arch: Add blackfin CPU choice.
PATCH v2 1/7 arch: Add blackfin CPU choice.
From: Sonic Zhang <[email protected]>
v2-changes:
none
v1-changes:
- is patch adds a Target CPU configuration option and uses it to select a -m option for gcc.
--- arch/Config.in.bfin | 97 +++++++++++++++++++++++++++++++++++++++++++++++---- 1 files changed, 90 insertions(+), 7 deletions(-)
arch/Config.in.bfin +90 -7
@@ -1,12 +1,65 @@ choice - prompt "Target ABI" + prompt "Target CPU" depends on BR2_bfin - default BR2_BFIN_FDPIC -config BR2_BFIN_FDPIC - bool "FDPIC" -config BR2_BFIN_FLAT - bool "FLAT" - select BR2_PREFER_STATIC_LIB + default BR2_bf609 + help + Specify target CPU +config BR2_bf606 + bool "bf606" +config BR2_bf607 + bool "bf607" +config BR2_bf608 + bool "bf608" +config BR2_bf609 + bool "bf609" +config BR2_bf512 + bool "bf512" +config BR2_bf514 + bool "bf514" +config BR2_bf516 + bool "bf516" +config BR2_bf518 + bool "bf518" +config BR2_bf522 + bool "bf522" +config BR2_bf523 + bool "bf523" +config BR2_bf524 + bool "bf524" +config BR2_bf525 + bool "bf525" +config BR2_bf526 + bool "bf526" +config BR2_bf527 + bool "bf527" +config BR2_bf531 + bool "bf531" +config BR2_bf532 + bool "bf532" +config BR2_bf533 + bool "bf533" +config BR2_bf534 + bool "bf534" +config BR2_bf536 + bool "bf536" +config BR2_bf537 + bool "bf537" +config BR2_bf538 + bool "bf538" +config BR2_bf539 + bool "bf539" +config BR2_bf542 + bool "bf542" +config BR2_bf544 + bool "bf544" +config BR2_bf547 + bool "bf547" +config BR2_bf548 + bool "bf548" +config BR2_bf549 + bool "bf549" +config BR2_bf561 + bool "bf561" endchoice config BR2_ARCH
@@ -14,3 +67,33 @@ config BR2_ARCH config BR2_ENDIAN default "LITTLE" + +config BR2_GCC_TARGET_CPU + default bf606 if BR2_bf606 + default bf607 if BR2_bf607 + default bf608 if BR2_bf608 + default bf609 if BR2_bf609 + default bf512 if BR2_bf512 + default bf514 if BR2_bf514 + default bf516 if BR2_bf516 + default bf518 if BR2_bf518 + default bf522 if BR2_bf522 + default bf523 if BR2_bf523 + default bf524 if BR2_bf524 + default bf525 if BR2_bf525 + default bf526 if BR2_bf526 + default bf527 if BR2_bf527 + default bf531 if BR2_bf531 + default bf532 if BR2_bf532 + default bf533 if BR2_bf533 + default bf534 if BR2_bf534 + default bf536 if BR2_bf536 + default bf537 if BR2_bf537 + default bf538 if BR2_bf538 + default bf539 if BR2_bf539 + default bf542 if BR2_bf542 + default bf544 if BR2_bf544 + default bf547 if BR2_bf547 + default bf548 if BR2_bf548 + default bf549 if BR2_bf549 + default bf561 if BR2_bf561
Signature · 1 line
1.7.0.4
PATCH v2 2/7 arch: toolchain: Introduce target CPU revision.
From: Sonic Zhang <[email protected]>
v2-changes:
none
v1-changes:
- Adds the possibility to have a free-form CPU revision string and append it to the target CPU. Only Blackfin actually uses this option.
--- arch/Config.in | 3 +++ arch/Config.in.bfin | 5 +++++ toolchain/toolchain-external/ext-tool.mk | 4 ++++ 3 files changed, 12 insertions(+), 0 deletions(-)
arch/Config.in +3 -0
@@ -174,6 +174,9 @@ config BR2_GCC_TARGET_ABI config BR2_GCC_TARGET_CPU string +config BR2_GCC_TARGET_CPU_REVISION + string + if BR2_arm || BR2_armeb source "arch/Config.in.arm" endif
arch/Config.in.bfin +5 -0
@@ -97,3 +97,8 @@ config BR2_GCC_TARGET_CPU default bf548 if BR2_bf548 default bf549 if BR2_bf549 default bf561 if BR2_bf561 + +config BR2_GCC_TARGET_CPU_REVISION + string "Target CPU revision" + help + Specify target cpu revision which is passed to compiler as a parameter.
toolchain/toolchain-external/ext-tool.mk +4 -0
@@ -134,7 +134,11 @@ TOOLCHAIN_EXTERNAL_WRAPPER_ARGS += \ endif CC_TARGET_TUNE_:=$(call qstrip,$(BR2_GCC_TARGET_TUNE)) +ifeq ($(BR2_GCC_TARGET_CPU_REVISION),) CC_TARGET_CPU_:=$(call qstrip,$(BR2_GCC_TARGET_CPU)) +else +CC_TARGET_CPU_:=$(call qstrip,$(BR2_GCC_TARGET_CPU)-$(BR2_GCC_TARGET_CPU_REVISION)) +endif CC_TARGET_ARCH_:=$(call qstrip,$(BR2_GCC_TARGET_ARCH)) CC_TARGET_ABI_:=$(call qstrip,$(BR2_GCC_TARGET_ABI))
Signature · 1 line
1.7.0.4
PATCH v2 2/7 arch: toolchain: Introduce target CPU revision.
Dear Sonic Zhang,
arch/Config.in.bfin config BR2_GCC_TARGET_CPU · line 104
Quoted 4 lines · Sonic Zhang
arch/Config.in.bfin +4
-0+config BR2_GCC_TARGET_CPU_REVISION + string "Target CPU revision" + help + Specify target cpu revision which is passed to compiler as a parameter.
I think this line is slightly too long, it should be wrapped. And I would like more details here in the help text. Maybe something like:
Specify a target CPU revision, which will be appended to the value of the -mcpu option. For example, if the selected CPU is bf607, and then selected CPU revision is "r2", then gcc will receive the -mcpu=bf607-r2 option.
Other than this detail, looks good to me.
Thanks!
Thomas
Signature · 4 lines
Thomas Petazzoni, Free Electrons Kernel, drivers, real-time and embedded Linux development, consulting, training and support. http://free-electrons.com
PATCH v2 2/7 arch: toolchain: Introduce target CPU revision.
Hi Thomas,
Quoted 9 lines · Thomas Petazzoni
Dear Sonic Zhang,
arch/Config.in.bfin config BR2_GCC_TARGET_CPU · line 104
Quoted 4 lines · Sonic Zhang
arch/Config.in.bfin +4
-0+config BR2_GCC_TARGET_CPU_REVISION + string "Target CPU revision" + help + Specify target cpu revision which is passed to compiler as a parameter.
Quoted 8 lines
I think this line is slightly too long, it should be wrapped. And I would like more details here in the help text. Maybe something like:
Specify a target CPU revision, which will be appended to the value of the -mcpu option. For example, if the selected CPU is bf607, and then selected CPU revision is "r2", then gcc will receive the -mcpu=bf607-r2 option.
OK.
Regards,
Sonic
PATCH v2 3/7 arch: toolchain: Introduce binary formats BINFMT_*.
In reply to Sonic ZhangFrom: Sonic Zhang <[email protected]>
v2-changes:
none
v1-changes:
- Just introduce the symbol and options in arch generic Config.in. Append FLAT format link flags to external toolchain wrapper.
--- arch/Config.in | 13 +++++++++++++ toolchain/toolchain-external/ext-tool.mk | 4 ++++ .../toolchain-external/ext-toolchain-wrapper.c | 3 +++ 3 files changed, 20 insertions(+), 0 deletions(-)
arch/Config.in +13 -0
@@ -177,6 +177,19 @@ config BR2_GCC_TARGET_CPU config BR2_GCC_TARGET_CPU_REVISION string +# Set up target binary format +choice + prompt "Target Binary Format" + default BR2_BINFMT_ELF +config BR2_BINFMT_ELF + bool "ELF" +config BR2_BINFMT_FDPIC + bool "FDPIC" +config BR2_BINFMT_FLAT + bool "FLAT" + select BR2_PREFER_STATIC_LIB +endchoice + if BR2_arm || BR2_armeb source "arch/Config.in.arm" endif
toolchain/toolchain-external/ext-tool.mk +4 -0
@@ -164,6 +164,10 @@ ifneq ($(CC_TARGET_ABI_),) TOOLCHAIN_EXTERNAL_CFLAGS += -mabi=$(CC_TARGET_ABI_) TOOLCHAIN_EXTERNAL_WRAPPER_ARGS += -DBR_ABI='"$(CC_TARGET_ABI_)"' endif +ifeq ($(BR2_BINFMT_FLAT),y) +TOOLCHAIN_EXTERNAL_CFLAGS += -Wl,-elf2flt +TOOLCHAIN_EXTERNAL_WRAPPER_ARGS += -DBR_BINFMT_FLAT +endif ifneq ($(BR2_TARGET_OPTIMIZATION),) TOOLCHAIN_EXTERNAL_CFLAGS += $(call qstrip,$(BR2_TARGET_OPTIMIZATION))
toolchain/toolchain-external/ext-toolchain-wrapper.c +3 -0
@@ -47,6 +47,9 @@ static char *predef_args[] = {#ifdef BR_64 "-m64",#endif +#ifdef BR_BINFMT_FLAT + "-Wl,-elf2flt", +#endif #ifdef BR_ADDITIONAL_CFLAGS BR_ADDITIONAL_CFLAGS#endif
Signature · 1 line
1.7.0.4
PATCH v2 3/7 arch: toolchain: Introduce binary formats BINFMT_*.
Dear Sonic Zhang,
Quoted 10 lines · Sonic Zhang
From: Sonic Zhang <[email protected]>
v2-changes:
nonev1-changes:
- Just introduce the symbol and options in arch generic Config.in. Append FLAT format link flags to external toolchain wrapper.Signed-off-by Sonic Zhang <[email protected]>
As detailed in my comments to PATCH 1/7, this patch needs to be refactored with BR2_BFIN_FDPIC/BR2_BFIN_FLAT to avoid breaking existing code.
Also, I believe we want to add some architecture dependencies here. For the beginning, we at least want something like:
+# Set up target binary format
+choice
+ prompt "Target Binary Format"
+ default BR2_BINFMT_ELF
+config BR2_BINFMT_ELF
+ bool "ELF"
+ depends on !BR2_bfin
+config BR2_BINFMT_FDPIC
+ bool "FDPIC"
+ depends on BR2_bfin
+config BR2_BINFMT_FLAT
+ bool "FLAT"
+ select BR2_PREFER_STATIC_LIB
+ depends on BR2_bfin
+endchoice
Of course, this might be extended later on if we enable FLAT support on other architectures. But for now, it ensures that users can't make an invalid choice (like chose FDPIC on ARM).
Thanks!
Thomas
Signature · 4 lines
Thomas Petazzoni, Free Electrons Kernel, drivers, real-time and embedded Linux development, consulting, training and support. http://free-electrons.com
PATCH v2 3/7 arch: toolchain: Introduce binary formats BINFMT_*.
Hi Thomas,
Quoted 12 lines · Thomas Petazzoni
Dear Sonic Zhang,
Quoted 9 lines · Sonic Zhang
From: Sonic Zhang <[email protected]>
v2-changes:
nonev1-changes:
- Just introduce the symbol and options in arch generic Config.in. Append FLAT format link flags to external toolchain wrapper.
Quoted 5 lines
Quoted 1 line
Signed-off-by Sonic Zhang <[email protected]>As detailed in my comments to PATCH 1/7, this patch needs to be refactored with BR2_BFIN_FDPIC/BR2_BFIN_FLAT to avoid breaking existing code.
OK.
Quoted 10 lines
Also, I believe we want to add some architecture dependencies here. For the beginning, we at least want something like:
arch/Config.in +5
-0+# Set up target binary format +choice + prompt "Target Binary Format" + default BR2_BINFMT_ELF +config BR2_BINFMT_ELF
arch/Config.in config BR2_GCC_TARGET_CPU · line 191
Quoted 10 lines
arch/Config.in +10
-0+ bool "ELF" + depends on !BR2_bfin +config BR2_BINFMT_FDPIC + bool "FDPIC" + depends on BR2_bfin +config BR2_BINFMT_FLAT + bool "FLAT" + select BR2_PREFER_STATIC_LIB + depends on BR2_bfin +endchoice
No problem.
Regards,
Sonic
PATCH v2 3/7 arch: toolchain: Introduce binary formats BINFMT_*.
In reply to Sonic ZhangDear Sonic Zhang,
arch/Config.in config BR2_GCC_TARGET_CPU · line 191
Quoted 11 lines · Sonic Zhang
arch/Config.in +11
-0+choice + prompt "Target Binary Format" + default BR2_BINFMT_ELF +config BR2_BINFMT_ELF + bool "ELF" +config BR2_BINFMT_FDPIC + bool "FDPIC" +config BR2_BINFMT_FLAT + bool "FLAT" + select BR2_PREFER_STATIC_LIB +endchoice
Maybe some help text would also be nice for each choice.
Thanks,
Thomas
Signature · 4 lines
Thomas Petazzoni, Free Electrons Kernel, drivers, real-time and embedded Linux development, consulting, training and support. http://free-electrons.com
PATCH v2 3/7 arch: toolchain: Introduce binary formats BINFMT_*.
Hi Thomas,
Quoted 4 lines · Thomas Petazzoni
Dear Sonic Zhang,
On Fri, 29 Mar 2013 17:50:39 +0800, Sonic Zhang wrote:
Quoted 14 lines
arch/Config.in config BR2_GCC_TARGET_CPU · line 191
Quoted 11 lines
arch/Config.in +11
-0+choice + prompt "Target Binary Format" + default BR2_BINFMT_ELF +config BR2_BINFMT_ELF + bool "ELF" +config BR2_BINFMT_FDPIC + bool "FDPIC" +config BR2_BINFMT_FLAT + bool "FLAT" + select BR2_PREFER_STATIC_LIB +endchoiceMaybe some help text would also be nice for each choice.
OK
Regards,
Sonic
PATCH v2 4/7 arch: toolchain: Introduce binary format FLAT types.
In reply to Sonic ZhangFrom: Sonic Zhang <[email protected]>
v2-changes:
none
v1-changes:
- Just introduce the symbol and options in arch generic Config.in. Append FLAT types specific link flags into external toolchain wrapper.
--- arch/Config.in | 13 +++++++++++++ toolchain/toolchain-external/ext-tool.mk | 8 ++++++++ .../toolchain-external/ext-toolchain-wrapper.c | 6 ++++++ 3 files changed, 27 insertions(+), 0 deletions(-)
arch/Config.in +13 -0
@@ -190,6 +190,19 @@ config BR2_BINFMT_FLAT select BR2_PREFER_STATIC_LIB endchoice +# Set up flat binary type +choice + prompt "FLAT Binary type" + depends on BR2_BINFMT_FLAT + default BR2_BINFMT_FLAT_ONE +config BR2_BINFMT_FLAT_ONE + bool "One binary" +config BR2_BINFMT_FLAT_SEP_DATA + bool "Separate data and code" +config BR2_BINFMT_FLAT_SHARED + bool "Shared binary" +endchoice + if BR2_arm || BR2_armeb source "arch/Config.in.arm" endif
toolchain/toolchain-external/ext-tool.mk +8 -0
@@ -168,6 +168,14 @@ ifeq ($(BR2_BINFMT_FLAT),y) TOOLCHAIN_EXTERNAL_CFLAGS += -Wl,-elf2flt TOOLCHAIN_EXTERNAL_WRAPPER_ARGS += -DBR_BINFMT_FLAT endif +ifeq ($(BR2_BINFMT_FLAT_SEP_DATA),y) +TOOLCHAIN_EXTERNAL_CFLAGS += -msep-data +TOOLCHAIN_EXTERNAL_WRAPPER_ARGS += -DBR_BINFMT_FLAT_SEP_DATA +endif +ifeq ($(BR2_BINFMT_FLAT_SHARED),y) +TOOLCHAIN_EXTERNAL_CFLAGS += -mid-shared-library -mshared-library-id=0 +TOOLCHAIN_EXTERNAL_WRAPPER_ARGS += -DBR_BINFMT_FLAT_SHARED +endif ifneq ($(BR2_TARGET_OPTIMIZATION),) TOOLCHAIN_EXTERNAL_CFLAGS += $(call qstrip,$(BR2_TARGET_OPTIMIZATION))
toolchain/toolchain-external/ext-toolchain-wrapper.c +6 -0
@@ -50,6 +50,12 @@ static char *predef_args[] = {#ifdef BR_BINFMT_FLAT "-Wl,-elf2flt",#endif +#ifdef BR_BINFMT_FLAT_SEP_DATA + "-msep-data", +#endif +#ifdef BR_BINFMT_FLAT_SHARED + "-mid-shared-library -mshared-library-id=0", +#endif #ifdef BR_ADDITIONAL_CFLAGS BR_ADDITIONAL_CFLAGS#endif
Signature · 1 line
1.7.0.4
PATCH v2 4/7 arch: toolchain: Introduce binary format FLAT types.
Dear Sonic Zhang,
arch/Config.in config BR2_BINFMT_FLAT · line 201
Quoted 9 lines · Sonic Zhang
arch/Config.in +9
-0+# Set up flat binary type +choice + prompt "FLAT Binary type" + depends on BR2_BINFMT_FLAT + default BR2_BINFMT_FLAT_ONE +config BR2_BINFMT_FLAT_ONE + bool "One binary" +config BR2_BINFMT_FLAT_SEP_DATA + bool "Separate data and code"
depends on BR2_bfin || BR2_m68k
I know I've suggested on PATCH 3/7 to make FLAT depend on bfin, so technically this isn't needed. But since -msep-data is really an option that is only available on bfin and m68k, while the flat format can be used potentially on other architectures, I'd prefer to make this architecture dependency explicit here as well.
arch/Config.in config BR2_BINFMT_FLAT · line 203
Quoted 2 lines
arch/Config.in +2
-0+config BR2_BINFMT_FLAT_SHARED + bool "Shared binary"
depends on BR2_bfin || BR2_m68k
Same reason: -mid-shared-library is only available for Blackfin and m68k.
Quoted 1 line
+endchoice
I would have originally thought of putting those choices directly within the list of BR2_BINFMT_* (i.e have a list containing elf, fdpic, flat-one, flat-sep-data, flat-shared). But maybe it makes more sense the way you did, some I'm fine with that.
Having some help text for each choice would be nice.
Thanks!
Thomas
Signature · 4 lines
Thomas Petazzoni, Free Electrons Kernel, drivers, real-time and embedded Linux development, consulting, training and support. http://free-electrons.com
PATCH v2 4/7 arch: toolchain: Introduce binary format FLAT types.
Hi Thomas,
Quoted 22 lines · Thomas Petazzoni
Dear Sonic Zhang,
arch/Config.in config BR2_BINFMT_FLAT · line 201
Quoted 9 lines · Sonic Zhang
arch/Config.in +9
-0+# Set up flat binary type +choice + prompt "FLAT Binary type" + depends on BR2_BINFMT_FLAT + default BR2_BINFMT_FLAT_ONE +config BR2_BINFMT_FLAT_ONE + bool "One binary" +config BR2_BINFMT_FLAT_SEP_DATA + bool "Separate data and code"depends on BR2_bfin || BR2_m68kI know I've suggested on PATCH 3/7 to make FLAT depend on bfin, so technically this isn't needed. But since -msep-data is really an option that is only available on bfin and m68k, while the flat format can be used potentially on other architectures, I'd prefer to make this architecture dependency explicit here as well.
Quoted 7 lines
arch/Config.in config BR2_BINFMT_FLAT · line 203
Quoted 2 lines
arch/Config.in +2
-0+config BR2_BINFMT_FLAT_SHARED + bool "Shared binary"depends on BR2_bfin || BR2_m68kSame reason: -mid-shared-library is only available for Blackfin and m68k.
OK.
Quoted 10 lines
Quoted 2 lines
+endchoice
I would have originally thought of putting those choices directly within the list of BR2_BINFMT_* (i.e have a list containing elf, fdpic, flat-one, flat-sep-data, flat-shared). But maybe it makes more sense the way you did, some I'm fine with that.
Having some help text for each choice would be nice.
OK.
Regards,
Sonic
PATCH v2 4/7 arch: toolchain: Introduce binary format FLAT types.
In reply to Thomas PetazzoniHi Thomas,
Just figure out the compiling flags for BINFMT_FLAT_SHARED and BINFMT_FLAT_SEP can't be put into the toolchain wrapper, because these flags are for application only, they can't be used to build Linux kernel.
-mid-shared-library -mshared-library-id=0
-msep-data
So, I have to move them back to file package/Makefile.in .
Regards,
Sonic
PATCH v2 4/7 arch: toolchain: Introduce binary format FLAT types.
Dear Sonic Zhang,
Quoted 9 lines · Sonic Zhang
Just figure out the compiling flags for
BINFMT_FLAT_SHAREDandBINFMT_FLAT_SEPcan't be put into the toolchain wrapper, because these flags are for application only, they can't be used to build Linux kernel.(quoted diff) +0
-2-mid-shared-library -mshared-library-id=0 -msep-dataSo, I have to move them back to file package/Makefile.in .
Hum, that is not really nice, because the idea of the external toolchain wrapper is to ensure that when the user will use output/host/usr/bin/<arch>-gcc, he will get binaries that are built for the same architecture/flags than the binaries built by Buildroot. That's more or less the whole idea of the external toolchain wrapper thing.
One option is to have something like:
ifeq ($(BR2_bfin),y)LINUX_MAKE_FLAGS =
KBUILD_CFLAGS = "$(TARGET_CFLAGS) -mno-id-shared-library ..."
endif
Not sure exactly if that's the right solution or not. Peter, you're the one who did the external toolchain wrapper originally. Could you comment on this specific issue?
Thanks!
Thomas
Signature · 4 lines
Thomas Petazzoni, Free Electrons Kernel, drivers, real-time and embedded Linux development, consulting, training and support. http://free-electrons.com
PATCH v2 4/7 arch: toolchain: Introduce binary format FLAT types.
Hi Thomas,
Quoted 23 lines · Thomas Petazzoni
Dear Sonic Zhang,
Quoted 9 lines · Sonic Zhang
Just figure out the compiling flags for
BINFMT_FLAT_SHAREDandBINFMT_FLAT_SEPcan't be put into the toolchain wrapper, because these flags are for application only, they can't be used to build Linux kernel.(quoted diff) +0
-2-mid-shared-library -mshared-library-id=0 -msep-dataSo, I have to move them back to file package/Makefile.in .
Hum, that is not really nice, because the idea of the external toolchain wrapper is to ensure that when the user will use output/host/usr/bin/<arch>-gcc, he will get binaries that are built for the same architecture/flags than the binaries built by Buildroot. That's more or less the whole idea of the external toolchain wrapper thing.
One option is to have something like:
Quoted 9 lines
ifeq ($(BR2_bfin),y)
LINUX_MAKE_FLAGS=KBUILD_CFLAGS = "$(TARGET_CFLAGS) -mno-id-shared-library ..."endif
Not sure exactly if that's the right solution or not. Peter, you're the one who did the external toolchain wrapper originally. Could you comment on this specific issue?
The LINUX_MAKE_FLAGS should not include the shared flat specific flags "-mno-id-shared-library ...". Your suggestion doesn't work when building Linux kernel. We need some ways to remove flags "-mno-id-shared-library ..." from the gcc wrapper for the shared flat binary format when building Linux kernel.
My new patch is to append the shared flat flags only when building applications.
Any better solution?
Regards,
Sonic
PATCH v2 4/7 arch: toolchain: Introduce binary format FLAT types.
Dear Sonic Zhang,
Quoted 9 lines · Sonic Zhang
Quoted 4 lines
Not sure exactly if that's the right solution or not. Peter, you're the one who did the external toolchain wrapper originally. Could you comment on this specific issue?
The
LINUX_MAKE_FLAGSshould not include the shared flat specific flags "-mno-id-shared-library ...". Your suggestion doesn't work when building Linux kernel. We need some ways to remove flags "-mno-id-shared-library ..." from the gcc wrapper for the shared flat binary format when building Linux kernel.
Hm? If you pass -mno-id-shared-library after -mid-shared-library, it will be just as if you hadn't passed the option at all. So if the external toolchain wrapper passes -mid-shared-library, but then the CFLAGS passed to the kernel override that by passing -mno-id-shared-library, it will be just as if the external wrapper hadn't passed -mid-shared-library.
Thomas
Signature · 4 lines
Thomas Petazzoni, Free Electrons Kernel, drivers, real-time and embedded Linux development, consulting, training and support. http://free-electrons.com
PATCH v2 4/7 arch: toolchain: Introduce binary format FLAT types.
Hi Thomas,
Quoted 14 lines · Thomas Petazzoni
Dear Sonic Zhang,
Quoted 9 lines · Sonic Zhang
Quoted 4 lines
Not sure exactly if that's the right solution or not. Peter, you're the one who did the external toolchain wrapper originally. Could you comment on this specific issue?
The
LINUX_MAKE_FLAGSshould not include the shared flat specific flags "-mno-id-shared-library ...". Your suggestion doesn't work when building Linux kernel. We need some ways to remove flags "-mno-id-shared-library ..." from the gcc wrapper for the shared flat binary format when building Linux kernel.
Quoted 8 lines
Hm? If you pass -mno-id-shared-library after -mid-shared-library, it will be just as if you hadn't passed the option at all. So if the external toolchain wrapper passes -mid-shared-library, but then the CFLAGS passed to the kernel override that by passing -mno-id-shared-library, it will be just as if the external wrapper hadn't passed -mid-shared-library.
The blackfin GNU toolchain doesn't understand flag "-mno-shared-library-id", "-mno-shared-library-id=" and "-mno-shared-library-id=0". Flag "-mshared-library-id=0" set in wrapper can't be canceled by passing "-mno-shared-library-id" later in linux.mk.
ifeq ($(BR2_bfin),y)LINUX_MAKE_FLAGS += KBUILD_CFLAGS="$(TARGET_CFLAGS) -mno-id-shared-library -mno-shared-library-id -mno-sep-data"
endif
PATH="/home/sonic/projects/buildroot/output/host/bin:/home/sonic/projects/buildroot/output/host/usr/bin:/home/sonic/projects/buildroot/output/host/usr/sbin/:/opt/arm/bin:/opt/uclinux/bfin-elf/bin/:/opt/uclinux/bfin-uclinux/bin/:/opt/uclinux/bfin-linux-uclibc/bin/:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games" PERLLIB="/home/sonic/projects/buildroot/output/host/usr/lib/perl"/usr/bin/make -j4
O="/home/sonic/projects/buildroot/output/build/linux-custom"
HOSTCC="/usr/bin/gcc" HOSTCFLAGS="" ARCH=blackfin
INSTALL_MOD_PATH=/home/sonic/projects/buildroot/output/target
CROSS_COMPILE="/home/sonic/projects/buildroot/output/host/usr/bin/bfin-uclinux-"
DEPMOD=/home/sonic/projects/buildroot/output/host/usr/sbin/depmod
KBUILD_CFLAGS="-D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE
-D_FILE_OFFSET_BITS=64 -pipe -Os -D__NOMMU__ -D__uClinux__
-mno-id-shared-library -mno-shared-library-id -mno-sep-data" -C
"linux/linux-kernel"/ uImage
make[1]: Entering directory
`/home/sonic/projects/buildroot/linux/linux-kernel'
GEN /home/sonic/projects/buildroot/output/build/linux-custom/Makefile CHK include/generated/uapi/linux/version.h CC scripts/mod/empty.o
cc1: error: unrecognized command line option "-mno-shared-library-id" /home/sonic/projects/buildroot/linux/linux-kernel/scripts/mod/empty.c:1:
error: -mshared-library-id= specified without -mid-shared-library make[4]: *** [scripts/mod/empty.o] Error 1
make[3]: *** [scripts/mod] Error 2
make[2]: *** [scripts] Error 2
Regards,
Sonic
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20130417/2e8a91f1/attachment.html>
PATCH v2 4/7 arch: toolchain: Introduce binary format FLAT types.
Hi Peter and Thomas,
Any further suggestion on how can I push forward this patch?
Compiler parameter -mno-xxx is not an option for blackfin specific flag id-shared-library=0.
Regards,
Sonic
Quoted 67 lines · Sonic Zhang
Hi Thomas,
Quoted 15 lines · Thomas Petazzoni
Dear Sonic Zhang,
Quoted 1 line · Sonic Zhang
Quoted 1 line
Not sure exactly if that's the right solution or not. Peter, you're
the
Quoted 8 lines
Quoted 3 lines
one who did the external toolchain wrapper originally. Could you comment on this specific issue?
The
LINUX_MAKE_FLAGSshould not include the shared flat specific flags "-mno-id-shared-library ...". Your suggestion doesn't work when building Linux kernel. We need some ways to remove flags "-mno-id-shared-library ..." from the gcc wrapper for the shared flat binary format when building Linux kernel.Quoted 8 lines
Hm? If you pass -mno-id-shared-library after -mid-shared-library, it will be just as if you hadn't passed the option at all. So if the external toolchain wrapper passes -mid-shared-library, but then the CFLAGS passed to the kernel override that by passing -mno-id-shared-library, it will be just as if the external wrapper hadn't passed -mid-shared-library.
The blackfin GNU toolchain doesn't understand flag "-mno-shared-library-id", "-mno-shared-library-id=" and "-mno-shared-library-id=0". Flag "-mshared-library-id=0" set in wrapper can't be canceled by passing "-mno-shared-library-id" later in linux.mk.
ifeq ($(BR2_bfin),y)
LINUX_MAKE_FLAGS+= KBUILD_CFLAGS="$(TARGET_CFLAGS) -mno-id-shared-library -mno-shared-library-id -mno-sep-data"
endifPATH="/home/sonic/projects/buildroot/output/host/bin:/home/sonic/projects/buildroot/output/host/usr/bin:/home/sonic/projects/buildroot/output/host/usr/sbin/:/opt/arm/bin:/opt/uclinux/bfin-elf/bin/:/opt/uclinux/bfin-uclinux/bin/:/opt/uclinux/bfin-linux-uclibc/bin/:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games" PERLLIB="/home/sonic/projects/buildroot/output/host/usr/lib/perl"
/usr/bin/make-j4
O="/home/sonic/projects/buildroot/output/build/linux-custom"
HOSTCC="/usr/bin/gcc" HOSTCFLAGS="" ARCH=blackfin
INSTALL_MOD_PATH=/home/sonic/projects/buildroot/output/target
CROSS_COMPILE="/home/sonic/projects/buildroot/output/host/usr/bin/bfin-uclinux-"
DEPMOD=/home/sonic/projects/buildroot/output/host/usr/sbin/depmod
KBUILD_CFLAGS="-D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE(quoted diff) +0
-2-D_FILE_OFFSET_BITS=64 -pipe -Os -D__NOMMU__ -D__uClinux__ -mno-id-shared-library -mno-shared-library-id -mno-sep-data" -C"linux/linux-kernel"/ uImage
make[1]: Entering directory `/home/sonic/projects/buildroot/linux/linux-kernel'GEN /home/sonic/projects/buildroot/output/build/linux-custom/Makefile CHK include/generated/uapi/linux/version.h CC scripts/mod/empty.occ1: error: unrecognized command line option "-mno-shared-library-id"
/home/sonic/projects/buildroot/linux/linux-kernel/scripts/mod/empty.c:1:
error: -mshared-library-id= specified without -mid-shared-library make[4]: *** [scripts/mod/empty.o] Error 1
make[3]: *** [scripts/mod] Error 2
make[2]: *** [scripts] Error 2
Quoted 4 lines
Regards,
Sonic
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20130423/d5499318/attachment-0001.html>
PATCH v2 5/7 package: Introduce package-specific BINFMT_FLAT options.
In reply to Sonic ZhangFrom: Sonic Zhang <[email protected]>
v2-changes:
- Fix typo error in macro BINFMT_FLAT
v1-changes:
- Add new option <PKG>_FLAT_STACKSIZE. The document needs to be updated.
--- package/pkg-autotools.mk | 5 +++++ package/pkg-generic.mk | 5 +++++ 2 files changed, 10 insertions(+), 0 deletions(-)
package/pkg-autotools.mk +5 -0
@@ -82,6 +82,11 @@ $(2)_CLEAN_OPT ?= clean $(2)_UNINSTALL_STAGING_OPT ?= DESTDIR=$$(STAGING_DIR) uninstall $(2)_UNINSTALL_TARGET_OPT ?= DESTDIR=$$(TARGET_DIR) uninstall +ifeq ($(BR2_BINFMT_FLAT),y) + ifneq ($$($(2)_FLAT_STACKSIZE),) + $(2)_CONF_ENV += LDFLAGS="$(TARGET_LDFLAGS) -Wl,-elf2flt=-s$$($(2)_FLAT_STACKSIZE)" + endif +endif # # Configure step. Only define it if not already defined by the package
package/pkg-generic.mk +5 -0
@@ -303,6 +303,11 @@ endif $(2)_REDISTRIBUTE ?= YES +ifeq ($(BR2_BINFMT_FLAT),y) + ifneq ($$($(2)_FLAT_STACKSIZE),) + $(2)_FLAT_LDFLAGS = -Wl,-elf2flt=-s$$($(2)_FLAT_STACKSIZE) + endif +endif $(2)_DEPENDENCIES ?= $(filter-out $(1),$(patsubst host-host-%,host-%,$(addprefix host-,$($(3)_DEPENDENCIES))))
Signature · 1 line
1.7.0.4
PATCH v2 5/7 package: Introduce package-specific BINFMT_FLAT options.
Dear Sonic Zhang,
Quoted 21 lines · Sonic Zhang
package/pkg-autotools.mk +5
-0@@ -82,6 +82,11 @@ $(2)_CLEAN_OPT ?= clean $(2)_UNINSTALL_STAGING_OPT ?= DESTDIR=$$(STAGING_DIR) uninstall $(2)_UNINSTALL_TARGET_OPT ?= DESTDIR=$$(TARGET_DIR) uninstall +ifeq ($(BR2_BINFMT_FLAT),y) + ifneq ($$($(2)_FLAT_STACKSIZE),) + $(2)_CONF_ENV += LDFLAGS="$(TARGET_LDFLAGS) -Wl,-elf2flt=-s$$($(2)_FLAT_STACKSIZE)" + endif +endif # # Configure step. Only define it if not already defined by the packagepackage/pkg-generic.mk +0
-0
package/pkg-generic.mk endif · line 312
Quoted 10 lines
package/pkg-generic.mk +5
-0@@ -303,6 +303,11 @@ endif $(2)_REDISTRIBUTE ?= YES +ifeq ($(BR2_BINFMT_FLAT),y) + ifneq ($$($(2)_FLAT_STACKSIZE),) + $(2)_FLAT_LDFLAGS = -Wl,-elf2flt=-s$$($(2)_FLAT_STACKSIZE) + endif +endif $(2)_DEPENDENCIES ?= $(filter-out $(1),$(patsubst host-host-%,host-%,$(addprefix host-,$($(3)_DEPENDENCIES))))
I understand the problem but I'm still not really happy with the solution here. Maybe Arnout will have some suggestions.
The problem I see with the pkg-autotools change is that some packages already override LDFLAGS in their <pkg>_CONF_ENV. For example:
NDISC6_CONF_ENV += LDFLAGS="$(TARGET_LDFLAGS) -lintl"
Since your <pkg>_CONF_ENV += assignment will be done after the ones coming from the package .mk file, you're going to remove the -lintl that had been added by the package.
For the pkg-generic change, I'm not entirely happy because it means each and every package should use <pkg>_FLAT_LDFLAGS, while we have told for a long time our packagers to just use $(TARGET_CONFIGURE_OPTS) to get all the right compiler/linker flags.
So I think it seems we somehow need to make $(TARGET_LDFLAGS) a per-package variable? Arnout what do you think about this? Do you have ideas?
Thanks!
Thomas
Signature · 4 lines
Thomas Petazzoni, Free Electrons Kernel, drivers, real-time and embedded Linux development, consulting, training and support. http://free-electrons.com
PATCH v2 5/7 package: Introduce package-specific BINFMT_FLAT options.
Quoted 34 lines · Thomas Petazzoni
Dear Sonic Zhang,
Quoted 30 lines · Sonic Zhang
package/pkg-autotools.mk +5
-0@@ -82,6 +82,11 @@ $(2)_CLEAN_OPT ?= clean $(2)_UNINSTALL_STAGING_OPT ?= DESTDIR=$$(STAGING_DIR) uninstall $(2)_UNINSTALL_TARGET_OPT ?= DESTDIR=$$(TARGET_DIR) uninstall +ifeq ($(BR2_BINFMT_FLAT),y) + ifneq ($$($(2)_FLAT_STACKSIZE),) + $(2)_CONF_ENV += LDFLAGS="$(TARGET_LDFLAGS) -Wl,-elf2flt=-s$$($(2)_FLAT_STACKSIZE)" + endif +endif # # Configure step. Only define it if not already defined by the packagepackage/pkg-generic.mk +5
-0@@ -303,6 +303,11 @@ endif $(2)_REDISTRIBUTE ?= YES +ifeq ($(BR2_BINFMT_FLAT),y) + ifneq ($$($(2)_FLAT_STACKSIZE),) + $(2)_FLAT_LDFLAGS = -Wl,-elf2flt=-s$$($(2)_FLAT_STACKSIZE) + endif +endif
Quoted 4 lines
package/pkg-generic.mk endif · line 312
Quoted 1 line
package/pkg-generic.mk +0
-0$(2)_DEPENDENCIES ?= $(filter-out $(1),$(patsubst host-host-%,host-%,$(addprefix host-,$($(3)_DEPENDENCIES))))I understand the problem but I'm still not really happy with the solution here. Maybe Arnout will have some suggestions.
I had exactly the same thoughts while reading the patch...
Quoted 9 lines
The problem I see with the pkg-autotools change is that some packages already override LDFLAGS in their <pkg>_CONF_ENV. For example:
NDISC6_CONF_ENV+= LDFLAGS="$(TARGET_LDFLAGS) -lintl"Since your <pkg>_CONF_ENV += assignment will be done after the ones coming from the package .mk file, you're going to remove the -lintl that had been added by the package.
Quoted 8 lines
For the pkg-generic change, I'm not entirely happy because it means each and every package should use <pkg>_FLAT_LDFLAGS, while we have told for a long time our packagers to just use $(TARGET_CONFIGURE_OPTS) to get all the right compiler/linker flags.
So I think it seems we somehow need to make $(TARGET_LDFLAGS) a per-package variable? Arnout what do you think about this? Do you have ideas?
In fact, it can easily be made a per-package variable. In the
definition of TARGET_LDFLAGS, we can add:
TARGET_LDFLAGS = ... \ $($(PKG)_LDFLAGS)
The $(PKG) is only evaluated when the TARGET_LDFLAGS variable is used.
Outside the scope of the generic rules, $(PKG) evaluates to nothing,
and _LDFLAGS is not defined, so the $($(PKG)_LDFLAGS) has no effect.
Inside the scope of the generic rules, $(PKG) is always defined. So we
just have to add the following to the pkg-generic macro:
ifeq ($(BR2_BINFMT_FLAT),y)
$(2)_LDFLAGS += -Wl,-elf2flt=-s$$($(2)_FLAT_STACKSIZE)
endif
As far as I can see, that should work nicely. No need to change
anything in autotools. Unfortunately, it does require changing the packages that have something like:
DOSFSTOOLS_LDFLAGS = $(TARGET_LDFLAGS)
Also, it is not very intuitive, because in a non-autotools package
where you can't use the LDFLAGS environment variable, you'll have something like:
AXEL_LDFLAGS = -lpthread
define AXEL_BUILD_CMDS
$(TARGET_CONFIGURE_OPTS) $(MAKE) \
LFLAGS="$(TARGET_LDFLAGS)" -C $(@D)endef
i.e. you're defining AXEL_LDFLAGS but using TARGET_LDFLAGS. Not a really big issue, I think, but not very nice either.
For this patch list, however, I think changing the 15 uses ofFOO_LDFLAGS is going a bit too far. So I propose to make the following change instead in package/Makefile.in:
ifeq ($(BR2_BINFMT_FLAT),y)
TARGET_LDFLAGS += $(if $($(PKG)_FLAT_STACKSIZE),\ -Wl,-elf2flt=-s$($(PKG)_FLAT_STACKSIZE))
endif
AFAICS, that should work for any package without any further change.
But needs to be tested, of course...
Regards,
Arnout
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7CB5 E4CC 6C2E EFD4 6E3D A754 F963 ECAB 2450 2F1F
PATCH v2 5/7 package: Introduce package-specific BINFMT_FLAT options.
Hi Arnout,
Quoted 105 lines · Arnout Vandecappelle
Quoted 36 lines · Thomas Petazzoni
Dear Sonic Zhang,
Quoted 31 lines · Sonic Zhang
package/pkg-autotools.mk +5
-1@@ -82,6 +82,11 @@ $(2)_CLEAN_OPT ?= clean $(2)_UNINSTALL_STAGING_OPT ?= DESTDIR=$$(STAGING_DIR) uninstall $(2)_UNINSTALL_TARGET_OPT ?= DESTDIR=$$(TARGET_DIR) uninstall +ifeq ($(BR2_BINFMT_FLAT),y) + ifneq ($$($(2)_FLAT_STACKSIZE),) + $(2)_CONF_ENV += LDFLAGS="$(TARGET_LDFLAGS) -Wl,-elf2flt=-s$$($(2)_FLAT_STACKSIZE)" + endif +endif # # Configure step. Only define it if not already defined by the packagepackage/pkg-generic.mk +5
-0@@ -303,6 +303,11 @@ endif $(2)_REDISTRIBUTE ?= YES +ifeq ($(BR2_BINFMT_FLAT),y) + ifneq ($$($(2)_FLAT_STACKSIZE),) + $(2)_FLAT_LDFLAGS = -Wl,-elf2flt=-s$$($(2)_FLAT_STACKSIZE) + endif +endifQuoted 6 lines
package/pkg-generic.mk endif · line 313
Quoted 2 lines
package/pkg-generic.mk +0
-0$(2)_DEPENDENCIES ?= $(filter-out $(1),$(patsubst host-host-%,host-%,$(addprefix host-,$($(3)_DEPENDENCIES))))I understand the problem but I'm still not really happy with the solution here. Maybe Arnout will have some suggestions.
I had exactly the same thoughts while reading the patch...
Quoted 9 lines
The problem I see with the pkg-autotools change is that some packages already override LDFLAGS in their <pkg>_CONF_ENV. For example:
NDISC6_CONF_ENV+= LDFLAGS="$(TARGET_LDFLAGS) -lintl"Since your <pkg>_CONF_ENV += assignment will be done after the ones coming from the package .mk file, you're going to remove the -lintl that had been added by the package.
Quoted 8 lines
For the pkg-generic change, I'm not entirely happy because it means each and every package should use <pkg>_FLAT_LDFLAGS, while we have told for a long time our packagers to just use $(TARGET_CONFIGURE_OPTS) to get all the right compiler/linker flags.
So I think it seems we somehow need to make $(TARGET_LDFLAGS) a per-package variable? Arnout what do you think about this? Do you have ideas?
In fact, it can easily be made a per-package variable. In the definition of
TARGET_LDFLAGS, we can add:TARGET_LDFLAGS = ... \ $($(PKG)_LDFLAGS)The $(PKG) is only evaluated when the
TARGET_LDFLAGSvariable is used.Outside the scope of the generic rules, $(PKG) evaluates to nothing, and
_LDFLAGSis not defined, so the $($(PKG)_LDFLAGS) has no effect.Inside the scope of the generic rules, $(PKG) is always defined. So we just
have to add the following to the pkg-generic macro:ifeq ($(BR2_BINFMT_FLAT),y)
$(2)_LDFLAGS += -Wl,-elf2flt=-s$$($(2)_FLAT_STACKSIZE)
endifAs far as I can see, that should work nicely. No need to change anything in
autotools. Unfortunately, it does require changing the packages that have something like:
DOSFSTOOLS_LDFLAGS= $(TARGET_LDFLAGS)Also, it is not very intuitive, because in a non-autotools package where
you can't use the LDFLAGS environment variable, you'll have something like:
AXEL_LDFLAGS= -lpthread
defineAXEL_BUILD_CMDS$(TARGET_CONFIGURE_OPTS) $(MAKE) \ LFLAGS="$(TARGET_LDFLAGS)" -C $(@D)endef
i.e. you're defining
AXEL_LDFLAGSbut usingTARGET_LDFLAGS. Not a really big issue, I think, but not very nice either.
Quoted 9 lines
For this patch list, however, I think changing the 15 uses of
FOO_LDFLAGS
is going a bit too far. So I propose to make the following change instead in package/Makefile.in:ifeq ($(BR2_BINFMT_FLAT),y)
TARGET_LDFLAGS += $(if $($(PKG)_FLAT_STACKSIZE),\ -Wl,-elf2flt=-s$($(PKG)_FLAT_STACKSIZE))endif
This doesn't work.
/home/sonic/projects/buildroot/output/host/usr/bin/bfin-uclinux-gcc -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -pipe -Os -D__NOMMU__ -D__uClinux__ -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -Wl --static bfin-dma.c
-o bfin-dma
cc1: error: unrecognized command line option "-Wl" make[1]: *** [bfin-dma] Error 1
In addition, if 2 generic makefile packages set different flat stack sizes, are both of these stack size flags added to the same TARGE_LDFLAGS macro with your change?
Regards,
Sonic
PATCH v2 5/7 package: Introduce package-specific BINFMT_FLAT options.
Quoted 3 lines · Sonic Zhang
Hi Arnout,
On Wed, Apr 10, 2013 at 1:47 PM, Arnout Vandecappelle <[email protected]> wrote:
[snip]
Quoted 13 lines
Quoted 5 lines
ifeq ($(BR2_BINFMT_FLAT),y)
TARGET_LDFLAGS += $(if $($(PKG)_FLAT_STACKSIZE),\ -Wl,-elf2flt=-s$($(PKG)_FLAT_STACKSIZE))endif
This doesn't work.
/home/sonic/projects/buildroot/output/host/usr/bin/bfin-uclinux-gcc(quoted diff) +0
-3-D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -pipe -Os -D__NOMMU__ -D__uClinux__ -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -Wl --static bfin-dma.c -o bfin-dma
Quoted 2 lines
cc1: error: unrecognized command line option "-Wl" make[1]: *** [bfin-dma] Error 1
Ah yes, of course, the comma. If should be:
ifeq ($(BR2_BINFMT_FLAT),y)
TARGET_LDFLAGS += $(if $($(PKG)_FLAT_STACKSIZE),\
-Wl$(comma)-elf2flt=-s$($(PKG)_FLAT_STACKSIZE))endif
Quoted 3 lines
In addition, if 2 generic makefile packages set different flat stack sizes, are both of these stack size flags added to the same
TARGE_LDFLAGSmacro with your change?
That is the magic of late binding in make. The value of TARGET_LDFLAGS
is only evaluated at the time when it is used, and it is only at that time that the $(PKG) is expanded. And if it is used in the body of a pattern rule, then the expansion only takes place when the rule is instantiated.
For example, there are two packages, FOO and BAR, that setFOO_FLAT_STACKSIZE and BAR_FLAT_STACKSIZE. They both use TARGET_LDFLAGS in their BUILD_CMDS. The BUILD_CMDS are used in the following pattern rule from pkg-generic.mk:
# Build
$(BUILD_DIR)/%/.stamp_built::
@$(call MESSAGE,"Building")
$($(PKG)_BUILD_CMDS)
$(foreach hook,$($(PKG)_POST_BUILD_HOOKS),$(call $(hook))$(sep))
$(Q)touch $@ When this rule is instantiated for package foo, then the PKG variable
is set to FOO. So $($(PKG)_BUILD_CMDS) expands to $(FOO_BUILD_CMDS), which expands to something like
$(MAKE) LDFLAGS="$(TARGET_LDFLAGS)"
which expands to
make -j8 LDFLAGS="$(call qstrip,$(BR2_TARGET_LDFLAGS))\
--static\
$(if $($(PKG)_FLAT_STACKSIZE),\
-Wl$(comma)-elf2flt=-s$($(PKG)_FLAT_STACKSIZE))"which expands to
make -j8 LDFLAGS="$(call qstrip,"")\
--static\
$(if $(FOO_FLAT_STACKSIZE),\
-Wl$(comma)-elf2flt=-s$($(PKG)_FLAT_STACKSIZE))"which expands to
make -j8 LDFLAGS="\
--static\
$(if 2048,\
-Wl$(comma)-elf2flt=-s$($(PKG)_FLAT_STACKSIZE))"which expands to
make -j8 LDFLAGS="\ --static\ -Wl$(comma)-elf2flt=-s$($(PKG)_FLAT_STACKSIZE)"
which expands to
make -j8 LDFLAGS="\ --static\ -W,-elf2flt=-s$(FOO_FLAT_STACKSIZE)"
which expands to
make -j8 LDFLAGS=" --static -W,-elf2flt=-s2048"
Regards,
Arnout
Signature · 6 lines
Arnout Vandecappelle arnout at mind be Senior Embedded Software Architect +32-16-286500 Essensium/Mind http://www.mind.be G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle GPG fingerprint: 7CB5 E4CC 6C2E EFD4 6E3D A754 F963 ECAB 2450 2F1F
PATCH v2 6/7 arch: Introduce blackfin-specific Makefile
In reply to Sonic ZhangFrom: Sonic Zhang <[email protected]>
v2-changes:
- Remove useless Macro __uClinux__
v1-changes:
- Create arch makefile.
- Create blackfin makefile.
- Add FDPIC and FLAT library options and makefile targets to install libraries for different binary formats into the same rootfs image.
- update the TARGET_EXCEPTIONS variable in support/scripts/graph-depends for new custom targets
--- Makefile | 2 ++ arch/Config.in.bfin | 21 +++++++++++++++++++++ arch/Makefile.in | 5 +++++ arch/Makefile.in.bfin | 22 ++++++++++++++++++++++ support/scripts/graph-depends | 2 ++ 5 files changed, 52 insertions(+), 0 deletions(-) create mode 100644 arch/Makefile.in create mode 100644 arch/Makefile.in.bfin
Makefile +2 -0
@@ -329,6 +329,8 @@ ifneq ($(PACKAGE_OVERRIDE_FILE),) -include $(PACKAGE_OVERRIDE_FILE) endif +include arch/Makefile.in + include package/*/*.mk include boot/common.mk
arch/Config.in.bfin +21 -0
@@ -62,6 +62,27 @@ config BR2_bf561 bool "bf561" endchoice +config BR2_BFIN_INSTALL_FDPIC_SHARED + depends on BR2_bfin && !BR2_BINFMT_FDPIC + bool "Install FDPIC shared libraries" + default y + help + Blackfin Linux kernel supports running both FDPIC and FLAT applications + concurrently if the binary format specific libraries installed properly. + This option allow developer to install FDPIC libraries into a buildroot + rootfs image built with binary format macro other than BR2_BINFMT_FDPIC. + +config BR2_BFIN_INSTALL_FLAT_SHARED + depends on BR2_bfin && !BR2_BINFMT_FLAT_SHARED + bool "Install FLAT shared libraries" + default y + help + Blackfin Linux kernel supports running both FDPIC and FLAT applications + concurrently if the binary format specific libraries installed properly. + This option allow developer to install FLAT libraries into a buildroot + rootfs image built with binary format macro other than + BR2_BINFMT_FLAT_SHARED. + config BR2_ARCH default "bfin"
arch/Makefile.in new file +5 -0
@@ -0,0 +1,5 @@ +# The architecture specific Makefile.in.$ARCH should be included only +# when the arch macro is selected. +ifeq ($(BR2_bfin),y) +include arch/Makefile.in.bfin +endif
arch/Makefile.in.bfin new file +22 -0
@@ -0,0 +1,22 @@ +TARGETS_BFIN-$(BR2_BFIN_INSTALL_FDPIC_SHARED) += target-shared-libs-fdpic +TARGETS_BFIN-$(BR2_BFIN_INSTALL_FLAT_SHARED) += target-shared-libs-flat +TARGETS += $(TARGETS_BFIN-y) + +CROSS_COMPILE_SHARED_FDPIC ?= bfin-linux-uclibc- +target-shared-libs-fdpic: + set -e; \ + t=`$(CROSS_COMPILE_SHARED_FDPIC)gcc $(CPUFLAGS) -print-file-name=libc.a`; \ + t=`dirname $$t`/../..; \ + for i in $$t/lib/*so*; do \ + i=`readlink -f "$$i"`; \ + soname=`$(CROSS_COMPILE_SHARED_FDPIC)readelf -d "$$i" | sed -n '/(SONAME)/s:.*[[]\(.*\)[]].*:\1:p'`; \ + $(INSTALL) -D $$i $(TARGET_DIR)/lib/$$soname; \ + done + +CROSS_COMPILE_SHARED_FLAT ?= bfin-uclinux- +target-shared-libs-flat: + set -e; \ + t=`$(CROSS_COMPILE_SHARED_FLAT)gcc $(CPUFLAGS) -mid-shared-library -print-file-name=libc`; \ + if [ -f $$t -a ! -h $$t ] ; then \ + $(INSTALL) -D $$t $(TARGET_DIR)/lib/lib1.so; \ + fi
support/scripts/graph-depends +2 -0
@@ -168,6 +168,8 @@ TARGET_EXCEPTIONS = [ "erase-fakeroots", "target-generic-hostname", "target-root-passwd", + "target-shared-libs-fdpic", + "target-shared-libs-flat", ] # In full mode, start with the result of get_targets() to get the main
Signature · 1 line
1.7.0.4
PATCH v2 6/7 arch: Introduce blackfin-specific Makefile
Dear Sonic Zhang,
arch/Config.in.bfin config BR2_bf561 · line 68
Quoted 4 lines · Sonic Zhang
arch/Config.in.bfin +4
-0+config BR2_BFIN_INSTALL_FDPIC_SHARED + depends on BR2_bfin && !BR2_BINFMT_FDPIC + bool "Install FDPIC shared libraries" + default y
Any reason to have this default to y ?
Review
Quoted 2 lines
(quoted diff) +2
-0+ help + Blackfin Linux kernel supports running both FDPIC and FLAT applications
The Blackfin Linux kernel ...
Review
Quoted 1 line
(quoted diff) +1
-0+ concurrently if the binary format specific libraries installed properly.
are installed properly.
arch/Config.in.bfin config BR2_bf561 · line 72
Quoted 1 line
arch/Config.in.bfin +1
-0+ This option allow developer to install FDPIC libraries into a buildroot
allows the developer
arch/Config.in.bfin config BR2_bf561 · line 73
Quoted 1 line
arch/Config.in.bfin +1
-0+ rootfs image built with binary format macro other than BR2_BINFMT_FDPIC.
built with a binary format that is not FDPIC.
arch/Config.in.bfin config BR2_bf561 · line 84
Quoted 11 lines
arch/Config.in.bfin +11
-0+ +config BR2_BFIN_INSTALL_FLAT_SHARED + depends on BR2_bfin && !BR2_BINFMT_FLAT_SHARED + bool "Install FLAT shared libraries" + default y + help + Blackfin Linux kernel supports running both FDPIC and FLAT applications + concurrently if the binary format specific libraries installed properly. + This option allow developer to install FLAT libraries into a buildroot + rootfs image built with binary format macro other than + BR2_BINFMT_FLAT_SHARED.
Same typos.
Quoted 18 lines
arch/Makefile.in.bfin new file +12
-0@@ -0,0 +1,22 @@ +TARGETS_BFIN-$(BR2_BFIN_INSTALL_FDPIC_SHARED) += target-shared-libs-fdpic +TARGETS_BFIN-$(BR2_BFIN_INSTALL_FLAT_SHARED) += target-shared-libs-flat +TARGETS += $(TARGETS_BFIN-y) + +CROSS_COMPILE_SHARED_FDPIC ?= bfin-linux-uclibc- +target-shared-libs-fdpic: + set -e; \ + t=`$(CROSS_COMPILE_SHARED_FDPIC)gcc $(CPUFLAGS) -print-file-name=libc.a`; \ + t=`dirname $$t`/../..; \ + for i in $$t/lib/*so*; do \ + i=`readlink -f "$$i"`; \ + soname=`$(CROSS_COMPILE_SHARED_FDPIC)readelf -d "$$i" | sed -n '/(SONAME)/s:.*[[]\(.*\)[]].*:\1:p'`; \
arch/Makefile.in.bfin line 22
Quoted 10 lines
arch/Makefile.in.bfin new file +10
-0+ $(INSTALL) -D $$i $(TARGET_DIR)/lib/$$soname; \ + done + +CROSS_COMPILE_SHARED_FLAT ?= bfin-uclinux- +target-shared-libs-flat: + set -e; \ + t=`$(CROSS_COMPILE_SHARED_FLAT)gcc $(CPUFLAGS) -mid-shared-library -print-file-name=libc`; \ + if [ -f $$t -a ! -h $$t ] ; then \ + $(INSTALL) -D $$t $(TARGET_DIR)/lib/lib1.so; \ + fi
I am still not happy with this, because it simply doesn't work unless you have the Blackfin toolchain installed globally on your system, which simply isn't the case for normal users.
So I believe we should extend the external toolchain logic to be able to use three variants of the Blackfin toolchain:
* Only FDPIC
* Only FLAT
* Both FLAT and FDPIC
And then, this logic should be moved in the external toolchain logic, where we already do the process of copying libraries from the toolchain to the target, and from the toolchain to the staging directory.
Also:
* $(CPUFLAGS) isn't defined anywhere
* the logic to determine which .so files to copy should use the $(LIB_EXTERNAL_LIBS) and $(USR_LIB_EXTERNAL_LIBS) that we already use in ext-tool.mk.
* I don't quite understand the logic you're using for the shared flat library copying. You use -print-file-name=libc, but then you copy it under the name 'lib1.so'. I think a comment above this part would be useful.
Thanks!
Thomas
Signature · 4 lines
Thomas Petazzoni, Free Electrons Kernel, drivers, real-time and embedded Linux development, consulting, training and support. http://free-electrons.com
PATCH v2 6/7 arch: Introduce blackfin-specific Makefile
Hi Thomas,
Quoted 10 lines · Thomas Petazzoni
Dear Sonic Zhang,
arch/Config.in.bfin config BR2_bf561 · line 68
Quoted 4 lines · Sonic Zhang
arch/Config.in.bfin +4
-0+config BR2_BFIN_INSTALL_FDPIC_SHARED + depends on BR2_bfin && !BR2_BINFMT_FDPIC + bool "Install FDPIC shared libraries" + default yAny reason to have this default to y ?
No. Can be removed.
Quoted 18 lines
Review
Quoted 3 lines
(quoted diff) +2
-0+ help + Blackfin Linux kernel supports running both FDPIC and FLAT applicationsThe Blackfin Linux kernel ...
Quoted 1 line
+ concurrently if the binary format specific libraries installed properly.
are installed properly.
Quoted 1 line
+ This option allow developer to install FDPIC libraries into a buildroot
allows the developer
Quoted 1 line
+ rootfs image built with binary format macro other than
BR2_BINFMT_FDPIC.built with a binary format that is not FDPIC.
Quoted 13 lines
arch/Config.in.bfin config BR2_bf561 · line 84
Quoted 11 lines
arch/Config.in.bfin +11
-0+ +config BR2_BFIN_INSTALL_FLAT_SHARED + depends on BR2_bfin && !BR2_BINFMT_FLAT_SHARED + bool "Install FLAT shared libraries" + default y + help + Blackfin Linux kernel supports running both FDPIC and FLAT applications + concurrently if the binary format specific libraries installed properly. + This option allow developer to install FLAT libraries into a buildroot + rootfs image built with binary format macro other than + BR2_BINFMT_FLAT_SHARED.Same typos.
Thanks.
Quoted 37 lines
Quoted 19 lines
arch/Makefile.in.bfin new file +12
-0@@ -0,0 +1,22 @@ +TARGETS_BFIN-$(BR2_BFIN_INSTALL_FDPIC_SHARED) += target-shared-libs-fdpic +TARGETS_BFIN-$(BR2_BFIN_INSTALL_FLAT_SHARED) += target-shared-libs-flat +TARGETS += $(TARGETS_BFIN-y) + +CROSS_COMPILE_SHARED_FDPIC ?= bfin-linux-uclibc- +target-shared-libs-fdpic: + set -e; \ + t=`$(CROSS_COMPILE_SHARED_FDPIC)gcc $(CPUFLAGS) -print-file-name=libc.a`; \ + t=`dirname $$t`/../..; \ + for i in $$t/lib/*so*; do \ + i=`readlink -f "$$i"`; \ + soname=`$(CROSS_COMPILE_SHARED_FDPIC)readelf -d "$$i" | sed -n '/(SONAME)/s:.*[[]\(.*\)[]].*:\1:p'`; \arch/Makefile.in.bfin line 22
Quoted 10 lines
arch/Makefile.in.bfin new file +10
-0+ $(INSTALL) -D $$i $(TARGET_DIR)/lib/$$soname; \ + done + +CROSS_COMPILE_SHARED_FLAT ?= bfin-uclinux- +target-shared-libs-flat: + set -e; \ + t=`$(CROSS_COMPILE_SHARED_FLAT)gcc $(CPUFLAGS) -mid-shared-library -print-file-name=libc`; \ + if [ -f $$t -a ! -h $$t ] ; then \ + $(INSTALL) -D $$t $(TARGET_DIR)/lib/lib1.so; \ + fiI am still not happy with this, because it simply doesn't work unless you have the Blackfin toolchain installed globally on your system, which simply isn't the case for normal users.
So I believe we should extend the external toolchain logic to be able to use three variants of the Blackfin toolchain:
Quoted 7 lines
* Only FDPIC
* Only FLAT
* Both FLAT and FDPICAnd then, this logic should be moved in the external toolchain logic, where we already do the process of copying libraries from the toolchain to the target, and from the toolchain to the staging directory.
Since FPDIC libraries are installed by default if BINFMT_FDPIC is selected, how about making INSTALL_FDPIC_LIB option depend on BINFMT_FLAT only? So do the INSTALL_FLAT_LIB option. And no "Both FLAT and FDPIC" is needed in this case.
Quoted 4 lines
Also:
* $(CPUFLAGS) isn't defined anywhere
Yes, forgot to define it after switching to the toolchain wrapper solution.
Quoted 3 lines
* the logic to determine which .so files to copy should use the $(LIB_EXTERNAL_LIBS) and $(USR_LIB_EXTERNAL_LIBS) that we already use in ext-tool.mk.
This only applies to the FDPIC libraries. The FLAT libraries use a different name policy.
Quoted 4 lines
* I don't quite understand the logic you're using for the shared flat library copying. You use -print-file-name=libc, but then you copy it under the name 'lib1.so'. I think a comment above this part would be useful.
The flat libraries are found and linked according to the index in it name "libN.so". Index 1 is reserved for the standard C library. Customer libraries can use 4 and above. That's why the libc is renamed to lib1.so in target rootfs.
Regards,
Sonic.
PATCH v2 6/7 arch: Introduce blackfin-specific Makefile
Dear Sonic Zhang,
Quoted 11 lines · Sonic Zhang
Quoted 11 lines
I am still not happy with this, because it simply doesn't work unless you have the Blackfin toolchain installed globally on your system, which simply isn't the case for normal users.
So I believe we should extend the external toolchain logic to be able to use three variants of the Blackfin toolchain:
* Only FDPIC
* Only FLAT
* Both FLAT and FDPIC
Quoted 7 lines
Quoted 3 lines
And then, this logic should be moved in the external toolchain logic, where we already do the process of copying libraries from the toolchain to the target, and from the toolchain to the staging directory.
Since FPDIC libraries are installed by default if
BINFMT_FDPICis selected, how about makingINSTALL_FDPIC_LIBoption depend onBINFMT_FLATonly? So do theINSTALL_FLAT_LIBoption.
Yes, this makes sense.
Quoted 1 line
And no "Both FLAT and FDPIC" is needed in this case.
In fact, I looked again at our support for Blackfin external toolchains, and I was wrong: we always install *both* the FDPIC and Flat shared libraries. So in fact, you don't need to handle the variants of the external toolchain.
Just move your logic into the external toolchain logic that copies libraries, and that should be fine. I believe it belongs to the external toolchain logic, because it is highly specific to Analog Devices toolchains. If we ever support Blackfin in the internal backend, it will always either be FDPIC *or* FLAT, but not both at the same time.
Quoted 6 lines
Quoted 3 lines
* the logic to determine which .so files to copy should use the $(LIB_EXTERNAL_LIBS) and $(USR_LIB_EXTERNAL_LIBS) that we already use in ext-tool.mk.
This only applies to the FDPIC libraries. The FLAT libraries use a different name policy.
Yes, sure, but I was referring to the FDPIC libraries in this part (sorry if it wasn't clear).
Quoted 9 lines
Quoted 4 lines
* I don't quite understand the logic you're using for the shared flat library copying. You use -print-file-name=libc, but then you copy it under the name 'lib1.so'. I think a comment above this part would be useful.
The flat libraries are found and linked according to the index in it name "libN.so". Index 1 is reserved for the standard C library. Customer libraries can use 4 and above. That's why the libc is renamed to lib1.so in target rootfs.
Oh, ok, interesting. Could you just copy/paste this explanation as a comment above the relevant part of the code?
Thanks a lot Sonic for your very quick feedback in this discussion, I think we're making great progress!
Thanks,
Thomas
Signature · 4 lines
Thomas Petazzoni, Free Electrons Kernel, drivers, real-time and embedded Linux development, consulting, training and support. http://free-electrons.com
PATCH v2 7/7 package: Introduce NOMMU symbol
In reply to Sonic ZhangFrom: Sonic Zhang <[email protected]>
v2-changes:
none
v1-chagnes:
- Introduce NOMMU symbol __NOMMU__
--- package/Makefile.in | 4 ++++ 1 files changed, 4 insertions(+), 0 deletions(-)
package/Makefile.in +4 -0
@@ -103,6 +103,10 @@ TARGET_CFLAGS = $(TARGET_CPPFLAGS) $(TARGET_ABI) $(TARGET_OPTIMIZATION) $(TARGET TARGET_CXXFLAGS = $(TARGET_CFLAGS) TARGET_LDFLAGS = $(call qstrip,$(BR2_TARGET_LDFLAGS)) +ifneq ($(BR2_USE_MMU), y) +TARGET_CFLAGS += -D__NOMMU__ +endif + ifeq ($(BR2_TOOLCHAIN_BUILDROOT)$(BR2_TOOLCHAIN_CTNG),y) TARGET_CROSS=$(HOST_DIR)/usr/bin/$(GNU_TARGET_NAME)- else
Signature · 1 line
1.7.0.4
PATCH v2 1/7 arch: Add blackfin CPU choice.
In reply to Sonic ZhangDear Sonic Zhang,
Thanks for continuing the work on this!
Quoted 19 lines · Sonic Zhang
From: Sonic Zhang <[email protected]>
v2-changes:
nonev1-changes:
- is patch adds a Target CPU configuration option and uses it to select a -m option for gcc.Signed-off-by Sonic Zhang <[email protected]>--- arch/Config.in.bfin | 97 +++++++++++++++++++++++++++++++++++++++++++++++---- 1 files changed, 90 insertions(+), 7 deletions(-)arch/Config.in.bfin +0
-0
arch/Config.in.bfin line 9
Quoted 10 lines
arch/Config.in.bfin +1
-7@@ -1,12 +1,65 @@ choice - prompt "Target ABI" + prompt "Target CPU" depends on BR2_bfin - default BR2_BFIN_FDPIC -config BR2_BFIN_FDPIC - bool "FDPIC" -config BR2_BFIN_FLAT - bool "FLAT" - select BR2_PREFER_STATIC_LIB
You can't just remove BR2_BFIN_FDPIC and BR2_BFIN_FLAT, they are used in the external toolchain logic of Buildroot:
$ git grep BR2_BFIN_FDPIC . arch/Config.in.bfin: default BR2_BFIN_FDPIC arch/Config.in.bfin:config BR2_BFIN_FDPIC toolchain/toolchain-external/Config.in: default "bfin-uclinux" if BR2_TOOLCHAIN_EXTERNAL_BLACKFIN_UCLINUX_2011R1 && !BR2_BFIN_FDPIC toolchain/toolchain-external/Config.in: default "bfin-linux-uclibc" if BR2_TOOLCHAIN_EXTERNAL_BLACKFIN_UCLINUX_2011R1 && BR2_BFIN_FDPIC toolchain/toolchain-external/Config.in: default "bfin-uclinux" if BR2_TOOLCHAIN_EXTERNAL_BLACKFIN_UCLINUX_2012R1 && !BR2_BFIN_FDPIC toolchain/toolchain-external/Config.in: default "bfin-linux-uclibc" if BR2_TOOLCHAIN_EXTERNAL_BLACKFIN_UCLINUX_2012R1 && BR2_BFIN_FDPIC toolchain/toolchain-external/Config.in: default "bfin-uclinux" if BR2_TOOLCHAIN_EXTERNAL_BLACKFIN_UCLINUX_2012R2 && !BR2_BFIN_FDPIC toolchain/toolchain-external/Config.in: default "bfin-linux-uclibc" if BR2_TOOLCHAIN_EXTERNAL_BLACKFIN_UCLINUX_2012R2 && BR2_BFIN_FDPIC
So, I think your patch 3/7 should be first in the series, and it should be followed by a patch that removes BR2_BFIN_FDPIC and BR2_BFIN_FLAT and use the new BR2_BINFMT_* symbols in the external toolchain logic.
Basically, your patch set should not break things! When you remove a symbol, make sure it isn't used anymore.
Other than that, I'm fine with this patch adding the blackfin CPU variants.
Best regards,
Thomas
Signature · 4 lines
Thomas Petazzoni, Free Electrons Kernel, drivers, real-time and embedded Linux development, consulting, training and support. http://free-electrons.com
PATCH v2 1/7 arch: Add blackfin CPU choice.
Hi Thomas,
Quoted 48 lines · Thomas Petazzoni
Dear Sonic Zhang,
Thanks for continuing the work on this!
Quoted 19 lines · Sonic Zhang
From: Sonic Zhang <[email protected]>
v2-changes:
nonev1-changes:
- is patch adds a Target CPU configuration option and uses it to select a -m option for gcc.Signed-off-by Sonic Zhang <[email protected]>--- arch/Config.in.bfin | 97 +++++++++++++++++++++++++++++++++++++++++++++++---- 1 files changed, 90 insertions(+), 7 deletions(-)arch/Config.in.bfin +0
-0arch/Config.in.bfin line 9
Quoted 10 lines
arch/Config.in.bfin +1
-7@@ -1,12 +1,65 @@ choice - prompt "Target ABI" + prompt "Target CPU" depends on BR2_bfin - default BR2_BFIN_FDPIC -config BR2_BFIN_FDPIC - bool "FDPIC" -config BR2_BFIN_FLAT - bool "FLAT" - select BR2_PREFER_STATIC_LIBYou can't just remove
BR2_BFIN_FDPICandBR2_BFIN_FLAT, they are used in the external toolchain logic of Buildroot:$ git grep BR2_BFIN_FDPIC . arch/Config.in.bfin: default BR2_BFIN_FDPIC arch/Config.in.bfin:config BR2_BFIN_FDPIC toolchain/toolchain-external/Config.in: default "bfin-uclinux" if BR2_TOOLCHAIN_EXTERNAL_BLACKFIN_UCLINUX_2011R1 && !BR2_BFIN_FDPIC toolchain/toolchain-external/Config.in: default "bfin-linux-uclibc" if BR2_TOOLCHAIN_EXTERNAL_BLACKFIN_UCLINUX_2011R1 && BR2_BFIN_FDPIC toolchain/toolchain-external/Config.in: default "bfin-uclinux" if BR2_TOOLCHAIN_EXTERNAL_BLACKFIN_UCLINUX_2012R1 && !BR2_BFIN_FDPIC toolchain/toolchain-external/Config.in: default "bfin-linux-uclibc" if BR2_TOOLCHAIN_EXTERNAL_BLACKFIN_UCLINUX_2012R1 && BR2_BFIN_FDPIC toolchain/toolchain-external/Config.in: default "bfin-uclinux" if BR2_TOOLCHAIN_EXTERNAL_BLACKFIN_UCLINUX_2012R2 && !BR2_BFIN_FDPIC toolchain/toolchain-external/Config.in: default "bfin-linux-uclibc" if BR2_TOOLCHAIN_EXTERNAL_BLACKFIN_UCLINUX_2012R2 && BR2_BFIN_FDPIC
Quoted 4 lines
So, I think your patch 3/7 should be first in the series, and it should be followed by a patch that removes
BR2_BFIN_FDPICandBR2_BFIN_FLATand use the new BR2_BINFMT_* symbols in the external toolchain logic.
OK. I will move 3/7 ahead.
Regards,
Sonic