| Message ID | 1322487768-19439-1-git-send-email-daniel.schwierzeck@googlemail.com |
|---|---|
| State | Superseded |
| Headers | show |
On 12/4/11 9:02 PM, Daniel Schwierzeck wrote: >> diff --git a/boards.cfg b/boards.cfg >> index c83d861..2cd917e 100644 >> --- a/boards.cfg >> +++ b/boards.cfg >> @@ -324,13 +324,13 @@ dbau1000 mips mips32 dbau1x00 - >> dbau1100 mips mips32 dbau1x00 - au1x00 dbau1x00:DBAU1100 >> dbau1500 mips mips32 dbau1x00 - au1x00 dbau1x00:DBAU1500 >> dbau1550 mips mips32 dbau1x00 - au1x00 dbau1x00:DBAU1550 >> -dbau1550_el mips mips32 dbau1x00 - au1x00 dbau1x00:DBAU1550 >> +dbau1550_el mips mips32 dbau1x00 - au1x00 dbau1x00:DBAU1550,SYS_LITTLE_ENDIAN This looks Ok. >> gth2 mips mips32 - - au1x00 >> incaip mips mips32 incaip - incaip >> incaip_100MHz mips mips32 incaip - incaip incaip:CPU_CLOCK_RATE=100000000 >> incaip_133MHz mips mips32 incaip - incaip incaip:CPU_CLOCK_RATE=133000000 >> incaip_150MHz mips mips32 incaip - incaip incaip:CPU_CLOCK_RATE=150000000 >> -pb1000 mips mips32 pb1x00 - au1x00 pb1x00:PB1000 >> +pb1000 mips mips32 pb1x00 - au1x00 pb1x00:PB1000,SYS_LITTLE_ENDIAN >> qemu_mips mips mips32 qemu-mips - - qemu-mips >> tb0229 mips mips32 >> vct_premium mips mips32 vct micronas - vct:VCT_PREMIUM But I don't see any reason CONFIG_SYS_LITTLE_ENDIAN is specified in boards.cfg. Just putting it in configs/pb1000.h is enough, isn't it? > Shinya, do you have an opinion about this? No, if it works for you, I'm fine. Thanks for tackling this issue. Does anyone disagree with this change? If not, I'll pick this up. >> diff --git a/examples/standalone/Makefile b/examples/standalone/Makefile >> index e23865b..eab23b4 100644 >> --- a/examples/standalone/Makefile >> +++ b/examples/standalone/Makefile >> @@ -88,6 +88,10 @@ endif >> CFLAGS_NTR := $(call cc-option,-fno-toplevel-reorder) >> CFLAGS += $(CFLAGS_NTR) >> >> +# Pass through endianess settings in LDFLAGS to LD s/endianess/endianness/
Dear Daniel Schwierzeck, In message <4EDB5F86.40605@googlemail.com> you wrote: > > Wolfgang, is there any reason why standalone apps are linked without > LDFLAGS (especially PLATFORM_LDFLAGS)? None that I know, at least. In general, standalone applications are a rarely used special feature, and the few people who actually ever used this for a purpose probably used custom code and custom linker scripts, so the mainline code is eventually not that good supproted. > This patch fixes a general issue with standalone apps on archs > supporting both endianess types when running MAKEALL. I guess this could > be relevant for ARM too if U-Boot has support for big-endian ARM boards. It does, but I doubt anybody ever used standalone applications on these. Best regards, Wolfgang Denk
Dear Shinya Kuribayashi, In message <4EDB8475.1080305@pobox.com> you wrote: > > >> -pb1000 mips mips32 pb1x00 - au1x00 pb1x00:PB1000 > >> +pb1000 mips mips32 pb1x00 - au1x00 pb1x00:PB1000,SYS_LITTLE_ENDIAN ... > But I don't see any reason CONFIG_SYS_LITTLE_ENDIAN is specified in > boards.cfg. Just putting it in configs/pb1000.h is enough, isn't it? Agreed. In almost all cases this should be done in the board config file only. Only if there are cases where a board suppors both LE and BE configurations (is there any such board??) then we should need a SYS_*_ENDIAN setting in boards.cfg (but even then, only one - not two). Best regards, Wolfgang Denk
On 12/04/2011 04:06 PM, Wolfgang Denk wrote: > Dear Daniel Schwierzeck, > > In message<4EDB5F86.40605@googlemail.com> you wrote: >> >> Wolfgang, is there any reason why standalone apps are linked without >> LDFLAGS (especially PLATFORM_LDFLAGS)? > > None that I know, at least. > > In general, standalone applications are a rarely used special feature, > and the few people who actually ever used this for a purpose probably > used custom code and custom linker scripts, so the mainline code is > eventually not that good supproted. the problem is those apps are always compiled. Maybe we should introduce a config option to enable building only if someone really needs them? > >> This patch fixes a general issue with standalone apps on archs >> supporting both endianess types when running MAKEALL. I guess this could >> be relevant for ARM too if U-Boot has support for big-endian ARM boards. > > It does, but I doubt anybody ever used standalone applications on > these. > > Best regards, > > Wolfgang Denk >
On 12/04/2011 04:08 PM, Wolfgang Denk wrote: > Dear Shinya Kuribayashi, > > In message<4EDB8475.1080305@pobox.com> you wrote: >> >>>> -pb1000 mips mips32 pb1x00 - au1x00 pb1x00:PB1000 >>>> +pb1000 mips mips32 pb1x00 - au1x00 pb1x00:PB1000,SYS_LITTLE_ENDIAN > ... >> But I don't see any reason CONFIG_SYS_LITTLE_ENDIAN is specified in >> boards.cfg. Just putting it in configs/pb1000.h is enough, isn't it? > > Agreed. In almost all cases this should be done in the board config > file only. Only if there are cases where a board suppors both LE and > BE configurations (is there any such board??) then we should need a > SYS_*_ENDIAN setting in boards.cfg (but even then, only one - not > two). I did this because of dbau1550 and dbau1550_el. For the pb1000 I'll move it to the board config. > > Best regards, > > Wolfgang Denk >
Dear Daniel Schwierzeck, In message <4EDC0A99.6090106@googlemail.com> you wrote: > > > In general, standalone applications are a rarely used special feature, > > and the few people who actually ever used this for a purpose probably > > used custom code and custom linker scripts, so the mainline code is > > eventually not that good supproted. > > the problem is those apps are always compiled. Maybe we should introduce > a config option to enable building only if someone really needs them? As is, they are rarely used, so they tend to collect bit-rot. If we stop to even compile them, they wil rot much faster. This is not a good idea. Best regards, Wolfgang Denk
Dear Daniel Schwierzeck, > On 12/04/2011 04:08 PM, Wolfgang Denk wrote: > > Dear Shinya Kuribayashi, > > > > In message<4EDB8475.1080305@pobox.com> you wrote: > >>>> -pb1000 mips mips32 pb1x00 > >>>> - au1x00 pb1x00:PB1000 +pb1000 > >>>> mips mips32 pb1x00 - au1x00 > >>>> pb1x00:PB1000,SYS_LITTLE_ENDIAN > > > > ... > > > >> But I don't see any reason CONFIG_SYS_LITTLE_ENDIAN is specified in > >> boards.cfg. Just putting it in configs/pb1000.h is enough, isn't it? > > > > Agreed. In almost all cases this should be done in the board config > > file only. Only if there are cases where a board suppors both LE and > > BE configurations (is there any such board??) then we should need a > > SYS_*_ENDIAN setting in boards.cfg (but even then, only one - not > > two). > > I did this because of dbau1550 and dbau1550_el. For the pb1000 I'll > move it to the board config. > > > Best regards, > > > > Wolfgang Denk Will we see V2 of this patch please? Best regards, Marek Vasut
diff --git a/README b/README index 07f1d11..468bfdf 100644 --- a/README +++ b/README @@ -374,6 +374,12 @@ The following options need to be configured: Defines the string to utilize when trying to match PCIe device tree nodes for the given platform. +- Generic CPU options: + CONFIG_SYS_BIG_ENDIAN, CONFIG_SYS_LITTLE_ENDIAN + + Defines the endianess of the CPU. Implementation of those + values is arch specific. + - Intel Monahans options: CONFIG_SYS_MONAHANS_RUN_MODE_OSC_RATIO diff --git a/arch/mips/cpu/mips32/config.mk b/arch/mips/cpu/mips32/config.mk index 4d1b273..a1cd590 100644 --- a/arch/mips/cpu/mips32/config.mk +++ b/arch/mips/cpu/mips32/config.mk @@ -27,14 +27,23 @@ # Note: Toolchains with binutils prior to v2.16 # are no longer supported by U-Boot MIPS tree! # -MIPSFLAGS = -march=mips32r2 +MIPSFLAGS := -march=mips32r2 +# Handle special prefix in ELDK 4.0 toolchain ifneq (,$(findstring 4KCle,$(CROSS_COMPILE))) -ENDIANNESS = -EL -else -ENDIANNESS = -EB +ENDIANNESS := -EL endif -MIPSFLAGS += $(ENDIANNESS) +ifdef CONFIG_SYS_LITTLE_ENDIAN +ENDIANNESS := -EL +endif + +ifdef CONFIG_SYS_BIG_ENDIAN +ENDIANNESS := -EB +endif + +# Default to EB if no endianess is configured +ENDIANNESS ?= -EB -PLATFORM_CPPFLAGS += $(MIPSFLAGS) +PLATFORM_CPPFLAGS += $(MIPSFLAGS) $(ENDIANNESS) +PLATFORM_LDFLAGS += $(ENDIANNESS) diff --git a/boards.cfg b/boards.cfg index c83d861..2cd917e 100644 --- a/boards.cfg +++ b/boards.cfg @@ -324,13 +324,13 @@ dbau1000 mips mips32 dbau1x00 - dbau1100 mips mips32 dbau1x00 - au1x00 dbau1x00:DBAU1100 dbau1500 mips mips32 dbau1x00 - au1x00 dbau1x00:DBAU1500 dbau1550 mips mips32 dbau1x00 - au1x00 dbau1x00:DBAU1550 -dbau1550_el mips mips32 dbau1x00 - au1x00 dbau1x00:DBAU1550 +dbau1550_el mips mips32 dbau1x00 - au1x00 dbau1x00:DBAU1550,SYS_LITTLE_ENDIAN gth2 mips mips32 - - au1x00 incaip mips mips32 incaip - incaip incaip_100MHz mips mips32 incaip - incaip incaip:CPU_CLOCK_RATE=100000000 incaip_133MHz mips mips32 incaip - incaip incaip:CPU_CLOCK_RATE=133000000 incaip_150MHz mips mips32 incaip - incaip incaip:CPU_CLOCK_RATE=150000000 -pb1000 mips mips32 pb1x00 - au1x00 pb1x00:PB1000 +pb1000 mips mips32 pb1x00 - au1x00 pb1x00:PB1000,SYS_LITTLE_ENDIAN qemu_mips mips mips32 qemu-mips - - qemu-mips tb0229 mips mips32 vct_premium mips mips32 vct micronas - vct:VCT_PREMIUM diff --git a/examples/standalone/Makefile b/examples/standalone/Makefile index e23865b..eab23b4 100644 --- a/examples/standalone/Makefile +++ b/examples/standalone/Makefile @@ -88,6 +88,10 @@ endif CFLAGS_NTR := $(call cc-option,-fno-toplevel-reorder) CFLAGS += $(CFLAGS_NTR) +# Pass through endianess settings in LDFLAGS to LD +LDFLAGS_ENDIAN += $(filter -EB,$(LDFLAGS)) +LDFLAGS_ENDIAN += $(filter -EL,$(LDFLAGS)) + all: $(obj).depend $(OBJS) $(LIB) $(SREC) $(BIN) $(ELF) ######################################################################### @@ -96,7 +100,7 @@ $(LIB): $(obj).depend $(LIBOBJS) $(ELF): $(obj)%: $(obj)%.o $(LIB) - $(LD) -g -Ttext $(CONFIG_STANDALONE_LOAD_ADDR) \ + $(LD) $(LDFLAGS_ENDIAN) -g -Ttext $(CONFIG_STANDALONE_LOAD_ADDR) \ -o $@ -e $(SYM_PREFIX)$(notdir $(<:.o=)) $< $(LIB) \ -L$(gcclibdir) -lgcc
Make endianess of target CPU configurable. Use the new config option for dbau1550_el and pb1000 boards. Adapt linking of standalone applications to pass through endianess options to LD. Build tested with: - ELDK 4 mips_4KC- and mips4KCle - Sourcery CodeBench Lite 2011.03-93 Signed-off-by: Daniel Schwierzeck <daniel.schwierzeck@googlemail.com> Cc: Thomas Lange <thomas@corelatus.se> Cc: Mike Frysinger <vapier@gentoo.org> Cc: Wolfgang Denk <wd@denx.de> --- Another try to finally fix this originated by discussion: [1] http://thread.gmane.org/gmane.comp.boot-loaders.u-boot/118111/focus=118122 Related discussions: [2] http://thread.gmane.org/gmane.comp.boot-loaders.u-boot/81572 [3] http://thread.gmane.org/gmane.comp.boot-loaders.u-boot/45404 README | 6 ++++++ arch/mips/cpu/mips32/config.mk | 21 +++++++++++++++------ boards.cfg | 4 ++-- examples/standalone/Makefile | 6 +++++- 4 files changed, 28 insertions(+), 9 deletions(-)