Re: Build regressions/improvements in v4.20-rc1

Re: Build regressions/improvements in v4.20-rc1

[Geert Uytterhoeven]

On Mon, Nov 5, 2018 at 11:07 PM Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> Below is the list of build error/warning regressions/improvements in
> v4.20-rc1[1] compared to v4.19[2].
>
> Summarized:
>   - build errors: +3/-0
>   - build warnings: +449/-2712
>
> Happy fixing! ;-)
>
> Thanks to the linux-next team for providing the build service.
>
> [1] http://kisskb.ellerman.id.au/kisskb/branch/linus/head/651022382c7f8da46cb4872a545ee1da6d097d2a/ (all 240 configs)
> [2] http://kisskb.ellerman.id.au/kisskb/branch/linus/head/84df9525b0c27f3ebc2ebb1864fa62a97fdedb7d/ (all 240 configs)
>
>
> *** ERRORS ***
>
> 3 error regressions:
>   + /kisskb/src/arch/s390/kernel/nospec-branch.c: error: macro "memcpy" passed 6 arguments, but takes just 3:  => 130:61

s390-all{mod,yes}config

>   + /kisskb/src/sound/pci/hda/patch_ca0132.c: error: implicit declaration of function 'pci_iomap' [-Werror=implicit-function-declaration]:  => 8799:3

sh4-all{mod,yes}config

Looks like d9b84a15892c0233 ("ALSA: hda: Fix implicit definition of
pci_iomap() on SH")
is not sufficient?

>   + {standard input}: Error: Unrecognized opcode: `slbfee.':  => 89, 37

powerpc-all{mod,yes}config

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

Build regressions/improvements in v4.20-rc1 (sound/pci/hda/patch_ca0132.c)

[Randy Dunlap]

On 11/5/18 2:12 PM, Geert Uytterhoeven wrote:
> On Mon, Nov 5, 2018 at 11:07 PM Geert Uytterhoeven <geert@linux-m68k.org> wrote:
>> Below is the list of build error/warning regressions/improvements in
>> v4.20-rc1[1] compared to v4.19[2].
>>
>> Summarized:
>>   - build errors: +3/-0
>>   - build warnings: +449/-2712
>>
>> Happy fixing! ;-)
>>
>> Thanks to the linux-next team for providing the build service.
>>
>> [1] http://kisskb.ellerman.id.au/kisskb/branch/linus/head/651022382c7f8da46cb4872a545ee1da6d097d2a/ (all 240 configs)
>> [2] http://kisskb.ellerman.id.au/kisskb/branch/linus/head/84df9525b0c27f3ebc2ebb1864fa62a97fdedb7d/ (all 240 configs)
>>
>>
>> *** ERRORS ***
>>
>>   + /kisskb/src/sound/pci/hda/patch_ca0132.c: error: implicit declaration of function 'pci_iomap' [-Werror=implicit-function-declaration]:  => 8799:3
> 
> sh4-all{mod,yes}config
> 
> Looks like d9b84a15892c0233 ("ALSA: hda: Fix implicit definition of
> pci_iomap() on SH")
> is not sufficient?

Different problem.  This is about "select":

config SND_SOC_ALL_CODECS
	tristate "Build all ASoC CODEC drivers"

That enables (sets):
	select SND_SOC_HDAC_HDA
which selects SND_HDA even though CONFIG_PCI is not enabled.

After SND_HDA is selected (above), the Kconfig symbols in
sound/pci/hda/Kconfig are available for enabling, so
SND_HDA_CODEC_CA0132 is enabled but will not build.


One simple solution (but possibly too naive) is:

---
 sound/soc/codecs/Kconfig |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

--- lnx-420-rc1.orig/sound/soc/codecs/Kconfig
+++ lnx-420-rc1/sound/soc/codecs/Kconfig
@@ -82,7 +82,7 @@ config SND_SOC_ALL_CODECS
 	select SND_SOC_ES7241
 	select SND_SOC_GTM601
 	select SND_SOC_HDAC_HDMI
-	select SND_SOC_HDAC_HDA
+	select SND_SOC_HDAC_HDA if PCI
 	select SND_SOC_ICS43432
 	select SND_SOC_INNO_RK3036
 	select SND_SOC_ISABELLE if I2C




-- 
~Randy

Re: Build regressions/improvements in v4.20-rc1 (sound/pci/hda/patch_ca0132.c)

[Geert Uytterhoeven]

Hi Randy,

On Tue, Nov 6, 2018 at 2:06 AM Randy Dunlap <rdunlap@infradead.org> wrote:
> On 11/5/18 2:12 PM, Geert Uytterhoeven wrote:
> > On Mon, Nov 5, 2018 at 11:07 PM Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> >> Below is the list of build error/warning regressions/improvements in
> >> v4.20-rc1[1] compared to v4.19[2].
> >>
> >> Summarized:
> >>   - build errors: +3/-0
> >>   - build warnings: +449/-2712
> >>
> >> Happy fixing! ;-)
> >>
> >> Thanks to the linux-next team for providing the build service.
> >>
> >> [1] http://kisskb.ellerman.id.au/kisskb/branch/linus/head/651022382c7f8da46cb4872a545ee1da6d097d2a/ (all 240 configs)
> >> [2] http://kisskb.ellerman.id.au/kisskb/branch/linus/head/84df9525b0c27f3ebc2ebb1864fa62a97fdedb7d/ (all 240 configs)
> >>
> >>
> >> *** ERRORS ***
> >>
> >>   + /kisskb/src/sound/pci/hda/patch_ca0132.c: error: implicit declaration of function 'pci_iomap' [-Werror=implicit-function-declaration]:  => 8799:3
> >
> > sh4-all{mod,yes}config
> >
> > Looks like d9b84a15892c0233 ("ALSA: hda: Fix implicit definition of
> > pci_iomap() on SH")
> > is not sufficient?
>
> Different problem.  This is about "select":
>
> config SND_SOC_ALL_CODECS
>         tristate "Build all ASoC CODEC drivers"
>
> That enables (sets):
>         select SND_SOC_HDAC_HDA
> which selects SND_HDA even though CONFIG_PCI is not enabled.
>
> After SND_HDA is selected (above), the Kconfig symbols in
> sound/pci/hda/Kconfig are available for enabling, so
> SND_HDA_CODEC_CA0132 is enabled but will not build.

Thanks for looking into this!

> One simple solution (but possibly too naive) is:
>
> ---
>  sound/soc/codecs/Kconfig |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> --- lnx-420-rc1.orig/sound/soc/codecs/Kconfig
> +++ lnx-420-rc1/sound/soc/codecs/Kconfig
> @@ -82,7 +82,7 @@ config SND_SOC_ALL_CODECS
>         select SND_SOC_ES7241
>         select SND_SOC_GTM601
>         select SND_SOC_HDAC_HDMI
> -       select SND_SOC_HDAC_HDA
> +       select SND_SOC_HDAC_HDA if PCI
>         select SND_SOC_ICS43432
>         select SND_SOC_INNO_RK3036
>         select SND_SOC_ISABELLE if I2C

I guess that will work. There are already plenty of "select foo if bar" lines.
However, looking at what else can enable SND_HDA, I think it should be

    select SND_SOC_HDAC_HDA if SND_PCI || ARCH_TEGRA

That still leaves the issue that pci_iomap() on SH should be an empty stub if
PCI is not available, like on other architectures.

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: [alsa-devel] Build regressions/improvements in v4.20-rc1 (sound/pci/hda/patch_ca0132.c)

[Pierre-Louis Bossart]

>>>>
>>>> *** ERRORS ***
>>>>
>>>>    + /kisskb/src/sound/pci/hda/patch_ca0132.c: error: implicit declaration of function 'pci_iomap' [-Werror=implicit-function-declaration]:  => 8799:3
>>> sh4-all{mod,yes}config
>>>
>>> Looks like d9b84a15892c0233 ("ALSA: hda: Fix implicit definition of
>>> pci_iomap() on SH")
>>> is not sufficient?
>> Different problem.  This is about "select":
>>
>> config SND_SOC_ALL_CODECS
>>          tristate "Build all ASoC CODEC drivers"
>>
>> That enables (sets):
>>          select SND_SOC_HDAC_HDA
>> which selects SND_HDA even though CONFIG_PCI is not enabled.
>>
>> After SND_HDA is selected (above), the Kconfig symbols in
>> sound/pci/hda/Kconfig are available for enabling, so
>> SND_HDA_CODEC_CA0132 is enabled but will not build.
> Thanks for looking into this!
>
>> One simple solution (but possibly too naive) is:
>>
>> ---
>>   sound/soc/codecs/Kconfig |    2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> --- lnx-420-rc1.orig/sound/soc/codecs/Kconfig
>> +++ lnx-420-rc1/sound/soc/codecs/Kconfig
>> @@ -82,7 +82,7 @@ config SND_SOC_ALL_CODECS
>>          select SND_SOC_ES7241
>>          select SND_SOC_GTM601
>>          select SND_SOC_HDAC_HDMI
>> -       select SND_SOC_HDAC_HDA
>> +       select SND_SOC_HDAC_HDA if PCI
>>          select SND_SOC_ICS43432
>>          select SND_SOC_INNO_RK3036
>>          select SND_SOC_ISABELLE if I2C
> I guess that will work. There are already plenty of "select foo if bar" lines.
> However, looking at what else can enable SND_HDA, I think it should be
>
>      select SND_SOC_HDAC_HDA if SND_PCI || ARCH_TEGRA

This codec can only be used by the Skylake driver (and the upcoming SOF 
one). For Tegra this module will never be used unless they follow the 
same path of enabling ASoC to deal with the HDaudio codecs instead of 
the legacy.

Likewise HDAC_HDMI will only work on Intel platforms for now.

>
> That still leaves the issue that pci_iomap() on SH should be an empty stub if
> PCI is not available, like on other architectures.

I thought Mark Brown provided a fix to SH maintainers?

Re: [alsa-devel] Build regressions/improvements in v4.20-rc1 (sound/pci/hda/patch_ca0132.c)

[Geert Uytterhoeven]

Hi Pierre,

On Tue, Nov 6, 2018 at 3:56 PM Pierre-Louis Bossart
<pierre-louis.bossart@linux.intel.com> wrote:
> >>>>
> >>>> *** ERRORS ***
> >>>>
> >>>>    + /kisskb/src/sound/pci/hda/patch_ca0132.c: error: implicit declaration of function 'pci_iomap' [-Werror=implicit-function-declaration]:  => 8799:3
> >>> sh4-all{mod,yes}config
> >>>
> >>> Looks like d9b84a15892c0233 ("ALSA: hda: Fix implicit definition of
> >>> pci_iomap() on SH")
> >>> is not sufficient?
> >> Different problem.  This is about "select":
> >>
> >> config SND_SOC_ALL_CODECS
> >>          tristate "Build all ASoC CODEC drivers"
> >>
> >> That enables (sets):
> >>          select SND_SOC_HDAC_HDA
> >> which selects SND_HDA even though CONFIG_PCI is not enabled.
> >>
> >> After SND_HDA is selected (above), the Kconfig symbols in
> >> sound/pci/hda/Kconfig are available for enabling, so
> >> SND_HDA_CODEC_CA0132 is enabled but will not build.
> > Thanks for looking into this!
> >
> >> One simple solution (but possibly too naive) is:
> >>
> >> ---
> >>   sound/soc/codecs/Kconfig |    2 +-
> >>   1 file changed, 1 insertion(+), 1 deletion(-)
> >>
> >> --- lnx-420-rc1.orig/sound/soc/codecs/Kconfig
> >> +++ lnx-420-rc1/sound/soc/codecs/Kconfig
> >> @@ -82,7 +82,7 @@ config SND_SOC_ALL_CODECS
> >>          select SND_SOC_ES7241
> >>          select SND_SOC_GTM601
> >>          select SND_SOC_HDAC_HDMI
> >> -       select SND_SOC_HDAC_HDA
> >> +       select SND_SOC_HDAC_HDA if PCI
> >>          select SND_SOC_ICS43432
> >>          select SND_SOC_INNO_RK3036
> >>          select SND_SOC_ISABELLE if I2C
> > I guess that will work. There are already plenty of "select foo if bar" lines.
> > However, looking at what else can enable SND_HDA, I think it should be
> >
> >      select SND_SOC_HDAC_HDA if SND_PCI || ARCH_TEGRA
>
> This codec can only be used by the Skylake driver (and the upcoming SOF
> one). For Tegra this module will never be used unless they follow the
> same path of enabling ASoC to deal with the HDaudio codecs instead of
> the legacy.
>
> Likewise HDAC_HDMI will only work on Intel platforms for now.
>
> >
> > That still leaves the issue that pci_iomap() on SH should be an empty stub if
> > PCI is not available, like on other architectures.
>
> I thought Mark Brown provided a fix to SH maintainers?

Indeed, https://patchwork.kernel.org/patch/10597409/

As usual, no response from the J^HSH maintainers :-(

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: [alsa-devel] Build regressions/improvements in v4.20-rc1 (sound/pci/hda/patch_ca0132.c)

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 744 bytes --]

On Tue, Nov 06, 2018 at 08:56:39AM -0600, Pierre-Louis Bossart wrote:

> > However, looking at what else can enable SND_HDA, I think it should be

> >      select SND_SOC_HDAC_HDA if SND_PCI || ARCH_TEGRA

> This codec can only be used by the Skylake driver (and the upcoming SOF
> one). For Tegra this module will never be used unless they follow the same
> path of enabling ASoC to deal with the HDaudio codecs instead of the legacy.

> Likewise HDAC_HDMI will only work on Intel platforms for now.

_ALL_CODECS is a build coverage option, not a useful configuration
option.  We should be building things in as many different
configurations as possible so that it's easier for people to get build
coverage when making subsystem wide changes.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [alsa-devel] Build regressions/improvements in v4.20-rc1 (sound/pci/hda/patch_ca0132.c)

[Takashi Iwai]

On Tue, 06 Nov 2018 02:04:47 +0100,
Randy Dunlap wrote:
> 
> On 11/5/18 2:12 PM, Geert Uytterhoeven wrote:
> > On Mon, Nov 5, 2018 at 11:07 PM Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> >> Below is the list of build error/warning regressions/improvements in
> >> v4.20-rc1[1] compared to v4.19[2].
> >>
> >> Summarized:
> >>   - build errors: +3/-0
> >>   - build warnings: +449/-2712
> >>
> >> Happy fixing! ;-)
> >>
> >> Thanks to the linux-next team for providing the build service.
> >>
> >> [1] http://kisskb.ellerman.id.au/kisskb/branch/linus/head/651022382c7f8da46cb4872a545ee1da6d097d2a/ (all 240 configs)
> >> [2] http://kisskb.ellerman.id.au/kisskb/branch/linus/head/84df9525b0c27f3ebc2ebb1864fa62a97fdedb7d/ (all 240 configs)
> >>
> >>
> >> *** ERRORS ***
> >>
> >>   + /kisskb/src/sound/pci/hda/patch_ca0132.c: error: implicit declaration of function 'pci_iomap' [-Werror=implicit-function-declaration]:  => 8799:3
> > 
> > sh4-all{mod,yes}config
> > 
> > Looks like d9b84a15892c0233 ("ALSA: hda: Fix implicit definition of
> > pci_iomap() on SH")
> > is not sufficient?
> 
> Different problem.  This is about "select":
> 
> config SND_SOC_ALL_CODECS
> 	tristate "Build all ASoC CODEC drivers"
> 
> That enables (sets):
> 	select SND_SOC_HDAC_HDA
> which selects SND_HDA even though CONFIG_PCI is not enabled.

Actually it is OK to enable CONFIG_SND_HDA_CODEC_CA0132 without
CONFIG_PCI.  IIRC, there was a system like that, too.
The commit above should have covered the build failure on SH, but
apparently isn't enough for some arch setups, as it seems.

The cause is clear now: pci_iomap() is defined in
asm-generic/pci_iomap.h only when CONFIG_GENERIC_PCI_IOMAP is
defined.  Including asm/io.h doesn't help unless CONFIG_PCI is set.

Below is a quick fix for this.


thanks,

Takashi

-- 8< --

From: Takashi Iwai <tiwai@suse.de>
Subject: [PATCH] ALSA: hda/ca0132 - Yet more fix on build breakage without PCI
 support

The recent change in CA0132 codec driver for supporting more
Creative boards includes the pci_iomap() call to get the extra
register accesses.  This is supposed to work on all archs and setups,
by the implicit assumption that every arch would provide a dummy
function returning NULL when no PCI is available.  But the reality
bites, of course; as Geert's regular build test shows, some configs
(at least SH4 without CONFIG_PCI) leads to a build error due to the
implicit function declaration.

So this is another attempt to fix the issue: now we add an ifdef
CONFIG_PCI line, so that pci_iomap() won't be called unless PCI is
really usable.  This should fall back to the standard quirk again with
a warning.

Fixes: d9b84a15892c0233 ("ALSA: hda: Fix implicit definition of pci_iomap() on SH")
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
---
 sound/pci/hda/patch_ca0132.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/sound/pci/hda/patch_ca0132.c b/sound/pci/hda/patch_ca0132.c
index 0a24037184c3..9ed808b45e75 100644
--- a/sound/pci/hda/patch_ca0132.c
+++ b/sound/pci/hda/patch_ca0132.c
@@ -8796,7 +8796,13 @@ static int patch_ca0132(struct hda_codec *codec)
 	}
 
 	if (spec->use_pci_mmio) {
+		/*
+		 * ifdef below needed due to lack of pci_iomap() decleration
+		 * for some archs when no PCI is defined
+		 */
+#ifdef CONFIG_PCI
 		spec->mem_base = pci_iomap(codec->bus->pci, 2, 0xC20);
+#endif
 		if (spec->mem_base == NULL) {
 			codec_warn(codec, "pci_iomap failed! Setting quirk to QUIRK_NONE.");
 			spec->quirk = QUIRK_NONE;
-- 
2.19.1

Re: [alsa-devel] Build regressions/improvements in v4.20-rc1 (sound/pci/hda/patch_ca0132.c)

[Geert Uytterhoeven]

Hi Iwai-san,

On Tue, Nov 6, 2018 at 5:18 PM Takashi Iwai <tiwai@suse.de> wrote:
> On Tue, 06 Nov 2018 02:04:47 +0100,
> Randy Dunlap wrote:
> >
> > On 11/5/18 2:12 PM, Geert Uytterhoeven wrote:
> > > On Mon, Nov 5, 2018 at 11:07 PM Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> > >> Below is the list of build error/warning regressions/improvements in
> > >> v4.20-rc1[1] compared to v4.19[2].
> > >>
> > >> Summarized:
> > >>   - build errors: +3/-0
> > >>   - build warnings: +449/-2712
> > >>
> > >> Happy fixing! ;-)
> > >>
> > >> Thanks to the linux-next team for providing the build service.
> > >>
> > >> [1] http://kisskb.ellerman.id.au/kisskb/branch/linus/head/651022382c7f8da46cb4872a545ee1da6d097d2a/ (all 240 configs)
> > >> [2] http://kisskb.ellerman.id.au/kisskb/branch/linus/head/84df9525b0c27f3ebc2ebb1864fa62a97fdedb7d/ (all 240 configs)
> > >>
> > >>
> > >> *** ERRORS ***
> > >>
> > >>   + /kisskb/src/sound/pci/hda/patch_ca0132.c: error: implicit declaration of function 'pci_iomap' [-Werror=implicit-function-declaration]:  => 8799:3
> > >
> > > sh4-all{mod,yes}config
> > >
> > > Looks like d9b84a15892c0233 ("ALSA: hda: Fix implicit definition of
> > > pci_iomap() on SH")
> > > is not sufficient?
> >
> > Different problem.  This is about "select":
> >
> > config SND_SOC_ALL_CODECS
> >       tristate "Build all ASoC CODEC drivers"
> >
> > That enables (sets):
> >       select SND_SOC_HDAC_HDA
> > which selects SND_HDA even though CONFIG_PCI is not enabled.
>
> Actually it is OK to enable CONFIG_SND_HDA_CODEC_CA0132 without
> CONFIG_PCI.  IIRC, there was a system like that, too.
> The commit above should have covered the build failure on SH, but
> apparently isn't enough for some arch setups, as it seems.
>
> The cause is clear now: pci_iomap() is defined in
> asm-generic/pci_iomap.h only when CONFIG_GENERIC_PCI_IOMAP is
> defined.  Including asm/io.h doesn't help unless CONFIG_PCI is set.
>
> Below is a quick fix for this.
>
>
> thanks,
>
> Takashi
>
> -- 8< --
>
> From: Takashi Iwai <tiwai@suse.de>
> Subject: [PATCH] ALSA: hda/ca0132 - Yet more fix on build breakage without PCI
>  support
>
> The recent change in CA0132 codec driver for supporting more
> Creative boards includes the pci_iomap() call to get the extra
> register accesses.  This is supposed to work on all archs and setups,
> by the implicit assumption that every arch would provide a dummy
> function returning NULL when no PCI is available.  But the reality
> bites, of course; as Geert's regular build test shows, some configs
> (at least SH4 without CONFIG_PCI) leads to a build error due to the
> implicit function declaration.
>
> So this is another attempt to fix the issue: now we add an ifdef
> CONFIG_PCI line, so that pci_iomap() won't be called unless PCI is
> really usable.  This should fall back to the standard quirk again with
> a warning.
>
> Fixes: d9b84a15892c0233 ("ALSA: hda: Fix implicit definition of pci_iomap() on SH")
> Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
> Cc: <stable@vger.kernel.org>
> Signed-off-by: Takashi Iwai <tiwai@suse.de>

Thanks for your patch!

> --- a/sound/pci/hda/patch_ca0132.c
> +++ b/sound/pci/hda/patch_ca0132.c
> @@ -8796,7 +8796,13 @@ static int patch_ca0132(struct hda_codec *codec)
>         }
>
>         if (spec->use_pci_mmio) {
> +               /*
> +                * ifdef below needed due to lack of pci_iomap() decleration
> +                * for some archs when no PCI is defined
> +                */
> +#ifdef CONFIG_PCI
>                 spec->mem_base = pci_iomap(codec->bus->pci, 2, 0xC20);
> +#endif

I'm sorry, but that is not a proper fix.
This should be fixed in the SH-specific code, to behave like other
architectures.

>                 if (spec->mem_base == NULL) {
>                         codec_warn(codec, "pci_iomap failed! Setting quirk to QUIRK_NONE.");
>                         spec->quirk = QUIRK_NONE;

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: [alsa-devel] Build regressions/improvements in v4.20-rc1 (sound/pci/hda/patch_ca0132.c)

[Takashi Iwai]

On Wed, 07 Nov 2018 09:44:25 +0100,
Geert Uytterhoeven wrote:
> 
> Hi Iwai-san,
> 
> On Tue, Nov 6, 2018 at 5:18 PM Takashi Iwai <tiwai@suse.de> wrote:
> > On Tue, 06 Nov 2018 02:04:47 +0100,
> > Randy Dunlap wrote:
> > >
> > > On 11/5/18 2:12 PM, Geert Uytterhoeven wrote:
> > > > On Mon, Nov 5, 2018 at 11:07 PM Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> > > >> Below is the list of build error/warning regressions/improvements in
> > > >> v4.20-rc1[1] compared to v4.19[2].
> > > >>
> > > >> Summarized:
> > > >>   - build errors: +3/-0
> > > >>   - build warnings: +449/-2712
> > > >>
> > > >> Happy fixing! ;-)
> > > >>
> > > >> Thanks to the linux-next team for providing the build service.
> > > >>
> > > >> [1] http://kisskb.ellerman.id.au/kisskb/branch/linus/head/651022382c7f8da46cb4872a545ee1da6d097d2a/ (all 240 configs)
> > > >> [2] http://kisskb.ellerman.id.au/kisskb/branch/linus/head/84df9525b0c27f3ebc2ebb1864fa62a97fdedb7d/ (all 240 configs)
> > > >>
> > > >>
> > > >> *** ERRORS ***
> > > >>
> > > >>   + /kisskb/src/sound/pci/hda/patch_ca0132.c: error: implicit declaration of function 'pci_iomap' [-Werror=implicit-function-declaration]:  => 8799:3
> > > >
> > > > sh4-all{mod,yes}config
> > > >
> > > > Looks like d9b84a15892c0233 ("ALSA: hda: Fix implicit definition of
> > > > pci_iomap() on SH")
> > > > is not sufficient?
> > >
> > > Different problem.  This is about "select":
> > >
> > > config SND_SOC_ALL_CODECS
> > >       tristate "Build all ASoC CODEC drivers"
> > >
> > > That enables (sets):
> > >       select SND_SOC_HDAC_HDA
> > > which selects SND_HDA even though CONFIG_PCI is not enabled.
> >
> > Actually it is OK to enable CONFIG_SND_HDA_CODEC_CA0132 without
> > CONFIG_PCI.  IIRC, there was a system like that, too.
> > The commit above should have covered the build failure on SH, but
> > apparently isn't enough for some arch setups, as it seems.
> >
> > The cause is clear now: pci_iomap() is defined in
> > asm-generic/pci_iomap.h only when CONFIG_GENERIC_PCI_IOMAP is
> > defined.  Including asm/io.h doesn't help unless CONFIG_PCI is set.
> >
> > Below is a quick fix for this.
> >
> >
> > thanks,
> >
> > Takashi
> >
> > -- 8< --
> >
> > From: Takashi Iwai <tiwai@suse.de>
> > Subject: [PATCH] ALSA: hda/ca0132 - Yet more fix on build breakage without PCI
> >  support
> >
> > The recent change in CA0132 codec driver for supporting more
> > Creative boards includes the pci_iomap() call to get the extra
> > register accesses.  This is supposed to work on all archs and setups,
> > by the implicit assumption that every arch would provide a dummy
> > function returning NULL when no PCI is available.  But the reality
> > bites, of course; as Geert's regular build test shows, some configs
> > (at least SH4 without CONFIG_PCI) leads to a build error due to the
> > implicit function declaration.
> >
> > So this is another attempt to fix the issue: now we add an ifdef
> > CONFIG_PCI line, so that pci_iomap() won't be called unless PCI is
> > really usable.  This should fall back to the standard quirk again with
> > a warning.
> >
> > Fixes: d9b84a15892c0233 ("ALSA: hda: Fix implicit definition of pci_iomap() on SH")
> > Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
> > Cc: <stable@vger.kernel.org>
> > Signed-off-by: Takashi Iwai <tiwai@suse.de>
> 
> Thanks for your patch!
> 
> > --- a/sound/pci/hda/patch_ca0132.c
> > +++ b/sound/pci/hda/patch_ca0132.c
> > @@ -8796,7 +8796,13 @@ static int patch_ca0132(struct hda_codec *codec)
> >         }
> >
> >         if (spec->use_pci_mmio) {
> > +               /*
> > +                * ifdef below needed due to lack of pci_iomap() decleration
> > +                * for some archs when no PCI is defined
> > +                */
> > +#ifdef CONFIG_PCI
> >                 spec->mem_base = pci_iomap(codec->bus->pci, 2, 0xC20);
> > +#endif
> 
> I'm sorry, but that is not a proper fix.
> This should be fixed in the SH-specific code, to behave like other
> architectures.

OK, this is the thing I didn't want to go deeply, as the problem is
messy...

The top of iceberg is that asm/pci_iomap.h defines the dummy function
of pci_iomap() only when CONFIG_PCI=n && CONFIG_GENERIC_PCI_IOMAP=y.
This has been so for long time since the commit 97a29d59fc22
    [PARISC] fix compile break caused by iomap: make IOPORT/PCI
    mapping function s conditional

Before that point, the header defined pci_iomap() and pci_ioumap()
whenever CONFIG_PCi=n.

And, looking at other arch codes, most of them define
CONFIG_GENERIC_PCI_IOMAP no matter whether CONFIG_PCI is set or not.
Actually this is supposedly OK, as the generic code (lib/pci_iomap.c)
contains a large ifdef CONFIG_PCI in it, so essentially it's empty
when CONFIG_PCI=n.

So, one possible fix would be to convert all fallout archs (alpha,
powerpc and sh) to define always CONFIG_GENERIC_PCI_IOMAP.  I'm not
100% sure whether this is safe, but we may try.

Another fix would be to revert the commit 97a29d59fc22, and let
asm/pci_iomap.h providing the dummy functions whenever CONFIG_PCi=n.
Along with it, we need to fix the arch side, the duplicated (and
useless) definitions in arch/parisc/lib/iomap.c.  Through a quick
look, arch/sparc also has the pci_iounmap() definition without
CONFIG_PCI check, so this needs to be fixed as well.

And, what makes the situation messy is that the handling of
pci_iounmap() is totally different from pci_iomap().  pci_iomap() is
covered by CONFIG_GENERIC_PCI_IOMAP while pci_iounmap() is by
CONFIG_GENERIC_IOMAP.  And, yet there are multiple dummy function
definitions of pci_iounmap().  Once in include/asm-generic/iomap.h,
and once in include/asm-generic/io.h.

Overall, there are a few rooms of cleanups / fixes, yeah.


thanks,

Takashi