2013-10-09 14:41:02

by Namjae Jeon

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

From: Namjae Jeon <[email protected]>

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 <[email protected]>
Signed-off-by: Amit Sahrawat <[email protected]>
---
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 37572c2..7dc689e 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 2211489..2f2a265 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

2013-10-27 13:42:05

by OGAWA Hirofumi

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

Namjae Jeon <[email protected]> writes:

> + 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;

Hm, mmu_private is not increased by write_begin()?
--
OGAWA Hirofumi <[email protected]>

2013-10-27 13:45:41

by OGAWA Hirofumi

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

OGAWA Hirofumi <[email protected]> writes:

> Namjae Jeon <[email protected]> writes:
>
>> + 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;
>
> Hm, mmu_private is not increased by write_begin()?

Ah, ok. I understood what this is doing. Please ignore this comment.
--
OGAWA Hirofumi <[email protected]>