This idea came after a particular workload requested
the quickack attribute set on routes, and a performance
drop was noticed for large bulk transfers.
For high throughput flows, it is best to use one cpu
running the user thread issuing socket system calls,
and a separate cpu to process incoming packets from BH context.
(With TSO/GRO, bottleneck is usually the 'user' cpu)
Problem is the user thread can spend a lot of time while holding
the socket lock, forcing BH handler to queue most of incoming
packets in the socket backlog.
Whenever the user thread releases the socket lock, it must first
process all accumulated packets in the backlog, potentially
adding latency spikes. Due to flood mitigation, having too many
packets in the backlog increases chance of unexpected drops.
Backlog processing unfortunately shifts a fair amount of cpu cycles
from the BH cpu to the 'user' cpu, thus reducing max throughput.
This patch takes advantage of the backlog processing,
and the fact that ACK are mostly cumulative.
The idea is to detect we are in the backlog processing
and defer all eligible ACK into a single one,
sent from tcp_release_cb().
This saves cpu cycles on both sides, and network resources.
Performance of a single TCP flow on a 200Gbit NIC:
- Throughput is increased by 20% (100Gbit -> 120Gbit).
- Number of generated ACK per second shrinks from 240,000 to 40,000.
- Number of backlog drops per second shrinks from 230 to 0.
Benchmark context:
- Regular netperf TCP_STREAM (no zerocopy)
- Intel(R) Xeon(R) Platinum 8481C (Saphire Rapids)
- MAX_SKB_FRAGS = 17 (~60KB per GRO packet)
This feature is guarded by a new sysctl, and enabled by default:
/proc/sys/net/ipv4/tcp_backlog_ack_defer
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Acked-by: Dave Taht <dave.taht@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
With modern NIC drivers shifting to full page allocations per
received frame, we face the following issue:
TCP has one per-netns sysctl used to tweak how to translate
a memory use into an expected payload (RWIN), in RX path.
tcp_win_from_space() implementation is limited to few cases.
For hosts dealing with various MSS, we either under estimate
or over estimate the RWIN we send to the remote peers.
For instance with the default sysctl_tcp_adv_win_scale value,
we expect to store 50% of payload per allocated chunk of memory.
For the typical use of MTU=1500 traffic, and order-0 pages allocations
by NIC drivers, we are sending too big RWIN, leading to potential
tcp collapse operations, which are extremely expensive and source
of latency spikes.
This patch makes sysctl_tcp_adv_win_scale obsolete, and instead
uses a per socket scaling factor, so that we can precisely
adjust the RWIN based on effective skb->len/skb->truesize ratio.
This patch alone can double TCP receive performance when receivers
are too slow to drain their receive queue, or by allowing
a bigger RWIN when MSS is close to PAGE_SIZE.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Link: https://lore.kernel.org/r/20230717152917.751987-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Under certain circumstances, the tcp receive buffer memory limit
set by autotuning (sk_rcvbuf) is increased due to incoming data
packets as a result of the window not closing when it should be.
This can result in the receive buffer growing all the way up to
tcp_rmem[2], even for tcp sessions with a low BDP.
To reproduce: Connect a TCP session with the receiver doing
nothing and the sender sending small packets (an infinite loop
of socket send() with 4 bytes of payload with a sleep of 1 ms
in between each send()). This will cause the tcp receive buffer
to grow all the way up to tcp_rmem[2].
As a result, a host can have individual tcp sessions with receive
buffers of size tcp_rmem[2], and the host itself can reach tcp_mem
limits, causing the host to go into tcp memory pressure mode.
The fundamental issue is the relationship between the granularity
of the window scaling factor and the number of byte ACKed back
to the sender. This problem has previously been identified in
RFC 7323, appendix F [1].
The Linux kernel currently adheres to never shrinking the window.
In addition to the overallocation of memory mentioned above, the
current behavior is functionally incorrect, because once tcp_rmem[2]
is reached when no remediations remain (i.e. tcp collapse fails to
free up any more memory and there are no packets to prune from the
out-of-order queue), the receiver will drop in-window packets
resulting in retransmissions and an eventual timeout of the tcp
session. A receive buffer full condition should instead result
in a zero window and an indefinite wait.
In practice, this problem is largely hidden for most flows. It
is not applicable to mice flows. Elephant flows can send data
fast enough to "overrun" the sk_rcvbuf limit (in a single ACK),
triggering a zero window.
But this problem does show up for other types of flows. Examples
are websockets and other type of flows that send small amounts of
data spaced apart slightly in time. In these cases, we directly
encounter the problem described in [1].
RFC 7323, section 2.4 [2], says there are instances when a retracted
window can be offered, and that TCP implementations MUST ensure
that they handle a shrinking window, as specified in RFC 1122,
section 4.2.2.16 [3]. All prior RFCs on the topic of tcp window
management have made clear that sender must accept a shrunk window
from the receiver, including RFC 793 [4] and RFC 1323 [5].
This patch implements the functionality to shrink the tcp window
when necessary to keep the right edge within the memory limit by
autotuning (sk_rcvbuf). This new functionality is enabled with
the new sysctl: net.ipv4.tcp_shrink_window
Additional information can be found at:
https://blog.cloudflare.com/unbounded-memory-usage-by-tcp-for-receive-buffers-and-how-we-fixed-it/
[1] https://www.rfc-editor.org/rfc/rfc7323#appendix-F
[2] https://www.rfc-editor.org/rfc/rfc7323#section-2.4
[3] https://www.rfc-editor.org/rfc/rfc1122#page-91
[4] https://www.rfc-editor.org/rfc/rfc793
[5] https://www.rfc-editor.org/rfc/rfc1323
Signed-off-by: Mike Freemon <mfreemon@cloudflare.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently the SYN RTO schedule follows an exponential backoff
scheme, which can be unnecessarily conservative in cases where
there are link failures. In such cases, it's better to
aggressively try to retransmit packets, so it takes routers
less time to find a repath with a working link.
We chose a default value for this sysctl of 4, to follow
the macOS and IOS backoff scheme of 1,1,1,1,1,2,4,8, ...
MacOS and IOS have used this backoff schedule for over
a decade, since before this 2009 IETF presentation
discussed the behavior:
https://www.ietf.org/proceedings/75/slides/tcpm-1.pdf
This commit makes the SYN RTO schedule start with a number of
linear backoffs given by the following sysctl:
* tcp_syn_linear_timeouts
This changes the SYN RTO scheme to be: init_rto_val for
tcp_syn_linear_timeouts, exp backoff starting at init_rto_val
For example if init_rto_val = 1 and tcp_syn_linear_timeouts = 2, our
backoff scheme would be: 1, 1, 1, 2, 4, 8, 16, ...
Signed-off-by: David Morley <morleyd@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Tested-by: David Morley <morleyd@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20230509180558.2541885-1-morleyd.kernel@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The maximum hash table size is 64K due to the nature of the protocol. [0]
It's smaller than TCP, and fewer sockets can cause a performance drop.
On an EC2 c5.24xlarge instance (192 GiB memory), after running iperf3 in
different netns, creating 32Mi sockets without data transfer in the root
netns causes regression for the iperf3's connection.
uhash_entries sockets length Gbps
64K 1 1 5.69
1Mi 16 5.27
2Mi 32 4.90
4Mi 64 4.09
8Mi 128 2.96
16Mi 256 2.06
32Mi 512 1.12
The per-netns hash table breaks the lengthy lists into shorter ones. It is
useful on a multi-tenant system with thousands of netns. With smaller hash
tables, we can look up sockets faster, isolate noisy neighbours, and reduce
lock contention.
The max size of the per-netns table is 64K as well. This is because the
possible hash range by udp_hashfn() always fits in 64K within the same
netns and we cannot make full use of the whole buckets larger than 64K.
/* 0 < num < 64K -> X < hash < X + 64K */
(num + net_hash_mix(net)) & mask;
Also, the min size is 128. We use a bitmap to search for an available
port in udp_lib_get_port(). To keep the bitmap on the stack and not
fire the CONFIG_FRAME_WARN error at build time, we round up the table
size to 128.
The sysctl usage is the same with TCP:
$ dmesg | cut -d ' ' -f 6- | grep "UDP hash"
UDP hash table entries: 65536 (order: 9, 2097152 bytes, vmalloc)
# sysctl net.ipv4.udp_hash_entries
net.ipv4.udp_hash_entries = 65536 # can be changed by uhash_entries
# sysctl net.ipv4.udp_child_hash_entries
net.ipv4.udp_child_hash_entries = 0 # disabled by default
# ip netns add test1
# ip netns exec test1 sysctl net.ipv4.udp_hash_entries
net.ipv4.udp_hash_entries = -65536 # share the global table
# sysctl -w net.ipv4.udp_child_hash_entries=100
net.ipv4.udp_child_hash_entries = 100
# ip netns add test2
# ip netns exec test2 sysctl net.ipv4.udp_hash_entries
net.ipv4.udp_hash_entries = 128 # own a per-netns table with 2^n buckets
We could optimise the hash table lookup/iteration further by removing
the netns comparison for the per-netns one in the future. Also, we
could optimise the sparse udp_hslot layout by putting it in udp_table.
[0]: https://lore.kernel.org/netdev/4ACC2815.7010101@gmail.com/
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We will soon introduce an optional per-netns hash table
for UDP.
This means we cannot use the global sk->sk_prot->h.udp_table
to fetch a UDP hash table.
Instead, set NULL to sk->sk_prot->h.udp_table for UDP and get
a proper table from net->ipv4.udp_table.
Note that we still need sk->sk_prot->h.udp_table for UDP LITE.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
PLB (Protective Load Balancing) is a host based mechanism for load
balancing across switch links. It leverages congestion signals(e.g. ECN)
from transport layer to randomly change the path of the connection
experiencing congestion. PLB changes the path of the connection by
changing the outgoing IPv6 flow label for IPv6 connections (implemented
in Linux by calling sk_rethink_txhash()). Because of this implementation
mechanism, PLB can currently only work for IPv6 traffic. For more
information, see the SIGCOMM 2022 paper:
https://doi.org/10.1145/3544216.3544226
This commit adds new sysctl knobs and sets their default values for
TCP PLB.
Signed-off-by: Mubashir Adnan Qureshi <mubashirq@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The more sockets we have in the hash table, the longer we spend looking
up the socket. While running a number of small workloads on the same
host, they penalise each other and cause performance degradation.
The root cause might be a single workload that consumes much more
resources than the others. It often happens on a cloud service where
different workloads share the same computing resource.
On EC2 c5.24xlarge instance (196 GiB memory and 524288 (1Mi / 2) ehash
entries), after running iperf3 in different netns, creating 24Mi sockets
without data transfer in the root netns causes about 10% performance
regression for the iperf3's connection.
thash_entries sockets length Gbps
524288 1 1 50.7
24Mi 48 45.1
It is basically related to the length of the list of each hash bucket.
For testing purposes to see how performance drops along the length,
I set 131072 (1Mi / 8) to thash_entries, and here's the result.
thash_entries sockets length Gbps
131072 1 1 50.7
1Mi 8 49.9
2Mi 16 48.9
4Mi 32 47.3
8Mi 64 44.6
16Mi 128 40.6
24Mi 192 36.3
32Mi 256 32.5
40Mi 320 27.0
48Mi 384 25.0
To resolve the socket lookup degradation, we introduce an optional
per-netns hash table for TCP, but it's just ehash, and we still share
the global bhash, bhash2 and lhash2.
With a smaller ehash, we can look up non-listener sockets faster and
isolate such noisy neighbours. In addition, we can reduce lock contention.
We can control the ehash size by a new sysctl knob. However, depending
on workloads, it will require very sensitive tuning, so we disable the
feature by default (net.ipv4.tcp_child_ehash_entries == 0). Moreover,
we can fall back to using the global ehash in case we fail to allocate
enough memory for a new ehash. The maximum size is 16Mi, which is large
enough that even if we have 48Mi sockets, the average list length is 3,
and regression would be less than 1%.
We can check the current ehash size by another read-only sysctl knob,
net.ipv4.tcp_ehash_entries. A negative value means the netns shares
the global ehash (per-netns ehash is disabled or failed to allocate
memory).
# dmesg | cut -d ' ' -f 5- | grep "established hash"
TCP established hash table entries: 524288 (order: 10, 4194304 bytes, vmalloc hugepage)
# sysctl net.ipv4.tcp_ehash_entries
net.ipv4.tcp_ehash_entries = 524288 # can be changed by thash_entries
# sysctl net.ipv4.tcp_child_ehash_entries
net.ipv4.tcp_child_ehash_entries = 0 # disabled by default
# ip netns add test1
# ip netns exec test1 sysctl net.ipv4.tcp_ehash_entries
net.ipv4.tcp_ehash_entries = -524288 # share the global ehash
# sysctl -w net.ipv4.tcp_child_ehash_entries=100
net.ipv4.tcp_child_ehash_entries = 100
# ip netns add test2
# ip netns exec test2 sysctl net.ipv4.tcp_ehash_entries
net.ipv4.tcp_ehash_entries = 128 # own a per-netns ehash with 2^n buckets
When more than two processes in the same netns create per-netns ehash
concurrently with different sizes, we need to guarantee the size in
one of the following ways:
1) Share the global ehash and create per-netns ehash
First, unshare() with tcp_child_ehash_entries==0. It creates dedicated
netns sysctl knobs where we can safely change tcp_child_ehash_entries
and clone()/unshare() to create a per-netns ehash.
2) Control write on sysctl by BPF
We can use BPF_PROG_TYPE_CGROUP_SYSCTL to allow/deny read/write on
sysctl knobs.
Note that the global ehash allocated at the boot time is spread over
available NUMA nodes, but inet_pernet_hashinfo_alloc() will allocate
pages for each per-netns ehash depending on the current process's NUMA
policy. By default, the allocation is done in the local node only, so
the per-netns hash table could fully reside on a random node. Thus,
depending on the NUMA policy the netns is created with and the CPU the
current thread is running on, we could see some performance differences
for highly optimised networking applications.
Note also that the default values of two sysctl knobs depend on the ehash
size and should be tuned carefully:
tcp_max_tw_buckets : tcp_child_ehash_entries / 2
tcp_max_syn_backlog : max(128, tcp_child_ehash_entries / 128)
As a bonus, we can dismantle netns faster. Currently, while destroying
netns, we call inet_twsk_purge(), which walks through the global ehash.
It can be potentially big because it can have many sockets other than
TIME_WAIT in all netns. Splitting ehash changes that situation, where
it's only necessary for inet_twsk_purge() to clean up TIME_WAIT sockets
in each netns.
With regard to this, we do not free the per-netns ehash in inet_twsk_kill()
to avoid UAF while iterating the per-netns ehash in inet_twsk_purge().
Instead, we do it in tcp_sk_exit_batch() after calling tcp_twsk_purge() to
keep it protocol-family-independent.
In the future, we could optimise ehash lookup/iteration further by removing
netns comparison for the per-netns ehash.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We will soon introduce an optional per-netns ehash and access hash
tables via net->ipv4.tcp_death_row->hashinfo instead of &tcp_hashinfo
in most places.
It could harm the fast path because dereferences of two fields in net
and tcp_death_row might incur two extra cache line misses. To save one
dereference, let's place tcp_death_row back in netns_ipv4 and fetch
hashinfo via net->ipv4.tcp_death_row"."hashinfo.
Note tcp_death_row was initially placed in netns_ipv4, and commit
fbb8295248 ("tcp: allocate tcp_death_row outside of struct netns_ipv4")
changed it to a pointer so that we can fire TIME_WAIT timers after freeing
net. However, we don't do so after commit 04c494e68a ("Revert "tcp/dccp:
get rid of inet_twsk_purge()""), so we need not define tcp_death_row as a
pointer.
Also, we move refcount_dec_and_test(&tw_refcount) from tcp_sk_exit() to
tcp_sk_exit_batch() as a debug check.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Because per host rate limiting has been proven problematic (side channel
attacks can be based on it), per host rate limiting of challenge acks ideally
should be per netns and turned off by default.
This is a long due followup of following commits:
083ae30828 ("tcp: enable per-socket rate limiting of all 'challenge acks'")
f2b2c582e8 ("tcp: mitigate ACK loops for connections as tcp_sock")
75ff39ccc1 ("tcp: make challenge acks less predictable")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Jason Baron <jbaron@akamai.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This patch adds missing includes to headers under include/net.
All these problems are currently masked by the existing users
including the missing dependency before the broken header.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Back when tcp_tso_autosize() and TCP pacing were introduced,
our focus was really to reduce burst sizes for long distance
flows.
The simple heuristic of using sk_pacing_rate/1024 has worked
well, but can lead to too small packets for hosts in the same
rack/cluster, when thousands of flows compete for the bottleneck.
Neal Cardwell had the idea of making the TSO burst size
a function of both sk_pacing_rate and tcp_min_rtt()
Indeed, for local flows, sending bigger bursts is better
to reduce cpu costs, as occasional losses can be repaired
quite fast.
This patch is based on Neal Cardwell implementation
done more than two years ago.
bbr is adjusting max_pacing_rate based on measured bandwidth,
while cubic would over estimate max_pacing_rate.
/proc/sys/net/ipv4/tcp_tso_rtt_log can be used to tune or disable
this new feature, in logarithmic steps.
Tested:
100Gbit NIC, two hosts in the same rack, 4K MTU.
600 flows rate-limited to 20000000 bytes per second.
Before patch: (TSO sizes would be limited to 20000000/1024/4096 -> 4 segments per TSO)
~# echo 0 >/proc/sys/net/ipv4/tcp_tso_rtt_log
~# nstat -n;perf stat ./super_netperf 600 -H otrv6 -l 20 -- -K dctcp -q 20000000;nstat|egrep "TcpInSegs|TcpOutSegs|TcpRetransSegs|Delivered"
96005
Performance counter stats for './super_netperf 600 -H otrv6 -l 20 -- -K dctcp -q 20000000':
65,945.29 msec task-clock # 2.845 CPUs utilized
1,314,632 context-switches # 19935.279 M/sec
5,292 cpu-migrations # 80.249 M/sec
940,641 page-faults # 14264.023 M/sec
201,117,030,926 cycles # 3049769.216 GHz (83.45%)
17,699,435,405 stalled-cycles-frontend # 8.80% frontend cycles idle (83.48%)
136,584,015,071 stalled-cycles-backend # 67.91% backend cycles idle (83.44%)
53,809,530,436 instructions # 0.27 insn per cycle
# 2.54 stalled cycles per insn (83.36%)
9,062,315,523 branches # 137422329.563 M/sec (83.22%)
153,008,621 branch-misses # 1.69% of all branches (83.32%)
23.182970846 seconds time elapsed
TcpInSegs 15648792 0.0
TcpOutSegs 58659110 0.0 # Average of 3.7 4K segments per TSO packet
TcpExtTCPDelivered 58654791 0.0
TcpExtTCPDeliveredCE 19 0.0
After patch:
~# echo 9 >/proc/sys/net/ipv4/tcp_tso_rtt_log
~# nstat -n;perf stat ./super_netperf 600 -H otrv6 -l 20 -- -K dctcp -q 20000000;nstat|egrep "TcpInSegs|TcpOutSegs|TcpRetransSegs|Delivered"
96046
Performance counter stats for './super_netperf 600 -H otrv6 -l 20 -- -K dctcp -q 20000000':
48,982.58 msec task-clock # 2.104 CPUs utilized
186,014 context-switches # 3797.599 M/sec
3,109 cpu-migrations # 63.472 M/sec
941,180 page-faults # 19214.814 M/sec
153,459,763,868 cycles # 3132982.807 GHz (83.56%)
12,069,861,356 stalled-cycles-frontend # 7.87% frontend cycles idle (83.32%)
120,485,917,953 stalled-cycles-backend # 78.51% backend cycles idle (83.24%)
36,803,672,106 instructions # 0.24 insn per cycle
# 3.27 stalled cycles per insn (83.18%)
5,947,266,275 branches # 121417383.427 M/sec (83.64%)
87,984,616 branch-misses # 1.48% of all branches (83.43%)
23.281200256 seconds time elapsed
TcpInSegs 1434706 0.0
TcpOutSegs 58883378 0.0 # Average of 41 4K segments per TSO packet
TcpExtTCPDelivered 58878971 0.0
TcpExtTCPDeliveredCE 9664 0.0
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Link: https://lore.kernel.org/r/20220309015757.2532973-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Different netns has different requirement on the setting of min_adv_mss
sysctl which the advertised MSS will be never lower than.
Enable min_adv_mss to be configured per network namespace.
Signed-off-by: xu xin <xu.xin16@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP ipv4 uses per-cpu/per-netns ctl sockets in order to send
RST and some ACK packets (on behalf of TIMEWAIT sockets).
This adds memory and cpu costs, which do not seem needed.
Now typical servers have 256 or more cores, this adds considerable
tax to netns users.
tcp sockets are used from BH context, are not receiving packets,
and do not store any persistent state but the 'struct net' pointer
in order to be able to use IPv4 output functions.
Note that I attempted a related change in the past, that had
to be hot-fixed in commit bdbbb8527b ("ipv4: tcp: get rid of ugly unicast_sock")
This patch could very well surface old bugs, on layers not
taking care of sk->sk_kern_sock properly.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Back in linux-2.6.25 (commit 4a6ad7a141 "[NETNS]: Make icmp_sk per namespace."),
we added private per-cpu/per-netns ipv4 icmp sockets.
This adds memory and cpu costs, which do not seem needed.
Now typical servers have 256 or more cores, this adds considerable
tax to netns users.
icmp sockets are used from BH context, are not receiving packets,
and do not store any persistent state but the 'struct net' pointer.
icmp_xmit_lock() already makes sure to lock the chosen per-cpu
socket.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch enables the sysctl mtu_expires to be configured per net
namespace.
Signed-off-by: xu xin <xu.xin16@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch enables the sysctl min_pmtu to be configured per net
namespace.
Signed-off-by: xu xin <xu.xin16@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
Before commit faa041a40b ("ipv4: Create cleanup helper for fib_nh")
changes to net->ipv4.fib_num_tclassid_users were protected by RTNL.
After the change, this is no longer the case, as free_fib_info_rcu()
runs after rcu grace period, without rtnl being held.
Fixes: faa041a40b ("ipv4: Create cleanup helper for fib_nh")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: David Ahern <dsahern@kernel.org>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit adds a new sysctl option: net.ipv4.tcp_migrate_req. If this
option is enabled or eBPF program is attached, we will be able to migrate
child sockets from a listener to another in the same reuseport group after
close() or shutdown() syscalls.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.co.jp>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Reviewed-by: Benjamin Herrenschmidt <benh@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Link: https://lore.kernel.org/bpf/20210612123224.12525-2-kuniyu@amazon.co.jp
A subsequent patch will add a new multipath hash policy where the packet
fields used for multipath hash calculation are determined by user space.
This patch adds a sysctl that allows user space to set these fields.
The packet fields are represented using a bitmask and are common between
IPv4 and IPv6 to allow user space to use the same numbering across both
protocols. For example, to hash based on standard 5-tuple:
# sysctl -w net.ipv4.fib_multipath_hash_fields=0x0037
net.ipv4.fib_multipath_hash_fields = 0x0037
The kernel rejects unknown fields, for example:
# sysctl -w net.ipv4.fib_multipath_hash_fields=0x1000
sysctl: setting key "net.ipv4.fib_multipath_hash_fields": Invalid argument
More fields can be added in the future, if needed.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
No longer needed, table pointer arg is now passed via netfilter core.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
By shuffling around some fields to remove 8 bytes of hole,
we can save one cache line.
pahole result before/after the patch :
/* size: 768, cachelines: 12, members: 139 */
/* sum members: 673, holes: 11, sum holes: 39 */
/* padding: 56 */
/* paddings: 2, sum paddings: 7 */
/* forced alignments: 1 */
->
/* size: 704, cachelines: 11, members: 139 */
/* sum members: 673, holes: 10, sum holes: 31 */
/* paddings: 2, sum paddings: 7 */
/* forced alignments: 1 */
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
struct inet_timewait_death_row uses two cache lines, because we want
tw_count to use a full cache line to avoid false sharing.
Rework its definition and placement in netns_ipv4 so that:
1) We add 60 bytes of padding after tw_count to avoid
false sharing, knowing that tcp_death_row will
have ____cacheline_aligned_in_smp attribute.
2) We do not risk padding before tcp_death_row, because
we move it at the beginning of netns_ipv4, even if new
fields are added later.
3) We do not waste 48 bytes of padding after it.
Note that I have not changed dccp.
pahole result for struct netns_ipv4 before/after the patch :
/* size: 832, cachelines: 13, members: 139 */
/* sum members: 721, holes: 12, sum holes: 95 */
/* padding: 16 */
/* paddings: 2, sum paddings: 55 */
->
/* size: 768, cachelines: 12, members: 139 */
/* sum members: 673, holes: 11, sum holes: 39 */
/* padding: 56 */
/* paddings: 2, sum paddings: 7 */
/* forced alignments: 1 */
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Section 8 of RFC 8335 specifies potential security concerns of
responding to PROBE requests, and states that nodes that support PROBE
functionality MUST be able to enable/disable responses and that
responses MUST be disabled by default
Signed-off-by: Andreas Roeseler <andreas.a.roeseler@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Many tcp sysctls are either bools or small ints that can fit into u8.
Reducing space taken by sysctls can save few cache line misses
when sending/receiving data while cpu caches are empty,
for example after cpu idle period.
This is hard to measure with typical network performance tests,
but after this patch, struct netns_ipv4 has shrunk
by three cache lines.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For these sysctls, their dedicated helpers have
to use proc_dou8vec_minmax().
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This sysctl uses ip_fwd_update_priority() helper,
so the conversion needs to change it.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
These sysctls that can fit in one byte instead of one int
are converted to save space and thus reduce cache line misses.
- icmp_echo_ignore_all, icmp_echo_ignore_broadcasts,
- icmp_ignore_bogus_error_responses, icmp_errors_use_inbound_ifaddr
- tcp_ecn, tcp_ecn_fallback
- ip_default_ttl, ip_no_pmtu_disc, ip_fwd_use_pmtu
- ip_nonlocal_bind, ip_autobind_reuse
- ip_dynaddr, ip_early_demux, raw_l3mdev_accept
- nexthop_compat_mode, fwmark_reflect
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After commit 098a697b49 ("tcp_metrics: Use a single hash table
for all network namespaces."), tcpm_hash_bucket is local to
net/ipv4/tcp_metrics.c
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After installing a route to the kernel, user space receives an
acknowledgment, which means the route was installed in the kernel,
but not necessarily in hardware.
The asynchronous nature of route installation in hardware can lead to a
routing daemon advertising a route before it was actually installed in
hardware. This can result in packet loss or mis-routed packets until the
route is installed in hardware.
It is also possible for a route already installed in hardware to change
its action and therefore its flags. For example, a host route that is
trapping packets can be "promoted" to perform decapsulation following
the installation of an IPinIP/VXLAN tunnel.
Emit RTM_NEWROUTE notifications whenever RTM_F_OFFLOAD/RTM_F_TRAP flags
are changed. The aim is to provide an indication to user-space
(e.g., routing daemons) about the state of the route in hardware.
Introduce a sysctl that controls this behavior.
Keep the default value at 0 (i.e., do not emit notifications) for several
reasons:
- Multiple RTM_NEWROUTE notification per-route might confuse existing
routing daemons.
- Convergence reasons in routing daemons.
- The extra notifications will negatively impact the insertion rate.
- Not all users are interested in these notifications.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Acked-by: Roopa Prabhu <roopa@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This commit adds a new TCP feature to reflect the tos value received in
SYN, and send it out on the SYN-ACK, and eventually set the tos value of
the established socket with this reflected tos value. This provides a
way to set the traffic class/QoS level for all traffic in the same
connection to be the same as the incoming SYN request. It could be
useful in data centers to provide equivalent QoS according to the
incoming request.
This feature is guarded by /proc/sys/net/ipv4/tcp_reflect_tos, and is by
default turned off.
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>
Add a sysctl to control hrtimer slack, default of 100 usec.
This gives the opportunity to reduce system overhead,
and help very short RTT flows.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current route nexthop API maintains user space compatibility
with old route API by default. Dumps and netlink notifications
support both new and old API format. In systems which have
moved to the new API, this compatibility mode cancels some
of the performance benefits provided by the new nexthop API.
This patch adds new sysctl nexthop_compat_mode which is on
by default but provides the ability to turn off compatibility
mode allowing systems to run entirely with the new routing
API. Old route API behaviour and support is not modified by this
sysctl.
Uses a single sysctl to cover both ipv4 and ipv6 following
other sysctls. Covers dumps and delete notifications as
suggested by David Ahern.
Signed-off-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit aacd9289af ("tcp: bind() use stronger
condition for bind_conflict") introduced a restriction to forbid to bind
SO_REUSEADDR enabled sockets to the same (addr, port) tuple in order to
assign ports dispersedly so that we can connect to the same remote host.
The change results in accelerating port depletion so that we fail to bind
sockets to the same local port even if we want to connect to the different
remote hosts.
You can reproduce this issue by following instructions below.
1. # sysctl -w net.ipv4.ip_local_port_range="32768 32768"
2. set SO_REUSEADDR to two sockets.
3. bind two sockets to (localhost, 0) and the latter fails.
Therefore, when ephemeral ports are exhausted, bind(0) should fallback to
the legacy behaviour to enable the SO_REUSEADDR option and make it possible
to connect to different remote (addr, port) tuples.
This patch allows us to bind SO_REUSEADDR enabled sockets to the same
(addr, port) only when net.ipv4.ip_autobind_reuse is set 1 and all
ephemeral ports are exhausted. This also allows connect() and listen() to
share ports in the following way and may break some applications. So the
ip_autobind_reuse is 0 by default and disables the feature.
1. setsockopt(sk1, SO_REUSEADDR)
2. setsockopt(sk2, SO_REUSEADDR)
3. bind(sk1, saddr, 0)
4. bind(sk2, saddr, 0)
5. connect(sk1, daddr)
6. listen(sk2)
If it is set 1, we can fully utilize the 4-tuples, but we should use
IP_BIND_ADDRESS_NO_PORT for bind()+connect() as possible.
The notable thing is that if all sockets bound to the same port have
both SO_REUSEADDR and SO_REUSEPORT enabled, we can bind sockets to an
ephemeral port and also do listen().
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch introduces a sysctl knob "net.ipv4.tcp_no_ssthresh_metrics_save"
that disables TCP ssthresh metrics cache by default. Other parts of TCP
metrics cache, e.g. rtt, cwnd, remain unchanged.
As modern networks becoming more and more dynamic, TCP metrics cache
today often causes more harm than benefits. For example, the same IP
address is often shared by different subscribers behind NAT in residential
networks. Even if the IP address is not shared by different users,
caching the slow-start threshold of a previous short flow using loss-based
congestion control (e.g. cubic) often causes the future longer flows of
the same network path to exit slow-start prematurely with abysmal
throughput.
Caching ssthresh is very risky and can lead to terrible performance.
Therefore it makes sense to make disabling ssthresh caching by
default and opt-in for specific networks by the administrators.
This practice also has worked well for several years of deployment with
CUBIC congestion control at Google.
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Kevin(Yudong) Yang <yyd@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current implementation of TCP MTU probing can considerably
underestimate the MTU on lossy connections allowing the MSS to get down to
48. We have found that in almost all of these cases on our networks these
paths can handle much larger MTUs meaning the connections are being
artificially limited. Even though TCP MTU probing can raise the MSS back up
we have seen this not to be the case causing connections to be "stuck" with
an MSS of 48 when heavy loss is present.
Prior to pushing out this change we could not keep TCP MTU probing enabled
b/c of the above reasons. Now with a reasonble floor set we've had it
enabled for the past 6 months.
The new sysctl will still default to TCP_MIN_SND_MSS (48), but gives
administrators the ability to control the floor of MSS probing.
Signed-off-by: Josh Hunt <johunt@akamai.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some TCP peers announce a very small MSS option in their SYN and/or
SYN/ACK messages.
This forces the stack to send packets with a very high network/cpu
overhead.
Linux has enforced a minimal value of 48. Since this value includes
the size of TCP options, and that the options can consume up to 40
bytes, this means that each segment can include only 8 bytes of payload.
In some cases, it can be useful to increase the minimal value
to a saner value.
We still let the default to 48 (TCP_MIN_SND_MSS), for compatibility
reasons.
Note that TCP_MAXSEG socket option enforces a minimal value
of (TCP_MIN_MSS). David Miller increased this minimal value
in commit c39508d6f1 ("tcp: Make TCP_MAXSEG minimum more correct.")
from 64 to 88.
We might in the future merge TCP_MIN_SND_MSS and TCP_MIN_MSS.
CVE-2019-11479 -- tcp mss hardcoded to 48
Signed-off-by: Eric Dumazet <edumazet@google.com>
Suggested-by: Jonathan Looney <jtl@netflix.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Tyler Hicks <tyhicks@canonical.com>
Cc: Bruce Curtis <brucec@netflix.com>
Cc: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Following patch will add rcu grace period before fqdir
rhashtable destruction, so we need to dynamically allocate
fqdir structures to not force expensive synchronize_rcu() calls
in netns dismantle path.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rename the @frags fields from structs netns_ipv4, netns_ipv6,
netns_nf_frag and netns_ieee802154_lowpan to @fqdir
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1) struct netns_frags is renamed to struct fqdir
This structure is really holding many frag queues in a hash table.
2) (struct inet_frag_queue)->net field is renamed to fqdir
since net is generally associated to a 'struct net' pointer
in networking stack.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
According to Amit Klein and Benny Pinkas, IP ID generation is too weak
and might be used by attackers.
Even with recent net_hash_mix() fix (netns: provide pure entropy for net_hash_mix())
having 64bit key and Jenkins hash is risky.
It is time to switch to siphash and its 128bit keys.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Amit Klein <aksecurity@gmail.com>
Reported-by: Benny Pinkas <benny@pinkas.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a sysctl raw_l3mdev_accept to control raw socket lookup in a manner
similar to use of tcp_l3mdev_accept for stream and of udp_l3mdev_accept
for datagram sockets. Have this default to enabled for reasons of
backwards compatibility. This is so as to specify the output device
with cmsg and IP_PKTINFO, but using a socket not bound to the
corresponding VRF. This allows e.g. older ping implementations to be
run with specifying the device but without executing it in the VRF.
If the option is disabled, packets received in a VRF context are only
handled by a raw socket bound to the VRF, and correspondingly packets
in the default VRF are only handled by a socket not bound to any VRF.
Signed-off-by: Mike Manning <mmanning@vyatta.att-mail.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Tested-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
