MIPS: fix build with binutils 2.24.51+

Starting with version 2.24.51.20140728 MIPS binutils complain loudly about mixing soft-float and hard-float object files, leading to this build failure since GCC is invoked with "-msoft-float" on MIPS:

{standard input}: Warning: .gnu_attribute 4,3 requires `softfloat'
LD arch/mips/alchemy/common/built-in.o
mipsel-softfloat-linux-gnu-ld: Warning: arch/mips/alchemy/common/built-in.o
uses -msoft-float (set by arch/mips/alchemy/common/prom.o), arch/mips/alchemy/common/sleeper.o uses -mhard-float

To fix this, we detect if GAS is new enough to support "-msoft-float" command option, and if it does, we can let GCC pass it to GAS; but then we also need to sprinkle the files which make use of floating point registers with the necessary ".set hardfloat" directives.

Signed-off-by Manuel Lauss <[email protected]>

---
Compiles with binutils 2.23 and current git head, tested with alchemy (mips32r1) and maltasmvp_defconfig (64bit)

Tests with MSA and other extensions also appreciated!

v6: #undef fp so that the preprocessor does not replace the fp in

	.set fp=64 with $30...  Fixes 64bit build.

[...]

Quoted 16 lines · Manuel Lauss

Starting with version 2.24.51.20140728 MIPS binutils complain loudly about mixing soft-float and hard-float object files, leading to this build failure since GCC is invoked with "-msoft-float" on MIPS:

{standard input}: Warning: .gnu_attribute 4,3 requires `softfloat'
LD arch/mips/alchemy/common/built-in.o
mipsel-softfloat-linux-gnu-ld: Warning: arch/mips/alchemy/common/built-in.o
uses -msoft-float (set by arch/mips/alchemy/common/prom.o), arch/mips/alchemy/common/sleeper.o uses -mhard-float

To fix this, we detect if GAS is new enough to support "-msoft-float" command option, and if it does, we can let GCC pass it to GAS; but then we also need to sprinkle the files which make use of floating point registers with the necessary ".set hardfloat" directives.

Signed-off-by Manuel Lauss <[email protected]>

Quoted 8 lines

---
Compiles with binutils 2.23 and current git head, tested with alchemy (mips32r1) and maltasmvp_defconfig (64bit)

Tests with MSA and other extensions also appreciated!

v6: #undef fp so that the preprocessor does not replace the fp in

      .set fp=64 with $30...  Fixes 64bit build.

Technically speaking, a maltasmvp_defconfig selects CONFIG_32BIT=y so it's still a 32-bit build.

Quoted 1 line

[...]

Ok the fp problem went away but I still have the even/odd errors with my tools

arch/mips/kernel/r4k_switch.S: Assembler messages:
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 1
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 3
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 5
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 7
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 9
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 11
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 13
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 15
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 17
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 19

The following patch did not help either:

arch/mips/Makefile +1 -1

index 58076472bdd8..b8bb7e170fee 100644
@@ -56,7 +56,7 @@ ifdef CONFIG_FUNCTION_GRAPH_TRACER
   endif
 endif
 cflags-y += $(call cc-option, -mno-check-zero-division)
-
+cflags-y += -mno-odd-spreg
 

This is with a regular maltasmvp_defconfig

I guess my gcc version is newer than yours. Matthew?

--
markos

-----Original Message-----
From: Manuel Lauss [mailto:[email protected]]
Sent: 31 October 2014 16:23
To: Markos Chandras
Cc: Linux-MIPS; Matthew Fortune; Maciej W. Rozycki; Ralf Baechle
Subject: Re: [RFC PATCH v6] MIPS: fix build with binutils 2.24.51+

I didn't encounter this error with what will be gcc-4.9.3.

Manuel

Quoted 8 lines · Markos Chandras

Quoted 7 lines · Manuel Lauss

Starting with version 2.24.51.20140728 MIPS binutils complain loudly about mixing soft-float and hard-float object files, leading to this build failure since GCC is invoked with "-msoft-float" on MIPS:

{standard input}: Warning: .gnu_attribute 4,3 requires `softfloat'
LD arch/mips/alchemy/common/built-in.o
mipsel-softfloat-linux-gnu-ld: Warning: arch/mips/alchemy/common/built-

in.o

Quoted 4 lines

Quoted 4 lines

uses -msoft-float (set by arch/mips/alchemy/common/prom.o), arch/mips/alchemy/common/sleeper.o uses -mhard-float

To fix this, we detect if GAS is new enough to support "-msoft-float"

command

Quoted 1 line

Quoted 1 line

option, and if it does, we can let GCC pass it to GAS; but then we

also need

Quoted 1 line

Quoted 1 line

to sprinkle the files which make use of floating point registers with

the

Quoted 5 lines

Quoted 5 lines

necessary ".set hardfloat" directives.

Signed-off-by Manuel Lauss <[email protected]>

---
Compiles with binutils 2.23 and current git head, tested with alchemy

(mips32r1)

Quoted 55 lines

Quoted 6 lines

and maltasmvp_defconfig (64bit)

Tests with MSA and other extensions also appreciated!

v6: #undef fp so that the preprocessor does not replace the fp in

      .set fp=64 with $30...  Fixes 64bit build.

Technically speaking, a maltasmvp_defconfig selects CONFIG_32BIT=y so it's still a 32-bit build.

Quoted 1 line

[...]

Ok the fp problem went away but I still have the even/odd errors with my tools

arch/mips/kernel/r4k_switch.S: Assembler messages:
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 1
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 3
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 5
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 7
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 9
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 11
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 13
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 15
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 17
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 19

The following patch did not help either:

arch/mips/Makefile +1 -1

index 58076472bdd8..b8bb7e170fee 100644
@@ -56,7 +56,7 @@ ifdef CONFIG_FUNCTION_GRAPH_TRACER
   endif
 endif
 cflags-y += $(call cc-option, -mno-check-zero-division)
-
+cflags-y += -mno-odd-spreg
 

This is with a regular maltasmvp_defconfig

I guess my gcc version is newer than yours. Matthew?

--
markos

Quoted 4 lines

Tests with MSA and other extensions also appreciated!

v6: #undef fp so that the preprocessor does not replace the fp in

	.set fp=64 with $30...  Fixes 64bit build.

Technically speaking, a maltasmvp_defconfig selects CONFIG_32BIT=y so it's still a 32-bit build.

Quoted 1 line

[...]

Ok the fp problem went away but I still have the even/odd errors with my tools

arch/mips/kernel/r4k_switch.S: Assembler messages:
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 1
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 3
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 5
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 7
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 9
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 11

arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 13
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 15
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 17
arch/mips/kernel/r4k_switch.S:81: Error: float register should be even, was 19

+(mips1) `cfc1 $2,$31'
make[1]: *** [arch/mips/math-emu/cp1emu.o] Error 1
make: *** [arch/mips/math-emu] Error 2
make: *** Waiting for unfinished jobs....

Quoted 4 lines

+(mips1) `cfc1 $2,$31'
make[1]: *** [arch/mips/math-emu/cp1emu.o] Error 1
make: *** [arch/mips/math-emu] Error 2
make: *** Waiting for unfinished jobs....

This is the offending code in cp1emu.c:

                        if (is_fpu_owner())
                                asm volatile(
                                        ".set push\n"
                                        "\t.set mips1\n"
                                        "\tcfc1\t%0,$31\n"
                                        "\t.set pop" : "=r" (fcr31));
                        else
                                fcr31 = current->thread.fpu.fcr31;
                        preempt_enable();

I'm not sure how this can have built with binutils 2.23 (as indicated by Manuel and not built with 2.24). The reason this works with the latest version of binutils 2.24.x is that cfc1 has been reclassified as not an FPU instruction.

This just needs the hardfloat annotation adding via the macro as in the other cases.

Thanks,
Matthew

Quoted 4 lines

+(mips1) `cfc1 $2,$31'
make[1]: *** [arch/mips/math-emu/cp1emu.o] Error 1
make: *** [arch/mips/math-emu] Error 2
make: *** Waiting for unfinished jobs....

This is the offending code in cp1emu.c:

                        if (is_fpu_owner())
                                asm volatile(
                                        ".set push\n"
                                        "\t.set mips1\n"
                                        "\tcfc1\t%0,$31\n"
                                        "\t.set pop" : "=r" (fcr31));
                        else
                                fcr31 = current->thread.fpu.fcr31;
                        preempt_enable();

I'm not sure how this can have built with binutils 2.23 (as indicated by Manuel and not built with 2.24). The reason this works with the latest version of binutils 2.24.x is that cfc1 has been reclassified as not an FPU instruction.

This just needs the hardfloat annotation adding via the macro as in the other cases.