Skip to content
  1. Jul 12, 2018
  3. Jul 11, 2018
    • Petr Machata's avatar
      selftests: forwarding: mirror_lib: Tighten up VLAN capture · db560d16
      Petr Machata authored 
      

The function do_test_span_vlan_dir_ips() is used for testing whether
mirrored packets are VLAN-encapsulated. But since it only considers
VLAN encapsulation, it may end up matching unmirrored ARP traffic as
well. One consequence is a rare failure of mirror_gre_vlan_bridge_1q's
test_gretap_untagged_egress. Decreasing ping cadence in mirror_test()
makes the problem easily reproducible.

Therefore tighten up the match criterion to only count those 802.1q
packets where the next header is IP.

Signed-off-by: default avatarPetr Machata <petrm@mellanox.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      db560d16
    • David S. Miller's avatar
      Merge branch 'cake-qdisc' · 5025b99c
      David S. Miller authored 
      

Toke Høiland-Jørgensen says:

====================
sched: Add Common Applications Kept Enhanced (cake) qdisc

This patch series adds the CAKE qdisc, and has been split up to ease
review.

I have attempted to split out each configurable feature into its own patch.
The first commit adds the base shaper and packet scheduler, while
subsequent commits add the optional features. The full userspace API and
most data structures are included in this commit, but options not
understood in the base version will be ignored.

The result of applying the entire series is identical to the out of tree
version that have seen extensive testing in previous deployments, most
notably as an out of tree patch to OpenWrt. However, note that I have only
compile tested the individual patches; so the whole series should be
considered as a unit.

---
Changelog

v19:
  - Rebase to current net-next.
  - Don't rely on the value of sch->q.qlen to break loops; fixes possible
    infinite loop on multi-queue devices.
  - Don't overwrite NAT flag when setting flow mode.

v18:
  - Rework classification logic in the diffserv case to always hash if
    filter doesn't select a queue, and to run TC filters before
    selecting the diffserv tin (allowing filter to influence this).
  - Make sure we always call qdisc_watchdog_init() in cake_init(), so we
    don't crash in cake_destroy().

v17:
  - Rebase to newest net-next and move the conntrack callback to
    nf_ct_hook
  - Fix a compile error when NF_CONNTRACK is unset.

v16:
  - Move conntrack lookup function into conntrack core and read it via
    RCU so it is only active when the nf_conntrack module is loaded.
    This avoids the module dependency on conntrack for NAT mode. Thanks
    to Pablo for the idea.

v15:
  - Handle ECN flags in ACK filter

v14:
  - Handle seqno wraps and DSACKs in ACK filter

v13:
  - Avoid ktime_t to scalar compares
  - Add class dumping and basic stats
  - Fail with ENOTSUPP when requesting NAT mode and conntrack is not
    available.
  - Parse all TCP options in ACK filter and make sure to only drop safe
    ones. Also handle SACK ranges properly.

v12:
  - Get rid of custom time typedefs. Use ktime_t for time and u64 for
    duration instead.

v11:
  - Fix overhead compensation calculation for GSO packets
  - Change configured rate to be u64 (I ran out of bits before I ran out
    of CPU when testing the effects of the above)

v10:
  - Christmas tree gardening (fix variable declarations to be in reverse
    line length order)

v9:
  - Remove duplicated checks around kvfree() and just call it
    unconditionally.
  - Don't pass __GFP_NOWARN when allocating memory
  - Move options in cake_dump() that are related to optional features to
    later patches implementing the features.
  - Support attaching filters to the qdisc and use the classification
    result to select flow queue.
  - Support overriding diffserv priority tin from skb->priority

v8:
  - Remove inline keyword from function definitions
  - Simplify ACK filter; remove the complex state handling to make the
    logic easier to follow. This will potentially be a bit less efficient,
    but I have not been able to measure a difference.

v7:
  - Split up patch into a series to ease review.
  - Constify the ACK filter.

v6:
  - Fix 6in4 encapsulation checks in ACK filter code
  - Checkpatch fixes

v5:
  - Refactor ACK filter code and hopefully fix the safety issues
    properly this time.

v4:
  - Only split GSO packets if shaping at speeds <= 1Gbps
  - Fix overhead calculation code to also work for GSO packets
  - Don't re-implement kvzalloc()
  - Remove local header include from out-of-tree build (fixes kbuild-bot
    complaint).
  - Several fixes to the ACK filter:
    - Check pskb_may_pull() before deref of transport headers.
    - Don't run ACK filter logic on split GSO packets
    - Fix TCP sequence number compare to deal with wraparounds

v3:
  - Use IS_REACHABLE() macro to fix compilation when sch_cake is
    built-in and conntrack is a module.
  - Switch the stats output to use nested netlink attributes instead
    of a versioned struct.
  - Remove GPL boilerplate.
  - Fix array initialisation style.

v2:
  - Fix kbuild test bot complaint
  - Clean up the netlink ABI
  - Fix checkpatch complaints
  - A few tweaks to the behaviour of cake based on testing carried out
    while writing the paper.
====================

Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      5025b99c
    • Toke Høiland-Jørgensen's avatar
      sch_cake: Conditionally split GSO segments · 0c850344
      Toke Høiland-Jørgensen authored 
      

At lower bandwidths, the transmission time of a single GSO segment can add
an unacceptable amount of latency due to HOL blocking. Furthermore, with a
software shaper, any tuning mechanism employed by the kernel to control the
maximum size of GSO segments is thrown off by the artificial limit on
bandwidth. For this reason, we split GSO segments into their individual
packets iff the shaper is active and configured to a bandwidth <= 1 Gbps.

Signed-off-by: default avatarToke Høiland-Jørgensen <toke@toke.dk>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      0c850344
    • Toke Høiland-Jørgensen's avatar
      sch_cake: Add overhead compensation support to the rate shaper · a729b7f0
      Toke Høiland-Jørgensen authored 
      

This commit adds configurable overhead compensation support to the rate
shaper. With this feature, userspace can configure the actual bottleneck
link overhead and encapsulation mode used, which will be used by the shaper
to calculate the precise duration of each packet on the wire.

This feature is needed because CAKE is often deployed one or two hops
upstream of the actual bottleneck (which can be, e.g., inside a DSL or
cable modem). In this case, the link layer characteristics and overhead
reported by the kernel does not match the actual bottleneck. Being able to
set the actual values in use makes it possible to configure the shaper rate
much closer to the actual bottleneck rate (our experience shows it is
possible to get with 0.1% of the actual physical bottleneck rate), thus
keeping latency low without sacrificing bandwidth.

The overhead compensation has three tunables: A fixed per-packet overhead
size (which, if set, will be accounted from the IP packet header), a
minimum packet size (MPU) and a framing mode supporting either ATM or PTM
framing. We include a set of common keywords in TC to help users configure
the right parameters. If no overhead value is set, the value reported by
the kernel is used.

Signed-off-by: default avatarToke Høiland-Jørgensen <toke@toke.dk>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      a729b7f0
    • Toke Høiland-Jørgensen's avatar
      sch_cake: Add DiffServ handling · 83f8fd69
      Toke Høiland-Jørgensen authored 
      

This adds support for DiffServ-based priority queueing to CAKE. If the
shaper is in use, each priority tier gets its own virtual clock, which
limits that tier's rate to a fraction of the overall shaped rate, to
discourage trying to game the priority mechanism.

CAKE defaults to a simple, three-tier mode that interprets most code points
as "best effort", but places CS1 traffic into a low-priority "bulk" tier
which is assigned 1/16 of the total rate, and a few code points indicating
latency-sensitive or control traffic (specifically TOS4, VA, EF, CS6, CS7)
into a "latency sensitive" high-priority tier, which is assigned 1/4 rate.
The other supported DiffServ modes are a 4-tier mode matching the 802.11e
precedence rules, as well as two 8-tier modes, one of which implements
strict precedence of the eight priority levels.

This commit also adds an optional DiffServ 'wash' mode, which will zero out
the DSCP fields of any packet passing through CAKE. While this can
technically be done with other mechanisms in the kernel, having the feature
available in CAKE significantly decreases configuration complexity; and the
implementation cost is low on top of the other DiffServ-handling code.

Filters and applications can set the skb->priority field to override the
DSCP-based classification into tiers. If TC_H_MAJ(skb->priority) matches
CAKE's qdisc handle, the minor number will be interpreted as a priority
tier if it is less than or equal to the number of configured priority
tiers.

Signed-off-by: default avatarToke Høiland-Jørgensen <toke@toke.dk>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      83f8fd69
    • Toke Høiland-Jørgensen's avatar
      sch_cake: Add NAT awareness to packet classifier · ea825115
      Toke Høiland-Jørgensen authored 
      

When CAKE is deployed on a gateway that also performs NAT (which is a
common deployment mode), the host fairness mechanism cannot distinguish
internal hosts from each other, and so fails to work correctly.

To fix this, we add an optional NAT awareness mode, which will query the
kernel conntrack mechanism to obtain the pre-NAT addresses for each packet
and use that in the flow and host hashing.

When the shaper is enabled and the host is already performing NAT, the cost
of this lookup is negligible. However, in unlimited mode with no NAT being
performed, there is a significant CPU cost at higher bandwidths. For this
reason, the feature is turned off by default.

Cc: netfilter-devel@vger.kernel.org
Signed-off-by: default avatarToke Høiland-Jørgensen <toke@toke.dk>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      ea825115
    • Toke Høiland-Jørgensen's avatar
      netfilter: Add nf_ct_get_tuple_skb global lookup function · b60a6040
      Toke Høiland-Jørgensen authored 
      

This adds a global netfilter function to extract a conntrack tuple from an
skb. The function uses a new function added to nf_ct_hook, which will try
to get the tuple from skb->_nfct, and do a full lookup if that fails. This
makes it possible to use the lookup function before the skb has passed
through the conntrack init hooks (e.g., in an ingress qdisc). The tuple is
copied to the caller to avoid issues with reference counting.

The function returns false if conntrack is not loaded, allowing it to be
used without incurring a module dependency on conntrack. This is used by
the NAT mode in sch_cake.

Cc: netfilter-devel@vger.kernel.org
Signed-off-by: default avatarToke Høiland-Jørgensen <toke@toke.dk>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      b60a6040
    • Toke Høiland-Jørgensen's avatar
      sch_cake: Add optional ACK filter · 8b713881
      Toke Høiland-Jørgensen authored 
      

The ACK filter is an optional feature of CAKE which is designed to improve
performance on links with very asymmetrical rate limits. On such links
(which are unfortunately quite prevalent, especially for DSL and cable
subscribers), the downstream throughput can be limited by the number of
ACKs capable of being transmitted in the *upstream* direction.

Filtering ACKs can, in general, have adverse effects on TCP performance
because it interferes with ACK clocking (especially in slow start), and it
reduces the flow's resiliency to ACKs being dropped further along the path.
To alleviate these drawbacks, the ACK filter in CAKE tries its best to
always keep enough ACKs queued to ensure forward progress in the TCP flow
being filtered. It does this by only filtering redundant ACKs. In its
default 'conservative' mode, the filter will always keep at least two
redundant ACKs in the queue, while in 'aggressive' mode, it will filter
down to a single ACK.

The ACK filter works by inspecting the per-flow queue on every packet
enqueue. Starting at the head of the queue, the filter looks for another
eligible packet to drop (so the ACK being dropped is always closer to the
head of the queue than the packet being enqueued). An ACK is eligible only
if it ACKs *fewer* bytes than the new packet being enqueued, including any
SACK options. This prevents duplicate ACKs from being filtered, to avoid
interfering with retransmission logic. In addition, we check TCP header
options and only drop those that are known to not interfere with sender
state. In particular, packets with unknown option codes are never dropped.

In aggressive mode, an eligible packet is always dropped, while in
conservative mode, at least two ACKs are kept in the queue. Only pure ACKs
(with no data segments) are considered eligible for dropping, but when an
ACK with data segments is enqueued, this can cause another pure ACK to
become eligible for dropping.

The approach described above ensures that this ACK filter avoids most of
the drawbacks of a naive filtering mechanism that only keeps flow state but
does not inspect the queue. This is the rationale for including the ACK
filter in CAKE itself rather than as separate module (as the TC filter, for
instance).

Our performance evaluation has shown that on a 30/1 Mbps link with a
bidirectional traffic test (RRUL), turning on the ACK filter on the
upstream link improves downstream throughput by ~20% (both modes) and
upstream throughput by ~12% in conservative mode and ~40% in aggressive
mode, at the cost of ~5ms of inter-flow latency due to the increased
congestion.

In *really* pathological cases, the effect can be a lot more; for instance,
the ACK filter increases the achievable downstream throughput on a link
with 100 Kbps in the upstream direction by an order of magnitude (from ~2.5
Mbps to ~25 Mbps).

Finally, even though we consider the ACK filter to be safer than most, we
do not recommend turning it on everywhere: on more symmetrical link
bandwidths the effect is negligible at best.

Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Signed-off-by: default avatarToke Høiland-Jørgensen <toke@toke.dk>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      8b713881
    • Toke Høiland-Jørgensen's avatar
      sch_cake: Add ingress mode · 7298de9c
      Toke Høiland-Jørgensen authored 
      

The ingress mode is meant to be enabled when CAKE runs downlink of the
actual bottleneck (such as on an IFB device). The mode changes the shaper
to also account dropped packets to the shaped rate, as these have already
traversed the bottleneck.

Enabling ingress mode will also tune the AQM to always keep at least two
packets queued *for each flow*. This is done by scaling the minimum queue
occupancy level that will disable the AQM by the number of active bulk
flows. The rationale for this is that retransmits are more expensive in
ingress mode, since dropped packets have to traverse the bottleneck again
when they are retransmitted; thus, being more lenient and keeping a minimum
number of packets queued will improve throughput in cases where the number
of active flows are so large that they saturate the bottleneck even at
their minimum window size.

This commit also adds a separate switch to enable ingress mode rate
autoscaling. If enabled, the autoscaling code will observe the actual
traffic rate and adjust the shaper rate to match it. This can help avoid
latency increases in the case where the actual bottleneck rate decreases
below the shaped rate. The scaling filters out spikes by an EWMA filter.

Signed-off-by: default avatarToke Høiland-Jørgensen <toke@toke.dk>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      7298de9c