Hans-Peter Jansen writes:
>
> Hi Nikita,
>
> while diggin' on the topic in the net, I came across
> http://namesys.com/bad-block-handling.html, which
> answers some of my questions, but generates new ones:
>
> 1.)
> What about giving a hint how to gather the blocksize?
> It's the commonly used 4096, I guess.
Block size is recorded in a super-block. You can get it by running
/bin/debugreiserfs
>
> 4. & 5.)
> How about loosing a word whats happening behind the scenes
> here? Right now, I interpret 4. as the necessary step to
> mark some blocks bad. But why marking the block as "used"
> first, and as free before running reiserfsck, then. Does
> this mean, reiserfsck cannot handle bad blocks yet, and I
> have to restore them afterwards, or what?
Yes, you are right, reiserfsck doesn't handle bad-blocks yet.
>
> What does the third parameter "set" in add-bad-block.c?
When you manually mark block as "used" in bitmap, special counter in
super-block is incremented (this is supposed to be used by reiserfsck in
a future). "set" option allows one to set this counter to given value.
> What about the foo file? What happens, if I remove it?
Nothing. This file is just a stub.
>
> My problem is, the fs is populated already, and I've found
> this in the logs today:
> Nov 8 13:51:48 shrek kernel: end_request: I/O error,
> dev 03:09 (hda), sector 40027056
> Nov 8 13:52:43 shrek kernel: hda: dma_intr: status=0x51
> { DriveReady SeekComplete Error }
> Nov 8 13:52:43 shrek kernel: hda: dma_intr: error=0x40
> { UncorrectableError }, LBAsect=85282224
>
> I was able to isolate two defect sectors on this hd with:
> dd if=/dev/hda bs=512 count=1 skip=85282224
> dd if=/dev/hda bs=512 count=1 skip=85282225
> giving I/O errors.
>
> It's a IBM-DTLA-307060. I've patched/compiled my 2.4.13-ac7
> with linux-2.4.8-reiser-badblocks-1.b.diff already, and
> will wait for any hopefully enlighting reply before messing
> with add-bad-block.
It looks like you are having bad-blocks already allocated for files or
meta-data.
You should perform the following:
- full backup (dd is inadequate for this, as it doesn't do seek on
output device when read from input device fails).
- /sbin/reiserfsck --check /device. If this fails you have got
bad-blocks in meta-data. If so, copy data from /device to
/good-device and run /sbin/reiserfsck --rebuild-tree /good-device.
If /sbin/reiserfsck --check /device didn't fail, find what files
contain corrupted blocks (by catting all files to /dev/null for
example), remove them, and mark bad-blocks as used.
>
> Thanks in advance,
> Hans-Peter
Nikita.
