Re: [PATCH 08/15] x86/alternatives: Teach text_poke_bp() to emulate instructions

[Peter Zijlstra]

On Tue, Jun 11, 2019 at 11:14:10AM -0400, Steven Rostedt wrote:
> On Wed, 05 Jun 2019 15:08:01 +0200
> Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> 
> > -void text_poke_bp(void *addr, const void *opcode, size_t len, void *handler)
> > +void text_poke_bp(void *addr, const void *opcode, size_t len, const void *emulate)
> >  {
> >  	unsigned char int3 = 0xcc;
> >  
> > -	bp_int3_handler = handler;
> > +	bp_int3_opcode = emulate ?: opcode;
> >  	bp_int3_addr = (u8 *)addr + sizeof(int3);
> >  	bp_patching_in_progress = true;
> >  
> >  	lockdep_assert_held(&text_mutex);
> >  
> >  	/*
> > +	 * poke_int3_handler() relies on @opcode being a 5 byte instruction;
> > +	 * notably a JMP, CALL or NOP5_ATOMIC.
> > +	 */
> > +	BUG_ON(len != 5);
> 
> If we have a bug on here, why bother with passing in len at all? Just
> force it to be 5.

Masami said the same.

> We could make it a WARN_ON() and return without doing anything.
> 
> This also prevents us from ever changing two byte jmps.

It doesn't; that is, we'd need to add emulation for the 3 byte jump, but
that'd be pretty trivial.