Re: [PATCH] x86, ioremap: use %pR in printk

[Ingo Molnar]

* Ingo Molnar <mingo@xxxxxxx> wrote:

> > No, you don't pass it a phys_addr_t or a resource_size_t, you pass 
> > it a struct resource *
> 
> oops ... i only looked at:
> 
> +       char sym[4*sizeof(resource_size_t) + 8];
> 
> and assumed that it was a resource_size_t printout format :-/
> 
> while printing out ranges is useful too, it's by far not the only 
> source of ugliness all around resource_size_t.

so how about something like the two patches below (ontop of Linus's 
patch): the first patch introduces a "small" resource pointer printout 
format: %pr - the little brother of %pR.

The output format is [0x00001234] - minimum width is 8.

The second patch takes advantage of it in ioremap.c.

Not tested yet but should work. One thing that would be nice is to 
extend it to phys_addr_t as well. Right now there's no guarantee that 
sizeof(resource_size_t) == sizeof(phys_addr_t).

What do you think? Passing it by reference makes it somewhat ugly [and 
restricts it to lvalues], but that's the safest we can get at the moment 
i guess, as it guarantees followup vararg parameter correctness, because 
gcc does not check these extensions.

Worst-case we get gibberish output if it's used incorrectly - but we 
dont get a kernel security hole if a %s follows it in the format string 
and an attacker can control some of the parameters, etc.

	Ingo

---------------->