Re: [PATCH 20/22] x86,word-at-a-time: Remove .fixup usage

[Peter Zijlstra]

On Fri, Nov 05, 2021 at 11:01:52AM -0700, Josh Poimboeuf wrote:
> On Fri, Nov 05, 2021 at 06:10:43PM +0100, Peter Zijlstra wrote:
> > +#ifdef CONFIG_CC_HAS_ASM_GOTO_OUTPUT
> > +
> > +static inline unsigned long load_unaligned_zeropad(const void *addr)
> > +{
> > +	unsigned long offset, data;
> > +	unsigned long ret;
> > +
> > +	asm_volatile_goto(
> > +		"1:	mov %[mem], %[ret]\n"
> > +
> > +		_ASM_EXTABLE(1b, %l[do_exception])
> > +
> > +		: [ret] "=&r" (ret)
> > +		: [mem] "m" (*(unsigned long *)addr)
> > +		: : do_exception);
> > +
> > +out:
> > +	return ret;
> > +
> > +do_exception: __cold;
> > +
> > +	offset = (unsigned long)addr & (sizeof(long) - 1);
> > +	addr = (void *)((unsigned long)addr & ~(sizeof(long) - 1));
> > +	data = *(unsigned long *)addr;
> > +	ret = data >> offset * 8;
> > +	goto out;
> 
> Superfluous goto, can just return here?

Can do I suppose.

> >  static inline unsigned long load_unaligned_zeropad(const void *addr)
> >  {
> > -	unsigned long ret, dummy;
> > +	unsigned long offset, data;
> > +	unsigned long ret, err = 0;
> >  
> > -	asm(
> > -		"1:\tmov %2,%0\n"
> > +	asm(	"1:	mov %[mem], %[ret]\n"
> >  		"2:\n"
> > -		".section .fixup,\"ax\"\n"
> > -		"3:\t"
> > -		"lea %2,%1\n\t"
> > -		"and %3,%1\n\t"
> > -		"mov (%1),%0\n\t"
> > -		"leal %2,%%ecx\n\t"
> > -		"andl %4,%%ecx\n\t"
> > -		"shll $3,%%ecx\n\t"
> > -		"shr %%cl,%0\n\t"
> > -		"jmp 2b\n"
> > -		".previous\n"
> > -		_ASM_EXTABLE(1b, 3b)
> > -		:"=&r" (ret),"=&c" (dummy)
> > -		:"m" (*(unsigned long *)addr),
> > -		 "i" (-sizeof(unsigned long)),
> > -		 "i" (sizeof(unsigned long)-1));
> > +
> > +		_ASM_EXTABLE_FAULT(1b, 2b)
> > +
> > +		: [ret] "=&r" (ret), "+a" (err)
> > +		: [mem] "m" (*(unsigned long *)addr));
> > +
> > +	if (unlikely(err)) {
> > +		offset = (unsigned long)addr & (sizeof(long) - 1);
> > +		addr = (void *)((unsigned long)addr & ~(sizeof(long) - 1));
> > +		data = *(unsigned long *)addr;
> > +		ret = data >> offset * 8;
> > +	}
> > +
> >  	return ret;
> 
> This adds a (normally not taken) conditional jump, would a straight jmp
> over the fixup not be better?
> 
> i.e.
> 
> 	1: mov %[mem], %[ret]
> 	jmp 2
> 	... fixup code ...
> 	2:

I really don't know, this way we can get the bulk of the text
out-of-line.