Messages in this thread | | | Date | Tue, 14 Oct 2003 11:19:20 +0100 | From | John Bradford <> | Subject | Re: Why are bad disk sectors numbered strangely, and what happens to them? |
| |
(I know I said my previous post was the last one on this subject, but we seem to have moved on to a slightly different area).
Quote from Hans Reiser <reiser@namesys.com>: > Perhaps we should tell people to first write to the bad block, and only > if the block remains bad after triggering the remapping by writing to it > should you make any effort to get the filesystem to remap it for you. > What do you think?
I'm not convinced that this belongs in the filesystem. I can see how it makes sense in some ways for magnetic disk devices, but that's not the filesystem's concern. How would we know that the write isn't being cached by hardware further along the line, for example? What are the negative effects of repeated writes if the filesystem is on flash, or a tape. A damaged tape could be damaged more by winding back and forth, for example, (OK, tape is a bad example, but some future storage technology that we don't know about could have an analogous problem. My point is that just because 99% of installations will use ReiserFS on disk device, is it right to put disk device specifics in the FS?).
Also, one corner case that occurs to me is that the first remapping worked, and then the newly allocated area went bad in the time before we verified it. Then it could look like a persistant fault, when it is infact it's two separate faults. Realistically, though, I suspect that is only likely to happen on a rapidly dieing disk, in which case there isn't much we can do anyway.
In general, though, the question is really, should ReiserFS be usable on a device which doesn't do it's own bad block handling? I suggest no.
The ultimate point is that only the drive firmware really knows what's going on, and it can make informed decisions based on things that nothing external to the drive knows about. How much error correction it needed to read a block, the number of errors per physical head, and per physical cylinder, etc. The filesystem can only generally make a decision based on whether there is an error or not.
> Rogier has not indicated that he has tried writing to the bad sector, > has he?
I don't think so.
John. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |