Skip to contentSkip to conversation

x86_64, -git: Error: unknown pseudo-op: `.cfi_signal_frame'

build bug Ingo Molnar 6 participants 8 messages Oct 16, 2006 – Oct 17, 2006

lore
Share & tools
lore
#0 lore raw

using latest -git i'm getting this build bug on gcc 3.4:

arch/x86_64/kernel/entry.S: Assembler messages:
arch/x86_64/kernel/entry.S:157: Error: unknown pseudo-op: .cfi_signal_frame' arch/x86_64/kernel/entry.S:215: Error: unknown pseudo-op: .cfi_signal_frame' arch/x86_64/kernel/entry.S:333: Error: unknown pseudo-op: .cfi_signal_frame' arch/x86_64/kernel/entry.S:548: Error: unknown pseudo-op: .cfi_signal_frame'

gcc version 3.4.0 20040129 (Red Hat Linux 3.4.0-0.3)

using gcc 4.1 it doesnt happen

gcc version 4.1.1 20060525 (Red Hat 4.1.1-1)

this is caused by the following commit:

commit adf1423698f00d00b267f7dca8231340ce7d65ef
Author: Jan Beulich <[email protected]>
Date: Tue Sep 26 10:52:41 2006 +0200

reverting that patch solves the build problem and the resulting kernel builds and boots fine.

Ingo

#1 lore raw
Quoted 15 lines · On Mon, 16 Oct 2006 08:10:37 +0200 Ingo Molnar <[email protected]> wrote:

using latest -git i'm getting this build bug on gcc 3.4:

arch/x86_64/kernel/entry.S: Assembler messages:
arch/x86_64/kernel/entry.S:157: Error: unknown pseudo-op: .cfi_signal_frame' arch/x86_64/kernel/entry.S:215: Error: unknown pseudo-op: .cfi_signal_frame' arch/x86_64/kernel/entry.S:333: Error: unknown pseudo-op: .cfi_signal_frame' arch/x86_64/kernel/entry.S:548: Error: unknown pseudo-op: .cfi_signal_frame'

gcc version 3.4.0 20040129 (Red Hat Linux 3.4.0-0.3)

using gcc 4.1 it doesnt happen

gcc version 4.1.1 20060525 (Red Hat 4.1.1-1)

Quoted 9 lines

this is caused by the following commit:

commit adf1423698f00d00b267f7dca8231340ce7d65ef
Author: Jan Beulich <[email protected]>
Date: Tue Sep 26 10:52:41 2006 +0200

reverting that patch solves the build problem and the resulting kernel builds and boots fine.

That patch has obvious copy-n-paste errors:

i386:

cflags-y += $(call as-instr,.cfi_startproc\n.cfi_endproc,-DCONFIG_AS_CFI=1,) AFLAGS += $(call as-instr,.cfi_startproc\n.cfi_endproc,-DCONFIG_AS_CFI=1,)

+# is .cfi_signal_frame supported too?
+cflags-y += $(call as-instr,.cfi_startproc\n.cfi_endproc,-DCONFIG_AS_CFI=1,) +AFLAGS += $(call as-instr,.cfi_startproc\n.cfi_endproc,-DCONFIG_AS_CFI=1,)

So that won't work.

x86_64 appears to get it right:

cflags-y += $(call as-instr,.cfi_startproc\n.cfi_endproc,-DCONFIG_AS_CFI=1,) AFLAGS += $(call as-instr,.cfi_startproc\n.cfi_endproc,-DCONFIG_AS_CFI=1,)

+# is .cfi_signal_frame supported too?
+cflags-y += $(call as-instr,.cfi_startproc\n.cfi_signal_frame\n.cfi_endproc,-DCONFIG_AS_CFI_SIGNAL_FRAME=1,) +AFLAGS += $(call as-instr,.cfi_startproc\n.cfi_signal_frame\n.cfi_endproc,-DCONFIG_AS_CFI_SIGNAL_FRAME=1,)

Later,

+#ifdef CONFIG_AS_CFI_SIGNAL_FRAME
+#define CFI_SIGNAL_FRAME .cfi_signal_frame
+#else
+#define CFI_SIGNAL_FRAME
+#endif

but it's obviously not working. Wanna debug it a bit?

#3 lore raw
In reply to Ingo Molnar
Quoted 16 lines · Ingo Molnar

using latest -git i'm getting this build bug on gcc 3.4:

arch/x86_64/kernel/entry.S: Assembler messages:
arch/x86_64/kernel/entry.S:157: Error: unknown pseudo-op: .cfi_signal_frame' arch/x86_64/kernel/entry.S:215: Error: unknown pseudo-op: .cfi_signal_frame' arch/x86_64/kernel/entry.S:333: Error: unknown pseudo-op: .cfi_signal_frame' arch/x86_64/kernel/entry.S:548: Error: unknown pseudo-op: .cfi_signal_frame'

gcc version 3.4.0 20040129 (Red Hat Linux 3.4.0-0.3)

using gcc 4.1 it doesnt happen

gcc version 4.1.1 20060525 (Red Hat 4.1.1-1)

this is caused by the following commit:

Quoted 8 lines

commit adf1423698f00d00b267f7dca8231340ce7d65ef
Author: Jan Beulich <[email protected]>
Date: Tue Sep 26 10:52:41 2006 +0200

reverting that patch solves the build problem and the resulting kernel builds and boots fine.

Ingo

This is a binutils/gas feature; gcc has nothing to do with it. I take it the gcc 4.1 system is FC5 with binutils 2.16.91.0.6-5 and the gcc 3.4 system is something older?

The kernel build system is supposed to detect gas support for .cfi_signal_frame -- and while the i386 test is obviously broken (the i386 test for the directive doesn't actually use the directive), the AMD64 test looks like it should work (assuming as-instr works).

Signature · 1 line
Nicholas Miell <[email protected]>
#4 lore raw
Quoted 4 lines · Nicholas Miell

The kernel build system is supposed to detect gas support for .cfi_signal_frame -- and while the i386 test is obviously broken (the i386 test for the directive doesn't actually use the directive), the AMD64 test looks like it should work (assuming as-instr works).

the broken one is:

GNU assembler 2.14.90.0.8 20040114

the working one is:

GNU assembler 2.16.91.0.6 20060212

i guess i found the reason for the breakage: i was using a cross-compiling setup via the following script:

make -j165 bzImage ARCH=x86_64 CC='distcc /opt/crosstool/gcc-4.0.2-glibc-2.3.6/x86_64-unknown-linux-gnu/bin/x86_64-unknown-linux-gnu-gcc'

Note that i override 'CC' instead of specifying a 'CROSS' prefix. I suspect this means as-instr does not switch over to the cross-environment and thus mis-detected the gas version?

Ingo

#5 lore raw
Quoted 3 lines · Ingo Molnar

Note that i override 'CC' instead of specifying a 'CROSS' prefix. I suspect this means as-instr does not switch over to the cross-environment and thus mis-detected the gas version?

this did not solve it either - it seems if both CROSS and CC are set then CC overrides it and CROSS is ignored? Removing the CC override solved the problem. But how do i insert the 'distcc' that way? Seems like a Kbuild breakage to me.

Ingo

#2 lore raw
In reply to Andrew Morton
Quoted 59 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line

Andrew Morton <[email protected]> 16.10.06 08:33 >>>

Quoted 15 lines · On Mon, 16 Oct 2006 08:10:37 +0200 Ingo Molnar <[email protected]> wrote:

using latest -git i'm getting this build bug on gcc 3.4:

arch/x86_64/kernel/entry.S: Assembler messages:
arch/x86_64/kernel/entry.S:157: Error: unknown pseudo-op: .cfi_signal_frame' arch/x86_64/kernel/entry.S:215: Error: unknown pseudo-op: .cfi_signal_frame' arch/x86_64/kernel/entry.S:333: Error: unknown pseudo-op: .cfi_signal_frame' arch/x86_64/kernel/entry.S:548: Error: unknown pseudo-op: .cfi_signal_frame'

gcc version 3.4.0 20040129 (Red Hat Linux 3.4.0-0.3)

using gcc 4.1 it doesnt happen

gcc version 4.1.1 20060525 (Red Hat 4.1.1-1)

Quoted 9 lines

this is caused by the following commit:

commit adf1423698f00d00b267f7dca8231340ce7d65ef
Author: Jan Beulich <[email protected]>
Date: Tue Sep 26 10:52:41 2006 +0200

reverting that patch solves the build problem and the resulting kernel builds and boots fine.

That patch has obvious copy-n-paste errors:

i386:

cflags-y += $(call as-instr,.cfi_startproc\n.cfi_endproc,-DCONFIG_AS_CFI=1,) AFLAGS += $(call as-instr,.cfi_startproc\n.cfi_endproc,-DCONFIG_AS_CFI=1,)

(quoted diff) +3 -0
+# is .cfi_signal_frame supported too?
+cflags-y += $(call as-instr,.cfi_startproc\n.cfi_endproc,-DCONFIG_AS_CFI=1,)
+AFLAGS += $(call as-instr,.cfi_startproc\n.cfi_endproc,-DCONFIG_AS_CFI=1,)

So that won't work.

x86_64 appears to get it right:

cflags-y += $(call as-instr,.cfi_startproc\n.cfi_endproc,-DCONFIG_AS_CFI=1,) AFLAGS += $(call as-instr,.cfi_startproc\n.cfi_endproc,-DCONFIG_AS_CFI=1,)

(quoted diff) +3 -0
+# is .cfi_signal_frame supported too?
+cflags-y += $(call as-instr,.cfi_startproc\n.cfi_signal_frame\n.cfi_endproc,-DCONFIG_AS_CFI_SIGNAL_FRAME=1,)
+AFLAGS += $(call as-instr,.cfi_startproc\n.cfi_signal_frame\n.cfi_endproc,-DCONFIG_AS_CFI_SIGNAL_FRAME=1,)

Later,

(quoted diff) +5 -0
+#ifdef CONFIG_AS_CFI_SIGNAL_FRAME
+#define CFI_SIGNAL_FRAME .cfi_signal_frame
+#else
+#define CFI_SIGNAL_FRAME
+#endif
Quoted 1 line

but it's obviously not working. Wanna debug it a bit?

That is nothing I added, I suppose Andi did. But the fix is obvious and trivial.

Jan

#6 lore raw
In reply to Ingo Molnar
Quoted 11 lines · Ingo Molnar
Quoted 3 lines · Ingo Molnar

Note that i override 'CC' instead of specifying a 'CROSS' prefix. I suspect this means as-instr does not switch over to the cross-environment and thus mis-detected the gas version?

this did not solve it either - it seems if both CROSS and CC are set then CC overrides it and CROSS is ignored? Removing the CC override solved the problem. But how do i insert the 'distcc' that way? Seems like a Kbuild breakage to me.

I will try to take a look when I'm back from vacation. In few hours I'm leaving of for France for 8 days with the family. If someone else beats me why I'm away it would be fine!

Sam

#7 lore raw
In reply to Ingo Molnar

Hallo.

Quoted 6 lines · Ingo Molnar
Quoted 3 lines · Ingo Molnar

Note that i override 'CC' instead of specifying a 'CROSS' prefix. I suspect this means as-instr does not switch over to the cross-environment and thus mis-detected the gas version?

Do you mean 'CROSS_COMPILE' ?

Quoted 4 lines

this did not solve it either - it seems if both CROSS and CC are set then CC overrides it and CROSS is ignored? Removing the CC override solved the problem. But how do i insert the 'distcc' that way? Seems like a Kbuild breakage to me.

CC with friends is set like this:

+--[linux/Makefile]
|[...]
|AS		= $(CROSS_COMPILE)as
|LD		= $(CROSS_COMPILE)ld
|CC		= $(CROSS_COMPILE)gcc
|CPP		= $(CC) -E
|[...]
+--

So, i think, you must set 'CROSS_COMPILE' first and then 'CC' like this:
(exporing PATH in shell, of course)
+--

$ make CROSS_COMPILE=x86_64-pc-linux- 'CC=distcc $(CROSS_COMPILE)gcc'
+--

(BTW, it's funny to see arch-*unknown*-linux... Like something is deeply hidden and is a very big secret ;)
____