- 18 Oct, 2017 11 commits
-
-
Vivien Didelot authored
Many portions of DSA core code require to get the dsa_port structure corresponding to a slave net_device. For this purpose, introduce a dsa_slave_to_port() helper. Signed-off-by:
Vivien Didelot <vivien.didelot@savoirfairelinux.com> Reviewed-by:
Florian Fainelli <f.fainelli@gmail.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Vivien Didelot authored
Both DSA slave create and destroy functions call call_dsa_notifiers with respectively DSA_PORT_REGISTER and DSA_PORT_UNREGISTER and the same dsa_notifier_register_info structure. Wrap this in a dsa_slave_notify helper so prevent cluttering these functions. Signed-off-by:
-
Vivien Didelot authored
When dsa_slave_create is called, the related port already has a CPU port assigned to it, available in its cpu_dp member. Use it instead of the unique tree cpu_dp. Signed-off-by:
-
Johannes Berg authored
When netlink_ack() reports an allocation error to the sending socket, there's no need to look up the sending socket since it's available in the SKB's CB. Use that instead of going to the trouble of looking it up. Note that the pointer is only available since Eric Biederman's commit 3fbc2905 ("netlink: Make the sending netlink socket availabe in NETLINK_CB") which is far newer than the original lookup code (Oct 2003) (though the field was called 'ssk' in that commit and only got renamed to 'sk' later, I'd actually argue 'ssk' was better - or perhaps it should've been 'source_sk' - since there are so many different 'sk's involved.) Signed-off-by:
Johannes Berg <johannes.berg@intel.com> Signed-off-by:
-
David S. Miller authored
Jesper Dangaard Brouer says: ==================== net: New bpf cpumap type for XDP_REDIRECT Introducing a new way to redirect XDP frames. Notice how no driver changes are necessary given the design of XDP_REDIRECT. This redirect map type is called 'cpumap', as it allows redirection XDP frames to remote CPUs. The remote CPU will do the SKB allocation and start the network stack invocation on that CPU. This is a scalability and isolation mechanism, that allow separating the early driver network XDP layer, from the rest of the netstack, and assigning dedicated CPUs for this stage. The sysadm control/configure the RX-CPU to NIC-RX queue (as usual) via procfs smp_affinity and how many queues are configured via ethtool --set-channels. Benchmarks show that a single CPU can handle approx 11Mpps. Thus, only assigning two NIC RX-queues (and two CPUs) is sufficient for handling 10Gbit/s wirespeed smallest packet 14.88Mpps. Reducing the number of queues have the advantage that more packets being "bulk" available per hard interrupt[1]. [1] https://www.netdevconf.org/2.1/papers/BusyPollingNextGen.pdf Use-cases: 1. End-host based pre-filtering for DDoS mitigation. This is fast enough to allow software to see and filter all packets wirespeed. Thus, no packets getting silently dropped by hardware. 2. Given NIC HW unevenly distributes packets across RX queue, this mechanism can be used for redistribution load across CPUs. This usually happens when HW is unaware of a new protocol. This resembles RPS (Receive Packet Steering), just faster, but with more responsibility placed on the BPF program for correct steering. 3. Auto-scaling or power saving via only activating the appropriate number of remote CPUs for handling the current load. The cpumap tracepoints can function as a feedback loop for this purpose. In V7, a --stress-mode was implemented for the samples program, which between each stats update, adds + removes CPUs from the map concurrently with traffic. I did find and fix some concurrency issues in the tear-down path, details in patch desc. The stress test have now been running for 15 hours without any issues, while being bombarded with 11.6 Mpps via pktgen_sample04_many_flows.sh. See individual patches for patchset-version changes. ==================== Signed-off-by:
-
Jesper Dangaard Brouer authored
This sample program show how to use cpumap and the associated tracepoints. It provides command line stats, which shows how the XDP-RX process, cpumap-enqueue and cpumap kthread dequeue is cooperating on a per CPU basis. It also utilize the xdp_exception and xdp_redirect_err transpoints to allow users quickly to identify setup issues. One issue with ixgbe driver is that the driver reset the link when loading XDP. This reset the procfs smp_affinity settings. Thus, after loading the program, these must be reconfigured. The easiest workaround it to reduce the RX-queue to e.g. two via: # ethtool --set-channels ixgbe1 combined 2 And then add CPUs above 0 and 1, like: # xdp_redirect_cpu --dev ixgbe1 --prog 2 --cpu 2 --cpu 3 --cpu 4 Another issue with ixgbe is that the page recycle mechanism is tied to the RX-ring size. And the default setting of 512 elements is too small. This is the same issue with regular devmap XDP_REDIRECT. To overcome this I've been using 1024 rx-ring size: # ethtool -G ixgbe1 rx 1024 tx 1024 V3: - whitespace cleanups - bpf tracepoint cannot access top part of struct V4: - report on kthread sched events, according to tracepoint change - report average bulk enqueue size V5: - bpf_map_lookup_elem on cpumap not allowed from bpf_prog use separate map to mark CPUs not available V6: - correct kthread sched summary output V7: - Added a --stress-mode for concurrently changing underlying cpumap Signed-off-by:
Jesper Dangaard Brouer <brouer@redhat.com> Signed-off-by:
-
Jesper Dangaard Brouer authored
This adds two tracepoint to the cpumap. One for the enqueue side trace_xdp_cpumap_enqueue() and one for the kthread dequeue side trace_xdp_cpumap_kthread(). To mitigate the tracepoint overhead, these are invoked during the enqueue/dequeue bulking phases, thus amortizing the cost. The obvious use-cases are for debugging and monitoring. The non-intuitive use-case is using these as a feedback loop to know the system load. One can imagine auto-scaling by reducing, adding or activating more worker CPUs on demand. V4: tracepoint remove time_limit info, instead add sched info V8: intro struct bpf_cpu_map_entry members cpu+map_id in this patch Signed-off-by:
-
Jesper Dangaard Brouer authored
This patch makes cpumap functional, by adding SKB allocation and invoking the network stack on the dequeuing CPU. For constructing the SKB on the remote CPU, the xdp_buff in converted into a struct xdp_pkt, and it mapped into the top headroom of the packet, to avoid allocating separate mem. For now, struct xdp_pkt is just a cpumap internal data structure, with info carried between enqueue to dequeue. If a driver doesn't have enough headroom it is simply dropped, with return code -EOVERFLOW. This will be picked up the xdp tracepoint infrastructure, to allow users to catch this. V2: take into account xdp->data_meta V4: - Drop busypoll tricks, keeping it more simple. - Skip RPS and Generic-XDP-recursive-reinjection, suggested by Alexei V5: correct RCU read protection around __netif_receive_skb_core. V6: Setting TASK_RUNNING vs TASK_INTERRUPTIBLE based on talk with Rik van Riel Signed-off-by:
-
Jesper Dangaard Brouer authored
This patch connects cpumap to the xdp_do_redirect_map infrastructure. Still no SKB allocation are done yet. The XDP frames are transferred to the other CPU, but they are simply refcnt decremented on the remote CPU. This served as a good benchmark for measuring the overhead of remote refcnt decrement. If driver page recycle cache is not efficient then this, exposes a bottleneck in the page allocator. A shout-out to MST's ptr_ring, which is the secret behind is being so efficient to transfer memory pointers between CPUs, without constantly bouncing cache-lines between CPUs. V3: Handle !CONFIG_BPF_SYSCALL pointed out by kbuild test robot. V4: Make Generic-XDP aware of cpumap type, but don't allow redirect yet, as implementation require a separate upstream discussion. V5: - Fix a maybe-uninitialized pointed out by kbuild test robot. - Restrict bpf-prog side access to cpumap, open when use-cases appear - Implement cpu_map_enqueue() as a more simple void pointer enqueue V6: - Allow cpumap type for usage in helper bpf_redirect_map, general bpf-prog side restriction moved to earlier patch. Signed-off-by:
-
Jesper Dangaard Brouer authored
The 'cpumap' is primarily used as a backend map for XDP BPF helper call bpf_redirect_map() and XDP_REDIRECT action, like 'devmap'. This patch implement the main part of the map. It is not connected to the XDP redirect system yet, and no SKB allocation are done yet. The main concern in this patch is to ensure the datapath can run without any locking. This adds complexity to the setup and tear-down procedure, which assumptions are extra carefully documented in the code comments. V2: - make sure array isn't larger than NR_CPUS - make sure CPUs added is a valid possible CPU V3: fix nitpicks from Jakub Kicinski <kubakici@wp.pl> V5: - Restrict map allocation to root / CAP_SYS_ADMIN - WARN_ON_ONCE if queue is not empty on tear-down - Return -EPERM on memlock limit instead of -ENOMEM - Error code in __cpu_map_entry_alloc() also handle ptr_ring_cleanup() - Moved cpu_map_enqueue() to next patch V6: all notice by Daniel Borkmann - Fix err return code in cpu_map_alloc() introduced in V5 - Move cpu_possible() check after max_entries boundary check - Forbid usage initially in check_map_func_compatibility() V7: - Fix alloc error path spotted by Daniel Borkmann - Did stress test adding+removing CPUs from the map concurrently - Fixed refcnt issue on cpu_map_entry, kthread started too soon - Make sure packets are flushed during tear-down, involved use of rcu_barrier() and kthread_run only exit after queue is empty - Fix alloc error path in __cpu_map_entry_alloc() for ptr_ring V8: - Nitpicking comments and gramma by Edward Cree - Fix missing semi-colon introduced in V7 due to rebasing - Move struct bpf_cpu_map_entry members cpu+map_id to tracepoint patch Signed-off-by:
Alexei Starovoitov <ast@kernel.org> Signed-off-by:
-
Joel Stanley authored
According to the ASPEED datasheet, gigabit speeds require a clock of 100MHz or higher. Other speeds require 25MHz or higher. This patch configures a 100MHz clock if the system has a direct-attached PHY, or 25MHz if the system is running NC-SI which is limited to 100MHz. There appear to be no other upstream users of the FTGMAC100 driver it is hard to know the clocking requirements of other platforms. Therefore a conservative approach was taken with enabling clocks. If the platform is not ASPEED, both requesting the clock and configuring the speed is skipped. Signed-off-by:
Joel Stanley <joel@jms.id.au> Tested-by:
Andrew Jeffery <andrew@aj.id.au> Signed-off-by:
-
- 17 Oct, 2017 1 commit
-
-
Henrik Austad authored
In commit 32302902 ("mqprio: Reserve last 32 classid values for HW traffic classes and misc IDs") sch_mqprio started using netdev_txq_to_tc to find the correct tc instead of dev->tc_to_txq[] However, when mqprio is compiled as a module, it cannot resolve the symbol, leading to this error: ERROR: "netdev_txq_to_tc" [net/sched/sch_mqprio.ko] undefined! This adds an EXPORT_SYMBOL() since the other user in the kernel (netif_set_xps_queue) is also EXPORT_SYMBOL() (and not _GPL) or in a sysfs-callback. Cc: Alexander Duyck <alexander.h.duyck@intel.com> Cc: Jesus Sanchez-Palencia <jesus.sanchez-palencia@intel.com> Cc: David S. Miller <davem@davemloft.net> Signed-off-by:
Henrik Austad <haustad@cisco.com> Reviewed-by:
Eric Dumazet <edumazet@google.com> Acked-by:
Alexander Duyck <alexander.h.duyck@intel.com> Signed-off-by:
-
- 16 Oct, 2017 28 commits
-
-
David S. Miller authored
Jiri Pirko says: ==================== mlxsw: GRE: Offload decap without encap Petr says: The current code doesn't offload GRE decapsulation unless there's a corresponding encapsulation route as well. While not strictly incorrect (when encap route is absent, the decap route traps traffic to CPU and the kernel handles it), it's a missed optimization opportunity. With this patchset, IPIP entries are created as soon as offloadable tunneling netdevice is created. This then leads to offloading of decap route, if one exists, or is added afterwards, even when no encap route is present. In Linux, when there is a decap route, matching IP-in-IP packets are always decapsulated. However, with IPv4 overlays in particular, whether the inner packet is then forwarded depends on setting of net.ipv4.conf.*.rp_filter. When RP filtering is turned on, inner packets aren't forwarded unless there's a matching encap route. The mlxsw driver doesn't reflect this behavior in other router interfaces, and thus it's not implemented for tunnel types either. A better support for this will be subject of follow-up work. ==================== Signed-off-by: -
Petr Machata authored
Formerly, IPIP entries were created lazily by next hops that referenced an offloadable IP-in-IP netdevice. However now that they are created eagerly as a reaction to events on such netdevices, the reference counting is useless. Hence drop it. The routes whose next hops reference an offloaded IP-in-IP netdevice actually linger around a bit after their device is unregistered. However, mlxsw_sp_ipip_entry_destroy() also destroys the backing loopback, and mlxsw_sp_rif_destroy() transitively (via mlxsw_sp_nexthop_rif_gone_sync()) calls mlxsw_sp_nexthop_ipip_fini(), which unlinks the IPIP entry from a next hop. Thus no dangling pointers are left behind for the brief window after netdevice is gone, but routes not yet. Signed-off-by:
Petr Machata <petrm@mellanox.com> Reviewed-by:
Ido Schimmel <idosch@mellanox.com> Signed-off-by:
Jiri Pirko <jiri@mellanox.com> Signed-off-by:
-
Petr Machata authored
IPIP entries are created as soon as an offloadable device is created. That means that when such a device is later moved to a different VRF, the loopback device that backs the tunnel is wrong. Thus when an offloadable encapsulating netdevice moves from one VRF to another, make sure that the loopback is updated as necessary. Signed-off-by:
-
Petr Machata authored
Current code for offloading IP-in-IP tunneling assumes that there is no decap without encap. But that's never true for IPv6 overlays, and is not true for IPv4 ones either, if net.ipv4.conf.*.rp_filter is unset. To support decap-only tunnels, an IPIP entry is now created as soon as an offloadable tunneling device is created. When that netdevice is up'd, a decap route is looked up and possibly offloaded. Thus decap is not handled implicitly as part of mlxsw_sp_ipip_entry_get() call anymore, but needs to be done explicitly after the get, if desired. Signed-off-by:
-
Petr Machata authored
Signed-off-by:
-
Petr Machata authored
So far, all netdevice notifications that the driver cared about were related to its own ports, and mlxsw_sp could be retrieved from the netdevice's private data. For IP-in-IP offloading however, the driver cares about events on foreign netdevices, and getting at mlxsw_sp or router data structures from the handler is inconvenient. Therefore move the netdevice notifier blocks from global scope to struct mlxsw_sp to allow retrieval from the notifier block pointer itself. Signed-off-by:
-
Jon Maloy authored
In commit 2f487712 ("tipc: guarantee that group broadcast doesn't bypass group unicast") there was introduced a last-minute rebasing error that broke non-group communication. We fix this here. Signed-off-by:
Jon Maloy <jon.maloy@ericsson.com> Signed-off-by:
-
David S. Miller authored
Florian Westphal says: ==================== net: core: rcuify rtnl af_ops None of the rtnl af_ops callbacks sleep, so they can be called while holding rcu read lock. Switch handling of af_ops to rcu. This would allow to later call af_ops functions without holding the rtnl mutex anymore. ==================== Signed-off-by: -
Florian Westphal authored
rtnl af_ops currently rely on rtnl mutex: unregister (called from module exit functions) takes the rtnl mutex and all users that do af_ops lookup also take the rtnl mutex. IOW, parallel rmmod will block until doit() callback is done. As none of the af_ops implementation sleep we can use rcu instead. doit functions that need the af_ops can now use rcu instead of the rtnl mutex provided the mutex isn't needed for other reasons. Signed-off-by:
Florian Westphal <fw@strlen.de> Signed-off-by:
-
Florian Westphal authored
next patch will rcu-ify rtnl af_ops, i.e. allow af_ops lookup and function calls with rcu read lock held instead of rtnl mutex. Signed-off-by:
-
Colin Ian King authored
The structure tcp_cdg is local to the source and does not need to be in global scope, so make it static. Cleans up sparse warning: symbol 'tcp_cdg' was not declared. Should it be static? Signed-off-by:
Colin Ian King <colin.king@canonical.com> Signed-off-by:
-
Arnd Bergmann authored
The notifier cause a link error when NET_DSA is a loadable module: drivers/net/ethernet/broadcom/bcmsysport.o: In function `bcm_sysport_remove': bcmsysport.c:(.text+0x1582): undefined reference to `unregister_dsa_notifier' drivers/net/ethernet/broadcom/bcmsysport.o: In function `bcm_sysport_probe': bcmsysport.c:(.text+0x278d): undefined reference to `register_dsa_notifier' This adds a dependency that forces the systemport driver to be a loadable module as well when that happens, but otherwise allows it to be built normally when DSA is either built-in or completely disabled. Fixes: d1565763 ("net: systemport: Establish lower/upper queue mapping") Signed-off-by:
Arnd Bergmann <arnd@arndb.de> Reviewed-by:
-
Sudip Mukherjee authored
Modify baycom driver to use the new parallel port device model. Signed-off-by:
Sudip Mukherjee <sudipm.mukherjee@gmail.com> Signed-off-by:
-
Gustavo A. R. Silva authored
In preparation to enabling -Wimplicit-fallthrough, mark switch cases where we are expecting to fall through. Notice that for options.c file, I placed the "fall through" comment on its own line, which is what GCC is expecting to find. Signed-off-by:
Gustavo A. R. Silva <garsilva@embeddedor.com> Signed-off-by:
-
Andy Shevchenko authored
This removes custom flag handling. Signed-off-by:
Andy Shevchenko <andriy.shevchenko@linux.intel.com> Signed-off-by:
-
Steven Rostedt (VMware) authored
All the trace events defined in include/trace/events/bpf.h are only used when CONFIG_BPF_SYSCALL is defined. But this file gets included by include/linux/bpf_trace.h which is included by the networking code with CREATE_TRACE_POINTS defined. If a trace event is created but not used it still has data structures and functions created for its use, even though nothing is using them. To not waste space, do not define the BPF trace events in bpf.h unless CONFIG_BPF_SYSCALL is defined. Signed-off-by:
Steven Rostedt (VMware) <rostedt@goodmis.org> Acked-by:
-
Wei Wang authored
In order to not dirty the cacheline too often, we try to only update dst->__use and dst->lastusetime at most once per jiffy. As dst->lastusetime is only used by ipv6 garbage collector, it should be good enough time resolution. And __use is only used in ipv6_route_seq_show() to show how many times a dst has been used. And as __use is not atomic_t right now, it does not show the precise number of usage times anyway. So we think it should be OK to only update it at most once per jiffy. According to my latest syn flood test on a machine with intel Xeon 6th gen processor and 2 10G mlx nics bonded together, each with 8 rx queues on 2 NUMA nodes: With this patch, the packet process rate increases from ~3.49Mpps to ~3.75Mpps with a 7% increase rate. Note: dst_use() is being renamed to dst_hold_and_use() to better specify the purpose of the function. Signed-off-by:
Wei Wang <weiwan@google.com> Acked-by:
Eric Dumazet <edumazet@googl.com> Signed-off-by:
-
Wei Wang authored
In fib6_locate(), we need to first make sure fn is not NULL before doing FIB6_SUBTREE(fn) to avoid crash. This fixes the following static checker warning: net/ipv6/ip6_fib.c:1462 fib6_locate() warn: variable dereferenced before check 'fn' (see line 1459) net/ipv6/ip6_fib.c 1458 if (src_len) { 1459 struct fib6_node *subtree = FIB6_SUBTREE(fn); ^^^^^^^^^^^^^^^^ We shifted this dereference 1460 1461 WARN_ON(saddr == NULL); 1462 if (fn && subtree) ^^ before the check for NULL. 1463 fn = fib6_locate_1(subtree, saddr, src_len, 1464 offsetof(struct rt6_info, rt6i_src) Fixes: 66f5d6ce ("ipv6: replace rwlock with rcu and spinlock in fib6_table") Reported-by:Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by:
Martin KaFai Lau <kafai@fb.com> Signed-off-by:
-
Abhijit Ayarekar authored
Update to llvm excludes assembly instructions. llvm git revision is below commit 65fad7c26569 ("bpf: add inline-asm support") This change will be part of llvm release 6.0 __ASM_SYSREG_H define is not required for native compile. -target switch includes appropriate target specific files while cross compiling Tested on x86 and arm64. Signed-off-by:Abhijit Ayarekar <abhijit.ayarekar@caviumnetworks.com> Acked-by:
Daniel Borkmann <daniel@iogearbox.net> Signed-off-by:
-
David S. Miller authored
Jiri Pirko <jiri@mellanox.com> ==================== net: sched: remove some tp->q usage In order to prepare for block sharing, tcf_proto instances need to be independent on particular qdisc instances. This patchset takes care of removal of couple occurrences of tp->q usage. ==================== Signed-off-by: -
Jiri Pirko authored
The callers have this info, they will pass it down to tcf_fill_node. Signed-off-by:
-
Jiri Pirko authored
Use tcf_block_q helper to get q pointer to be used for direct call of sch_tree_lock/unlock instead of tcf_tree_lock/unlock. Signed-off-by:
-
Jiri Pirko authored
Use helper to get q pointer per block. Signed-off-by:
-
Jiri Pirko authored
tc_u_common is now per-q. With blocks, it has to be converted to be per-block. Signed-off-by:
-
Jiri Pirko authored
Instead of using tp->q, use block to get the net pointer. Signed-off-by:
-
Jiri Pirko authored
Whenever the block->q is set, it can be used instead of tp->q as it contains the same value. When it is not set, which can't happen now but it might happen with the follow-up shared blocks introduction, the class is not set in the result. That would lead to a class lookup instead of direct class pointer use for classful qdiscs. However, it is not planned to support classful qdisqs sharing filter blocks, so that may never happen. Signed-off-by:
-
Jiri Pirko authored
These helpers allows to get a q and netdev pointers for given block easily. Signed-off-by:
-
Jiri Pirko authored
Store net pointer in the block structure. Along the way, introduce qdisc_net helper which allows to easily obtain net pointer for qdisc instance. Signed-off-by:
-
