- 19 3月, 2015 27 次提交
-
-
由 David Ahern 提交于
Implement the phys_port_name operation. Port names are pulled from the rocker hardware model in qemu and default to the qemu name + port id. e.g., sw1p1: flags=4098<BROADCAST,MULTICAST> mtu 1500 ether 52:54:00:12:35:01 txqueuelen 1000 (Ethernet) RX packets 0 bytes 0 (0.0 B) RX errors 0 dropped 0 overruns 0 frame 0 TX packets 0 bytes 0 (0.0 B) TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0 where 'sw1' comes from the qemu command line -device rocker,name=sw1, and 'p1' is port 1. Patch is adapted from Scott's phys_port_id patch. Signed-off-by: NDavid Ahern <dsahern@gmail.com> Acked-by: NScott Feldman <sfeldma@gmail.com> Acked-by: NJiri Pirko <jiri@resnulli.us> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 David Ahern 提交于
Similar to port id allow netdevices to specify port names and export the name via sysfs. Drivers can implement the netdevice operation to assist udev in having sane default names for the devices using the rule: $ cat /etc/udev/rules.d/80-net-setup-link.rules SUBSYSTEM=="net", ACTION=="add", ATTR{phys_port_name}!="", NAME="$attr{phys_port_name}" Use of phys_name versus phys_id was suggested-by Jiri Pirko. Signed-off-by: NDavid Ahern <dsahern@gmail.com> Acked-by: NJiri Pirko <jiri@resnulli.us> Acked-by: NScott Feldman <sfeldma@gmail.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Marcelo Ricardo Leitner 提交于
Currently, if a multicast join operation fail, the vxlan interface will be UP but not functional, without even a log message informing the user. Now that we can grab socket lock after already having rntl, we don't need to defer socket creation and multicast operations. By not deferring we can do proper error reporting to the user through ip exit code. This patch thus removes all deferred work that vxlan had and put it back inline. Now the socket will only be created, bound and join multicast group when one bring the interface up, and will undo all that as soon as one put the interface down. As vxlan_sock_hold() is not used after this patch, it was removed too. Signed-off-by: NMarcelo Ricardo Leitner <marcelo.leitner@gmail.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Marcelo Ricardo Leitner 提交于
in favor of their inner __ ones, which doesn't grab rtnl. As these functions need to operate on a locked socket, we can't be grabbing rtnl by then. It's too late and doing so causes reversed locking. So this patch: - move rtnl handling to callers instead while already fixing some reversed locking situations, like on vxlan and ipvs code. - renames __ ones to not have the __ mark: __ip_mc_{join,leave}_group -> ip_mc_{join,leave}_group __ipv6_sock_mc_{join,drop} -> ipv6_sock_mc_{join,drop} Signed-off-by: NMarcelo Ricardo Leitner <marcelo.leitner@gmail.com> Acked-by: NHannes Frederic Sowa <hannes@stressinduktion.org> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Marcelo Ricardo Leitner 提交于
There are some setsockopt operations in ipv4 and ipv6 that are grabbing rtnl after having grabbed the socket lock. Yet this makes it impossible to do operations that have to lock the socket when already within a rtnl protected scope, like ndo dev_open and dev_stop. We normally take coarse grained locks first but setsockopt inverted that. So this patch invert the lock logic for these operations and makes setsockopt grab rtnl if it will be needed prior to grabbing socket lock. Signed-off-by: NMarcelo Ricardo Leitner <marcelo.leitner@gmail.com> Acked-by: NHannes Frederic Sowa <hannes@stressinduktion.org> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 David S. Miller 提交于
Eric Dumazet says: ==================== inet: tcp listener refactoring, part 13 inet_hash functions are in a bad state : Too much IPv6/IPv4 copy/pasting. Lets refactor a bit. Idea is that we do not want to have an equivalent of inet_csk(sk)->icsk_af_ops for request socks in order to be able to use the right variant. In this patch series, I started to let IPv6/IPv4 converge to common helpers. Idea is to use ipv6_addr_set_v4mapped() even for AF_INET sockets, so that we can test if (sk->sk_family == AF_INET6 && !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) to tell if we deal with an IPv6 socket, or IPv4 one, at least in slow paths. Ideally, we could save 8 bytes per struct sock_common, if we alias skc_daddr & skc_rcv_saddr to skc_v6_daddr[3]/skc_v6_rcv_saddr[3]. ==================== Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
In order to be able to use sk_ehashfn() for request socks, we need to initialize their IPv6/IPv4 addresses. Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
We now always call __inet_hash_nolisten(), no need to pass it as an argument. Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
We can now use inet_hash() and __inet_hash() instead of private functions. Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
Intent is to converge IPv4 & IPv6 inet_hash functions to factorize code. IPv4 sockets initialize sk_rcv_saddr and sk_v6_daddr in this patch, thanks to new sk_daddr_set() and sk_rcv_saddr_set() helpers. __inet6_hash can now use sk_ehashfn() instead of a private inet6_sk_ehashfn() and will simply use __inet_hash() in a following patch. Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
Goal is to unify IPv4/IPv6 inet_hash handling, and use common helpers for all kind of sockets (full sockets, timewait and request sockets) inet_sk_ehashfn() becomes sk_ehashfn() but still only copes with IPv4 Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
const qualifiers ease code review by making clear which objects are not written in a function. Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 David S. Miller 提交于
Or Gerlitz says: ==================== Add max rate TXQ attribute Add the ability to set a max-rate limitation for TX queues. The attribute name is maxrate and the units are Mbs, to make it similar to the existing max-rate limitation knobs (ETS and SRIOV ndo calls). changes from V2: - added Documentation (thanks Florian and Tom) - rebased to latest net-next to comply with the swdev ndo removal - addressed more feedback from Dave on the comments style changes from V1: - addressed feedback from Dave changes from V0: - addressed feedback from Sergei John Fastabend (1): net: Add max rate tx queue attribute ==================== Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Or Gerlitz 提交于
Add ndo_set_tx_maxrate support. To support per tx queue maxrate limit, we use the update-qp firmware command to do run-time rate setting for the qp that serves this tx ring. Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com> Signed-off-by: NIdo Shamay <idos@mellanox.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Or Gerlitz 提交于
Add the low-level device commands and definitions used for QP max-rate limiting. This is done through the following elements: - read rate-limit device caps in QUERY_DEV_CAP: number of different rates and the min/max rates in Kbs/Mbs/Gbs units - enhance the QP context struct to contain rate limit units and value - allow to do run time rate-limit setting to QPs through the update-qp firmware command - QP rate-limiting is disallowed for VFs Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 John Fastabend 提交于
This adds a tx_maxrate attribute to the tx queue sysfs entry allowing for max-rate limiting. Along with DCB-ETS and BQL this provides another knob to tune queue performance. The limit units are Mbps. By default it is disabled. To disable the rate limitation after it has been set for a queue, it should be set to zero. Signed-off-by: NJohn Fastabend <john.r.fastabend@intel.com> Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 David S. Miller 提交于
Herbert Xu says: ==================== rhashtable: Kill redundant shift parameter I was trying to squeeze bucket_table->rehash in by downsizing bucket_table->size, only to find that my spot had been taken over by bucket_table->shift. These patches kill shift and makes me feel better :) v2 corrects the typo in the test_rhashtable changelog and also notes the min_shift parameter in the tipc patch changelog. ==================== Acked-by: NThomas Graf <tgraf@suug.ch> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Herbert Xu 提交于
Now that nobody uses max_shift and min_shift, we can safely remove them. Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Herbert Xu 提交于
This patch converts test_rhashtable to use rhashtable max_size instead of the obsolete max_shift. Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Herbert Xu 提交于
This patch converts tipc to use rhashtable max/min_size instead of the obsolete max/min_shift. Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Herbert Xu 提交于
This patch converts netlink to use rhashtable max_size instead of the obsolete max_shift. Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Herbert Xu 提交于
This patch adds the parameters max_size and min_size which are meant to replace max_shift and min_shift. Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Herbert Xu 提交于
Keeping both size and shift is silly. We only need one. Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 David S. Miller 提交于
Keyur Chudgar says: ==================== drivers: net: xgene: Add second SGMII based 1G interface This patch adds support for second SGMII based 1G interface. ==================== Signed-off-by: NKeyur Chudgar <kchudgar@apm.com> Signed-off-by: NIyappan Subramanian <isubramanian@apm.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Keyur Chudgar 提交于
- Added resource initialization based on port-id field - Enabled second SGMII 1G interface Signed-off-by: NKeyur Chudgar <kchudgar@apm.com> Signed-off-by: NIyappan Subramanian <isubramanian@apm.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Keyur Chudgar 提交于
- Added new SGMII node for port 1 - Added port-id field Signed-off-by: NKeyur Chudgar <kchudgar@apm.com> Signed-off-by: NIyappan Subramanian <isubramanian@apm.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Keyur Chudgar 提交于
Signed-off-by: NKeyur Chudgar <kchudgar@apm.com> Signed-off-by: NIyappan Subramanian <isubramanian@apm.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
- 18 3月, 2015 13 次提交
-
-
由 David S. Miller 提交于
Ying Xue says: ==================== tipc: fix netns refcnt leak The series aims to eliminate the issue of netns refcount leak. But during fixing it, another two additional problems are found. So all of known issues associated with the netns refcnt leak are resolved at the same time in the patchset. ==================== Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Ying Xue 提交于
The TIPC topology server is a per namespace service associated with the tipc name {1, 1}. When a namespace is deleted, that name must be withdrawn before we call sk_release_kernel because the kernel socket release is done in init_net and trying to withdraw a TIPC name published in another namespace will fail with an error as: [ 170.093264] Unable to remove local publication [ 170.093264] (type=1, lower=1, ref=2184244004, key=2184244005) We fix this by breaking the association between the topology server name and socket before calling sk_release_kernel. Signed-off-by: NYing Xue <ying.xue@windriver.com> Reviewed-by: NErik Hugne <erik.hugne@ericsson.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Ying Xue 提交于
[ 28.531768] ============================================= [ 28.532322] [ INFO: possible recursive locking detected ] [ 28.532322] 3.19.0+ #194 Not tainted [ 28.532322] --------------------------------------------- [ 28.532322] insmod/583 is trying to acquire lock: [ 28.532322] (&(&nseq->lock)->rlock){+.....}, at: [<ffffffffa000d219>] tipc_nametbl_remove_publ+0x49/0x2e0 [tipc] [ 28.532322] [ 28.532322] but task is already holding lock: [ 28.532322] (&(&nseq->lock)->rlock){+.....}, at: [<ffffffffa000e0dc>] tipc_nametbl_stop+0xfc/0x1f0 [tipc] [ 28.532322] [ 28.532322] other info that might help us debug this: [ 28.532322] Possible unsafe locking scenario: [ 28.532322] [ 28.532322] CPU0 [ 28.532322] ---- [ 28.532322] lock(&(&nseq->lock)->rlock); [ 28.532322] lock(&(&nseq->lock)->rlock); [ 28.532322] [ 28.532322] *** DEADLOCK *** [ 28.532322] [ 28.532322] May be due to missing lock nesting notation [ 28.532322] [ 28.532322] 3 locks held by insmod/583: [ 28.532322] #0: (net_mutex){+.+.+.}, at: [<ffffffff8163e30f>] register_pernet_subsys+0x1f/0x50 [ 28.532322] #1: (&(&tn->nametbl_lock)->rlock){+.....}, at: [<ffffffffa000e091>] tipc_nametbl_stop+0xb1/0x1f0 [tipc] [ 28.532322] #2: (&(&nseq->lock)->rlock){+.....}, at: [<ffffffffa000e0dc>] tipc_nametbl_stop+0xfc/0x1f0 [tipc] [ 28.532322] [ 28.532322] stack backtrace: [ 28.532322] CPU: 1 PID: 583 Comm: insmod Not tainted 3.19.0+ #194 [ 28.532322] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2007 [ 28.532322] ffffffff82394460 ffff8800144cb928 ffffffff81792f3e 0000000000000007 [ 28.532322] ffffffff82394460 ffff8800144cba28 ffffffff810a8080 ffff8800144cb998 [ 28.532322] ffffffff810a4df3 ffff880013e9cb10 ffffffff82b0d330 ffff880013e9cb38 [ 28.532322] Call Trace: [ 28.532322] [<ffffffff81792f3e>] dump_stack+0x4c/0x65 [ 28.532322] [<ffffffff810a8080>] __lock_acquire+0x740/0x1ca0 [ 28.532322] [<ffffffff810a4df3>] ? __bfs+0x23/0x270 [ 28.532322] [<ffffffff810a7506>] ? check_irq_usage+0x96/0xe0 [ 28.532322] [<ffffffff810a8a73>] ? __lock_acquire+0x1133/0x1ca0 [ 28.532322] [<ffffffffa000d219>] ? tipc_nametbl_remove_publ+0x49/0x2e0 [tipc] [ 28.532322] [<ffffffff810a9c0c>] lock_acquire+0x9c/0x140 [ 28.532322] [<ffffffffa000d219>] ? tipc_nametbl_remove_publ+0x49/0x2e0 [tipc] [ 28.532322] [<ffffffff8179c41f>] _raw_spin_lock_bh+0x3f/0x50 [ 28.532322] [<ffffffffa000d219>] ? tipc_nametbl_remove_publ+0x49/0x2e0 [tipc] [ 28.532322] [<ffffffffa000d219>] tipc_nametbl_remove_publ+0x49/0x2e0 [tipc] [ 28.532322] [<ffffffffa000e11e>] tipc_nametbl_stop+0x13e/0x1f0 [tipc] [ 28.532322] [<ffffffffa000dfe5>] ? tipc_nametbl_stop+0x5/0x1f0 [tipc] [ 28.532322] [<ffffffffa0004bab>] tipc_init_net+0x13b/0x150 [tipc] [ 28.532322] [<ffffffffa0004a75>] ? tipc_init_net+0x5/0x150 [tipc] [ 28.532322] [<ffffffff8163dece>] ops_init+0x4e/0x150 [ 28.532322] [<ffffffff810aa66d>] ? trace_hardirqs_on+0xd/0x10 [ 28.532322] [<ffffffff8163e1d3>] register_pernet_operations+0xf3/0x190 [ 28.532322] [<ffffffff8163e31e>] register_pernet_subsys+0x2e/0x50 [ 28.532322] [<ffffffffa002406a>] tipc_init+0x6a/0x1000 [tipc] [ 28.532322] [<ffffffffa0024000>] ? 0xffffffffa0024000 [ 28.532322] [<ffffffff810002d9>] do_one_initcall+0x89/0x1c0 [ 28.532322] [<ffffffff811b7cb0>] ? kmem_cache_alloc_trace+0x50/0x1b0 [ 28.532322] [<ffffffff810e725b>] ? do_init_module+0x2b/0x200 [ 28.532322] [<ffffffff810e7294>] do_init_module+0x64/0x200 [ 28.532322] [<ffffffff810e9353>] load_module+0x12f3/0x18e0 [ 28.532322] [<ffffffff810e5890>] ? show_initstate+0x50/0x50 [ 28.532322] [<ffffffff810e9a19>] SyS_init_module+0xd9/0x110 [ 28.532322] [<ffffffff8179f3b3>] sysenter_dispatch+0x7/0x1f Before tipc_purge_publications() calls tipc_nametbl_remove_publ() to remove a publication with a name sequence, the name sequence's lock is held. However, when tipc_nametbl_remove_publ() calling tipc_nameseq_remove_publ() to remove the publication, it first tries to query name sequence instance with the publication, and then holds the lock of the found name sequence. But as the lock may be already taken in tipc_purge_publications(), deadlock happens like above scenario demonstrated. As tipc_nameseq_remove_publ() doesn't grab name sequence's lock, the deadlock can be avoided if it's directly invoked by tipc_purge_publications(). Fixes: 97ede29e ("tipc: convert name table read-write lock to RCU") Signed-off-by: NYing Xue <ying.xue@windriver.com> Reviewed-by: NErik Hugne <erik.hugne@ericsson.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Ying Xue 提交于
When the TIPC module is loaded, we launch a topology server in kernel space, which in its turn is creating TIPC sockets for communication with topology server users. Because both the socket's creator and provider reside in the same module, it is necessary that the TIPC module's reference count remains zero after the server is started and the socket created; otherwise it becomes impossible to perform "rmmod" even on an idle module. Currently, we achieve this by defining a separate "tipc_proto_kern" protocol struct, that is used only for kernel space socket allocations. This structure has the "owner" field set to NULL, which restricts the module reference count from being be bumped when sk_alloc() for local sockets is called. Furthermore, we have defined three kernel-specific functions, tipc_sock_create_local(), tipc_sock_release_local() and tipc_sock_accept_local(), to avoid the module counter being modified when module local sockets are created or deleted. This has worked well until we introduced name space support. However, after name space support was introduced, we have observed that a reference count leak occurs, because the netns counter is not decremented in tipc_sock_delete_local(). This commit remedies this problem. But instead of just modifying tipc_sock_delete_local(), we eliminate the whole parallel socket handling infrastructure, and start using the regular sk_create_kern(), kernel_accept() and sk_release_kernel() calls. Since those functions manipulate the module counter, we must now compensate for that by explicitly decrementing the counter after module local sockets are created, and increment it just before calling sk_release_kernel(). Fixes: a62fbcce ("tipc: make subscriber server support net namespace") Signed-off-by: NYing Xue <ying.xue@windriver.com> Reviewed-by: NJon Maloy <jon.maloy@ericson.com> Reviewed-by: NErik Hugne <erik.hugne@ericsson.com> Reported-by: NCong Wang <cwang@twopensource.com> Tested-by: NErik Hugne <erik.hugne@ericsson.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 David S. Miller 提交于
Eric Dumazet says: ==================== inet: tcp listener refactoring, part 12 By adding a pointer back to listener, we are preparing synack rtx handling to no longer be governed by listener keepalive timer, as this is the most problematic source of contention on listener spinlock. Note that TCP FastOpen had such pointer anyway, so we make it generic. ==================== Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
While testing last patch series, I found req sock refcounting was wrong. We must set skc_refcnt to 1 for all request socks added in hashes, but also on request sockets created by FastOpen or syncookies. It is tricky because we need to defer this initialization so that future RCU lookups do not try to take a refcount on a not yet fully initialized request socket. Also get rid of ireq_refcnt alias. Signed-off-by: NEric Dumazet <edumazet@google.com> Fixes: 13854e5a ("inet: add proper refcounting to request sock") Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
It is not because a TCP listener is FastOpen ready that all incoming sockets actually used FastOpen. Avoid taking queue->fastopenq->lock if not needed. Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
The listener field in struct tcp_request_sock is a pointer back to the listener. We now have req->rsk_listener, so TCP only needs one boolean and not a full pointer. Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
Once we'll be able to lookup request sockets in ehash table, we'll need to get access to listener which created this request. This avoid doing a lookup to find the listener, which benefits for a more solid SO_REUSEPORT, and is needed once we no longer queue request sock into a listener private queue. Note that 'struct tcp_request_sock'->listener could be reduced to a single bit, as TFO listener should match req->rsk_listener. TFO will no longer need to hold a reference on the listener. Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
inet_reqsk_alloc() is becoming fat and should not be inlined. Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
listener socket can be used to set net pointer, and will be later used to hold a reference on listener. Add a const qualifier to first argument (struct request_sock_ops *), and factorize all write_pnet(&ireq->ireq_net, sock_net(sk)); Signed-off-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 David S. Miller 提交于
Eric Dumazet says: ==================== inet: tcp listener refactoring, part 11 Before inserting request sockets into general (ehash) table, we need to prepare netfilter to cope with them, as they are not full sockets. I'll later change xt_socket to get full support, including for request sockets (NEW_SYN_RECV) Save 8 bytes in inet_request_sock on 64bit arches. We'll soon add a pointer to the listener socket. I included two TCP changes in this patch series. ==================== Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-
由 Eric Dumazet 提交于
tcp_oow_rate_limited() is hardly used in fast path, there is no point inlining it. Signed-of-by: NEric Dumazet <edumazet@google.com> Signed-off-by: NDavid S. Miller <davem@davemloft.net>
-