From mboxrd@z Thu Jan 1 00:00:00 1970 From: Stephen Hemminger Subject: Re: [PATCH 1/3] Kernel interfaces for multiqueue aware socket Date: Wed, 15 Dec 2010 17:23:58 -0800 Message-ID: <20101215172358.64a611c8@nehalam> References: <46a08278c2ba21737528eb4b77391a7e8bc88000.1292405004.git.fenghua.yu@intel.com> <1292446118.2603.11.camel@edumazet-laptop> <20101216011425.GA17446@linux-os.sc.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Eric Dumazet , "David S. Miller" , "Fastabend, John R" , "Tang, Xinan" , Junchang Wang , netdev , linux-kernel To: Fenghua Yu Return-path: In-Reply-To: <20101216011425.GA17446@linux-os.sc.intel.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Wed, 15 Dec 2010 17:14:25 -0800 =46enghua Yu wrote: > On Wed, Dec 15, 2010 at 12:48:38PM -0800, Eric Dumazet wrote: > > Le mercredi 15 d=E9cembre 2010 =E0 12:02 -0800, Fenghua Yu a =E9cri= t : > > > From: Fenghua Yu > > >=20 > > > Multiqueue and multicore provide packet parallel processing metho= dology. > > > Current kernel and network drivers place one queue on one core. B= ut the higher > > > level socket doesn't know multiqueue. Current socket only can rec= eive or send > > > packets through one network interfaces. In some cases e.g. multi = bpf filter > > > tcpdump and snort, a lot of contentions come from socket operatio= ns like ring > > > buffer. Even if the application itself has been fully parallelize= d and run on > > > multi-core systems and NIC handlex tx/rx in multiqueue in paralle= l, network layer > > > and NIC device driver assemble packets to a single, serialized qu= eue. Thus the > > > application cannot actually run in parallel in high speed. > > >=20 > > > To break the serialized packets assembling bottleneck in kernel, = one way is to > > > allow socket to know multiqueue associated with a NIC interface. = So each socket > > > can handle tx/rx in one queue in parallel. > > >=20 > > > Kernel provides several interfaces by which sockets can be bound = to rx/tx queues. > > > User applications can configure socket by providing several socke= ts that each > > > bound to a single queue, applications can get data from kernel in= parallel. After > > > that, competitions mentioned above can be removed. > > >=20 > > > With this patch, the user-space receiving speed on a Intel SR1690= server with > > > a single L5640 6-core processor and a single ixgbe-based NIC goes= from 0.73Mpps > > > to 4.20Mpps, nearly a linear speedup. A Intel SR1625 server two E= 5530 4-core > > > processors and a single ixgbe-based NIC goes from 0.80Mpps to 4.6= Mpps. We noticed > > > the performance penalty comes from NUMA memory allocation. > > >=20 > >=20 > > ??? please elaborate on these NUMA memory allocations. This should = be OK > > after commit 564824b0c52c34692d (net: allocate skbs on local node) > >=20 > > > This patch set provides kernel ioctl interfaces for user space. U= ser space can > > > either directly call the interfaces or libpcap interfaces can be = further provided > > > on the top of the kernel ioctl interfaces. > >=20 > > So, say we have 8 queues, you want libpcap opens 8 sockets, and bin= d > > them to each queue. Add a bpf filter to each one of them. This seem= s not > > generic way, because it wont work for an UDP socket for example. >=20 > This only works for AF_PACKET like this patch set shows. >=20 > > And you already can do this using SKF_AD_QUEUE (added in commit > > d19742fb) >=20 > SKF_AD_QUEUE doesn't know number of rx queues. Thus user application = can't > specify right SKF_AD_QUEUE. >=20 > SKF_AD_QUEUE only works for rx. There is no queue bound interfaces fo= r tx. >=20 > I can change the patch set to use SKF_AD_QUEUE by removing the set rx= queue > interface and still keep interfaces of > #define SIOGNUMRXQUEUE 0x8939 /* Get number of rx queues. */ > #define SIOGNUMTXQUEUE 0x893A /* Get number of tx queues. */ > #define SIOSTXQUEUEMAPPING 0x893C /* Set tx queue mapping. */ > #define SIOGRXQUEUEMAPPING 0x893D /* Get rx queue mapping. */ > #define SIOGTXQUEUEMAPPING 0x893E /* Get tx queue mapping. */ >=20 > >=20 > > Also your AF_PACKET patch only address mmaped sockets. > >=20 > The new patch set will use SKF_AD_QUEUE for rx. So it won't be limite= d to mmaped > sockets. Do we really want to expose this kind of internals to userspace? The problem is once exposed, it becomes a kernel ABI and can not ever change.=20 --=20