Re: [PATCH v1 3/5] mm, uaccess: trigger might_sleep() in might_fault() when pagefaults are disabled

[David Hildenbrand <dahi@linux.vnet.ibm.com>]

> From: David Hildenbrand [...
> > > This should be likely() instead of unlikely(), no?
> > > I'd rather write this
> > >
> > > 	if (pagefault_disabled())
> > >  		return;
> > > 	__might_sleep(file, line, 0);
> > >
> > > and leave the likely stuff completely away.
> > 
> > Makes perfect sense!
> 
> From my experience of getting (an older version of) gcc to emit
> 'correctly' statically predicted branches I found that code that
> looks like (I don't think return/goto make any difference):
> 
> 	If (unlikely(condition)) {
> 		code;
> 	}
> 	more_code;
> 
> is compile with a forwards conditional branch (ie ignoring the unlikely()).
> Similarly 'if () continue' is likely to generate a 'predicted taken'
> backwards conditional branch.
> 
> To get the desired effect you need a non-empty 'else' part, an assembler
> comment will suffice, eg: asm volatile("# comment").
> 
> 	David
> 
> 
> 

Thanks for the hint David!

I'm going to drop that unlikely and simply replace in_atomic() by
pagefault_disabled() - will also keep the change minimal!

David