Zhang Yi wrote:
>> From: Zhang Yi <yi.zhang@huawei.com>
>>
>> Now the beginning of the first four functions in ext4_fallocate() (punch
>> hole, zero range, insert range and collapse range) are almost the same,
>> they need to wait for the dio to finish, get filemap invalidate lock,
>> write back dirty data and finally drop page cache. Factor out a common
>> helper to do these work can reduce a lot of the redundant code.
>>
>> Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
> 
> I like that we factor out this functionality in a common helper. But see
> below:
> 
>> @@ -4731,6 +4707,52 @@ static int ext4_fallocate_check(struct inode *inode, int mode,
>>  	return 0;
>>  }
>>  
>> +int ext4_prepare_falloc(struct file *file, loff_t start, loff_t end, int mode)
>> +{
>> +	struct inode *inode = file_inode(file);
>> +	struct address_space *mapping = inode->i_mapping;
>> +	int ret;
>> +
>> +	/* Wait all existing dio workers, newcomers will block on i_rwsem */
>> +	inode_dio_wait(inode);
>> +	ret = file_modified(file);
>> +	if (ret)
>> +		return ret;
>> +
>> +	/*
>> +	 * Prevent page faults from reinstantiating pages we have released
>> +	 * from page cache.
>> +	 */
>> +	filemap_invalidate_lock(mapping);
>> +
>> +	ret = ext4_break_layouts(inode);
>> +	if (ret)
>> +		goto failed;
>> +
>> +	/*
>> +	 * Write data that will be zeroed to preserve them when successfully
>> +	 * discarding page cache below but fail to convert extents.
>> +	 */
>> +	ret = filemap_write_and_wait_range(mapping, start, end);
> 
> The comment is somewhat outdated now.

Sure, will update it in next iteration.

> Also the range is wrong for collapse
> and insert range. There we need to writeout data upto the EOF because we
> truncate it below.
> 

For collapse and insert range, I passed the length LLONG_MAX, which is
the same as before, this should've upto the EOF, so I think it's
right, or am I missing something?

ext4_collapse_range():

-	start = round_down(offset, PAGE_SIZE);
-	/* Write out all dirty pages */
-	ret = filemap_write_and_wait_range(mapping, start, LLONG_MAX);
+	ret = ext4_prepare_falloc(file, round_down(offset, PAGE_SIZE),
+				  LLONG_MAX, FALLOC_FL_COLLAPSE_RANGE);


ext4_insert_range():

-	start = round_down(offset, PAGE_SIZE);
-	/* Write out all dirty pages */
-	ret = filemap_write_and_wait_range(mapping, start, LLONG_MAX);
+	ret = ext4_prepare_falloc(file, round_down(offset, PAGE_SIZE),
+				  LLONG_MAX, FALLOC_FL_INSERT_RANGE);

Thanks,
Yi.

> 
>> +	if (ret)
>> +		goto failed;
>> +
>> +	/*
>> +	 * For insert range and collapse range, COWed private pages should
>> +	 * be removed since the file's logical offset will be changed, but
>> +	 * punch hole and zero range doesn't.
>> +	 */
>> +	if (mode & (FALLOC_FL_INSERT_RANGE | FALLOC_FL_COLLAPSE_RANGE))
>> +		truncate_pagecache(inode, start);
>> +	else
>> +		truncate_pagecache_range(inode, start, end);
>> +
>> +	return 0;
>> +failed:
>> +	filemap_invalidate_unlock(mapping);
>> +	return ret;
>> +}
>
Jan Kara Sept. 24, 2024, 10:05 a.m. UTC | #3 
On Tue 24-09-24 16:11:08, Zhang Yi wrote:
> On 2024/9/23 16:54, Jan Kara wrote:
> > On Wed 04-09-24 14:29:25, Zhang Yi wrote:
> > Also the range is wrong for collapse
> > and insert range. There we need to writeout data upto the EOF because we
> > truncate it below.
> > 
> 
> For collapse and insert range, I passed the length LLONG_MAX, which is
> the same as before, this should've upto the EOF, so I think it's
> right, or am I missing something?
> 
> ext4_collapse_range():
> 
> -	start = round_down(offset, PAGE_SIZE);
> -	/* Write out all dirty pages */
> -	ret = filemap_write_and_wait_range(mapping, start, LLONG_MAX);
> +	ret = ext4_prepare_falloc(file, round_down(offset, PAGE_SIZE),
> +				  LLONG_MAX, FALLOC_FL_COLLAPSE_RANGE);
> 
> 
> ext4_insert_range():
> 
> -	start = round_down(offset, PAGE_SIZE);
> -	/* Write out all dirty pages */
> -	ret = filemap_write_and_wait_range(mapping, start, LLONG_MAX);
> +	ret = ext4_prepare_falloc(file, round_down(offset, PAGE_SIZE),
> +				  LLONG_MAX, FALLOC_FL_INSERT_RANGE);
> 

Ah sorry, I've missed these bits. So we should be fine.

								Honza
diff mbox series

Patch

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index e8d7965f62c4..281fab9abc42 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -3696,6 +3696,8 @@  extern int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start,
 				 ext4_lblk_t end);
 extern void ext4_ext_init(struct super_block *);
 extern void ext4_ext_release(struct super_block *);
+extern int ext4_prepare_falloc(struct file *file, loff_t start, loff_t end,
+			       int mode);
 extern long ext4_fallocate(struct file *file, int mode, loff_t offset,
 			  loff_t len);
 extern int ext4_convert_unwritten_extents(handle_t *handle, struct inode *inode,
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 91e509201915..eee63e92dcc6 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -4558,34 +4558,10 @@  static long ext4_zero_range(struct file *file, loff_t offset,
 			return ret;
 	}
 
-	/* Wait all existing dio workers, newcomers will block on i_rwsem */
-	inode_dio_wait(inode);
-
-	ret = file_modified(file);
+	ret = ext4_prepare_falloc(file, offset, end - 1, FALLOC_FL_ZERO_RANGE);
 	if (ret)
 		return ret;
 
-	/*
-	 * Prevent page faults from reinstantiating pages we have released
-	 * from page cache.
-	 */
-	filemap_invalidate_lock(mapping);
-
-	ret = ext4_break_layouts(inode);
-	if (ret)
-		goto out_invalidate_lock;
-
-	/*
-	 * Write data that will be zeroed to preserve them when successfully
-	 * discarding page cache below but fail to convert extents.
-	 */
-	ret = filemap_write_and_wait_range(mapping, offset, end - 1);
-	if (ret)
-		goto out_invalidate_lock;
-
-	/* Now release the pages and zero block aligned part of pages */
-	truncate_pagecache_range(inode, offset, end - 1);
-
 	flags = EXT4_GET_BLOCKS_CREATE_UNWRIT_EXT;
 	/* Preallocate the range including the unaligned edges */
 	if (offset & (blocksize - 1) || end & (blocksize - 1)) {
@@ -4731,6 +4707,52 @@  static int ext4_fallocate_check(struct inode *inode, int mode,
 	return 0;
 }
 
+int ext4_prepare_falloc(struct file *file, loff_t start, loff_t end, int mode)
+{
+	struct inode *inode = file_inode(file);
+	struct address_space *mapping = inode->i_mapping;
+	int ret;
+
+	/* Wait all existing dio workers, newcomers will block on i_rwsem */
+	inode_dio_wait(inode);
+	ret = file_modified(file);
+	if (ret)
+		return ret;
+
+	/*
+	 * Prevent page faults from reinstantiating pages we have released
+	 * from page cache.
+	 */
+	filemap_invalidate_lock(mapping);
+
+	ret = ext4_break_layouts(inode);
+	if (ret)
+		goto failed;
+
+	/*
+	 * Write data that will be zeroed to preserve them when successfully
+	 * discarding page cache below but fail to convert extents.
+	 */
+	ret = filemap_write_and_wait_range(mapping, start, end);
+	if (ret)
+		goto failed;
+
+	/*
+	 * For insert range and collapse range, COWed private pages should
+	 * be removed since the file's logical offset will be changed, but
+	 * punch hole and zero range doesn't.
+	 */
+	if (mode & (FALLOC_FL_INSERT_RANGE | FALLOC_FL_COLLAPSE_RANGE))
+		truncate_pagecache(inode, start);
+	else
+		truncate_pagecache_range(inode, start, end);
+
+	return 0;
+failed:
+	filemap_invalidate_unlock(mapping);
+	return ret;
+}
+
 /*
  * preallocate space for a file. This implements ext4's fallocate file
  * operation, which gets called from sys_fallocate system call.
@@ -5284,39 +5306,20 @@  static int ext4_collapse_range(struct file *file, loff_t offset, loff_t len)
 	ext4_lblk_t start_lblk, end_lblk;
 	handle_t *handle;
 	unsigned int credits;
-	loff_t start, new_size;
+	loff_t new_size;
 	int ret;
 
 	trace_ext4_collapse_range(inode, offset, len);
 	WARN_ON_ONCE(!inode_is_locked(inode));
 
-	/* Wait for existing dio to complete */
-	inode_dio_wait(inode);
-
-	ret = file_modified(file);
-	if (ret)
-		return ret;
-
-	/*
-	 * Prevent page faults from reinstantiating pages we have released from
-	 * page cache.
-	 */
-	filemap_invalidate_lock(mapping);
-
-	ret = ext4_break_layouts(inode);
-	if (ret)
-		goto out_invalidate_lock;
-
 	/*
 	 * Need to round down offset to be aligned with page size boundary
 	 * for page size > block size.
 	 */
-	start = round_down(offset, PAGE_SIZE);
-	/* Write out all dirty pages */
-	ret = filemap_write_and_wait_range(mapping, start, LLONG_MAX);
+	ret = ext4_prepare_falloc(file, round_down(offset, PAGE_SIZE),
+				  LLONG_MAX, FALLOC_FL_COLLAPSE_RANGE);
 	if (ret)
-		goto out_invalidate_lock;
-	truncate_pagecache(inode, start);
+		return ret;
 
 	credits = ext4_writepage_trans_blocks(inode);
 	handle = ext4_journal_start(inode, EXT4_HT_TRUNCATE, credits);
@@ -5386,38 +5389,18 @@  static int ext4_insert_range(struct file *file, loff_t offset, loff_t len)
 	ext4_lblk_t start_lblk, len_lblk, ee_start_lblk = 0;
 	unsigned int credits, ee_len;
 	int ret, depth, split_flag = 0;
-	loff_t start;
 
 	trace_ext4_insert_range(inode, offset, len);
 	WARN_ON_ONCE(!inode_is_locked(inode));
 
-	/* Wait for existing dio to complete */
-	inode_dio_wait(inode);
-
-	ret = file_modified(file);
-	if (ret)
-		return ret;
-
-	/*
-	 * Prevent page faults from reinstantiating pages we have released from
-	 * page cache.
-	 */
-	filemap_invalidate_lock(mapping);
-
-	ret = ext4_break_layouts(inode);
-	if (ret)
-		goto out_invalidate_lock;
-
 	/*
 	 * Need to round down to align start offset to page size boundary
 	 * for page size > block size.
 	 */
-	start = round_down(offset, PAGE_SIZE);
-	/* Write out all dirty pages */
-	ret = filemap_write_and_wait_range(mapping, start, LLONG_MAX);
+	ret = ext4_prepare_falloc(file, round_down(offset, PAGE_SIZE),
+				  LLONG_MAX, FALLOC_FL_INSERT_RANGE);
 	if (ret)
-		goto out_invalidate_lock;
-	truncate_pagecache(inode, start);
+		return ret;
 
 	credits = ext4_writepage_trans_blocks(inode);
 	handle = ext4_journal_start(inode, EXT4_HT_TRUNCATE, credits);
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 57636c656fa5..4b7f8fcaa5c2 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -3953,33 +3953,10 @@  int ext4_punch_hole(struct file *file, loff_t offset, loff_t length)
 			return ret;
 	}
 
-	/* Wait all existing dio workers, newcomers will block on i_rwsem */
-	inode_dio_wait(inode);
-
-	ret = file_modified(file);
+	ret = ext4_prepare_falloc(file, offset, end - 1, FALLOC_FL_PUNCH_HOLE);
 	if (ret)
 		return ret;
 
-	/*
-	 * Prevent page faults from reinstantiating pages we have released from
-	 * page cache.
-	 */
-	filemap_invalidate_lock(mapping);
-
-	ret = ext4_break_layouts(inode);
-	if (ret)
-		goto out_invalidate_lock;
-
-	/* Write out all dirty pages to avoid race conditions */
-	if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) {
-		ret = filemap_write_and_wait_range(mapping, offset, end - 1);
-		if (ret)
-			goto out_invalidate_lock;
-	}
-
-	/* Now release the pages and zero block aligned part of pages*/
-	truncate_pagecache_range(inode, offset, end - 1);
-
 	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
 		credits = ext4_writepage_trans_blocks(inode);
 	else