The ip command would always lookup the network device index
even when not necessary. This slows down operations like creating
lots of VLAN's.
David reported the original issue, this is an alternative patch
that solves it in a slightly more general method.
Using iproute2 to create a bridge and add 4094 vlans to it can take from
2 to 3 *minutes*. The reason is the extraneous call to ll_name_to_index.
ll_name_to_index results in an ioctl(SIOCGIFINDEX) call which in turn
invokes dev_load. If the index does not exist, which it won't when
creating a new link, dev_load calls modprobe twice -- once for
netdev-NAME and again for NAME. This is unnecessary overhead for each
link create.
When ip link is invoked for a new device, there is no reason to
call ll_name_to_index for the new device. With this patch, creating
a bridge and adding 4094 vlans takes less than 3 *seconds*.
old:
# time ip -batch ip-vlan.batch
real 3m13.727s
user 0m0.076s
sys 0m1.959s
new:
# time ip -batch ip-vlan.batch
real 0m3.222s
user 0m0.044s
sys 0m1.777s
Reported-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
rta_expires is a signed int; print it as one.
Fixes: 663c3cb231 ("iproute: implement JSON and color output")
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Currently NETNS_RUN_DIR is hardcoded and refers to /var/run/netns.
However, some systems (e.g. Android) doesn't have /var
which results in error attempts to create network namespaces on these
systems. This change makes NETNS_RUN_DIR configurable at build time
by allowing to pass environment variable to make command.
Also, this change makes /etc/netns directory configurable through
NETNS_ETC_DIR environment variable.
For example: ./configure && NETNS_RUN_DIR=/mnt/vendor/netns make
Tested: verified that iproute2 with configuration mentioned above
creates namespaces in /mnt/vendor/netns
Signed-off-by: Pavel Maltsev <pavelm@google.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Steve Wise says:
====================
This series enhances the iproute2 rdma tool to include displaying
driver-specific resource attributes. It is the user-space part of the
kernel driver resource tracking series that has been accepted for merging
into linux-4.18 [1]
If there are no additional review comments, it can now be merged, I think.
Changes since v2:
- resync rdma_netlink.h to fix uapi break
Changes since v1:
- commit log editorial fixes
- cite kernel commits that updated rdma_netlink.h in the
iproute2 commit syncing this header
- reorder stack definitions ala "reverse christmas tree"
- correctly handle unknown driver attributes when printing
Changes since v0/rfc:
- changed "provider" to "driver" based on kernel side changes
- updated man pages
- removed "RFC" tag
Thanks,
Steve.
[1] https://www.spinics.net/lists/linux-rdma/msg64199.html
====================
Signed-off-by: David Ahern <dsahern@gmail.com>
Update the man pages for the resource attributes as well
as the driver-specific attributes.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
This enhancement allows printing rdma device-specific state, if provided
by the kernel. This is done in a generic manner, so rdma tool doesn't
need to know about the details of every type of rdma device.
Driver attributes for a rdma resource are in the form of <key,
[print_type], value> tuples, where the key is a string and the value can
be any supported driver attribute. The print_type attribute, if present,
provides a print format to use vs the standard print format for the type.
For example, the default print type for a PROVIDER_S32 value is "%d ",
but "0x%x " if the print_type of PRINT_TYPE_HEX is included inthe tuple.
Driver resources are only printed when the -dd flag is present.
If -p is present, then the output is formatted to not exceed 80 columns,
otherwise it is printed as a single row to be grep/awk friendly.
Example output:
# rdma resource show qp lqpn 1028 -dd -p
link cxgb4_0/- lqpn 1028 rqpn 0 type RC state RTS rq-psn 0 sq-psn 0 path-mig-state MIGRATED pid 0 comm [nvme_rdma]
sqid 1028 flushed 0 memsize 123968 cidx 85 pidx 85 wq_pidx 106 flush_cidx 85 in_use 0
size 386 flags 0x0 rqid 1029 memsize 16768 cidx 43 pidx 41 wq_pidx 171 msn 44 rqt_hwaddr 0x2a8a5d00
rqt_size 256 in_use 128 size 130
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
We make it easier for users to correlate between 128-bit node
identities and 32-bit node hash number by extending the 'node list'
command to also show the hash number.
We also improve the 'nametable show' command to show the node identity
instead of the node hash number. Since the former potentially is much
longer than the latter, we make room for it by eliminating the (to the
user) irrelevant publication key. We also reorder some of the columns so
that the node id comes last, since this looks nicer and is more logical.
Signed-off-by: David Ahern <dsahern@gmail.com>
In order to make TDC tests match the output patterns, the missing space
character must be added in the mode output string.
Fixes: 8744c5d338 ("tc: jsonify ife action")
Signed-off-by: Roman Mashak <mrv@mojatatu.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Currently there is no way to log offloading errors if the rule is not
explicitly marked as skip_sw, making it hard for other applications such
as Open vSwitch to log why a given could not be offloaded.
This patch adds support for signaling the kernel that more verbose
logging is wanted, which now will include such messages.
Signed-off-by: Marcelo Ricardo Leitner <mleitner@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Allowing 0% is sometimes useful for example in netem loss and drop
or perhaps dropping all traffic in a HTB bin.
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=199745
Reported-by: stuartmarsden@gmail.com
Fixes: 927e3cfb52 ("tc: B.W limits can now be specified in %.")
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
As the kernel code says, limit is actually the amount of packets it can
hold queued at a time, as per:
static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
struct sk_buff **to_free)
{
...
if (unlikely(sch->q.qlen >= sch->limit))
return qdisc_drop_all(skb, sch, to_free);
So lets fix the description of the field in the man page.
Signed-off-by: Marcelo Ricardo Leitner <mleitner@redhat.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
As the kernel code says, limit is actually the amount of packets it can
hold queued at a time, as per:
static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
struct sk_buff **to_free)
{
...
if (unlikely(sch->q.qlen >= sch->limit))
return qdisc_drop_all(skb, sch, to_free);
So lets fix the description of the field in the man page.
Signed-off-by: Marcelo Ricardo Leitner <mleitner@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Users have reported a regression due to ip now dropping capabilities
unconditionally.
zerotier-one VPN and VirtualBox use ambient capabilities in their
binary and then fork out to ip to set routes and links, and this
does not work anymore.
As a workaround, do not drop caps if CAP_NET_ADMIN (the most common
capability used by ip) is set with the INHERITABLE flag.
Users that want ip vrf exec to work do not need to set INHERITABLE,
which will then only set when the calling program had privileges to
give itself the ambient capability.
Fixes: ba2fc55b99 ("Drop capabilities if not running ip exec vrf with libcap")
Signed-off-by: Luca Boccassi <bluca@debian.org>
In this commit we introduce the ability to set and get
MTU for UDP media and bearer.
For set and get properties such as tolerance, window and priority,
we already do:
$ tipc media set PPROPERTY media MEDIA
$ tipc media get PPROPERTY media MEDIA
$ tipc bearer set OPTION media MEDIA ARGS
$ tipc bearer get [OPTION] media MEDIA ARGS
The same has been extended for MTU, with an exception to support
only media type UDP.
Acked-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: GhantaKrishnamurthy MohanKrishna <mohan.krishna.ghanta.krishnamurthy@ericsson.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Ss was using slabinfo to try and intuit TCP statistics.
The slabinfo changed several times since 2.4 and all these statistics
are broken by renames and slab merging. Plus slabinfo does not exist
at all if kernel is compiled with SLUB option.
Rather than trying to fix kernel, just trim away the no longer
valid statistics.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
The iproute2 header files must be complete to allow builds on
other places where some of the headers are not present.
For example, iproute2 is built on Windows Services for Linux
as a test tool. With the partial addition of rdma it was broken.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
This makes rdma/include/uapi/rdma headers align with those produced
by doing make headers_install from upstream (Linus) tree.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Perf arrays are handled specially by the kernel, don't request
offload even when used by an offloaded program.
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David Ahern <dsahern@gmail.com>
Continue parsing a multipath payload as long as another nexthop can fit
in the payload.
# ip route add 192.0.2.0/24 nexthop dev dummy0 nexthop dev dummy1
Before:
# ip route show 192.0.2.0/24
192.0.2.0/24
nexthop dev dummy0 weight 1
After:
# ip route show 192.0.2.0/24
192.0.2.0/24
nexthop dev dummy0 weight 1
nexthop dev dummy1 weight 1
Fixes: f48e14880a ("iproute: refactor multipath print")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Explicit link with pthread is not needed when linking dynamically. Even
static link with recent libdb does not pull in the code that uses
pthread. Finally, the configure check introduced in commit a25df4887d
(configure: Check for Berkeley DB for arpd compilation) does not add
-lpthread to its link command.
This change allows arpd build with toolchains that do not provide
threads support.
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Debian does not distribute libdb4.x-dev for quite some time now. Current
stable carries libdb5.3-dev. Update the wording accordingly.
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
print_uint() will silently promote its variable type to uint64_t, but there
is nothing that ensures that the format string specifier passed along with
it fits (and the function name suggest to pass "%u").
Fix this by changing print_uint() to use a native 'unsigned int' type, and
introduce a separate print_u64() function for printing 64-bit values. All
call sites that were actually printing 64-bit values using print_uint() are
converted to use print_u64() instead.
Since print_int() was already using native int types, just add a
print_s64() to match, but don't convert any call sites. For symmetry,
also add a print_luint() method (with no users).
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
The dash printed by the ingress qdisc breaks JSON output, so only print it
in regular output mode.
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Currently, iproute allows setting those flags, but it's impossible to
clear them, since their current value is fetched from the kernel and
then we OR in the additional flags passed on the command line.
Add no* variants to allow clearing them.
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: David Ahern <dsahern@gmail.com>
GRE tunnels are currently only documented together with IPIP and SIT
tunnels, but they actually have very different configuration
options. Let's separate them.
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: David Ahern <dsahern@gmail.com>
Commit 6c4b672738 ("iplink_geneve: Get rid of inet_get_addr()")
inadvertently changed the parameter to addattr_l() resulting in:
addattr_l ERROR: message exceeded bound of 4
when remote is specified.
Fixes: 6c4b672738 ("iplink_geneve: Get rid of inet_get_addr()")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Quentin Monnet <quentin.monnet@netronome.com>
In theory, the path for BPF could exceed the 4K PATH_MAX.
In practice, not really possible. But shut up gcc.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Like kernel net-next commit 72f6d71e491e6 ("vxlan: add ttl inherit support"),
vxlan ttl inherit should means inherit the inner protocol's ttl value.
But currently when we add vxlan with "ttl inherit", we only set ttl 0,
which is actually use whatever default value instead of inherit the inner
protocol's ttl value.
To make a difference with ttl inherit and ttl == 0, we add an attribute
IFLA_VXLAN_TTL_INHERIT when "ttl inherit" specified. And use "ttl auto"
to means "use whatever default value", the same behavior with ttl == 0.
Reported-by: Jianlin Shi <jishi@redhat.com>
Suggested-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Thomas reported a change in behavior with respect to autodectecting
address families. Specifically, 'ip ro add default via fe80::1'
syntax was failing to treat fe80::1 as an IPv6 address as it did in
prior releases. The root causes appears to be a change in family when
the default keyword is parsed.
'default', 'any' and 'all' are relevant outside of AF_INET. Leave the
family arg as is for these when setting addr.
Fixes: 93fa12418d ("utils: Always specify family and ->bytelen in get_prefix_1()")
Reported-by: Thomas Deutschmann <whissi@gentoo.org>
Signed-off-by: David Ahern <dsahern@gmail.com>
Cc: Serhey Popovych <serhe.popovych@gmail.com>
Attempt to add a multipath route where a nexthop definition refers to a
non-existent device causes 'ip' to crash and burn due to stack buffer
overflow:
# ip -6 route add fd00::1/64 nexthop dev fake1
Cannot find device "fake1"
Cannot find device "fake1"
Cannot find device "fake1"
...
Segmentation fault (core dumped)
Don't ignore errors from the helper routine that parses the nexthop
definition, and abort immediately if parsing fails.
Signed-off-by: Jakub Sitnicki <jkbs@redhat.com>
