Re: printk()s of user-supplied strings

[Willy Tarreau]

On Mon, Aug 28, 2006 at 01:17:43PM +0200, Krzysztof Halasa wrote:
> Willy Tarreau <w@xxxxxx> writes:
> 
> > Well, I'm not sure about this. Nearly all patches which get merged pass
> > through a public review first, and when you see how many replies you get
> > for and 'else' and and 'if' on two different lines, I expect lots of
> > spontaneous replies such as "use %S for user-supplied strings".
> 
> I wouldn't rely on that.
> 
> >> A solution would be to normally use "%S" and only use
> >> "%s" where "%S" wouldn't work.  In that case, we could as well swap "%s"
> >> and "%S", though - hardening the existing "%s" and introducing "%S" for
> >> those callers that depend on the old behavior.
> 
> I think it's the way to go.
> 
> > I'd rather not change "%s" semantics if we introduce another specifier
> > which does exactly what we would expect "%s" to do.
> 
> Both would be equivalent in most cases. It's better to use "%s" for
> most cases (either secured or not) and leave "%S" for the bunch of
> special cases whose authors better know what are they doing.
> 
> > I will try your proposal to retain the trailing '\n' unescaped.
> 
> I think with "%s" and "%S" this is no longer needed.

Yes it will be for compatibility reasons : we for sure will not fix all
users of "%s" quickly, so we will have to do our best not to break them.
If it was easy to find them all, we could replace "%s" with "%S" everywhere
and make "%S" the escaped one.

But well, I believe that you convinced me that escaping the "%s" and providing
a new "%S" for secure or special usages might be the way to go.

I will propose a patch soon.

> Krzysztof Halasa

thanks,
willy

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/