[U-Boot] [PATCH] Link the correct libgcc.a on ARMv4T architecture
If we use $(CC -print-libgcc-file-name) without option -march=armv4t, we will link the wrong version libgcc.a(for ARMv5). That causes cpu core raising excptions when u-boot running on the board.
Signed-off-by:Paul Hsu pochun.hsu@gmail.com --- Makefile | 15 +++++++++++++++ 1 files changed, 15 insertions(+), 0 deletions(-)
diff --git a/Makefile b/Makefile index 4f1cb1b..978c9cd 100644 --- a/Makefile +++ b/Makefile @@ -259,6 +259,17 @@ LIBS := $(addprefix $(obj),$(LIBS)) LIBBOARD = board/$(BOARDDIR)/lib$(BOARD).a LIBBOARD := $(addprefix $(obj),$(LIBBOARD))
+#check ARMV4T +ifdef CPU +ifeq ("$(CPU)","arm920t") +ARMV4T := yes +endif +ifeq ("$(CPU)","arm720t") +ARMV4T := yes +endif +endif + + # Add GCC lib ifdef USE_PRIVATE_LIBGCC ifeq ("$(USE_PRIVATE_LIBGCC)", "yes") @@ -267,8 +278,12 @@ else PLATFORM_LIBGCC = -L $(USE_PRIVATE_LIBGCC) -lgcc endif else +ifeq ("$(ARMV4T)","yes") +PLATFORM_LIBGCC = -L $(shell dirname `$(CC) $(CFLAGS) -print-libgcc-file-name -march=armv4t`) -lgcc +else PLATFORM_LIBGCC = -L $(shell dirname `$(CC) $(CFLAGS) -print-libgcc-file-name`) -lgcc endif +endif PLATFORM_LIBS += $(PLATFORM_LIBGCC) export PLATFORM_LIBS
Dear Paul Hsu,
In message 1283932498-8989-1-git-send-email-pochun.hsu@gmail.com you wrote:
If we use $(CC -print-libgcc-file-name) without option -march=armv4t, we will link the wrong version libgcc.a(for ARMv5). That causes cpu core raising excptions when u-boot running on the board.
On which tool chain is this causing problems?
And how does this problem manifest?
Signed-off-by:Paul Hsu pochun.hsu@gmail.com
Makefile | 15 +++++++++++++++ 1 files changed, 15 insertions(+), 0 deletions(-)
NAK. If really needed, this needs to be fixed in ARM specific config files / Makefiles, not in the top level Makefile.
Best regards,
Wolfgang Denk
On Wed, Sep 8, 2010 at 4:16 PM, Wolfgang Denk wd@denx.de wrote:
Dear Paul Hsu,
In message 1283932498-8989-1-git-send-email-pochun.hsu@gmail.com you wrote:
If we use $(CC -print-libgcc-file-name) without option -march=armv4t, we will link the wrong version libgcc.a(for ARMv5). That causes cpu core raising excptions when u-boot running on the board.
On which tool chain is this causing problems?
I use toolchain released by code sourcery(Sourcery G++ Lite 2010q1-202 for ARM GNU/Linux).
$ /opt/arm-2010q1/bin/arm-none-linux-gnueabi-gcc -print-libgcc-file-name -march=armv4t /opt/arm-2010q1/bin/../lib/gcc/arm-none-linux-gnueabi/4.4.1/armv4t/libgcc.a This libgcc.a is a correct one we need on ARMv4T.
If you use the same command wihtout -march=armv4t $ /opt/arm-2010q1/bin/arm-none-linux-gnueabi-gcc -print-libgcc-file-name /opt/arm-2010q1/bin/../lib/gcc/arm-none-linux-gnueabi/4.4.1/libgcc.a This is not the one we need.
And how does this problem manifest?
Signed-off-by:Paul Hsu pochun.hsu@gmail.com
Makefile | 15 +++++++++++++++ 1 files changed, 15 insertions(+), 0 deletions(-)
NAK. If really needed, this needs to be fixed in ARM specific config files / Makefiles, not in the top level Makefile.
I agree that we should find a better place to put ARM specific config in. But currently, top level Makefile is the one I found. Any suggestion?
BR, Paul
Best regards,
Wolfgang Denk
-- DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de core error - bus dumped
Dear Paul Hsu,
In message AANLkTin=dbkiHNGHO2W+knxaC3mk9TFJacQD9d8BwTHf@mail.gmail.com you wrote:
On which tool chain is this causing problems?
I use toolchain released by code sourcery(Sourcery G++ Lite 2010q1-202 for ARM GNU/Linux).
Thanks for the information.
And how does this problem manifest?
You did not answer this question. Do you bet build errors? Or does the build complete, but U-Boot does not work correctly?
NAK. If really needed, this needs to be fixed in ARM specific config files / Makefiles, not in the top level Makefile.
I agree that we should find a better place to put ARM specific config in. But currently, top level Makefile is the one I found. Any suggestion?
arch/arm/config.mk ?
Best regards,
Wolfgang Denk
On Wed, Sep 8, 2010 at 5:05 PM, Wolfgang Denk wd@denx.de wrote:
Dear Paul Hsu,
In message AANLkTin=dbkiHNGHO2W+knxaC3mk9TFJacQD9d8BwTHf@mail.gmail.com you wrote:
On which tool chain is this causing problems?
I use toolchain released by code sourcery(Sourcery G++ Lite 2010q1-202 for ARM GNU/Linux).
Thanks for the information.
And how does this problem manifest?
You did not answer this question. Do you bet build errors? Or does the build complete, but U-Boot does not work correctly?
Build process is complete, but U-Boot does not work correctly. Because of linking the wrong version libgcc.a, which contains armv5 instruction such that arm core raises a undefined instruction exception then code hanged.
NAK. If really needed, this needs to be fixed in ARM specific config files / Makefiles, not in the top level Makefile.
I agree that we should find a better place to put ARM specific config in. But currently, top level Makefile is the one I found. Any suggestion?
arch/arm/config.mk ?
I found another way to solve this issue.
In arch/arm/cpu/arm920t/config.mk, change this line from PLATFORM_CPPFLAGS += -march=armv4 to PLATFORM_CPPFLAGS += -march=armv4t
Then everything goes well. This solution won't make any modification on top level Makefile.
But I don't know exactly that the solution will incur other side effect or not?
BTW, If this solution is the right thing to do, shall I issue another patch?
BR, Paul
Best regards,
Wolfgang Denk
-- DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de If some day we are defeated, well, war has its fortunes, good and bad. -- Commander Kor, "Errand of Mercy", stardate 3201.7
participants (2)
-
Paul Hsu -
Wolfgang Denk