• Bug#264870: Initial boot fails (monitor goes off)

    From Joey Hess@1:229/2 to Timo Jyrinki on Tue Aug 10 22:20:04 2004
    XPost: linux.debian.maint.boot
    From: [email protected]

    Timo Jyrinki wrote:
    I wanted to try sarge by swapping out my main hard drive and putting in
    a spare 10GB drive just to see how it's like. Installation was easy
    (despite my need for manually configure network settings) and base installation went fine, but...

    Rebooting didn't finish. The kernel booted, but after a while I just got
    my monitor shut off, and trying to switch between consoles didn't help.
    ctrl-alt-del initiated reboot sequence so I didn't need to do a cold
    boot, but that was it. The same results were had in the "recovery"-mode.
    I guess the installation of other packages (besides the base system)
    should have started (?), but I have no idea what happened after the
    screen went blank.

    I have a Radeon 9600XT graphics adapter, which might be the cause of incompabilities, ie. framebuffer driver crashing the display or
    XFree86's radeon-driver doing the same. Though does Debian use fb and
    does the XFree86 even start at that time (first boot after base install)?

    The last messages I see when booting are:
    1) various USB messages (over one full screen of them)
    2) just before monitor shutting itself, something about
    "xxxx
    net
    pci" or something like that... hotplug things? Those are shown only
    for a fraction of a second, after the USB messages, before the monitor
    goes off.

    It seems that most of the failures people are getting installing rc1 are because hotplug loads unsafe kernel modules, including hardware
    watchdogs, and frame buffer modules that don't work right. It doesn't
    help that as you noticed hotplug brings in all these unnecessary modules
    even in single user mode.

    Your problem sounds like hotplug loading a bad frame buffer module. Try
    booting in emergency mode ("linux emergency"), remount the root
    filesystem read-write ("mount -o remount,rw /"), remove
    /etc/rcS.d/S40hotplug, then you should be able to boot without this
    problem. If you can send us your lspci and lspci -n output, we can
    probably determine which module hotplug is loading to cause your
    problem, and fix it.

    --
    see shy jo

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.2.5 (GNU/Linux)

    iD8DBQFBGSdgd8HHehbQuO8RAqwRAJ9p/EyKeHUbzBqhO+6cPpRm3DQNugCePXFT bre4AS8iW27/pnTeEPhZuF4=
    =rNFh
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)
  • From Timo Jyrinki@1:229/2 to Joey Hess on Tue Aug 10 23:30:14 2004
    From: [email protected]

    On Tuesday 10 August 2004 22:52, Joey Hess wrote:
    It seems that most of the failures people are getting installing rc1 are because hotplug loads unsafe kernel modules, including hardware
    watchdogs, and frame buffer modules that don't work right. It doesn't
    help that as you noticed hotplug brings in all these unnecessary modules
    even in single user mode.

    Your problem sounds like hotplug loading a bad frame buffer module. Try booting in emergency mode ("linux emergency"), remount the root
    filesystem read-write ("mount -o remount,rw /"), remove /etc/rcS.d/S40hotplug, then you should be able to boot without this
    problem. If you can send us your lspci and lspci -n output, we can
    probably determine which module hotplug is loading to cause your
    problem, and fix it.

    Thanks, I'm writing this now on Debian. The removal of S40hotplug allowed
    sarge to boot properly and finish installation. The lspci and lspci -n
    outputs are similar to those I posted from SUSE just before your e-mail, so
    you can check this bug's page for those. The only exception is that USB Controller lines were a bit more verbose in Debian with "VT82xxxxx UHCI USB
    1.1 Controller" included.

    As a side note, booting in emergency mode required not exactly that, it required adding 'emergency' to the grub's kernel line.

    ---

    Additional note regarding installation: exim's configuration, which seemingly comes up also when you just select the desktop packages, is probably odd for many people. It didn't even state what exim is at first, and doesn't say something like "it's safe to select the default here if you don't know what's this all about".

    Other than this, Debian installation went nicely. One rant is that KDE wasn't installed with the Finnish language even though I chose it as the
    installation language, and what's worse Finnish keyboard map was not
    configured in KDE. But this is probably not a place for this.

    -TJ


    --
    To UNSUBSCRIBE, email to [email protected]
    with a subject of "unsubscribe". Trouble? Contact [email protected]

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)
  • From Marco d'Itri@1:229/2 to Timo Jyrinki on Wed Aug 11 01:10:12 2004
    From: [email protected]

    On Aug 10, Timo Jyrinki <[email protected]> wrote:

    On Tuesday 10 August 2004 22:52, Joey Hess wrote:
    Your problem sounds like hotplug loading a bad frame buffer module. Try
    Maybe, but the default configuration will not load any module for the
    display PCI class.

    If you want to help finding which module breaks your system please
    run *AS A USER* "/etc/hotplug/pci.rc start", it should display which
    modules are going to be loaded (and fail loading them, because you are
    not root). Then try to manually load each of them with modprobe and
    report which one crashes your system.

    --
    ciao, |
    Marco | [7498 st7.3aS4sJrB2]


    --
    To UNSUBSCRIBE, email to [email protected]
    with a subject of "unsubscribe". Trouble? Contact [email protected]

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)
  • From Joey Hess@1:229/2 to Marco d'Itri on Wed Aug 11 02:00:20 2004
    From: [email protected]

    Marco d'Itri wrote:
    Maybe, but the default configuration will not load any module for the
    display PCI class.

    I know that's not true, because hotplug on fresh d-i installs has taken
    to loading the matroxfb module for my test laptop, which yields a very
    nice highres console. And when I disable hotplug, it stops being loaded.
    Maybe you have a bug?

    If you want to help finding which module breaks your system please
    run *AS A USER* "/etc/hotplug/pci.rc start", it should display which
    modules are going to be loaded (and fail loading them, because you are
    not root). Then try to manually load each of them with modprobe and
    report which one crashes your system.

    Ah, so that's how to get the list.

    --
    see shy jo

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.2.5 (GNU/Linux)

    iD8DBQFBGV3hd8HHehbQuO8RAgwRAJ0TBuAdU0FCqoj9iM02PP3YetBl3ACgrVyc KZtQ19kTElpgPgQCGDoGVus=
    =5yrF
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)
  • From Marco d'Itri@1:229/2 to Joey Hess on Wed Aug 11 02:30:12 2004
    From: [email protected]

    On Aug 11, Joey Hess <[email protected]> wrote:

    Marco d'Itri wrote:
    Maybe, but the default configuration will not load any module for the display PCI class.
    I know that's not true, because hotplug on fresh d-i installs has taken
    to loading the matroxfb module for my test laptop, which yields a very
    nice highres console. And when I disable hotplug, it stops being loaded. Maybe you have a bug?
    I don't know... Fumitoshi UKAI worked on this. Please attach your /etc/default/hotplug file, it should contain IGNORE_PCI_CLASS_DISPLAY=true.

    --
    ciao, |
    Marco | [7501 plPB419iLJITM]

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.2.5 (GNU/Linux)

    iD8DBQFBGWNuFGfw2OHuP7ERAtAYAKCSxglYRP+G2zo5dKQ+oguAXC/v8wCeMusu 07PJNfE9Zg+O6svAASg7nY8=
    =z3ZJ
    -----END PGP SIGNATURE-----

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)
  • From Fumitoshi UKAI@1:229/2 to Marco d'Itri on Wed Aug 11 06:20:07 2004
    From: [email protected]

    At Wed, 11 Aug 2004 02:08:14 +0200,
    Marco d'Itri wrote:

    Marco d'Itri wrote:
    Maybe, but the default configuration will not load any module for the display PCI class.
    I know that's not true, because hotplug on fresh d-i installs has taken
    to loading the matroxfb module for my test laptop, which yields a very
    nice highres console. And when I disable hotplug, it stops being loaded. Maybe you have a bug?
    I don't know... Fumitoshi UKAI worked on this. Please attach your /etc/default/hotplug file, it should contain IGNORE_PCI_CLASS_DISPLAY=true.

    Note that in hotplug 0.0.20040329-12 (which probably used in d-i rc1), IGNORE_PCI_CLASS_DISPLAY is effective on only 2.6 kernel, since it checks
    sysfs pci device class and this is not used on 2.4 kernel.
    In hotplug 0.0.20040329-14 (which in sid), IGNORE_PCI_CLASS_DISPLAY
    should be effective on 2.4 kernel too, because it now see whether module driver's path is in drivers/video or not by IGNORE_PCI_CLASS_DISPLAY.

    Also note that 0.0.20040329-12 has minor bug to check blacklist for some modules (such as i810_tco, confusion '-' and '_' in module name, both may be used), which also has been fixed in 0.0.20040329-14.

    Regards,
    Fumitoshi UKAI


    --
    To UNSUBSCRIBE, email to [email protected]
    with a subject of "unsubscribe". Trouble? Contact [email protected]

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)
  • From Timo Jyrinki@1:229/2 to Marco d'Itri on Wed Aug 11 10:10:10 2004
    From: [email protected]

    Marco d'Itri wrote:
    If you want to help finding which module breaks your system please
    run *AS A USER* "/etc/hotplug/pci.rc start", it should display which
    modules are going to be loaded (and fail loading them, because you are
    not root). Then try to manually load each of them with modprobe and
    report which one crashes your system.

    It gave the following ones: radeonfb, usb_uhci ("blacklisted"),
    via-ircc, 3c59x (already loaded), ohci1394, emu10k1-gp, btaudio and bttv.

    And yes, 'modprobe radeonfb' crashed the monitor.

    /etc/default/hotplug:
    # hotplug configuraton file
    # See also /usr/share/doc/hotplug/README.Debian
    # This file is automatically generated by hotplug package
    #
    # STATIC_MODULE_LIST="..."
    # - which modules are preloaded?
    # deprecated. if you need to load at start time, use /etc/modules instead
    # it will be removed in future
    STATIC_MODULE_LIST=""
    #
    # X11_USBMICE_HACK
    # - use USB mouse with X11?
    # deprecated. if you need to load at start time, use /etc/modules instead
    # it will be removed in future
    X11_USBMICE_HACK=false
    #
    # NET_AGENT_POLICY
    # - how to manage network interfaces with ifupdown? 'all', 'auto' or
    'hotplug'
    NET_AGENT_POLICY=hotplug
    #
    # IGNORE_PCI_CLASS_DISPLAY
    # - ignore PCI_CLASS_DISPLAY?
    IGNORE_PCI_CLASS_DISPLAY=true
    #

    So, as Fumitoshi said, IGNORE_PCI_CLASS_DISPLAY doesn't seem to work.
    Radeon 9500-9800 gfx cards are quite common, so this will hopefully get
    fixed.

    -TJ


    --
    To UNSUBSCRIBE, email to [email protected]
    with a subject of "unsubscribe". Trouble? Contact [email protected]

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)
  • From James Strandboge@1:229/2 to All on Thu Aug 12 15:20:06 2004
    From: [email protected]

    I encountered this bug as well. I did a 'dist-upgrade' from woody with
    a 2.4 kernel, rebooted and got a garbled display.

    Turned out to be that IGNORE_PCI_CLASS_DISPLAY doesn't work, and loading
    rivafb was the problem. Installing 0.0.20040329-14 from unstable fixed
    the problem.

    IMO, this bug priority should be raised. I was lucky enough to have a
    dual boot (sarge and woody) machine that I could use to diagnose the
    problem-- but with the console being unusable even in single user mode,
    many will just give up on sarge.

    Jamie

    --
    Targeted Performance Partners, LLC
    Web: http://www.tpptraining.com
    E-mail: [email protected]
    Tel: (585) 271-8370
    Fax: (585) 271-8373



    --
    To UNSUBSCRIBE, email to [email protected]
    with a subject of "unsubscribe". Trouble? Contact [email protected]

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)
  • From James Strandboge@1:229/2 to Marco d'Itri on Thu Aug 12 17:40:09 2004
    From: [email protected]

    On Thu, 2004-08-12 at 10:51, Marco d'Itri wrote:
    retitle 264870 radeonfb breaks the system
    thanks

    Note that on my system, it is 'rivafb'. The original poster had
    problems with 'radeonfb'.

    Jamie




    --
    To UNSUBSCRIBE, email to [email protected]
    with a subject of "unsubscribe". Trouble? Contact [email protected]

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)
  • From Marco d'Itri@1:229/2 to James Strandboge on Thu Aug 12 17:50:07 2004
    From: [email protected]

    retitle 264870 radeonfb breaks the system
    thanks

    On Aug 12, James Strandboge <[email protected]> wrote:

    Turned out to be that IGNORE_PCI_CLASS_DISPLAY doesn't work, and loading rivafb was the problem. Installing 0.0.20040329-14 from unstable fixed
    the problem.
    So the bug with $IGNORE_PCI_CLASS_DISPLAY and 2.4 kernels is fixed in
    unstable.
    I encourage the d-i people to reassign this bug to the appropriate
    kernel package.

    --
    ciao, |
    Marco | [7517 sci/TikRFdYRI]


    --
    To UNSUBSCRIBE, email to [email protected]
    with a subject of "unsubscribe". Trouble? Contact [email protected]

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)