[Clonezilla-live] Cloning FreeBSD partitons: DMA disk error

[Marcin C. <sa...@sa...>]

Hello,

I recently gave clonezilla-live-20100921-lucid.iso on my Vista/FreeBSD laptop.
I am booting it from CD. 

I wanted to clone my internal 114473MB SATA drive (that has 3 bad sectors):

ad4: 114473MB <FUJITSU MHW2120BH 00000012> at ata2-master UDMA100 SATA

to the external one (connected by Firewire or USB)

da2: 131071MB (268435455 512 byte sectors: 255H 63S/T 16709C)

Here's the layout for the source disk as shown by FreeBSD's gpart utility:


Script started on Sat Oct 30 23:46:11 2010
radziecki$ gpart list ad4
Geom name: ad4
fwheads: 16
fwsectors: 63
last: 234441647
first: 63
entries: 4
scheme: MBR
Providers:
1. Name: ad4s1
   Mediasize: 83575296 (80M)
   Sectorsize: 512
   Mode: r0w0e0
   rawtype: 4
   length: 83575296
   offset: 32256
   type: !4
   index: 1
   end: 163295
   start: 63
2. Name: ad4s2
   Mediasize: 40000028672 (37G)
   Sectorsize: 512
   Mode: r0w0e0
   rawtype: 7
   length: 40000028672
   offset: 10000269312
   type: ntfs
   index: 2
   end: 97656831
   start: 19531776
3. Name: ad4s3
   Mediasize: 70031213056 (65G)
   Sectorsize: 512
   Mode: r3w3e5
   attrib: active
   rawtype: 165
   length: 70031213056
   offset: 50000297984
   type: freebsd
   index: 3
   end: 234436544
   start: 97656832
4. Name: ad4s4
   Mediasize: 9916268544 (9.2G)
   Sectorsize: 512
   Mode: r1w1e2
   rawtype: 165
   length: 9916268544
   offset: 83607552
   type: freebsd
   index: 4
   end: 19531007
   start: 163296
Consumers:
1. Name: ad4
   Mediasize: 120034123776 (112G)
   Sectorsize: 512
   Mode: r4w4e11

radziecki$ gpart list ad4s3
Geom name: ad4s3
fwheads: 16
fwsectors: 63
last: 136779712
first: 0
entries: 8
scheme: BSD
Providers:
1. Name: ad4s3a
   Mediasize: 536870912 (512M)
   Sectorsize: 512
   Mode: r1w1e1
   rawtype: 7
   length: 536870912
   offset: 0
   type: freebsd-ufs
   index: 1
   end: 1048575
   start: 0
2. Name: ad4s3b
   Mediasize: 536870912 (512M)
   Sectorsize: 512
   Mode: r1w1e0
   rawtype: 1
   length: 536870912
   offset: 536870912
   type: freebsd-swap
   index: 2
   end: 2097151
   start: 1048576
3. Name: ad4s3d
   Mediasize: 68957471232 (64G)
   Sectorsize: 512
   Mode: r1w1e1
   rawtype: 7
   length: 68957471232
   offset: 1073741824
   type: freebsd-ufs
   index: 4
   end: 136779712
   start: 2097152
Consumers:
1. Name: ad4s3
   Mediasize: 70031213056 (65G)
   Sectorsize: 512
   Mode: r3w3e5

radziecki$ gpart list ad4s4
Geom name: ad4s4
fwheads: 16
fwsectors: 63
last: 19367711
first: 0
entries: 8
scheme: BSD
Providers:
1. Name: ad4s4b
   Mediasize: 2621440000 (2.4G)
   Sectorsize: 512
   Mode: r0w0e0
   rawtype: 1
   length: 2621440000
   offset: 65536
   type: freebsd-swap
   index: 2
   end: 5120127
   start: 128
2. Name: ad4s4e
   Mediasize: 7294763008 (6.8G)
   Sectorsize: 512
   Mode: r1w1e1
   rawtype: 7
   length: 7294763008
   offset: 2621505536
   type: freebsd-ufs
   index: 5
   end: 19367711
   start: 5120128
Consumers:
1. Name: ad4s4
   Mediasize: 9916268544 (9.2G)
   Sectorsize: 512
   Mode: r1w1e2

So we have DOS FAT, NTFS, FreeBSD UFS and ZFS there, plus
two swap partitions (ad4s4b and ad4s4e are not really used).

My first problem during first two runs of a full
disk-to-disk copy with clonezilla was that clone
aborted with the message that Linux could not find /dev/sdb5
partition (/dev/sda was the source and /dev/sdb was the target).

Looks like all MBR partitions (/dev/sdb1..sdb4) got recognized
by Linux on a target, but sdb5..sdb8 were missing, although
I am sure I requested recreation of the partition table according
to the source.

Things got better after 2nd or 3rd attempt and I am not seeing this
anymore (maybe I will try to reproduce this later again). 

But now I am facing a problem related to the disk errors:
they are now harmless since ZFS worked them around, but they
still cause problems with imaging software.

I was using Ghost 2003 for DOS previously and it was enough
to tell it to ignore checksums. Ghost is very slow on this 
machine (it uses INT 13h access since the USB drive is not 
recognized by DOS ASPI drivers) so it takes ages to clone
few GBs, but it works. 

What I am seeing with clonezilla (no TUI/GUI mode, verbose) is this
(taken from /var/log/messages, not from the screen where it's
different, well):

t 30 13:32:01 lucid kernel: [ 7333.292762] sd 2:0:0:0: [sda] Unhandled sense code
Oct 30 13:32:01 lucid kernel: [ 7333.292766] sd 2:0:0:0: [sda] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
Oct 30 13:32:01 lucid kernel: [ 7333.292773] sd 2:0:0:0: [sda] Sense Key : Medium Error [current] [descriptor]
Oct 30 13:32:01 lucid kernel: [ 7333.292782] Descriptor sense data with sense descriptors (in hex):
Oct 30 13:32:01 lucid kernel: [ 7333.292787]         72 03 13 00 00 00 00 0c 00 0a 80 00 00 00 00 00 
Oct 30 13:32:01 lucid kernel: [ 7333.292809]         06 66 85 c6 
Oct 30 13:32:01 lucid kernel: [ 7333.292818] sd 2:0:0:0: [sda] Add. Sense: Address mark not found for data field
Oct 30 13:32:01 lucid kernel: [ 7333.292828] sd 2:0:0:0: [sda] CDB: Read(10): 28 00 06 66 85 c6 00 00 02 00

I did two runs today:

(1) http://saper.info/files/ocs/ocs_onthefly_local.WRMtsE/ 
is a partition-to-partition copy, the command is here:

http://saper.info/files/ocs/ocs-onthefly-2010-10-30-11-34

(2) http://saper.info/files/ocs/ocs_onthefly_local.xdkVmV/
is a disk-to-disk copy, the exact command is here:

http://saper.info/files/ocs/ocs-onthefly-2010-10-30-12-13

Unfortunately, screen output does not get logged into /tmp
files, I attach a full /var/log/messages dump from those
two runs (they were executed by restarting clonezilla
and not rebooting):

http://saper.info/files/ocs/messages

If there is some way to have the clonezilla console 
output logged to a file, please advise (except
from starting the whole thing from command line of course...)

Those DMA/disk error seem to halt the replication process
forever - I get them over and over (one night was not enough
to push the process further). 

Is there any way to skip over bad blocks? (as you
see, I tried the "-rescue" option).

//Marcin