| Message ID | 1453936230-7305-1-git-send-email-dann.frazier@canonical.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, Jan 27, 2016 at 04:10:30PM -0700, dann frazier wrote: > 16K page size is an optional feature of the architecture, and is not > supported by the X-Gene SoC family. > > Signed-off-by: dann frazier <dann.frazier@canonical.com> > --- > arch/arm64/Kconfig.platforms | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/arch/arm64/Kconfig.platforms b/arch/arm64/Kconfig.platforms > index 21074f6..b241290 100644 > --- a/arch/arm64/Kconfig.platforms > +++ b/arch/arm64/Kconfig.platforms > @@ -133,6 +133,7 @@ config ARCH_VEXPRESS > > config ARCH_XGENE > bool "AppliedMicro X-Gene SOC Family" > + depends on !ARM64_16K_PAGES > help > This enables support for AppliedMicro X-Gene SOC Family This would be the case on Juno as well. But maybe at some point the X-Gene family would gain a CPU with such support. Anyway, I would rather make 16K pages depend on EXPERT, make it a bit harder to enable. We've had a few questions recently about enabling it on CPUs that don't have such feature.
On Thu, Jan 28, 2016 at 10:34:41AM +0000, Catalin Marinas wrote: > On Wed, Jan 27, 2016 at 04:10:30PM -0700, dann frazier wrote: > > 16K page size is an optional feature of the architecture, and is not > > supported by the X-Gene SoC family. > > > > Signed-off-by: dann frazier <dann.frazier@canonical.com> > > --- > > arch/arm64/Kconfig.platforms | 1 + > > 1 file changed, 1 insertion(+) > > > > diff --git a/arch/arm64/Kconfig.platforms b/arch/arm64/Kconfig.platforms > > index 21074f6..b241290 100644 > > --- a/arch/arm64/Kconfig.platforms > > +++ b/arch/arm64/Kconfig.platforms > > @@ -133,6 +133,7 @@ config ARCH_VEXPRESS > > > > config ARCH_XGENE > > bool "AppliedMicro X-Gene SOC Family" > > + depends on !ARM64_16K_PAGES > > help > > This enables support for AppliedMicro X-Gene SOC Family > > This would be the case on Juno as well. But maybe at some point the > X-Gene family would gain a CPU with such support. > > Anyway, I would rather make 16K pages depend on EXPERT, make it a bit > harder to enable. We've had a few questions recently about enabling it > on CPUs that don't have such feature. I thought there was also a suggestion that we could fail gracefully in the EFI stub if we detected an unsupported page size? Will
On Thu, Jan 28, 2016 at 11:08:20AM +0000, Will Deacon wrote: > On Thu, Jan 28, 2016 at 10:34:41AM +0000, Catalin Marinas wrote: > > On Wed, Jan 27, 2016 at 04:10:30PM -0700, dann frazier wrote: > > > 16K page size is an optional feature of the architecture, and is not > > > supported by the X-Gene SoC family. > > > > > > Signed-off-by: dann frazier <dann.frazier@canonical.com> > > > --- > > > arch/arm64/Kconfig.platforms | 1 + > > > 1 file changed, 1 insertion(+) > > > > > > diff --git a/arch/arm64/Kconfig.platforms b/arch/arm64/Kconfig.platforms > > > index 21074f6..b241290 100644 > > > --- a/arch/arm64/Kconfig.platforms > > > +++ b/arch/arm64/Kconfig.platforms > > > @@ -133,6 +133,7 @@ config ARCH_VEXPRESS > > > > > > config ARCH_XGENE > > > bool "AppliedMicro X-Gene SOC Family" > > > + depends on !ARM64_16K_PAGES > > > help > > > This enables support for AppliedMicro X-Gene SOC Family > > > > This would be the case on Juno as well. But maybe at some point the > > X-Gene family would gain a CPU with such support. > > > > Anyway, I would rather make 16K pages depend on EXPERT, make it a bit > > harder to enable. We've had a few questions recently about enabling it > > on CPUs that don't have such feature. > > I thought there was also a suggestion that we could fail gracefully in > the EFI stub if we detected an unsupported page size? Yup. There are other things we could/should test, too. I believe Jeremy had a patch, but due to churn in that area it didn't get picked up. Mark.
On Wed, Jan 27, 2016 at 04:10:30PM -0700, dann frazier wrote: > 16K page size is an optional feature of the architecture, Technically, support for any particular page size and endianness configuration is an optional architectural feature. I don't think we can capture this information for each SoC family or specific SoC. The best thing we can do is early checks in the EFI stub (and potentially bootloaders) to check whether the kernel's configured page size and endianness is implemented. Thanks, Mark. > and is not supported by the X-Gene SoC family. > > Signed-off-by: dann frazier <dann.frazier@canonical.com> > --- > arch/arm64/Kconfig.platforms | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/arch/arm64/Kconfig.platforms b/arch/arm64/Kconfig.platforms > index 21074f6..b241290 100644 > --- a/arch/arm64/Kconfig.platforms > +++ b/arch/arm64/Kconfig.platforms > @@ -133,6 +133,7 @@ config ARCH_VEXPRESS > > config ARCH_XGENE > bool "AppliedMicro X-Gene SOC Family" > + depends on !ARM64_16K_PAGES > help > This enables support for AppliedMicro X-Gene SOC Family > > -- > 2.7.0 > > > _______________________________________________ > linux-arm-kernel mailing list > linux-arm-kernel@lists.infradead.org > http://lists.infradead.org/mailman/listinfo/linux-arm-kernel >
On 28/01/16 11:08, Will Deacon wrote: > On Thu, Jan 28, 2016 at 10:34:41AM +0000, Catalin Marinas wrote: >> On Wed, Jan 27, 2016 at 04:10:30PM -0700, dann frazier wrote: >>> 16K page size is an optional feature of the architecture, and is not >>> supported by the X-Gene SoC family. This enables support for AppliedMicro X-Gene SOC Family >> >> This would be the case on Juno as well. But maybe at some point the >> X-Gene family would gain a CPU with such support. >> >> Anyway, I would rather make 16K pages depend on EXPERT, make it a bit >> harder to enable. We've had a few questions recently about enabling it >> on CPUs that don't have such feature. > > I thought there was also a suggestion that we could fail gracefully in > the EFI stub if we detected an unsupported page size? Yes, there was. But then it was also recommended [1] to add a new stubbed function for both arm & arm64, where it got lost. May be we should revive it. https://lkml.org/lkml/2015/10/15/489 Thanks Suzuki
On 28 January 2016 at 12:27, Suzuki K. Poulose <Suzuki.Poulose@arm.com> wrote: > On 28/01/16 11:08, Will Deacon wrote: >> >> On Thu, Jan 28, 2016 at 10:34:41AM +0000, Catalin Marinas wrote: >>> >>> On Wed, Jan 27, 2016 at 04:10:30PM -0700, dann frazier wrote: >>>> >>>> 16K page size is an optional feature of the architecture, and is not >>>> supported by the X-Gene SoC family. > > > This enables support for AppliedMicro X-Gene SOC Family >>> >>> >>> This would be the case on Juno as well. But maybe at some point the >>> X-Gene family would gain a CPU with such support. >>> >>> Anyway, I would rather make 16K pages depend on EXPERT, make it a bit >>> harder to enable. We've had a few questions recently about enabling it >>> on CPUs that don't have such feature. >> >> >> I thought there was also a suggestion that we could fail gracefully in >> the EFI stub if we detected an unsupported page size? > > > Yes, there was. But then it was also recommended [1] to add a new stubbed > function > for both arm & arm64, where it got lost. May be we should revive it. > > https://lkml.org/lkml/2015/10/15/489 > I can pick this up
On 28/01/16 11:30, Ard Biesheuvel wrote: > On 28 January 2016 at 12:27, Suzuki K. Poulose <Suzuki.Poulose@arm.com> wrote: >> On 28/01/16 11:08, Will Deacon wrote: >>> I thought there was also a suggestion that we could fail gracefully in >>> the EFI stub if we detected an unsupported page size? >> >> >> Yes, there was. But then it was also recommended [1] to add a new stubbed >> function >> for both arm & arm64, where it got lost. May be we should revive it. >> >> https://lkml.org/lkml/2015/10/15/489 >> > > I can pick this up > Thanks Ard. Cheers Suzuki
On 01/28/2016 05:20 AM, Mark Rutland wrote: > On Thu, Jan 28, 2016 at 11:08:20AM +0000, Will Deacon wrote: >> On Thu, Jan 28, 2016 at 10:34:41AM +0000, Catalin Marinas wrote: >> I thought there was also a suggestion that we could fail gracefully in >> the EFI stub if we detected an unsupported page size? > > Yup. There are other things we could/should test, too. > > I believe Jeremy had a patch, but due to churn in that area it didn't > get picked up. I will clean-up and re-post that patch in the future too. It was fairly trivial so if someone feels the need for immediately... https://lkml.org/lkml/2015/10/14/948 IMHO, The main sticking point initially was that the correct place for the check was a moving target. That problem has been fixed, so it should be easy to move it.
diff --git a/arch/arm64/Kconfig.platforms b/arch/arm64/Kconfig.platforms index 21074f6..b241290 100644 --- a/arch/arm64/Kconfig.platforms +++ b/arch/arm64/Kconfig.platforms @@ -133,6 +133,7 @@ config ARCH_VEXPRESS config ARCH_XGENE bool "AppliedMicro X-Gene SOC Family" + depends on !ARM64_16K_PAGES help This enables support for AppliedMicro X-Gene SOC Family
16K page size is an optional feature of the architecture, and is not supported by the X-Gene SoC family. Signed-off-by: dann frazier <dann.frazier@canonical.com> --- arch/arm64/Kconfig.platforms | 1 + 1 file changed, 1 insertion(+)