| Message ID | 20220510022055.67582-1-joel@jms.id.au |
|---|---|
| State | Accepted |
| Headers | show |
| Series | linux: Fix powerpc64le defconfig selection | expand |
On 10/05/2022 04:20, Joel Stanley wrote: > The default defconfig target for the 64 bit powerpc kernel is > ppc64_defconfig, the big endian configuration. > > When building for powerpc64le users want the little endian kernel as > they can't boot LE userspace on a BE kernel. > > Fix up the defconfig used in this case. This will avoid the following > autobuilder failure: > > VDSO32A arch/powerpc/kernel/vdso32/sigtramp.o > cc1: error: ‘-m32’ not supported in this configuratioin > make[4]: *** [arch/powerpc/kernel/vdso32/Makefile:49: arch/powerpc/kernel/vdso32/sigtramp.o] Error 1 > > http://autobuild.buildroot.net/results/dd76d53bab56470c0b83e296872d7bb90f9e8296/ > > Note that the failure indicates the toolchain is configured to disable > the 32 bit target, causing the kernel to fail when building the 32 bit > VDSO. This is only a problem on the BE kernel as the LE kernel disables > CONFIG_COMPAT, aka 32 bit userspace support, by default. > > Signed-off-by: Joel Stanley <joel@jms.id.au> Applied to master, thanks. However, the defconfig mechanism for *all* powerpc seems pretty broken. Here's what we have in 5.16, before that there was something similar: # If we're on a ppc/ppc64/ppc64le machine use that defconfig, otherwise just use # ppc64_defconfig because we have nothing better to go on. uname := $(shell uname -m) KBUILD_DEFCONFIG := $(if $(filter ppc%,$(uname)),$(uname),ppc64)_defconfig So I guess we should use a specific defconfig for *all* powerpc. The arch-default defconfig is generally not really reliable, for example for arm it always takes v7_multi, but that won't work for v7m targets... Regards, Arnout > --- > Alternatively, we could disalow BR2_LINUX_KERNEL_USE_ARCH_DEFAULT_CONFIG > when BR2_powerpc64le=y. > > linux/linux.mk | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/linux/linux.mk b/linux/linux.mk > index b27e436b0d57..67832127d92e 100644 > --- a/linux/linux.mk > +++ b/linux/linux.mk > @@ -307,7 +307,11 @@ endif > ifeq ($(BR2_LINUX_KERNEL_USE_DEFCONFIG),y) > LINUX_KCONFIG_DEFCONFIG = $(call qstrip,$(BR2_LINUX_KERNEL_DEFCONFIG))_defconfig > else ifeq ($(BR2_LINUX_KERNEL_USE_ARCH_DEFAULT_CONFIG),y) > +ifeq ($(BR2_powerpc64le),y) > +LINUX_KCONFIG_DEFCONFIG = ppc64le_defconfig > +else > LINUX_KCONFIG_DEFCONFIG = defconfig > +endif > else ifeq ($(BR2_LINUX_KERNEL_USE_CUSTOM_CONFIG),y) > LINUX_KCONFIG_FILE = $(call qstrip,$(BR2_LINUX_KERNEL_CUSTOM_CONFIG_FILE)) > endif
On Fri, 13 May 2022 at 21:52, Arnout Vandecappelle <arnout@mind.be> wrote: > > > > On 10/05/2022 04:20, Joel Stanley wrote: > > The default defconfig target for the 64 bit powerpc kernel is > > ppc64_defconfig, the big endian configuration. > > > > When building for powerpc64le users want the little endian kernel as > > they can't boot LE userspace on a BE kernel. > > > > Fix up the defconfig used in this case. This will avoid the following > > autobuilder failure: > > > > VDSO32A arch/powerpc/kernel/vdso32/sigtramp.o > > cc1: error: ‘-m32’ not supported in this configuratioin > > make[4]: *** [arch/powerpc/kernel/vdso32/Makefile:49: arch/powerpc/kernel/vdso32/sigtramp.o] Error 1 > > > > http://autobuild.buildroot.net/results/dd76d53bab56470c0b83e296872d7bb90f9e8296/ > > > > Note that the failure indicates the toolchain is configured to disable > > the 32 bit target, causing the kernel to fail when building the 32 bit > > VDSO. This is only a problem on the BE kernel as the LE kernel disables > > CONFIG_COMPAT, aka 32 bit userspace support, by default. > > > > Signed-off-by: Joel Stanley <joel@jms.id.au> > > Applied to master, thanks. However, the defconfig mechanism for *all* powerpc > seems pretty broken. Here's what we have in 5.16, before that there was > something similar: > > # If we're on a ppc/ppc64/ppc64le machine use that defconfig, otherwise just use > # ppc64_defconfig because we have nothing better to go on. > uname := $(shell uname -m) > KBUILD_DEFCONFIG := $(if $(filter ppc%,$(uname)),$(uname),ppc64)_defconfig Yes, mpe and I discussed this. x86 does a similar thing: ifeq ($(ARCH),x86) ifeq ($(shell uname -m),x86_64) KBUILD_DEFCONFIG := x86_64_defconfig else KBUILD_DEFCONFIG := i386_defconfig endif else KBUILD_DEFCONFIG := $(ARCH)_defconfig endif If you were building on an eg. Arm or PowerPC host with ARCH=x86, you would get unexpected results. > > So I guess we should use a specific defconfig for *all* powerpc. > > The arch-default defconfig is generally not really reliable, for example for > arm it always takes v7_multi, but that won't work for v7m targets... I noticed that. It is similarly broken if you expected a kernel that worked on armv5. Should we disable BR2_LINUX_KERNEL_USE_ARCH_DEFAULT_CONFIG for platforms that don't have a single defconfig? The other option I considered was removing ARCH_DEFAULT_CONFIG and providing a default for platforms where we could suggest one. It would look something like this: --- a/linux/Config.in +++ b/linux/Config.in @@ -176,10 +176,15 @@ endchoice config BR2_LINUX_KERNEL_DEFCONFIG string "Defconfig name" depends on BR2_LINUX_KERNEL_USE_DEFCONFIG + default "ppc64le_defconfig" if BR2_powerpc_powerpc64le + default "ppc64_defconfig" if BR2_powerpc_powerpc64 + default "ppc_defconfig" if BR2_powerpc_powerpc + default "mutli_v5_defconfig" if BR2_ARM_CPU_ARMV5 + default "mutli_v7_defconfig" if BR2_ARM_CPU_ARMV7A + default "defconfig" help - Name of the kernel defconfig file to use, without the - trailing _defconfig. The defconfig is located in - arch/<arch>/configs in the kernel tree. + Name of the kernel defconfig file to use. The defconfig is located in + arch/<arch>/configs in the kernel tree It would need to change the option to include the _defconfig (in order to support plain old "defconfig"), or we could add some logic in the .mk to add _defconfig only if the name is not "defconfig".
On 16/05/2022 15:17, Michael Ellerman wrote: > Arnout Vandecappelle <arnout@mind.be> writes: >> On 10/05/2022 04:20, Joel Stanley wrote: >>> The default defconfig target for the 64 bit powerpc kernel is >>> ppc64_defconfig, the big endian configuration. >>> >>> When building for powerpc64le users want the little endian kernel as >>> they can't boot LE userspace on a BE kernel. >>> >>> Fix up the defconfig used in this case. This will avoid the following >>> autobuilder failure: >>> >>> VDSO32A arch/powerpc/kernel/vdso32/sigtramp.o >>> cc1: error: ‘-m32’ not supported in this configuratioin >>> make[4]: *** [arch/powerpc/kernel/vdso32/Makefile:49: arch/powerpc/kernel/vdso32/sigtramp.o] Error 1 >>> >>> http://autobuild.buildroot.net/results/dd76d53bab56470c0b83e296872d7bb90f9e8296/ >>> >>> Note that the failure indicates the toolchain is configured to disable >>> the 32 bit target, causing the kernel to fail when building the 32 bit >>> VDSO. This is only a problem on the BE kernel as the LE kernel disables >>> CONFIG_COMPAT, aka 32 bit userspace support, by default. >>> >>> Signed-off-by: Joel Stanley <joel@jms.id.au> >> >> Applied to master, thanks. However, the defconfig mechanism for *all* powerpc >> seems pretty broken. Here's what we have in 5.16, before that there was >> something similar: >> >> # If we're on a ppc/ppc64/ppc64le machine use that defconfig, otherwise just use >> # ppc64_defconfig because we have nothing better to go on. >> uname := $(shell uname -m) >> KBUILD_DEFCONFIG := $(if $(filter ppc%,$(uname)),$(uname),ppc64)_defconfig >> >> So I guess we should use a specific defconfig for *all* powerpc. >> >> The arch-default defconfig is generally not really reliable, for example for >> arm it always takes v7_multi, but that won't work for v7m targets... > > There's a fundamental problem that just the "arch" is not sufficient > detail when you're building a kernel. Yes, which is pretty much unavoidable. > Two CPUs that implement the same user-visible "arch" may differ enough > at the kernel level to require a different defconfig. > > Having said that I think we could handle this better in the powerpc > kernel. Other arches allow specifying a different value for ARCH, which > then is fed into the defconfig. I don't know if it's worth bothering with that. It certainly would not make our life easier, because it would mean we need to set ARCH correctly. If we can do that, we can just as well set the defconfig correctly. > That way you could at least pass ARCH=ppc/ppc64/ppc64le, and get an > appropriate defconfig. > > I'll work on some kernel changes for that. I think the most important thing is that it makes no sense to rely on uname when ARCH and/or CROSS_COMPILE are set. Regards, Arnout > > cheers
On 17/05/2022 08:13, Joel Stanley wrote: > On Fri, 13 May 2022 at 21:52, Arnout Vandecappelle <arnout@mind.be> wrote: [snip] >> So I guess we should use a specific defconfig for *all* powerpc. >> >> The arch-default defconfig is generally not really reliable, for example for >> arm it always takes v7_multi, but that won't work for v7m targets... > > I noticed that. It is similarly broken if you expected a kernel that > worked on armv5. > > Should we disable BR2_LINUX_KERNEL_USE_ARCH_DEFAULT_CONFIG for > platforms that don't have a single defconfig? > > The other option I considered was removing ARCH_DEFAULT_CONFIG and > providing a default for platforms where we could suggest one. It would > look something like this: > > --- a/linux/Config.in > +++ b/linux/Config.in > @@ -176,10 +176,15 @@ endchoice > config BR2_LINUX_KERNEL_DEFCONFIG > string "Defconfig name" > depends on BR2_LINUX_KERNEL_USE_DEFCONFIG > + default "ppc64le_defconfig" if BR2_powerpc_powerpc64le > + default "ppc64_defconfig" if BR2_powerpc_powerpc64 > + default "ppc_defconfig" if BR2_powerpc_powerpc > + default "mutli_v5_defconfig" if BR2_ARM_CPU_ARMV5 > + default "mutli_v7_defconfig" if BR2_ARM_CPU_ARMV7A > + default "defconfig" > help > - Name of the kernel defconfig file to use, without the > - trailing _defconfig. The defconfig is located in > - arch/<arch>/configs in the kernel tree. > + Name of the kernel defconfig file to use. The defconfig is located in > + arch/<arch>/configs in the kernel tree > > It would need to change the option to include the _defconfig (in order > to support plain old "defconfig"), or we could add some logic in the > .mk to add _defconfig only if the name is not "defconfig". I like that idea! It allows us to gradually complete the list of defconfigs with better defaults. We could also add a condition after the "defconfig" to allow it only for the cases which are known to actually work (not much more than aarch64 and s390x, I think...). Note that changing it so that _defconfig needs to be added is not acceptable, because it breaks existing configs too much. We can easily work around that by special-casing the "defconfig" string in the .mk file. Regards, Arnout
>>>>> "Joel" == Joel Stanley <joel@jms.id.au> writes: > The default defconfig target for the 64 bit powerpc kernel is > ppc64_defconfig, the big endian configuration. > When building for powerpc64le users want the little endian kernel as > they can't boot LE userspace on a BE kernel. > Fix up the defconfig used in this case. This will avoid the following > autobuilder failure: > VDSO32A arch/powerpc/kernel/vdso32/sigtramp.o > cc1: error: ‘-m32’ not supported in this configuratioin > make[4]: *** [arch/powerpc/kernel/vdso32/Makefile:49: > arch/powerpc/kernel/vdso32/sigtramp.o] Error 1 > http://autobuild.buildroot.net/results/dd76d53bab56470c0b83e296872d7bb90f9e8296/ > Note that the failure indicates the toolchain is configured to disable > the 32 bit target, causing the kernel to fail when building the 32 bit > VDSO. This is only a problem on the BE kernel as the LE kernel disables > CONFIG_COMPAT, aka 32 bit userspace support, by default. > Signed-off-by: Joel Stanley <joel@jms.id.au> > --- > Alternatively, we could disalow BR2_LINUX_KERNEL_USE_ARCH_DEFAULT_CONFIG > when BR2_powerpc64le=y. Committed to 2022.02.x, thanks.
diff --git a/linux/linux.mk b/linux/linux.mk index b27e436b0d57..67832127d92e 100644 --- a/linux/linux.mk +++ b/linux/linux.mk @@ -307,7 +307,11 @@ endif ifeq ($(BR2_LINUX_KERNEL_USE_DEFCONFIG),y) LINUX_KCONFIG_DEFCONFIG = $(call qstrip,$(BR2_LINUX_KERNEL_DEFCONFIG))_defconfig else ifeq ($(BR2_LINUX_KERNEL_USE_ARCH_DEFAULT_CONFIG),y) +ifeq ($(BR2_powerpc64le),y) +LINUX_KCONFIG_DEFCONFIG = ppc64le_defconfig +else LINUX_KCONFIG_DEFCONFIG = defconfig +endif else ifeq ($(BR2_LINUX_KERNEL_USE_CUSTOM_CONFIG),y) LINUX_KCONFIG_FILE = $(call qstrip,$(BR2_LINUX_KERNEL_CUSTOM_CONFIG_FILE)) endif
The default defconfig target for the 64 bit powerpc kernel is ppc64_defconfig, the big endian configuration. When building for powerpc64le users want the little endian kernel as they can't boot LE userspace on a BE kernel. Fix up the defconfig used in this case. This will avoid the following autobuilder failure: VDSO32A arch/powerpc/kernel/vdso32/sigtramp.o cc1: error: ‘-m32’ not supported in this configuratioin make[4]: *** [arch/powerpc/kernel/vdso32/Makefile:49: arch/powerpc/kernel/vdso32/sigtramp.o] Error 1 http://autobuild.buildroot.net/results/dd76d53bab56470c0b83e296872d7bb90f9e8296/ Note that the failure indicates the toolchain is configured to disable the 32 bit target, causing the kernel to fail when building the 32 bit VDSO. This is only a problem on the BE kernel as the LE kernel disables CONFIG_COMPAT, aka 32 bit userspace support, by default. Signed-off-by: Joel Stanley <joel@jms.id.au> --- Alternatively, we could disalow BR2_LINUX_KERNEL_USE_ARCH_DEFAULT_CONFIG when BR2_powerpc64le=y. linux/linux.mk | 4 ++++ 1 file changed, 4 insertions(+)