Re: reiser4 w. cryptcompress data-corruption (?!)

[Edward Shishkin <edward.shishkin@gmail.com>]

Mat wrote:

>Hi Edward,
>
>I just ran into an (seemingly) data corruption incident:
>
>here the steps to (hopefully) reproduce:
>
>1. boot up system with reiser4-formatted root-partition (cryptcompress lzo1,
>algorithm)
>2. work with it, openoffice, gnome-desktop (I had compiz activated, but I'm sure
>that's only worth a side-note)
>3. plug in an usbstick with a large enough capacity (e.g. 4 GB) or a large
>enough external usb-harddrive, now format it via
>mkfs.ntfs
>
><-- here was the point where I made the "mistake" ;), since I was concerned of
>the usbstick' life (read- and write-cycles are limited so I wanted to forcefully
>quit mkfs.ntfs and re-execute it with mkfs.ntfs -f  after that didn't help and
>the progress number stayed at around 81% I unplugged the usbstick ;)
>
>4. this should lead to a similar error message like the following in dmesg:
>
>[12423.120846] reiser4[mkfs.ntfs(22617)]: dc_check_checksum
>(fs/reiser4/plugin/file/cryptcompress.c:971)[edward-156]:
>[12423.120848] WARNING: Bad disk cluster checksum -1991768807, (should be
>1988809339) Fsck?
>[12423.120850] 
>[12423.121158] reiser4[mkfs.ntfs(22617)]: reiser4_inflate_cluster
>(fs/reiser4/plugin/file/cryptcompress.c:1138)[edward-1460]:
>[12423.121159] WARNING: Inode 75477: disk cluster 0 looks corrupted
>[12423.121307] reiser4[mkfs.ntfs(22617)]: dc_check_checksum
>(fs/reiser4/plugin/file/cryptcompress.c:971)[edward-156]:
>[12423.121308] WARNING: Bad disk cluster checksum -1991768807, (should be
>1988809339) Fsck?
>[12423.121309] 
>[12423.121472] reiser4[mkfs.ntfs(22617)]: reiser4_inflate_cluster
>(fs/reiser4/plugin/file/cryptcompress.c:1138)[edward-1460]:
>[12423.121473] WARNING: Inode 75477: disk cluster 0 looks corrupted
>  
>

Interesting, I'll try to reproduce it...

Btw, it would be nice to figure out (by find -inum 75477)
what is that file (It may shed some light to this corruption).

>5. safe your data & reboot your system with magic sysrq key
>6. I booted into a livecd & checked the reiser4-partition, there was one error
>(corrupted cluster or something, I'm apologize for not having noted it down ;( )
>after that fsck.reiser4 --build-fs
>showed thousands of errors
>  
>

Yes, the kernel doesn't support i_bytes and i_blocks of compressed files
(because of performance reason), but fsck expects this to be set properly.
The option -n suppresses the screams..

>I'm sure it would have fixed the filesystem, however I'm dependent on this box
>right now & need to work, so I played back a tarball with the system & here I am
>  
>

So eventually fsck.reiser4 --build-fs reported that fs is consistent?

>I have to give you guys involed in reiserfs creation a compliment: my /home
>partition suffered from no (0 == zero) data corruption from this incident, where
>other filesystem surely would have corrupted some files (I already had similar
>cases with other filesystems losing almost all data ;( )
>  
>

Thanks for your experience.
Edward.

>I hope you're able to reproduce this error
>
>comments:
>this is the 2nd case of ntfs-related access/contact leading to data corruption,
>I think Microsoft has included a secret data corruption mechanism in ntfs to
>destroy other opensource filesystems when accessed outside of windows ;)
>
>Thanks for your help & work
>
>Mat
>
>-
>To unsubscribe from this list: send the line "unsubscribe reiserfs-devel" in
>the body of a message to majordomo@vger.kernel.org
>More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
>  
>