The commit 7aae231ac9 ("bpf: tcp: Limit calling some tcp cc functions to CONFIG_DYNAMIC_FTRACE")
added CONFIG_DYNAMIC_FTRACE guard because pahole was only generating
btf for ftrace-able functions. The ftrace filter had already been
removed from pahole, so the CONFIG_DYNAMIC_FTRACE guard can be
removed.
The commit 569c484f99 ("bpf: Limit static tcp-cc functions in the .BTF_ids list to x86")
has added CONFIG_X86 guard because it failed the powerpc arch which
prepended a "." to the local static function, so "cubictcp_init" becomes
".cubictcp_init". "__bpf_kfunc" has been added to kfunc
since then and it uses the __unused compiler attribute.
There is an existing
"__bpf_kfunc static u32 bpf_kfunc_call_test_static_unused_arg(u32 arg, u32 unused)"
test in bpf_testmod.c to cover the static kfunc case.
cross compile on ppc64 with CONFIG_DYNAMIC_FTRACE disabled:
> readelf -s vmlinux | grep cubictcp_
56938: c00000000144fd00 184 FUNC LOCAL DEFAULT 2 cubictcp_cwnd_event [<localentry>: 8]
56939: c00000000144fdb8 200 FUNC LOCAL DEFAULT 2 cubictcp_recalc_[...] [<localentry>: 8]
56940: c00000000144fe80 296 FUNC LOCAL DEFAULT 2 cubictcp_init [<localentry>: 8]
56941: c00000000144ffa8 228 FUNC LOCAL DEFAULT 2 cubictcp_state [<localentry>: 8]
56942: c00000000145008c 1908 FUNC LOCAL DEFAULT 2 cubictcp_cong_avoid [<localentry>: 8]
56943: c000000001450800 1644 FUNC LOCAL DEFAULT 2 cubictcp_acked [<localentry>: 8]
> bpftool btf dump file vmlinux | grep cubictcp_
[51540] FUNC 'cubictcp_acked' type_id=38137 linkage=static
[51541] FUNC 'cubictcp_cong_avoid' type_id=38122 linkage=static
[51543] FUNC 'cubictcp_cwnd_event' type_id=51542 linkage=static
[51544] FUNC 'cubictcp_init' type_id=9186 linkage=static
[51545] FUNC 'cubictcp_recalc_ssthresh' type_id=35021 linkage=static
[51547] FUNC 'cubictcp_state' type_id=38141 linkage=static
The patch removed both config guards.
Cc: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://lore.kernel.org/r/20240322191433.4133280-1-martin.lau@linux.dev
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This commit marks kfuncs as such inside the .BTF_ids section. The upshot
of these annotations is that we'll be able to automatically generate
kfunc prototypes for downstream users. The process is as follows:
1. In source, use BTF_KFUNCS_START/END macro pair to mark kfuncs
2. During build, pahole injects into BTF a "bpf_kfunc" BTF_DECL_TAG for
each function inside BTF_KFUNCS sets
3. At runtime, vmlinux or module BTF is made available in sysfs
4. At runtime, bpftool (or similar) can look at provided BTF and
generate appropriate prototypes for functions with "bpf_kfunc" tag
To ensure future kfunc are similarly tagged, we now also return error
inside kfunc registration for untagged kfuncs. For vmlinux kfuncs,
we also WARN(), as initcall machinery does not handle errors.
Signed-off-by: Daniel Xu <dxu@dxuuu.xyz>
Acked-by: Benjamin Tissoires <bentiss@kernel.org>
Link: https://lore.kernel.org/r/e55150ceecbf0a5d961e608941165c0bee7bc943.1706491398.git.dxu@dxuuu.xyz
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Now that we have the __bpf_kfunc tag, we should use add it to all
existing kfuncs to ensure that they'll never be elided in LTO builds.
Signed-off-by: David Vernet <void@manifault.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Stanislav Fomichev <sdf@google.com>
Link: https://lore.kernel.org/bpf/20230201173016.342758-4-void@manifault.com
Instead of populating multiple sets to indicate some attribute and then
researching the same BTF ID in them, prepare a single unified BTF set
which indicates whether a kfunc is allowed to be called, and also its
attributes if any at the same time. Now, only one call is needed to
perform the lookup for both kfunc availability and its attributes.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20220721134245.2450-4-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
hystart_ack_delay() had the assumption that a TSO packet
would not be bigger than GSO_MAX_SIZE.
This will no longer be true.
We should use sk->sk_gso_max_size instead.
This reduces chances of spurious Hystart ACK train detections.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Alexander Duyck <alexanderduyck@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
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>
Completely remove the old code for check_kfunc_call to help it work
with modules, and also the callback itself.
The previous commit adds infrastructure to register all sets and put
them in vmlinux or module BTF, and concatenates all related sets
organized by the hook and the type. Once populated, these sets remain
immutable for the lifetime of the struct btf.
Also, since we don't need the 'owner' module anywhere when doing
check_kfunc_call, drop the 'btf_modp' module parameter from
find_kfunc_desc_btf.
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20220114163953.1455836-4-memxor@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
While testing BIG TCP patch series, I was expecting that TCP_RR workloads
with 80KB requests/answers would send one 80KB TSO packet,
then being received as a single GRO packet.
It turns out this was not happening, and the root cause was that
cubic Hystart ACK train was triggering after a few (2 or 3) rounds of RPC.
Hystart was wrongly setting CWND/SSTHRESH to 30, while my RPC
needed a budget of ~20 segments.
Ideally these TCP_RR flows should not exit slow start.
Cubic Hystart should reset itself at each round, instead of assuming
every TCP flow is a bulk one.
Note that even after this patch, Hystart can still trigger, depending
on scheduling artifacts, but at a higher CWND/SSTHRESH threshold,
keeping optimal TSO packet sizes.
Tested:
ip link set dev eth0 gro_ipv6_max_size 131072 gso_ipv6_max_size 131072
nstat -n; netperf -H ... -t TCP_RR -l 5 -- -r 80000,80000 -K cubic; nstat|egrep "Ip6InReceives|Hystart|Ip6OutRequests"
Before:
8605
Ip6InReceives 87541 0.0
Ip6OutRequests 129496 0.0
TcpExtTCPHystartTrainDetect 1 0.0
TcpExtTCPHystartTrainCwnd 30 0.0
After:
8760
Ip6InReceives 88514 0.0
Ip6OutRequests 87975 0.0
Fixes: ae27e98a51 ("[TCP] CUBIC v2.3")
Co-developed-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Soheil Hassas Yeganeh <soheil@google.com>
Link: https://lore.kernel.org/r/20211123202535.1843771-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This commit moves BTF ID lookup into the newly added registration
helper, in a way that the bbr, cubic, and dctcp implementation set up
their sets in the bpf_tcp_ca kfunc_btf_set list, while the ones not
dependent on modules are looked up from the wrapper function.
This lifts the restriction for them to be compiled as built in objects,
and can be loaded as modules if required. Also modify Makefile.modfinal
to call resolve_btfids for each module.
Note that since kernel kfunc_ids never overlap with module kfunc_ids, we
only match the owner for module btf id sets.
See following commits for background on use of:
CONFIG_X86 ifdef:
569c484f99 (bpf: Limit static tcp-cc functions in the .BTF_ids list to x86)
CONFIG_DYNAMIC_FTRACE ifdef:
7aae231ac9 (bpf: tcp: Limit calling some tcp cc functions to CONFIG_DYNAMIC_FTRACE)
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211002011757.311265-6-memxor@gmail.com
The cubic functions in tcp_cubic.c are using the bictcp prefix as
in tcp_bic.c. This patch gives it the proper name cubictcp
because the later patch will allow the bpf prog to directly
call the cubictcp implementation. Renaming them will avoid
the name collision when trying to find the intended
one to call during bpf prog load time.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210325015155.1545532-1-kafai@fb.com
Mirja Kuehlewind reported a bug in Linux TCP CUBIC Hystart, where
Hystart HYSTART_DELAY mechanism can exit Slow Start spuriously on an
ACK when the minimum rtt of a connection goes down. From inspection it
is clear from the existing code that this could happen in an example
like the following:
o The first 8 RTT samples in a round trip are 150ms, resulting in a
curr_rtt of 150ms and a delay_min of 150ms.
o The 9th RTT sample is 100ms. The curr_rtt does not change after the
first 8 samples, so curr_rtt remains 150ms. But delay_min can be
lowered at any time, so delay_min falls to 100ms. The code executes
the HYSTART_DELAY comparison between curr_rtt of 150ms and delay_min
of 100ms, and the curr_rtt is declared far enough above delay_min to
force a (spurious) exit of Slow start.
The fix here is simple: allow every RTT sample in a round trip to
lower the curr_rtt.
Fixes: ae27e98a51 ("[TCP] CUBIC v2.3")
Reported-by: Mirja Kuehlewind <mirja.kuehlewind@ericsson.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
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>
For years we disabled Hystart ACK train detection at Google
because it was fooled by TCP pacing.
ACK train detection uses a simple heuristic, detecting if
we receive ACK past half the RTT, to exit slow start before
hitting the bottleneck and experience massive drops.
But pacing by design might delay packets up to RTT/2,
so we need to tweak the Hystart logic to be aware of this
extra delay.
Tested:
Added a 100 usec delay at receiver.
Before:
nstat -n;for f in {1..10}; do ./super_netperf 1 -H lpaa24 -l -4000000; done;nstat|egrep "Hystart"
9117
7057
9553
8300
7030
6849
9533
10126
6876
8473
TcpExtTCPHystartTrainDetect 10 0.0
TcpExtTCPHystartTrainCwnd 1230 0.0
After :
nstat -n;for f in {1..10}; do ./super_netperf 1 -H lpaa24 -l -4000000; done;nstat|egrep "Hystart"
9845
10103
10866
11096
11936
11487
11773
12188
11066
11894
TcpExtTCPHystartTrainDetect 10 0.0
TcpExtTCPHystartTrainCwnd 6462 0.0
Disabling Hystart ACK Train detection gives similar numbers
echo 2 >/sys/module/tcp_cubic/parameters/hystart_detect
nstat -n;for f in {1..10}; do ./super_netperf 1 -H lpaa24 -l -4000000; done;nstat|egrep "Hystart"
11173
10954
12455
10627
11578
11583
11222
10880
10665
11366
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
After switching ca->delay_min to usec resolution, we exit
slow start prematurely for very low RTT flows, setting
snd_ssthresh to 20.
The reason is that delay_min is fed with RTT of small packet
trains. Then as cwnd is increased, TCP sends bigger TSO packets.
LRO/GRO aggregation and/or interrupt mitigation strategies
on receiver tend to inflate RTT samples.
Fix this by adding to delay_min the expected delay of
two TSO packets, given current pacing rate.
Tested:
Sender uses pfifo_fast qdisc
Before :
$ nstat -n;for f in {1..10}; do ./super_netperf 1 -H lpaa24 -l -4000000; done;nstat|egrep "Hystart"
11348
11707
11562
11428
11773
11534
9878
11693
10597
10968
TcpExtTCPHystartTrainDetect 10 0.0
TcpExtTCPHystartTrainCwnd 200 0.0
After :
$ nstat -n;for f in {1..10}; do ./super_netperf 1 -H lpaa24 -l -4000000; done;nstat|egrep "Hystart"
14877
14517
15797
18466
17376
14833
17558
17933
16039
18059
TcpExtTCPHystartTrainDetect 10 0.0
TcpExtTCPHystartTrainCwnd 1670 0.0
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current 1ms clock feeds ca->round_start, ca->delay_min,
ca->last_ack.
This is quite problematic for data-center flows, where delay_min
is way below 1 ms.
This means Hystart Train detection triggers every time jiffies value
is updated, since "((s32)(now - ca->round_start) > ca->delay_min >> 4)"
expression becomes true.
This kind of random behavior can be solved by reusing the existing
usec timestamp that TCP keeps in tp->tcp_mstamp
Note that a followup patch will tweak things a bit, because
during slow start, GRO aggregation on receivers naturally
increases the RTT as TSO packets gradually come to ~64KB size.
To recap, right after this patch CUBIC Hystart train detection
is more aggressive, since short RTT flows might exit slow start at
cwnd = 20, instead of being possibly unbounded.
Following patch will address this problem.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If we initialize ca->curr_rtt to ~0U, we do not need to test
for zero value in hystart_update()
We only read ca->curr_rtt if at least HYSTART_MIN_SAMPLES have
been processed, and thus ca->curr_rtt will have a sane value.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We do not care which bit in ca->found is set.
We avoid accessing hystart and hystart_detect unless really needed,
possibly avoiding one cache line miss.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add SPDX license identifiers to all files which:
- Have no license information of any form
- Have MODULE_LICENCE("GPL*") inside which was used in the initial
scan/conversion to ignore the file
These files fall under the project license, GPL v2 only. The resulting SPDX
license identifier is:
GPL-2.0-only
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Most TCP congestion controls are using identical logic to undo
cwnd except BBR. This patch consolidates these similar functions
to the one used currently by Reno and others.
Suggested-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use tcp_jiffies32 instead of tcp_time_stamp, since
tcp_time_stamp will soon be only used for TCP TS option.
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>
Use tcp_jiffies32 instead of tcp_time_stamp to feed
tp->lsndtime.
tcp_time_stamp will soon be a litle bit more expensive
than simply reading 'jiffies'.
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>
Replace 2 arguments (cnt and rtt) in the congestion control modules'
pkts_acked() function with a struct. This will allow adding more
information without having to modify existing congestion control
modules (tcp_nv in particular needs bytes in flight when packet
was sent).
As proposed by Neal Cardwell in his comments to the tcp_nv patch.
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We want to to make TCP stack preemptible, as draining prequeue
and backlog queues can take lot of time.
Many SNMP updates were assuming that BH (and preemption) was disabled.
Need to convert some __NET_INC_STATS() calls to NET_INC_STATS()
and some __TCP_INC_STATS() to TCP_INC_STATS()
Before using this_cpu_ptr(net->ipv4.tcp_sk) in tcp_v4_send_reset()
and tcp_v4_send_ack(), we add an explicit preempt disabled section.
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>
Rename NET_INC_STATS_BH() to __NET_INC_STATS()
and NET_ADD_STATS_BH() to __NET_ADD_STATS()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tracking idle time in bictcp_cwnd_event() is imprecise, as epoch_start
is normally set at ACK processing time, not at send time.
Doing a proper fix would need to add an additional state variable,
and does not seem worth the trouble, given CUBIC bug has been there
forever before Jana noticed it.
Let's simply not set epoch_start in the future, otherwise
bictcp_update() could overflow and CUBIC would again
grow cwnd too fast.
This was detected thanks to a packetdrill test Neal wrote that was flaky
before applying this fix.
Fixes: 30927520db ("tcp_cubic: better follow cubic curve after idle period")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Cc: Jana Iyengar <jri@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jana Iyengar found an interesting issue on CUBIC :
The epoch is only updated/reset initially and when experiencing losses.
The delta "t" of now - epoch_start can be arbitrary large after app idle
as well as the bic_target. Consequentially the slope (inverse of
ca->cnt) would be really large, and eventually ca->cnt would be
lower-bounded in the end to 2 to have delayed-ACK slow-start behavior.
This particularly shows up when slow_start_after_idle is disabled
as a dangerous cwnd inflation (1.5 x RTT) after few seconds of idle
time.
Jana initial fix was to reset epoch_start if app limited,
but Neal pointed out it would ask the CUBIC algorithm to recalculate the
curve so that we again start growing steeply upward from where cwnd is
now (as CUBIC does just after a loss). Ideally we'd want the cwnd growth
curve to be the same shape, just shifted later in time by the amount of
the idle period.
Reported-by: Jana Iyengar <jri@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: Sangtae Ha <sangtae.ha@gmail.com>
Cc: Lawrence Brakmo <lawrence@brakmo.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a helper to test the slow start condition in various congestion
control modules and other places. This is to prepare a slight improvement
in policy as to exactly when to slow start.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Nandita Dukkipati <nanditad@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 814d488c61 ("tcp: fix the timid additive increase on stretch
ACKs") fixed a bug where tcp_cong_avoid_ai() would either credit a
connection with an increase of snd_cwnd_cnt, or increase snd_cwnd, but
not both, resulting in cwnd increasing by 1 packet on at most every
alternate invocation of tcp_cong_avoid_ai().
Although the commit correctly implemented the CUBIC algorithm, which
can increase cwnd by as much as 1 packet per 1 packet ACKed (2x per
RTT), in practice that could be too aggressive: in tests on network
paths with small buffers, YouTube server retransmission rates nearly
doubled.
This commit restores CUBIC to a maximum cwnd growth rate of 1 packet
per 2 packets ACKed (1.5x per RTT). In YouTube tests this restored
retransmit rates to low levels.
Testing: This patch has been tested in datacenter netperf transfers
and live youtube.com and google.com servers.
Fixes: 9cd981dcf1 ("tcp: fix stretch ACK bugs in CUBIC")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes a bug in CUBIC that causes cwnd to increase slightly
too slowly when multiple ACKs arrive in the same jiffy.
If cwnd is supposed to increase at a rate of more than once per jiffy,
then CUBIC was sometimes too slow. Because the bic_target is
calculated for a future point in time, calculated with time in
jiffies, the cwnd can increase over the course of the jiffy while the
bic_target calculated as the proper CUBIC cwnd at time
t=tcp_time_stamp+rtt does not increase, because tcp_time_stamp only
increases on jiffy tick boundaries.
So since the cnt is set to:
ca->cnt = cwnd / (bic_target - cwnd);
as cwnd increases but bic_target does not increase due to jiffy
granularity, the cnt becomes too large, causing cwnd to increase
too slowly.
For example:
- suppose at the beginning of a jiffy, cwnd=40, bic_target=44
- so CUBIC sets:
ca->cnt = cwnd / (bic_target - cwnd) = 40 / (44 - 40) = 40/4 = 10
- suppose we get 10 acks, each for 1 segment, so tcp_cong_avoid_ai()
increases cwnd to 41
- so CUBIC sets:
ca->cnt = cwnd / (bic_target - cwnd) = 41 / (44 - 41) = 41 / 3 = 13
So now CUBIC will wait for 13 packets to be ACKed before increasing
cwnd to 42, insted of 10 as it should.
The fix is to avoid adjusting the slope (determined by ca->cnt)
multiple times within a jiffy, and instead skip to compute the Reno
cwnd, the "TCP friendliness" code path.
Reported-by: Eyal Perry <eyalpe@mellanox.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Change CUBIC to properly handle stretch ACKs in additive increase mode
by passing in the count of ACKed packets to tcp_cong_avoid_ai().
In addition, because we are now precisely accounting for stretch ACKs,
including delayed ACKs, we can now remove the delayed ACK tracking and
estimation code that tracked recent delayed ACK behavior in
ca->delayed_ack.
Reported-by: Eyal Perry <eyalpe@mellanox.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
LRO, GRO, delayed ACKs, and middleboxes can cause "stretch ACKs" that
cover more than the RFC-specified maximum of 2 packets. These stretch
ACKs can cause serious performance shortfalls in common congestion
control algorithms that were designed and tuned years ago with
receiver hosts that were not using LRO or GRO, and were instead
politely ACKing every other packet.
This patch series fixes Reno and CUBIC to handle stretch ACKs.
This patch prepares for the upcoming stretch ACK bug fix patches. It
adds an "acked" parameter to tcp_cong_avoid_ai() to allow for future
fixes to tcp_cong_avoid_ai() to correctly handle stretch ACKs, and
changes all congestion control algorithms to pass in 1 for the ACKed
count. It also changes tcp_slow_start() to return the number of packet
ACK "credits" that were not processed in slow start mode, and can be
processed by the congestion control module in additive increase mode.
In future patches we will fix tcp_cong_avoid_ai() to handle stretch
ACKs, and fix Reno and CUBIC handling of stretch ACKs in slow start
and additive increase mode.
Reported-by: Eyal Perry <eyalpe@mellanox.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In commit 2b4636a5f8 ("tcp_cubic: make the delay threshold of HyStart
less sensitive"), HYSTART_DELAY_MIN was changed to 4 ms.
The remaining problem is that using delay_min + (delay_min/16) as the
threshold is too sensitive.
6.25 % of variation is too small for rtt above 60 ms, which are not
uncommon.
Lets use 12.5 % instead (delay_min + (delay_min/8))
Tested:
80 ms RTT between peers, FQ/pacing packet scheduler on sender.
10 bulk transfers of 10 seconds :
nstat >/dev/null
for i in `seq 1 10`
do
netperf -H remote -- -k THROUGHPUT | grep THROUGHPUT
done
nstat | grep Hystart
With the 6.25 % threshold :
THROUGHPUT=20.66
THROUGHPUT=249.38
THROUGHPUT=254.10
THROUGHPUT=14.94
THROUGHPUT=251.92
THROUGHPUT=237.73
THROUGHPUT=19.18
THROUGHPUT=252.89
THROUGHPUT=21.32
THROUGHPUT=15.58
TcpExtTCPHystartTrainDetect 2 0.0
TcpExtTCPHystartTrainCwnd 4756 0.0
TcpExtTCPHystartDelayDetect 5 0.0
TcpExtTCPHystartDelayCwnd 180 0.0
With the 12.5 % threshold
THROUGHPUT=251.09
THROUGHPUT=247.46
THROUGHPUT=250.92
THROUGHPUT=248.91
THROUGHPUT=250.88
THROUGHPUT=249.84
THROUGHPUT=250.51
THROUGHPUT=254.15
THROUGHPUT=250.62
THROUGHPUT=250.89
TcpExtTCPHystartTrainDetect 1 0.0
TcpExtTCPHystartTrainCwnd 3175 0.0
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Tested-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When deploying FQ pacing, one thing we noticed is that CUBIC Hystart
triggers too soon.
Having SNMP counters to have an idea of how often the various Hystart
methods trigger is useful prior to any modifications.
This patch adds SNMP counters tracking, how many time "ack train" or
"Delay" based Hystart triggers, and cumulative sum of cwnd at the time
Hystart decided to end SS (Slow Start)
myhost:~# nstat -a | grep Hystart
TcpExtTCPHystartTrainDetect 9 0.0
TcpExtTCPHystartTrainCwnd 20650 0.0
TcpExtTCPHystartDelayDetect 10 0.0
TcpExtTCPHystartDelayCwnd 360 0.0
->
Train detection was triggered 9 times, and average cwnd was
20650/9=2294,
Delay detection was triggered 10 times and average cwnd was 36
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix places where there is space before tab, long lines, and
awkward if(){, double spacing etc. Add blank line after declaration/initialization.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/ethernet/altera/altera_sgdma.c
net/netlink/af_netlink.c
net/sched/cls_api.c
net/sched/sch_api.c
The netlink conflict dealt with moving to netlink_capable() and
netlink_ns_capable() in the 'net' tree vs. supporting 'tc' operations
in non-init namespaces. These were simple transformations from
netlink_capable to netlink_ns_capable.
The Altera driver conflict was simply code removal overlapping some
void pointer cast cleanups in net-next.
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit e114a710aa ("tcp: fix cwnd limited checking to improve
congestion control") obsoleted in_flight parameter from
tcp_is_cwnd_limited() and its callers.
This patch does the removal as promised.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit b9f47a3aae (tcp_cubic: limit delayed_ack ratio to prevent
divide error) try to prevent divide error, but there is still a little
chance that delayed_ack can reach zero. In case the param cnt get
negative value, then ratio+cnt would overflow and may happen to be zero.
As a result, min(ratio, ACK_RATIO_LIMIT) will calculate to be zero.
In some old kernels, such as 2.6.32, there is a bug that would
pass negative param, which then ultimately leads to this divide error.
commit 5b35e1e6e9 (tcp: fix tcp_trim_head() to adjust segment count
with skb MSS) fixed the negative param issue. However,
it's safe that we fix the range of delayed_ack as well,
to make sure we do not hit a divide by zero.
CC: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Liu Yu <allanyuliu@tencent.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>
Upcoming congestion controls for TCP require usec resolution for RTT
estimations. Millisecond resolution is simply not enough these days.
FQ/pacing in DC environments also require this change for finer control
and removal of bimodal behavior due to the current hack in
tcp_update_pacing_rate() for 'small rtt'
TCP_CONG_RTT_STAMP is no longer needed.
As Julian Anastasov pointed out, we need to keep user compatibility :
tcp_metrics used to export RTT and RTTVAR in msec resolution,
so we added RTT_US and RTTVAR_US. An iproute2 patch is needed
to use the new attributes if provided by the kernel.
In this example ss command displays a srtt of 32 usecs (10Gbit link)
lpk51:~# ./ss -i dst lpk52
Netid State Recv-Q Send-Q Local Address:Port Peer
Address:Port
tcp ESTAB 0 1 10.246.11.51:42959
10.246.11.52:64614
cubic wscale:6,6 rto:201 rtt:0.032/0.001 ato:40 mss:1448
cwnd:10 send
3620.0Mbps pacing_rate 7240.0Mbps unacked:1 rcv_rtt:993 rcv_space:29559
Updated iproute2 ip command displays :
lpk51:~# ./ip tcp_metrics | grep 10.246.11.52
10.246.11.52 age 561.914sec cwnd 10 rtt 274us rttvar 213us source
10.246.11.51
Old binary displays :
lpk51:~# ip tcp_metrics | grep 10.246.11.52
10.246.11.52 age 561.914sec cwnd 10 rtt 250us rttvar 125us source
10.246.11.51
With help from Julian Anastasov, Stephen Hemminger and Yuchung Cheng
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Larry Brakmo <brakmo@google.com>
Cc: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
Slow start now increases cwnd by 1 if an ACK acknowledges some packets,
regardless the number of packets. Consequently slow start performance
is highly dependent on the degree of the stretch ACKs caused by
receiver or network ACK compression mechanisms (e.g., delayed-ACK,
GRO, etc). But slow start algorithm is to send twice the amount of
packets of packets left so it should process a stretch ACK of degree
N as if N ACKs of degree 1, then exits when cwnd exceeds ssthresh. A
follow up patch will use the remainder of the N (if greater than 1)
to adjust cwnd in the congestion avoidance phase.
In addition this patch retires the experimental limited slow start
(LSS) feature. LSS has multiple drawbacks but questionable benefit. The
fractional cwnd increase in LSS requires a loop in slow start even
though it's rarely used. Configuring such an increase step via a global
sysctl on different BDPS seems hard. Finally and most importantly the
slow start overshoot concern is now better covered by the Hybrid slow
start (hystart) enabled by default.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While investigating about strange increase of retransmit rates
on hosts ~24 days after boot, Van found hystart was disabled
if ca->epoch_start was 0, as following condition is true
when tcp_time_stamp high order bit is set.
(s32)(tcp_time_stamp - ca->epoch_start) < HZ
Quoting Van :
At initialization & after every loss ca->epoch_start is set to zero so
I believe that the above line will turn off hystart as soon as the 2^31
bit is set in tcp_time_stamp & hystart will stay off for 24 days.
I think we've observed that cubic's restart is too aggressive without
hystart so this might account for the higher drop rate we observe.
Diagnosed-by: Van Jacobson <vanj@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit 17a6e9f1aa ("tcp_cubic: fix clock dependency") added an
overflow error in bictcp_update() in following code :
/* change the unit from HZ to bictcp_HZ */
t = ((tcp_time_stamp + msecs_to_jiffies(ca->delay_min>>3) -
ca->epoch_start) << BICTCP_HZ) / HZ;
Because msecs_to_jiffies() being unsigned long, compiler does
implicit type promotion.
We really want to constrain (tcp_time_stamp - ca->epoch_start)
to a signed 32bit value, or else 't' has unexpected high values.
This bugs triggers an increase of retransmit rates ~24 days after
boot [1], as the high order bit of tcp_time_stamp flips.
[1] for hosts with HZ=1000
Big thanks to Van Jacobson for spotting this problem.
Diagnosed-by: Van Jacobson <vanj@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes CUBIC so that cwnd reductions made during RTOs can be
undone (just as they already can be undone when using the default/Reno
behavior).
When undoing cwnd reductions, BIC-derived congestion control modules
were restoring the cwnd from last_max_cwnd. There were two problems
with using last_max_cwnd to restore a cwnd during undo:
(a) last_max_cwnd was set to 0 on state transitions into TCP_CA_Loss
(by calling the module's reset() functions), so cwnd reductions from
RTOs could not be undone.
(b) when fast_covergence is enabled (which it is by default)
last_max_cwnd does not actually hold the value of snd_cwnd before the
loss; instead, it holds a scaled-down version of snd_cwnd.
This patch makes the following changes:
(1) upon undo, revert snd_cwnd to ca->loss_cwnd, which is already, as
the existing comment notes, the "congestion window at last loss"
(2) stop forgetting ca->loss_cwnd on TCP_CA_Loss events
(3) use ca->last_max_cwnd to check if we're in slow start
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: Sangtae Ha <sangtae.ha@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP Cubic keeps a metric that estimates the amount of delayed
acknowledgements to use in adjusting the window. If an abnormally
large number of packets are acknowledged at once, then the update
could wrap and reach zero. This kind of ACK could only
happen when there was a large window and huge number of
ACK's were lost.
This patch limits the value of delayed ack ratio. The choice of 32
is just a conservative value since normally it should be range of
1 to 4 packets.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
HyStart sets the initial exit point of slow start.
Suppose that HyStart exits at 0.5BDP in a BDP network and no history exists.
If the BDP of a network is large, CUBIC's initial cwnd growth may be
too conservative to utilize the link.
CUBIC increases the cwnd 20% per RTT in this case.
Signed-off-by: Sangtae Ha <sangtae.ha@gmail.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make HyStart less sensitive to abrupt delay variations due to buffer bloat.
Signed-off-by: Sangtae Ha <sangtae.ha@gmail.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Reported-by: Lucas Nussbaum <lucas.nussbaum@loria.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a refined version of an earlier patch by Lucas Nussbaum.
Cubic needs RTT values in milliseconds. If HZ < 1000 then
the values will be too coarse.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Reported-by: Lucas Nussbaum <lucas.nussbaum@loria.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
The hystart code was written with assumption that HZ=1000.
Replace the use of jiffies with bictcp_clock as a millisecond
real time clock.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Reported-by: Lucas Nussbaum <lucas.nussbaum@loria.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
