What was the reason for enabling Xen on all i386 kernels? - Debian

This is a discussion on What was the reason for enabling Xen on all i386 kernels? - Debian ; I would like to know why it was decided to enable Xen on all i386 kernels. The changelog simply states it was done, which isn't exactly helpful, other than to show it apparently was on purpose and not an accident. ...

+ Reply to Thread
Results 1 to 8 of 8

Thread: What was the reason for enabling Xen on all i386 kernels?

  1. What was the reason for enabling Xen on all i386 kernels?

    I would like to know why it was decided to enable Xen on all i386
    kernels. The changelog simply states it was done, which isn't exactly
    helpful, other than to show it apparently was on purpose and not an
    accident.

    Meanwhile amd64 kernels don't have it enabled. Why the difference?

    Perhaps I am wrong, but I suspect the majority of users don't give a
    darn about xen support, while on the other hand quite a lot of people
    are quite annoyed at loosing the ability to use a lot of kernel modules
    that made their machines actually do what they wanted to do. Certainly
    the nvidia and ati drivers are broken by this, and I am not sure how
    much convincing it would take to get either of them to fix it. I
    wouldn't be surprised if ndiswrapper is broken by this either, although
    I haven't used that lately.

    So really what is the point of making all kernels xen enabled when
    almost noone will actually use that feature, while at the same time
    causing lots of grief for a much larger group of users? Was having
    seperate xen flavour kernels really that big a deal?

    The only reason I seem to have found so far is in bug 473645 which to me
    looks like this was done entirely to make life easier for the debian
    installer team so that they can do testing under xen. If that's the
    case, well the d-i team just took priority over actually using the
    system by a much larger group of users. Sure the d-i is important, but
    so is actually being able to use the system. Could we at least have
    some kernel images built without xen, even if they aren't the default so
    that we can still use the system for what we want without all users
    having to go and start building custom kernels. I for one am not
    looking forward to going back to that, but if I have to, then I will. I
    highly think this change should be reconsidered.

    --
    Len Sorensen


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

  2. Re: What was the reason for enabling Xen on all i386 kernels?

    On Wed, May 21, 2008 at 12:29:52PM -0400, Lennart Sorensen wrote:
    > I would like to know why it was decided to enable Xen on all i386
    > kernels. The changelog simply states it was done, which isn't exactly
    > helpful, other than to show it apparently was on purpose and not an
    > accident.


    Because it is supported.

    > Meanwhile amd64 kernels don't have it enabled. Why the difference?


    Because it is not supported.

    > Perhaps I am wrong, but I suspect the majority of users don't give a
    > darn about xen support, while on the other hand quite a lot of people
    > are quite annoyed at loosing the ability to use a lot of kernel modules
    > that made their machines actually do what they wanted to do.


    The Xen support only adds something, it does not remove something which
    was there before.

    > Certainly
    > the nvidia and ati drivers are broken by this, and I am not sure how
    > much convincing it would take to get either of them to fix it. I
    > wouldn't be surprised if ndiswrapper is broken by this either, although
    > I haven't used that lately.


    Please define "broken". If they would be broken by paravirt support,
    they would already fail to work with 2.6.24. Please take a look at
    #481485.

    > So really what is the point of making all kernels xen enabled when
    > almost noone will actually use that feature, while at the same time
    > causing lots of grief for a much larger group of users? Was having
    > seperate xen flavour kernels really that big a deal?


    Please show evidence that the support breaks something except random
    scripts for non-free software.

    Bastian

    --
    Prepare for tomorrow -- get ready.
    -- Edith Keeler, "The City On the Edge of Forever",
    stardate unknown


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

  3. Re: What was the reason for enabling Xen on all i386 kernels?

    On Wed, May 21, 2008 at 12:29:52PM -0400, Lennart Sorensen wrote:
    > I would like to know why it was decided to enable Xen on all i386
    > kernels. The changelog simply states it was done, which isn't exactly
    > helpful, other than to show it apparently was on purpose and not an
    > accident.


    the reason seems quite evident.
    to be able to have any debian kernel as xen guest.
    it is a frequent feature request we received.

    > Meanwhile amd64 kernels don't have it enabled. Why the difference?


    because paravirt upstream support is currently only x86_32
    once ported to amd64 it will enabled there too.
    but haven't yet landed for 2.6.26, so still..


    please bug your vendor for those.
    ati is already mostly supported by the free radeon driver,
    use latest of it.
    anyway nvidia hasn't released any official driver for 2.6.25 afaik.

    best regards

    --
    maks


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

  4. Re: What was the reason for enabling Xen on all i386 kernels?

    On Wed, May 21, 2008 at 06:33:12PM +0200, maximilian attems wrote:
    > the reason seems quite evident.
    > to be able to have any debian kernel as xen guest.
    > it is a frequent feature request we received.


    Well I would like to request a kernel without this feature again. It
    isn't as if a 686-noxen kernel would be a big load on the x86 buildd's
    and the disk space isn't particularly significant and it would save a
    lot of users the trouble of having to configure and compile their own
    kernels to work around this change.

    > because paravirt upstream support is currently only x86_32
    > once ported to amd64 it will enabled there too.
    > but haven't yet landed for 2.6.26, so still..
    >
    >
    > please bug your vendor for those.
    > ati is already mostly supported by the free radeon driver,
    > use latest of it.
    > anyway nvidia hasn't released any official driver for 2.6.25 afaik.


    The 173.08 driver works fine with 2.6.25, as long as xen isn't enabled
    (in which case compiling the driver fails).

    Unfortunately I don't have much hope for xen support in either the ati
    or nvidia drivers until some day redhat or suse decides to make it
    mandetory, and I actually doubt they are going to do that.

    So at the moment there are two fixes:

    make nvidia/ati fix it: Not seeming very likely.
    make the debian kernel team fix it: Simple and actually possible, other
    than having to potentially deal with politics and ideologies.

    I suppose another option is to setup another unofficial repository and
    start offering kernels for those people that want to still be able to
    use their systems until someday when just maybe xen support is added to
    the various nonfree drivers.

    --
    Len Sorensen


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

  5. Re: What was the reason for enabling Xen on all i386 kernels?

    On Wed, May 21, 2008 at 06:47:25PM +0200, Bastian Blank wrote:
    > Because it is not supported.


    OK, that is a good reason.

    > The Xen support only adds something, it does not remove something which
    > was there before.


    I was under the impression it changed some of the ways to access
    physical addresses. I guess for the dom0 kernel that wouldn't really
    make sense though.

    > Please define "broken". If they would be broken by paravirt support,
    > they would already fail to work with 2.6.24. Please take a look at
    > #481485.


    OK, I will try and work from that.

    And broken as in "it won't compile".

    > Please show evidence that the support breaks something except random
    > scripts for non-free software.


    So xen support should not break any driver? If that is the case, I will
    continue to try and fix the compile problem and try and make it build
    and then see if it does work. Certainly everything I found about it so
    far indicated that the xen changes would break driver access to the
    hardware in some cases, but I guess believing random stuff on the
    internet could be a bad idea. I am willing to try fixing it some more.

    --
    Len Sorensen


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

  6. Re: What was the reason for enabling Xen on all i386 kernels?

    On Wed, May 21, 2008 at 02:17:17PM -0400, Lennart Sorensen wrote:
    >
    > make nvidia/ati fix it: Not seeming very likely.


    why?

    their scripts are buggy, beat them to it.
    you seem to have enough energy.


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

  7. Re: What was the reason for enabling Xen on all i386 kernels?

    On Wed, May 21, 2008 at 08:25:44PM +0200, maximilian attems wrote:
    > why?
    >
    > their scripts are buggy, beat them to it.
    > you seem to have enough energy.


    I am doing that now. I was under the (apparently mistaken) impression
    that the xen features would break the driver's access to the hardware
    without changes.

    I guess the real case is that the driver should work on a kernel with
    xen support, just as long as it is the main kernel and not a xen guest
    kernel. That would mean checking for CONFIG_XEN is not the correct way
    to detect whether you are in a xen guest or running on dom0.

    I am about to test a build against the 2.6.25-2-686 kernel using the
    173.08 driver with a patch I just made. I hope it works.

    --
    Len Sorensen


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

  8. Re: What was the reason for enabling Xen on all i386 kernels?

    On Wed, May 21, 2008 at 03:31:45PM -0400, Lennart Sorensen wrote:
    > I am doing that now. I was under the (apparently mistaken) impression
    > that the xen features would break the driver's access to the hardware
    > without changes.
    >
    > I guess the real case is that the driver should work on a kernel with
    > xen support, just as long as it is the main kernel and not a xen guest
    > kernel. That would mean checking for CONFIG_XEN is not the correct way
    > to detect whether you are in a xen guest or running on dom0.
    >
    > I am about to test a build against the 2.6.25-2-686 kernel using the
    > 173.08 driver with a patch I just made. I hope it works.


    So it booted, it started X with the nvidia driver. Yay!

    I will now stop whining and go finish the patch.

    Thanks for the push in the right direction.

    --
    Len Sorensen


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

+ Reply to Thread