Date:   Wed, 13 Apr 2022 17:37:22 +0900
From:   Masahiro Yamada <masahiroy@...nel.org>
To:     Libo Chen <libo.chen@...cle.com>
Cc:     Randy Dunlap <rdunlap@...radead.org>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        "Peter Zijlstra (Intel)" <peterz@...radead.org>,
        Ingo Molnar <mingo@...nel.org>,
        Vlastimil Babka <vbabka@...e.cz>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Linux Kbuild mailing list <linux-kbuild@...r.kernel.org>,
        linux-arm-kernel <linux-arm-kernel@...ts.infradead.org>,
        linux-arch <linux-arch@...r.kernel.org>
Subject: Re: [PATCH RESEND 1/1] lib/Kconfig: remove DEBUG_PER_CPU_MAPS
 dependency for CPUMASK_OFFSTACK

On Wed, Apr 13, 2022 at 3:56 PM Libo Chen <libo.chen@...cle.com> wrote:
>
> Hi Randy
>
> On 4/12/22 22:54, Randy Dunlap wrote:
> > Hi Libo,
> >
> > On 4/12/22 19:34, Libo Chen wrote:
> >>
> >> On 4/12/22 19:13, Randy Dunlap wrote:
> >>> Hi,
> >>>
> >>> On 4/12/22 18:35, Libo Chen wrote:
> >>>> Hi Randy,
> >>>>
> >>>> On 4/12/22 17:18, Randy Dunlap wrote:
> >>>>> Hi--
> >>>>>
> >>>>> On 4/12/22 16:15, Libo Chen wrote:
> >>>>>> Forcing CPUMASK_OFFSTACK to be conditoned on DEBUG_PER_CPU_MAPS doesn't
> >>>>>> make a lot of sense nowaday. Even the original patch dating back to 2008,
> >>>>>> aab46da0520a ("cpumask: Add CONFIG_CPUMASK_OFFSTACK") didn't give any
> >>>>>> rationale for such dependency.
> >>>>>>
> >>>>>> Nowhere in the code supports the presumption that DEBUG_PER_CPU_MAPS is
> >>>>>> necessary for CONFIG_CPUMASK_OFFSTACK. Make no mistake, it's good to
> >>>>>> have DEBUG_PER_CPU_MAPS for debugging purpose or precaution, but it's
> >>>>>> simply not a hard requirement for CPUMASK_OFFSTACK. Moreover, x86 Kconfig
> >>>>>> already can set CPUMASK_OFFSTACK=y without DEBUG_PER_CPU_MAPS=y.
> >>>>>> There is no reason other architectures cannot given the fact that they
> >>>>>> have even fewer, if any, arch-specific CONFIG_DEBUG_PER_CPU_MAPS code than
> >>>>>> x86.
> >>>>>>
> >>>>>> Signed-off-by: Libo Chen <libo.chen@...cle.com>
> >>>>>> ---
> >>>>>>     lib/Kconfig | 2 +-
> >>>>>>     1 file changed, 1 insertion(+), 1 deletion(-)
> >>>>>>
> >>>>>> diff --git a/lib/Kconfig b/lib/Kconfig
> >>>>>> index 087e06b4cdfd..7209039dfb59 100644
> >>>>>> --- a/lib/Kconfig
> >>>>>> +++ b/lib/Kconfig
> >>>>>> @@ -511,7 +511,7 @@ config CHECK_SIGNATURE
> >>>>>>         bool
> >>>>>>       config CPUMASK_OFFSTACK
> >>>>>> -    bool "Force CPU masks off stack" if DEBUG_PER_CPU_MAPS
> >>>>> This "if" dependency only controls whether the Kconfig symbol's prompt is
> >>>>> displayed (presented) in kconfig tools. Removing it makes the prompt always
> >>>>> be displayed.
> >>>>>
> >>>>> Any architecture could select (should be able to) CPUMASK_OFFSTACK independently
> >>>>> of DEBUG_PER_CPU_MAPS.
> >>>> Do you mean changing arch/xxxx/Kconfig to select CPUMASK_OFFSTACK under some config xxx? That will work but it requires code changes for each architecture.
> >>>> But if you are talking about setting CONFIG_CPUMASK_OFFSTACK=y without CONFIG_DEBUG_PER_CPU_MAPS directly in config file, I have tried, it doesn't work.
> >>> I'm just talking about the Kconfig change below.  Not talking about whatever else
> >>> it might require per architecture.
> >>>
> >>> But you say you have tried that and it doesn't work. What part of it doesn't work?
> >>> The Kconfig part or some code execution?
> >> oh the Kconfig part. For example, make olddefconfig on a config file with CPUMASK_OFFSTACK=y only turns off CPUMASK_OFFSTACK unless I explicitly set DEBUG_PER_CPU_MAPS=y
> > I can enable CPUMASK_OFFSTACK for arm64 without having DEBUG_PER_CPU_MAPS enabled.
> > (with a patch, of course.)
> > It builds OK. I don't know if it will run OK.
>
> I am a little confused, did you succeed with your patch (replacing "if"
> with "depends on") or my patch (removing "if")? Because I definitely
> cannot enable CPUMASK_OFFSTACK for arm64 without DEBUG_PER_CPUMAPS
> enabled using your change.
> > I think that you are arguing for a patch like this:
>
> I am actually arguing for the opposite, I don't think CPUMASK_OFFSTACK
> should require DEBUG_PER_CPU_MAPS. They should be separate and
> independent to each other. So removing "if ..." should be enough in my
> opinion.
> > --- a/lib/Kconfig
> > +++ b/lib/Kconfig
> > @@ -511,7 +511,8 @@ config CHECK_SIGNATURE
> >       bool
> >
> >   config CPUMASK_OFFSTACK
> > -     bool "Force CPU masks off stack" if DEBUG_PER_CPU_MAPS
> > +     bool "Force CPU masks off stack"
> > +     depends on DEBUG_PER_CPU_MAPS
>
> This forces every arch to enable DEBUG_PER_CPU_MAPS if they want to
> enable CPUMASK_OFFSTACK, it's even stronger than "if". My whole argument
> is CPUMASK_OFFSTACK should be enable/disabled independent of
> DEBUG_PER_CPU_MASK
> >       help
> >         Use dynamic allocation for cpumask_var_t, instead of putting
> >         them on the stack.  This is a bit more expensive, but avoids
> >
> >
> > As I said earlier, the "if" on the "bool" line just controls the prompt message.
> > This patch make CPUMASK_OFFSTACK require DEBUG_PER_CPU_MAPS -- which might be overkill.
> >
>
> Okay I understand now "if" on the "boot" is not a dependency and it only
> controls the prompt message, then the question is why we cannot enable
> CPUMASK_OFFSTACK without DEBUG_PER_CPU_MAPS if it only controls prompt
> message? Is it not the behavior we expect?
>


    config CPUMASK_OFFSTACK
            bool "Force CPU masks off stack" if DEBUG_PER_CPU_MAPS

... is equivalent to this:

    config CPUMASK_OFFSTACK
            bool
            prompt "Force CPU masks off stack" if DEBUG_PER_CPU_MAPS



When DEBUG_PER_CPU_MAPS is disabled, the prompt line is ignored,
and CPUMASK_OFFSTACK becomes a user-unconfigurable option.


Other options still can select it,
but users cannot enable it directly from the prompt.

I see x86 and powerpc do this.

$ kgrep 'select CPUMASK_OFFSTACK'
./arch/x86/Kconfig:946: select CPUMASK_OFFSTACK
./arch/powerpc/Kconfig:164: select CPUMASK_OFFSTACK if NR_CPUS >= 8192







> Libo
>
> >> Libo
> >>> I'll test the Kconfig part of it later (in a few hours).
> >>>
> >>>> Libo
> >>>>> Is there another problem here?
> >>>>>
> >>>>>> +    bool "Force CPU masks off stack"
> >>>>>>         help
> >>>>>>           Use dynamic allocation for cpumask_var_t, instead of putting
> >>>>>>           them on the stack.  This is a bit more expensive, but avoids
>


-- 
Best Regards
Masahiro Yamada

Powered by blists - more mailing lists