Re: [RFC][PATCH] printk: Add %pb to print bitmaps

[Ingo Molnar <mingo@kernel.org>]

* Peter Zijlstra <peterz@infradead.org> wrote:

> On Wed, 2012-05-09 at 15:44 +0200, Peter Zijlstra wrote:
> > On Wed, 2012-05-09 at 15:36 +0200, Ingo Molnar wrote:
> > > * Peter Zijlstra <peterz@infradead.org> wrote:
> > > 
> > > > @@ -857,15 +860,19 @@ int kptr_restrict __read_mostly;
> > > >   *       correctness of the format string and va_list arguments.
> > > >   * - 'K' For a kernel pointer that should be hidden from unprivileged users
> > > >   * - 'NF' For a netdev_features_t
> > > > + * - 'b' For a bitmap, consumes 2 args, second is int
> > > 
> > > hm, won't the second arg confuse gcc's printf format checker?
> > 
> > 
> > Ah, yes, I suppose I could abuse something like %*pb. Let me try that.
> 
> I guess I should use %.*pb and keep the field_width in case someone
> manages to actually make bitmap_scnlistprintf() conform to it. The
> precision is unused anyway.

That's a cute trick, and it's intuitive as well.

> +	case 'b':
> +		{
> +			int bits, len;
> +
> +			switch (fmt[1]) {
> +			case 'c':
> +				bits = nr_cpumask_bits;
> +				break;
> +			case 'n':
> +				bits = MAX_NUMNODES;
> +				break;
> +			default:
> +				bits = spec->precision;
> +				break;

So, if someone specifies an incomplete "%pb" format - fmt[1] 
will be 0 and we take precision as the length - presumably also 
0. We stick that 0 into:

> +			len = bitmap_scnlistprintf(buf, end - buf, ptr, bits);

Will that work?

Provided it all tests out fine for you it looks good to me.

I'd probably write the switch statement as:

			switch (fmt[1]) {
			case 'c': bits = nr_cpumask_bits; break;
			case 'n': bits = MAX_NUMNODES;    break;
			default:  bits = spec->precision; break;
			}

... but that's a small detail and a matter of taste in any case.

Thanks,

	Ingo