Re: [patch V4 part 3 17/29] x86/entry: Convert Invalid Opcode exception to IDTENTRY

[Andy Lutomirski]

> On May 14, 2020, at 5:33 AM, Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
> 
> ïAndy Lutomirski <luto@xxxxxxxxxx> writes:
> 
>>> On Tue, May 5, 2020 at 7:15 AM Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
>>> 
>>> From: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
>>> 
>>> Convert #UD to IDTENTRY:
>>>  - Implement the C entry point with DEFINE_IDTENTRY
>>>  - Emit the ASM stub with DECLARE_IDTENTRY
>>>  - Remove the ASM idtentry in 64bit
>>>  - Remove the open coded ASM entry code in 32bit
>>>  - Fixup the XEN/PV code
>>>  - Fixup the FOOF bug call in fault.c
>>>  - Remove the old prototyoes
>>> 
>>> No functional change.
>> 
>> I think there *is* a functional change:
>> 
>> 
>>> --- a/arch/x86/mm/fault.c
>>> +++ b/arch/x86/mm/fault.c
>>> @@ -567,7 +567,7 @@ static int is_f00f_bug(struct pt_regs *r
>>>                nr = (address - idt_descr.address) >> 3;
>>> 
>>>                if (nr == 6) {
>>> -                       do_invalid_op(regs, 0);
>>> +                       handle_invalid_op(regs);
>> 
>> I suspect the old code was wrong and no one noticed because no one has
>> a F00F-buggy machine any more.
> 
> I don't think so. It's really just the same thing. The old #UD C
> function was:
> 
> void do_invalid_op(struct pt_regs *regs, long error_code)
> {
>    do_error_trap(regs, error_code, "invalid opcode", X86_TRAP_UD,
>                      SIGILL, ILL_ILLOPN, (void __user *)uprobe_get_trap_addr(regs));
> }
> 
> after expanding the DO_ERROR() muck.
> 
> The new one does:
> 
> void handle_invalid_op(struct pt_regs *regs)
> {
>    do_error_trap(regs, 0, "invalid opcode", X86_TRAP_UD, SIGILL,
>              ILL_ILLOPN, error_get_trap_addr(regs));
> }
> 
> which is exactly the same except for the error code being hardcoded to
> zero in handle_invalid_op() because #UD does not have one.
> 
>> So maybe document that you fixed up the F00F bug, too.  Otherwise:
> 
> Not sure what to document :)

Duh, right, never mind.  Here I was thinking the old C handlers did all the entry/exit work.

> 
> Thanks,
> 
>        tglx