Le 28/01/2020 à 19:39, Nick Hudson a écrit : > On 28/01/2020 17:47, Maxime Villard wrote: >> @@ -460,8 +460,7 @@ cpu_setup_id(struct cpu_info *ci) >> >> id->ac_aa64mmfr0 = reg_id_aa64mmfr0_el1_read(); >> id->ac_aa64mmfr1 = reg_id_aa64mmfr1_el1_read(); >> - /* Only in ARMv8.2. */ >> - id->ac_aa64mmfr2 = 0 /* reg_id_aa64mmfr2_el1_read() */; >> + id->ac_aa64mmfr2 = reg_id_aa64mmfr2_el1_read(); >> >> id->ac_mvfr0 = reg_mvfr0_el1_read(); >> id->ac_mvfr1 = reg_mvfr1_el1_read(); > > I didn't ok this bit...
Verily you did; I sent you this patch as-is two weeks ago. > This needs to be conditional on the CPU we're running on. ID_AA64MMFR2_EL1 is res0 on < ARMv8.2, reading it is therefore not a problem.
