insert NOP after STI

[Freax13]

Insert a NOP instruction after STI to make sure the interrupt shadow doesn't extend past the enable function.

From the AMD64 APM, Volume 3:

If STI sets the IF flag and IF was initially clear, then interrupts
are not enabled until after the instruction following STI. Thus, if
IF is 0, this code will not allow an INTR to happen:

STI
CLI

In the following sequence, INTR will be allowed to happen only after the NOP.

STI
NOP
CLI

[phil-opp]

The change itself looks good to me, but I'm not sure if this is a breaking change. There might be some use cases that rely on the interrupt shadow, which this change might break in subtle ways.

[Freax13]

I don't think one could have reliably relied on the interrupt shadow after the enable function. Rust doesn't guarantee inlining, so one would always run the risk of the next instruction after STI being ret.
Note that we do have enable_and_hlt for STI+HLT with the guarantee that HLT is in the interrupt shadow.

[phil-opp]

Good point! We should update the method documentation though to document that the method does a sti+nop and explain why (interrupt shadow). Maybe also link enable_and_hlt

[phil-opp]

Thanks!

[Freax13]

Thanks for the quick review!