On Fri, 2007-08-03 at 11:55 +1000, Michael Neuling wrote:
> We sometimes change the vmalloc segment in slb_flush_and_rebolt but we
> never updated with slb shadow buffer. This fixes it. Thanks to paulus
> for finding this.
>
> Also added some write barriers to ensure the shadow buffer is always
We sometimes change the vmalloc segment in slb_flush_and_rebolt but we
never updated with slb shadow buffer. This fixes it. Thanks to paulus
for finding this.
Also added some write barriers to ensure the shadow buffer is always
valid.
Signed-off-by: Michael Neuling <[EMAIL PROTECTED]>
---
Integ
> On Thu, 2007-08-02 at 19:03 +1000, Michael Neuling wrote:
> > >
> > > Ok, that was missing from your description :-)
> >
> > Sorry... so ditch the barriers?
>
> As you like. The reason why you can ditch them is purely because you
> know for sure that the only case the firmware will access tho
On Thu, 2007-08-02 at 19:03 +1000, Michael Neuling wrote:
> >
> > Ok, that was missing from your description :-)
>
> Sorry... so ditch the barriers?
As you like. The reason why you can ditch them is purely because you
know for sure that the only case the firmware will access those shadows
from
> On Thu, 2007-08-02 at 18:56 +1000, Michael Neuling wrote:
> > > > But even in the case of a checkpoint restart, the ordering will be
> > > > preserved as the NIA we get as part of the checkpoint will have
> > all
> > > > previous instructions complete and none of the following
> > instructions
>
On Thu, 2007-08-02 at 18:56 +1000, Michael Neuling wrote:
> > > But even in the case of a checkpoint restart, the ordering will be
> > > preserved as the NIA we get as part of the checkpoint will have
> all
> > > previous instructions complete and none of the following
> instructions
> > > started.
> >
> > But even in the case of a checkpoint restart, the ordering will be
> > preserved as the NIA we get as part of the checkpoint will have all
> > previous instructions complete and none of the following instructions
> > started.
>
> Instruction completion isn't enough to ensure storage order
On Thu, 2007-08-02 at 15:56 +1000, Michael Neuling wrote:
>
> But even in the case of a checkpoint restart, the ordering will be
> preserved as the NIA we get as part of the checkpoint will have all
> previous instructions complete and none of the following instructions
> started.
Instruction com
> > --- linux-2.6-ozlabs.orig/arch/powerpc/kernel/entry_64.S
> > +++ linux-2.6-ozlabs/arch/powerpc/kernel/entry_64.S
> > @@ -389,7 +389,9 @@ BEGIN_FTR_SECTION
> > ld r9,PACA_SLBSHADOWPTR(r13)
> > li r12,0
> > std r12,SLBSHADOW_STACKESID(r9) /* Clear ESID */
> > + eieio
>
> The first block of code detects the need for a demotion and changes the
> global mmu_vmalloc_psize, along with changing the value on the local CPU
> (and flushing it's SLB). But other CPUs still have the old values.
>
> The second block of code checks if the global value matches the per-CPU
> v
On Thu, 2007-08-02 at 09:32 +1000, Michael Neuling wrote:
> > On Wed, 2007-08-01 at 16:02 +1000, Michael Neuling wrote:
> > > We sometimes change the vmalloc segment in slb_flush_and_rebolt but we
> > > never updated with slb shadow buffer. This fixes it. Thanks to paulus
> > > for finding this.
> On Wed, 2007-08-01 at 16:02 +1000, Michael Neuling wrote:
> > We sometimes change the vmalloc segment in slb_flush_and_rebolt but we
> > never updated with slb shadow buffer. This fixes it. Thanks to paulus
> > for finding this.
> >
> > Also added some write barriers to ensure the shadow buffe
On Wed, 2007-08-01 at 16:02 +1000, Michael Neuling wrote:
> We sometimes change the vmalloc segment in slb_flush_and_rebolt but we
> never updated with slb shadow buffer. This fixes it. Thanks to paulus
> for finding this.
>
> Also added some write barriers to ensure the shadow buffer is always
On Wed, 2007-08-01 at 16:02 +1000, Michael Neuling wrote:
> --- linux-2.6-ozlabs.orig/arch/powerpc/kernel/entry_64.S
> +++ linux-2.6-ozlabs/arch/powerpc/kernel/entry_64.S
> @@ -389,7 +389,9 @@ BEGIN_FTR_SECTION
> ld r9,PACA_SLBSHADOWPTR(r13)
> li r12,0
> std r12,SLBS
We sometimes change the vmalloc segment in slb_flush_and_rebolt but we
never updated with slb shadow buffer. This fixes it. Thanks to paulus
for finding this.
Also added some write barriers to ensure the shadow buffer is always
valid.
Signed-off-by: Michael Neuling <[EMAIL PROTECTED]>
---
> >
Michael Neuling writes:
> + slb_shadow_update(mk_esid_data(VMALLOC_START, 1),
> + mk_vsid_data(VMALLOC_START, vflags), 1);
Could you re-jig slb_shadow_update to take ea, slot and vflags, and
call mk_[ev]sid_data itself, rather than exposing mk_esid_data a
We sometimes change the vmalloc segment in slb_flush_and_rebolt but we
never updated with slb shadow buffer. This fixes it. Thanks to paulus
for finding this.
Also added some write barriers to ensure the shadow buffer is always
valid.
Signed-off-by: Michael Neuling <[EMAIL PROTECTED]>
---
Paulu
17 matches
Mail list logo
