For high speed adapter like Mellanox CX-5 card, it can reach upto
100 Gbits per second bandwidth. Currently htb already supports 64bit rate
in tc utility. However police action rate and peakrate are still limited
to 32bit value (upto 32 Gbits per second). Taking advantage of the 2 new
attributes TCA_POLICE_RATE64 and TCA_POLICE_PEAKRATE64 from kernel,
tc can use them to break the 32bit limit, and still keep the backward
binary compatibility.
Tested-by: David Dai <zdai@linux.vnet.ibm.com>
Signed-off-by: David Dai <zdai@linux.vnet.ibm.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
The revert of batchsize accidently reverted more than it should
and broke shared block functionality. Fix this by restoring the
original functionality.
To reproduce:
dst_ip 192.0.2.0/24 action drop
Unknown filter "block", hence option "10" is unparsable
Fixes: e991c04d64 ("Revert "tc: Add batchsize feature for filter and actions"")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
This adds support for setting the txtime_delay parameter which is useful
for the txtime offload mode of taprio.
Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
This allows a new parameter, flags, to be passed to taprio. Currently, it
only supports enabling the txtime-assist mode. But, we plan to add
different modes for taprio (e.g. hardware offloading) and this parameter
will be useful in enabling those modes.
Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
ETF Qdisc currently checks for a socket with SO_TXTIME socket option. If
either is not present, the packet is dropped. In the future commits, we
want other Qdiscs to add packet with launchtime to the ETF Qdisc. Also,
there are some packets (e.g. ICMP packets) which may not have a socket
associated with them. So, add an option to skip this check.
Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
New tc action to send packets to conntrack module, commit
them, and set a zone, labels, mark, and nat on the connection.
It can also clear the packet's conntrack state by using clear.
Usage:
ct clear
ct commit [force] [zone] [mark] [label] [nat]
ct [nat] [zone]
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Yossi Kuperman <yossiku@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Strict netlink validation now requires this flag on all nested
attributes, add it for action options.
Signed-off-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
parse_percent() currently allows to specify negative percentages
or value above 100%. However this does not seems to make sense,
as the function is used for probabilities or bandiwidth rates.
Moreover, using negative values leads to erroneous results
(using Bernoulli loss model as example):
$ ip link add test type dummy
$ ip link set test up
$ tc qdisc add dev test root netem loss gemodel -10% limit 10
$ tc qdisc show dev test
qdisc netem 800c: root refcnt 2 limit 10 loss gemodel p 90% r 10% 1-h 100% 1-k 0%
Using values above 100% we have instead:
$ ip link add test type dummy
$ ip link set test up
$ tc qdisc add dev test root netem loss gemodel 140% limit 10
$ tc qdisc show dev test
qdisc netem 800f: root refcnt 2 limit 10 loss gemodel p 40% r 60% 1-h 100% 1-k 0%
This commit changes parse_percent() with a check to ensure
percentage values stay between 1.0 and 0.0.
parse_percent_rate() function, which already employs a similar
check, is adjusted accordingly.
With this check in place, we have:
$ ip link add test type dummy
$ ip link set test up
$ tc qdisc add dev test root netem loss gemodel -10% limit 10
Illegal "loss gemodel p"
Fixes: 927e3cfb52 ("tc: B.W limits can now be specified in %.")
Signed-off-by: Andrea Claudi <aclaudi@redhat.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Many tc modules were printing error messages to stdout.
This is problematic if using JSON or other output formats.
Change all these places to use fprintf(stderr, ...) instead.
Also, remove unnecessary initialization and places
where else is used after error return.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Create a new action type for TC that allows the pushing, popping, and
modifying of MPLS headers.
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Add 32-bit missing mask attribute in iproute2/tc, which has been long
supported by the kernel side.
v2: print value in hex with print_hex() as suggested by Stephen Hemminger.
Signed-off-by: Roman Mashak <mrv@mojatatu.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
As the man page for tc netem states:
To use the Bernoulli model, the only needed parameter is p while the
others will be set to the default values r=1-p, 1-h=1 and 1-k=0.
However r parameter is erroneusly set to 1, and not to 1-p.
Fix this using the same approach of the 4-state loss model.
Fixes: 3c7950af59 ("netem: add support for 4 state and GE loss model")
Signed-off-by: Andrea Claudi <aclaudi@redhat.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Add JSON output support to q_netem.
The normal output is untouched.
In JSON output always use seconds as the base of time units,
and non-percentage numbers (0.01 instead of 1%). Try to always
report the fields, even if they are zero.
All this should make the output more machine-friendly.
v2: less macroes
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Add a new parameter '-Numeric' to show the number of protocol, scope,
dsfield, etc directly instead of converting it to human readable name.
Do the same on tc and ss.
This patch is based on David Ahern's previous patch.
Suggested-by: Phil Sutter <phil@nwl.cc>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
ctinfo is a tc action restoring data stored in conntrack marks to
various fields. At present it has two independent modes of operation,
restoration of DSCP into IPv4/v6 diffserv and restoration of conntrack
marks into packet skb marks.
It understands a number of parameters specific to this action in
additional to the usual action syntax. Each operating mode is
independent of the other so all options are optional, however not
specifying at least one mode is a bit pointless.
Usage: ... ctinfo [dscp mask [statemask]] [cpmark [mask]] [zone ZONE]
[CONTROL] [index <INDEX>]
DSCP mode
dscp enables copying of a DSCP stored in the conntrack mark into the
ipv4/v6 diffserv field. The mask is a 32bit field and specifies where
in the conntrack mark the DSCP value is located. It must be 6
contiguous bits long. eg. 0xfc000000 would restore the DSCP from the
upper 6 bits of the conntrack mark.
The DSCP copying may be optionally controlled by a statemask. The
statemask is a 32bit field, usually with a single bit set and must not
overlap the dscp mask. The DSCP restore operation will only take place
if the corresponding bit/s in conntrack mark ANDed with the statemask
yield a non zero result.
eg. dscp 0xfc000000 0x01000000 would retrieve the DSCP from the top 6
bits, whilst using bit 25 as a flag to do so. Bit 26 is unused in this
example.
CPMARK mode
cpmark enables copying of the conntrack mark to the packet skb mark. In
this mode it is completely equivalent to the existing act_connmark
action. Additional functionality is provided by the optional mask
parameter, whereby the stored conntrack mark is logically ANDed with the
cpmark mask before being stored into skb mark. This allows shared usage
of the conntrack mark between applications.
eg. cpmark 0x00ffffff would restore only the lower 24 bits of the
conntrack mark, thus may be useful in the event that the upper 8 bits
are used by the DSCP function.
Usage: ... ctinfo [dscp mask [statemask]] [cpmark [mask]] [zone ZONE]
[CONTROL] [index <INDEX>]
where :
dscp MASK is the bitmask to restore DSCP
STATEMASK is the bitmask to determine conditional restoring
cpmark MASK mask applied to restored packet mark
ZONE is the conntrack zone
CONTROL := reclassify | pipe | drop | continue | ok |
goto chain <CHAIN_INDEX>
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
the following TDC test case:
b776 - Replace simple action with invalid goto chain control
checks if the kernel correctly validates the 'goto chain' control action,
when it is specified in 'act_simple' rules. The test systematically fails
because the control action is hardcoded in parse_simple(), i.e. it is not
parsed by command line arguments, so its value is constantly TC_ACT_PIPE.
Because of that, the following command:
# tc action add action simple sdata "test" drop index 7
installs an 'act_simple' rule that never drops packets, and whose 'index'
is the first IDR available, plus an 'act_gact' rule with 'index' equal to
7, that drops packets.
Use parse_action_control_dflt(), like we did on many other TC actions, to
make the control action configurable also with 'act_simple'. The expected
results of test b776 are summarized below:
iproute2
v kernel->| 5.1-rc2 (and previous) | 5.1-rc3 (and subsequent)
------------------+-------------------------+-------------------------
5.1.0 | FAIL (bad IDR) | FAIL (bad IDR)
5.1.0(patched) | FAIL (no rule/bad sdata)| PASS
Changes since v1:
- reword commit message, thanks Stephen Hemminger
Fixes: 087f46ee4e ("tc: introduce simple action")
CC: Andrea Claudi <aclaudi@redhat.com>
CC: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
The following operation fails:
% sudo tc actions add action pipe index 1
% sudo tc filter add dev lo parent ffff: \
protocol ip pref 10 u32 match ip src 127.0.0.2 \
flowid 1:10 action gact index 1
Bad action type index
Usage: ... gact <ACTION> [RAND] [INDEX]
Where: ACTION := reclassify | drop | continue | pass | pipe |
goto chain <CHAIN_INDEX> | jump <JUMP_COUNT>
RAND := random <RANDTYPE> <ACTION> <VAL>
RANDTYPE := netrand | determ
VAL : = value not exceeding 10000
JUMP_COUNT := Absolute jump from start of action list
INDEX := index value used
However, passing a control action of gact rule during filter binding works:
% sudo tc filter add dev lo parent ffff: \
protocol ip pref 10 u32 match ip src 127.0.0.2 \
flowid 1:10 action gact pipe index 1
Binding by reference, i.e. by index, has to consistently work with
any tc action.
Since tc is sensitive to the order of keywords passed on the command line,
we can teach gact to skip parsing arguments as soon as it sees 'gact'
followed by 'index' keyword.
Signed-off-by: Roman Mashak <mrv@mojatatu.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
sscanf truncates read port values silently without any error. As sscanf
man says:
(...) sscanf() conform to C89 and C99 and POSIX.1-2001. These standards
do not specify the ERANGE error.
Replace sscanf with safer get_be16 that returns error when value is out
of range.
Example:
tc filter add dev eth0 protocol ip parent ffff: prio 1 flower ip_proto
tcp dst_port 70000 hw_tc 1
Would result in filter for port 4464 without any warning.
Fixes: 8930840e67 ("tc: flower: Classify packets based port ranges")
Signed-off-by: Lukasz Czapnik <lukasz.czapnik@intel.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
The mirred act admits an optional control action, defaulting
to TC_ACT_PIPE. The parsing code currently emits an error message
if the control action is not provided on the command line, even
if the command itself completes with no error.
This change shuts down the error message, using the appropriate
parsing helper.
Fixes: e67aba5595 ("tc: actions: add helpers to parse and print control actions")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Every tool in the iproute2 package have one or more function to show
an help message to the user. Some of these functions print the help
line by line with a series of printf call, e.g. ip/xfrm_state.c does
60 fprintf calls.
If we group all the calls to a single one and just concatenate strings,
we save a lot of libc calls and thus object size. The size difference
of the compiled binaries calculated with bloat-o-meter is:
ip/ip:
add/remove: 0/0 grow/shrink: 5/15 up/down: 103/-4796 (-4693)
Total: Before=672591, After=667898, chg -0.70%
ip/rtmon:
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-54 (-54)
Total: Before=48879, After=48825, chg -0.11%
tc/tc:
add/remove: 0/2 grow/shrink: 31/10 up/down: 882/-6133 (-5251)
Total: Before=351912, After=346661, chg -1.49%
bridge/bridge:
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-459 (-459)
Total: Before=70502, After=70043, chg -0.65%
misc/lnstat:
add/remove: 0/1 grow/shrink: 1/0 up/down: 48/-486 (-438)
Total: Before=9960, After=9522, chg -4.40%
tipc/tipc:
add/remove: 0/0 grow/shrink: 1/1 up/down: 18/-62 (-44)
Total: Before=79182, After=79138, chg -0.06%
While at it, indent some strings which were starting at column 0,
and use tabs where possible, to have a consistent style across helps.
Signed-off-by: Matteo Croce <mcroce@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
This allows a cycle-time and a cycle-time-extension to be specified.
Specifying a cycle-time will truncate that cycle, so when that instant
is reached, the cycle will start from its beginning.
A cycle-time-extension may cause the last entry of a cycle, just
before the start of a new schedule (the base-time of the "admin"
schedule) to be extended by at maximum "cycle-time-extension"
nanoseconds. The idea of this feauture, as described by the IEEE
802.1Q, is too avoid too narrow gate states.
Example:
tc qdisc change dev IFACE parent root handle 100 taprio \
sched-entry S 0x1 1000000 \
sched-entry S 0x0 2000000 \
sched-entry S 0x1 3000000 \
sched-entry S 0x0 4000000 \
cycle-time-extension 100000 \
cycle-time 9000000 \
base-time 12345678900000000
Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
This allows for a new schedule to be specified during runtime, without
removing the current one.
For that, the semantics of the 'tc qdisc change' operation in the
context of taprio is that if "change" is called and there is a running
schedule, a new schedule is created and the base-time (let's call it
X) of this new schedule is used so at instant X, it becomes the
"current" schedule. So, in short, "change" doesn't change the current
schedule, it creates a new one and sets it up to it becomes the
current one at some point.
In IEEE 802.1Q terms, it means that we have support for the
"Oper" (current and read-only) and "Admin" (future and mutable)
schedules.
Example of creating the first schedule, then adding a new one:
(1)
tc qdisc add dev IFACE parent root handle 100 taprio \
num_tc 1 \
map 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 \
queues 1@0 \
sched-entry S 0x1 1000000 \
sched-entry S 0x0 2000000 \
sched-entry S 0x1 3000000 \
sched-entry S 0x0 4000000 \
base-time 100000000 \
clockid CLOCK_TAI
(2)
tc qdisc change dev IFACE parent root handle 100 taprio \
base-time 7500000000000 \
sched-entry S 0x0 5000000 \
sched-entry S 0x1 5000000 \
It was necessary to fix a bug, so the clockid doesn't need to be
specified when changing the schedule.
Most of the changes are related to make it easier to reuse the same
function for printing the "admin" and "oper" schedules.
Signed-off-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
sch_plug can be used to perform functional qdisc unit tests
controlling explicitly the queuing behaviour from user-space.
Plug support lacks since its introduction in 2012. This change
introduces basic support, to control the tc status.
v1 -> v2:
- use the SPDX identifier
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
This adds support for the newly added fwmark option to CAKE, which allows
overriding the tin selection from the per-packet firewall marks. The fwmark
field is a bitmask that is applied to the fwmark to select the tin.
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
tc_pie_xstats->prob has a maximum value of (2^64 - 1).
Signed-off-by: Leslie Monis <lesliemonis@gmail.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
action m_connmark returns error messages identifying itself as the
'simple' action instead of 'connmark' action. e.g.
tc filter add dev eth0 protocol all u32 match u32 0 0 flowid 1:1 \
action connmark index wrong
simple: Illegal "index"
bad action parsing
parse_action: bad value (3:connmark)!
Illegal "action"
In what is most likely a copy/paste error from the simple action example
code, fix connmark error messages to identify themselves as coming from
connmark.
tc filter add dev eth0 protocol all u32 match u32 0 0 flowid 1:1 \
action connmark index wrong
connmark: Illegal "index"
bad action parsing
parse_action: bad value (3:connmark)!
Illegal "action"
While we're here also fixup the 'Illegal "Zone"' error code to say
'Illegal "zone"' instead of 'Illegal "index"'
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Tc pedit action with more than two ip6 munge in a row cause infinite
loop.
Example:
$ tc filter add dev eth0 protocol ipv6 parent ffff: \
flower ip_proto sctp \
action pedit ex \
munge ip6 hoplimit set 0x1 \
munge ip6 src set 2001:0db8:0:f101::1 \
munge that cause infinite loop
The example command never returns, instead of failing with parse error
as expected. Pedit ipv6 structure has wrong id, which leads to the
creation linked list with one node in tc/m_pedit.c:get_pedit_kind(),
referring to itself. This node is created if command have two ip6 munge
in a row, and any third ip6 munge will cause infinite loop.
Changing this id from "ipv6" to "ip6" solves the problem.
Fixes: f3e1b2448a ("pedit: Introduce ipv6 support")
Signed-off-by: Dmytro Linkin <dmitrolin@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
As /sys/class/net/<iface>/speed indicates a value in Mbits/sec, the
conversion is necessary to create the correct limits.
This guarantees the same result for the following commands in an
1000Mbit/sec device:
tc class add ... htb rate 500Mbit
tc class add ... htb rate 50%
Fixes: 927e3cfb52 ("tc: B.W limits can now be specified in %.")
Signed-off-by: Marcos Antonio Moraes <marcos.antonio@digirati.com.br>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
The parse_percent_rate function assumed the buffer was 20 characters.
Better to pass length in case the size ever changes.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
If value passed to parse_percent was not valid, it would
leak the dynamic allocation from sscanf.
Fixes: 927e3cfb52 ("tc: B.W limits can now be specified in %.")
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
unlike other TC actions already supporting JSON printout, 'csum' does not
print the value of TCA_KIND in the 'kind' property: remove 'csum' word
from 'csum' property, and add a separate 'kind' property containing the
action name. The human-readable printout is preserved.
Tested with:
# ./tdc.py -c csum
Cc: Andrea Claudi <aclaudi@redhat.com>
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
The kernel (and iproute2) don't use the if (NULL == x) style
and instead prefer if (!x)
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David Ahern <dsahern@gmail.com>
David Dai
David Ahern
Stephen Hemminger
Ido Schimmel
Vedang Patel
Vinicius Costa Gomes
Paul Blakey
Andrea Claudi
John Hurley
Roman Mashak
Jakub Kicinski
Hangbin Liu
Kevin Darbyshire-Bryant
Davide Caratti
Lukasz Czapnik
Paolo Abeni
Matteo Croce
Toke Høiland-Jørgensen
Leslie Monis
Dmytro Linkin
Marcos Antonio Moraes