Messages in this thread | | | Date | Sun, 6 Jun 1999 13:20:59 +0200 | From | Andi Kleen <> | Subject | Filesystem corruptions with 2.3.4 |
| |
Hallo Linus,
Alan told me to send 2.3 ext2 corruption reports to you.
I see filesystem corruptions with 2.3.4/i386/SMP/egcs 1.1.2-release compiled.
For example my kernel patch directory (a few hundred files) disappeared after I copied it around with cp -a from another fs; now I have an empty directory that I cannot delete with rmdir because it has a link count of 17. I tried to reset the link count to 3 with debugfs, but an fsck afterwards changed it back to 17:
drwxrwxr-x 17 andi andi 8192 May 19 11:01 STRANGE-FILE
debugfs: stat src/STRANGE-FILE Inode: 118773 Type: directory Mode: 0775 Flags: 0x0 Version: 1 User: 500 Group: 500 Size: 8192 File ACL: 0 Directory ACL: 0 Links: 17 Blockcount: 16 Fragment: Address: 0 Number: 0 Size: 0 ctime: 0x3751a10b -- Sun May 30 22:35:23 1999 atime: 0x375a566a -- Sun Jun 6 13:07:22 1999 mtime: 0x37427e03 -- Wed May 19 11:01:55 1999 BLOCKS: 954669 960720 TOTAL: 2
Another occurrence is that my kernel compiles die. I checked out a fresh kernel tree from a CVS server and started a -j4 compile. It died then in the make depend phase: it seems the drivers/net was completely missing (I'm pretty sure that the CVS checkout worked correctly)
Here the same effect:
kali% ls -l net total 0 kali% ls -ld net drwxrwxr-x 6 andi andi 4096 Jun 6 02:03 net kali% rmdir net rmdir: net: Directory not empty Inode: 80957 Type: directory Mode: 0775 Flags: 0x0 Version: 1 User: 500 Group: 500 Size: 4096 File ACL: 0 Directory ACL: 0 Links: 6 Blockcount: 8 Fragment: Address: 0 Number: 0 Size: 0 ctime: 0x3759bae5 -- Sun Jun 6 02:03:49 1999 atime: 0x375a57d0 -- Sun Jun 6 13:13:20 1999 mtime: 0x3759bae5 -- Sun Jun 6 02:03:49 1999 BLOCKS: 659758 TOTAL: 1
Again the high link count. Something is fishy.
I use ext2 file systems with 4K blocks on IDE disks; using extended IDE with the ALI15x3 dma driver. Haven't seen any corruption on 1K ext2s yet (so perhaps the 4K block size makes the block more likely to happen?)
-Andi
-- This is like TV. I don't like TV.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |