Re: [Xen-devel] [PATCH] xen/x86: Implement enable_nmis() in C

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 15/03/18 17:02, Jan Beulich wrote:
>>>> On 15.03.18 at 17:43, <andrew.cooper3@xxxxxxxxxx> wrote:
>> +static inline void enable_nmis(void)
>> +{
>> +    unsigned long tmp;
>> +
>> +    asm volatile ( "mov %%rsp, %[sp]     \n\t"
>> +                   "push %[ss]           \n\t"
>> +                   "push %[sp]           \n\t"
>> +                   "pushf                \n\t"
>> +                   "push %[cs]           \n\t"
>> +                   "lea 1f(%%rip), %[ip] \n\t"
>> +                   "push %[ip]           \n\t"
>> +                   "iretq; 1:            \n\t"
>> +                   : [sp] "=r" (tmp), [ip] "=r" (tmp)
> Strictly speaking this needs to be "=&r" in both cases. That'll
> guarantee the compiler to pick two distinct registers (not sure
> how that ends up being with the code you have), which is
> more than we need want. How about having just a single
> [tmp] output?
>
> With at least the missing & added
> Reviewed-by: Jan Beulich <jbeulich@xxxxxxxx>

In this case, everything works fine even if the compiler picks the same
register.  GCC 7.3 picks %rax for sp and %rdx for.

Then again, we can get away with a single tmp, so I'll switch to that.

~Andrew

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel