Skip to contentSkip to conversation

fat: additions to support fat_fallocate

PATCH Superseded Namjae Jeon 2 participants 3 messages Oct 17, 2014 – Oct 23, 2014 VFS

v6 of 8 Superseded Latest is v8 →

lore
Share & tools
lore
#0 lore raw

v6 cover fat: additions to support fat_fallocate

This patch set provides support for doing fallocate operation on FAT filesystem.

v6:
- Remove usage of i_disksize
- Align all fallocate code sequence to use i_blocks as
this is more or less can be used to denote the on-disksize. Also, i_blocks is common in update path of cluster allocation via. fat_chain_add() both from normal write path and fallocate() operation

v5:
- change all routines to make ->i_disksize aligned by cluster size.
- merge fat_bmap2() and fat_get_block_bmap().

v4:
- ->i_disksize is aligned by block size in fill_inode.
- use i_disksize when computing the number of cluster for fallocate.
- make fat_get_block_bmap to avoid race issue.
- use __fat_write_inode() with inode_needs_sync() in evict.

v3:
- fix i_disksize and mmu_private comments(uninit/initialized)
- remove unneeded condition in fat_bmap.
- use mmu_private instead of fclus in fat_fallocate.
- make function mark buffer new for block in fallocated region.
- use flags for bmap.
- use mmu_private instead of i_size in direct_IO.
- use flags and remove i_mutex in bmap.
- update vfat documentation.

v2:
- remove a stary unlock(from Dan Carpenter)
- check ->mmu_private and ->i_disksize to truncate uninitialized data.

Namjae Jeon (4):
fat: add fat_fallocate operation
fat: skip cluster allocation on fallocated region
fat: permit to return phy block number by fibmap in fallocated region Documentation/filesystems/vfat.txt: update the limitation for fat
fallocate

Signature · 1 line
1.7.9.5
#1 lore raw
Quoted 2 lines · Namjae Jeon

This patch set provides support for doing fallocate operation on FAT filesystem.

I'm still not reviewing deeply though, basically looks like good to me. But I'm still not testing this one at all, sorry.

If not tested yet, we would be able to start test, e.g. stress.

Thanks.

Quoted 37 lines

v6:
- Remove usage of i_disksize
- Align all fallocate code sequence to use i_blocks as
this is more or less can be used to denote the on-disksize. Also, i_blocks is common in update path of cluster allocation via. fat_chain_add() both from normal write path and fallocate() operation

v5:
- change all routines to make ->i_disksize aligned by cluster size.
- merge fat_bmap2() and fat_get_block_bmap().

v4:
- ->i_disksize is aligned by block size in fill_inode.
- use i_disksize when computing the number of cluster for fallocate.
- make fat_get_block_bmap to avoid race issue.
- use __fat_write_inode() with inode_needs_sync() in evict.

v3:
- fix i_disksize and mmu_private comments(uninit/initialized)
- remove unneeded condition in fat_bmap.
- use mmu_private instead of fclus in fat_fallocate.
- make function mark buffer new for block in fallocated region.
- use flags for bmap.
- use mmu_private instead of i_size in direct_IO.
- use flags and remove i_mutex in bmap.
- update vfat documentation.

v2:
- remove a stary unlock(from Dan Carpenter)
- check ->mmu_private and ->i_disksize to truncate uninitialized data.

Namjae Jeon (4):
fat: add fat_fallocate operation
fat: skip cluster allocation on fallocated region
fat: permit to return phy block number by fibmap in fallocated region Documentation/filesystems/vfat.txt: update the limitation for fat
fallocate

Signature · 1 line
OGAWA Hirofumi <[email protected]>
#2 lore raw
Quoted 9 lines
Quoted 2 lines · Namjae Jeon

This patch set provides support for doing fallocate operation on FAT filesystem.

I'm still not reviewing deeply though, basically looks like good to me. But I'm still not testing this one at all, sorry.

If not tested yet, we would be able to start test, e.g. stress.

Hi, OGAWA.
Okay, I will share stress test result(per sector size) using fsx, fsstress, etc. And will fix the review points of 2/4, 3/4 patch also.

Thanks for review!

Quoted 43 lines

Thanks.

Quoted 31 lines

v6:
- Remove usage of i_disksize
- Align all fallocate code sequence to use i_blocks as
this is more or less can be used to denote the on-disksize. Also, i_blocks is common in update path of cluster allocation via. fat_chain_add() both from normal write path and fallocate() operation

v5:
- change all routines to make ->i_disksize aligned by cluster size.
- merge fat_bmap2() and fat_get_block_bmap().

v4:
- ->i_disksize is aligned by block size in fill_inode.
- use i_disksize when computing the number of cluster for fallocate.
- make fat_get_block_bmap to avoid race issue.
- use __fat_write_inode() with inode_needs_sync() in evict.

v3:
- fix i_disksize and mmu_private comments(uninit/initialized)
- remove unneeded condition in fat_bmap.
- use mmu_private instead of fclus in fat_fallocate.
- make function mark buffer new for block in fallocated region.
- use flags for bmap.
- use mmu_private instead of i_size in direct_IO.
- use flags and remove i_mutex in bmap.
- update vfat documentation.

v2:
- remove a stary unlock(from Dan Carpenter)
- check ->mmu_private and ->i_disksize to truncate uninitialized data.

Quoted 6 lines

Namjae Jeon (4):
fat: add fat_fallocate operation
fat: skip cluster allocation on fallocated region
fat: permit to return phy block number by fibmap in fallocated region Documentation/filesystems/vfat.txt: update the limitation for fat
fallocate

--
OGAWA Hirofumi <[email protected]>