tcp_set_default_congestion_control() is netns-safe in that it writes
to &net->ipv4.tcp_congestion_control, but it also sets
ca->flags |= TCP_CONG_NON_RESTRICTED which is not namespaced.
This has the unintended side-effect of changing the global
net.ipv4.tcp_allowed_congestion_control sysctl, despite the fact that it
is read-only: 97684f0970 ("net: Make tcp_allowed_congestion_control
readonly in non-init netns")
Resolve this netns "leak" by only allowing the init netns to set the
default algorithm to one that is restricted. This restriction could be
removed if tcp_allowed_congestion_control were namespace-ified in the
future.
This bug was uncovered with
https://github.com/JonathonReinhart/linux-netns-sysctl-verify
Fixes: 6670e15244 ("tcp: Namespace-ify sysctl_tcp_default_congestion_control")
Signed-off-by: Jonathon Reinhart <jonathon.reinhart@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When setting congestion control via a BPF program it is seen that the
SYN/ACK for packets within a given flow will not include the ECT0 flag. A
bit of simple printk debugging shows that when this is configured without
BPF we will see the value INET_ECN_xmit value initialized in
tcp_assign_congestion_control however when we configure this via BPF the
socket is in the closed state and as such it isn't configured, and I do not
see it being initialized when we transition the socket into the listen
state. The result of this is that the ECT0 bit is configured based on
whatever the default state is for the socket.
Any easy way to reproduce this is to monitor the following with tcpdump:
tools/testing/selftests/bpf/test_progs -t bpf_tcp_ca
Without this patch the SYN/ACK will follow whatever the default is. If dctcp
all SYN/ACK packets will have the ECT0 bit set, and if it is not then ECT0
will be cleared on all SYN/ACK packets. With this patch applied the SYN/ACK
bit matches the value seen on the other packets in the given stream.
Fixes: 91b5b21c7c ("bpf: Add support for changing congestion control")
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Simplify tcp_set_congestion_control() by removing the initialization
code path for the !load case.
There are only two call sites for tcp_set_congestion_control(). The
EBPF call site is the only one that passes load=false; it also passes
cap_net_admin=true. Because of that, the exact same behavior can be
achieved by removing the special if (!load) branch of the logic. Both
before and after this commit, the EBPF case will call
bpf_try_module_get(), and if that succeeds then call
tcp_reinit_congestion_control() or if that fails then return EBUSY.
Note that this returns the logic to a structure very similar to the
structure before:
commit 91b5b21c7c ("bpf: Add support for changing congestion control")
except that the CAP_NET_ADMIN status is passed in as a function
argument.
This clean-up was suggested by Martin KaFai Lau.
Suggested-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Cc: Lawrence Brakmo <brakmo@fb.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Kevin Yang <yyd@google.com>
Now that the previous patches ensure that all call sites for
tcp_set_congestion_control() want to initialize congestion control, we
can simplify tcp_set_congestion_control() by removing the reinit
argument and the code to support it.
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Kevin Yang <yyd@google.com>
Cc: Lawrence Brakmo <brakmo@fb.com>
Change tcp_init_transfer() to only initialize congestion control if it
has not been initialized already.
With this new approach, we can arrange things so that if the EBPF code
sets the congestion control by calling setsockopt(TCP_CONGESTION) then
tcp_init_transfer() will not re-initialize the CC module.
This is an approach that has the following beneficial properties:
(1) This allows CC module customizations made by the EBPF called in
tcp_init_transfer() to persist, and not be wiped out by a later
call to tcp_init_congestion_control() in tcp_init_transfer().
(2) Does not flip the order of EBPF and CC init, to avoid causing bugs
for existing code upstream that depends on the current order.
(3) Does not cause 2 initializations for for CC in the case where the
EBPF called in tcp_init_transfer() wants to set the CC to a new CC
algorithm.
(4) Allows follow-on simplifications to the code in net/core/filter.c
and net/ipv4/tcp_cong.c, which currently both have some complexity
to special-case CC initialization to avoid double CC
initialization if EBPF sets the CC.
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Kevin Yang <yyd@google.com>
Cc: Lawrence Brakmo <brakmo@fb.com>
This patch makes "struct tcp_congestion_ops" to be the first user
of BPF STRUCT_OPS. It allows implementing a tcp_congestion_ops
in bpf.
The BPF implemented tcp_congestion_ops can be used like
regular kernel tcp-cc through sysctl and setsockopt. e.g.
[root@arch-fb-vm1 bpf]# sysctl -a | egrep congestion
net.ipv4.tcp_allowed_congestion_control = reno cubic bpf_cubic
net.ipv4.tcp_available_congestion_control = reno bic cubic bpf_cubic
net.ipv4.tcp_congestion_control = bpf_cubic
There has been attempt to move the TCP CC to the user space
(e.g. CCP in TCP). The common arguments are faster turn around,
get away from long-tail kernel versions in production...etc,
which are legit points.
BPF has been the continuous effort to join both kernel and
userspace upsides together (e.g. XDP to gain the performance
advantage without bypassing the kernel). The recent BPF
advancements (in particular BTF-aware verifier, BPF trampoline,
BPF CO-RE...) made implementing kernel struct ops (e.g. tcp cc)
possible in BPF. It allows a faster turnaround for testing algorithm
in the production while leveraging the existing (and continue growing)
BPF feature/framework instead of building one specifically for
userspace TCP CC.
This patch allows write access to a few fields in tcp-sock
(in bpf_tcp_ca_btf_struct_access()).
The optional "get_info" is unsupported now. It can be added
later. One possible way is to output the info with a btf-id
to describe the content.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20200109003508.3856115-1-kafai@fb.com
snprintf returns the number of chars that would be written, not number
of chars that were actually written. As such, 'offs' may get larger than
'tbl.maxlen', causing the 'tbl.maxlen - offs' being < 0, and since the
parameter is size_t, it would overflow.
Since using scnprintf may hide the limit error, while the buffer is still
enough now, let's just add a WARN_ON_ONCE in case it reach the limit
in future.
v2: Use WARN_ON_ONCE as Jiri and Eric suggested.
Suggested-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Neal reported incorrect use of ns_capable() from bpf hook.
bpf_setsockopt(...TCP_CONGESTION...)
-> tcp_set_congestion_control()
-> ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)
-> ns_capable_common()
-> current_cred()
-> rcu_dereference_protected(current->cred, 1)
Accessing 'current' in bpf context makes no sense, since packets
are processed from softirq context.
As Neal stated : The capability check in tcp_set_congestion_control()
was written assuming a system call context, and then was reused from
a BPF call site.
The fix is to add a new parameter to tcp_set_congestion_control(),
so that the ns_capable() call is only performed under the right
context.
Fixes: 91b5b21c7c ("bpf: Add support for changing congestion control")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Lawrence Brakmo <brakmo@fb.com>
Reported-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Lawrence Brakmo <brakmo@fb.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 EXPORT_.*_SYMBOL_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>
Make default TCP default congestion control to a per namespace
value. This changes default congestion control to a pointer to congestion ops
(rather than implicit as first element of available lsit).
The congestion control setting of new namespaces is inherited
from the current setting of the root namespace.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are a few bugs around refcnt handling in the new BPF congestion
control setsockopt:
- The new ca is assigned to icsk->icsk_ca_ops even in the case where we
cannot get a reference on it. This would lead to a use after free,
since that ca is going away soon.
- Changing the congestion control case doesn't release the refcnt on
the previous ca.
- In the reinit case, we first leak a reference on the old ca, then we
call tcp_reinit_congestion_control on the ca that we have just
assigned, leading to deinitializing the wrong ca (->release of the
new ca on the old ca's data) and releasing the refcount on the ca
that we actually want to use.
This is visible by building (for example) BIC as a module and setting
net.ipv4.tcp_congestion_control=bic, and using tcp_cong_kern.c from
samples/bpf.
This patch fixes the refcount issues, and moves reinit back into tcp
core to avoid passing a ca pointer back to BPF.
Fixes: 91b5b21c7c ("bpf: Add support for changing congestion control")
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Acked-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using ssthresh to revert cwnd is less reliable when ssthresh is
bounded to 2 packets. This patch uses an existing variable in TCP
"prior_cwnd" that snapshots the cwnd right before entering fast
recovery and RTO recovery in Reno. This fixes the issue discussed
in netdev thread: "A buggy behavior for Linux TCP Reno and HTCP"
https://www.spinics.net/lists/netdev/msg444955.html
Suggested-by: Neal Cardwell <ncardwell@google.com>
Reported-by: Wei Sun <unlcsewsun@gmail.com>
Signed-off-by: Yuchung Cheng <ncardwell@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Added support for changing congestion control for SOCK_OPS bpf
programs through the setsockopt bpf helper function. It also adds
a new SOCK_OPS op, BPF_SOCK_OPS_NEEDS_ECN, that is needed for
congestion controls, like dctcp, that need to enable ECN in the
SYN packets.
Signed-off-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When the sender switches its congestion control during loss
recovery, if the recovery is spurious then it may incorrectly
revert cwnd and ssthresh to the older values set by a previous
congestion control. Consider a congestion control (like BBR)
that does not use ssthresh and keeps it infinite: the connection
may incorrectly revert cwnd to an infinite value when switching
from BBR to another congestion control.
This patch fixes it by disallowing such cwnd undo operation
upon switching congestion control. Note that undo_marker
is not reset s.t. the packets that were incorrectly marked
lost would be corrected. We only avoid undoing the cwnd in
tcp_undo_cwnd_reduction().
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Soheil Hassas Yeganeh <soheil@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>
Always zero out ca_priv data in tcp_assign_congestion_control() so that
ca_priv data is cleared out during socket creation.
Also always zero out ca_priv data in tcp_reinit_congestion_control() so
that when cc algorithm is changed, ca_priv data is cleared out as well.
We should still zero out ca_priv data even in TCP_CLOSE state because
user could call connect() on AF_UNSPEC to disconnect the socket and
leave it in TCP_CLOSE state and later call setsockopt() to switch cc
algorithm on this socket.
Fixes: 2b0a8c9ee ("tcp: add CDG congestion control")
Reported-by: Andrey Konovalov <andreyknvl@google.com>
Signed-off-by: Wei Wang <weiwan@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
All conflicts were simple overlapping changes except perhaps
for the Thunder driver.
That driver has a change_mtu method explicitly for sending
a message to the hardware. If that fails it returns an
error.
Normally a driver doesn't need an ndo_change_mtu method becuase those
are usually just range changes, which are now handled generically.
But since this extra operation is needed in the Thunder driver, it has
to stay.
However, if the message send fails we have to restore the original
MTU before the change because the entire call chain expects that if
an error is thrown by ndo_change_mtu then the MTU did not change.
Therefore code is added to nicvf_change_mtu to remember the original
MTU, and to restore it upon nicvf_update_hw_max_frs() failue.
Signed-off-by: David S. Miller <davem@davemloft.net>
The undo_cwnd fallback in the stack doubles cwnd based on ssthresh,
which un-does reno halving behaviour.
It seems more appropriate to let congctl algorithms pair .ssthresh
and .undo_cwnd properly. Add a 'tcp_reno_undo_cwnd' function and wire it
up for all congestion algorithms that used to rely on the fallback.
Cc: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
We need to zero out the private data area when application switches
connection to different algorithm (TCP_CONGESTION setsockopt).
When congestion ops get assigned at connect time everything is already
zeroed because sk_alloc uses GFP_ZERO flag. But in the setsockopt case
this contains whatever previous cc placed there.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit introduces an optional new "omnipotent" hook,
cong_control(), for congestion control modules. The cong_control()
function is called at the end of processing an ACK (i.e., after
updating sequence numbers, the SACK scoreboard, and loss
detection). At that moment we have precise delivery rate information
the congestion control module can use to control the sending behavior
(using cwnd, TSO skb size, and pacing rate) in any CA state.
This function can also be used by a congestion control that prefers
not to use the default cwnd reduction approach (i.e., the PRR
algorithm) during CA_Recovery to control the cwnd and sending rate
during loss recovery.
We take advantage of the fact that recent changes defer the
retransmission or transmission of new data (e.g. by F-RTO) in recovery
until the new tcp_cong_control() function is run.
With this commit, we only run tcp_update_pacing_rate() if the
congestion control is not using this new API. New congestion controls
which use the new API do not want the TCP stack to run the default
pacing rate calculation and overwrite whatever pacing rate they have
chosen at initialization time.
Signed-off-by: Van Jacobson <vanj@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Nandita Dukkipati <nanditad@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
SYNACK packets might be sent without holding socket lock.
For DCTCP/ECN sake, we should call INET_ECN_xmit() while
socket lock is owned, and only when we init/change congestion control.
This also fixies a bug if congestion module is changed from
dctcp to another one on a listener : we now clear ECN bits
properly.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently, the following case doesn't use DCTCP, even if it should:
A responder has f.e. Cubic as system wide default, but for a specific
route to the initiating host, DCTCP is being set in RTAX_CC_ALGO. The
initiating host then uses DCTCP as congestion control, but since the
initiator sets ECT(0), tcp_ecn_create_request() doesn't set ecn_ok,
and we have to fall back to Reno after 3WHS completes.
We were thinking on how to solve this in a minimal, non-intrusive
way without bloating tcp_ecn_create_request() needlessly: lets cache
the CA ecn option flag in RTAX_FEATURES. In other words, when ECT(0)
is set on the SYN packet, set ecn_ok=1 iff route RTAX_FEATURES
contains the unexposed (internal-only) DST_FEATURE_ECN_CA. This allows
to only do a single metric feature lookup inside tcp_ecn_create_request().
Joint work with Florian Westphal.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the original design slow start is only used to raise cwnd
when cwnd is stricly below ssthresh. It makes little sense
to slow start when cwnd == ssthresh: especially
when hystart has set ssthresh in the initial ramp, or after
recovery when cwnd resets to ssthresh. Not doing so will
also help reduce the buffer bloat slightly.
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>
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>
Linux 3.17 and earlier are explicitly engineered so that if the app
doesn't specifically request a CC module on a listener before the SYN
arrives, then the child gets the system default CC when the connection
is established. See tcp_init_congestion_control() in 3.17 or earlier,
which says "if no choice made yet assign the current value set as
default". The change ("net: tcp: assign tcp cong_ops when tcp sk is
created") altered these semantics, so that children got their parent
listener's congestion control even if the system default had changed
after the listener was created.
This commit returns to those original semantics from 3.17 and earlier,
since they are the original semantics from 2007 in 4d4d3d1e8 ("[TCP]:
Congestion control initialization."), and some Linux congestion
control workflows depend on that.
In summary, if a listener socket specifically sets TCP_CONGESTION to
"x", or the route locks the CC module to "x", then the child gets
"x". Otherwise the child gets current system default from
net.ipv4.tcp_congestion_control. That's the behavior in 3.17 and
earlier, and this commit restores that.
Fixes: 55d8694fa8 ("net: tcp: assign tcp cong_ops when tcp sk is created")
Cc: Florian Westphal <fw@strlen.de>
Cc: Daniel Borkmann <dborkman@redhat.com>
Cc: Glenn Judd <glenn.judd@morganstanley.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/ethernet/emulex/benet/be_main.c
net/core/sysctl_net_core.c
net/ipv4/inet_diag.c
The be_main.c conflict resolution was really tricky. The conflict
hunks generated by GIT were very unhelpful, to say the least. It
split functions in half and moved them around, when the real actual
conflict only existed solely inside of one function, that being
be_map_pci_bars().
So instead, to resolve this, I checked out be_main.c from the top
of net-next, then I applied the be_main.c changes from 'net' since
the last time I merged. And this worked beautifully.
The inet_diag.c and sysctl_net_core.c conflicts were simple
overlapping changes, and were easily to resolve.
Signed-off-by: David S. Miller <davem@davemloft.net>
The recent change to tcp_cong_avoid_ai() to handle stretch ACKs
introduced a bug where snd_cwnd_cnt could accumulate a very large
value while w was large, and then if w was reduced snd_cwnd could be
incremented by a large delta, leading to a large burst and high packet
loss. This was tickled when CUBIC's bictcp_update() sets "ca->cnt =
100 * cwnd".
This bug crept in while preparing the upstream version of
814d488c61.
Testing: This patch has been tested in datacenter netperf transfers
and live youtube.com and google.com servers.
Fixes: 814d488c61 ("tcp: fix the timid additive increase on stretch ACKs")
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 message isn't really needed it justs waits time/space.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/vxlan.c
drivers/vhost/net.c
include/linux/if_vlan.h
net/core/dev.c
The net/core/dev.c conflict was the overlap of one commit marking an
existing function static whilst another was adding a new function.
In the include/linux/if_vlan.h case, the type used for a local
variable was changed in 'net', whereas the function got rewritten
to fix a stacked vlan bug in 'net-next'.
In drivers/vhost/net.c, Al Viro's iov_iter conversions in 'net-next'
overlapped with an endainness fix for VHOST 1.0 in 'net'.
In drivers/net/vxlan.c, vxlan_find_vni() added a 'flags' parameter
in 'net-next' whereas in 'net' there was a bug fix to pass in the
correct network namespace pointer in calls to this function.
Signed-off-by: David S. Miller <davem@davemloft.net>
Change Reno to properly handle stretch ACKs in additive increase mode
by passing in the count of ACKed packets to tcp_cong_avoid_ai().
In addition, if snd_cwnd crosses snd_ssthresh during slow start
processing, and we then exit slow start mode, we need to carry over
any remaining "credit" for packets ACKed and apply that to additive
increase by passing this remaining "acked" count to
tcp_cong_avoid_ai().
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>
tcp_cong_avoid_ai() was too timid (snd_cwnd increased too slowly) on
"stretch ACKs" -- cases where the receiver ACKed more than 1 packet in
a single ACK. For example, suppose w is 10 and we get a stretch ACK
for 20 packets, so acked is 20. We ought to increase snd_cwnd by 2
(since acked/w = 20/10 = 2), but instead we were only increasing cwnd
by 1. This patch fixes that behavior.
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>
This patch adds necessary infrastructure to the congestion control
framework for later per route congestion control support.
For a per route congestion control possibility, our aim is to store
a unique u32 key identifier into dst metrics, which can then be
mapped into a tcp_congestion_ops struct. We argue that having a
RTAX key entry is the most simple, generic and easy way to manage,
and also keeps the memory footprint of dst entries lower on 64 bit
than with storing a pointer directly, for example. Having a unique
key id also allows for decoupling actual TCP congestion control
module management from the FIB layer, i.e. we don't have to care
about expensive module refcounting inside the FIB at this point.
We first thought of using an IDR store for the realization, which
takes over dynamic assignment of unused key space and also performs
the key to pointer mapping in RCU. While doing so, we stumbled upon
the issue that due to the nature of dynamic key distribution, it
just so happens, arguably in very rare occasions, that excessive
module loads and unloads can lead to a possible reuse of previously
used key space. Thus, previously stale keys in the dst metric are
now being reassigned to a different congestion control algorithm,
which might lead to unexpected behaviour. One way to resolve this
would have been to walk FIBs on the actually rare occasion of a
module unload and reset the metric keys for each FIB in each netns,
but that's just very costly.
Therefore, we argue a better solution is to reuse the unique
congestion control algorithm name member and map that into u32 key
space through jhash. For that, we split the flags attribute (as it
currently uses 2 bits only anyway) into two u32 attributes, flags
and key, so that we can keep the cacheline boundary of 2 cachelines
on x86_64 and cache the precalculated key at registration time for
the fast path. On average we might expect 2 - 4 modules being loaded
worst case perhaps 15, so a key collision possibility is extremely
low, and guaranteed collision-free on LE/BE for all in-tree modules.
Overall this results in much simpler code, and all without the
overhead of an IDR. Due to the deterministic nature, modules can
now be unloaded, the congestion control algorithm for a specific
but unloaded key will fall back to the default one, and on module
reload time it will switch back to the expected algorithm
transparently.
Joint work with Florian Westphal.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can just move this to an extra function and make the code
a bit more readable, no functional change.
Joint work with Florian Westphal.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
No caller uses the return value, so make this function return void.
Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Split assignment and initialization from one into two functions.
This is required by followup patches that add Datacenter TCP
(DCTCP) congestion control algorithm - we need to be able to
determine if the connection is moderated by DCTCP before the
3WHS has finished.
As we walk the available congestion control list during the
assignment, we are always guaranteed to have Reno present as
it's fixed compiled-in. Therefore, since we're doing the
early assignment, we don't have a real use for the Reno alias
tcp_init_congestion_ops anymore and can thus remove it.
Actual usage of the congestion control operations are being
made after the 3WHS has finished, in some cases however we
can access get_info() via diag if implemented, therefore we
need to zero out the private area for those modules.
Joint work with Daniel Borkmann and Glenn Judd.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: Glenn Judd <glenn.judd@morganstanley.com>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
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>
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>
Yuchung discovered tcp_is_cwnd_limited() was returning false in
slow start phase even if the application filled the socket write queue.
All congestion modules take into account tcp_is_cwnd_limited()
before increasing cwnd, so this behavior limits slow start from
probing the bandwidth at full speed.
The problem is that even if write queue is full (aka we are _not_
application limited), cwnd can be under utilized if TSO should auto
defer or TCP Small queues decided to hold packets.
So the in_flight can be kept to smaller value, and we can get to the
point tcp_is_cwnd_limited() returns false.
With TCP Small Queues and FQ/pacing, this issue is more visible.
We fix this by having tcp_cwnd_validate(), which is supposed to track
such things, take into account unsent_segs, the number of segs that we
are not sending at the moment due to TSO or TSQ, but intend to send
real soon. Then when we are cwnd-limited, remember this fact while we
are processing the window of ACKs that comes back.
For example, suppose we have a brand new connection with cwnd=10; we
are in slow start, and we send a flight of 9 packets. By the time we
have received ACKs for all 9 packets we want our cwnd to be 18.
We implement this by setting tp->lsnd_pending to 9, and
considering ourselves to be cwnd-limited while cwnd is less than
twice tp->lsnd_pending (2*9 -> 18).
This makes tcp_is_cwnd_limited() more understandable, by removing
the GSO/TSO kludge, that tried to work around the issue.
Note the in_flight parameter can be removed in a followup cleanup
patch.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/wireless/ath/ath9k/recv.c
drivers/net/wireless/mwifiex/pcie.c
net/ipv6/sit.c
The SIT driver conflict consists of a bug fix being done by hand
in 'net' (missing u64_stats_init()) whilst in 'net-next' a helper
was created (netdev_alloc_pcpu_stats()) which takes care of this.
The two wireless conflicts were overlapping changes.
Signed-off-by: David S. Miller <davem@davemloft.net>
tcp_is_cwnd_limited() allows GSO/TSO enabled flows to increase
their cwnd to allow a full size (64KB) TSO packet to be sent.
Non GSO flows only allow an extra room of 3 MSS.
For most flows with a BDP below 10 MSS, this results in a bloat
of cwnd reaching 90, and an inflate of RTT.
Thanks to TSO auto sizing, we can restrict the bloat to the number
of MSS contained in a TSO packet (tp->xmit_size_goal_segs), to keep
original intent without performance impact.
Because we keep cwnd small, it helps to keep TSO packet size to their
optimal value.
Example for a 10Mbit flow, with low TCP Small queue limits (no more than
2 skb in qdisc/device tx ring)
Before patch :
lpk51:~# ./ss -i dst lpk52:44862 | grep cwnd
cubic wscale:6,6 rto:215 rtt:15.875/2.5 mss:1448 cwnd:96
ssthresh:96
send 70.1Mbps unacked:14 rcv_space:29200
After patch :
lpk51:~# ./ss -i dst lpk52:52916 | grep cwnd
cubic wscale:6,6 rto:206 rtt:5.206/0.036 mss:1448 cwnd:15
ssthresh:14
send 33.4Mbps unacked:4 rcv_space:29200
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Nandita Dukkipati <nanditad@google.com>
Cc: Van Jacobson <vanj@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 684bad1107 "tcp: use PRR to reduce cwin in CWR state" removed all
calls to min_cwnd, so we can safely remove it.
Also, remove tcp_reno_min_cwnd because it was only used for min_cwnd.
Signed-off-by: Stanislav Fomichev <stfomichev@yandex-team.ru>
Acked-by: Yuchung Cheng <ycheng@google.com>
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>
TCP Appropriate Byte Count was added by me, but later disabled.
There is no point in maintaining it since it is a potential source
of bugs and Linux already implements other better window protection
heuristics.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since commit 9dc274151a (tcp: fix ABC in tcp_slow_start()),
a nul snd_cwnd triggers an infinite loop in tcp_slow_start()
Avoid this infinite loop and log a one time error for further
analysis. FRTO code is suspected to cause this bug.
Reported-by: Pasi Kärkkäinen <pasik@iki.fi>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull trivial branch from Jiri Kosina:
"Usual stuff -- comment/printk typo fixes, documentation updates, dead
code elimination."
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/trivial: (39 commits)
HOWTO: fix double words typo
x86 mtrr: fix comment typo in mtrr_bp_init
propagate name change to comments in kernel source
doc: Update the name of profiling based on sysfs
treewide: Fix typos in various drivers
treewide: Fix typos in various Kconfig
wireless: mwifiex: Fix typo in wireless/mwifiex driver
messages: i2o: Fix typo in messages/i2o
scripts/kernel-doc: check that non-void fcts describe their return value
Kernel-doc: Convention: Use a "Return" section to describe return values
radeon: Fix typo and copy/paste error in comments
doc: Remove unnecessary declarations from Documentation/accounting/getdelays.c
various: Fix spelling of "asynchronous" in comments.
Fix misspellings of "whether" in comments.
eisa: Fix spelling of "asynchronous".
various: Fix spelling of "registered" in comments.
doc: fix quite a few typos within Documentation
target: iscsi: fix comment typos in target/iscsi drivers
treewide: fix typo of "suport" in various comments and Kconfig
treewide: fix typo of "suppport" in various comments
...
Allow an unpriviled user who has created a user namespace, and then
created a network namespace to effectively use the new network
namespace, by reducing capable(CAP_NET_ADMIN) and
capable(CAP_NET_RAW) calls to be ns_capable(net->user_ns,
CAP_NET_ADMIN), or capable(net->user_ns, CAP_NET_RAW) calls.
Settings that merely control a single network device are allowed.
Either the network device is a logical network device where
restrictions make no difference or the network device is hardware NIC
that has been explicity moved from the initial network namespace.
In general policy and network stack state changes are allowed
while resource control is left unchanged.
Allow creating raw sockets.
Allow the SIOCSARP ioctl to control the arp cache.
Allow the SIOCSIFFLAG ioctl to allow setting network device flags.
Allow the SIOCSIFADDR ioctl to allow setting a netdevice ipv4 address.
Allow the SIOCSIFBRDADDR ioctl to allow setting a netdevice ipv4 broadcast address.
Allow the SIOCSIFDSTADDR ioctl to allow setting a netdevice ipv4 destination address.
Allow the SIOCSIFNETMASK ioctl to allow setting a netdevice ipv4 netmask.
Allow the SIOCADDRT and SIOCDELRT ioctls to allow adding and deleting ipv4 routes.
Allow the SIOCADDTUNNEL, SIOCCHGTUNNEL and SIOCDELTUNNEL ioctls for
adding, changing and deleting gre tunnels.
Allow the SIOCADDTUNNEL, SIOCCHGTUNNEL and SIOCDELTUNNEL ioctls for
adding, changing and deleting ipip tunnels.
Allow the SIOCADDTUNNEL, SIOCCHGTUNNEL and SIOCDELTUNNEL ioctls for
adding, changing and deleting ipsec virtual tunnel interfaces.
Allow setting the MRT_INIT, MRT_DONE, MRT_ADD_VIF, MRT_DEL_VIF, MRT_ADD_MFC,
MRT_DEL_MFC, MRT_ASSERT, MRT_PIM, MRT_TABLE socket options on multicast routing
sockets.
Allow setting and receiving IPOPT_CIPSO, IP_OPT_SEC, IP_OPT_SID and
arbitrary ip options.
Allow setting IP_SEC_POLICY/IP_XFRM_POLICY ipv4 socket option.
Allow setting the IP_TRANSPARENT ipv4 socket option.
Allow setting the TCP_REPAIR socket option.
Allow setting the TCP_CONGESTION socket option.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
