CD media burnt using TAO mode fails to boot

CD media burnt using TAO mode fails to boot

[Alba Pompeo]

Hello,

there is a recurrent problem bug in the kernel that makes a CD-R fail
to boot when burnt in TAO mode.

[    9.698776] sr 1:0:0:0: [sr0] tag#12 UNKNOWN(0x2003) Result:
hostbyte=0x00 driverbyte=0x08
[    9.698784] sr 1:0:0:0: [sr0] tag#12 Sense Key : 0x4 [current]
[    9.698788] sr 1:0:0:0: [sr0] tag#12 ASC=0x3e ASCQ=0x2
[    9.698795] sr 1:0:0:0: [sr0] tag#12 CDB: opcode=0x28 28 00 00 00
ae 96 00 00 02 00
[    9.698798] blk_update_request: I/O error, dev sr0, sector 178776

My best guess would be that this confuses the entity among the ISO
payload, which is in charge for creating /dev/disk/by-label/ links
while the system boots up.

I posted about this on bugzilla too -
https://bugzilla.kernel.org/show_bug.cgi?id=118621

If any more information is needed, let me know.
Thank you.

Re: CD media burnt using TAO mode fails to boot

[Thomas Schmitt]

Hi,

(I am currently not subscribed to linux-scsi but will do if a
 discussion arises. Just give me a note by direct mail.)

Alba Pompeo wrote:
> [    9.698784] sr 1:0:0:0: [sr0] tag#12 Sense Key : 0x4 [current]
> [    9.698788] sr 1:0:0:0: [sr0] tag#12 ASC=0x3e ASCQ=0x2
> [    9.698795] sr 1:0:0:0: [sr0] tag#12 CDB: opcode=0x28 28 00 00 00 ae 96 00 00 02 00

  4 3E 02 TIMEOUT ON LOGICAL UNIT

Quite a harsh reaction for an attempt to read the TAO roun-out blocks.
Normal would be rather something like

  5 30 02 CANNOT READ MEDIUM  INCOMPATIBLE FORMAT
  5 64 00 ILLEGAL MODE FOR THIS TRACK


> My best guess would be that this confuses the entity among the ISO
> payload, which is in charge for creating /dev/disk/by-label/ links
> while the system boots up.

At least with Archlinux it is/was the program blkid.

A few weeks ago i posted to linux-scsi 

  Date: Sun, 28 Feb 2016 14:41:59 +0100
  Subject: sr: Failures to read TAO CD track ends
  Message-Id: <6531578277915568387@scdbackup.webframe.org>

  http://marc.info/?l=linux-scsi&m=145666692729714&w=2

with CD drive observations and remarks about drivers/scsi/sr.c and
its strange theory of 75 block fuzziness of READ CAPACITY.

The problem is rather about the last two blocks which are reported
to be part of a TAO track. They need to be handled mistrustingly,
because the CD drives show no uniform behavior with counting the
unreadable run-out blocks as part of readable payload or not.

As of april 19 2016 the problem gets circumvented in upstream
libblkid:
  https://github.com/karelzak/util-linux/commit/bfebe74e3babe8c188a51269acc2673f1aea283f
At least for me it solved the failure of blkid on a TAO CD.

So it might be worth to inform the publishers of CD-sized ISOs that
they can close the pitfall by updating the blkid program which they
put into their ISOs.


Have a nice day :)

Thomas