On Wed, Sep 28, 2022 at 10:58:37PM +0200, Stefano Brivio wrote:
> On Wed, 28 Sep 2022 14:33:32 +1000
> David Gibson <david(a)gibson.dropbear.id.au> wrote:
> 
> > Recent versions of cppcheck give a warning due to the NULL buffer passed
> > to recv() in tcp_sock_consume().  Since this apparently works, I assume
> > it's actually valid,
> 
> Yes, given that we use MSG_TRUNC to discard socket buffers, I thought
> it's cleaner to avoid supplying a data buffer altogether.
> 
> POSIX doesn't specify MSG_TRUNC, and whether the buffer can be NULL
> isn't specified in Linux documentation, but it works reliably (the
> kernel won't even look at it).

Ah, interesting.  I've update the comment and commit message with this
detail.

> This was actually the first step of a long-overdue plan: if you observe
> this through perf(1), you'll see some overhead in the kernel, which
> looks a bit more than a reasonable expectation for pure syscall
> overhead.
> 
> So, I planned to check if we can simplify the kernel path if no buffer
> is passed. I'll track this somewhere next week.
> 
> Even the day we add a vhost-user back-end, this overhead is still going
> to be there, as that wouldn't change anything host-side.
> 
> > but cppcheck doesn't know that recv() can take a NULL
> > buffer.  So, use a suppression to get rid of the error.
> 
> valgrind didn't know either, by the way -- see the corresponding
> suppression in test/valgrind.supp.
> 

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson