live session from usb key take very long to start (40 min.) due to probing of non-existent floppy

From Kern.log:

Mar 21 21:57:22 ubuntu kernel: [ 20.788412] end_request: I/O error, dev fd0, sector 0
Mar 21 21:57:22 ubuntu kernel: [ 32.869106] end_request: I/O error, dev fd0, sector 0
Mar 21 21:57:22 ubuntu kernel: [ 32.869109] Buffer I/O error on device fd0, logical block 0

Mar 21 21:57:22 ubuntu kernel: [ 2209.279269] end_request: I/O error, dev fd0, sector 0
Mar 21 21:57:22 ubuntu kernel: [ 2209.279271] Buffer I/O error on device fd0, logical block 0

Look at the times.

I confirm having the same problem with Natty Beta1 the iso 20110328

I confirm having the same problem with Natty Beta1 iso 20110329.1

I'm having the same exact problem with the iso 20110330 of Natty Beta 1 (testing)

Thanks for you report.

Is there of floppy drive in this system ?

On 03/31/2011 09:34 PM, Jean-Baptiste Lallement wrote:
> Thanks for you report.
you're welcome !

> Is there of floppy drive in this system ?
>
>

No I do not have a floppy but I have a card reader/writer.

--
Carla Sella
email: <email address hidden>
https://launchpad.net/~carla-sella

This sounds reminiscent of bug 539515. Carla, can you go into your BIOS (when you boot you'll see something like, "Press F2 for options.") and see if you can disable the floppy controller? Do let us know whether or not that fixes the issue.

Thanks!

On 04/01/2011 02:04 PM, Evan Dandrea wrote:
> This sounds reminiscent of bug 539515. Carla, can you go into your BIOS
> (when you boot you'll see something like, "Press F2 for options.") and
> see if you can disable the floppy controller? Do let us know whether or
> not that fixes the issue.
>
> Thanks!
>
Hello Evan,
It worked ! :)
I entered my BIOS and disabled the entry "Legacy Diskette A", rebooted
with the usb persistent key and my PC started within a few seconds.

--
Carla Sella
email: <email address hidden>
https://launchpad.net/~carla-sella

Hello Evan,

we found this bug during the ISO Testing (another guy on the Italia testing team mailing list said to have reproduced it). As far as I can see it's a regression because bug #539515 claim to be fixed in Maverick. What do you recommend, should it be assigned to udisks package?

I finally found an old computer which still has a floppy controller (I don't have an actual floppy anywhere, but having one wouldn't expose this bug at all). I can reproduce the problem there.

Nothing in udisks changed since maverick, and as the problem only occurs on the live CD, with persistency, then the bug is in casper somewhere. I debugged this a bit and indeed found that find_cow_device() iterates over all block devices and runs blkid and other tests on them; this will time out badly.

In udev and udisks we set up the rules to avoid touching a floppy device automatically at all costs, for this very reason. casper must do the same. I figure it wouldn't be very useful to have a persistency file on a floppy disk anyway.

I definitively found one place where this wreaks havoc, and I'll upload a fix for that in a minute. It remains to be seen whether this is the only place where things go wrong, but as the scripts are very hard to test without a cow (chicken-egg problem), I'll upload this and test tomorrow's daily CD.

This bug was fixed in the package casper - 1.263

---------------
casper (1.263) natty; urgency=low

  [ Evan Dandrea ]
  * Suppress Log Out item in the live session (LP: #750140).

  [ Martin Pitt ]
  * scripts/casper-helpers, find_cow_device(): Do not ever probe floppy
    devices. There are lots of computers which still have a floppy controller
    (and thus a /dev/fd0), but no actual floppy. Trying to send anything to
    /dev/fd0 will result in very long timeouts. Here this causes booting the
    live CD to take more than 30 minutes. (See also LP #539515 for the
    corresponding bug in udev/udisks). (LP: #739774)
 -- Martin Pitt <email address hidden> Mon, 04 Apr 2011 16:12:34 +0200

I tested today's desktop image, and it's a lot better now. I still get about three floppy accesses from somewhere, but now the live system starts within two or three minutes.

I tested today.s live 64 bit desktop image with and without the floppy controller enabled and the system started in both cases in 2-3 minutes.
So everything's ok.
Thank you.
Carla