We'll need this to handle ICMP errors for tunnels without a sending socket
(i.e. FoU and GUE). There, we might have to look up different types of IP
tunnels, registered as network protocols, before we get a match, so we
want this for the error handlers of IPPROTO_IPIP and IPPROTO_IPV6 in both
inet_protos and inet6_protos. These error codes will be used in the next
patch.
For consistency, return sensible error codes in protocol error handlers
whenever handlers can't handle errors because, even if valid, they don't
match a protocol or any of its states.
This has no effect on existing error handling paths.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
In some scenarios, the GRO engine can assemble an UDP GRO packet
that ultimately lands on a non GRO-enabled socket.
This patch tries to address the issue explicitly checking for the UDP
socket features before enqueuing the packet, and eventually segmenting
the unexpected GRO packet, as needed.
We must also cope with re-insertion requests: after segmentation the
UDP code calls the helper introduced by the previous patches, as needed.
Segmentation is performed by a common helper, which takes care of
updating socket and protocol stats is case of failure.
rfc v3 -> v1
- fix compile issues with rxrpc
- when gso_segment returns NULL, treat is as an error
- added 'ipv4' argument to udp_rcv_segment()
rfc v2 -> rfc v3
- moved udp_rcv_segment() into net/udp.h, account errors to socket
and ns, always return NULL or segs list
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ensure an unbound datagram skt is chosen when not in a VRF. The check
for a device match in compute_score() for UDP must be performed when
there is no device match. For this, a failure is returned when there is
no device match. This ensures that bound sockets are never selected,
even if there is no unbound socket.
Allow IPv6 packets to be sent over a datagram skt bound to a VRF. These
packets are currently blocked, as flowi6_oif was set to that of the
master vrf device, and the ipi6_ifindex is that of the slave device.
Allow these packets to be sent by checking the device with ipi6_ifindex
has the same L3 scope as that of the bound device of the skt, which is
the master vrf device. Note that this check always succeeds if the skt
is unbound.
Even though the right datagram skt is now selected by compute_score(),
a different skt is being returned that is bound to the wrong vrf. The
difference between these and stream sockets is the handling of the skt
option for SO_REUSEPORT. While the handling when adding a skt for reuse
correctly checks that the bound device of the skt is a match, the skts
in the hashslot are already incorrect. So for the same hash, a skt for
the wrong vrf may be selected for the required port. The root cause is
that the skt is immediately placed into a slot when it is created,
but when the skt is then bound using SO_BINDTODEVICE, it remains in the
same slot. The solution is to move the skt to the correct slot by
forcing a rehash.
Signed-off-by: Mike Manning <mmanning@vyatta.att-mail.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Tested-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Avoid the socket lookup cost in udp_gro_receive if no socket has a
udp tunnel callback configured.
udp_sk(sk)->gro_receive requires a registration with
setup_udp_tunnel_sock, which enables the static key.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Simple overlapping changes in stmmac driver.
Adjust skb_gro_flush_final_remcsum function signature to make GRO list
changes in net-next, as per Stephen Rothwell's example merge
resolution.
Signed-off-by: David S. Miller <davem@davemloft.net>
The poll() changes were not well thought out, and completely
unexplained. They also caused a huge performance regression, because
"->poll()" was no longer a trivial file operation that just called down
to the underlying file operations, but instead did at least two indirect
calls.
Indirect calls are sadly slow now with the Spectre mitigation, but the
performance problem could at least be largely mitigated by changing the
"->get_poll_head()" operation to just have a per-file-descriptor pointer
to the poll head instead. That gets rid of one of the new indirections.
But that doesn't fix the new complexity that is completely unwarranted
for the regular case. The (undocumented) reason for the poll() changes
was some alleged AIO poll race fixing, but we don't make the common case
slower and more complex for some uncommon special case, so this all
really needs way more explanations and most likely a fundamental
redesign.
[ This revert is a revert of about 30 different commits, not reverted
individually because that would just be unnecessarily messy - Linus ]
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Manage pending per-NAPI GRO packets via list_head.
Return an SKB pointer from the GRO receive handlers. When GRO receive
handlers return non-NULL, it means that this SKB needs to be completed
at this time and removed from the NAPI queue.
Several operations are greatly simplified by this transformation,
especially timing out the oldest SKB in the list when gro_count
exceeds MAX_GRO_SKBS, and napi_gro_flush() which walks the queue
in reverse order.
Signed-off-by: David S. Miller <davem@davemloft.net>
After commit 6b229cf77d ("udp: add batching to udp_rmem_release()")
the sk_rmem_alloc field does not measure exactly anymore the
receive queue length, because we batch the rmem release. The issue
is really apparent only after commit 0d4a6608f6 ("udp: do rmem bulk
free even if the rx sk queue is empty"): the user space can easily
check for an empty socket with not-0 queue length reported by the 'ss'
tool or the procfs interface.
We need to use a custom UDP helper to report the correct queue length,
taking into account the forward allocation deficit.
Reported-by: trevor.francis@46labs.com
Fixes: 6b229cf77d ("UDP: add batching to udp_rmem_release()")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull networking updates from David Miller:
1) Add Maglev hashing scheduler to IPVS, from Inju Song.
2) Lots of new TC subsystem tests from Roman Mashak.
3) Add TCP zero copy receive and fix delayed acks and autotuning with
SO_RCVLOWAT, from Eric Dumazet.
4) Add XDP_REDIRECT support to mlx5 driver, from Jesper Dangaard
Brouer.
5) Add ttl inherit support to vxlan, from Hangbin Liu.
6) Properly separate ipv6 routes into their logically independant
components. fib6_info for the routing table, and fib6_nh for sets of
nexthops, which thus can be shared. From David Ahern.
7) Add bpf_xdp_adjust_tail helper, which can be used to generate ICMP
messages from XDP programs. From Nikita V. Shirokov.
8) Lots of long overdue cleanups to the r8169 driver, from Heiner
Kallweit.
9) Add BTF ("BPF Type Format"), from Martin KaFai Lau.
10) Add traffic condition monitoring to iwlwifi, from Luca Coelho.
11) Plumb extack down into fib_rules, from Roopa Prabhu.
12) Add Flower classifier offload support to igb, from Vinicius Costa
Gomes.
13) Add UDP GSO support, from Willem de Bruijn.
14) Add documentation for eBPF helpers, from Quentin Monnet.
15) Add TLS tx offload to mlx5, from Ilya Lesokhin.
16) Allow applications to be given the number of bytes available to read
on a socket via a control message returned from recvmsg(), from
Soheil Hassas Yeganeh.
17) Add x86_32 eBPF JIT compiler, from Wang YanQing.
18) Add AF_XDP sockets, with zerocopy support infrastructure as well.
From Björn Töpel.
19) Remove indirect load support from all of the BPF JITs and handle
these operations in the verifier by translating them into native BPF
instead. From Daniel Borkmann.
20) Add GRO support to ipv6 gre tunnels, from Eran Ben Elisha.
21) Allow XDP programs to do lookups in the main kernel routing tables
for forwarding. From David Ahern.
22) Allow drivers to store hardware state into an ELF section of kernel
dump vmcore files, and use it in cxgb4. From Rahul Lakkireddy.
23) Various RACK and loss detection improvements in TCP, from Yuchung
Cheng.
24) Add TCP SACK compression, from Eric Dumazet.
25) Add User Mode Helper support and basic bpfilter infrastructure, from
Alexei Starovoitov.
26) Support ports and protocol values in RTM_GETROUTE, from Roopa
Prabhu.
27) Support bulking in ->ndo_xdp_xmit() API, from Jesper Dangaard
Brouer.
28) Add lots of forwarding selftests, from Petr Machata.
29) Add generic network device failover driver, from Sridhar Samudrala.
* ra.kernel.org:/pub/scm/linux/kernel/git/davem/net-next: (1959 commits)
strparser: Add __strp_unpause and use it in ktls.
rxrpc: Fix terminal retransmission connection ID to include the channel
net: hns3: Optimize PF CMDQ interrupt switching process
net: hns3: Fix for VF mailbox receiving unknown message
net: hns3: Fix for VF mailbox cannot receiving PF response
bnx2x: use the right constant
Revert "net: sched: cls: Fix offloading when ingress dev is vxlan"
net: dsa: b53: Fix for brcm tag issue in Cygnus SoC
enic: fix UDP rss bits
netdev-FAQ: clarify DaveM's position for stable backports
rtnetlink: validate attributes in do_setlink()
mlxsw: Add extack messages for port_{un, }split failures
netdevsim: Add extack error message for devlink reload
devlink: Add extack to reload and port_{un, }split operations
net: metrics: add proper netlink validation
ipmr: fix error path when ipmr_new_table fails
ip6mr: only set ip6mr_table from setsockopt when ip6mr_new_table succeeds
net: hns3: remove unused hclgevf_cfg_func_mta_filter
netfilter: provide udp*_lib_lookup for nf_tproxy
qed*: Utilize FW 8.37.2.0
...
Pull aio updates from Al Viro:
"Majority of AIO stuff this cycle. aio-fsync and aio-poll, mostly.
The only thing I'm holding back for a day or so is Adam's aio ioprio -
his last-minute fixup is trivial (missing stub in !CONFIG_BLOCK case),
but let it sit in -next for decency sake..."
* 'work.aio-1' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (46 commits)
aio: sanitize the limit checking in io_submit(2)
aio: fold do_io_submit() into callers
aio: shift copyin of iocb into io_submit_one()
aio_read_events_ring(): make a bit more readable
aio: all callers of aio_{read,write,fsync,poll} treat 0 and -EIOCBQUEUED the same way
aio: take list removal to (some) callers of aio_complete()
aio: add missing break for the IOCB_CMD_FDSYNC case
random: convert to ->poll_mask
timerfd: convert to ->poll_mask
eventfd: switch to ->poll_mask
pipe: convert to ->poll_mask
crypto: af_alg: convert to ->poll_mask
net/rxrpc: convert to ->poll_mask
net/iucv: convert to ->poll_mask
net/phonet: convert to ->poll_mask
net/nfc: convert to ->poll_mask
net/caif: convert to ->poll_mask
net/bluetooth: convert to ->poll_mask
net/sctp: convert to ->poll_mask
net/tipc: convert to ->poll_mask
...
Variants of proc_create{,_data} that directly take a struct seq_operations
and deal with network namespaces in ->open and ->release. All callers of
proc_create + seq_open_net converted over, and seq_{open,release}_net are
removed entirely.
Signed-off-by: Christoph Hellwig <hch@lst.de>
This patch is meant to allow us to avoid having to recompute the checksum
from scratch and have it passed as a parameter.
Instead of taking that approach we can take advantage of the fact that the
length that was used to compute the existing checksum is included in the
UDP header.
Finally to avoid the need to invert the result we can just call csum16_add
and csum16_sub directly. By doing this we can avoid a number of
instructions in the loop that is handling segmentation.
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is no point in passing MSS as a parameter for for the GSO
segmentation call as it is already available via the shared info for the
skb itself.
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow specifying segment size in the send call.
The new control message performs the same function as socket option
UDP_SEGMENT while avoiding the extra system call.
[ Export udp_cmsg_send for ipv6. -DaveM ]
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement generic segmentation offload support for udp datagrams. A
follow-up patch adds support to the protocol stack to generate such
packets.
UDP GSO is not UFO. UFO fragments a single large datagram. GSO splits
a large payload into a number of discrete UDP datagrams.
The implementation adds a GSO type SKB_UDP_GSO_L4 to differentiate it
from UFO (SKB_UDP_GSO).
IPPROTO_UDPLITE is excluded, as that protocol has no gso handler
registered.
[ Export __udp_gso_segment for ipv6. -DaveM ]
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
== The problem ==
See description of the problem in the initial patch of this patch set.
== The solution ==
The patch provides much more reliable in-kernel solution for the 2nd
part of the problem: making outgoing connecttion from desired IP.
It adds new attach types `BPF_CGROUP_INET4_CONNECT` and
`BPF_CGROUP_INET6_CONNECT` for program type
`BPF_PROG_TYPE_CGROUP_SOCK_ADDR` that can be used to override both
source and destination of a connection at connect(2) time.
Local end of connection can be bound to desired IP using newly
introduced BPF-helper `bpf_bind()`. It allows to bind to only IP though,
and doesn't support binding to port, i.e. leverages
`IP_BIND_ADDRESS_NO_PORT` socket option. There are two reasons for this:
* looking for a free port is expensive and can affect performance
significantly;
* there is no use-case for port.
As for remote end (`struct sockaddr *` passed by user), both parts of it
can be overridden, remote IP and remote port. It's useful if an
application inside cgroup wants to connect to another application inside
same cgroup or to itself, but knows nothing about IP assigned to the
cgroup.
Support is added for IPv4 and IPv6, for TCP and UDP.
IPv4 and IPv6 have separate attach types for same reason as sys_bind
hooks, i.e. to prevent reading from / writing to e.g. user_ip6 fields
when user passes sockaddr_in since it'd be out-of-bound.
== Implementation notes ==
The patch introduces new field in `struct proto`: `pre_connect` that is
a pointer to a function with same signature as `connect` but is called
before it. The reason is in some cases BPF hooks should be called way
before control is passed to `sk->sk_prot->connect`. Specifically
`inet_dgram_connect` autobinds socket before calling
`sk->sk_prot->connect` and there is no way to call `bpf_bind()` from
hooks from e.g. `ip4_datagram_connect` or `ip6_datagram_connect` since
it'd cause double-bind. On the other hand `proto.pre_connect` provides a
flexible way to add BPF hooks for connect only for necessary `proto` and
call them at desired time before `connect`. Since `bpf_bind()` is
allowed to bind only to IP and autobind in `inet_dgram_connect` binds
only port there is no chance of double-bind.
bpf_bind() sets `force_bind_address_no_port` to bind to only IP despite
of value of `bind_address_no_port` socket field.
bpf_bind() sets `with_lock` to `false` when calling to __inet_bind()
and __inet6_bind() since all call-sites, where bpf_bind() is called,
already hold socket lock.
Signed-off-by: Andrey Ignatov <rdna@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Currently no error is emitted, but this infrastructure will
used by the next patch to allow source address validation
for mcast sockets.
Since early demux can do a route lookup and an ipv4 route
lookup can return an error code this is consistent with the
current ipv4 route infrastructure.
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, in the udp6 code, the dst cookie is not initialized/updated
concurrently with the RX dst used by early demux.
As a result, the dst_check() in the early_demux path always fails,
the rx dst cache is always invalidated, and we can't really
leverage significant gain from the demux lookup.
Fix it adding udp6 specific variant of sk_rx_dst_set() and use it
to set the dst cookie when the dst entry is really changed.
The issue is there since the introduction of early demux for ipv6.
Fixes: 5425077d73 ("net: ipv6: Add early demux handler for UDP unicast")
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
copy_linear_skb() is broken; both of its callers actually
expect 'len' to be the amount we are trying to copy,
not the offset of the end.
Fix it keeping the meanings of arguments in sync with what the
callers (both of them) expect.
Also restore a saner behavior on EFAULT (i.e. preserving
the iov_iter position in case of failure):
The commit fd851ba9ca ("udp: harden copy_linear_skb()")
avoids the more destructive effect of the buggy
copy_linear_skb(), e.g. no more invalid memory access, but
said function still behaves incorrectly: when peeking with
offset it can fail with EINVAL instead of copying the
appropriate amount of memory.
Reported-by: Sasha Levin <alexander.levin@verizon.com>
Fixes: b65ac44674 ("udp: try to avoid 2 cache miss on dequeue")
Fixes: fd851ba9ca ("udp: harden copy_linear_skb()")
Signed-off-by: Al Viro <viro@ZenIV.linux.org.uk>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Tested-by: Sasha Levin <alexander.levin@verizon.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a second device index, sdif, to udp socket lookups. sdif is the
index for ingress devices enslaved to an l3mdev. It allows the lookups
to consider the enslaved device as well as the L3 domain when searching
for a socket.
Early demux lookups are handled in the next patch as part of INET_MATCH
changes.
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a second device index, sdif, to udp socket lookups. sdif is the
index for ingress devices enslaved to an l3mdev. It allows the lookups
to consider the enslaved device as well as the L3 domain when searching
for a socket.
Early demux lookups are handled in the next patch as part of INET_MATCH
changes.
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When an early demuxed packet reaches __udp6_lib_lookup_skb(), the
sk reference is retrieved and used, but the relevant reference
count is leaked and the socket destructor is never called.
Beyond leaking the sk memory, if there are pending UDP packets
in the receive queue, even the related accounted memory is leaked.
In the long run, this will cause persistent forward allocation errors
and no UDP skbs (both ipv4 and ipv6) will be able to reach the
user-space.
Fix this by explicitly accessing the early demux reference before
the lookup, and properly decreasing the socket reference count
after usage.
Also drop the skb_steal_sock() in __udp6_lib_lookup_skb(), and
the now obsoleted comment about "socket cache".
The newly added code is derived from the current ipv4 code for the
similar path.
v1 -> v2:
fixed the __udp6_lib_rcv() return code for resubmission,
as suggested by Eric
Reported-by: Sam Edwards <CFSworks@gmail.com>
Reported-by: Marc Haber <mh+netdev@zugschlus.de>
Fixes: 5425077d73 ("net: ipv6: Add early demux handler for UDP unicast")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Paul Moore reported a SELinux/IP_PASSSEC regression
caused by missing skb->sp at recvmsg() time. We need to
preserve the skb head state to process the IP_CMSG_PASSSEC
cmsg.
With this commit we avoid releasing the skb head state in the
BH even if a secpath is attached to the current skb, and stores
the skb status (with/without head states) in the scratch area,
so that we can access it at skb deallocation time, without
incurring in cache-miss penalties.
This also avoids misusing the skb CB for ipv6 packets,
as introduced by the commit 0ddf3fb2c4 ("udp: preserve
skb->dst if required for IP options processing").
Clean a bit the scratch area helpers implementation, to
reduce the code differences between 32 and 64 bits build.
Reported-by: Paul Moore <paul@paul-moore.com>
Fixes: 0a463c78d2 ("udp: avoid a cache miss on dequeue")
Fixes: 0ddf3fb2c4 ("udp: preserve skb->dst if required for IP options processing")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Tested-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
So that they can be later used by the IPv6 code, too.
Also lift the comments a bit.
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
under udp flood the sk_receive_queue spinlock is heavily contended.
This patch try to reduce the contention on such lock adding a
second receive queue to the udp sockets; recvmsg() looks first
in such queue and, only if empty, tries to fetch the data from
sk_receive_queue. The latter is spliced into the newly added
queue every time the receive path has to acquire the
sk_receive_queue lock.
The accounting of forward allocated memory is still protected with
the sk_receive_queue lock, so udp_rmem_release() needs to acquire
both locks when the forward deficit is flushed.
On specific scenarios we can end up acquiring and releasing the
sk_receive_queue lock multiple times; that will be covered by
the next patch
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Certain system process significant unconnected UDP workload.
It would be preferrable to disable UDP early demux for those systems
and enable it for TCP only.
By disabling UDP demux, we see these slight gains on an ARM64 system-
782 -> 788Mbps unconnected single stream UDPv4
633 -> 654Mbps unconnected UDPv4 different sources
The performance impact can change based on CPU architecure and cache
sizes. There will not much difference seen if entire UDP hash table
is in cache.
Both sysctls are enabled by default to preserve existing behavior.
v1->v2: Change function pointer instead of adding conditional as
suggested by Stephen.
v2->v3: Read once in callers to avoid issues due to compiler
optimizations. Also update commit message with the tests.
v3->v4: Store and use read once result instead of querying pointer
again incorrectly.
v4->v5: Refactor to avoid errors due to compilation with IPV6={m,n}
Signed-off-by: Subash Abhinov Kasiviswanathan <subashab@codeaurora.org>
Suggested-by: Eric Dumazet <edumazet@google.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: Tom Herbert <tom@herbertland.com>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
We pass these per-protocol equal functions around in various places, but
we can just have one function that checks the sk->sk_family and then do
the right comparison function. I've also changed the ipv4 version to
not cast to inet_sock since it is unneeded.
Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The commit 850cbaddb5 ("udp: use it's own memory accounting schema")
assumes that the socket proto has memory accounting enabled,
but this is not the case for UDPLITE.
Fix it enabling memory accounting for UDPLITE and performing
fwd allocated memory reclaiming on socket shutdown.
UDP and UDPLITE share now the same memory accounting limits.
Also drop the backlog receive operation, since is no more needed.
Fixes: 850cbaddb5 ("udp: use it's own memory accounting schema")
Reported-by: Andrei Vagin <avagin@gmail.com>
Suggested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A new argument is added to __skb_recv_datagram to provide
an explicit skb destructor, invoked under the receive queue
lock.
The UDP protocol uses such argument to perform memory
reclaiming on dequeue, so that the UDP protocol does not
set anymore skb->desctructor.
Instead explicit memory reclaiming is performed at close() time and
when skbs are removed from the receive queue.
The in kernel UDP protocol users now need to call a
skb_recv_udp() variant instead of skb_recv_datagram() to
properly perform memory accounting on dequeue.
Overall, this allows acquiring only once the receive queue
lock on dequeue.
Tested using pktgen with random src port, 64 bytes packet,
wire-speed on a 10G link as sender and udp_sink as the receiver,
using an l4 tuple rxhash to stress the contention, and one or more
udp_sink instances with reuseport.
nr sinks vanilla patched
1 440 560
3 2150 2300
6 3650 3800
9 4450 4600
12 6250 6450
v1 -> v2:
- do rmem and allocated memory scheduling under the receive lock
- do bulk scheduling in first_packet_length() and in udp_destruct_sock()
- avoid the typdef for the dequeue callback
Suggested-by: Eric Dumazet <edumazet@google.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Mostly simple overlapping changes.
For example, David Ahern's adjacency list revamp in 'net-next'
conflicted with an adjacency list traversal bug fix in 'net'.
Signed-off-by: David S. Miller <davem@davemloft.net>
Avoid using the generic helpers.
Use the receive queue spin lock to protect the memory
accounting operation, both on enqueue and on dequeue.
On dequeue perform partial memory reclaiming, trying to
leave a quantum of forward allocated memory.
On enqueue use a custom helper, to allow some optimizations:
- use a plain spin_lock() variant instead of the slightly
costly spin_lock_irqsave(),
- avoid dst_force check, since the calling code has already
dropped the skb dst
- avoid orphaning the skb, since skb_steal_sock() already did
the work for us
The above needs custom memory reclaiming on shutdown, provided
by the udp_destruct_sock().
v5 -> v6:
- don't orphan the skb on enqueue
v4 -> v5:
- replace the mem_lock with the receive queue spin lock
- ensure that the bh is always allowed to enqueue at least
a skb, even if sk_rcvbuf is exceeded
v3 -> v4:
- reworked memory accunting, simplifying the schema
- provide an helper for both memory scheduling and enqueuing
v1 -> v2:
- use a udp specific destrctor to perform memory reclaiming
- remove a couple of helpers, unneeded after the above cleanup
- do not reclaim memory on dequeue if not under memory
pressure
- reworked the fwd accounting schema to avoid potential
integer overflow
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Baozeng Ding reported KASAN traces showing uses after free in
udp_lib_get_port() and other related UDP functions.
A CONFIG_DEBUG_PAGEALLOC=y kernel would eventually crash.
I could write a reproducer with two threads doing :
static int sock_fd;
static void *thr1(void *arg)
{
for (;;) {
connect(sock_fd, (const struct sockaddr *)arg,
sizeof(struct sockaddr_in));
}
}
static void *thr2(void *arg)
{
struct sockaddr_in unspec;
for (;;) {
memset(&unspec, 0, sizeof(unspec));
connect(sock_fd, (const struct sockaddr *)&unspec,
sizeof(unspec));
}
}
Problem is that udp_disconnect() could run without holding socket lock,
and this was causing list corruptions.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Baozeng Ding <sploving1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This implements SOCK_DESTROY for UDP sockets similar to what was done
for TCP with commit c1e64e298b ("net: diag: Support destroying TCP
sockets.") A process with a UDP socket targeted for destroy is awakened
and recvmsg fails with ECONNABORTED.
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In commit e6afc8ace6 ("udp: remove headers from UDP packets before
queueing"), udp_csum_pull_header() helper was added but missed fact
that CHECKSUM_UNNECESSARY packets were now converted to CHECKSUM_NONE
and skb->csum_valid was set to 1 for them.
Since csum_partial() is quite expensive, even for 8-byte area, it is
worth adding a test.
We also can use skb->data instead of udp_hdr() as we are pulling
UDP headers, as it is sightly faster.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is nothing related to BH in SNMP counters anymore,
since linux-3.0.
Rename helpers to use __ prefix instead of _BH prefix,
for contexts where preemption is disabled.
This more closely matches convention used to update
percpu variables.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rename UDP_INC_STATS_BH() to __UDP_INC_STATS(),
and UDP6_INC_STATS_BH() to __UDP6_INC_STATS()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the old days (before linux-3.0), SNMP counters were duplicated,
one for user context, and one for BH context.
After commit 8f0ea0fe3a ("snmp: reduce percpu needs by 50%")
we have a single copy, and what really matters is preemption being
enabled or disabled, since we use this_cpu_inc() or __this_cpu_inc()
respectively.
We therefore kill SNMP_INC_STATS_USER(), SNMP_ADD_STATS_USER(),
NET_INC_STATS_USER(), NET_ADD_STATS_USER(), SCTP_INC_STATS_USER(),
SNMP_INC_STATS64_USER(), SNMP_ADD_STATS64_USER(), TCP_ADD_STATS_USER(),
UDP_INC_STATS_USER(), UDP6_INC_STATS_USER(), and XFRM_INC_STATS_USER()
Following patches will rename __BH helpers to make clear their
usage is not tied to BH being disabled.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds GRO functions (gro_receive and gro_complete) to UDP
sockets. udp_gro_receive is changed to perform socket lookup on a
packet. If a socket is found the related GRO functions are called.
This features obsoletes using UDP offload infrastructure for GRO
(udp_offload). This has the advantage of not being limited to provide
offload on a per port basis, GRO is now applied to whatever individual
UDP sockets are bound to. This also allows the possbility of
"application defined GRO"-- that is we can attach something like
a BPF program to a UDP socket to perfrom GRO on an application
layer protocol.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add externally visible functions to lookup a UDP socket by skb. This
will be used for GRO in UDP sockets. These functions also check
if skb->dst is set, and if it is not skb->dev is used to get dev_net.
This allows calling lookup functions before dst has been set on the
skbuff.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove UDP transport headers before queueing packets for reception.
This change simplifies a follow-up patch to add MSG_PEEK support.
Signed-off-by: Sam Kumar <samanthakumar@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tom Herbert would like not touching UDP socket refcnt for encapsulated
traffic. For this to happen, we need to use normal RCU rules, with a grace
period before freeing a socket. UDP sockets are not short lived in the
high usage case, so the added cost of call_rcu() should not be a concern.
This actually removes a lot of complexity in UDP stack.
Multicast receives no longer need to hold a bucket spinlock.
Note that ip early demux still needs to take a reference on the socket.
Same remark for functions used by xt_socket and xt_PROXY netfilter modules,
but this might be changed later.
Performance for a single UDP socket receiving flood traffic from
many RX queues/cpus.
Simple udp_rx using simple recvfrom() loop :
438 kpps instead of 374 kpps : 17 % increase of the peak rate.
v2: Addressed Willem de Bruijn feedback in multicast handling
- keep early demux break in __udp4_lib_demux_lookup()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Tom Herbert <tom@herbertland.com>
Cc: Willem de Bruijn <willemb@google.com>
Tested-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to support fast reuseport lookups in TCP, the hash function
defined in struct proto must be capable of returning an error code.
This patch changes the function signature of all related hash functions
to return an integer and handles or propagates this return value at
all call sites.
Signed-off-by: Craig Gallek <kraig@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Expose socket options for setting a classic or extended BPF program
for use when selecting sockets in an SO_REUSEPORT group. These options
can be used on the first socket to belong to a group before bind or
on any socket in the group after bind.
This change includes refactoring of the existing sk_filter code to
allow reuse of the existing BPF filter validation checks.
Signed-off-by: Craig Gallek <kraig@google.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Include a struct sock_reuseport instance when a UDP socket binds to
a specific address for the first time with the reuseport flag set.
When selecting a socket for an incoming UDP packet, use the information
available in sock_reuseport if present.
This required adding an additional field to the UDP source address
equality function to differentiate between exact and wildcard matches.
The original use case allowed wildcard matches when checking for
existing port uses during bind. The new use case of adding a socket
to a reuseport group requires exact address matching.
Performance test (using a machine with 2 CPU sockets and a total of
48 cores): Create reuseport groups of varying size. Use one socket
from this group per user thread (pinning each thread to a different
core) calling recvmmsg in a tight loop. Record number of messages
received per second while saturating a 10G link.
10 sockets: 18% increase (~2.8M -> 3.3M pkts/s)
20 sockets: 14% increase (~2.9M -> 3.3M pkts/s)
40 sockets: 13% increase (~3.0M -> 3.4M pkts/s)
This work is based off a similar implementation written by
Ying Cai <ycai@google.com> for implementing policy-based reuseport
selection.
Signed-off-by: Craig Gallek <kraig@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
