Re: [PATCH 17/21] x86/acpi: Convert indirect jump to retpoline

[Andrew Cooper]

On 14/01/2021 23:47, Josh Poimboeuf wrote:
> On Thu, Jan 14, 2021 at 10:59:39PM +0000, Andrew Cooper wrote:
>> On 14/01/2021 19:40, Josh Poimboeuf wrote:
>>> It's kernel policy to not have (unannotated) indirect jumps because of
>>> Spectre v2.  This one's probably harmless, but better safe than sorry.
>>> Convert it to a retpoline.
>>>
>>> Cc: "Rafael J. Wysocki" <rjw@xxxxxxxxxxxxx>
>>> Cc: Len Brown <len.brown@xxxxxxxxx>
>>> Cc: Pavel Machek <pavel@xxxxxx>
>>> Signed-off-by: Josh Poimboeuf <jpoimboe@xxxxxxxxxx>
>>> ---
>>>  arch/x86/kernel/acpi/wakeup_64.S | 3 ++-
>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/arch/x86/kernel/acpi/wakeup_64.S b/arch/x86/kernel/acpi/wakeup_64.S
>>> index 5d3a0b8fd379..0b371580e620 100644
>>> --- a/arch/x86/kernel/acpi/wakeup_64.S
>>> +++ b/arch/x86/kernel/acpi/wakeup_64.S
>>> @@ -7,6 +7,7 @@
>>>  #include <asm/msr.h>
>>>  #include <asm/asm-offsets.h>
>>>  #include <asm/frame.h>
>>> +#include <asm/nospec-branch.h>
>>>  
>>>  # Copyright 2003 Pavel Machek <pavel@xxxxxxx
>>>  
>>> @@ -39,7 +40,7 @@ SYM_FUNC_START(wakeup_long64)
>>>  	movq	saved_rbp, %rbp
>>>  
>>>  	movq	saved_rip, %rax
>>> -	jmp	*%rax
>>> +	JMP_NOSPEC rax
>>>  SYM_FUNC_END(wakeup_long64)
>> I suspect this won't work as you intend.
>>
>> wakeup_long64() still executes on the low mappings, not the high
>> mappings, so the `jmp __x86_indirect_thunk_rax` under this JMP_NOSPEC
>> will wander off into the weeds.
>>
>> This is why none of the startup "jmps from weird contexts onto the high
>> mappings" have been retpolined-up.
> D'oh.  Of course it wouldn't be that easy.  I suppose the other two
> retpoline patches (15 and 21) are bogus as well.

If by 21 you mean 19, then most likely, yes.  They're all low=>high
jumps in weird contexts.

~Andrew