Help me before I die again. (bad sector stuff)

[Chris Torek]

In article <2390 at mandrill.CWRU.Edu> edguer at mandrill (Apostle of Zeus) writes:
>	up2c: hard error sn38696 cn=241 tn=4 sn=9
>I decided to mark the sector as bad using bad144(8).  
>	bad144 rm03 up2
>returned a two page list - 126 entries.  The entries were not in
>order, and of course there is only room for 126 entries in the bad
>sector list.  Two strikes.
>From speaking with Keith Bostick, it would appear that something
>overwrote the bad sector table.  So I decided to simply create a new
>one.

(Bostic: no `k'.)  Most likely, your file system uses the whole area in
up2c.  Since the C partition in the `up' driver includes the bad sector
table, this is a bad thing.  If you have not changed the size of your
file system, you may lose that area again.  You can check the size of
your file system with `dumpfs' (multiply `blocks' times `fsize/512'
to get number of sectors).  Your best bet is not to use the last cylinder
of such a device.  (The same goes for `hp' and `rk' disks.  `ra' disks
are generally safe; the bad sector area is beyond the reported end of
media.  The distributed RA60 or RA80 tables are, however, wrong.  I
forget which it is; probably RA60.)

>According to bad144(8)
>	"The -f option may be used to mark the new bad sectors as 'bad'"

... but only if the device driver supports it ...

>So I tried
>	bad144 -f rm03 up2 1752457552 38696
>It didn't work.  Instead I received the error message:
>	ioctl: no such device

... and apparently the up driver does not.

>... I double checked the table and 38696 was still in the table.
>SO, HOW DO I MARK A SECTOR BAD ??????

Rewrite the sector header.  If the driver will not let you do it,
you will have to reformat (or change the driver).  Some disks do
not allow rewriting individual sector headers, making this particularly
hard.

>Since bad144 wasn't helping me I thought I would try using badsect so...
>	mount up2c /usr2
>	mkdir /usr2/BAD
>	cd /usr2
>	badsect /usr2/BAD 38696
>Instead of solving my problem I got back an error message
>	block 38696 in non-data area: cannot attach
>Any ideas on what that means?

That means block 38696 is in one of the inode or cylinder group areas,
and cannot be allocated as a file block.

>My eventual solution was to reformat the disk.

Probably the easiest fix.  Just make sure your file system does not
overlap the bad sector table!  It is generally best not to use the
`c' partitions for file systems; when we redid the partitioning on
our two large Vaxen, I made `h' partitions for `all but the bad
sector table' file systems:

brillig% disklabel hp4		# (this is a 4.3BSD-tahoe command)
# /dev/rhp4c:
type: SMD
disk: eagle
label: 
flags: badsect
bytes/sector: 512
sectors/track: 48
tracks/cylinder: 20
sectors/cylinder: 960
cylinders: 842
rpm: 3961
interleave: 1
trackskew: 0
cylinderskew: 0
headswitch: 0		# milliseconds
track-to-track seek: 0	# milliseconds [these are not used yet anyway -Chris]
drivedata: 0 3 8 15 

8 partitions:
#        size   offset    fstype   [fsize bsize   cpg]
  a:    63360        0    unused      512  4096       	# (Cyl.    0 - 65)
  b:    63360    63360      swap                    	# (Cyl.   66 - 131)
  c:   808320        0    unused      512  4096       	# (Cyl.    0 - 841)
  d:   680640   126720    unused      512  4096       	# (Cyl.  132 - 840)
  h:   807360        0    4.2BSD     1024  8192    16 	# (Cyl.    0 - 840)
-- 
In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7163)
Domain:	chris at mimsy.umd.edu	Path:	uunet!mimsy!chris