Check Condition on CDB: 0x28

[Thierry Lacoste]

I'm running NetBSD 3.0/sparc64 on a SUN Ultra1.
The machine is more than 10 years old but still doing it's job.
Once in a while I have the following errors (not very often as
you can see).

What does it mean? Should I worry?
It this my disk wearing out?

Regards,
Thierry.

May 23 03:15:57 vega /netbsd: sd0(esp0:0:0:0): Check Condition on CDB: 0x28
00 00 31 5f 84 00 00 04 00
May 23 03:15:57 vega /netbsd: SENSE KEY: Recovered Error
May 23 03:15:57 vega /netbsd: INFO FIELD: 3235717
May 23 03:15:57 vega /netbsd: ASC/ASCQ: ID CRC Or ECC Error
May 23 03:15:57 vega /netbsd: FRU CODE: 0xd7
May 23 03:15:57 vega /netbsd: SKSV: Actual Retry Count: 1

May 31 03:17:53 vega /netbsd: sd0(esp0:0:0:0): Check Condition on CDB: 0x28
00 00 30 a5 20 00 00 20 00
May 31 03:17:53 vega /netbsd: SENSE KEY: Recovered Error
May 31 03:17:53 vega /netbsd: INFO FIELD: 3188021
May 31 03:17:53 vega /netbsd: ASC/ASCQ: ID CRC Or ECC Error
May 31 03:17:53 vega /netbsd: FRU CODE: 0xd7
May 31 03:17:53 vega /netbsd: SKSV: Actual Retry Count: 1

Dec 12 03:20:31 vega /netbsd: sd0(esp0:0:0:0): Check Condition on CDB: 0x28
00 00 35 fc 20 00 00 20 00
Dec 12 03:20:31 vega /netbsd: SENSE KEY: Recovered Error
Dec 12 03:20:31 vega /netbsd: INFO FIELD: 3537970
Dec 12 03:20:31 vega /netbsd: ASC/ASCQ: ID CRC Or ECC Error
Dec 12 03:20:31 vega /netbsd: FRU CODE: 0xd7
Dec 12 03:20:31 vega /netbsd: SKSV: Actual Retry Count: 1

[Greg Troxel]

Yes, your disk is starting to have trouble. But, a reread succeeds.
You might try writing over those blocks since they just might not hav
been written for years. But it's probably best to get a new disk.

[Thierry Lacoste]

Post by Greg Troxel
You might try writing over those blocks since they just might not hav
been written for years.

Please excuse my ignorance but how do I do that?

The faulty sectors are given by the INFO FIELD number? Correct?

Regards,
Thierry.

[Greg Troxel]

The faulty sectors are given by the INFO FIELD number? Correct?

I think so, but writing them is very tricky; you'll have to read them
and then write them back, and you should have the fs mounted read only
when you do this.

It would be cool to have a program to read every sector and rewrite it
on a raw partition to freshen up bits on disks.

But my experience is that once you see any sign of flakiness the right
thing to do is to get a new disk and move your bits over before it is
too late.

[Julian Coleman]

Post by Greg Troxel
It would be cool to have a program to read every sector and rewrite it
on a raw partition to freshen up bits on disks.

Won't (for example) `dd if=/dev/rsd0a of=/dev/rsd0a bs=32k conv=notrunc`
do just that? You probably don't need the `conv=notrunc' when using the
raw partition.

J

PS. I haven't tried this on a raw partition, just on a file where it
works perfectly.

[Greg Troxel]

Won't (for example) `dd if=/dev/rsd0a of=/dev/rsd0a bs=32k conv=notrunc`
do just that? You probably don't need the `conv=notrunc' when using the
raw partition.

I suppose it would, but that seems scary. I'd like to see something
that retries the reads, perhaps many times, falling back to single
sectors from blocks, and only rewrites areas that read ok, saving a
log of non-rewritten sectors. I've just never gotten around to
writing this....