>> [ 550.569802] GPR16: 0000000000000000 0000000000000000 0000000000000000
>> 0000000000000000
>> [ 550.569802] GPR20: 0000000000000000 0000000000000000 0000000000000000
>> 0000000000000000
>> [ 550.569802] GPR24: 0000000000000002 0000000000000001 0000000002002000
>> 0000000002802000
>> [ 550.569802] GPR28: 0000000000000000 0000000000000800 c000000ae08bbe80
>> 0000000000040080
>> [ 550.569899] NIP [c0000000000171dc] arch_local_irq_restore+0x22c/0x230
>> [ 550.569909] LR [c000000000033240]
>> interrupt_exit_user_prepare_main+0x150/0x260
>> [ 550.569919] Call Trace:
>> [ 550.569925] [c000000ae08bbd80] [c000000000033240]
>> interrupt_exit_user_prepare_main+0x150/0x260
>> [ 550.569937] [c000000ae08bbde0] [c000000000033744]
>> syscall_exit_prepare+0x74/0x150
>> [ 550.569948] [c000000ae08bbe10] [c00000000000c758]
>> system_call_common+0xf8/0x268
>
> Yeah this looks like a different issue. Is there a test running which
> flips the security mitigations rapidly? There is a race window with
Yes, powerpc/security/mitigation-patching.sh. This test enables/disables
various supported mitigations (parallel execution).
> the the static branch causing exit_must_hard_disable() returning two
> different values.
>
> We should update they key while single threaded AFAIKS.
Thanks. I tested with this fix. The test ran correctly without a crash.
> diff --git a/arch/powerpc/lib/feature-fixups.c
> b/arch/powerpc/lib/feature-fixups.c
> index 57c6bb802f6c..a7cb317e7039 100644
> --- a/arch/powerpc/lib/feature-fixups.c
> +++ b/arch/powerpc/lib/feature-fixups.c
> @@ -232,11 +232,22 @@ static DEFINE_MUTEX(exit_flush_lock);
>
> static int __do_stf_barrier_fixups(void *data)
> {
> - enum stf_barrier_type *types = data;
> + enum stf_barrier_type types = *(enum stf_barrier_type *)data;
>
> do_stf_entry_barrier_fixups(*types);
> do_stf_exit_barrier_fixups(*types);
>
*types should be changed to “types” to avoid build failure.
