[GIT PULL] for-2.6.32/bug-fixes for 2.6.32 pvops tree.

[GIT PULL] for-2.6.32/bug-fixes for 2.6.32 pvops tree.

[Konrad Rzeszutek Wilk]

Hey Jeremy,

Please pull this git:

 git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen.git for-2.6.32/bug-fixes

It has three bootup-fixes that have gatted 2.6.32 tree from booting on some of
my machines:
 - hvc_console would blow up during bootup with the irq handler hitting a
   NULL pointer
 - On Intel i915, if one was to use dom0_mem=1500M the "flush" page would be
   incorrectly allocated in System RAM, and ioremap trying to use the PFN
   instead of the MFN ending disastrously. The fix is to reserve the
   System RAM E820 areas so the reserve API does not use it.
 - On IBM Summit machines (I've a x3850) it would end up using the native_apic_read
   causing a blowup. Back-ported how we are guard ourselves against this in 2.6.38

All of these fixes are in 2.6.39-rc0 (or on the way). The machines that boot now
are: IBM x3850, Intel DX58SO,, and G31M-ES2L/G31M-ES2L.

Konrad Rzeszutek Wilk (3):
      xen/hvc: Disable probe_irq_on/off from poking the hvc-console IRQ line.
      xen/e820: Don't mark balloon memory as E820_UNUSABLE when running as guest and fix overflow.
      xen/ioapic: Allow set_fixmap to set FIX_APIC_BASE to dummy mapping.

Zhang, Fengzhe (1):
      xen/setup: Inhibit resource API from using System RAM E820 gaps as PCI mem gaps.

 arch/x86/kernel/acpi/boot.c |    4 ----
 arch/x86/xen/mmu.c          |    4 ++++
 arch/x86/xen/setup.c        |    9 +++++++++
 drivers/char/hvc_xen.c      |    2 ++
 4 files changed, 15 insertions(+), 4 deletions(-)

P.S.
There is one more fix, which is a back-port of Stefano's patch consulting the memblock
(or e820_reserve_range) API when clearing the PMDs. That looks to be fixing some folks
builds, but I will let Stefano post that one.

Re: [GIT PULL] for-2.6.32/bug-fixes for 2.6.32 pvops tree.

[Stefano Stabellini]

On Thu, 17 Mar 2011, Konrad Rzeszutek Wilk wrote:
> Hey Jeremy,
> 
> Please pull this git:
> 
>  git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen.git for-2.6.32/bug-fixes
> 
> It has three bootup-fixes that have gatted 2.6.32 tree from booting on some of
> my machines:
>  - hvc_console would blow up during bootup with the irq handler hitting a
>    NULL pointer
>  - On Intel i915, if one was to use dom0_mem=1500M the "flush" page would be
>    incorrectly allocated in System RAM, and ioremap trying to use the PFN
>    instead of the MFN ending disastrously. The fix is to reserve the
>    System RAM E820 areas so the reserve API does not use it.
>  - On IBM Summit machines (I've a x3850) it would end up using the native_apic_read
>    causing a blowup. Back-ported how we are guard ourselves against this in 2.6.38
> 
> All of these fixes are in 2.6.39-rc0 (or on the way). The machines that boot now
> are: IBM x3850, Intel DX58SO,, and G31M-ES2L/G31M-ES2L.
> 
> Konrad Rzeszutek Wilk (3):
>       xen/hvc: Disable probe_irq_on/off from poking the hvc-console IRQ line.
>       xen/e820: Don't mark balloon memory as E820_UNUSABLE when running as guest and fix overflow.
>       xen/ioapic: Allow set_fixmap to set FIX_APIC_BASE to dummy mapping.
> 
> Zhang, Fengzhe (1):
>       xen/setup: Inhibit resource API from using System RAM E820 gaps as PCI mem gaps.
> 
>  arch/x86/kernel/acpi/boot.c |    4 ----
>  arch/x86/xen/mmu.c          |    4 ++++
>  arch/x86/xen/setup.c        |    9 +++++++++
>  drivers/char/hvc_xen.c      |    2 ++
>  4 files changed, 15 insertions(+), 4 deletions(-)
> 
> P.S.
> There is one more fix, which is a back-port of Stefano's patch consulting the memblock
> (or e820_reserve_range) API when clearing the PMDs. That looks to be fixing some folks
> builds, but I will let Stefano post that one.

I am just waiting for the upstream guys to ack my fix because I would
like to have the same fix in both trees.
I should be able to send a git pull request in the next few days.