Hi Zach,
Thanks for your help on the VMI-related 2.6.27.x regression. I'm hoping
to now pick your mind on a similar but unrelated issue:
Norman has found that with certain kernel configurations, 2.6.27 (plain,
as released by Linus) does not initialize VMI properly, witnessed as
missing initialization messages in the kernel logs. Other configurations
work fine. These are the messages that go missing:
VMI: Found VMware, Inc. Hypervisor OPROM, API version 3.0, ROM
version 1.0
vmi: registering clock event vmi-timer. mult=12582912 shift=22
Booting paravirtualized kernel on vmi
Do you have any immediate ideas as to what the responsible config
option(s) might be? It may well be a configuration error, but if that's
the case then perhaps we could produce a CONFIG_VMI documentation patch
for that.
Here is the bug report:
https://bugs.gentoo.org/show_bug.cgi?id=250094
Here is the working config:
https://bugs.gentoo.org/attachment.cgi?id=174901&action=view
and here is the broken config:
https://bugs.gentoo.org/attachment.cgi?id=174870&action=view
If you don't have any immediate ideas, say so and hopefully Norman will
then be able to spend some time "bisecting" the config differences.
cheers
Daniel
On Wed, 2008-12-10 at 15:32 -0800, Daniel Drake wrote:
> Hi Zach,
>
> Thanks for your help on the VMI-related 2.6.27.x regression. I'm hoping
> to now pick your mind on a similar but unrelated issue:
>
> Norman has found that with certain kernel configurations, 2.6.27 (plain,
> as released by Linus) does not initialize VMI properly, witnessed as
> missing initialization messages in the kernel logs. Other configurations
> work fine. These are the messages that go missing:
>
> VMI: Found VMware, Inc. Hypervisor OPROM, API version 3.0, ROM
> version 1.0
> vmi: registering clock event vmi-timer. mult=12582912 shift=22
> Booting paravirtualized kernel on vmi
>
> Do you have any immediate ideas as to what the responsible config
> option(s) might be? It may well be a configuration error, but if that's
> the case then perhaps we could produce a CONFIG_VMI documentation patch
> for that.
Make sure you do a clean compile for each build - ideally use
KBUILD_OUTPUT=/some/dir and keep the kernel tree completely clean at all
times.
It doesn't look like a configuration error. Both configurations share
this:
CONFIG_PARAVIRT_GUEST=y
# CONFIG_XEN is not set
CONFIG_XEN_SAVE_RESTORE=y
CONFIG_VMI=y
# CONFIG_KVM_CLOCK is not set
# CONFIG_KVM_GUEST is not set
CONFIG_PARAVIRT=y
# CONFIG_PARAVIRT_CLOCK is not set
CONFIG_PARAVIRT_DEBUG=y
Which should be fine. However, when upgrading kernels, you should always sanitize your .config - just run make menuconfig, save and exit, and it will fix any deselected options.
Could be this .config isn't valid under 2.6.27.