Garbage returned in nanosecond component of timestamps
[Fedora 14 with vendor kernel 2.6.35.6-48.fc14.i686.PAE]
When nanosecond timestamp resolution isn't supported on an ext4 partition (inode size = 128), stat() appears to be returning uninitialized garbage in the nanosecond component of timestamps.
See the following tests.
Expected: All timestamps should end with '.0000000000'.
Actual: '.0026721629' popping up on a different file each time.
Quoted 9 lines
# tune2fs -l /dev/md0 | fgrep 'Inode size' Inode size: 128 # echo 3 >/proc/sys/vm/drop_caches # find / -xdev -type f -printf '%t - %p\n' | fgrep -v '.0000000000' Wed Nov 4 16:53:20.0026721629 2009 - /boot/oldconfigs/config-2.6.27.38-170.2.113.fc10.i686.PAE # echo 3 >/proc/sys/vm/drop_caches # find / -xdev -type f -printf '%t - %p\n' | fgrep -v '.0000000000' Thu Nov 8 20:38:30.0026721629 2007 - /boot/oldconfigs/config-2.6.23.1-49.fc8PAE # echo 3 >/proc/sys/vm/drop_caches
Quoted 8 lines
# find / -xdev -type f -printf '%t - %p\n' | fgrep -v '.0000000000' Thu Aug 12 17:44:51.0026721629 2010 - /lib/firmware/emi26/bitstream.fw # echo 3 >/proc/sys/vm/drop_caches # find / -xdev -type f -printf '%t - %p\n' | fgrep -v '.0000000000' Thu Aug 12 17:44:51.0026721629 2010 - /lib/firmware/slicoss/gbdownload.sys # echo 3 >/proc/sys/vm/drop_caches # find / -xdev -type f -printf '%t - %p\n' | fgrep -v '.0000000000' Fri Nov 5 01:16:32.0026721629 2010 - /boot/grub/e2fs_stage1_5
Shouldn't EXT4_INODE_GET_XTIME zero out tv_nsec when EXT4_FITS_IN_INODE evaluates to false...?
Quoted 7 lines
#define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode) \ do { \ (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime); \ if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \ ext4_decode_extra_time(&(inode)->xtime, \ raw_inode->xtime ## _extra); \} while (0)
Signature · 1 line
Jordan Russell
Quoted 29 lines · Jordan Russell
[Fedora 14 with vendor kernel 2.6.35.6-48.fc14.i686.PAE]
When nanosecond timestamp resolution isn't supported on an ext4 partition (inode size = 128),
stat()appears to be returning uninitialized garbage in the nanosecond component of timestamps.See the following tests.
Expected: All timestamps should end with '.0000000000'.
Actual: '.0026721629' popping up on a different file each time.Quoted 9 lines
# tune2fs -l /dev/md0 | fgrep 'Inode size' Inode size: 128 # echo 3 >/proc/sys/vm/drop_caches # find / -xdev -type f -printf '%t - %p\n' | fgrep -v '.0000000000' Wed Nov 4 16:53:20.0026721629 2009 - /boot/oldconfigs/config-2.6.27.38-170.2.113.fc10.i686.PAE # echo 3 >/proc/sys/vm/drop_caches # find / -xdev -type f -printf '%t - %p\n' | fgrep -v '.0000000000' Thu Nov 8 20:38:30.0026721629 2007 - /boot/oldconfigs/config-2.6.23.1-49.fc8PAE # echo 3 >/proc/sys/vm/drop_cachesQuoted 8 lines
# find / -xdev -type f -printf '%t - %p\n' | fgrep -v '.0000000000' Thu Aug 12 17:44:51.0026721629 2010 - /lib/firmware/emi26/bitstream.fw # echo 3 >/proc/sys/vm/drop_caches # find / -xdev -type f -printf '%t - %p\n' | fgrep -v '.0000000000' Thu Aug 12 17:44:51.0026721629 2010 - /lib/firmware/slicoss/gbdownload.sys # echo 3 >/proc/sys/vm/drop_caches # find / -xdev -type f -printf '%t - %p\n' | fgrep -v '.0000000000' Fri Nov 5 01:16:32.0026721629 2010 - /boot/grub/e2fs_stage1_5
Quoted 2 lines
Shouldn't
EXT4_INODE_GET_XTIMEzero outtv_nsecwhenEXT4_FITS_IN_INODEevaluates to false...?
Yep I think so, good catch...
-Eric
Quoted 8 lines
Quoted 7 lines
#define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode) \ do { \ (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime); \ if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \ ext4_decode_extra_time(&(inode)->xtime, \ raw_inode->xtime ## _extra); \} while (0)
PATCH 1/1 ext4: zero out nanosecond timestamps for small inodes
In reply to Jordan RussellAs reported by Jordan Russell:
When nanosecond timestamp resolution isn't supported on an ext4 partition (inode size = 128), stat() appears to be returning uninitialized garbage in the nanosecond component of timestamps.
EXT4_INODE_GET_XTIME should zero out tv_nsec when EXT4_FITS_IN_INODE evaluates to false.
---
fs/ext4/ext4.h +4 -0
@@ -709,6 +709,8 @@ do { \ if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \ext4_decode_extra_time (&(inode)->xtime, \ raw_inode->xtime ## _extra); \ +else \ + (inode)->xtime.tv_nsec =0 ; \ }while (0 ) #define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode) \
@@ -719,6 +721,8 @@ do { \ if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \ext4_decode_extra_time (&(einode)->xtime, \ raw_inode->xtime ## _extra); \ +else \ + (einode)->xtime.tv_nsec =0 ; \ } while (0)#define i_disk_version osd1.linux1.l_i_version
Ted Ts'o
maintainerRe: [PATCH] ext4: zero out nanosecond timestamps for small inodes
Quoted 11 lines · Eric Sandeen
As reported by Jordan Russell:
When nanosecond timestamp resolution isn't supported on an ext4 partition (inode size = 128),
stat()appears to be returning uninitialized garbage in the nanosecond component of timestamps.
EXT4_INODE_GET_XTIMEshould zero outtv_nsecwhenEXT4_FITS_IN_INODEevaluates to false.Reported-by Jordan Russell <[email protected]>Signed-of-by Eric Sandeen <[email protected]>
Thanks, added to the ext4 patch queue.
- Ted