Sparc d-i terminal weirdness? - Debian

This is a discussion on Sparc d-i terminal weirdness? - Debian ; Hi there, I just built a netboot image from lenny using the source from these packages: linux-2.6_2.6.26-5.dsc debian-installer_20080522.dsc linux-kernel-di-sparc-2.6_1.37.dsc and kernel-wedge version 2.52. I retrieved my kernel config from here: http://merkel.debian.org/~jurij/2.6....sparc64-smp.gz My only differences being I added in lines: CONFIG_SUN_LDOMS=y ...

+ Reply to Thread
Results 1 to 10 of 10

Thread: Sparc d-i terminal weirdness?

  1. Sparc d-i terminal weirdness?

    Hi there,

    I just built a netboot image from lenny using the source from these packages:

    linux-2.6_2.6.26-5.dsc
    debian-installer_20080522.dsc
    linux-kernel-di-sparc-2.6_1.37.dsc

    and kernel-wedge version 2.52.

    I retrieved my kernel config from here:
    http://merkel.debian.org/~jurij/2.6....sparc64-smp.gz

    My only differences being I added in lines:

    CONFIG_SUN_LDOMS=y
    CONFIG_SUNVDC=m
    CONFIG_SUNVNET=m

    Anyways, the image builds fine and boots. The problem being, I get to
    the debian-installer menu, and at the point where I need to choose the
    language, the keypresses don't hit the installer. If I boot with
    BOOT_DEBUG=3, every time I get a shell, all the keys work (i.e., I can
    go into nano and the arrow keys, tab, and enter work fine). But when
    I go into the installer, enter just moves down a line, the arrow keys
    produce their familiar A, B, control chars, and tab literally prints a
    tab.

    Does anyone know where I went wrong here? I don't know if its a bug,
    or something I've done wrong. I've built the installer twice now
    using different kernel build methods (kpkg vs fakeroot debian/rules
    binary). I've also built other images in the past that do not have
    this problem. Would really appreciate some help, as I'm tearing my
    hair out here...

    Thanks,
    Mike.


    --
    To UNSUBSCRIBE, email to debian-boot-REQUEST@lists.debian.org
    with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

  2. Re: Sparc d-i terminal weirdness?

    On Thursday 16 October 2008, Mike Grice wrote:
    > debian-installer_20080522.dsc


    This is ancient.
    Suggest you try current subversion repository instead:
    svn co svn://svn.debian.org/svn/d-i/trunk/installer/

    I have no idea what Jurij's config contains. Suggest you use the config
    from the standard 2.6.26 kernel instead (/boot/config*).

    Does the current daily built image for sparc have the same problem?

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.9 (GNU/Linux)

    iEYEABECAAYFAkj2/KoACgkQgm/Kwh6ICoRBVQCgp+DBRRQ+7HrcjRZiL6IFESbX
    nP8AnR0SG97NjKp5UOosvp5aJl+bhPP+
    =Oku/
    -----END PGP SIGNATURE-----


  3. Re: Sparc d-i terminal weirdness?

    On Thu, Oct 16, 2008 at 9:34 AM, Frans Pop wrote:
    > On Thursday 16 October 2008, Mike Grice wrote:
    >> debian-installer_20080522.dsc

    >
    > This is ancient.
    > Suggest you try current subversion repository instead:
    > svn co svn://svn.debian.org/svn/d-i/trunk/installer/


    Will try that now, thanks.

    > I have no idea what Jurij's config contains. Suggest you use the config
    > from the standard 2.6.26 kernel instead (/boot/config*).


    I believe its meant to be the 'standard config' but I'll take a look
    at the one that came with the CD.

    > Does the current daily built image for sparc have the same problem?


    I couldn't find this until just now (I think the link was a 404 for a
    while), but having just tried this now, I *do* get the same problem
    with the daily image.

    Mike


    --
    To UNSUBSCRIBE, email to debian-boot-REQUEST@lists.debian.org
    with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

  4. Re: Sparc d-i terminal weirdness?

    On Thu, Oct 16, 2008 at 10:00 AM, Mike Grice wrote:
    > On Thu, Oct 16, 2008 at 9:34 AM, Frans Pop wrote:
    >> On Thursday 16 October 2008, Mike Grice wrote:
    >>> debian-installer_20080522.dsc

    >>
    >> This is ancient.
    >> Suggest you try current subversion repository instead:
    >> svn co svn://svn.debian.org/svn/d-i/trunk/installer/

    >
    > Will try that now, thanks.


    Ok, having rebuilt with the SVN trunk version, I get the same error...

    >> I have no idea what Jurij's config contains. Suggest you use the config
    >> from the standard 2.6.26 kernel instead (/boot/config*).

    >
    > I believe its meant to be the 'standard config' but I'll take a look
    > at the one that came with the CD.


    This is my next (and only) option, but if I have the same problem with
    my and the official version, maybe theres a problem outside my config?

    Thanks for your help so far!

    Mike.

    >> Does the current daily built image for sparc have the same problem?

    >
    > I couldn't find this until just now (I think the link was a 404 for a
    > while), but having just tried this now, I *do* get the same problem
    > with the daily image.
    >
    > Mike



    --
    To UNSUBSCRIBE, email to debian-boot-REQUEST@lists.debian.org
    with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

  5. Re: Sparc d-i terminal weirdness?

    On Thursday 16 October 2008, Mike Grice wrote:
    > > Does the current daily built image for sparc have the same problem?

    >
    > I couldn't find this until just now (I think the link was a 404 for a
    > while), but having just tried this now, I *do* get the same problem
    > with the daily image.


    I'm fairly sure that nothing has changed in the installer recently, so
    either your system or keyboard has never been supported or, if earlier
    versions of the installer did work, there is a regression in the kernel.

    Does the Etch version of the installer work?

    As the keyboard does still work in the first debug shell, you could try
    running through the installer's initialization scripts manually to see
    where it stops working.

    Does the keyboard still work in the second debug shell (i.e. after udev
    has been started)?

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.9 (GNU/Linux)

    iEYEABECAAYFAkj4XdMACgkQgm/Kwh6ICoS7jwCfU64tV/RgsnuTJ9zkoZdInwJO
    828AoLaG+CPXP5Qc15OGUXZUuoz731Hp
    =FXM+
    -----END PGP SIGNATURE-----


  6. Re: Sparc d-i terminal weirdness?

    On Fri, Oct 17, 2008 at 9:41 AM, Frans Pop wrote:
    > On Thursday 16 October 2008, Mike Grice wrote:
    >> > Does the current daily built image for sparc have the same problem?

    >>
    >> I couldn't find this until just now (I think the link was a 404 for a
    >> while), but having just tried this now, I *do* get the same problem
    >> with the daily image.

    >
    > I'm fairly sure that nothing has changed in the installer recently, so
    > either your system or keyboard has never been supported or, if earlier
    > versions of the installer did work, there is a regression in the kernel.


    Hi, sorry about the delay, I've been on holiday...

    > Does the Etch version of the installer work?


    No. I'm actually doing a netboot into a SUN LDOM. This shouldn't be
    any different from a normal SPARC install, except that the disk and
    net are accessed through new modules. The reason I was rolling my own
    d-i is because the base lenny installer (the rc) does not have these
    two modules, sunvnet and sunvdc, installed. The rc netboot image does
    work as expected, up until the point it (correctly) can't find the
    network card.

    > As the keyboard does still work in the first debug shell, you could try
    > running through the installer's initialization scripts manually to see
    > where it stops working.


    This is weird. I have booted with BOOT_DEBUG before, but I didn't
    think to run the scripts manually. Strangely, if I run them manually,
    it works (although the display looks a bit corrupted (like its not
    redrawn properly, but fine enough to use by hand)).

    > Does the keyboard still work in the second debug shell (i.e. after udev
    > has been started)?


    Yes, this has always been the case.

    So, my difference is:

    - the lenny rc works but has missing drivers (im prepared to roll my
    own to get around this until debian include them by default)
    - the daily (and rolling my own) don't seem to work (with the console
    badness) unless I step through manually?

    Would love to hear your thoughts...

    Mike


    --
    To UNSUBSCRIBE, email to debian-boot-REQUEST@lists.debian.org
    with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

  7. Re: Sparc d-i terminal weirdness?

    On Monday 03 November 2008, Mike Grice wrote:
    > > Does the Etch version of the installer work?

    >
    > No. I'm actually doing a netboot into a SUN LDOM. This shouldn't be
    > any different from a normal SPARC install, except that the disk and
    > net are accessed through new modules. The reason I was rolling my own
    > d-i is because the base lenny installer (the rc) does not have these
    > two modules, sunvnet and sunvdc, installed. The rc netboot image does
    > work as expected, up until the point it (correctly) can't find the
    > network card.


    I've checked and AFAICT these modules are currently not enabled in the
    kernel-image packages for Lenny. This means that, to get this supported
    in the installer, you'll have to do two things:
    - file a wishlist bug report against linux-2.6 requesting those modules
    to be activated for sparc
    - wait until you get an acknowledgement for that report with the bug
    number
    - file a second wishlist BR against linux-kernel-di-sparc-2.6 requesting
    to include the modules in the installer with a reference to the first
    bug report

    > This is weird. I have booted with BOOT_DEBUG before, but I didn't
    > think to run the scripts manually. Strangely, if I run them manually,
    > it works (although the display looks a bit corrupted (like its not
    > redrawn properly, but fine enough to use by hand)).


    Are you sure you ran the scripts the same way as during init? Some scripts
    need to be sourced instead of called normally. As you can get further by
    doing things manually, I'd expect it to be possible to trace the problem
    spot.

    I have absolutely no ideas about this. In cases like this it's basically
    up to the person who sees the problem to try to trace its origin. You
    could try asking on the debian-sparc mailing list or even the sparc list
    where kernel developers hang out (sparclinux AT vger.kernel.org).

    I don't have access to a Sparc LDOM, but even if I did I would not look
    into this myself beyond my general suggestions up to this point as I
    don't consider myself a sparc porter.
    If you do find out what the cause of the problem is, please let us know.
    If there is something we can change in the installer to fix it or work
    around it, we will be most happy to do so.

    Cheers,
    FJP


    --
    To UNSUBSCRIBE, email to debian-boot-REQUEST@lists.debian.org
    with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

  8. Re: Sparc d-i terminal weirdness?

    On Wed, Nov 5, 2008 at 9:27 PM, Frans Pop wrote:
    > On Monday 03 November 2008, Mike Grice wrote:
    >> > Does the Etch version of the installer work?

    >>
    >> No. I'm actually doing a netboot into a SUN LDOM. This shouldn't be
    >> any different from a normal SPARC install, except that the disk and
    >> net are accessed through new modules. The reason I was rolling my own
    >> d-i is because the base lenny installer (the rc) does not have these
    >> two modules, sunvnet and sunvdc, installed. The rc netboot image does
    >> work as expected, up until the point it (correctly) can't find the
    >> network card.

    >
    > I've checked and AFAICT these modules are currently not enabled in the
    > kernel-image packages for Lenny. This means that, to get this supported
    > in the installer, you'll have to do two things:
    > - file a wishlist bug report against linux-2.6 requesting those modules
    > to be activated for sparc
    > - wait until you get an acknowledgement for that report with the bug
    > number
    > - file a second wishlist BR against linux-kernel-di-sparc-2.6 requesting
    > to include the modules in the installer with a reference to the first
    > bug report


    Ok, thanks for that! I've added more text to the original bug I filed
    against the kernel (being new to the debian BR system, I filed it at a
    higher importance... it was reassigned to wishlist. The bug report is
    501651.

    I then filed a new bug as you said against the installer referencing
    501651. This new bug is 504742.

    >> This is weird. I have booted with BOOT_DEBUG before, but I didn't
    >> think to run the scripts manually. Strangely, if I run them manually,
    >> it works (although the display looks a bit corrupted (like its not
    >> redrawn properly, but fine enough to use by hand)).

    >
    > Are you sure you ran the scripts the same way as during init? Some scripts
    > need to be sourced instead of called normally. As you can get further by
    > doing things manually, I'd expect it to be possible to trace the problem
    > spot.


    Ok, I got a little further. I started running the wrapper scripts the
    same way init calls them (via reopen-console). The first part of the
    installer, /sbin/debian-installer-startup, works fine. When I run the
    second part, /sbin/debian-installer itself, things break.

    I'm not sure where, but its hard to recreate the exact environment (as
    its called via reopen-console). I'll have to do some digging, maybe I
    can write a bunch of similar scripts just using the contents of
    reopen-console with the sourcing part included.

    > I have absolutely no ideas about this. In cases like this it's basically
    > up to the person who sees the problem to try to trace its origin. You
    > could try asking on the debian-sparc mailing list or even the sparc list
    > where kernel developers hang out (sparclinux AT vger.kernel.org).


    I might try the debian-sparc list if this doesn't pan out... thanks!

    > I don't have access to a Sparc LDOM, but even if I did I would not look
    > into this myself beyond my general suggestions up to this point as I
    > don't consider myself a sparc porter.
    > If you do find out what the cause of the problem is, please let us know.
    > If there is something we can change in the installer to fix it or work
    > around it, we will be most happy to do so.


    Will do, and thanks for all your help.

    Mike


    --
    To UNSUBSCRIBE, email to debian-boot-REQUEST@lists.debian.org
    with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

  9. Re: Sparc d-i terminal weirdness?

    On Thursday 06 November 2008, Mike Grice wrote:
    > Ok, I got a little further. I started running the wrapper scripts the
    > same way init calls them (via reopen-console). The first part of the
    > installer, /sbin/debian-installer-startup, works fine. When I run the
    > second part, /sbin/debian-installer itself, things break.
    >
    > I'm not sure where, but its hard to recreate the exact environment (as
    > its called via reopen-console). I'll have to do some digging, maybe I
    > can write a bunch of similar scripts just using the contents of
    > reopen-console with the sourcing part included.


    OK, that makes sense. reopen-console is a recent addition in D-I, but is
    required to support both real terminals and e.g. serial console. It is
    very much possible that it lacks support for certain architectures,
    especially for special use-cases like virtual machines.

    I suggest that you open a BR at severity important against rootskel (the
    D-I component that includes reopen-console) with your current analysis
    and follow-up to that with anything further you discover.

    reopen-console itself is just a shell script, so it can be debugged by
    adding a 'set -x' during the early BOOT_DEBUG shells.

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.9 (GNU/Linux)

    iEYEABECAAYFAkkS6hwACgkQgm/Kwh6ICoQutwCgm/iRemUahJ4yI6yP9zlCQMO9
    u2EAn24SgsDX9GoEvy0ecSyDfyZAAIQi
    =DO9G
    -----END PGP SIGNATURE-----


  10. Re: Sparc d-i terminal weirdness?

    On Thu, Nov 6, 2008 at 12:59 PM, Frans Pop wrote:
    > On Thursday 06 November 2008, Mike Grice wrote:
    >> Ok, I got a little further. I started running the wrapper scripts the
    >> same way init calls them (via reopen-console). The first part of the
    >> installer, /sbin/debian-installer-startup, works fine. When I run the
    >> second part, /sbin/debian-installer itself, things break.
    >>
    >> I'm not sure where, but its hard to recreate the exact environment (as
    >> its called via reopen-console). I'll have to do some digging, maybe I
    >> can write a bunch of similar scripts just using the contents of
    >> reopen-console with the sourcing part included.

    >
    > OK, that makes sense. reopen-console is a recent addition in D-I, but is
    > required to support both real terminals and e.g. serial console. It is
    > very much possible that it lacks support for certain architectures,
    > especially for special use-cases like virtual machines.
    >
    > I suggest that you open a BR at severity important against rootskel (the
    > D-I component that includes reopen-console) with your current analysis
    > and follow-up to that with anything further you discover.


    I've done that now (504721). I'll poke more into the debug output if
    I get time tomorrow.

    > reopen-console itself is just a shell script, so it can be debugged by
    > adding a 'set -x' during the early BOOT_DEBUG shells.


    Thanks again,

    Mike.


    --
    To UNSUBSCRIBE, email to debian-boot-REQUEST@lists.debian.org
    with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

+ Reply to Thread