netdev - Re: [PATCH bpf 2/2] libbpf: remove dependency on barrier.h in xsk.h

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <621a9d70-5397-da90-a68c-6f12f5cbd3b2@iogearbox.net>
Date:   Tue, 9 Apr 2019 11:10:59 +0200
From:   Daniel Borkmann <daniel@...earbox.net>
To:     Magnus Karlsson <magnus.karlsson@...el.com>, bjorn.topel@...el.com,
        ast@...nel.org, netdev@...r.kernel.org
Cc:     bpf@...r.kernel.org, bruce.richardson@...el.com,
        ciara.loftus@...el.com, ilias.apalodimas@...aro.org,
        xiaolong.ye@...el.com, ferruh.yigit@...el.com,
        qi.z.zhang@...el.com, georgmueller@....net, will.deacon@....com
Subject: Re: [PATCH bpf 2/2] libbpf: remove dependency on barrier.h in xsk.h

On 04/09/2019 08:44 AM, Magnus Karlsson wrote:
> The use of smp_rmb() and smp_wmb() creates a Linux header dependency
> on barrier.h that is uneccessary in most parts. This patch implements
> the two small defines that are needed from barrier.h. As a bonus, the
> new implementations are faster than the default ones as they default
> to sfence and lfence for x86, while we only need a compiler barrier in
> our case. Just as it is when the same ring access code is compiled in
> the kernel.
> 
> Fixes: 1cad07884239 ("libbpf: add support for using AF_XDP sockets")
> Signed-off-by: Magnus Karlsson <magnus.karlsson@...el.com>
> ---
>  tools/lib/bpf/xsk.h | 19 +++++++++++++++++--
>  1 file changed, 17 insertions(+), 2 deletions(-)
> 
> diff --git a/tools/lib/bpf/xsk.h b/tools/lib/bpf/xsk.h
> index 3638147..317b44f 100644
> --- a/tools/lib/bpf/xsk.h
> +++ b/tools/lib/bpf/xsk.h
> @@ -39,6 +39,21 @@ DEFINE_XSK_RING(xsk_ring_cons);
>  struct xsk_umem;
>  struct xsk_socket;
>  
> +#if !defined bpf_smp_rmb && !defined bpf_smp_wmb
> +# if defined(__i386__) || defined(__x86_64__)
> +#  define bpf_smp_rmb() asm volatile("" : : : "memory")
> +#  define bpf_smp_wmb() asm volatile("" : : : "memory")
> +# elif defined(__aarch64__)
> +#  define bpf_smp_rmb() asm volatile("dmb ishld" : : : "memory")
> +#  define bpf_smp_wmb() asm volatile("dmb ishst" : : : "memory")
> +# elif defined(__arm__)
> +#  define bpf_smp_rmb() asm volatile("dmb ish" : : : "memory")
> +#  define bpf_smp_wmb() asm volatile("dmb ishst" : : : "memory")
> +# else
> +#  error Architecture not supported by the XDP socket code in libbpf.
> +# endif
> +#endif

Hmm, bit unfortunate, but I can see why it's needed inside the header here
as it's installed into the system and needs to be included by applications.
Fine with me. In any case, I still think we should also fix it for tooling
infra to avoid others running into the same issue, lemme resurrect [0] from
back in the days.

One question though regarding above arm32 implementation. You match these
barriers against the ones from af_xdp in kernel. So smp_* variants from the
arch/arm/include/asm/barrier.h header need to be matched. This selects the
dmb(ish) and dmb(ishst) variant. The implementation of dmb() however is
quite CPU dependent, the above assumes to match with __LINUX_ARM_ARCH__ >= 7.
Perhaps this assumption needs at minimum a comment to make it clear for
developers.

  [0] https://lore.kernel.org/netdev/20181017144156.16639-2-daniel@iogearbox.net/

>  static inline __u64 *xsk_ring_prod__fill_addr(struct xsk_ring_prod *fill,
>  					      __u32 idx)
>  {
> @@ -119,7 +134,7 @@ static inline void xsk_ring_prod__submit(struct xsk_ring_prod *prod, size_t nb)
>  	/* Make sure everything has been written to the ring before signalling
>  	 * this to the kernel.
>  	 */
> -	smp_wmb();
> +	bpf_smp_wmb();
>  
>  	*prod->producer += nb;
>  }
> @@ -133,7 +148,7 @@ static inline size_t xsk_ring_cons__peek(struct xsk_ring_cons *cons,
>  		/* Make sure we do not speculatively read the data before
>  		 * we have received the packet buffers from the ring.
>  		 */
> -		smp_rmb();
> +		bpf_smp_rmb();
>  
>  		*idx = cons->cached_cons;
>  		cons->cached_cons += entries;
>