Re: [PATCH v3 06/10] lib/refcount: Move saturation warnings out of line

[Will Deacon]

On Thu, Oct 10, 2019 at 01:48:20PM -0700, Kees Cook wrote:
> On Mon, Oct 07, 2019 at 04:46:59PM +0100, Will Deacon wrote:
> > Having the refcount saturation and warnings inline bloats the text,
> > despite the fact that these paths should never be executed in normal
> > operation.
> > 
> > Move the refcount saturation and warnings out of line to reduce the
> > image size when refcount_t checking is enabled. Relative to an x86_64
> > defconfig, the sizes reported by bloat-o-meter are:
> > 
> >  # defconfig+REFCOUNT_FULL, inline saturation (i.e. before this patch)
> >  Total: Before=14762076, After=14915442, chg +1.04%
> > 
> >  # defconfig+REFCOUNT_FULL, out-of-line saturation (i.e. after this patch)
> >  Total: Before=14762076, After=14835497, chg +0.50%
> 
> The downside of this change is that this means we get one warning per
> refcount_saturation_type, where as before it was once per refcount
> usage. I think, ultimately, this is okay, but it is a behavioral change
> that might be worth pointing out.

Good point; I'll add something to the commit message.

> > diff --git a/lib/refcount.c b/lib/refcount.c
> > index 3a534fbebdcc..6a61d39f9390 100644
> > --- a/lib/refcount.c
> > +++ b/lib/refcount.c
> > @@ -8,6 +8,34 @@
> >  #include <linux/spinlock.h>
> >  #include <linux/bug.h>
> >  
> > +#define REFCOUNT_WARN(str)	WARN_ONCE(1, "refcount_t: " str ".\n")
> 
> This define seems like overkill for just adding a prefix to 5 uses...

I dunno. It doesn't hurt and I did get bored of typing that prefix.

> > +void refcount_warn_saturate(refcount_t *r, enum refcount_saturation_type t)
> > +{
> > +	refcount_set(r, REFCOUNT_SATURATED);
> > +
> > +	switch (t) {
> > +	case REFCOUNT_ADD_NOT_ZERO_OVF:
> > +		REFCOUNT_WARN("saturated; leaking memory");
> > +		break;
> > +	case REFCOUNT_ADD_OVF:
> > +		REFCOUNT_WARN("saturated; leaking memory");
> > +		break;
> > +	case REFCOUNT_ADD_UAF:
> > +		REFCOUNT_WARN("addition on 0; use-after-free");
> > +		break;
> > +	case REFCOUNT_SUB_UAF:
> > +		REFCOUNT_WARN("underflow; use-after-free");
> > +		break;
> > +	case REFCOUNT_DEC_LEAK:
> > +		REFCOUNT_WARN("decrement hit 0; leaking memory");
> 
> Even the longest doesn't line wrap:
> 
> 		WARN_ONCE(1, "refcount_t: decrement hit 0; leaking memory\n");
> 
> > +		break;
> > +	default:
> > +		WARN_ON(1);
> 
> This deserves a string too, yes?
> 
> 		WARN_ONCE(1, "refcount_t: unknown saturation event!?\n");

Ok.

> 
> > +	}
> > +}
> > +EXPORT_SYMBOL(refcount_warn_saturate);
> > +
> >  /**
> >   * refcount_dec_if_one - decrement a refcount if it is 1
> >   * @r: the refcount
> > -- 
> > 2.23.0.581.g78d2f28ef7-goog
> > 
> 
> Otherwise, okay, I grudgingly accept the loss of warnings when running
> the lkdtm tests in order to gain back some text size... :)

Thanks :)

Will