Messages in this thread | | | From | "Kiniger, Karl (MED)" <> | Date | Wed, 2 Apr 2003 19:41:17 +0200 | Subject | Re: how to interpret ide error messages (2.4) |
| |
On Wed, Apr 02, 2003 at 02:31:09PM +0100, Alan Cox wrote: > On Wed, 2003-04-02 at 13:23, Kiniger, Karl (MED) wrote: > > kernel: hdc: dma_intr: status=0x51 { DriveReady SeekComplete Error } > > kernel: hdc: dma_intr: error=0x01 { AddrMarkNotFound }, LBAsect=20300322, sector=1263288 > > kernel: hdc: dma_intr: status=0x51 { DriveReady SeekComplete Error } > > The drive could not find the requested sector. That normally means bad > things but for some drivers can also mean the controller asked for a > totally bogon sector number > > > kernel: hdc: dma_intr: error=0x40 { UncorrectableError }, LBAsect=20803307, sector=1766272 > > kernel: end_request: I/O error, dev 16:04 (hdc), sector 1766272 > > Unrecoverable data error. > > > The affected sectors dont generate any error messages if I read them today... > > On errors the next write to a bad sector will typically remap it > transparently to another spare block on the disk. Read obviously cannot > do the same. That would mean that if for example clearcase ignored the > I/O error and wrote back what it thought it saw but did not that it may > have recovered the sector with invalid data. Its also possible of course > clearcase actually handles I/O errors properly (which is hard).
What is giving me an alarm signal is:
Since it is a raid1 I expected user space not being affected. (The other drive did not show any error messages since installation, they are Maxtors 6Y120L0 (120 GB) cooled quite well) So I thought that ClearCase should not have seen any error return code.
In the meantime I have just re-synced the faulty drive and there were no write error messages and the bad block seems to be properly re-mapped
> > Consult the clearcase support I guess, there should be tools to verify > your clearcase datasets. You might also want to force an fsck on your
fortunately only one replica packet seems to be corrupted and I can request it again - the database seems ok after a run of the equivalent of fsck.
> file systems while the box is down for disk replacement to check > everything out.
there is still interest what "sector=1766272" actually means, e.g given a partition table and LBAsect, how to calculate 'sector' or vice versa.
Thanks,
Karl
-- Karl Kiniger karl.kiniger@med.ge.com GE Medical Kretztechnik Tiefenbach 15 A-4871 Zipf Tel: (++43) 7682-3800-710 Fax (++43) 7682-3800-47 - 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/
| |