[PATCH v4 5/5] arm64/kernel: enable A53 erratum #8434319 handling at runtime

[Ard Biesheuvel]

On 8 March 2018 at 13:49, Mark Rutland <mark.rutland at arm.com> wrote:
> On Thu, Mar 08, 2018 at 01:46:34PM +0000, Ard Biesheuvel wrote:
>> On 8 March 2018 at 13:45, Will Deacon <will.deacon at arm.com> wrote:
>> > On Tue, Mar 06, 2018 at 05:15:35PM +0000, Ard Biesheuvel wrote:
>> >> Omit patching of ADRP instruction at module load time if the current
>> >> CPUs are not susceptible to the erratum.
>> >
>> > [...]
>> >
>> >> diff --git a/arch/arm64/kernel/module-plts.c b/arch/arm64/kernel/module-plts.c
>> >> index 534bf1d47119..1a583ccace00 100644
>> >> --- a/arch/arm64/kernel/module-plts.c
>> >> +++ b/arch/arm64/kernel/module-plts.c
>> >> @@ -158,7 +158,8 @@ static unsigned int count_plts(Elf64_Sym *syms, Elf64_Rela *rela, int num,
>> >>                       break;
>> >>               case R_AARCH64_ADR_PREL_PG_HI21_NC:
>> >>               case R_AARCH64_ADR_PREL_PG_HI21:
>> >> -                     if (!IS_ENABLED(CONFIG_ARM64_ERRATUM_843419))
>> >> +                     if (!IS_ENABLED(CONFIG_ARM64_ERRATUM_843419) ||
>> >> +                         !cpus_have_const_cap(ARM64_WORKAROUND_843419))
>> >>                               break;
>> >>
>> >>                       /*
>> >> diff --git a/arch/arm64/kernel/module.c b/arch/arm64/kernel/module.c
>> >> index 89217704944e..47b40aaa1a5d 100644
>> >> --- a/arch/arm64/kernel/module.c
>> >> +++ b/arch/arm64/kernel/module.c
>> >> @@ -201,6 +201,7 @@ static int reloc_insn_imm(enum aarch64_reloc_op op, __le32 *place, u64 val,
>> >>  static int reloc_insn_adrp(struct module *mod, __le32 *place, u64 val)
>> >>  {
>> >>       if (!IS_ENABLED(CONFIG_ARM64_ERRATUM_843419) ||
>> >> +         !cpus_have_const_cap(ARM64_WORKAROUND_843419) ||
>> >
>> > Mind if I drop the IS_ENABLED check here and in the hunk above? The
>> > const_cap check along should be sufficient, no?
>> >
>>
>> Without the IS_ENABLED() check, the code will always be present in the
>> object file.
>>
>> I have no strong preference either way, though.
>
> As with other case, perhaps fold this into a helper in
> <asm/cpufeature.h> ?
>
> static inline bool system_needs_arm64_workaround_843419()
> {
>         return (IS_ENABLED(CONFIG_ARM64_ERRATUM_843419) &&
>                 cpus_have_const_cap(ARM64_WORKAROUND_843419))
> }
>
> ... then use the inverse in the cases above.
>

I'm fine with adding a helper, but
'system_needs_arm64_workaround_843419' is a bit misleading, given that
it returns false if the system needs it but support is compiled out.