linux.git/fs/fat, branch v2.6.16

[PATCH] fat: Fix truncate() write ordering

2006-02-03T16:32:10+00:00

The truncate() should write the file size before writing the new EOF entry.
This patch fixes it.

This bug was pointed out by Machida Hiroyuki.

Signed-off-by: OGAWA Hirofumi 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] fat: Replace an own implementation with ll_rw_block(SWRITE,)

2006-02-03T16:32:10+00:00

This patch replaces an own implementation with LL_RW_BLOCK(SWRITE,) which was
newly added.

Signed-off-by: OGAWA Hirofumi 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] capable/capability.h (fs/)

2006-01-12T02:42:13+00:00

fs: Use  where capable() is used.

Signed-off-by: Randy Dunlap 
Acked-by: Tim Schmielau 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] mutex subsystem, semaphore to mutex: VFS, ->i_sem

2006-01-09T23:59:24+00:00

This patch converts the inode semaphore to a mutex. I have tested it on
XFS and compiled as much as one can consider on an ia64. Anyway your
luck with it might be different.

Modified-by: Ingo Molnar 

(finished the conversion)

Signed-off-by: Jes Sorensen 
Signed-off-by: Ingo Molnar

[PATCH] fat: support a truncate() for expanding size (generic_cont_expand)

2006-01-09T04:13:47+00:00

This patch changes generic_cont_expand(), in order to share the code
with fatfs.

  - Use vmtruncate() if ->prepare_write() returns a error.

Even if ->prepare_write() returns an error, it may already have added some
blocks.  So, this truncates blocks outside of ->i_size by vmtruncate().

  - Add generic_cont_expand_simple().

The generic_cont_expand_simple() assumes that ->prepare_write() can handle
the block boundary.  With this, we don't need to care the extra byte.

And for expanding a file size by truncate(), fatfs uses the
added generic_cont_expand_simple().

Signed-off-by: OGAWA Hirofumi 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] fat: support ->direct_IO()

2006-01-09T04:13:46+00:00

This patch add to support of ->direct_IO() for mostly read.

The user of this seems to want to use for streaming read.  So, current direct
I/O has limitation, it can only overwrite.  (For write operation, mainly we
need to handle the hole etc..)

Signed-off-by: OGAWA Hirofumi 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] fat: s/EXPORT_SYMBOL/EXPORT_SYMBOL_GPL/

2006-01-09T04:13:46+00:00

All EXPORT_SYMBOL of fatfs is only for vfat/msdos. _GPL would be proper.

Signed-off-by: OGAWA Hirofumi 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] fat: add the read/writepages()

2006-01-09T04:13:46+00:00

Signed-off-by: OGAWA Hirofumi 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] fat: use sb_find_get_block() instead of sb_getblk()

2006-01-09T04:13:46+00:00

We don't need to allocate buffer for checking the buffer is uptodate.  This
use sb_find_get_block() instead, and if it returns NULL it's not uptodate.

Signed-off-by: OGAWA Hirofumi 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] fat: move fat_clusters_flush() to write_super()

2006-01-09T04:13:46+00:00

It is overkill to update the FS_INFO whenever modifying
prev_free/free_clusters, because those are just a hint.

So, this patch uses ->write_super() for updating FS_INFO instead.

Signed-off-by: OGAWA Hirofumi 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds