Compile issues with se7206_defconfig

Compile issues with se7206_defconfig

[Nick Krause]

There seems to me a build issue with this config. I will paste my logs
for this build below this message.
Cheers Nick
scripts/mod/empty.c:1:0: error: cannot use atomic model soft-imask in user mode
 /* empty file to figure out endianness / word size */
 ^
make[2]: *** [scripts/mod/empty.o] Error 1
make[2]: *** Waiting for unfinished jobs....
make[1]: *** [scripts/mod] Error 2
make[1]: *** Waiting for unfinished jobs....
  CHK     include/generated/utsrelease.h
  UPD     include/generated/utsrelease.h
  Generating include/generated/machtypes.h
  CC      kernel/bounds.s
kernel/bounds.c:1:0: error: cannot use atomic model soft-imask in user mode
 /*
 ^
make[1]: *** [kernel/bounds.s] Error 1
make: *** [prepare0] Error 2
make: *** Waiting for unfinished jobs....
make: *** [scripts] Error 2

Re: Compile issues with se7206_defconfig

[Guenter Roeck]

On 07/07/2014 07:56 PM, Nick Krause wrote:
> There seems to me a build issue with this config. I will paste my logs
> for this build below this message.
> Cheers Nick
> scripts/mod/empty.c:1:0: error: cannot use atomic model soft-imask in user mode
>   /* empty file to figure out endianness / word size */
>   ^
> make[2]: *** [scripts/mod/empty.o] Error 1
> make[2]: *** Waiting for unfinished jobs....
> make[1]: *** [scripts/mod] Error 2
> make[1]: *** Waiting for unfinished jobs....
>    CHK     include/generated/utsrelease.h
>    UPD     include/generated/utsrelease.h
>    Generating include/generated/machtypes.h
>    CC      kernel/bounds.s
> kernel/bounds.c:1:0: error: cannot use atomic model soft-imask in user mode
>   /*
>   ^
> make[1]: *** [kernel/bounds.s] Error 1
> make: *** [prepare0] Error 2
> make: *** Waiting for unfinished jobs....
> make: *** [scripts] Error 2

What kernel version ? I see a build failure in mainline, but it is different.

net/built-in.o: In function `__sk_run_filter':
filter.c:(.text+0x3156c): undefined reference to `__fpscr_values'
filter.c:(.text+0x31760): undefined reference to `__fpscr_values'
filter.c:(.text+0x31c8c): undefined reference to `__fpscr_values'
filter.c:(.text+0x31f00): undefined reference to `__fpscr_values'
filter.c:(.text+0x323b0): undefined reference to `__fpscr_values'

Guenter

Fwd: Compile issues with se7206_defconfig

[Nick Krause]

 Guenter Roeck work  on Tue, Jul 8, 2014 at 12:03 AM
Subject: Re: Compile issues with se7206_defconfig
To: Nick Krause <xerofoify@gmail.com>, "linux-kernel@vger.kernel.org"
<linux-kernel@vger.kernel.org>


On 07/07/2014 07:56 PM, Nick Krause wrote:
>
> There seems to me a build issue with this config. I will paste my logs
> for this build below this message.
> Cheers Nick
> scripts/mod/empty.c:1:0: error: cannot use atomic model soft-imask in user mode
>   /* empty file to figure out endianness / word size */
>   ^
> make[2]: *** [scripts/mod/empty.o] Error 1
> make[2]: *** Waiting for unfinished jobs....
> make[1]: *** [scripts/mod] Error 2
> make[1]: *** Waiting for unfinished jobs....
>    CHK     include/generated/utsrelease.h
>    UPD     include/generated/utsrelease.h
>    Generating include/generated/machtypes.h
>    CC      kernel/bounds.s
> kernel/bounds.c:1:0: error: cannot use atomic model soft-imask in user mode
>   /*
>   ^
> make[1]: *** [kernel/bounds.s] Error 1
> make: *** [prepare0] Error 2
> make: *** Waiting for unfinished jobs....
> make: *** [scripts] Error 2


What kernel version ? I see a build failure in mainline, but it is different.

net/built-in.o: In function `__sk_run_filter':
filter.c:(.text+0x3156c): undefined reference to `__fpscr_values'
filter.c:(.text+0x31760): undefined reference to `__fpscr_values'
filter.c:(.text+0x31c8c): undefined reference to `__fpscr_values'
filter.c:(.text+0x31f00): undefined reference to `__fpscr_values'
filter.c:(.text+0x323b0): undefined reference to `__fpscr_values'

Guenter

I am running mainline too. I will retest and see if I get the same
results as you.
Cheers Nick

Re: Compile issues with se7206_defconfig

[Geert Uytterhoeven]

On Tue, Jul 8, 2014 at 6:03 AM, Guenter Roeck <linux@roeck-us.net> wrote:
> On 07/07/2014 07:56 PM, Nick Krause wrote:
>> There seems to me a build issue with this config. I will paste my logs
>> for this build below this message.
>> Cheers Nick
>> scripts/mod/empty.c:1:0: error: cannot use atomic model soft-imask in user
>> mode
>>   /* empty file to figure out endianness / word size */
>>   ^
>> make[2]: *** [scripts/mod/empty.o] Error 1
>> make[2]: *** Waiting for unfinished jobs....
>> make[1]: *** [scripts/mod] Error 2
>> make[1]: *** Waiting for unfinished jobs....
>>    CHK     include/generated/utsrelease.h
>>    UPD     include/generated/utsrelease.h
>>    Generating include/generated/machtypes.h
>>    CC      kernel/bounds.s
>> kernel/bounds.c:1:0: error: cannot use atomic model soft-imask in user
>> mode
>>   /*
>>   ^
>> make[1]: *** [kernel/bounds.s] Error 1
>> make: *** [prepare0] Error 2
>> make: *** Waiting for unfinished jobs....
>> make: *** [scripts] Error 2
>
> What kernel version ? I see a build failure in mainline, but it is
> different.

Indeed.

> net/built-in.o: In function `__sk_run_filter':
> filter.c:(.text+0x3156c): undefined reference to `__fpscr_values'
> filter.c:(.text+0x31760): undefined reference to `__fpscr_values'
> filter.c:(.text+0x31c8c): undefined reference to `__fpscr_values'
> filter.c:(.text+0x31f00): undefined reference to `__fpscr_values'
> filter.c:(.text+0x323b0): undefined reference to `__fpscr_values'

Toolchain issue: the compiler doesn't recognize the option to disable
the FPU, so it isn't passed, cfr.
www.spinics.net/lists/netdev/msg278300.html

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: Compile issues with se7206_defconfig

[Guenter Roeck]

On Tue, Jul 08, 2014 at 10:08:46AM +0200, Geert Uytterhoeven wrote:
> On Tue, Jul 8, 2014 at 6:03 AM, Guenter Roeck <linux@roeck-us.net> wrote:
> > On 07/07/2014 07:56 PM, Nick Krause wrote:
> >> There seems to me a build issue with this config. I will paste my logs
> >> for this build below this message.
> >> Cheers Nick
> >> scripts/mod/empty.c:1:0: error: cannot use atomic model soft-imask in user
> >> mode
> >>   /* empty file to figure out endianness / word size */
> >>   ^
> >> make[2]: *** [scripts/mod/empty.o] Error 1
> >> make[2]: *** Waiting for unfinished jobs....
> >> make[1]: *** [scripts/mod] Error 2
> >> make[1]: *** Waiting for unfinished jobs....
> >>    CHK     include/generated/utsrelease.h
> >>    UPD     include/generated/utsrelease.h
> >>    Generating include/generated/machtypes.h
> >>    CC      kernel/bounds.s
> >> kernel/bounds.c:1:0: error: cannot use atomic model soft-imask in user
> >> mode
> >>   /*
> >>   ^
> >> make[1]: *** [kernel/bounds.s] Error 1
> >> make: *** [prepare0] Error 2
> >> make: *** Waiting for unfinished jobs....
> >> make: *** [scripts] Error 2
> >
> > What kernel version ? I see a build failure in mainline, but it is
> > different.
> 
> Indeed.
> 
> > net/built-in.o: In function `__sk_run_filter':
> > filter.c:(.text+0x3156c): undefined reference to `__fpscr_values'
> > filter.c:(.text+0x31760): undefined reference to `__fpscr_values'
> > filter.c:(.text+0x31c8c): undefined reference to `__fpscr_values'
> > filter.c:(.text+0x31f00): undefined reference to `__fpscr_values'
> > filter.c:(.text+0x323b0): undefined reference to `__fpscr_values'
> 
> Toolchain issue: the compiler doesn't recognize the option to disable
> the FPU, so it isn't passed, cfr.
> www.spinics.net/lists/netdev/msg278300.html
> 

Seems that similar problems were fixed earlier without requiring a toolchain
change. What is different this time (besides the apparent difficulty to build a
new toolchain) ? Or am I missing something ?

Thanks,
Guenter

Re: Compile issues with se7206_defconfig

[Geert Uytterhoeven]

On Tue, Jul 8, 2014 at 6:45 PM, Guenter Roeck <linux@roeck-us.net> wrote:
>> > net/built-in.o: In function `__sk_run_filter':
>> > filter.c:(.text+0x3156c): undefined reference to `__fpscr_values'
>> > filter.c:(.text+0x31760): undefined reference to `__fpscr_values'
>> > filter.c:(.text+0x31c8c): undefined reference to `__fpscr_values'
>> > filter.c:(.text+0x31f00): undefined reference to `__fpscr_values'
>> > filter.c:(.text+0x323b0): undefined reference to `__fpscr_values'
>>
>> Toolchain issue: the compiler doesn't recognize the option to disable
>> the FPU, so it isn't passed, cfr.
>> www.spinics.net/lists/netdev/msg278300.html
>
> Seems that similar problems were fixed earlier without requiring a toolchain
> change. What is different this time (besides the apparent difficulty to build a
> new toolchain) ? Or am I missing something ?

sh4-linux-gcc doesn't recognize -m2a-nofpu, so that is not passed.
It does recognize -m4a-nofpu, but when that's passed, it complains
about opcode not valid for this cpu variant in inline asm/

Ah, it also recognizes -m4-nofpu. If I add that, se7206_defconfig builds.
I assume a proper sh2a-linux toolchain will not recognize the -m4-nofpu,
so it won't be passed due to $(call cc-option).

Does this sound like an acceptable solution? If yes, I'll send a patch.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: Compile issues with se7206_defconfig

[Guenter Roeck]

On Tue, Jul 08, 2014 at 09:24:49PM +0200, Geert Uytterhoeven wrote:
> On Tue, Jul 8, 2014 at 6:45 PM, Guenter Roeck <linux@roeck-us.net> wrote:
> >> > net/built-in.o: In function `__sk_run_filter':
> >> > filter.c:(.text+0x3156c): undefined reference to `__fpscr_values'
> >> > filter.c:(.text+0x31760): undefined reference to `__fpscr_values'
> >> > filter.c:(.text+0x31c8c): undefined reference to `__fpscr_values'
> >> > filter.c:(.text+0x31f00): undefined reference to `__fpscr_values'
> >> > filter.c:(.text+0x323b0): undefined reference to `__fpscr_values'
> >>
> >> Toolchain issue: the compiler doesn't recognize the option to disable
> >> the FPU, so it isn't passed, cfr.
> >> www.spinics.net/lists/netdev/msg278300.html
> >
> > Seems that similar problems were fixed earlier without requiring a toolchain
> > change. What is different this time (besides the apparent difficulty to build a
> > new toolchain) ? Or am I missing something ?
> 
> sh4-linux-gcc doesn't recognize -m2a-nofpu, so that is not passed.
> It does recognize -m4a-nofpu, but when that's passed, it complains
> about opcode not valid for this cpu variant in inline asm/
> 
> Ah, it also recognizes -m4-nofpu. If I add that, se7206_defconfig builds.
> I assume a proper sh2a-linux toolchain will not recognize the -m4-nofpu,
> so it won't be passed due to $(call cc-option).
> 
> Does this sound like an acceptable solution? If yes, I'll send a patch.
> 
Sounds good to me.

Guenter

Fwd: Compile issues with se7206_defconfig

[Nick Krause]

 Guenter Roeck  wrote on Tue, Jul 8, 2014 at 3:30 PM
Subject: Re: Compile issues with se7206_defconfig
To: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Nick Krause <xerofoify@gmail.com>, "linux-kernel@vger.kernel.org"
<linux-kernel@vger.kernel.org>


On Tue, Jul 08, 2014 at 09:24:49PM +0200, Geert Uytterhoeven wrote:
> On Tue, Jul 8, 2014 at 6:45 PM, Guenter Roeck <linux@roeck-us.net> wrote:
> >> > net/built-in.o: In function `__sk_run_filter':
> >> > filter.c:(.text+0x3156c): undefined reference to `__fpscr_values'
> >> > filter.c:(.text+0x31760): undefined reference to `__fpscr_values'
> >> > filter.c:(.text+0x31c8c): undefined reference to `__fpscr_values'
> >> > filter.c:(.text+0x31f00): undefined reference to `__fpscr_values'
> >> > filter.c:(.text+0x323b0): undefined reference to `__fpscr_values'
> >>
> >> Toolchain issue: the compiler doesn't recognize the option to disable
> >> the FPU, so it isn't passed, cfr.
> >> www.spinics.net/lists/netdev/msg278300.html
> >
> > Seems that similar problems were fixed earlier without requiring a toolchain
> > change. What is different this time (besides the apparent difficulty to build a
> > new toolchain) ? Or am I missing something ?
>
> sh4-linux-gcc doesn't recognize -m2a-nofpu, so that is not passed.
> It does recognize -m4a-nofpu, but when that's passed, it complains
> about opcode not valid for this cpu variant in inline asm/
>
> Ah, it also recognizes -m4-nofpu. If I add that, se7206_defconfig builds.
> I assume a proper sh2a-linux toolchain will not recognize the -m4-nofpu,
> so it won't be passed due to $(call cc-option).
>
> Does this sound like an acceptable solution? If yes, I'll send a patch.
>
Sounds good to me.

Guenter

Please send the patch to the maintainers of this and cc me so I can test it to
see if it works.
Cheers Nick