Linux kernel -stable discussions
 help / color / mirror / Atom feed
From: Nathan Chancellor <nathan@kernel.org>
To: Huacai Chen <chenhuacai@kernel.org>
Cc: Huacai Chen <chenhuacai@loongson.cn>,
	Thomas Bogendoerfer <tsbogend@alpha.franken.de>,
	linux-mips@vger.kernel.org, Jiaxun Yang <jiaxun.yang@flygoat.com>,
	stable@vger.kernel.org,
	Nick Desaulniers <ndesaulniers@google.com>
Subject: Re: [PATCH] MIPS: Loongson: Fix build error when make modules_install
Date: Tue, 27 Jun 2023 14:28:39 -0700	[thread overview]
Message-ID: <20230627212839.GA1806408@dev-arch.thelio-3990X> (raw)
In-Reply-To: <CAAhV-H6nyXa+wG-J50d=FrHX=4saVNAePW8HHQ2hm+EsGR9Umw@mail.gmail.com>

Hi Huacai,

+ Masahiro

On Tue, Jun 27, 2023 at 11:11:27AM +0800, Huacai Chen wrote:
> Hi, Nathan,
> 
> On Tue, Jun 27, 2023 at 12:07 AM Nathan Chancellor <nathan@kernel.org> wrote:
> >
> > On Mon, Jun 26, 2023 at 03:50:47PM +0800, Huacai Chen wrote:
> > > After commit 0e96ea5c3eb5904e5dc2f ("MIPS: Loongson64: Clean up use of
> > > cc-ifversion") we get a build error when make modules_install:
> > >
> > > cc1: error: '-mloongson-mmi' must be used with '-mhard-float'
> > >
> > > The reason is when make modules_install, 'call cc-option' doesn't work
> > > in $(KBUILD_CFLAGS) of 'CHECKFLAGS'. Then there is no -mno-loongson-mmi
> > > applied and -march=loongson3a enable MMI instructions.
> >
> > The first sentence does not make much sense to me, specifically "in
> > $(KBUILD_CFLAGS) of 'CHECKFLAGS'". What configuration and build command
> > reproduces this? I do not see how '-mno-loongson-mmi' would fail to get
> > added to cflags-y after 0e96ea5c3eb5, which should have had no
> > functional change... I don't want to hang this change up since there is
> > real breakage but I want to make sure we fully understand why
> > 0e96ea5c3eb5 broke things and why this patch resolves it.
> Please use loongson3_defconfig to build a loongson kernel with
> toolchains from here [1]:
> 'make' will succeed, but there is a build error when 'make
> modules_install'. And you should be careful because 'make
> modules_install' doesn't stop when the error occurs.

Excellent, thank you! I understand what is going on here and your patch
should work to resolve it (although I think the commit message should be
flushed out a little more with the following details) but I am curious
if Masahiro has any thoughts around this.

As you note, the error message comes from the CHECKFLAGS invocation of
$(CC) but it has no impact on the final result of modules_install, it is
purely a cosmetic issue from what I can tell. The error occurs because
cc-option is defined in scripts/Makefile.compiler, which is not included
in Makefile when running modules_install, as install targets are not
supposed to require the compiler; see commit 805b2e1d427a ("kbuild:
include Makefile.compiler only when compiler is needed"). As a result,
the call to check for '-mno-loongson-mmi' just never happens.

It would nice if '-mno-loongson-mmi' could be added unconditionally when
using GCC but I can see that the flag has only existed since 9.x, so we
do need to keep the cc-option call.

I am fine with your change as long as it includes some of the above
information (basically noting that while the original change should have
been equivalent, the requirement of '-mno-loongson-mmi' when using
certain Loongson '-march=' values with '-msoft-float' means that those
Loongson '-march=' values need to be called with cc-option as well),
even if clang will incur two more cc-option calls as a result (not the
end of the world).

Additionally, it seems like the same issue will occur when running
modules_install when CONFIG_CPU_LOONGSON2E or CONFIG_CPU_LOONGSON2F are
enabled, which I guess I also broke in commit 13ceb48bc19c ("MIPS:
Loongson2ef: Remove unnecessary {as,cc}-option calls") :/

Sorry again for the breakage and thanks for the fix!

Cheers,
Nathan

> > > Fix this by partially reverting to the old logic, use 'call cc-option'
> > > to conditionally apply -march=loongson3a and -march=mips64r2.
> > >
> > > Fixes: 0e96ea5c3eb5904e5dc2f ("MIPS: Loongson64: Clean up use of cc-ifversion")
> > > Cc: stable@vger.kernel.org
> > > Cc: Nathan Chancellor <nathan@kernel.org>
> > > Cc: Nick Desaulniers <ndesaulniers@google.com>
> > > Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
> > > ---
> > >  arch/mips/Makefile | 5 +----
> > >  1 file changed, 1 insertion(+), 4 deletions(-)
> > >
> > > diff --git a/arch/mips/Makefile b/arch/mips/Makefile
> > > index a7a4ee66a9d3..7fb76d12829e 100644
> > > --- a/arch/mips/Makefile
> > > +++ b/arch/mips/Makefile
> > > @@ -186,11 +186,8 @@ cflags-$(CONFIG_CPU_LOONGSON2F) += -march=loongson2f -Wa,--trap
> > >  # Some -march= flags enable MMI instructions, and GCC complains about that
> > >  # support being enabled alongside -msoft-float. Thus explicitly disable MMI.
> > >  cflags-$(CONFIG_CPU_LOONGSON2EF) += $(call cc-option,-mno-loongson-mmi)
> > > -ifdef CONFIG_CPU_LOONGSON64
> > >  cflags-$(CONFIG_CPU_LOONGSON64)      += -Wa,--trap
> > > -cflags-$(CONFIG_CC_IS_GCC) += -march=loongson3a
> > > -cflags-$(CONFIG_CC_IS_CLANG) += -march=mips64r2
> > > -endif
> > > +cflags-$(CONFIG_CPU_LOONGSON64) += $(call cc-option,-march=loongson3a,-march=mips64r2)
> > >  cflags-$(CONFIG_CPU_LOONGSON64) += $(call cc-option,-mno-loongson-mmi)
> > >
> > >  cflags-$(CONFIG_CPU_R4000_WORKAROUNDS)       += $(call cc-option,-mfix-r4000,)
> > > --
> > > 2.39.3
> > >

  reply	other threads:[~2023-06-27 21:28 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-06-26  7:50 [PATCH] MIPS: Loongson: Fix build error when make modules_install Huacai Chen
2023-06-26 16:07 ` Nathan Chancellor
2023-06-27  3:11   ` Huacai Chen
2023-06-27 21:28     ` Nathan Chancellor [this message]
2023-06-28  7:36       ` Huacai Chen

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20230627212839.GA1806408@dev-arch.thelio-3990X \
    --to=nathan@kernel.org \
    --cc=chenhuacai@kernel.org \
    --cc=chenhuacai@loongson.cn \
    --cc=jiaxun.yang@flygoat.com \
    --cc=linux-mips@vger.kernel.org \
    --cc=ndesaulniers@google.com \
    --cc=stable@vger.kernel.org \
    --cc=tsbogend@alpha.franken.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox