Expose TCP rx queue accessor and cleanup, so that TLS can
decrypt directly from the TCP queue. The expectation
is that the caller can access the skb returned from
tcp_recv_skb() and up to inq bytes worth of data (some
of which may be in ->next skbs) and then call
tcp_read_done() when data has been consumed.
The socket lock must be held continuously across
those two operations.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
While reading sysctl_tcp_adv_win_scale, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its reader.
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While reading sysctl_tcp_slow_start_after_idle, it can be changed
concurrently. Thus, we need to add READ_ONCE() to its readers.
Fixes: 35089bb203 ("[TCP]: Add tcp_slow_start_after_idle sysctl.")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While reading sysctl_tcp_notsent_lowat, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its reader.
Fixes: c9bee3b7fd ("tcp: TCP_NOTSENT_LOWAT socket option")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While reading these sysctl knobs, they can be changed concurrently.
Thus, we need to add READ_ONCE() to their readers.
- tcp_retries1
- tcp_retries2
- tcp_orphan_retries
- tcp_fin_timeout
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While reading sysctl_tcp_keepalive_(time|probes|intvl), they can be changed
concurrently. Thus, we need to add READ_ONCE() to their readers.
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit e21145a987 ("ipv4: namespacify ip_early_demux sysctl knob") made
it possible to enable/disable early_demux on a per-netns basis. Then, we
introduced two knobs, tcp_early_demux and udp_early_demux, to switch it for
TCP/UDP in commit dddb64bcb3 ("net: Add sysctl to toggle early demux for
tcp and udp"). However, the .proc_handler() was wrong and actually
disabled us from changing the behaviour in each netns.
We can execute early_demux if net.ipv4.ip_early_demux is on and each proto
.early_demux() handler is not NULL. When we toggle (tcp|udp)_early_demux,
the change itself is saved in each netns variable, but the .early_demux()
handler is a global variable, so the handler is switched based on the
init_net's sysctl variable. Thus, netns (tcp|udp)_early_demux knobs have
nothing to do with the logic. Whether we CAN execute proto .early_demux()
is always decided by init_net's sysctl knob, and whether we DO it or not is
by each netns ip_early_demux knob.
This patch namespacifies (tcp|udp)_early_demux again. For now, the users
of the .early_demux() handler are TCP and UDP only, and they are called
directly to avoid retpoline. So, we can remove the .early_demux() handler
from inet6?_protos and need not dereference them in ip6?_rcv_finish_core().
If another proto needs .early_demux(), we can restore it at that time.
Fixes: dddb64bcb3 ("net: Add sysctl to toggle early demux for tcp and udp")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://lore.kernel.org/r/20220713175207.7727-1-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently both splice() and sockmap use ->read_sock() to
read skb from receive queue, but for sockmap we only read
one entire skb at a time, so ->read_sock() is too conservative
to use. Introduce a new proto_ops ->read_skb() which supports
this sematic, with this we can finally pass the ownership of
skb to recv actors.
For non-TCP protocols, all ->read_sock() can be simply
converted to ->read_skb().
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20220615162014.89193-3-xiyou.wangcong@gmail.com
This patch inroduces tcp_read_skb() based on tcp_read_sock(),
a preparation for the next patch which actually introduces
a new sock ops.
TCP is special here, because it has tcp_read_sock() which is
mainly used by splice(). tcp_read_sock() supports partial read
and arbitrary offset, neither of them is needed for sockmap.
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20220615162014.89193-2-xiyou.wangcong@gmail.com
Daniel Borkmann says:
====================
pull-request: bpf-next 2022-06-17
We've added 72 non-merge commits during the last 15 day(s) which contain
a total of 92 files changed, 4582 insertions(+), 834 deletions(-).
The main changes are:
1) Add 64 bit enum value support to BTF, from Yonghong Song.
2) Implement support for sleepable BPF uprobe programs, from Delyan Kratunov.
3) Add new BPF helpers to issue and check TCP SYN cookies without binding to a
socket especially useful in synproxy scenarios, from Maxim Mikityanskiy.
4) Fix libbpf's internal USDT address translation logic for shared libraries as
well as uprobe's symbol file offset calculation, from Andrii Nakryiko.
5) Extend libbpf to provide an API for textual representation of the various
map/prog/attach/link types and use it in bpftool, from Daniel Müller.
6) Provide BTF line info for RV64 and RV32 JITs, and fix a put_user bug in the
core seen in 32 bit when storing BPF function addresses, from Pu Lehui.
7) Fix libbpf's BTF pointer size guessing by adding a list of various aliases
for 'long' types, from Douglas Raillard.
8) Fix bpftool to readd setting rlimit since probing for memcg-based accounting
has been unreliable and caused a regression on COS, from Quentin Monnet.
9) Fix UAF in BPF cgroup's effective program computation triggered upon BPF link
detachment, from Tadeusz Struk.
10) Fix bpftool build bootstrapping during cross compilation which was pointing
to the wrong AR process, from Shahab Vahedi.
11) Fix logic bug in libbpf's is_pow_of_2 implementation, from Yuze Chi.
12) BPF hash map optimization to avoid grabbing spinlocks of all CPUs when there
is no free element. Also add a benchmark as reproducer, from Feng Zhou.
13) Fix bpftool's codegen to bail out when there's no BTF, from Michael Mullin.
14) Various minor cleanup and improvements all over the place.
* https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (72 commits)
bpf: Fix bpf_skc_lookup comment wrt. return type
bpf: Fix non-static bpf_func_proto struct definitions
selftests/bpf: Don't force lld on non-x86 architectures
selftests/bpf: Add selftests for raw syncookie helpers in TC mode
bpf: Allow the new syncookie helpers to work with SKBs
selftests/bpf: Add selftests for raw syncookie helpers
bpf: Add helpers to issue and check SYN cookies in XDP
bpf: Allow helpers to accept pointers with a fixed size
bpf: Fix documentation of th_len in bpf_tcp_{gen,check}_syncookie
selftests/bpf: add tests for sleepable (uk)probes
libbpf: add support for sleepable uprobe programs
bpf: allow sleepable uprobe programs to attach
bpf: implement sleepable uprobes by chaining gps
bpf: move bpf_prog to bpf.h
libbpf: Fix internal USDT address translation logic for shared libraries
samples/bpf: Check detach prog exist or not in xdp_fwd
selftests/bpf: Avoid skipping certain subtests
selftests/bpf: Fix test_varlen verification failure with latest llvm
bpftool: Do not check return value from libbpf_set_strict_mode()
Revert "bpftool: Use libbpf 1.0 API mode instead of RLIMIT_MEMLOCK"
...
====================
Link: https://lore.kernel.org/r/20220617220836.7373-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The new helpers bpf_tcp_raw_{gen,check}_syncookie_ipv{4,6} allow an XDP
program to generate SYN cookies in response to TCP SYN packets and to
check those cookies upon receiving the first ACK packet (the final
packet of the TCP handshake).
Unlike bpf_tcp_{gen,check}_syncookie these new helpers don't need a
listening socket on the local machine, which allows to use them together
with synproxy to accelerate SYN cookie generation.
Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://lore.kernel.org/r/20220615134847.3753567-4-maximmi@nvidia.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Each protocol having a ->memory_allocated pointer gets a corresponding
per-cpu reserve, that following patches will use.
Instead of having reserved bytes per socket,
we want to have per-cpu reserves.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Logic added in commit f35f821935 ("tcp: defer skb freeing after socket
lock is released") helped bulk TCP flows to move the cost of skbs
frees outside of critical section where socket lock was held.
But for RPC traffic, or hosts with RFS enabled, the solution is far from
being ideal.
For RPC traffic, recvmsg() has to return to user space right after
skb payload has been consumed, meaning that BH handler has no chance
to pick the skb before recvmsg() thread. This issue is more visible
with BIG TCP, as more RPC fit one skb.
For RFS, even if BH handler picks the skbs, they are still picked
from the cpu on which user thread is running.
Ideally, it is better to free the skbs (and associated page frags)
on the cpu that originally allocated them.
This patch removes the per socket anchor (sk->defer_list) and
instead uses a per-cpu list, which will hold more skbs per round.
This new per-cpu list is drained at the end of net_action_rx(),
after incoming packets have been processed, to lower latencies.
In normal conditions, skbs are added to the per-cpu list with
no further action. In the (unlikely) cases where the cpu does not
run net_action_rx() handler fast enough, we use an IPI to raise
NET_RX_SOFTIRQ on the remote cpu.
Also, we do not bother draining the per-cpu list from dev_cpu_dead()
This is because skbs in this list have no requirement on how fast
they should be freed.
Note that we can add in the future a small per-cpu cache
if we see any contention on sd->defer_lock.
Tested on a pair of hosts with 100Gbit NIC, RFS enabled,
and /proc/sys/net/ipv4/tcp_rmem[2] tuned to 16MB to work around
page recycling strategy used by NIC driver (its page pool capacity
being too small compared to number of skbs/pages held in sockets
receive queues)
Note that this tuning was only done to demonstrate worse
conditions for skb freeing for this particular test.
These conditions can happen in more general production workload.
10 runs of one TCP_STREAM flow
Before:
Average throughput: 49685 Mbit.
Kernel profiles on cpu running user thread recvmsg() show high cost for
skb freeing related functions (*)
57.81% [kernel] [k] copy_user_enhanced_fast_string
(*) 12.87% [kernel] [k] skb_release_data
(*) 4.25% [kernel] [k] __free_one_page
(*) 3.57% [kernel] [k] __list_del_entry_valid
1.85% [kernel] [k] __netif_receive_skb_core
1.60% [kernel] [k] __skb_datagram_iter
(*) 1.59% [kernel] [k] free_unref_page_commit
(*) 1.16% [kernel] [k] __slab_free
1.16% [kernel] [k] _copy_to_iter
(*) 1.01% [kernel] [k] kfree
(*) 0.88% [kernel] [k] free_unref_page
0.57% [kernel] [k] ip6_rcv_core
0.55% [kernel] [k] ip6t_do_table
0.54% [kernel] [k] flush_smp_call_function_queue
(*) 0.54% [kernel] [k] free_pcppages_bulk
0.51% [kernel] [k] llist_reverse_order
0.38% [kernel] [k] process_backlog
(*) 0.38% [kernel] [k] free_pcp_prepare
0.37% [kernel] [k] tcp_recvmsg_locked
(*) 0.37% [kernel] [k] __list_add_valid
0.34% [kernel] [k] sock_rfree
0.34% [kernel] [k] _raw_spin_lock_irq
(*) 0.33% [kernel] [k] __page_cache_release
0.33% [kernel] [k] tcp_v6_rcv
(*) 0.33% [kernel] [k] __put_page
(*) 0.29% [kernel] [k] __mod_zone_page_state
0.27% [kernel] [k] _raw_spin_lock
After patch:
Average throughput: 73076 Mbit.
Kernel profiles on cpu running user thread recvmsg() looks better:
81.35% [kernel] [k] copy_user_enhanced_fast_string
1.95% [kernel] [k] _copy_to_iter
1.95% [kernel] [k] __skb_datagram_iter
1.27% [kernel] [k] __netif_receive_skb_core
1.03% [kernel] [k] ip6t_do_table
0.60% [kernel] [k] sock_rfree
0.50% [kernel] [k] tcp_v6_rcv
0.47% [kernel] [k] ip6_rcv_core
0.45% [kernel] [k] read_tsc
0.44% [kernel] [k] _raw_spin_lock_irqsave
0.37% [kernel] [k] _raw_spin_lock
0.37% [kernel] [k] native_irq_return_iret
0.33% [kernel] [k] __inet6_lookup_established
0.31% [kernel] [k] ip6_protocol_deliver_rcu
0.29% [kernel] [k] tcp_rcv_established
0.29% [kernel] [k] llist_reverse_order
v2: kdoc issue (kernel bots)
do not defer if (alloc_cpu == smp_processor_id()) (Paolo)
replace the sk_buff_head with a single-linked list (Jakub)
add a READ_ONCE()/WRITE_ONCE() for the lockless read of sd->defer_list
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Link: https://lore.kernel.org/r/20220422201237.416238-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
I had this bug sitting for too long in my pile, it is time to fix it.
Thanks to Doug Porter for reminding me of it!
We had various attempts in the past, including commit
0cbe6a8f08 ("tcp: remove SOCK_QUEUE_SHRUNK"),
but the issue is that TCP stack currently only generates
EPOLLOUT from input path, when tp->snd_una has advanced
and skb(s) cleaned from rtx queue.
If a flow has a big RTT, and/or receives SACKs, it is possible
that the notsent part (tp->write_seq - tp->snd_nxt) reaches 0
and no more data can be sent until tp->snd_una finally advances.
What is needed is to also check if POLLOUT needs to be generated
whenever tp->snd_nxt is advanced, from output path.
This bug triggers more often after an idle period, as
we do not receive ACK for at least one RTT. tcp_notsent_lowat
could be a fraction of what CWND and pacing rate would allow to
send during this RTT.
In a followup patch, I will remove the bogus call
to tcp_chrono_stop(sk, TCP_CHRONO_SNDBUF_LIMITED)
from tcp_check_space(). Fact that we have decided to generate
an EPOLLOUT does not mean the application has immediately
refilled the transmit queue. This optimistic call
might have been the reason the bug seemed not too serious.
Tested:
200 ms rtt, 1% packet loss, 32 MB tcp_rmem[2] and tcp_wmem[2]
$ echo 500000 >/proc/sys/net/ipv4/tcp_notsent_lowat
$ cat bench_rr.sh
SUM=0
for i in {1..10}
do
V=`netperf -H remote_host -l30 -t TCP_RR -- -r 10000000,10000 -o LOCAL_BYTES_SENT | egrep -v "MIGRATED|Bytes"`
echo $V
SUM=$(($SUM + $V))
done
echo SUM=$SUM
Before patch:
$ bench_rr.sh
130000000
80000000
140000000
140000000
140000000
140000000
130000000
40000000
90000000
110000000
SUM=1140000000
After patch:
$ bench_rr.sh
430000000
590000000
530000000
450000000
450000000
350000000
450000000
490000000
480000000
460000000
SUM=4680000000 # This is 410 % of the value before patch.
Fixes: c9bee3b7fd ("tcp: TCP_NOTSENT_LOWAT socket option")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Doug Porter <dsp@fb.com>
Cc: Soheil Hassas Yeganeh <soheil@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If an ACK (s)acks multiple skbs, we favor the information
from the most recently sent skb by choosing the skb with
the highest prior_delivered count. But in the interval
between receiving ACKs, we send multiple skbs with the same
prior_delivered, because the tp->delivered only changes
when we receive an ACK.
We used RACK's solution, copying tcp_rack_sent_after() as
tcp_skb_sent_after() helper to determine "which packet was
sent last?". Later, we will use tcp_skb_sent_after() instead
in RACK.
Fixes: b9f64820fb ("tcp: track data delivery rate for a TCP connection")
Signed-off-by: Pengcheng Yang <yangpc@wangsu.com>
Cc: Paolo Abeni <pabeni@redhat.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Tested-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/1650422081-22153-1-git-send-email-yangpc@wangsu.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The internal recvmsg() functions have two parameters 'flags' and 'noblock'
that were merged inside skb_recv_datagram(). As a follow up patch to commit
f4b41f062c ("net: remove noblock parameter from skb_recv_datagram()")
this patch removes the separate 'noblock' parameter for recvmsg().
Analogue to the referenced patch for skb_recv_datagram() the 'flags' and
'noblock' parameters are unnecessarily split up with e.g.
err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
flags & ~MSG_DONTWAIT, &addr_len);
or in
err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udp_recvmsg,
sk, msg, size, flags & MSG_DONTWAIT,
flags & ~MSG_DONTWAIT, &addr_len);
instead of simply using only flags all the time and check for MSG_DONTWAIT
where needed (to preserve for the formerly separated no(n)block condition).
Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Link: https://lore.kernel.org/r/20220411124955.154876-1-socketcan@hartkopp.net
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The congestion status of a tcp flow may be updated since there
is congestion between tcp sender and receiver. It makes sense to
add tracepoint for congestion status set function to summate cc
status duration and evaluate the performance of network
and congestion algorithm. the backgound of this patch is below.
Link: https://github.com/iovisor/bcc/pull/3899
Signed-off-by: Ping Gan <jacky_gam_2001@163.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20220406010956.19656-1-jacky_gam_2001@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We had various bugs over the years with code
breaking the assumption that tp->snd_cwnd is greater
than zero.
Lately, syzbot reported the WARN_ON_ONCE(!tp->prior_cwnd) added
in commit 8b8a321ff7 ("tcp: fix zero cwnd in tcp_cwnd_reduction")
can trigger, and without a repro we would have to spend
considerable time finding the bug.
Instead of complaining too late, we want to catch where
and when tp->snd_cwnd is set to an illegal value.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Suggested-by: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Link: https://lore.kernel.org/r/20220405233538.947344-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When CONFIG_TCP_MD5SIG isn't enabled, there is a compilation bug due to
the fact that the static inline definition of tcp_inbound_md5_hash() has
an unexpected semicolon. Remove it.
Fixes: 1330b6ef33 ("skb: make drop reason booleanable")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://lore.kernel.org/r/20220309122012.668986-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We have a number of cases where function returns drop/no drop
decision as a boolean. Now that we want to report the reason
code as well we have to pass extra output arguments.
We can make the reason code evaluate correctly as bool.
I believe we're good to reorder the reasons as they are
reported to user space as strings.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The functions do essentially the same work to verify TCP-MD5 sign.
Code can be merged into one family-independent function in order to
reduce copy'n'paste and generated code.
Later with TCP-AO option added, this will allow to create one function
that's responsible for segment verification, that will have all the
different checks for MD5/AO/non-signed packets, which in turn will help
to see checks for all corner-cases in one function, rather than spread
around different families and functions.
Cc: Eric Dumazet <edumazet@google.com>
Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Signed-off-by: Dmitry Safonov <dima@arista.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20220223175740.452397-1-dima@arista.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pass the address of drop_reason to tcp_add_backlog() to store the
reasons for skb drops when fails. Following drop reasons are
introduced:
SKB_DROP_REASON_SOCKET_BACKLOG
Reviewed-by: Mengen Sun <mengensun@tencent.com>
Reviewed-by: Hao Peng <flyingpeng@tencent.com>
Signed-off-by: Menglong Dong <imagedong@tencent.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
When setting RTO through BPF program, some SYN ACK packets were unaffected
and continued to use TCP_TIMEOUT_INIT constant. This patch adds timeout
option to struct request_sock. Option is initialized with TCP_TIMEOUT_INIT
and is reassigned through BPF using tcp_timeout_init call. SYN ACK
retransmits now use newly added timeout option.
Signed-off-by: Akhmat Karakotov <hmukos@yandex-team.ru>
Acked-by: Martin KaFai Lau <kafai@fb.com>
v2:
- Add timeout option to struct request_sock. Do not call
tcp_timeout_init on every syn ack retransmit.
v3:
- Use unsigned long for min. Bound tcp_timeout_init to TCP_RTO_MAX.
v4:
- Refactor duplicate code by adding reqsk_timeout function.
Signed-off-by: David S. Miller <davem@davemloft.net>
When compiling the kernel with CONFIG_INET disabled, the
sk_defer_free_flush() should be defined as a nop.
This resolves the following compilation error:
ld: net/core/sock.o: in function `sk_defer_free_flush':
./include/net/tcp.h:1378: undefined reference to `__sk_defer_free_flush'
Fixes: 79074a72d3 ("net: Flush deferred skb free on socket destroy")
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Gal Pressman <gal@nvidia.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20220120123440.9088-1-gal@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
tcp recvmsg() (or rx zerocopy) spends a fair amount of time
freeing skbs after their payload has been consumed.
A typical ~64KB GRO packet has to release ~45 page
references, eventually going to page allocator
for each of them.
Currently, this freeing is performed while socket lock
is held, meaning that there is a high chance that
BH handler has to queue incoming packets to tcp socket backlog.
This can cause additional latencies, because the user
thread has to process the backlog at release_sock() time,
and while doing so, additional frames can be added
by BH handler.
This patch adds logic to defer these frees after socket
lock is released, or directly from BH handler if possible.
Being able to free these skbs from BH handler helps a lot,
because this avoids the usual alloc/free assymetry,
when BH handler and user thread do not run on same cpu or
NUMA node.
One cpu can now be fully utilized for the kernel->user copy,
and another cpu is handling BH processing and skb/page
allocs/frees (assuming RFS is not forcing use of a single CPU)
Tested:
100Gbit NIC
Max throughput for one TCP_STREAM flow, over 10 runs
MTU : 1500
Before: 55 Gbit
After: 66 Gbit
MTU : 4096+(headers)
Before: 82 Gbit
After: 95 Gbit
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
tcp_segs_in() can be called from BH, while socket spinlock
is held but socket owned by user, eventually reading these
fields from tcp_get_info()
Found by code inspection, no need to backport this patch
to older kernels.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Track skbs containing only zerocopy data and avoid charging them to
kernel memory to correctly account the memory utilization for
msg_zerocopy. All of the data in such skbs is held in user pages which
are already accounted to user. Before this change, they are charged
again in kernel in __zerocopy_sg_from_iter. The charging in kernel is
excessive because data is not being copied into skb frags. This
excessive charging can lead to kernel going into memory pressure
state which impacts all sockets in the system adversely. Mark pure
zerocopy skbs with a SKBFL_PURE_ZEROCOPY flag and remove
charge/uncharge for data in such skbs.
Initially, an skb is marked pure zerocopy when it is empty and in
zerocopy path. skb can then change from a pure zerocopy skb to mixed
data skb (zerocopy and copy data) if it is at tail of write queue and
there is room available in it and non-zerocopy data is being sent in
the next sendmsg call. At this time sk_mem_charge is done for the pure
zerocopied data and the pure zerocopy flag is unmarked. We found that
this happens very rarely on workloads that pass MSG_ZEROCOPY.
A pure zerocopy skb can later be coalesced into normal skb if they are
next to each other in queue but this patch prevents coalescing from
happening. This avoids complexity of charging when skb downgrades from
pure zerocopy to mixed. This is also rare.
In sk_wmem_free_skb, if it is a pure zerocopy skb, an sk_mem_uncharge
for SKB_TRUESIZE(skb_end_offset(skb)) is done for sk_mem_charge in
tcp_skb_entail for an skb without data.
Testing with the msg_zerocopy.c benchmark between two hosts(100G nics)
with zerocopy showed that before this patch the 'sock' variable in
memory.stat for cgroup2 that tracks sum of sk_forward_alloc,
sk_rmem_alloc and sk_wmem_queued is around 1822720 and with this
change it is 0. This is due to no charge to sk_forward_alloc for
zerocopy data and shows memory utilization for kernel is lowered.
With this commit we don't see the warning we saw in previous commit
which resulted in commit 84882cf72c.
Signed-off-by: Talal Ahmad <talalahmad@google.com>
Acked-by: Arjun Roy <arjunroy@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
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>
Track skbs with only zerocopy data and avoid charging them to kernel
memory to correctly account the memory utilization for msg_zerocopy.
All of the data in such skbs is held in user pages which are already
accounted to user. Before this change, they are charged again in
kernel in __zerocopy_sg_from_iter. The charging in kernel is
excessive because data is not being copied into skb frags. This
excessive charging can lead to kernel going into memory pressure
state which impacts all sockets in the system adversely. Mark pure
zerocopy skbs with a SKBFL_PURE_ZEROCOPY flag and remove
charge/uncharge for data in such skbs.
Initially, an skb is marked pure zerocopy when it is empty and in
zerocopy path. skb can then change from a pure zerocopy skb to mixed
data skb (zerocopy and copy data) if it is at tail of write queue and
there is room available in it and non-zerocopy data is being sent in
the next sendmsg call. At this time sk_mem_charge is done for the pure
zerocopied data and the pure zerocopy flag is unmarked. We found that
this happens very rarely on workloads that pass MSG_ZEROCOPY.
A pure zerocopy skb can later be coalesced into normal skb if they are
next to each other in queue but this patch prevents coalescing from
happening. This avoids complexity of charging when skb downgrades from
pure zerocopy to mixed. This is also rare.
In sk_wmem_free_skb, if it is a pure zerocopy skb, an sk_mem_uncharge
for SKB_TRUESIZE(MAX_TCP_HEADER) is done for sk_mem_charge in
tcp_skb_entail for an skb without data.
Testing with the msg_zerocopy.c benchmark between two hosts(100G nics)
with zerocopy showed that before this patch the 'sock' variable in
memory.stat for cgroup2 that tracks sum of sk_forward_alloc,
sk_rmem_alloc and sk_wmem_queued is around 1822720 and with this
change it is 0. This is due to no charge to sk_forward_alloc for
zerocopy data and shows memory utilization for kernel is lowered.
Signed-off-by: Talal Ahmad <talalahmad@google.com>
Acked-by: Arjun Roy <arjunroy@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
sk_wmem_free_skb() is only used by TCP.
Rename it to make this clear, and move its declaration to
include/net/tcp.h
Signed-off-by: Talal Ahmad <talalahmad@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Acked-by: Arjun Roy <arjunroy@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
All tcp_remove_empty_skb() callers now use tcp_write_queue_tail()
for the skb argument, we can therefore factorize code.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sk_stream_alloc_skb() is only used by TCP.
Rename it to make this clear, and move its declaration
to include/net/tcp.h
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Multiple VRFs are generally meant to be "separate" but right now md5
keys for the default VRF also affect connections inside VRFs if the IP
addresses happen to overlap.
So far the combination of TCP_MD5SIG_FLAG_IFINDEX with tcpm_ifindex == 0
was an error, accept this to mean "key only applies to default VRF".
This is what applications using VRFs for traffic separation want.
Signed-off-by: Leonard Crestez <cdleonard@gmail.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use of percpu_counter structure to track count of orphaned
sockets is causing problems on modern hosts with 256 cpus
or more.
Stefan Bach reported a serious spinlock contention in real workloads,
that I was able to reproduce with a netfilter rule dropping
incoming FIN packets.
53.56% server [kernel.kallsyms] [k] queued_spin_lock_slowpath
|
---queued_spin_lock_slowpath
|
--53.51%--_raw_spin_lock_irqsave
|
--53.51%--__percpu_counter_sum
tcp_check_oom
|
|--39.03%--__tcp_close
| tcp_close
| inet_release
| inet6_release
| sock_close
| __fput
| ____fput
| task_work_run
| exit_to_usermode_loop
| do_syscall_64
| entry_SYSCALL_64_after_hwframe
| __GI___libc_close
|
--14.48%--tcp_out_of_resources
tcp_write_timeout
tcp_retransmit_timer
tcp_write_timer_handler
tcp_write_timer
call_timer_fn
expire_timers
__run_timers
run_timer_softirq
__softirqentry_text_start
As explained in commit cf86a086a1 ("net/dst: use a smaller percpu_counter
batch for dst entries accounting"), default batch size is too big
for the default value of tcp_max_orphans (262144).
But even if we reduce batch sizes, there would still be cases
where the estimated count of orphans is beyond the limit,
and where tcp_too_many_orphans() has to call the expensive
percpu_counter_sum_positive().
One solution is to use plain per-cpu counters, and have
a timer to periodically refresh this cache.
Updating this cache every 100ms seems about right, tcp pressure
state is not radically changing over shorter periods.
percpu_counter was nice 15 years ago while hosts had less
than 16 cpus, not anymore by current standards.
v2: Fix the build issue for CONFIG_CRYPTO_DEV_CHELSIO_TLS=m,
reported by kernel test robot <lkp@intel.com>
Remove unused socket argument from tcp_too_many_orphans()
Fixes: dd24c00191 ("net: Use a percpu_counter for orphan_count")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Stefan Bach <sfb@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When user sets SO_RESERVE_MEM socket option, in order to utilize the
reserved memory when in memory pressure state, we adjust rcv_ssthresh
according to the available reserved memory for the socket, instead of
using 4 * advmss always.
Signed-off-by: Wei Wang <weiwan@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to track CE marks per rate sample (one round trip), TCP needs a
per-skb header field to record the tp->delivered_ce count when the skb
was sent. To make space, we replace the "last_in_flight" field which is
used exclusively for NV congestion control. The stat needed by NV can be
alternatively approximated by existing stats tcp_sock delivered and
mss_cache.
This patch counts the number of packets delivered which have CE marks in
the rate sample, using similar approach of delivery accounting.
Cc: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Luke Hsiao <lukehsiao@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After the previous patch the mentioned helper is
used only inside its compilation unit: let's make
it static.
RFC -> v1:
- preserve the tcp_build_frag() helper (Eric)
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
the tcp_skb_entail() helper is actually skb_entail(), renamed
to provide proper scope.
The two helper will be used by the next patch.
RFC -> v1:
- rename skb_entail to tcp_skb_entail (Eric)
Acked-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Andrii Nakryiko says:
====================
bpf-next 2021-07-30
We've added 64 non-merge commits during the last 15 day(s) which contain
a total of 83 files changed, 5027 insertions(+), 1808 deletions(-).
The main changes are:
1) BTF-guided binary data dumping libbpf API, from Alan.
2) Internal factoring out of libbpf CO-RE relocation logic, from Alexei.
3) Ambient BPF run context and cgroup storage cleanup, from Andrii.
4) Few small API additions for libbpf 1.0 effort, from Evgeniy and Hengqi.
5) bpf_program__attach_kprobe_opts() fixes in libbpf, from Jiri.
6) bpf_{get,set}sockopt() support in BPF iterators, from Martin.
7) BPF map pinning improvements in libbpf, from Martynas.
8) Improved module BTF support in libbpf and bpftool, from Quentin.
9) Bpftool cleanups and documentation improvements, from Quentin.
10) Libbpf improvements for supporting CO-RE on old kernels, from Shuyi.
11) Increased maximum cgroup storage size, from Stanislav.
12) Small fixes and improvements to BPF tests and samples, from various folks.
* https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (64 commits)
tools: bpftool: Complete metrics list in "bpftool prog profile" doc
tools: bpftool: Document and add bash completion for -L, -B options
selftests/bpf: Update bpftool's consistency script for checking options
tools: bpftool: Update and synchronise option list in doc and help msg
tools: bpftool: Complete and synchronise attach or map types
selftests/bpf: Check consistency between bpftool source, doc, completion
tools: bpftool: Slightly ease bash completion updates
unix_bpf: Fix a potential deadlock in unix_dgram_bpf_recvmsg()
libbpf: Add btf__load_vmlinux_btf/btf__load_module_btf
tools: bpftool: Support dumping split BTF by id
libbpf: Add split BTF support for btf__load_from_kernel_by_id()
tools: Replace btf__get_from_id() with btf__load_from_kernel_by_id()
tools: Free BTF objects at various locations
libbpf: Rename btf__get_from_id() as btf__load_from_kernel_by_id()
libbpf: Rename btf__load() as btf__load_into_kernel()
libbpf: Return non-null error on failures in libbpf_find_prog_btf_id()
bpf: Emit better log message if bpf_iter ctx arg btf_id == 0
tools/resolve_btfids: Emit warnings and patch zero id for missing symbols
bpf: Increase supported cgroup storage value size
libbpf: Fix race when pinning maps in parallel
...
====================
Link: https://lore.kernel.org/r/20210730225606.1897330-1-andrii@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
A following patch will create a separate struct to store extra
bpf_iter state and it will embed the existing tcp_iter_state like this:
struct bpf_tcp_iter_state {
struct tcp_iter_state state;
/* More bpf_iter specific states here ... */
}
As a prep work, this patch removes the
"struct tcp_seq_afinfo *bpf_seq_afinfo" where its purpose is
to tell if it is iterating from bpf_iter instead of proc fs.
Currently, if "*bpf_seq_afinfo" is not NULL, it is iterating from
bpf_iter. The kernel should not filter by the addr family and
leave this filtering decision to the bpf prog.
Instead of adding a "*bpf_seq_afinfo" pointer, this patch uses the
"seq->op == &bpf_iter_tcp_seq_ops" test to tell if it is iterating
from the bpf iter.
The bpf_iter_(init|fini)_tcp() is left here to prepare for
the change of a following patch.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Kuniyuki Iwashima <kuniyu@amazon.co.jp>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20210701200554.1034982-1-kafai@fb.com
After cited commit, sysctl_tcp_fastopen_blackhole_timeout is no longer
a global variable.
Fixes: 3733be14a3 ("ipv4: Namespaceify tcp_fastopen_blackhole_timeout knob")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Haishuang Yan <yanhaishuang@cmss.chinamobile.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Acked-by: Wei Wang <weiwan@google.com>
Link: https://lore.kernel.org/r/20210719092028.3016745-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The MPTCP receive path is hooked only into the TCP slow-path.
The DSS presence allows plain MPTCP traffic to hit that
consistently.
Since commit e1ff9e82e2 ("net: mptcp: improve fallback to TCP"),
when an MPTCP socket falls back to TCP, it can hit the TCP receive
fast-path, and delay or stop triggering the event notification.
Address the issue explicitly disabling the header prediction
for MPTCP sockets.
Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/200
Fixes: e1ff9e82e2 ("net: mptcp: improve fallback to TCP")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
MPTCP is builtin, so no need to add EXPORT_SYMBOL()s.
It will be used to support SO_TIMESTAMP(NS) ancillary
messages in the mptcp receive path.
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
