natty gcc-4.5 kernel (2.6.37-rc1) build failure

[Robert Nelson]

Binary package hint: gcc-4.5

On Natty, I'm seeing build failure of the Linux Kernel, where i wasn't seeing them on Maverick, I can't replicate the issue with the maverick x86 gcc-4.5 armel cross compiler in the maverick repo..

Full logs of my builds: http://rcn-ee.homeip.net:81/dl/farm/log/

Natty usually fails with:

arch/arm/mach-omap2/omap-headsmp.S: Assembler messages:
arch/arm/mach-omap2/omap-headsmp.S:36: Error: selected processor does not support ARM mode `smc #0'
make[3]: *** [arch/arm/mach-omap2/omap-headsmp.o] Error 1
make[2]: *** [arch/arm/mach-omap2] Error 2
make[2]: *** Waiting for unfinished jobs....

To replicate, in either a natty chroot or running on an armv7 target board:

With the latest 2.6.37-rc1 tar

make ARCH=arm CROSS_COMPILE= omap2plus_defconfig

make ARCH=arm CROSS_COMPILE= menuconfig

then drop omap2/armv6 support, now only omap3/4/armv7-a enabled

System Type -> TI OMAP2/3/4 Specific Features -> (uncheck omap2)

Otherwise config: http://pastebin.com/WHy9RrqS

make ARCH=arm CROSS_COMPILE= uImage

Error:

  CC arch/arm/mach-omap2/omap-smp.o
  AS arch/arm/mach-omap2/omap-headsmp.o
arch/arm/mach-omap2/omap-headsmp.S: Assembler messages:
arch/arm/mach-omap2/omap-headsmp.S:36: Error: selected processor does not support ARM mode `smc #0'
make[1]: *** [arch/arm/mach-omap2/omap-headsmp.o] Error 1
make: *** [arch/arm/mach-omap2] Error 2

full build log: http://pastebin.com/cTWaxbwg

From http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=arch/arm/mach-omap2/Makefile;h=60e51bcf53bd2e053a384826d45ac760eb30588d;hb=HEAD#l29

AFLag settings: :=-Wa,-march=armv7-a

GCC:
voodoo@beagle-256mb-1:~/omap4-kernel/linux-2.6.37-rc1$ gcc -v
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/arm-linux-gnueabi/4.5.1/lto-wrapper
Target: arm-linux-gnueabi
Configured with: ../src/configure -v --with-pkgversion='Ubuntu/Linaro 4.5.1-9ubuntu1' --with-bugurl=file:///usr/share/doc/gcc-4.5/README.Bugs --enable-languages=c,c++,fortran,objc,obj-c++ --prefix=/usr --program-suffix=-4.5 --enable-shared --enable-multiarch --enable-linker-build-id --with-system-zlib --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --with-gxx-include-dir=/usr/include/c++/4.5 --libdir=/usr/lib --enable-nls --with-sysroot=/ --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-plugin --enable-objc-gc --disable-sjlj-exceptions --with-arch=armv7-a --with-float=softfp --with-fpu=vfpv3-d16 --with-mode=thumb --disable-werror --enable-checking=release --build=arm-linux-gnueabi --host=arm-linux-gnueabi --target=arm-linux-gnueabi
Thread model: posix
gcc version 4.5.1 (Ubuntu/Linaro 4.5.1-9ubuntu1)

[Matthias Klose (doko)]

that might be related to bug #673085, however somebody did tell me that the kernel builds with
-Wa,-mimplicit-it=thumb by default.

[Robert Nelson (robertcnelson)]

Hi Matthias,

Thank you for taking some time to look at this..

Versions:
gcc-4.5: gcc version 4.5.1 (Ubuntu/Linaro 4.5.1-9ubuntu1)
gcc-4.4: gcc version 4.4.5 (Ubuntu/Linaro 4.4.5-7ubuntu1)
as: GNU assembler version 2.20.90

So grepping the (arch/arm/Makfile) there's one occurance of 'mimplicit-it', and it's in the "build kernel in thumb2 mode config" which i usually don't enable..

ifeq ($(CONFIG_THUMB2_KERNEL),y)
AFLAGS_AUTOIT :=$(call as-option,-Wa$(comma)-mimplicit-it=always,-Wa$(comma)-mauto-it)
AFLAGS_NOWARN :=$(call as-option,-Wa$(comma)-mno-warn-deprecated,-Wa$(comma)-W)
CFLAGS_THUMB2 :=-mthumb $(AFLAGS_AUTOIT) $(AFLAGS_NOWARN)
AFLAGS_THUMB2 :=$(CFLAGS_THUMB2) -Wa$(comma)-mthumb
endif

So with that config enabled, it fails much sooner at:

  AS arch/arm/nwfpe/entry.o
arch/arm/nwfpe/entry.S: Assembler messages:
arch/arm/nwfpe/entry.S:96: Error: cannot use post-indexing with this instruction -- `ldrt r6,[r5],#4'
make[1]: *** [arch/arm/nwfpe/entry.o] Error 1
make: *** [arch/arm/nwfpe] Error 2

So playing around with different things, i forced gcc-4.4 over gcc-4.5 in the kernel Makfile with:

HOSTCC = gcc-4.4
HOSTCXX = g++-4.4

and it still fails...

  AS arch/arm/mach-omap2/omap-headsmp.o
arch/arm/mach-omap2/omap-headsmp.S: Assembler messages:
arch/arm/mach-omap2/omap-headsmp.S:36: Error: selected processor does not support ARM mode `smc #0'
make[1]: *** [arch/arm/mach-omap2/omap-headsmp.o] Error 1
make: *** [arch/arm/mach-omap2] Error 2

Makes me wonder if this might be a binutils issue instead of gcc-4.5?

Regards,

[Robert Nelson (robertcnelson)]

Clarity Note: the playing around with gcc-4.4 is back with the original config mentioned in this bug.. (thumb2 kernel disabled)

[Anand Gadiyar (gadiyar)]

Does reverting commit df571c4ab32da9bc05a7db98516fc06ef3cdc5d1 (omap4: Fix build break when used with gcc-4.4.1 (2009-q3) ) help?

[Robert Nelson (robertcnelson)]

Hi Anand,

No luck, using 2.6.37-rc3, and manually reverting that commit (changed smc #0 to smc)..

  CC arch/arm/mach-omap2/prm44xx.o
  CC arch/arm/mach-omap2/mcbsp.o
  CC arch/arm/mach-omap2/omap-smp.o
  AS arch/arm/mach-omap2/omap-headsmp.o
arch/arm/mach-omap2/omap-headsmp.S: Assembler messages:
arch/arm/mach-omap2/omap-headsmp.S:36: Error: selected processor does not support ARM mode `smc'
make[1]: *** [arch/arm/mach-omap2/omap-headsmp.o] Error 1
make: *** [arch/arm/mach-omap2] Error 2
voodoo@beagle-256mb-1:~/omap4-kernel/linux-2.6.37-rc3$ cat arch/arm/mach-omap2/omap-headsmp.S | grep smc
 smc @ read from AuxCoreBoot0

using natty's: gcc version 4.5.2 (Ubuntu/Linaro 4.5.1-10ubuntu3)

[Robert Nelson (robertcnelson)]

It might be the wrong rabbit hole, but there were some tweaks to 'smc' in binutils's changelog:

http://bazaar.launchpad.net/~vcs-imports/binutils/master/revision/22100

http://sourceware.org/cgi-bin/cvsweb.cgi/src/gas/config/tc-arm.c.diff?r1=1.464&r2=1.465&cvsroot=src&f=h

[Michael Hope (michaelh1)]

My reading of this is that 'smc' isn't in the base Cortex-A profile but requires the security extensions present. I think this can be enabled using '-march=armv7-a+sec' or (preferably) adding '.arch_extension sec' to the assembly code.

[Robert Nelson (robertcnelson)]

Interesting Michael,

  AS arch/arm/mach-omap2/omap-headsmp.o
  CC arch/arm/mach-omap2/omap-hotplug.o
  AS arch/arm/mach-omap2/omap44xx-smc.o
  CC arch/arm/mach-omap2/omap4-common.o
  AS arch/arm/mach-omap2/sram34xx.o

diff --git a/arch/arm/mach-omap2/Makefile b/arch/arm/mach-omap2/Makefile
index 60e51bc..1b931b4 100644
--- a/arch/arm/mach-omap2/Makefile
+++ b/arch/arm/mach-omap2/Makefile
@@ -26,8 +26,8 @@ obj-$(CONFIG_LOCAL_TIMERS) += timer-mpu.o
 obj-$(CONFIG_HOTPLUG_CPU) += omap-hotplug.o
 obj-$(CONFIG_ARCH_OMAP4) += omap44xx-smc.o omap4-common.o

-AFLAGS_omap-headsmp.o :=-Wa,-march=armv7-a
-AFLAGS_omap44xx-smc.o :=-Wa,-march=armv7-a
+AFLAGS_omap-headsmp.o :=-Wa,-march=armv7-a+sec
+AFLAGS_omap44xx-smc.o :=-Wa,-march=armv7-a+sec

 # Functions loaded to SRAM
 obj-$(CONFIG_ARCH_OMAP2420) += sram242x.o

I'll play around with it a little more when i get home in just a little bit... This shouldn't break anything before natty?

[Robert Nelson (robertcnelson)]

it'll take 4-5 hours for build and test of the image..

One big limitation with this patch, it breaks older binutils..

Assembler messages:
Error: unknown architecture `armv7-a+sec'

Error: unrecognized option -march=armv7-a+sec
make[1]: *** [arch/arm/mach-omap2/omap-headsmp.o] Error 1
make: *** [arch/arm/mach-omap2] Error 2
make: *** Waiting for unfinished jobs....

This was with maverick's cross gcc
gcc version 4.5.1 (Ubuntu/Linaro 4.5.1-7ubuntu1)
GNU assembler version 2.20.51 (arm-linux-gnueabi) using BFD version (GNU Binutils for Ubuntu) 2.20.51.20100908

[John Rigby (jcrigby)]

The following is compatible with old and new binutils:

diff --git a/arch/arm/mach-omap2/Makefile b/arch/arm/mach-omap2/Makefile
index 60e51bc..ee9ef4f 100644
--- a/arch/arm/mach-omap2/Makefile
+++ b/arch/arm/mach-omap2/Makefile
@@ -26,8 +26,9 @@ obj-$(CONFIG_LOCAL_TIMERS) += timer-mpu.o
 obj-$(CONFIG_HOTPLUG_CPU) += omap-hotplug.o
 obj-$(CONFIG_ARCH_OMAP4) += omap44xx-smc.o omap4-common.o

-AFLAGS_omap-headsmp.o :=-Wa,-march=armv7-a
-AFLAGS_omap44xx-smc.o :=-Wa,-march=armv7-a
+plus_sec := $(call as-instr,.arch_extension sec,+sec)
+AFLAGS_omap-headsmp.o :=-Wa,-march=armv7-a$(plus_sec)
+AFLAGS_omap44xx-smc.o :=-Wa,-march=armv7-a$(plus_sec)

 # Functions loaded to SRAM
 obj-$(CONFIG_ARCH_OMAP2420) += sram242x.o

[Matthias Klose (doko)]

closing the gcc-4.5 task

[Ricardo Salveti (rsalveti)]

The change at comment #10 is the one now being used by natty's omap 4 kernel: http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-natty.git;a=commitdiff;h=1ebe13dea156358f1044a7a755f02945ad3890e4;hp=f31ff7673ef77d04e74e1a1e45bb2940781f35f0

Plus http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-natty.git;a=commitdiff;h=99340d7b23f3df75d9b752fb25088a10064cd0a6;hp=9523bd6810e0229a8147e8c3d6498e2f40b38974

[Ricardo Salveti (rsalveti)]

Robert, were you able to check if these patches fix your issue?

[Robert Nelson (robertcnelson)]

Hi Ricardo,

Yeap, this patch fixes the issue i was seeing with building both 2.6.35 and 2.6.37-rc's in natty.. and no regressions on the old userspaces, (lenny/karmic)..

I haven't seen a need for the 2nd one, "AFLAGS_sleep44xx.o :=-Wa,-march=armv7-a$(plus_sec)" as i don't have the omap4 power management stuff as you guys..

With the merge 2.6.38 merge on the horizon, we better get this in.. ;)

[Jeremy Foshee (jeremyfoshee)]

Hi Robert,

Please be sure to confirm this issue exists with the latest development release of Ubuntu. ISO CD images are available from http://cdimage.ubuntu.com/daily/current/ . If the issue remains, please run the following command from a Terminal (Applications->Accessories->Terminal). It will automatically gather and attach updated debug information to this report.

apport-collect -p linux 669912

Also, if you could test the latest upstream kernel available that would be great. It will allow additional upstream developers to examine the issue. Refer to https://wiki.ubuntu.com/KernelMainlineBuilds . Once you've tested the upstream kernel, please remove the 'needs-upstream-testing' tag. This can be done by clicking on the yellow pencil icon next to the tag located at the bottom of the bug description and deleting the 'needs-upstream-testing' text. Please let us know your results.

Thanks in advance.

    [This is an automated message. Apologies if it has reached you inappropriately; please just reply to this message indicating so.]