[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <aMrFp/2TY2BYQ76W@e129823.arm.com>
Date: Wed, 17 Sep 2025 15:28:55 +0100
From: Yeoreum Yun <yeoreum.yun@....com>
To: Dave Martin <Dave.Martin@....com>
Cc: catalin.marinas@....com, will@...nel.org, broonie@...nel.org,
oliver.upton@...ux.dev, anshuman.khandual@....com, robh@...nel.org,
james.morse@....com, mark.rutland@....com, joey.gouly@....com,
ahmed.genidi@....com, kevin.brodsky@....com,
scott@...amperecomputing.com, mbenes@...e.cz,
james.clark@...aro.org, frederic@...nel.org, rafael@...nel.org,
pavel@...nel.org, ryan.roberts@....com, suzuki.poulose@....com,
maz@...nel.org, linux-arm-kernel@...ts.infradead.org,
linux-kernel@...r.kernel.org, linux-pm@...r.kernel.org,
kvmarm@...ts.linux.dev
Subject: Re: [PATCH v4 2/5] arm64: initialise SCTLR2_ELx register at boot time
Hi Dave,
[...]
> > diff --git a/arch/arm64/include/asm/assembler.h b/arch/arm64/include/asm/assembler.h
> > index 23be85d93348..c25c2aed5125 100644
> > --- a/arch/arm64/include/asm/assembler.h
> > +++ b/arch/arm64/include/asm/assembler.h
> > @@ -738,6 +738,21 @@ alternative_endif
> > set_sctlr sctlr_el2, \reg
> > .endm
> >
> > +/* Set SCTLR2_ELx to the @reg value. */
> > +.macro set_sctlr2_elx, el, reg, tmp
> > + mrs_s \tmp, SYS_ID_AA64MMFR3_EL1
> > + ubfx \tmp, \tmp, #ID_AA64MMFR3_EL1_SCTLRX_SHIFT, #4
> > + cbz \tmp, .Lskip_sctlr2_\@
> > + .if \el == 2
> > + msr_s SYS_SCTLR2_EL2, \reg
> > + .elseif \el == 12
> > + msr_s SYS_SCTLR2_EL12, \reg
> > + .else
> > + msr_s SYS_SCTLR2_EL1, \reg
> > + .endif
> > +.Lskip_sctlr2_\@:
> > +.endm
> > +
>
> You were right that just doing
>
> msr_s SYS_SCTLR_\el, \reg
>
> here doesn't work. It looks like we rely on the C preprocessor to
> expand the SYS_FOO_REG names to numeric sysreg encodings. By the time
> the assembler macros are expanded, it is too late to construct sysreg
> names -- the C preprocessor only runs once, before the assembler.
>
> So, your code here looks reasonable.
>
> But, it will still silently do the wrong thing if \el is empty or
> garbage, so it is probably worth adding an error check:
>
> .else
> .error "Bad EL \"\el\" in set_sctlr2_elx"
> .endif
>
>
> Also, looking at all this again, the "1", "2" and "12" suffixes are not
> really numbers: SYS_REG_EL02 would definitely not be the same thing as
> SYS_REG_EL2 (although there is no _EL02 version of this particular
> register).
>
> So, can you use .ifc to do a string comparison instead?
>
> If might be a good idea to migrate other macros that use an "el"
> argument to do something similar -- although that probably doesn't
> belong in this series.
>
> The assembler seems to have no ".elseifc" directive, so you'll need
> separate nested .ifc blocks:
>
> .ifc \el,2
> msr_s SYS_SCTLR2_EL2, \reg
> .else
> .ifc \el,12
> msr_s SYS_SCTLR2_EL12, \reg
> .else
> .ifc \el,1
> msr_s SYS_SCTLR2_EL1, \reg
> .else
> .error "Bad EL \"\el\" in set_sctlr2_elx"
> .endif
> .endif
> .endif
>
> (Note, I've not tested this approach. If you can think of a better
> way, please feel free to suggest.)
>
Sorry for late reply. but when I find some usage like above.
I couldn't find any usage for this except this macro.
In case of entry, since it just only chekc for "el0" case
I think it doesn't need to apply this for them.
So, let me apply this for set_sctlr2_elx only right now.
when some new register requires this kind of pattern,
let's apply at that time more generally.
Thanks.
--
Sincerely,
Yeoreum Yun
Powered by blists - more mailing lists