[PATCH v2 3/5] fat: zero out seek range on _fat_get_block

[PATCH v2 3/5] fat: zero out seek range on _fat_get_block

[Namjae Jeon]

From: Namjae Jeon <namjae.jeon@samsung.com>

For normal buffered write operations, normally if we try to write to an
offset > than file size, it does a cont_expand_zero till that offset.
Now, in case of fallocated regions, since the blocks are already allocated.
So, make it zero out that buffers for those blocks till the seek'ed offset.

Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Amit Sahrawat <a.sahrawat@samsung.com>
---
 fs/fat/cache.c |    3 ++-
 fs/fat/inode.c |   15 +++++++++++++--
 2 files changed, 15 insertions(+), 3 deletions(-)

diff --git a/fs/fat/cache.c b/fs/fat/cache.c
index a132666..c56bd7e 100644
--- a/fs/fat/cache.c
+++ b/fs/fat/cache.c
@@ -334,7 +334,8 @@ int fat_bmap(struct inode *inode, sector_t sector, sector_t *phys,
 		 */
 		last_block = (MSDOS_I(inode)->i_disksize + (blocksize - 1))
 			>> blocksize_bits;
-		if (sector >= last_block)
+		if (sector >= last_block &&
+		    MSDOS_I(inode)->mmu_private == MSDOS_I(inode)->i_disksize)
 			return 0;
 	}
 
diff --git a/fs/fat/inode.c b/fs/fat/inode.c
index 655dcc7..12a37a9 100644
--- a/fs/fat/inode.c
+++ b/fs/fat/inode.c
@@ -61,15 +61,26 @@ static inline int __fat_get_block(struct inode *inode, sector_t iblock,
 	struct super_block *sb = inode->i_sb;
 	struct msdos_sb_info *sbi = MSDOS_SB(sb);
 	unsigned long mapped_blocks;
-	sector_t phys;
+	sector_t phys, last_block, disk_block;
 	int err, offset;
+	const unsigned long blocksize = sb->s_blocksize;
+	const unsigned char blocksize_bits = sb->s_blocksize_bits;
 
 	err = fat_bmap(inode, iblock, &phys, &mapped_blocks, create);
 	if (err)
 		return err;
 	if (phys) {
-		map_bh(bh_result, sb, phys);
 		*max_blocks = min(mapped_blocks, *max_blocks);
+		last_block = (MSDOS_I(inode)->mmu_private + (blocksize - 1))
+			>> blocksize_bits;
+		disk_block = (MSDOS_I(inode)->i_disksize + (blocksize - 1))
+			>> blocksize_bits;
+		if (iblock >= last_block && iblock <= disk_block) {
+			MSDOS_I(inode)->mmu_private +=
+				*max_blocks << blocksize_bits;
+			set_buffer_new(bh_result);
+		}
+		map_bh(bh_result, sb, phys);
 		return 0;
 	}
 	if (!create)
-- 
1.7.9.5

Re: [PATCH v2 3/5] fat: zero out seek range on _fat_get_block

[OGAWA Hirofumi]

Namjae Jeon <linkinjeon@gmail.com> writes:

> --- a/fs/fat/cache.c
> +++ b/fs/fat/cache.c
> @@ -334,7 +334,8 @@ int fat_bmap(struct inode *inode, sector_t sector, sector_t *phys,
>  		 */
>  		last_block = (MSDOS_I(inode)->i_disksize + (blocksize - 1))
>  			>> blocksize_bits;
> -		if (sector >= last_block)
> +		if (sector >= last_block &&
> +		    MSDOS_I(inode)->mmu_private == MSDOS_I(inode)->i_disksize)
>  			return 0;

This is strange check. Why do we want to map even if "sector > ->i_disksize"
(when ->mmu_private != ->i_disksize)?


Another point, for ->bmap(), maybe we will need new flag to read beyond
->i_size? Because ->bmap() returns physical address even if > ->i_size.

> diff --git a/fs/fat/inode.c b/fs/fat/inode.c
> index 655dcc7..12a37a9 100644
> --- a/fs/fat/inode.c
> +++ b/fs/fat/inode.c
> @@ -61,15 +61,26 @@ static inline int __fat_get_block(struct inode *inode, sector_t iblock,
>  	struct super_block *sb = inode->i_sb;
>  	struct msdos_sb_info *sbi = MSDOS_SB(sb);
>  	unsigned long mapped_blocks;
> -	sector_t phys;
> +	sector_t phys, last_block, disk_block;
>  	int err, offset;
> +	const unsigned long blocksize = sb->s_blocksize;
> +	const unsigned char blocksize_bits = sb->s_blocksize_bits;
>  
>  	err = fat_bmap(inode, iblock, &phys, &mapped_blocks, create);
>  	if (err)
>  		return err;
>  	if (phys) {
> -		map_bh(bh_result, sb, phys);
>  		*max_blocks = min(mapped_blocks, *max_blocks);
> +		last_block = (MSDOS_I(inode)->mmu_private + (blocksize - 1))
> +			>> blocksize_bits;
> +		disk_block = (MSDOS_I(inode)->i_disksize + (blocksize - 1))
> +			>> blocksize_bits;
> +		if (iblock >= last_block && iblock <= disk_block) {
> +			MSDOS_I(inode)->mmu_private +=
> +				*max_blocks << blocksize_bits;
> +			set_buffer_new(bh_result);
> +		}
> +		map_bh(bh_result, sb, phys);
>  		return 0;
>  	}
>  	if (!create)

Let's give the name for this operation for readability, and make it
static function. With this, I think this one much readable.

static void check_fallocate_region()
{
	/* code here */
}
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: [PATCH v2 3/5] fat: zero out seek range on _fat_get_block

[Namjae Jeon]

2013/12/1, OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>:
> Namjae Jeon <linkinjeon@gmail.com> writes:
>
>> --- a/fs/fat/cache.c
>> +++ b/fs/fat/cache.c
>> @@ -334,7 +334,8 @@ int fat_bmap(struct inode *inode, sector_t sector,
>> sector_t *phys,
>>  		 */
>>  		last_block = (MSDOS_I(inode)->i_disksize + (blocksize - 1))
>>  			>> blocksize_bits;
>> -		if (sector >= last_block)
>> +		if (sector >= last_block &&
>> +		    MSDOS_I(inode)->mmu_private == MSDOS_I(inode)->i_disksize)
>>  			return 0;
>
> This is strange check. Why do we want to map even if "sector >
> ->i_disksize"
> (when ->mmu_private != ->i_disksize)?
Yes, Don't want it. I will fix it.
>
>
> Another point, for ->bmap(), maybe we will need new flag to read beyond
> ->i_size? Because ->bmap() returns physical address even if > ->i_size.
Okay, I will change it as your comment.
>
>> diff --git a/fs/fat/inode.c b/fs/fat/inode.c
>> index 655dcc7..12a37a9 100644
>> --- a/fs/fat/inode.c
>> +++ b/fs/fat/inode.c
>> @@ -61,15 +61,26 @@ static inline int __fat_get_block(struct inode *inode,
>> sector_t iblock,
>>  	struct super_block *sb = inode->i_sb;
>>  	struct msdos_sb_info *sbi = MSDOS_SB(sb);
>>  	unsigned long mapped_blocks;
>> -	sector_t phys;
>> +	sector_t phys, last_block, disk_block;
>>  	int err, offset;
>> +	const unsigned long blocksize = sb->s_blocksize;
>> +	const unsigned char blocksize_bits = sb->s_blocksize_bits;
>>
>>  	err = fat_bmap(inode, iblock, &phys, &mapped_blocks, create);
>>  	if (err)
>>  		return err;
>>  	if (phys) {
>> -		map_bh(bh_result, sb, phys);
>>  		*max_blocks = min(mapped_blocks, *max_blocks);
>> +		last_block = (MSDOS_I(inode)->mmu_private + (blocksize - 1))
>> +			>> blocksize_bits;
>> +		disk_block = (MSDOS_I(inode)->i_disksize + (blocksize - 1))
>> +			>> blocksize_bits;
>> +		if (iblock >= last_block && iblock <= disk_block) {
>> +			MSDOS_I(inode)->mmu_private +=
>> +				*max_blocks << blocksize_bits;
>> +			set_buffer_new(bh_result);
>> +		}
>> +		map_bh(bh_result, sb, phys);
>>  		return 0;
>>  	}
>>  	if (!create)
>
> Let's give the name for this operation for readability, and make it
> static function. With this, I think this one much readable.
>
> static void check_fallocate_region()
> {
> 	/* code here */
> }
Okay, I will make it.
Thanks for review.
> --
> OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
>