| Message ID | 20161123130840.1877-1-npiggin@gmail.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
On Thu, Nov 24, 2016 at 12:08 AM, Nicholas Piggin <npiggin@gmail.com> wrote: > Question, are there any fundamental reasons we shouldn't use the ELFv2 > ABI to build big endian kernels if the compiler supports it? > Does this have implications w.r.t interfaces to 1. openfirmware/skiboot 2. glibc/vdso Keen to find out as well Balbir Singh.
On Thu, Nov 24, 2016 at 12:08:40AM +1100, Nicholas Piggin wrote: > Question, are there any fundamental reasons we shouldn't use the ELFv2 > ABI to build big endian kernels if the compiler supports it? No one uses ELFv2 for BE in production, and it isn't thoroughly tested at all, not even regularly tested. "Not supported", as far as GCC is concerned (or any of the distros AFAIK). There are no fundamental reasons of course, ABIs are largely just conventions, not laws of nature. Segher
On Wed, 23 Nov 2016 08:38:34 -0600 Segher Boessenkool <segher@kernel.crashing.org> wrote: > On Thu, Nov 24, 2016 at 12:08:40AM +1100, Nicholas Piggin wrote: > > Question, are there any fundamental reasons we shouldn't use the ELFv2 > > ABI to build big endian kernels if the compiler supports it? > > No one uses ELFv2 for BE in production, and it isn't thoroughly tested > at all, not even regularly tested. "Not supported", as far as GCC is > concerned (or any of the distros AFAIK). > > There are no fundamental reasons of course, ABIs are largely just > conventions, not laws of nature. That's a very good reason! I didn't think of that, I'll drop the idea. Thanks, Nick
On Thu, Nov 24, 2016 at 1:38 AM, Segher Boessenkool <segher@kernel.crashing.org> wrote: > On Thu, Nov 24, 2016 at 12:08:40AM +1100, Nicholas Piggin wrote: >> Question, are there any fundamental reasons we shouldn't use the ELFv2 >> ABI to build big endian kernels if the compiler supports it? > > No one uses ELFv2 for BE in production, and it isn't thoroughly tested > at all, not even regularly tested. "Not supported", as far as GCC is > concerned (or any of the distros AFAIK). Is this actually unsupported by gcc? The ppc64 musl libc port is ABI v2 only so they use it on BE too. Buildroot forces ABI v2 to be used for all of userspace when musl is selected as the libc for this reason so it's not completely used in the wild. It's still pretty niche though...
On Thu, Nov 24, 2016 at 04:37:50PM +1100, Oliver O'Halloran wrote: > On Thu, Nov 24, 2016 at 1:38 AM, Segher Boessenkool > <segher@kernel.crashing.org> wrote: > > On Thu, Nov 24, 2016 at 12:08:40AM +1100, Nicholas Piggin wrote: > >> Question, are there any fundamental reasons we shouldn't use the ELFv2 > >> ABI to build big endian kernels if the compiler supports it? > > > > No one uses ELFv2 for BE in production, and it isn't thoroughly tested > > at all, not even regularly tested. "Not supported", as far as GCC is > > concerned (or any of the distros AFAIK). > > Is this actually unsupported by gcc? It may or may not work. We of course try to keep it working, or make it work if it doesn't now. But it isn't regularly tested, and it isn't a target that is considered for the release criteria (see https://gcc.gnu.org/gcc-7/criteria.html -- powerpc64{,le}-linux, i.e. ABIv1 for BE, ABIv2 for LE). It is supported as in, "it is a feature a user can enable". It is unsupported as in, "we do not know if it works". > The ppc64 musl libc port is ABI v2 only so they use it on BE too. Interesting! I've never seen a PowerPC musl in the wild though. > Buildroot forces ABI v2 to be used > for all of userspace when musl is selected as the libc for this reason > so it's not completely used in the wild. It's still pretty niche > though... The point remains that the kernel will be the only non-niche user of ABIv2 on BE if they choose to use it. I don't expect many problems, but it is uncharted territory. YMMV, etc. We (GCC) of course will be happy if you (Linux) wants to be guinea pig for this. But don't say we didn't warn you ;-) Segher
Segher Boessenkool <segher@kernel.crashing.org> writes: > On Thu, Nov 24, 2016 at 04:37:50PM +1100, Oliver O'Halloran wrote: >> On Thu, Nov 24, 2016 at 1:38 AM, Segher Boessenkool >> <segher@kernel.crashing.org> wrote: >> > On Thu, Nov 24, 2016 at 12:08:40AM +1100, Nicholas Piggin wrote: >> >> Question, are there any fundamental reasons we shouldn't use the ELFv2 >> >> ABI to build big endian kernels if the compiler supports it? >> > >> > No one uses ELFv2 for BE in production, and it isn't thoroughly tested >> > at all, not even regularly tested. "Not supported", as far as GCC is >> > concerned (or any of the distros AFAIK). >> >> Is this actually unsupported by gcc? > > It may or may not work. We of course try to keep it working, or make > it work if it doesn't now. But it isn't regularly tested, and it isn't > a target that is considered for the release criteria (see > https://gcc.gnu.org/gcc-7/criteria.html -- powerpc64{,le}-linux, i.e. > ABIv1 for BE, ABIv2 for LE). It doesn't actually say that though. It just says powerpc64-unknown-linux-gnu. So how is someone, say the musl folks, supposed to know that BE ABIv2 is not supported? cheers
On Fri, Nov 25, 2016 at 09:22:16AM +1100, Michael Ellerman wrote: > >> >> Question, are there any fundamental reasons we shouldn't use the ELFv2 > >> >> ABI to build big endian kernels if the compiler supports it? > >> > > >> > No one uses ELFv2 for BE in production, and it isn't thoroughly tested > >> > at all, not even regularly tested. "Not supported", as far as GCC is > >> > concerned (or any of the distros AFAIK). > >> > >> Is this actually unsupported by gcc? > > > > It may or may not work. We of course try to keep it working, or make > > it work if it doesn't now. But it isn't regularly tested, and it isn't > > a target that is considered for the release criteria (see > > https://gcc.gnu.org/gcc-7/criteria.html -- powerpc64{,le}-linux, i.e. > > ABIv1 for BE, ABIv2 for LE). > > It doesn't actually say that though. It just says > powerpc64-unknown-linux-gnu. So how is someone, say the musl folks, > supposed to know that BE ABIv2 is not supported? Because their target is powerpc64*-*-linux-musl instead? It is not on the release criteria list, it is not something we make any claims about. How would you know -m32 -mlittle is not well tested at all? It is in much the same boat: unusual combinations of options, and unusual configurations, are not well tested. You have to build a separate C library just to get started with it, that should tell you there are some rough waters ahead! Which isn't to say you should not do this -- just think twice before doing so. And wear a life vest. Segher
On Thu, 24 Nov 2016 17:17:16 -0600 Segher Boessenkool <segher@kernel.crashing.org> wrote: > On Fri, Nov 25, 2016 at 09:22:16AM +1100, Michael Ellerman wrote: > > >> >> Question, are there any fundamental reasons we shouldn't use the ELFv2 > > >> >> ABI to build big endian kernels if the compiler supports it? > > >> > > > >> > No one uses ELFv2 for BE in production, and it isn't thoroughly tested > > >> > at all, not even regularly tested. "Not supported", as far as GCC is > > >> > concerned (or any of the distros AFAIK). > > >> > > >> Is this actually unsupported by gcc? > > > > > > It may or may not work. We of course try to keep it working, or make > > > it work if it doesn't now. But it isn't regularly tested, and it isn't > > > a target that is considered for the release criteria (see > > > https://gcc.gnu.org/gcc-7/criteria.html -- powerpc64{,le}-linux, i.e. > > > ABIv1 for BE, ABIv2 for LE). > > > > It doesn't actually say that though. It just says > > powerpc64-unknown-linux-gnu. So how is someone, say the musl folks, > > supposed to know that BE ABIv2 is not supported? > > Because their target is powerpc64*-*-linux-musl instead? It is not on > the release criteria list, it is not something we make any claims about. > > How would you know -m32 -mlittle is not well tested at all? It is in much > the same boat: unusual combinations of options, and unusual configurations, > are not well tested. You have to build a separate C library just to get > started with it, that should tell you there are some rough waters ahead! > > Which isn't to say you should not do this -- just think twice before > doing so. And wear a life vest. Other than curiosity, only two reasons for the kernel to enable it: either it helps end users, or it allows us to get rid of elfv1 support (eventually). Both would require gcc to have some base amount of testing. So unless an interested BE user pops up, it remains a curiosity. Thanks, Nick
Segher Boessenkool <segher@kernel.crashing.org> writes: > On Fri, Nov 25, 2016 at 09:22:16AM +1100, Michael Ellerman wrote: >> >> >> Question, are there any fundamental reasons we shouldn't use the ELFv2 >> >> >> ABI to build big endian kernels if the compiler supports it? >> >> > >> >> > No one uses ELFv2 for BE in production, and it isn't thoroughly tested >> >> > at all, not even regularly tested. "Not supported", as far as GCC is >> >> > concerned (or any of the distros AFAIK). >> >> >> >> Is this actually unsupported by gcc? >> > >> > It may or may not work. We of course try to keep it working, or make >> > it work if it doesn't now. But it isn't regularly tested, and it isn't >> > a target that is considered for the release criteria (see >> > https://gcc.gnu.org/gcc-7/criteria.html -- powerpc64{,le}-linux, i.e. >> > ABIv1 for BE, ABIv2 for LE). >> >> It doesn't actually say that though. It just says >> powerpc64-unknown-linux-gnu. So how is someone, say the musl folks, >> supposed to know that BE ABIv2 is not supported? > > Because their target is powerpc64*-*-linux-musl instead? It is not on > the release criteria list, it is not something we make any claims about. > > How would you know -m32 -mlittle is not well tested at all? It is in much > the same boat: unusual combinations of options, and unusual configurations, > are not well tested. You have to build a separate C library just to get > started with it, that should tell you there are some rough waters ahead! Yeah OK, I guess that's the part that's confusing me. Because we're hairy chested kernel programmers we don't actually build a C library at all. So we don't get that hint that we're using an odd combination, we just configure the toolchain, build it and off we go. cheers
Nicholas Piggin <npiggin@gmail.com> writes: > On Thu, 24 Nov 2016 17:17:16 -0600 > Segher Boessenkool <segher@kernel.crashing.org> wrote: > >> On Fri, Nov 25, 2016 at 09:22:16AM +1100, Michael Ellerman wrote: >> > >> >> Question, are there any fundamental reasons we shouldn't use the ELFv2 >> > >> >> ABI to build big endian kernels if the compiler supports it? >> > >> > >> > >> > No one uses ELFv2 for BE in production, and it isn't thoroughly tested >> > >> > at all, not even regularly tested. "Not supported", as far as GCC is >> > >> > concerned (or any of the distros AFAIK). >> > >> >> > >> Is this actually unsupported by gcc? >> > > >> > > It may or may not work. We of course try to keep it working, or make >> > > it work if it doesn't now. But it isn't regularly tested, and it isn't >> > > a target that is considered for the release criteria (see >> > > https://gcc.gnu.org/gcc-7/criteria.html -- powerpc64{,le}-linux, i.e. >> > > ABIv1 for BE, ABIv2 for LE). >> > >> > It doesn't actually say that though. It just says >> > powerpc64-unknown-linux-gnu. So how is someone, say the musl folks, >> > supposed to know that BE ABIv2 is not supported? >> >> Because their target is powerpc64*-*-linux-musl instead? It is not on >> the release criteria list, it is not something we make any claims about. >> >> How would you know -m32 -mlittle is not well tested at all? It is in much >> the same boat: unusual combinations of options, and unusual configurations, >> are not well tested. You have to build a separate C library just to get >> started with it, that should tell you there are some rough waters ahead! >> >> Which isn't to say you should not do this -- just think twice before >> doing so. And wear a life vest. > > Other than curiosity, only two reasons for the kernel to enable it: > either it helps end users, or it allows us to get rid of elfv1 support > (eventually). Both would require gcc to have some base amount of testing. I think it might be worth adding as option, as long as it's not too intrusive. We could then put it in our CI and at least keep an eye on whether it continues to work. Using ABIv1 does have user visible effects, ie. dot symbols show up in traces and so on. So getting rid of that would be nice. Having said that, perf and other tools are currently built to assume BE==ABIv1, so breaking that assumption would possibly cause more trouble. I guess the other question is when did ABIv2 land in the toolchain, ie. how many years do we have to wait until we can mandate it. cheers
On Fri, 25 Nov 2016 13:08:19 +1100 Michael Ellerman <mpe@ellerman.id.au> wrote: > Nicholas Piggin <npiggin@gmail.com> writes: > > > On Thu, 24 Nov 2016 17:17:16 -0600 > > Segher Boessenkool <segher@kernel.crashing.org> wrote: > > > >> On Fri, Nov 25, 2016 at 09:22:16AM +1100, Michael Ellerman wrote: > >> > >> >> Question, are there any fundamental reasons we shouldn't use the ELFv2 > >> > >> >> ABI to build big endian kernels if the compiler supports it? > >> > >> > > >> > >> > No one uses ELFv2 for BE in production, and it isn't thoroughly tested > >> > >> > at all, not even regularly tested. "Not supported", as far as GCC is > >> > >> > concerned (or any of the distros AFAIK). > >> > >> > >> > >> Is this actually unsupported by gcc? > >> > > > >> > > It may or may not work. We of course try to keep it working, or make > >> > > it work if it doesn't now. But it isn't regularly tested, and it isn't > >> > > a target that is considered for the release criteria (see > >> > > https://gcc.gnu.org/gcc-7/criteria.html -- powerpc64{,le}-linux, i.e. > >> > > ABIv1 for BE, ABIv2 for LE). > >> > > >> > It doesn't actually say that though. It just says > >> > powerpc64-unknown-linux-gnu. So how is someone, say the musl folks, > >> > supposed to know that BE ABIv2 is not supported? > >> > >> Because their target is powerpc64*-*-linux-musl instead? It is not on > >> the release criteria list, it is not something we make any claims about. > >> > >> How would you know -m32 -mlittle is not well tested at all? It is in much > >> the same boat: unusual combinations of options, and unusual configurations, > >> are not well tested. You have to build a separate C library just to get > >> started with it, that should tell you there are some rough waters ahead! > >> > >> Which isn't to say you should not do this -- just think twice before > >> doing so. And wear a life vest. > > > > Other than curiosity, only two reasons for the kernel to enable it: > > either it helps end users, or it allows us to get rid of elfv1 support > > (eventually). Both would require gcc to have some base amount of testing. > > I think it might be worth adding as option, as long as it's not too > intrusive. We could then put it in our CI and at least keep an eye on > whether it continues to work. If you're willing to take it, sure. I'll resubmit it with a default-n config option hidden away somewhere. > > Using ABIv1 does have user visible effects, ie. dot symbols show up in > traces and so on. So getting rid of that would be nice. > > Having said that, perf and other tools are currently built to assume > BE==ABIv1, so breaking that assumption would possibly cause more > trouble. We can make it clear it's experimental/not supported, but things like that could probably be cleaned up slowly. They should depend on elf version rather than endian anyway. > > I guess the other question is when did ABIv2 land in the toolchain, ie. > how many years do we have to wait until we can mandate it. I think gcc 4.9, binutils 2.24, early 2014. But if we've never been testing with those older toolchains we're a bit behind the 8 ball. Still, the second best time to plant the tree is now... Thanks, Nick
Nicholas Piggin <npiggin@gmail.com> writes: > On Fri, 25 Nov 2016 13:08:19 +1100 > Michael Ellerman <mpe@ellerman.id.au> wrote: > >> Nicholas Piggin <npiggin@gmail.com> writes: >> >> > On Thu, 24 Nov 2016 17:17:16 -0600 >> > Segher Boessenkool <segher@kernel.crashing.org> wrote: >> > >> >> On Fri, Nov 25, 2016 at 09:22:16AM +1100, Michael Ellerman wrote: >> >> > >> >> Question, are there any fundamental reasons we shouldn't use the ELFv2 >> >> > >> >> ABI to build big endian kernels if the compiler supports it? >> >> > >> > >> >> > >> > No one uses ELFv2 for BE in production, and it isn't thoroughly tested >> >> > >> > at all, not even regularly tested. "Not supported", as far as GCC is >> >> > >> > concerned (or any of the distros AFAIK). >> >> > >> >> >> > >> Is this actually unsupported by gcc? >> >> > > >> >> > > It may or may not work. We of course try to keep it working, or make >> >> > > it work if it doesn't now. But it isn't regularly tested, and it isn't >> >> > > a target that is considered for the release criteria (see >> >> > > https://gcc.gnu.org/gcc-7/criteria.html -- powerpc64{,le}-linux, i.e. >> >> > > ABIv1 for BE, ABIv2 for LE). >> >> > >> >> > It doesn't actually say that though. It just says >> >> > powerpc64-unknown-linux-gnu. So how is someone, say the musl folks, >> >> > supposed to know that BE ABIv2 is not supported? >> >> >> >> Because their target is powerpc64*-*-linux-musl instead? It is not on >> >> the release criteria list, it is not something we make any claims about. >> >> >> >> How would you know -m32 -mlittle is not well tested at all? It is in much >> >> the same boat: unusual combinations of options, and unusual configurations, >> >> are not well tested. You have to build a separate C library just to get >> >> started with it, that should tell you there are some rough waters ahead! >> >> >> >> Which isn't to say you should not do this -- just think twice before >> >> doing so. And wear a life vest. >> > >> > Other than curiosity, only two reasons for the kernel to enable it: >> > either it helps end users, or it allows us to get rid of elfv1 support >> > (eventually). Both would require gcc to have some base amount of testing. >> >> I think it might be worth adding as option, as long as it's not too >> intrusive. We could then put it in our CI and at least keep an eye on >> whether it continues to work. > > If you're willing to take it, sure. I'll resubmit it with a default-n > config option hidden away somewhere. I think so, unless we find that the size of the patch grows once we start testing it more. >> Using ABIv1 does have user visible effects, ie. dot symbols show up in >> traces and so on. So getting rid of that would be nice. >> >> Having said that, perf and other tools are currently built to assume >> BE==ABIv1, so breaking that assumption would possibly cause more >> trouble. > > We can make it clear it's experimental/not supported, but things like > that could probably be cleaned up slowly. They should depend on elf > version rather than endian anyway. They should, but that assumes you have the vmlinux ELF, or some other way to query the ABI of the kernel which may not always be the case. But it's probably solvable. >> I guess the other question is when did ABIv2 land in the toolchain, ie. >> how many years do we have to wait until we can mandate it. > > I think gcc 4.9, binutils 2.24, early 2014. > > But if we've never been testing with those older toolchains we're a bit > behind the 8 ball. Still, the second best time to plant the tree is now... True and true. cheers
On Fri, Nov 25, 2016 at 01:02:58PM +1100, Michael Ellerman wrote: > Because we're hairy chested kernel programmers we don't actually build a > C library at all. So we don't get that hint that we're using an odd > combination, we just configure the toolchain, build it and off we go. Right, I keep forgetting that kernel developers never read any documentation either :-P Segher
diff --git a/arch/powerpc/Makefile b/arch/powerpc/Makefile index 902da6e..b4867fc 100644 --- a/arch/powerpc/Makefile +++ b/arch/powerpc/Makefile @@ -72,18 +72,19 @@ GNUTARGET := powerpc MULTIPLEWORD := -mmultiple endif +ifeq ($(CONFIG_PPC64),y) +cflags-y += $(call cc-option,-mabi=elfv2) +aflags-y += $(call cc-option,-mabi=elfv2) +endif + cflags-$(CONFIG_CPU_BIG_ENDIAN) += $(call cc-option,-mbig-endian) -cflags-$(CONFIG_CPU_BIG_ENDIAN) += $(call cc-option,-mabi=elfv1) cflags-$(CONFIG_CPU_LITTLE_ENDIAN) += -mlittle-endian -cflags-$(CONFIG_CPU_LITTLE_ENDIAN) += -mabi=elfv2 ifneq ($(cc-name),clang) cflags-$(CONFIG_CPU_LITTLE_ENDIAN) += -mno-strict-align endif aflags-$(CONFIG_CPU_BIG_ENDIAN) += $(call cc-option,-mbig-endian) -aflags-$(CONFIG_CPU_BIG_ENDIAN) += $(call cc-option,-mabi=elfv1) aflags-$(CONFIG_CPU_LITTLE_ENDIAN) += -mlittle-endian -aflags-$(CONFIG_CPU_LITTLE_ENDIAN) += -mabi=elfv2 ifeq ($(HAS_BIARCH),y) override AS += -a$(BITS) @@ -113,14 +114,8 @@ endif endif CFLAGS-$(CONFIG_PPC64) := $(call cc-option,-mtraceback=no) -ifeq ($(CONFIG_CPU_LITTLE_ENDIAN),y) CFLAGS-$(CONFIG_PPC64) += $(call cc-option,-mabi=elfv2,$(call cc-option,-mcall-aixdesc)) AFLAGS-$(CONFIG_PPC64) += $(call cc-option,-mabi=elfv2) -else -CFLAGS-$(CONFIG_PPC64) += $(call cc-option,-mcall-aixdesc) -CFLAGS-$(CONFIG_PPC64) += $(call cc-option,-mabi=elfv1) -AFLAGS-$(CONFIG_PPC64) += $(call cc-option,-mabi=elfv1) -endif CFLAGS-$(CONFIG_PPC64) += $(call cc-option,-mcmodel=medium,$(call cc-option,-mminimal-toc)) CFLAGS-$(CONFIG_PPC64) += $(call cc-option,-mno-pointers-to-nested-functions) CFLAGS-$(CONFIG_PPC32) := -ffixed-r2 $(MULTIPLEWORD) diff --git a/arch/powerpc/boot/Makefile b/arch/powerpc/boot/Makefile index 1d50d41..ef2499b 100644 --- a/arch/powerpc/boot/Makefile +++ b/arch/powerpc/boot/Makefile @@ -33,7 +33,7 @@ BOOTCFLAGS += -m64 endif ifdef CONFIG_CPU_BIG_ENDIAN BOOTCFLAGS += -mbig-endian -BOOTCFLAGS += $(call cc-option,-mabi=elfv1) +BOOTCFLAGS += $(call cc-option,-mabi=elfv2) else BOOTCFLAGS += -mlittle-endian BOOTCFLAGS += $(call cc-option,-mabi=elfv2)