Skip to contentSkip to conversation

inconsistent timestamp update in rename() of xfs/fat/gfs2/ramfs/jffs2...

DISCUSSION Taesoo Kim 2 participants 2 messages Mar 20, 2015 VFS

lore
Share & tools
lore

Taesoo Kim

author
#0 lore raw

Hi all,

We've cross-checking patches from ext3/ext4, and found out inconsistent implementations of other fs. We want to ask whether this is intended or unexpected behavior. We will be able to send patches as soon as confirmed/acknowledged.

Ref.

(ext4) 53b7e9f6807c1274eee19201396b4c2b5f721553
(ext3) 0b23076988b44b2c165e060248345de6f2337387

| ext3/4: fix update of mtime and ctime on rename

---------------------------------------------------------------------
We summarized our finding:
(* means what we believe is correct beahvoir)

<Linux 4.0-rc2>
ramfs affs fsplus
vfs xfs fat gfs2 jffs2 hfsh
operation | * | | | | | | | | |
===========================================================

new_inode->i_ctime | V | - | - | V | V | - | - | V | - | -
old_inode->i_ctime | V | V | V | - | V | V | - | - | - | -
old_dir->i_ctime   | V | V | V | V | V | V | V | - | V | V
old_dir->i_mtime   | V | V | V | V | V | V | V | - | V | V
new_dir->i_ctime   | V | V | V | V | - | V | V | - | V | V
new_dir->i_mtime   | V | V | V | V | - | V | V | - | V | V
new_dir->i_atime   | - | - | - | V | - | - | - | - | - | -

* Correct behavior in

ext2, ext3, ext4, btrfs, f2fs, nilfs, reiserfs, cifs, exofs, jfs, ubifs, ocfs and minix
---------------------------------------------------------------------

Thanks,
Taesoo

#1 lore raw
Quoted 18 lines · Taesoo Kim

Hi all,

We've cross-checking patches from ext3/ext4, and found out inconsistent implementations of other fs. We want to ask whether this is intended or unexpected behavior. We will be able to send patches as soon as confirmed/acknowledged.

Ref.

(ext4) 53b7e9f6807c1274eee19201396b4c2b5f721553
(ext3) 0b23076988b44b2c165e060248345de6f2337387

| ext3/4: fix update of mtime and ctime on rename

---------------------------------------------------------------------
We summarized our finding:
(* means what we believe is correct beahvoir)

Quoted 6 lines

<Linux 4.0-rc2>
ramfs affs fsplus
vfs xfs fat gfs2 jffs2 hfsh
operation | * | | | | | | | | |
===========================================================
new_inode->i_ctime | V | - | - | V | V | - | - | V | - | -

This timestamp behaviour is undefined by posix, therefore all filesystems are behaving "correctly" according to the POSIX specification regardless of whether this timestamp is updated or not.

http://pubs.opengroup.org/onlinepubs/9699919799/functions/rename.html

....
Upon successful completion, rename() shall mark for update the last data modification and last file status change timestamps of the parent directory of each file.
....
APPLICATION USAGE

Some implementations mark for update the last file status change timestamp of renamed files and some do not. Applications which make use of the last file status change timestamp may behave differently with respect to renamed files unless they are designed to allow for either behavior.
....

Cheers,

Dave.

Signature · 2 lines
Dave Chinner
[email protected]