The main purpose of locking the memory is to ensure that OVS can keep
doing what it did before in case of increased memory pressure, e.g.,
during VM ingest / migration. Fulfilling this requirement can be
achieved without locking all the allocated memory, but only the pages
already accessed in the past (faulted in). Processing of the new
traffic involves new memory allocations. Latency on these operations
can't be guaranteed by the locking. The main difference would be
the pre-faulting of the stack memory. However, in order to revalidate
or process upcalls on the same traffic, the same amount of stack is
likely needed, so all the necessary memory will already be faulted in.
Switch 'mlockall' to MCL_ONFAULT to avoid consuming unnecessarily
large amounts of RAM on systems with high core counts. For example,
in a densely populated OVN cluster this saves about 650 MB of RAM per
node on a system with 64 cores. This equates to 320 GB of allocated
but unused RAM in a 500 node cluster.
This also makes OVS better suited by default for small systems with
limited amount of memory.
The MCL_ONFAULT flag was introduced in Linux kernel 4.4 and wasn't
available at the time of '--mlockall' introduction, but we can use it
now. Falling back to an old way of locking in case we're running on
an older kernel just in case.
Only locking the faulted in pages also makes locking compatible with
vhost post-copy live migration by default, because we'll no longer
pre-fault all the guest's memory. Post-copy relies on userfaultfd
to work on shared huge pages, which is only available in 4.11+ kernels.
So, technically, it should not be possible for MCL_ONFAULT to fail and
the call without it to succeed. But keeping the check just in case
for now.
Acked-by: Simon Horman <horms@ovn.org>
Acked-by: Eelco Chaudron <echaudro@redhat.com>
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
Extract checking for a given kernel version to a separate function.
It will be used also in the next patch.
Acked-by: Mike Pattrick <mkp@redhat.com>
Acked-by: Eelco Chaudron <echaudro@redhat.com>
Acked-by: Aaron Conole <aconole@redhat.com>
Signed-off-by: Felix Huettner <felix.huettner@mail.schwarz>
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
The make clang-analyze target reports an 'Dereference of null
pointer' and an 'Uninitialized argument value' issue due to
it assumes some function can return NULL.
This patch annotates these functions, so the static analyzer
is aware of this.
Acked-by: Dumitru Ceara <dceara@redhat.com>
Acked-by: Simon Horman <horms@ovn.org>
Signed-off-by: Eelco Chaudron <echaudro@redhat.com>
The default Linux timer slack groups timer expires into 50 uS intervals.
With some traffic patterns this can mean that returning to process
packets after a sleep takes too long and packets are dropped.
Add a helper to util.c and set use it to reduce the timer slack
for PMD threads, so that sleeps with smaller resolutions can be done
to prevent sleeping for too long.
Fixes: de3bbdc479a9 ("dpif-netdev: Add PMD load based sleeping.")
Reported-at: https://mail.openvswitch.org/pipermail/ovs-dev/2023-January/401121.html
Reported-by: Ilya Maximets <i.maximets@ovn.org>
Signed-off-by: David Marchand <david.marchand@redhat.com>
Co-authored-by: Kevin Traynor <ktraynor@redhat.com>
Signed-off-by: Kevin Traynor <ktraynor@redhat.com>
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
xnanosleep forces the thread into quiesce state in anticipation that
it will be sleeping for a considerable time and that the thread may
need to quiesce before the sleep is finished.
In some cases, a very short sleep may be requested and in that case
the overhead of going to into quiesce state may be unnecessary.
To allow for those cases add a xnanosleep_no_quiesce() variant.
Suggested-by: Ilya Maximets <i.maximets@ovn.org>
Reviewed-by: David Marchand <david.marchand@redhat.com>
Signed-off-by: Kevin Traynor <ktraynor@redhat.com>
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
This macro is handy when it comes paste two tokens when one or both
are macros.
Rename CURSOR_JOIN() to OVS_JOIN() and move it to util.h so that it can
be reused.
Signed-off-by: Paolo Valerio <pvalerio@redhat.com>
Acked-by: Gaetan Rivet <grive@u256.net>
Acked-by: Aaron Conole <aconole@redhat.com>
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
The thread-local data allocators can't increment coverage counters
because this can cause reentrancy. Until now, this code has used
explicit calls to malloc(). This code replaces them by calls to the
new functions. This will make it easier in an upcoming patch to update
all the code that can run out of memory.
Signed-off-by: Ben Pfaff <blp@ovn.org>
Acked-by: Ilya Maximets <i.maximets@ovn.org>
While committing set() actions, commit() could wildcard all the fields
that are same in match key and in the set action. This leads to
situation where mask after commit could actually contain less bits
than it was before. And if set action was partially committed, all
the fields that were the same will be cleared out from the matching key
resulting in the incorrect (too wide) flow.
For example, for the flow that matches on both src and dst mac
addresses, if the dst mac is the same and only src should be changed
by the set() action, destination address will be wildcarded in the
match key and will never be matched, i.e. flows with any destination
mac will match, which is not correct.
Setting OF rule:
in_port=1,dl_src=50:54:00:00:00:09 actions=mod_dl_dst(50:54:00:00:00:0a),output(2)
Sending following packets on port 1:
1. eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0a),eth_type(0x0800)
2. eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0c),eth_type(0x0800)
3. eth(src=50:54:00:00:00:0b,dst=50:54:00:00:00:0c),eth_type(0x0800)
Resulted datapath flows:
eth(dst=50:54:00:00:00:0c),<...>, actions:set(eth(dst=50:54:00:00:00:0a)),2
eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0a),<...>, actions:2
The first flow doesn't have any match on source MAC address and the
third packet successfully matched on it while it must be dropped.
Fix that by updating the match mask with only the new bits set by
commit(), but keeping those that were cleared (OR operation).
With fix applied, resulted correct flows are:
eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0a),<...>, actions:2
eth(src=50:54:00:00:00:09,dst=50:54:00:00:00:0c),<...>,
actions:set(eth(dst=50:54:00:00:00:0a)),2
eth(src=50:54:00:00:00:0b),<...>, actions:drop
The code before commit dbf4a92800d0 was not able to reduce the mask,
it was only possible to expand it to exact match, so it was OK to
update original matching mask with the new value in all cases.
Fixes: dbf4a92800d0 ("odp-util: Do not rewrite fields with the same values as matched")
Reported-at: https://bugzilla.redhat.com/show_bug.cgi?id=1854376
Acked-by: Eli Britstein <elibr@mellanox.com>
Tested-by: Adrián Moreno <amorenoz@redhat.com>
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
The patch introduces experimental AF_XDP support for OVS netdev.
AF_XDP, the Address Family of the eXpress Data Path, is a new Linux socket
type built upon the eBPF and XDP technology. It is aims to have comparable
performance to DPDK but cooperate better with existing kernel's networking
stack. An AF_XDP socket receives and sends packets from an eBPF/XDP program
attached to the netdev, by-passing a couple of Linux kernel's subsystems
As a result, AF_XDP socket shows much better performance than AF_PACKET
For more details about AF_XDP, please see linux kernel's
Documentation/networking/af_xdp.rst. Note that by default, this feature is
not compiled in.
Signed-off-by: William Tu <u9012063@gmail.com>
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Userspace datapath needs to traverse through miniflow values many times. In
this process, 'count_1bits' operation for 'Flowmap' significantly impact
performance. On arm, this function was defined by portable implementation
because gcc for arm does not support popcnt feature.
But in the aarch64, VCNT neon instruction can accelerate "count_1bits".
From Gcc-7, the built-in function is implemented with neon intruction.
In this patch, count_1bits function will be impelmented with gcc built-in
from gcc-7 on, and with neon intrinsics in gcc-6.
Performance test was run in two aarch64 machines. In the NIC2NIC test, one
tuple dpcls lookup case achieves around 4% throughput improvement and
10(average) tuples case achieves around 5% improvement.
Tested-by: Malvika Gupta <malvika.gupta@arm.com>
Signed-off-by: Yanqin Wei <Yanqin.Wei@arm.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
Needed for the future post-copy live migration support for
vhost-user ports.
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
This allows a controller to change the name of OpenFlow flow tables in the
OVS software switch.
CC: Brad Cowie <brad@cowie.nz>
Acked-by: Justin Pettit <jpettit@ovn.org>
Signed-off-by: Ben Pfaff <blp@ovn.org>
Introduce new 'OVS_CTL_TIMEOUT' environment variable
that, if set, will be used as a default timeout for
OVS control utilities. Setting it in 'atlocal.in' will
cover all the hangs inside the testsuite, even when
utils called in a subshell.
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
abs_file_name() believed that a file name that begins with / or contains :
is absolute and that any other file name is relative. On Windows, this is
wrong in at least the following ways:
* / and \ are interchangeable on Windows.
* A name that begins with \\ or // is also absolute.
* A name that begins with X: but not X:\ is not absolute.
* A name with : in some position other than the second position is
not absolute (although it might not be valid either?).
Furthermore, Windows has more than one current working directory (one per
volume letter), so trying to make a file name absolute by just prefixing
the current working directory for the current volume results in silliness.
This patch attempts to fix the problem.
This makes OVS link against shlwapi, which is needed to use
PathIsRelative().
Found by inspection.
Acked-by: Alin Gabriel Serdean <aserdean@ovn.org>
Signed-off-by: Ben Pfaff <blp@ovn.org>
Daniel Alvarez Sanchez reported a significant overall speedup in ovn-northd
due to a similar patch.
Reported-by: Daniel Alvarez Sanchez <dalvarez@redhat.com>
Reported-at: https://mail.openvswitch.org/pipermail/ovs-discuss/2018-February/046120.html
Signed-off-by: Ben Pfaff <blp@ovn.org>
Acked-by: Daniel Alvarez <dalvarez@redhat.com>
Reviewed-by: Yifeng Sun <pkusunyifeng@gmail.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
- New get_custom_stats interface function is added to netdev. It
allows particular netdev implementation to expose custom
counters in dictionary format (counter name/counter value).
- New statistics are retrieved using experimenter code and
are printed as a result to ofctl dump-ports.
- New counters are available for OpenFlow 1.4+.
- New statistics are printed to output via ofctl only if those
are present in reply message.
- New statistics definition is added to include/openflow/intel-ext.h.
- Custom statistics are implemented only for dpdk-physical
port type.
- DPDK-physical implementation uses xstats to collect statistics.
Only dropped and error counters are exposed.
Co-authored-by: Ben Pfaff <blp@ovn.org>
Signed-off-by: Ben Pfaff <blp@ovn.org>
Signed-off-by: Michal Weglicki <michalx.weglicki@intel.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
FreeBSD insists that <sys/types.h> be included before <netinet/in.h> and
that <netinet/in.h> be included before <arpa/inet.h>. This adds guards to
the "sparse" headers to yield a warning if this order is violated. This
commit also adjusts the order of many #includes to suit this requirement.
Signed-off-by: Ben Pfaff <blp@ovn.org>
Acked-by: Justin Pettit <jpettit@ovn.org>
Will be used to convert strings to unsigned long long.
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
Acked-by: Jan Scheurich <jan.scheurich@ericsson.com>
This commit introduces xnanosleep() for the threads needing high
resolution sleep timeouts.
usleep() that provides microsecond granularity is deprecated and threads
wanting sub-second(ms,us,ns) granularity can use this implementation.
Signed-off-by: Bhanuprakash Bodireddy <bhanuprakash.bodireddy@intel.com>
Acked-by: Alin Gabriel Serdean <aserdean@ovn.org>
Signed-off-by: Ben Pfaff <blp@ovn.org>
This patch fixes three C++ compilation errors when it includes
"lib/packets.h".
1) Fix in "include/openvswitch/util.h" is to avoid duplicated
named_member__ in struct pkt_metadata.
2) Fix in "lib/packets.h" is because designated initializers are not
implemented in GNU C++ [1].
3) Fix in "lib/util.h" is because __builtin_types_compatible_p and
__builtin_choose_expr are only supported in GCC. I use one solution
for C++ that is type-safe and works at compile time from [2].
[1]: https://gcc.gnu.org/onlinedocs/gcc/Designated-Inits.html
[2]: https://goo.gl/xNe48A
Signed-off-by: Yi-Hung Wei <yihung.wei@gmail.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
PADDED_MEMBERS_CACHELINE_MARKER macro introduces a way to mark
cachelines.
This macro expands to an anonymous union containing cacheline marker,
members in nested anonymous structure, followed by array of bytes that
is multiple of UNIT bytes.
Signed-off-by: Bhanuprakash Bodireddy <bhanuprakash.bodireddy@intel.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
This makes it easy for callers to choose all-ones or all-zeros based on
a parameter instead of choice of function.
Signed-off-by: Ben Pfaff <blp@ovn.org>
Acked-by: Andy Zhou <azhou@ovn.org>
As part of retrieving system statistics, process status APIs along with
helper functions were implemented. Some of them are very generic and can
be reused by other subsystems.
Move the APIs in system-stats.c to process.c and util.c and make them
available. This patch doesn't change any functionality.
Signed-off-by: Bhanuprakash Bodireddy <bhanuprakash.bodireddy@intel.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
Eliminate a number of instances of undefined behavior related to
passing NULL in parameters having "nonnull" annotations.
Found with gcc's undefined behavior sanitizer.
Signed-off-by: Lance Richardson <lrichard@redhat.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
When both arguments to ovs_strlcpy() are character arrays, it makes sense
to just pass the smaller of their sizes as the overall size. It's
somewhat error-prone and definitely redundant to write that by hand, so
this commit adds a new macro that does it automatically.
Signed-off-by: Ben Pfaff <blp@ovn.org>
Acked-by: Andy Zhou <azhou@ovn.org>
Adding / removing a range of integers to a column accepting a set of
integers requires enumarating all of the integers. This patch simplifies
it by introducing 'range' concept to the database commands. Two integers
separated by a hyphen represent an inclusive range.
The patch adds positive and negative tests for the new syntax.
The patch was tested by 'make check'. Covarage was tested by
'make check-lcov'.
Signed-off-by: Lukasz Rzasik <lukasz.rzasik@gmail.com>
Suggested-by: <my_ovs_discuss@yahoo.com>
Suggested-by: Ben Pfaff <blp@ovn.org>
Signed-off-by: Ben Pfaff <blp@ovn.org>
This commit adds the conntrack module.
It is a connection tracker that resides entirely in userspace. Its
primary user will be the dpif-netdev datapath.
The module main goal is to provide conntrack_execute(), which offers a
convenient interface to implement the datapath ct() action.
The conntrack module uses two submodules to deal with the l4 protocol
details (conntrack-other for UDP and ICMP, conntrack-tcp for TCP).
The conntrack-tcp submodule implementation is adapted from FreeBSD's pf
subsystem, therefore it's BSD licensed. It has been slightly altered to
match the OVS coding style and to allow the pickup of already
established connections.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Acked-by: Antonio Fischetti <antonio.fischetti@intel.com>
Acked-by: Joe Stringer <joe@ovn.org>
Implementation of 'nullable_string_is_equal()' moved to util.c and
reused inside dpif-netdev.
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Daniele Di Proietto <diproiettod@vmware.com>
To easily allow both in- and out-of-tree building of the Python
wrapper for the OVS JSON parser (e.g. w/ pip), move json.h to
include/openvswitch. This also requires moving lib/{hmap,shash}.h.
Both hmap.h and shash.h were #include-ing "util.h" even though the
headers themselves did not use anything from there, but rather from
include/openvswitch/util.h. Fixing that required including util.h
in several C files mostly due to OVS_NOT_REACHED and things like
xmalloc.
Signed-off-by: Terry Wilson <twilson@redhat.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
Commit f2d105b5 (ofproto-dpif-xlate: xlate ct_{mark, label} correctly.)
introduced the ovs_u128_and() function. It directly takes ovs_u128
values as arguments instead of pointers to them. As this is a bit more
direct way to deal with 128-bit values, modify the other utility
functions to do the same.
Signed-off-by: Justin Pettit <jpettit@ovn.org>
Acked-by: Joe Stringer <joe@ovn.org>
When translating multiple ct actions in a row which include modification
of ct_mark or ct_labels, these fields could be incorrectly translated
into datapath actions, resulting in modification of these fields for
entries when the OpenFlow rules didn't actually specify the change.
For instance, the following OpenFlow actions:
ct(zone=1,commit,exec(set_field(1->ct_mark))),ct(zone=2,table=1),...
Would translate into the datapath actions:
ct(zone=1,commit,mark=1),ct(zone=2,mark=1),recirc(...),...
This commit fixes the issue by zeroing the wildcards for these fields
prior to performing nested actions translation (and restoring
afterwards). As such, these fields do not hold both the match and the
field modification values at the same time. As a result, the ct_mark and
ct_labels don't leak from one ct action to the next.
Fixes: 8e53fe8cf7a1 ("Add connection tracking mark support.")
Fixes: 9daf23484fb1 ("Add connection tracking label support.")
Signed-off-by: Joe Stringer <joe@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>
Removed redundant #includes and moved some macros to different file
scope
Signed-off-by: Ben Warren <ben@skyportsystems.com>
Acked-by: Ryan Moats <rmoats@us.ibm.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
This was here to help out older compilers that do not include va_copy(),
which was new in C99. All the compilers we care about these days (GCC,
Clang, MSVC) do support va_copy(), so remove it.
Signed-off-by: Ben Pfaff <blp@ovn.org>
Acked-by: Russell Bryant <russell@ovn.org>
Most of the list code is properly namespaced, so is OK to move to the
global export directory. Some "lib/util.h" code had to move to the
other directory as well, but I've tried to make that as small as
possible
Signed-off-by: Ben Warren <ben@skyportsystems.com>
Acked-by: Ryan Moats <rmoats@us.ibm.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
Add an option to ovs-ofctl utility so as to obtain colorized output in
tty, for easier reading. Currently, only the dump-flows command supports
colors.
A new `--color` option has been added to ovs-ofctl so as to indicate
whether color markers should be used or not. It can be set to `always`
(force colors), `never` (no colors) or `auto` (use colors only if output
is a tty). If provided without any value, it is the same as `auto`. If
the option is not provided at all, colors are disabled by default.
Examples:
This first call will output colorized flows:
ovs-ofctl dump-flows br0 --color=always
These two calls will produce colorized output on a tty, but they will
not use color markers if the output is redirected to a file or piped
into another command:
ovs-ofctl dump-flows br0 --color=auto
ovs-ofctl dump-flows br0 --color
These two calls will not use color markers:
ovs-ofctl dump-flows br0 --color=never
ovs-ofctl dump-flows br0
The result of this option is stored into a variable which is to be
forwarded (in next commits) as a function argument until it reaches the
functions that print the elements of the flows.
Signed-off-by: Quentin Monnet <quentin.monnet@6wind.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
Make miniflow_pad_to_64() a little more robust with regards to updates to
struct flow by passing the last field, whose end should be considered for
padding, rather than the next field, whose start should be considered.
Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>
The previous definitions of these variables using designated
initializers caused a variety of issues when attempting to compile with
MSVC, particularly if including these headers from C++ code. By defining
them like this, we can appease MSVC and keep the definitions the same on
all platforms.
VMware-BZ: #1517163
Suggested-by: Yin Lin <linyi@vmware.com>
Signed-off-by: Joe Stringer <joe@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>
These types will be used by the following patches to ensure a consistent
wire format for 128-bit connection tracking labels. Common functions for
comparison, endian translation, etc. are provided.
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
These functions could only work with 32-bit integers because of their
special cases for an argument of value 0. However, none of the existing
users depended on this special case, and some of the users did try to use
these functions with 64-bit integer arguments. Thus, this commit changes
them to support 64-bit integer arguments and drops the special cases for
zero.
This fixes a latent bug that applied rightmost_1bit_idx() to an ofpact
bitmap, which only becomes visible when an OFPACT_* with value greater than
32 is included in the bitmap.
Reported-by: Kyle Upton <kupton@baymicrosystems.com>
Reported-at: http://openvswitch.org/pipermail/dev/2015-September/060128.html
Signed-off-by: Ben Pfaff <blp@nicira.com>
This allows OVN to take advantage of the client scalability changes
that have been committed to ovsdb-server on master recently.
Conflicts:
Makefile.am
lib/learn.c
Debian likes to enable -Wformat-zero-length, even over our code trying to
disable it. It isn't too hard to make our code warning-free against this
option, so this commit both stops disabling it and fixes the warnings.
The first fix is to change set_subprogram_name() to take a plain string
instead of a format string, and to adjust its few callers. This fixes one
warning since one of those callers passed in an empty string.
The second fix is to remove a test for ovs_scan() against an empty string.
I couldn't find a way to avoid a warning for this test, and it isn't too
valuable in any case.
This allows us to drop filtering for -Wformat from the Debian rules file,
so this commit removes it.
Signed-off-by: Ben Pfaff <blp@nicira.com>
This function doesn't need to be exported in the public OVS headers, and
it had an inconsistent name compared to uuid_equals(). Rename and move.
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
Geneve options are variable length and up to 124 bytes long, which means
that they can't be easily manipulated by the integer string functions
like we do for other fields. This adds a few helper routines to make
these operations easier.
Signed-off-by: Jesse Gross <jesse@nicira.com>
Acked-by: Andy Zhou <azhou@nicira.com>
