UDP reuseport groups can hold a mix unconnected and connected sockets.
Ensure that connections only receive all traffic to their 4-tuple.
Fast reuseport returns on the first reuseport match on the assumption
that all matches are equal. Only if connections are present, return to
the previous behavior of scoring all sockets.
Record if connections are present and if so (1) treat such connected
sockets as an independent match from the group, (2) only return
2-tuple matches from reuseport and (3) do not return on the first
2-tuple reuseport match to allow for a higher scoring match later.
New field has_conns is set without locks. No other fields in the
bitmap are modified at runtime and the field is only ever set
unconditionally, so an RMW cannot miss a change.
Fixes: e32ea7e747 ("soreuseport: fast reuseport UDP socket selection")
Link: http://lkml.kernel.org/r/CA+FuTSfRP09aJNYRt04SS6qj22ViiOEWaWmLAwX0psk8-PGNxw@mail.gmail.com
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: Craig Gallek <kraig@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Enable setting skb->mark for UDP and RAW sockets using cmsg.
This is analogous to existing support for TOS, TTL, txtime, etc.
Packet sockets already support this as of commit c7d39e3263
("packet: support per-packet fwmark for af_packet sendmsg").
Similar to other fields, implement by
1. initialize the sockcm_cookie.mark from socket option sk_mark
2. optionally overwrite this in ip_cmsg_send/ip6_datagram_send_ctl
3. initialize inet_cork.mark from sockcm_cookie.mark
4. initialize each (usually just one) skb->mark from inet_cork.mark
Step 1 is handled in one location for most protocols by ipcm_init_sk
as of commit 351782067b ("ipv4: ipcm_cookie initializers").
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
skb_warn_bad_offload and netdev_rx_csum_fault trigger on hard to debug
issues. Dump more state and the header.
Optionally dump the entire packet and linear segment. This is required
to debug checksum bugs that may include bytes past skb_tail_pointer().
Both call sites call this function inside a net_ratelimit() block.
Limit full packet log further to a hard limit of can_dump_full (5).
Based on an earlier patch by Cong Wang, see link below.
Changes v1 -> v2
- dump frag_list only on full_pkt
Link: https://patchwork.ozlabs.org/patch/1000841/
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Processes can request ipv6 flowlabels with cmsg IPV6_FLOWINFO.
If not set, by default an autogenerated flowlabel is selected.
Explicit flowlabels require a control operation per label plus a
datapath check on every connection (every datagram if unconnected).
This is particularly expensive on unconnected sockets multiplexing
many flows, such as QUIC.
In the common case, where no lease is exclusive, the check can be
safely elided, as both lease request and check trivially succeed.
Indeed, autoflowlabel does the same even with exclusive leases.
Elide the check if no process has requested an exclusive lease.
fl6_sock_lookup previously returns either a reference to a lease or
NULL to denote failure. Modify to return a real error and update
all callers. On return NULL, they can use the label and will elide
the atomic_dec in fl6_sock_release.
This is an optimization. Robust applications still have to revert to
requesting leases if the fast path fails due to an exclusive lease.
Changes RFC->v1:
- use static_key_false_deferred to rate limit jump label operations
- call static_key_deferred_flush to stop timers on exit
- move decrement out of RCU context
- defer optimization also if opt data is associated with a lease
- updated all fp6_sock_lookup callers, not just udp
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Kselftest can be run against older kernels. Instead of failing hard
when a feature is unsupported, return the KSFT_SKIP exit code.
Specifically, do not fail hard on missing udp zerocopy.
The udp gso bench test runs multiple test cases from a single script.
Fail if any case fails, else return skip if any test is skipped.
Link: https://lore.kernel.org/lkml/20190618171516.GA17547@kroah.com/
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
One warning each on signedness, unused variable and return type.
Fixes: 10fbcdd12a ("selftests/net: add TFO key rotation selftest")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
NAPI tx mode improves TCP behavior by enabling TCP small queues (TSQ).
TSQ reduces queuing ("bufferbloat") and burstiness.
Previous measurements have shown significant improvement for
TCP_STREAM style workloads. Such as those in commit 86a5df1495
("Merge branch 'virtio-net-tx-napi'").
There has been uncertainty about smaller possible regressions in
latency due to increased reliance on tx interrupts.
The above results did not show that, nor did I observe this when
rerunning TCP_RR on Linux 5.1 this week on a pair of guests in the
same rack. This may be subject to other settings, notably interrupt
coalescing.
In the unlikely case of regression, we have landed a credible runtime
solution. Ethtool can configure it with -C tx-frames [0|1] as of
commit 0c465be183 ("virtio_net: ethtool tx napi configuration").
NAPI tx mode has been the default in Google Container-Optimized OS
(COS) for over half a year, as of release M70 in October 2018,
without any negative reports.
Link: https://marc.info/?l=linux-netdev&m=149305618416472
Link: https://lwn.net/Articles/507065/
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Deferred static key clean_acked_data_enabled uses the deferred
variants of dec and flush. Do the same for inc.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This interface is currently only defined if CONFIG_JUMP_LABEL. Make it
available also when jump labels are off.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The below patch fixes an incorrect zerocopy refcnt increment when
appending with MSG_MORE to an existing zerocopy udp skb.
send(.., MSG_ZEROCOPY | MSG_MORE); // refcnt 1
send(.., MSG_ZEROCOPY | MSG_MORE); // refcnt still 1 (bar frags)
But it missed that zerocopy need not be passed at the first send. The
right test whether the uarg is newly allocated and thus has extra
refcnt 1 is not !skb, but !skb_zcopy.
send(.., MSG_MORE); // <no uarg>
send(.., MSG_ZEROCOPY); // refcnt 1
Fixes: 100f6d8e09 ("net: correct zerocopy refcnt with udp MSG_MORE")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rollover used to use a complex RCU mechanism for assignment, which had
a race condition. The below patch fixed the bug and greatly simplified
the logic.
The feature depends on fanout, but the state is private to the socket.
Fanout_release returns f only when the last member leaves and the
fanout struct is to be freed.
Destroy rollover unconditionally, regardless of fanout state.
Fixes: 57f015f5ec ("packet: fix crash in fanout_demux_rollover()")
Reported-by: syzbot <syzkaller@googlegroups.com>
Diagnosed-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP zerocopy takes a uarg reference for every skb, plus one for the
tcp_sendmsg_locked datapath temporarily, to avoid reaching refcnt zero
as it builds, sends and frees skbs inside its inner loop.
UDP and RAW zerocopy do not send inside the inner loop so do not need
the extra sock_zerocopy_get + sock_zerocopy_put pair. Commit
52900d22288ed ("udp: elide zerocopy operation in hot path") introduced
extra_uref to pass the initial reference taken in sock_zerocopy_alloc
to the first generated skb.
But, sock_zerocopy_realloc takes this extra reference at the start of
every call. With MSG_MORE, no new skb may be generated to attach the
extra_uref to, so refcnt is incorrectly 2 with only one skb.
Do not take the extra ref if uarg && !tcp, which implies MSG_MORE.
Update extra_uref accordingly.
This conditional assignment triggers a false positive may be used
uninitialized warning, so have to initialize extra_uref at define.
Changes v1->v2: fix typo in Fixes SHA1
Fixes: 52900d2228 ("udp: elide zerocopy operation in hot path")
Reported-by: syzbot <syzkaller@googlegroups.com>
Diagnosed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Test the IPv6 flowlabel control and datapath interfaces:
Acquire and release the right to use flowlabels with socket option
IPV6_FLOWLABEL_MGR.
Then configure flowlabels on send and read them on recv with cmsg
IPV6_FLOWINFO. Also verify auto-flowlabel if not explicitly set.
This helped identify the issue fixed in commit 95c169251b ("ipv6:
invert flowlabel sharing check in process and user mode")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The SO_TXTIME API enables packet tranmission with delayed delivery.
This is currently supported by the ETF and FQ packet schedulers.
Evaluate the interface with both schedulers. Install the scheduler
and send a variety of packets streams: without delay, with one
delayed packet, with multiple ordered delays and with reordering.
Verify that packets are released by the scheduler in expected order.
The ETF qdisc requires a timestamp in the future on every packet. It
needs a delay on the qdisc else the packet is dropped on dequeue for
having a delivery time in the past. The test value is experimentally
derived. ETF requires clock_id CLOCK_TAI. It checks this base and
drops for non-conformance.
The FQ qdisc expects clock_id CLOCK_MONOTONIC, the base used by TCP
as of commit fb420d5d91 ("tcp/fq: move back to CLOCK_MONOTONIC").
Within a flow there is an expecation of ordered delivery, as shown by
delivery times of test 4. The FQ qdisc does not require all packets to
have timestamps and does not drop for non-conformance.
The large (msec) delays are chosen to avoid flakiness.
Output:
SO_TXTIME ipv6 clock monotonic
payload:a delay:28 expected:0 (us)
SO_TXTIME ipv4 clock monotonic
payload:a delay:38 expected:0 (us)
SO_TXTIME ipv6 clock monotonic
payload:a delay:40 expected:0 (us)
SO_TXTIME ipv4 clock monotonic
payload:a delay:33 expected:0 (us)
SO_TXTIME ipv6 clock monotonic
payload:a delay:10120 expected:10000 (us)
SO_TXTIME ipv4 clock monotonic
payload:a delay:10102 expected:10000 (us)
[.. etc ..]
OK. All tests passed
Changes v1->v2: update commit message output
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Zerocopy skbs without completion notification were added for packet
sockets with PACKET_TX_RING user buffers. Those signal completion
through the TP_STATUS_USER bit in the ring. Zerocopy annotation was
added only to avoid premature notification after clone or orphan, by
triggering a copy on these paths for these packets.
The mechanism had to define a special "no-uarg" mode because packet
sockets already use skb_uarg(skb) == skb_shinfo(skb)->destructor_arg
for a different pointer.
Before deferencing skb_uarg(skb), verify that it is a real pointer.
Fixes: 5cd8d46ea1 ("packet: copy user buffers before orphan or clone")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Packet sockets in datagram mode take a destination address. Verify its
length before passing to dev_hard_header.
Prior to 2.6.14-rc3, the send code ignored sll_halen. This is
established behavior. Directly compare msg_namelen to dev->addr_len.
Change v1->v2: initialize addr in all paths
Fixes: 6b8d95f179 ("packet: validate address length if non-zero")
Suggested-by: David Laight <David.Laight@aculab.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Packet send checks that msg_name is at least sizeof sockaddr_ll.
Packet recv must return at least this length, so that its output
can be passed unmodified to packet send.
This ceased to be true since adding support for lladdr longer than
sll_addr. Since, the return value uses true address length.
Always return at least sizeof sockaddr_ll, even if address length
is shorter. Zero the padding bytes.
Change v1->v2: do not overwrite zeroed padding again. use copy_len.
Fixes: 0fb375fb9b ("[AF_PACKET]: Allow for > 8 byte hardware addresses.")
Suggested-by: David Laight <David.Laight@aculab.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A request for a flowlabel fails in process or user exclusive mode must
fail if the caller pid or uid does not match. Invert the test.
Previously, the test was unsafe wrt PID recycling, but indeed tested
for inequality: fl1->owner != fl->owner
Fixes: 4f82f45730 ("net ip6 flowlabel: Make owner a union of struct pid* and kuid_t")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
So far, all BPF tc tunnel testcases encapsulate in the same network
protocol. Add an encap testcase that requires updating skb->protocol.
The 6in4 tunnel encapsulates an IPv6 packet inside an IPv4 tunnel.
Verify that bpf_skb_net_grow correctly updates skb->protocol to
select the right protocol handler in __netif_receive_skb_core.
The BPF program should also manually update the link layer header to
encode the right network protocol.
Changes v1->v2
- improve documentation of non-obvious logic
Signed-off-by: Willem de Bruijn <willemb@google.com>
Tested-by: Alan Maguire <alan.maguire@oracle.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Some tunnels, like sit, change the network protocol of packet.
If so, update skb->protocol to match the new type.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Alan Maguire <alan.maguire@oracle.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The ENCAP flags in bpf_skb_adjust_room are ignored on decap with
bpf_skb_net_shrink. Reserve these bits for future use.
Fixes: 868d523535 ("bpf: add bpf_skb_adjust_room encap flags")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Alan Maguire <alan.maguire@oracle.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
These three variables are set in one branch and used in another with
the same condition. But on some architectures they still generate
compiler warnings of the kind:
warning: 'inner_trans' may be used uninitialized in this function [-Wmaybe-uninitialized]
Silence these false positives. Use the straightforward approach to
always initialize them, if a bit superfluous.
Fixes: 868d523535 ("bpf: add bpf_skb_adjust_room encap flags")
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Make the tests correctly annotate skbs with tunnel metadata.
This makes the gso tests succeed. Enable them.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Lower route MTU to ensure packets fit in device MTU after encap, then
skip the gso_size changes.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Avoid moving the network layer header when prefixing tunnel headers.
This avoids an explicit call to bpf_skb_store_bytes and an implicit
move of the network header bytes in bpf_skb_adjust_room.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Sync include/uapi/linux/bpf.h with tools/
Changes
v1->v2:
- BPF_F_ADJ_ROOM_MASK moved, no longer in this commit
v2->v3:
- BPF_F_ADJ_ROOM_ENCAP_L3_MASK moved, no longer in this commit
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
When pushing tunnel headers, annotate skbs in the same way as tunnel
devices.
For GSO packets, the network stack requires certain fields set to
segment packets with tunnel headers. gro_gse_segment depends on
transport and inner mac header, for instance.
Add an option to pass this information.
Remove the restriction on len_diff to network header length, which
is too short, e.g., for GRE protocols.
Changes
v1->v2:
- document new flags
- BPF_F_ADJ_ROOM_MASK moved
v2->v3:
- BPF_F_ADJ_ROOM_ENCAP_L3_MASK moved
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
bpf_skb_adjust_room adjusts gso_size of gso packets to account for the
pushed or popped header room.
This is not allowed with UDP, where gso_size delineates datagrams. Add
an option to avoid these updates and allow this call for datagrams.
It can also be used with TCP, when MSS is known to allow headroom,
e.g., through MSS clamping or route MTU.
Changes v1->v2:
- document flag BPF_F_ADJ_ROOM_FIXED_GSO
- do not expose BPF_F_ADJ_ROOM_MASK through uapi, as it may change.
Link: https://patchwork.ozlabs.org/patch/1052497/
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
bpf_skb_adjust_room net allows inserting room in an skb.
Existing mode BPF_ADJ_ROOM_NET inserts room after the network header
by pulling the skb, moving the network header forward and zeroing the
new space.
Add new mode BPF_ADJUST_ROOM_MAC that inserts room after the mac
header. This allows inserting tunnel headers in front of the network
header without having to recreate the network header in the original
space, avoiding two copies.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Segmentation offload takes a longer path. Verify that the feature
works with large packets.
The test succeeds if not setting dodgy in bpf_skb_adjust_room, as veth
TSO is permissive.
If not setting SKB_GSO_DODGY, this enables tunneled TSO offload on
supporting NICs.
The feature sets SKB_GSO_DODGY because the caller is untrusted. As a
result the packets traverse through the gso stack at least up to TCP.
And fail the gso_type validation, such as the skb->encapsulation check
in gre_gso_segment and the gso_type checks introduced in commit
418e897e07 ("gso: validate gso_type on ipip style tunnel").
This will be addressed in a follow-on feature patch. In the meantime,
disable the new gso tests.
Changes v1->v2:
- not all netcat versions support flag '-q', use timeout instead
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
GRE is a commonly used protocol. Add GRE cases for both IPv4 and IPv6.
It also inserts different sized headers, which can expose some
unexpected edge cases.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The test only uses ipv4 so far, expand to ipv6.
This is mostly a boilerplate near copy of the ipv4 path.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The bpf tunnel test encapsulates using bpf, then decapsulates using
a standard tunnel device to verify correctness.
Once encap is verified, also test decap, by replacing the tunnel
device on decap with another bpf program.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Validate basic tunnel encapsulation using ipip.
Set up two namespaces connected by veth. Connect a client and server.
Do this with and without bpf encap.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
bpf_skb_adjust_room calls skb_cow on grow.
This expensive operation can be avoided in the fast path when the only
other clone has released the header. This is the common case for TCP,
where one headerless clone is kept on the retransmit queue.
It is safe to do so even when touching the gso fields in skb_shinfo.
Regular tunnel encap with iptunnel_handle_offloads takes the same
optimization.
The tcp stack unclones in the unlikely case that it accesses these
fields through headerless clones packets on the retransmit queue (see
__tcp_retransmit_skb).
If any other clones are present, e.g., from packet sockets,
skb_cow_head returns the same value as skb_cow().
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
BPF can adjust gso only for tcp bytestreams. Fail on other gso types.
But only on gso packets. It does not touch this field if !gso_size.
Fixes: b90efd2258 ("bpf: only adjust gso_size on bytestream protocols")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This header defines the BPF functions enumerated in uapi/linux.bpf.h
in a callable format. Expand to include all registered functions.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
GSO packets with vnet_hdr must conform to a small set of gso_types.
The below commit uses flow dissection to drop packets that do not.
But it has false positives when the skb is not fully initialized.
Dissection needs skb->protocol and skb->network_header.
Infer skb->protocol from gso_type as the two must agree.
SKB_GSO_UDP can use both ipv4 and ipv6, so try both.
Exclude callers for which network header offset is not known.
Fixes: d5be7f632b ("net: validate untrusted gso packets without csum offload")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 121d57af30 ("gso: validate gso_type in GSO handlers") added
gso_type validation to existing gso_segment callback functions, to
filter out illegal and potentially dangerous SKB_GSO_DODGY packets.
Convert tunnels that now call inet_gso_segment and ipv6_gso_segment
directly to have their own callbacks and extend validation to these.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Syzkaller again found a path to a kernel crash through bad gso input.
By building an excessively large packet to cause an skb field to wrap.
If VIRTIO_NET_HDR_F_NEEDS_CSUM was set this would have been dropped in
skb_partial_csum_set.
GSO packets that do not set checksum offload are suspicious and rare.
Most callers of virtio_net_hdr_to_skb already pass them to
skb_probe_transport_header.
Move that test forward, change it to detect parse failure and drop
packets on failure as those cleary are not one of the legitimate
VIRTIO_NET_HDR_GSO types.
Fixes: bfd5f4a3d6 ("packet: Add GSO/csum offload support.")
Fixes: f43798c276 ("tun: Allow GSO using virtio_net_hdr")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
bpf_skb_change_proto and bpf_skb_adjust_room change skb header length.
For GSO packets they adjust gso_size to maintain the same MTU.
The gso size can only be safely adjusted on bytestream protocols.
Commit d02f51cbcf ("bpf: fix bpf_skb_adjust_net/bpf_skb_proto_xlat
to deal with gso sctp skbs") excluded SKB_GSO_SCTP.
Since then type SKB_GSO_UDP_L4 has been added, whose contents are one
gso_size unit per datagram. Also exclude these.
Move from a blacklist to a whitelist check to future proof against
additional such new GSO types, e.g., for fraglist based GRO.
Fixes: bec1f6f697 ("udp: generate gso with UDP_SEGMENT")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Accept MSG_ZEROCOPY in all the TCP states that allow sendmsg. Remove
the explicit check for ESTABLISHED and CLOSE_WAIT states.
This requires correctly handling zerocopy state (uarg, sk_zckey) in
all paths reachable from other TCP states. Such as the EPIPE case
in sk_stream_wait_connect, which a sendmsg() in incorrect state will
now hit. Most paths are already safe.
Only extension needed is for TCP Fastopen active open. This can build
an skb with data in tcp_send_syn_data. Pass the uarg along with other
fastopen state, so that this skb also generates a zerocopy
notification on release.
Tested with active and passive tcp fastopen packetdrill scripts at
https://github.com/wdebruij/packetdrill/commit/1747eef03d25a2404e8132817d0f1244fd6f129d
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Syzkaller was able to construct a packet of negative length by
redirecting from bpf_prog_test_run_skb with BPF_PROG_TYPE_LWT_XMIT:
BUG: KASAN: slab-out-of-bounds in memcpy include/linux/string.h:345 [inline]
BUG: KASAN: slab-out-of-bounds in skb_copy_from_linear_data include/linux/skbuff.h:3421 [inline]
BUG: KASAN: slab-out-of-bounds in __pskb_copy_fclone+0x2dd/0xeb0 net/core/skbuff.c:1395
Read of size 4294967282 at addr ffff8801d798009c by task syz-executor2/12942
kasan_report.cold.9+0x242/0x309 mm/kasan/report.c:412
check_memory_region_inline mm/kasan/kasan.c:260 [inline]
check_memory_region+0x13e/0x1b0 mm/kasan/kasan.c:267
memcpy+0x23/0x50 mm/kasan/kasan.c:302
memcpy include/linux/string.h:345 [inline]
skb_copy_from_linear_data include/linux/skbuff.h:3421 [inline]
__pskb_copy_fclone+0x2dd/0xeb0 net/core/skbuff.c:1395
__pskb_copy include/linux/skbuff.h:1053 [inline]
pskb_copy include/linux/skbuff.h:2904 [inline]
skb_realloc_headroom+0xe7/0x120 net/core/skbuff.c:1539
ipip6_tunnel_xmit net/ipv6/sit.c:965 [inline]
sit_tunnel_xmit+0xe1b/0x30d0 net/ipv6/sit.c:1029
__netdev_start_xmit include/linux/netdevice.h:4325 [inline]
netdev_start_xmit include/linux/netdevice.h:4334 [inline]
xmit_one net/core/dev.c:3219 [inline]
dev_hard_start_xmit+0x295/0xc90 net/core/dev.c:3235
__dev_queue_xmit+0x2f0d/0x3950 net/core/dev.c:3805
dev_queue_xmit+0x17/0x20 net/core/dev.c:3838
__bpf_tx_skb net/core/filter.c:2016 [inline]
__bpf_redirect_common net/core/filter.c:2054 [inline]
__bpf_redirect+0x5cf/0xb20 net/core/filter.c:2061
____bpf_clone_redirect net/core/filter.c:2094 [inline]
bpf_clone_redirect+0x2f6/0x490 net/core/filter.c:2066
bpf_prog_41f2bcae09cd4ac3+0xb25/0x1000
The generated test constructs a packet with mac header, network
header, skb->data pointing to network header and skb->len 0.
Redirecting to a sit0 through __bpf_redirect_no_mac pulls the
mac length, even though skb->data already is at skb->network_header.
bpf_prog_test_run_skb has already pulled it as LWT_XMIT !is_l2.
Update the offset calculation to pull only if skb->data differs
from skb->network_header, which is not true in this case.
The test itself can be run only from commit 1cf1cae963 ("bpf:
introduce BPF_PROG_TEST_RUN command"), but the same type of packets
with skb at network header could already be built from lwt xmit hooks,
so this fix is more relevant to that commit.
Also set the mac header on redirect from LWT_XMIT, as even after this
change to __bpf_redirect_no_mac that field is expected to be set, but
is not yet in ip_finish_output2.
Fixes: 3a0af8fd61 ("bpf: BPF for lightweight tunnel infrastructure")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
On multiqueue network devices, RPS maps are configured independently
for each receive queue through /sys/class/net/$DEV/queues/rx-*.
On virtio-net currently all packets use the map from rx-0, because the
real rx queue is not known at time of map lookup by get_rps_cpu.
Call skb_record_rx_queue in the driver rx path to make lookup work.
Recording the receive queue has ramifications beyond RPS, such as in
sticky load balancing decisions for sockets (skb_tx_hash) and XPS.
Reported-by: Mark Hlady <mhlady@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Failure __ip_append_data triggers udp_flush_pending_frames, but these
tests happen later. The skb must be freed directly.
Fixes: bec1f6f697 ("udp: generate gso with UDP_SEGMENT")
Reported-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A network device stack with multiple layers of bonding devices can
trigger a false positive lockdep warning. Adding lockdep nest levels
fixes this. Update the level on both enslave and unlink, to avoid the
following series of events ..
ip netns add test
ip netns exec test bash
ip link set dev lo addr 00:11:22:33:44:55
ip link set dev lo down
ip link add dev bond1 type bond
ip link add dev bond2 type bond
ip link set dev lo master bond1
ip link set dev bond1 master bond2
ip link set dev bond1 nomaster
ip link set dev bond2 master bond1
.. from still generating a splat:
[ 193.652127] ======================================================
[ 193.658231] WARNING: possible circular locking dependency detected
[ 193.664350] 4.20.0 #8 Not tainted
[ 193.668310] ------------------------------------------------------
[ 193.674417] ip/15577 is trying to acquire lock:
[ 193.678897] 00000000a40e3b69 (&(&bond->stats_lock)->rlock#3/3){+.+.}, at: bond_get_stats+0x58/0x290
[ 193.687851]
but task is already holding lock:
[ 193.693625] 00000000807b9d9f (&(&bond->stats_lock)->rlock#2/2){+.+.}, at: bond_get_stats+0x58/0x290
[..]
[ 193.851092] lock_acquire+0xa7/0x190
[ 193.855138] _raw_spin_lock_nested+0x2d/0x40
[ 193.859878] bond_get_stats+0x58/0x290
[ 193.864093] dev_get_stats+0x5a/0xc0
[ 193.868140] bond_get_stats+0x105/0x290
[ 193.872444] dev_get_stats+0x5a/0xc0
[ 193.876493] rtnl_fill_stats+0x40/0x130
[ 193.880797] rtnl_fill_ifinfo+0x6c5/0xdc0
[ 193.885271] rtmsg_ifinfo_build_skb+0x86/0xe0
[ 193.890091] rtnetlink_event+0x5b/0xa0
[ 193.894320] raw_notifier_call_chain+0x43/0x60
[ 193.899225] netdev_change_features+0x50/0xa0
[ 193.904044] bond_compute_features.isra.46+0x1ab/0x270
[ 193.909640] bond_enslave+0x141d/0x15b0
[ 193.913946] do_set_master+0x89/0xa0
[ 193.918016] do_setlink+0x37c/0xda0
[ 193.921980] __rtnl_newlink+0x499/0x890
[ 193.926281] rtnl_newlink+0x48/0x70
[ 193.930238] rtnetlink_rcv_msg+0x171/0x4b0
[ 193.934801] netlink_rcv_skb+0xd1/0x110
[ 193.939103] rtnetlink_rcv+0x15/0x20
[ 193.943151] netlink_unicast+0x3b5/0x520
[ 193.947544] netlink_sendmsg+0x2fd/0x3f0
[ 193.951942] sock_sendmsg+0x38/0x50
[ 193.955899] ___sys_sendmsg+0x2ba/0x2d0
[ 193.960205] __x64_sys_sendmsg+0xad/0x100
[ 193.964687] do_syscall_64+0x5a/0x460
[ 193.968823] entry_SYSCALL_64_after_hwframe+0x49/0xbe
Fixes: 7e2556e400 ("bonding: avoid lockdep confusion in bond_get_stats()")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 2efd4fca70 ("ip: in cmsg IP(V6)_ORIGDSTADDR call
pskb_may_pull") avoided a read beyond the end of the skb linear
segment by calling pskb_may_pull.
That function can trigger a BUG_ON in pskb_expand_head if the skb is
shared, which it is when when peeking. It can also return ENOMEM.
Avoid both by switching to safer skb_header_pointer.
Fixes: 2efd4fca70 ("ip: in cmsg IP(V6)_ORIGDSTADDR call pskb_may_pull")
Reported-by: syzbot <syzkaller@googlegroups.com>
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
KMSAN detected read beyond end of buffer in vti and sit devices when
passing truncated packets with PF_PACKET. The issue affects additional
ip tunnel devices.
Extend commit 76c0ddd8c3 ("ip6_tunnel: be careful when accessing the
inner header") and commit ccfec9e5cb ("ip_tunnel: be careful when
accessing the inner header").
Move the check to a separate helper and call at the start of each
ndo_start_xmit function in net/ipv4 and net/ipv6.
Minor changes:
- convert dev_kfree_skb to kfree_skb on error path,
as dev_kfree_skb calls consume_skb which is not for error paths.
- use pskb_network_may_pull even though that is pedantic here,
as the same as pskb_may_pull for devices without llheaders.
- do not cache ipv6 hdrs if used only once
(unsafe across pskb_may_pull, was more relevant to earlier patch)
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The BPF flow dissector expects either skb->sk or skb->dev set on
all skbs. Delay flow dissection until after skb->dev is set.
This requires calling from within an rcu read-side critical section.
That is fine, see also the call from tun_xdp_one.
Fixes: d0e13a1488 ("flow_dissector: lookup netns by skb->sk if skb->dev is NULL")
Reported-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Packet sockets may call dev_header_parse with NULL daddr. Make
lowpan_header_ops.create fail.
Fixes: 87a93e4ece ("ieee802154: change needed headroom/tailroom")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Alexander Aring <aring@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Validate packet socket address length if a length is given. Zero
length is equivalent to not setting an address.
Fixes: 99137b7888 ("packet: validate address length")
Reported-by: Ido Schimmel <idosch@idosch.org>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Packet sockets with SOCK_DGRAM may pass an address for use in
dev_hard_header. Ensure that it is of sufficient length.
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Virtio-net devices negotiate LRO support with the host.
Display the initially negotiated state with ethtool -k.
Also allow configuring it with ethtool -K, reusing the existing
virtnet_set_guest_offloads helper that configures LRO for XDP.
This is conditional on VIRTIO_NET_F_CTRL_GUEST_OFFLOADS.
Virtio-net negotiates TSO4 and TSO6 separately, but ethtool does not
distinguish between the two. Display LRO as on only if any offload
is active.
RTNL is held while calling virtnet_set_features, same as on the path
from virtnet_xdp_set.
Changes v1 -> v2
- allow ethtool config (-K) only if VIRTIO_NET_F_CTRL_GUEST_OFFLOADS
- show LRO as enabled if any LRO variant is enabled
- do not allow configuration while XDP is active
- differentiate current features from the capable set, to restore
on XDP down only those features that were active on XDP up
- move test out of VIRTIO_NET_F_CSUM/TSO branch, which is tx only
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Run the transmit timestamp tests as part of kselftests.
Add a txtimestamp.sh test script that runs most variants:
ipv4/ipv6, tcp/udp/raw/raw_ipproto/pf_packet, data/nodata,
setsockopt/cmsg. The script runs tests with netem delays.
Refine txtimestamp.c to validate results. Take expected
netem delays as input and compare against real timestamps.
To run without dependencies, add a listener socket to be
able to connect in the case of TCP.
Add the timestamping directory to the kselftests Makefile.
Build all the binaries. Only run verified txtimestamp.sh.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Expand the transmit timestamp regression test with support for
missing protocols: ipv6 datagram and raw and pf_packet.
Also refine resolve_hostname to independently request AF_INET or
AF_INET6 addresses. Else, ipv4 addresses may be returned as AF_INET6.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 3dd17e63f5 ("sock: accept SO_TIMESTAMPING flags in socket
cmsg") added support for passing tx timestamping options per-call
in sendmsg.
Expand the txtimestamp test with support for this feature.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
SOF_TIMESTAMPING_OPT_ID is supported on TCP, UDP and RAW sockets.
But it was missing on RAW with IPPROTO_IP, PF_PACKET and CAN.
Add skb_setup_tx_timestamp that configures both tx_flags and tskey
for these paths that do not need corking or use bytestream keys.
Fixes: 09c2d251b7 ("net-timestamp: add key to disambiguate concurrent datagrams")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Raw sockets support tx timestamping, but one case is missing.
IPPROTO_RAW takes a separate packet construction path. raw_send_hdrinc
has an explicit call to sock_tx_timestamp, but rawv6_send_hdrinc does
not. Add it.
Fixes: 11878b40ed ("net-timestamp: SOCK_RAW and PING timestamping")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
extra_uref is used in __ip(6)_append_data only if uarg is set.
Smatch sees that the variable is passed to sock_zerocopy_put_abort.
This function accesses it only when uarg is set, but smatch cannot
infer this.
Make this dependency explicit.
Fixes: 52900d2228 ("udp: elide zerocopy operation in hot path")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Both msg_zerocopy and udpgso_bench have udp zerocopy variants.
Exercise these as part of the standard kselftest run.
With udp, msg_zerocopy has no control channel. Ensure that the
receiver exits after the sender by accounting for the initial
delay in starting them (in msg_zerocopy.sh).
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With MSG_ZEROCOPY, each skb holds a reference to a struct ubuf_info.
Release of its last reference triggers a completion notification.
The TCP stack in tcp_sendmsg_locked holds an extra ref independent of
the skbs, because it can build, send and free skbs within its loop,
possibly reaching refcount zero and freeing the ubuf_info too soon.
The UDP stack currently also takes this extra ref, but does not need
it as all skbs are sent after return from __ip(6)_append_data.
Avoid the extra refcount_inc and refcount_dec_and_test, and generally
the sock_zerocopy_put in the common path, by passing the initial
reference to the first skb.
This approach is taken instead of initializing the refcount to 0, as
that would generate error "refcount_t: increment on 0" on the
next skb_zcopy_set.
Changes
v3 -> v4
- Move skb_zcopy_set below the only kfree_skb that might cause
a premature uarg destroy before skb_zerocopy_put_abort
- Move the entire skb_shinfo assignment block, to keep that
cacheline access in one place
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>
Extend zerocopy to udp sockets. Allow setting sockopt SO_ZEROCOPY and
interpret flag MSG_ZEROCOPY.
This patch was previously part of the zerocopy RFC patchsets. Zerocopy
is not effective at small MTU. With segmentation offload building
larger datagrams, the benefit of page flipping outweights the cost of
generating a completion notification.
tools/testing/selftests/net/msg_zerocopy.sh after applying follow-on
test patch and making skb_orphan_frags_rx same as skb_orphan_frags:
ipv4 udp -t 1
tx=191312 (11938 MB) txc=0 zc=n
rx=191312 (11938 MB)
ipv4 udp -z -t 1
tx=304507 (19002 MB) txc=304507 zc=y
rx=304507 (19002 MB)
ok
ipv6 udp -t 1
tx=174485 (10888 MB) txc=0 zc=n
rx=174485 (10888 MB)
ipv6 udp -z -t 1
tx=294801 (18396 MB) txc=294801 zc=y
rx=294801 (18396 MB)
ok
Changes
v1 -> v2
- Fixup reverse christmas tree violation
v2 -> v3
- Split refcount avoidance optimization into separate patch
- Fix refcount leak on error in fragmented case
(thanks to Paolo Abeni for pointing this one out!)
- Fix refcount inc on zero
- Test sock_flag SOCK_ZEROCOPY directly in __ip_append_data.
This is needed since commit 5cf4a8532c ("tcp: really ignore
MSG_ZEROCOPY if no SO_ZEROCOPY") did the same for tcp.
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>
Packet sockets with PACKET_TX_RING send skbs with user data in frags.
Before commit 5cd8d46ea1 ("packet: copy user buffers before orphan
or clone") ring slots could be released prematurely, possibly allowing
a process to overwrite data still in flight.
This test opens two packet sockets, one to send and one to read.
The sender has a tx ring of one slot. It sends two packets with
different payload, then reads both and verifies their payload.
Before the above commit, both receive calls return the same data as
the send calls use the same buffer. From the commit, the clone
needed for looping onto a packet socket triggers an skb_copy_ubufs
to create a private copy. The separate sends each arrive correctly.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In ip packet generation, pagedlen is initialized for each skb at the
start of the loop in __ip(6)_append_data, before label alloc_new_skb.
Depending on compiler options, code can be generated that jumps to
this label, triggering use of an an uninitialized variable.
In practice, at -O2, the generated code moves the initialization below
the label. But the code should not rely on that for correctness.
Fixes: 15e36f5b8e ("udp: paged allocation with gso")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
tpacket_snd sends packets with user pages linked into skb frags. It
notifies that pages can be reused when the skb is released by setting
skb->destructor to tpacket_destruct_skb.
This can cause data corruption if the skb is orphaned (e.g., on
transmit through veth) or cloned (e.g., on mirror to another psock).
Create a kernel-private copy of data in these cases, same as tun/tap
zerocopy transmission. Reuse that infrastructure: mark the skb as
SKBTX_ZEROCOPY_FRAG, which will trigger copy in skb_orphan_frags(_rx).
Unlike other zerocopy packets, do not set shinfo destructor_arg to
struct ubuf_info. tpacket_destruct_skb already uses that ptr to notify
when the original skb is released and a timestamp is recorded. Do not
change this timestamp behavior. The ubuf_info->callback is not needed
anyway, as no zerocopy notification is expected.
Mark destructor_arg as not-a-uarg by setting the lower bit to 1. The
resulting value is not a valid ubuf_info pointer, nor a valid
tpacket_snd frame address. Add skb_zcopy_.._nouarg helpers for this.
The fix relies on features introduced in commit 52267790ef ("sock:
add MSG_ZEROCOPY"), so can be backported as is only to 4.14.
Tested with from `./in_netns.sh ./txring_overwrite` from
http://github.com/wdebruij/kerneltools/tests
Fixes: 69e3c75f4d ("net: TX_RING and packet mmap")
Reported-by: Anand H. Krishnan <anandhkrishnan@gmail.com>
Signed-off-by: Willem de Bruijn <willemb@google.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>
BPF flow dissectors are configured per network namespace.
__skb_flow_dissect looks up the netns through dev_net(skb->dev).
In some dissector paths skb->dev is NULL, such as for Unix sockets.
In these cases fall back to looking up the netns by socket.
Analyzing the codepaths leading to __skb_flow_dissect I did not find
a case where both skb->dev and skb->sk are NULL. Warn and fall back to
standard flow dissector if one is found.
Fixes: d58e468b11 ("flow_dissector: implements flow dissector BPF hook")
Reported-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
If boolean CONFIG_BPF_SYSCALL is enabled, kernel/bpf/syscall.c will
call flow_dissector functions from net/core/flow_dissector.c.
This causes this build failure if CONFIG_NET is disabled:
kernel/bpf/syscall.o: In function `__x64_sys_bpf':
syscall.c:(.text+0x3278): undefined reference to
`skb_flow_dissector_bpf_prog_attach'
syscall.c:(.text+0x3310): undefined reference to
`skb_flow_dissector_bpf_prog_detach'
kernel/bpf/syscall.o:(.rodata+0x3f0): undefined reference to
`flow_dissector_prog_ops'
kernel/bpf/verifier.o:(.rodata+0x250): undefined reference to
`flow_dissector_verifier_ops'
Analogous to other optional BPF program types in syscall.c, add stubs
if the relevant functions are not compiled and move the BPF_PROG_TYPE
definition in the #ifdef CONFIG_NET block.
Fixes: d58e468b11 ("flow_dissector: implements flow dissector BPF hook")
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Convert pr_info to net_info_ratelimited to limit the total number of
synflood warnings.
Commit 946cedccbd ("tcp: Change possible SYN flooding messages")
rate limits synflood warnings to one per listener.
Workloads that open many listener sockets can still see a high rate of
log messages. Syzkaller is one frequent example.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The kbuild robot shows build failure on machines without CONFIG_SMP:
drivers/net/virtio_net.c:1916:10: error:
implicit declaration of function 'cpumask_next_wrap'
cpumask_next_wrap is exported from lib/cpumask.o, which has
lib-$(CONFIG_SMP) += cpumask.o
same as other functions, also define it as static inline in the
NR_CPUS==1 branch in include/linux/cpumask.h.
If wrap is true and next == start, return nr_cpumask_bits, or 1.
Else wrap across the range of valid cpus, here [0].
Fixes: 2ca653d607 ("virtio_net: Stripe queue affinities across cores.")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Tested-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
TPACKET_V3 stores variable length frames in fixed length blocks.
Blocks must be able to store a block header, optional private space
and at least one minimum sized frame.
Frames, even for a zero snaplen packet, store metadata headers and
optional reserved space.
In the block size bounds check, ensure that the frame of the
chosen configuration fits. This includes sockaddr_ll and optional
tp_reserve.
Syzbot was able to construct a ring with insuffient room for the
sockaddr_ll in the header of a zero-length frame, triggering an
out-of-bounds write in dev_parse_header.
Convert the comparison to less than, as zero is a valid snap len.
This matches the test for minimum tp_frame_size immediately below.
Fixes: f6fb8f100b ("af-packet: TPACKET_V3 flexible buffer implementation.")
Fixes: eb73190f4f ("net/packet: refine check for priv area size")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Syzbot reported a read beyond the end of the skb head when returning
IPV6_ORIGDSTADDR:
BUG: KMSAN: kernel-infoleak in put_cmsg+0x5ef/0x860 net/core/scm.c:242
CPU: 0 PID: 4501 Comm: syz-executor128 Not tainted 4.17.0+ #9
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS
Google 01/01/2011
Call Trace:
__dump_stack lib/dump_stack.c:77 [inline]
dump_stack+0x185/0x1d0 lib/dump_stack.c:113
kmsan_report+0x188/0x2a0 mm/kmsan/kmsan.c:1125
kmsan_internal_check_memory+0x138/0x1f0 mm/kmsan/kmsan.c:1219
kmsan_copy_to_user+0x7a/0x160 mm/kmsan/kmsan.c:1261
copy_to_user include/linux/uaccess.h:184 [inline]
put_cmsg+0x5ef/0x860 net/core/scm.c:242
ip6_datagram_recv_specific_ctl+0x1cf3/0x1eb0 net/ipv6/datagram.c:719
ip6_datagram_recv_ctl+0x41c/0x450 net/ipv6/datagram.c:733
rawv6_recvmsg+0x10fb/0x1460 net/ipv6/raw.c:521
[..]
This logic and its ipv4 counterpart read the destination port from
the packet at skb_transport_offset(skb) + 4.
With MSG_MORE and a local SOCK_RAW sender, syzbot was able to cook a
packet that stores headers exactly up to skb_transport_offset(skb) in
the head and the remainder in a frag.
Call pskb_may_pull before accessing the pointer to ensure that it lies
in skb head.
Link: http://lkml.kernel.org/r/CAF=yD-LEJwZj5a1-bAAj2Oy_hKmGygV6rsJ_WOrAYnv-fnayiQ@mail.gmail.com
Reported-by: syzbot+9adb4b567003cac781f0@syzkaller.appspotmail.com
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The udpgso benchmark compares various configurations of UDP and TCP.
Including one that is not upstream, udp zerocopy. This is a leftover
from the earlier RFC patchset.
The test is part of kselftests and run in continuous spinners. Remove
the failing case to make the test start passing.
Fixes: 3a687bef14 ("selftests: udp gso benchmark")
Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If variable length link layer headers result in a packet shorter
than dev->hard_header_len, reset the network header offset. Else
skb->mac_len may exceed skb->len after skb_mac_reset_len.
packet_sendmsg_spkt already has similar logic.
Fixes: b84bbaf7a6 ("packet: in packet_snd start writing at link layer allocation")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
skb_shinfo(skb)->tx_flags is derived from sk->sk_tsflags, possibly
after modification by __sock_cmsg_send, by calling sock_tx_timestamp.
The IPv4 and IPv6 paths do this conversion differently. In IPv4, the
individual protocols that support tx timestamps call this function
and store the result in ipc.tx_flags. In IPv6, sock_tx_timestamp is
called in __ip6_append_data.
There is no need to store both tx_flags and ts_flags in the cookie
as one is derived from the other. Convert when setting up the cork
and remove the redundant field. This is similar to IPv6, only have
the conversion happen only once per datagram, in ip(6)_setup_cork.
Also change __ip6_append_data to match __ip_append_data. Only update
tskey if timestamping is enabled with OPT_ID. The SOCK_.. test is
redundant: only valid protocols can have non-zero cork->tx_flags.
After this change the IPv4 and IPv6 logic is the same.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ipcm_cookie includes sockcm_cookie. Do the same for ipcm6_cookie.
This reduces the number of arguments that need to be passed around,
applies ipcm6_init to all cookie fields at once and reduces code
differentiation between ipv4 and ipv6.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Initialize the cookie in one location to reduce code duplication and
avoid bugs from inconsistent initialization, such as that fixed in
commit 9887cba199 ("ip: limit use of gso_size to udp").
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Initialize the cookie in one location to reduce code duplication and
avoid bugs from inconsistent initialization, such as that fixed in
commit 9887cba199 ("ip: limit use of gso_size to udp").
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Initialize the cookie in one location to reduce code duplication and
avoid bugs from inconsistent initialization, such as that fixed in
commit 9887cba199 ("ip: limit use of gso_size to udp").
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ipcm(6)_cookie field gso_size is set only in the udp path. The ip
layer copies this to cork only if sk_type is SOCK_DGRAM. This check
proved too permissive. Ping and l2tp sockets have the same type.
Limit to sockets of type SOCK_DGRAM and protocol IPPROTO_UDP to
exclude ping sockets.
v1 -> v2
- remove irrelevant whitespace changes
Fixes: bec1f6f697 ("udp: generate gso with UDP_SEGMENT")
Reported-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tun, tap, virtio, packet and uml vector all use struct virtio_net_hdr
to communicate packet metadata to userspace.
For skbuffs with vlan, the first two return the packet as it may have
existed on the wire, inserting the VLAN tag in the user buffer. Then
virtio_net_hdr.csum_start needs to be adjusted by VLAN_HLEN bytes.
Commit f09e2249c4 ("macvtap: restore vlan header on user read")
added this feature to macvtap. Commit 3ce9b20f19 ("macvtap: Fix
csum_start when VLAN tags are present") then fixed up csum_start.
Virtio, packet and uml do not insert the vlan header in the user
buffer.
When introducing virtio_net_hdr_from_skb to deduplicate filling in
the virtio_net_hdr, the variant from macvtap which adds VLAN_HLEN was
applied uniformly, breaking csum offset for packets with vlan on
virtio and packet.
Make insertion of VLAN_HLEN optional. Convert the callers to pass it
when needed.
Fixes: e858fae2b0 ("virtio_net: use common code for virtio_net_hdr and skb GSO conversion")
Fixes: 1276f24eee ("packet: use common code for virtio_net_hdr and skb GSO conversion")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add regression tests for PF_PACKET transmission using packet_snd.
The TPACKET ring interface has tests for transmission and reception.
This is an initial stab at the same for the send call based interface.
Packets are sent over loopback, then read twice. The entire packet is
read from another packet socket and compared. The packet is also
verified to arrive at a UDP socket for protocol conformance.
The test sends a packet over loopback, testing the following options
(not the full cross-product):
- SOCK_DGRAM
- SOCK_RAW
- vlan tag
- qdisc bypass
- bind() and sendto()
- virtio_net_hdr
- csum offload (NOT actual csum feature, ignored on loopback)
- gso
Besides these basic functionality tests, the test runs from a set
of bounds checks, positive and negative. Running over loopback, which
has dev->min_header_len, it cannot generate variable length hhlen.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Verify that udpgso can generate segments smaller than device mtu, down
to the extreme case of 1B gso_size.
Verify that irrespective of gso_size, udpgso restricts the number of
segments it will generate per call (UDP_MAX_SEGMENTS).
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The existing msg_zerocopy test takes additional protocol arguments.
Add a variant that takes no arguments and runs all supported variants.
Call this from kselftest.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A precondition check in ip_recv_error triggered on an otherwise benign
race. Remove the warning.
The warning triggers when passing an ipv6 socket to this ipv4 error
handling function. RaceFuzzer was able to trigger it due to a race
in setsockopt IPV6_ADDRFORM.
---
CPU0
do_ipv6_setsockopt
sk->sk_socket->ops = &inet_dgram_ops;
---
CPU1
sk->sk_prot->recvmsg
udp_recvmsg
ip_recv_error
WARN_ON_ONCE(sk->sk_family == AF_INET6);
---
CPU0
do_ipv6_setsockopt
sk->sk_family = PF_INET;
This socket option converts a v6 socket that is connected to a v4 peer
to an v4 socket. It updates the socket on the fly, changing fields in
sk as well as other structs. This is inherently non-atomic. It races
with the lockless udp_recvmsg path.
No other code makes an assumption that these fields are updated
atomically. It is benign here, too, as ip_recv_error cares only about
the protocol of the skbs enqueued on the error queue, for which
sk_family is not a precise predictor (thanks to another isue with
IPV6_ADDRFORM).
Link: http://lkml.kernel.org/r/20180518120826.GA19515@dragonet.kaist.ac.kr
Fixes: 7ce875e5ec ("ipv4: warn once on passing AF_INET6 socket to ip_recv_error")
Reported-by: DaeRyong Jeong <threeearcat@gmail.com>
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit b84bbaf7a6 ("packet: in packet_snd start writing at link
layer allocation") ensures that packet_snd always starts writing
the link layer header in reserved headroom allocated for this
purpose.
This is needed because packets may be shorter than hard_header_len,
in which case the space up to hard_header_len may be zeroed. But
that necessary padding is not accounted for in skb->len.
The fix, however, is buggy. It calls skb_push, which grows skb->len
when moving skb->data back. But in this case packet length should not
change.
Instead, call skb_reserve, which moves both skb->data and skb->tail
back, without changing length.
Fixes: b84bbaf7a6 ("packet: in packet_snd start writing at link layer allocation")
Reported-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Until the udp receive stack supports large packets (UDP GRO), GSO
packets must not loop from the egress to the ingress path.
Revert the change that added NETIF_F_GSO_UDP_L4 to various virtual
devices through NETIF_F_GSO_ENCAP_ALL as this included devices that
may loop packets, such as veth and macvlan.
Instead add it to specific devices that forward to another device's
egress path, bonding and team.
Fixes: 83aa025f53 ("udp: add gso support to virtual devices")
CC: Alexander Duyck <alexander.duyck@gmail.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
UDP GSO delays final datagram construction to the GSO layer. This
conflicts with protocol transformations.
Fixes: bec1f6f697 ("udp: generate gso with UDP_SEGMENT")
CC: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Device features may change during transmission. In particular with
corking, a device may toggle scatter-gather in between allocating
and writing to an skb.
Do not unconditionally assume that !NETIF_F_SG at write time implies
that the same held at alloc time and thus the skb has sufficient
tailroom.
This issue predates git history.
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Reported-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Packet sockets allow construction of packets shorter than
dev->hard_header_len to accommodate protocols with variable length
link layer headers. These packets are padded to dev->hard_header_len,
because some device drivers interpret that as a minimum packet size.
packet_snd reserves dev->hard_header_len bytes on allocation.
SOCK_DGRAM sockets call skb_push in dev_hard_header() to ensure that
link layer headers are stored in the reserved range. SOCK_RAW sockets
do the same in tpacket_snd, but not in packet_snd.
Syzbot was able to send a zero byte packet to a device with massive
116B link layer header, causing padding to cross over into skb_shinfo.
Fix this by writing from the start of the llheader reserved range also
in the case of packet_snd/SOCK_RAW.
Update skb_set_network_header to the new offset. This also corrects
it for SOCK_DGRAM, where it incorrectly double counted reserve due to
the skb_push in dev_hard_header.
Fixes: 9ed988cd59 ("packet: validate variable length ll headers")
Reported-by: syzbot+71d74a5406d02057d559@syzkaller.appspotmail.com
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Syzbot managed to send a udp gso packet without checksum offload into
the gso stack by disabling tx checksum (UDP_NO_CHECK6_TX). This
triggered the skb_warn_bad_offload.
RIP: 0010:skb_warn_bad_offload+0x2bc/0x600 net/core/dev.c:2658
skb_gso_segment include/linux/netdevice.h:4038 [inline]
validate_xmit_skb+0x54d/0xd90 net/core/dev.c:3120
__dev_queue_xmit+0xbf8/0x34c0 net/core/dev.c:3577
dev_queue_xmit+0x17/0x20 net/core/dev.c:3618
UDP_NO_CHECK6_TX sets skb->ip_summed to CHECKSUM_NONE just after the
udp gso integrity checks in udp_(v6_)send_skb. Extend those checks to
catch and fail in this case.
After the integrity checks jump directly to the CHECKSUM_PARTIAL case
to avoid reading the no_check_tx flags again (a TOCTTOU race).
Fixes: bec1f6f697 ("udp: generate gso with UDP_SEGMENT")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
UDP GSO needs to export __udp_gso_segment to call it from ipv6.
I accidentally exported static ipv4 function __udp4_gso_segment.
Remove that EXPORT_SYMBOL_GPL.
Fixes: ee80d1ebe5 ("udp: add udp gso")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Send udp data between a source and sink, optionally with udp gso.
The two processes are expected to be run on separate hosts.
A script is included that runs them together over loopback in a
single namespace for functionality testing.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Corked sockets take a different path to construct a udp datagram than
the lockless fast path. Test this alternate path.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Connected sockets use path mtu instead of device mtu.
Test this path by inserting a route mtu that is lower than the device
mtu. Verify that the path mtu for the connection matches this lower
number, then run the same test as in the connectionless case.
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Willem de Bruijn