Call for Debian Installer testing, before D-I Beta1 release - Debian

This is a discussion on Call for Debian Installer testing, before D-I Beta1 release - Debian ; -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hello, We're starting the work for D-I Beta1 release and we would like ask to everyone to test the installer as much as possible, especially in not so common architectures where we receive less ...

+ Reply to Thread
Results 1 to 18 of 18

Thread: Call for Debian Installer testing, before D-I Beta1 release

  1. Call for Debian Installer testing, before D-I Beta1 release

    -----BEGIN PGP SIGNED MESSAGE-----
    Hash: SHA1

    Hello,

    We're starting the work for D-I Beta1 release and we would like ask to
    everyone to test the installer as much as possible, especially in not
    so common architectures where we receive less general testing.

    In the current timeline, we will stop to accept fixes for the
    installer in few days so any important issues need to be fixed _now_.

    You can find the latest installer, for your pet architecture, at
    http://www.debian.org/devel/debian-installer/ and if you find any
    issues file a installation report.

    Thanks in advance and let's make Lenny the best release ever made!

    Cheers,
    Otavio Salvador
    on behalf of Debian Installer Team

    - --
    O T A V I O S A L V A D O R
    - ---------------------------------------------
    E-mail: otavio@debian.org UIN: 5906116
    GNU/Linux User: 239058 GPG ID: 49A5F855
    Home Page: http://otavio.ossystems.com.br
    - ---------------------------------------------
    "Microsoft sells you Windows ... Linux gives
    you the whole house."
    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.6 (GNU/Linux)
    Comment: Processed by Mailcrypt 3.5.8+

    iD8DBQFHp5L1LqiZQEml+FURAm5PAKCf8IPsN9zFHhth5pio7N Y3MpEgdQCeIIn5
    RAYua80aQbLTjk9BDfupWRU=
    =cTEq
    -----END PGP SIGNATURE-----


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

  2. Re: Call for Debian Installer testing, before D-I Beta1 release

    Hi All

    On Feb 4, 2008 3:34 PM, Otavio Salvador wrote:

    > We're starting the work for D-I Beta1 release and we would like ask to
    > everyone to test the installer as much as possible, especially in not
    > so common architectures where we receive less general testing.


    Installation on the Linksys NSLU2 will be broken until nslu2-utils
    0.10+r71-14 [1] transitions into testing. This version of nslu2-utils
    is only required for DSFG installer images (i.e. no firmware) and for
    installer images using kernel versions 2.6.24 and greater.

    Gordon

    [1] http://packages.qa.debian.org/n/nslu2-utils.html

    --
    Gordon Farquharson
    GnuPG Key ID: 32D6D676


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

  3. Re: Call for Debian Installer testing, before D-I Beta1 release

    Hello,

    Just to mention that I've just tested the image with a USB braille
    device, and it just works great

    Samuel


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

  4. Re: Call for Debian Installer testing, before D-I Beta1 release

    On Monday 04 February 2008, Otavio Salvador wrote:
    > We're starting the work for D-I Beta1 release and we would like ask to
    > everyone to test the installer as much as possible, especially in not
    > so common architectures where we receive less general testing.


    I have run netboot installation tests for hppa, s390 and sparc.
    For all I have built the images myself mainly because there are no current
    daily images for s390 (build failures, looks like for some time already)
    and sparc (no new builds since end of Jan).

    s390 and sparc: successful installs, no issues

    hppa
    The first part of the installation runs fine, but I'm running into a weird
    failure at the start of partman. The failure seems to be not in partman
    itself, but in main-menu. I'm fairly sure that partman-base.postinst is
    never even started.

    Here's part of the log:
    [hard disk detection completed]
    Feb 10 23:48:43 main-menu[3851]: INFO: Falling back to the package
    description for palo-installer
    Feb 10 23:50:29 main-menu[3851]: INFO: Falling back to the package
    description for auto-install
    Feb 10 23:50:29 main-menu[3851]: INFO: Falling back to the package
    description for ai-choosers
    [main menu displayed]
    Feb 10 23:50:29 main-menu[3851]: INFO: Menu item 'partman-base' selected
    Feb 10 23:50:29 main-menu[3851]: INFO: Falling back to the package
    description for kernel-image-2.6.22-3-parisc-di
    Feb 10 23:50:29 main-menu[3851]: INFO: Falling back to the package
    description for kernel-image-2.6.22-3-parisc64-di
    Feb 10 23:50:29 main-menu[3851]: INFO: Falling back to the package
    description for kernel-image-2.6.22-3-parisc-di
    Feb 10 23:50:29 main-menu[3851]: INFO: Falling back to the package
    description for kernel-image-2.6.22-3-parisc64-di
    Feb 10 23:50:29 main-menu[3851]: INFO: Falling back to the package
    description for partman-palo
    [!!! this message repeats A LOT !!!]
    Feb 10 23:50:29 main-menu[3851]: INFO: Falling back to the package
    description for partman-palo
    Feb 10 23:50:29 main-menu[3851]: WARNING **: Deep recursion configuring
    package partman-base (dep loop?)
    Feb 10 23:50:29 main-menu[3851]: WARNING **: Menu item 'partman-base'
    failed.

    Note that there seems to be something weird with the menu description for
    partman palo even before partman is started. As the data for partman-palo
    looks fine in the dpkg status file that could be just a symptom, and not
    the cause.

    At the moment I haven't got a clue what's going on here.

    I'm going to try with a custom build of partman-palo and probably
    partman-base, and possibly with some additional debugging in main-menu or
    using strace to see if I can get any further than this.

    Cheers,
    FJP

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.6 (GNU/Linux)

    iD8DBQBHr6Wvgm/Kwh6ICoQRAhfGAKC2ptt835cQNZs1wJ2Xs8zh6tz4IQCgl76d
    sbdS/LA7KR/t3iyHGDyyqMo=
    =axyK
    -----END PGP SIGNATURE-----


  5. Re: Call for Debian Installer testing, before D-I Beta1 release

    On Monday 11 February 2008, Frans Pop wrote:
    > hppa
    > The first part of the installation runs fine, but I'm running into a
    > weird failure at the start of partman. The failure seems to be not in
    > partman itself, but in main-menu. I'm fairly sure that
    > partman-base.postinst is never even started.


    Of course, it does not help if you include ancient udebs in your image
    because your build environment isn't clean...

    Without those the install went fine on my HPPA box.

    Cheers,
    FJP

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.6 (GNU/Linux)

    iD8DBQBHsLF7gm/Kwh6ICoQRAvbZAKCluWIyVE9GRBBmJxpQ/eAzTyPF3wCeJOaY
    68CsyPoPOibd8XMvl+SrtEM=
    =ISDy
    -----END PGP SIGNATURE-----


  6. Re: Call for Debian Installer testing, before D-I Beta1 release

    On Wed, Feb 6, 2008 at 6:56 AM, Gordon Farquharson
    wrote:

    > Installation on the Linksys NSLU2 will be broken until nslu2-utils
    > 0.10+r71-14 [1] transitions into testing. This version of nslu2-utils
    > is only required for DSFG installer images (i.e. no firmware) and for
    > installer images using kernel versions 2.6.24 and greater.


    The "and" above should really be an "or".

    The daily installer images now work.

    Gordon

    --
    Gordon Farquharson
    GnuPG Key ID: 32D6D676


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

  7. Re: Call for Debian Installer testing, before D-I Beta1 release

    On Mon, Feb 04, 2008 at 08:34:54PM -0200, Otavio Salvador wrote:
    > We're starting the work for D-I Beta1 release and we would like ask to
    > everyone to test the installer as much as possible, especially in not
    > so common architectures where we receive less general testing.


    Otavio,

    On Feb 9, I pulled the 150MB "net install" ISO image, burned a CD
    and tested that today on my j6000 (hppa, aka parisc). Install
    options selected were "/home" and used dhcp to config the network.
    (I had to setup a dhcpd on my laptop, which was a NAT'd gateway).
    Install succeeded and system rebooted to a 32-bit UP kernel.
    (2.6.22-3-parisc #2 Mon Nov 12 20:42:16 CET 2007). This box only
    has 2GB of RAM installed and 2 CPUs.
    So despite ability to run 64-bit kernel, 32-bit kernel is fine.

    SMP would have been appropriate though. Installing SMP kernel is not
    a critical issue. Folks can easily figure out how to install an SMP
    kernel and SMP kernel might not be as stable for some configurations.
    So leaving that to the user is fine with me. I just want to share how
    detection might be implemented.


    Kernel at boot time signals the additional CPUs in two ways:
    ....
    Searching for devices...
    Found devices:
    ....
    6. Duet W+ at 0xfffa0000 [32] { 0, 0x0, 0x5d4, 0x00004 }
    7. Duet W+ at 0xfffa2000 [34] { 0, 0x0, 0x5d4, 0x00004 }
    8. Memory at 0xfed10200 [49] { 1, 0x0, 0x00a, 0x00009 }
    Enabling regular chassis codes support v0.05
    CONFIG_SMP=n ignoring additional CPUs
    CPU: probe of 34 failed with error 1
    ....

    First, the probe failure and then the fact that multiple CPUs are listed.

    Runtime test for the hppa installer might be to scrounge in
    /sys/bus/parisc/devices for "hw_type" contains "0x00".

    e.g.:
    j6k:~# fgrep 0x00 /sys/bus/parisc/devices/*/hw_type
    /sys/bus/parisc/devices/32/hw_type:0x00
    /sys/bus/parisc/devices/34/hw_type:0x00


    > In the current timeline, we will stop to accept fixes for the
    > installer in few days so any important issues need to be fixed _now_.
    >
    > You can find the latest installer, for your pet architecture, at
    > http://www.debian.org/devel/debian-installer/ and if you find any
    > issues file a installation report.


    BTW, the "News" section of that web page needs love.

    [21 Jul 2007] Support for sparc32 dropped
    [20 Jul 2007] Experimental support for Serial ATA RAID

    isn't exactly news anymore

    And this is my installation report. I didn't see any other guidance
    on where/how to submit the installation reports.

    > Thanks in advance and let's make Lenny the best release ever made!


    thanks too!
    grant

    >
    > Cheers,
    > Otavio Salvador
    > on behalf of Debian Installer Team
    >
    > - --
    > O T A V I O S A L V A D O R
    > - ---------------------------------------------
    > E-mail: otavio@debian.org UIN: 5906116
    > GNU/Linux User: 239058 GPG ID: 49A5F855
    > Home Page: http://otavio.ossystems.com.br
    > - ---------------------------------------------
    > "Microsoft sells you Windows ... Linux gives
    > you the whole house."
    > -----BEGIN PGP SIGNATURE-----
    > Version: GnuPG v1.4.6 (GNU/Linux)
    > Comment: Processed by Mailcrypt 3.5.8+
    >
    > iD8DBQFHp5L1LqiZQEml+FURAm5PAKCf8IPsN9zFHhth5pio7N Y3MpEgdQCeIIn5
    > RAYua80aQbLTjk9BDfupWRU=
    > =cTEq
    > -----END PGP SIGNATURE-----
    >
    >
    > --
    > To UNSUBSCRIBE, email to debian-hppa-REQUEST@lists.debian.org
    > with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org



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

  8. Re: Call for Debian Installer testing, before D-I Beta1 release

    Grant Grundler writes:

    <...>
    > SMP would have been appropriate though. Installing SMP kernel is not
    > a critical issue. Folks can easily figure out how to install an SMP
    > kernel and SMP kernel might not be as stable for some configurations.
    > So leaving that to the user is fine with me. I just want to share how
    > detection might be implemented.


    Right. base-installer was suppose to deal nicely with it for you.

    > Runtime test for the hppa installer might be to scrounge in
    > /sys/bus/parisc/devices for "hw_type" contains "0x00".


    > e.g.:
    > j6k:~# fgrep 0x00 /sys/bus/parisc/devices/*/hw_type
    > /sys/bus/parisc/devices/32/hw_type:0x00
    > /sys/bus/parisc/devices/34/hw_type:0x00


    I don't have access to get this data however today's base-installer
    checks it on /proc/cpuinfo.

    <...>
    > And this is my installation report. I didn't see any other guidance
    > on where/how to submit the installation reports.


    Please, go in your installed system and use the installation-report
    script to send us a full installation report, that way we can get more
    information about how things went there.

    I've checked base-installer code and it has code to handle hppa smp
    machines. Probably 2.6.22 had a problem to enable your second
    processor and then this might be fixed once we go to 2.6.24.

    Filling a full installation report will allow us to check if this code
    has issues or if it wasn't properly detected.

    --
    O T A V I O S A L V A D O R
    ---------------------------------------------
    E-mail: otavio@debian.org UIN: 5906116
    GNU/Linux User: 239058 GPG ID: 49A5F855
    Home Page: http://otavio.ossystems.com.br
    ---------------------------------------------
    "Microsoft sells you Windows ... Linux gives
    you the whole house."


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

  9. Re: Call for Debian Installer testing, before D-I Beta1 release

    On Monday 18 February 2008, Grant Grundler wrote:
    > Runtime test for the hppa installer might be to scrounge in
    > /sys/bus/parisc/devices for "hw_type" contains "0x00".
    >
    > e.g.:
    > j6k:~# fgrep 0x00 /sys/bus/parisc/devices/*/hw_type
    > /sys/bus/parisc/devices/32/hw_type:0x00
    > /sys/bus/parisc/devices/34/hw_type:0x00


    Yes, that seems like it could be a reliable test. At least it works on my
    PARISC box too. I'll implement something in base-installer.

    > BTW, the "News" section of that web page needs love.
    >
    > [21 Jul 2007] Support for sparc32 dropped
    > [20 Jul 2007] Experimental support for Serial ATA RAID
    >
    > isn't exactly news anymore


    No, but it's still the most recent news we have (until Beta1 gets released),
    so I don't see anything wrong with it either...

    Cheers,
    FJP

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.6 (GNU/Linux)

    iD8DBQBHuZv7gm/Kwh6ICoQRAjatAKCv1h6X9pSNpbM1fGqqDDsnEKjTugCeOc8D
    0dEhLj0HQr1O5QXtA8K5iQY=
    =1QW4
    -----END PGP SIGNATURE-----


  10. Re: Call for Debian Installer testing, before D-I Beta1 release

    On Monday 18 February 2008, Otavio Salvador wrote:
    > Grant Grundler writes:
    > > e.g.:
    > > j6k:~# fgrep 0x00 /sys/bus/parisc/devices/*/hw_type
    > > /sys/bus/parisc/devices/32/hw_type:0x00
    > > /sys/bus/parisc/devices/34/hw_type:0x00

    >
    > I don't have access to get this data however today's base-installer
    > checks it on /proc/cpuinfo.


    Which will not work because that test only works if you're _already running_
    an SMP kernel and the installer doesn't. Looking in sysfs seems to be a
    much more solid solution.

    > > And this is my installation report. I didn't see any other guidance
    > > on where/how to submit the installation reports.


    See the link in "After using the Debian-Installer, please send us an
    installation report, even if there weren't any problems." on the D-I
    homepage (the same one with the outdated news ;-)

    > I've checked base-installer code and it has code to handle hppa smp
    > machines. Probably 2.6.22 had a problem to enable your second
    > processor and then this might be fixed once we go to 2.6.24.


    That really has nothing to do with it.

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.6 (GNU/Linux)

    iD8DBQBHuZ0Wgm/Kwh6ICoQRAi5HAKCSyB0aXRO9JFaJIbtaoW9yhr9DpgCggald
    NRPZiMrqd1JfvjYg5kX+0ec=
    =QITI
    -----END PGP SIGNATURE-----


  11. Re: Call for Debian Installer testing, before D-I Beta1 release

    Frans Pop writes:

    > On Monday 18 February 2008, Grant Grundler wrote:
    >> Runtime test for the hppa installer might be to scrounge in
    >> /sys/bus/parisc/devices for "hw_type" contains "0x00".
    >>
    >> e.g.:
    >> j6k:~# fgrep 0x00 /sys/bus/parisc/devices/*/hw_type
    >> /sys/bus/parisc/devices/32/hw_type:0x00
    >> /sys/bus/parisc/devices/34/hw_type:0x00

    >
    > Yes, that seems like it could be a reliable test. At least it works on my
    > PARISC box too. I'll implement something in base-installer.


    I did it just after answer his mail and was going to ask someone to
    test. Could you see the attached patch?

    diff --git a/packages/base-installer/debian/changelog b/packages/base-installer/debian/changelog
    index f1ee8d4..8ddfa62 100644
    --- a/packages/base-installer/debian/changelog
    +++ b/packages/base-installer/debian/changelog
    @@ -1,10 +1,15 @@
    base-installer (1.88) UNRELEASED; urgency=low

    + [ Joey Hess ]
    * Demote Transmeta Crusoe systems to use the 486 kernel.
    Based on #464962, Crusoe lacks support for a long NOP instruction,
    which gcc has begun using when optimising for the 686.

    - -- Joey Hess Tue, 12 Feb 2008 18:04:46 -0500
    + [ Otavio Salvador ]
    + * Improve hppa SMP kernel test. Thanks to Grant Grundler for his
    + suggestions.
    +
    + -- Otavio Salvador Mon, 18 Feb 2008 09:58:00 -0300

    base-installer (1.87) unstable; urgency=low

    diff --git a/packages/base-installer/kernel/hppa.sh b/packages/base-installer/kernel/hppa.sh
    index 476b7a6..ac8265c 100644
    --- a/packages/base-installer/kernel/hppa.sh
    +++ b/packages/base-installer/kernel/hppa.sh
    @@ -14,8 +14,7 @@ arch_check_usable_kernel () {
    }

    arch_get_kernel () {
    - CPUS="$(grep ^processor "$CPUINFO" | tail -n 1 | cut -d: -f2)"
    - if [ -z "$CPUS" ] || [ "$CPUS" -ne 0 ]; then
    + if grep -q 0x00 /sys/bus/parisc/devices/*/hw_type; then
    echo "linux-image-$KERNEL_MAJOR-$1-smp"
    fi
    echo "linux-image-$KERNEL_MAJOR-$1"


    --
    O T A V I O S A L V A D O R
    ---------------------------------------------
    E-mail: otavio@debian.org UIN: 5906116
    GNU/Linux User: 239058 GPG ID: 49A5F855
    Home Page: http://otavio.ossystems.com.br
    ---------------------------------------------
    "Microsoft sells you Windows ... Linux gives
    you the whole house."


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

  12. Re: Call for Debian Installer testing, before D-I Beta1 release

    On Monday 18 February 2008, Otavio Salvador wrote:
    > + if grep -q 0x00 /sys/bus/parisc/devices/*/hw_type; then


    You're only testing for the presence of _a_ processor here. You're not
    testing _how many_ there are.

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.6 (GNU/Linux)

    iD8DBQBHuaT6gm/Kwh6ICoQRAhkVAJ0cKZG5ddpDYKQ83m17HqMj4WR2ZQCfaAeo
    U03f4wP8jAZJyM6f5MgvHXk=
    =9A9Z
    -----END PGP SIGNATURE-----


  13. Re: Call for Debian Installer testing, before D-I Beta1 release

    Frans Pop writes:

    > On Monday 18 February 2008, Otavio Salvador wrote:
    >> I did it just after answer his mail and was going to ask someone to
    >> test. Could you see the attached patch?

    >
    > I've about to commit a different solution where /var/numcpus is set in
    > rootskel. Main reason is that that will work better with the base-installer
    > kernel testsuite.


    Please go ahead.

    --
    O T A V I O S A L V A D O R
    ---------------------------------------------
    E-mail: otavio@debian.org UIN: 5906116
    GNU/Linux User: 239058 GPG ID: 49A5F855
    Home Page: http://otavio.ossystems.com.br
    ---------------------------------------------
    "Microsoft sells you Windows ... Linux gives
    you the whole house."


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

  14. Re: Call for Debian Installer testing, before D-I Beta1 release

    On Monday 18 February 2008, Otavio Salvador wrote:
    > I did it just after answer his mail and was going to ask someone to
    > test. Could you see the attached patch?


    I've about to commit a different solution where /var/numcpus is set in
    rootskel. Main reason is that that will work better with the base-installer
    kernel testsuite.

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.6 (GNU/Linux)

    iD8DBQBHuaRQgm/Kwh6ICoQRAjuwAKCe6EkLCL/CWJmAixhoUVsu8gnlgwCeOndk
    F2SAt7c4O9Q51CGsJnKm158=
    =697y
    -----END PGP SIGNATURE-----


  15. Re: Call for Debian Installer testing, before D-I Beta1 release

    On Mon, Feb 18, 2008 at 09:54:22AM -0300, Otavio Salvador wrote:
    > <...>
    > > And this is my installation report. I didn't see any other guidance
    > > on where/how to submit the installation reports.

    >
    > Please, go in your installed system and use the installation-report
    > script to send us a full installation report, that way we can get more
    > information about how things went there.


    I'm not near the system for the next couple of days but will do that
    when I get back - thanks for pointing that out.

    >
    > I've checked base-installer code and it has code to handle hppa smp
    > machines. Probably 2.6.22 had a problem to enable your second
    > processor and then this might be fixed once we go to 2.6.24.


    Well, historically, hppa has had problems with SMP kernels. AFAIK, the worst
    of those have been fixed and SMP seems (in general) stable.

    > Filling a full installation report will allow us to check if this code
    > has issues or if it wasn't properly detected.


    ok - will do. Expect that within a week.

    thanks,
    grant
    >
    > --
    > O T A V I O S A L V A D O R
    > ---------------------------------------------
    > E-mail: otavio@debian.org UIN: 5906116
    > GNU/Linux User: 239058 GPG ID: 49A5F855
    > Home Page: http://otavio.ossystems.com.br
    > ---------------------------------------------
    > "Microsoft sells you Windows ... Linux gives
    > you the whole house."



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

  16. Re: Call for Debian Installer testing, before D-I Beta1 release

    On Wednesday 20 February 2008, Grant Grundler wrote:
    > > Please, go in your installed system and use the installation-report
    > > script to send us a full installation report, that way we can get more
    > > information about how things went there.

    >
    > I'm not near the system for the next couple of days but will do that
    > when I get back - thanks for pointing that out.


    That's not really needed anymore as I've already made the necessary changes
    to detect smp support. They will only be uploaded after the Beta1 release
    though.
    See also my replies to your original mail (not CCed).

    Grant, do you happen to know a similar trick to detect 64-bit support even
    when the installer is running with a 32-bit kernel?
    Currently the installer will only select a 64-bit kernel if the user changed
    to the 64-bit D-I kernel when the installer was booted (which can be done
    by answering Yes to the question asked by the firmware whether you want to
    interact with the boot loader).

    Cheers,
    FJP

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.6 (GNU/Linux)

    iD8DBQBHu/vRgm/Kwh6ICoQRAoS1AJ0c/5VZHMYXSG7gmwYYD6gT4zhO4ACcC8B6
    0MfgE8w5Sw48HMlt+2LdnbA=
    =YvP3
    -----END PGP SIGNATURE-----


  17. Re: Call for Debian Installer testing, before D-I Beta1 release

    On Wed, Feb 20, 2008 at 11:07:03AM +0100, Frans Pop wrote:
    > On Wednesday 20 February 2008, Grant Grundler wrote:
    > > > Please, go in your installed system and use the installation-report
    > > > script to send us a full installation report, that way we can get more
    > > > information about how things went there.

    > >
    > > I'm not near the system for the next couple of days but will do that
    > > when I get back - thanks for pointing that out.

    >
    > That's not really needed anymore as I've already made the necessary changes
    > to detect smp support. They will only be uploaded after the Beta1 release
    > though.
    > See also my replies to your original mail (not CCed).


    ok - cool

    >
    > Grant, do you happen to know a similar trick to detect 64-bit support even
    > when the installer is running with a 32-bit kernel?


    Hrm...yes. I explain the "trick" and also some caveats on why I would
    rather leave this user selectable.

    grundler@gsyprf11:~$ fgrep "PA-RISC 2.0" /proc/cpuinfo
    cpu family : PA-RISC 2.0
    cpu family : PA-RISC 2.0

    Anything with PA 2.0 CPU can run 64-bit.

    In general, we want to run 32-bit kernel because:
    o PA-RISC to date _only_ has a 32-bit usr space. So all syscalls to
    a 64-bit kernel go through a "wrapper" to convert 32-bit user space
    arguments to 64-bit kernel calling conventions.
    (some work is being done to enable 64-bit user space but it's not
    ready yet).
    o pointers are smaller and thus data structures are more compact.
    Simple applications are _slower_ with 64-bit.

    We must/can run 64-bit IFF:
    o PAT PDC (aka BIOS) - palo boot loader detects PAT PDC and will
    attempt to boot a 64-bit kernel. No changes needed for this.

    o chunks of physical memory are mapped above 4GB.
    Different platforms will have different physical memory maps.
    /proc/meminfo is of no help here.
    But /proc/iomem might be helpful:
    grundler@gsyprf11:~$ fgrep "System RAM" /proc/iomem
    00000000-efffffff : System RAM
    100000000-1ffffffff : System RAM
    10f0000000-10ffffffff : System RAM

    These are the physical addresses of RAM for an "A500" (aka rp2470)
    with 8GB of RAM:
    grundler@gsyprf11:~$ cat /proc/meminfo
    MemTotal: 8239436 kB
    MemFree: 414672 kB
    Buffers: 1812020 kB
    Cached: 4587320 kB
    SwapCached: 72 kB
    Active: 4434600 kB
    Inactive: 2115708 kB
    ....

    This is really only a problem for "Astro" (e.g. C3000) chipset workstations.
    They have PA2.0 CPUs and can have physical RAM above 4GB address.

    > Currently the installer will only select a 64-bit kernel if the user changed
    > to the 64-bit D-I kernel when the installer was booted (which can be done
    > by answering Yes to the question asked by the firmware whether you want to
    > interact with the boot loader).


    Unless someone wants to code up the above rules, my preference would be
    to leave this up to the user. We can revisit this issue when 64-bit user
    space is working.

    thanks,
    grant

    >
    > Cheers,
    > FJP




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

  18. Re: Call for Debian Installer testing, before D-I Beta1 release

    On Sunday 24 February 2008, Grant Grundler wrote:
    > > Grant, do you happen to know a similar trick to detect 64-bit support
    > > even when the installer is running with a 32-bit kernel?

    >
    > Hrm...yes. I explain the "trick" and also some caveats on why I would
    > rather leave this user selectable.


    OK. In that case I won't bother about this (for now).

    Thanks,
    FJP

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.6 (GNU/Linux)

    iD8DBQBHxHj1gm/Kwh6ICoQRAp1KAJ9kRFJG3VyEnEWCYhU4s/490Z+aVACgpz2i
    KRvdg3z0dQA9+YIXDqKNXj8=
    =QgPn
    -----END PGP SIGNATURE-----


+ Reply to Thread