This reverts commit 8bdfe1313894047d44349fa4cf4402970865950f.
I failed to see that lib/dpif-netdev.c actually needs the concurrency
provided by pvector prior to this change. More specifically, when a
subtable is removed, concurrent lookups may skip over another subtable
swapped in to the place of the removed subtable in the vector.
Since this was the only use of the non-concurrent pvector, it is
cleaner to revert the whole patch.
Reported-by: Jan Scheurich <jan.scheurich@ericsson.com>
Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Acked-by: Daniele Di Proietto <diproiettod@vmware.com>
This patch adds some comments to the dpcls_lookup() funtion,
which is one of the most important places where the Userspace
wildcard matching happens.
The purpose is to give some more explanations on its design
and also on how it works.
Signed-off-by: Antonio Fischetti <antonio.fischetti@intel.com>
Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Acked-by: Jarno Rajahalme <jarno@ovn.org>
Revalidation should work in case of 'dynamic_txqs == true'.
Fixes: 324c8374852a ("dpif-netdev: XPS (Transmit Packet Steering) implementation.")
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Daniele Di Proietto <diproiettod@vmware.com>
PMD threads use pvectors but do not need the overhead of the
concurrent version. Expose the non-concurrent version for
that use.
Note that struct pvector is renamed as struct cpvector (for concurrent
priority vector), and the former struct pvector_impl is now struct
pvector.
Signed-off-by: Jarno Rajahalme <jarno@ovn.org>
Acked-by: Ben Pfaff <blp@ovn.org>
Now that dpif_execute has a 'flow' member, it's pretty easy to access a
the flow (or the matching megaflow) in dp_execute_cb().
This means that's not necessary anymore for the connection tracker to
reextract 'dl_type' from the packet, it can be passed as a parameter.
This change means that we have to complicate sightly test-conntrack to
group the packets by dl_type before passing them to the connection
tracker.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Acked-by: Joe Stringer <joe@ovn.org>
New functions are implemented in the conntrack module to support this.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Acked-by: Flavio Leitner <fbl@sysclose.org>
New functions are implemented in the conntrack module to support this.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Acked-by: Flavio Leitner <fbl@sysclose.org>
This commit implements the OVS_ACTION_ATTR_CT action in dpif-netdev.
To allow ofproto-dpif to detect the conntrack feature, flow_put will not
discard anymore flows with ct_* fields set. We still shouldn't allow
flows with NAT bits set, since there is no support for NAT.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Acked-by: Flavio Leitner <fbl@sysclose.org>
Acked-by: Antonio Fischetti <antonio.fischetti@intel.com>
Instead of using the internal type, use the port_open_type when creating the
local port. That makes sure that whenever dpif_port_query is used, the netdev
open_type is returned instead of the "internal" type.
For other ports, that is already the case, as the netdev type is used when
creating the dp_netdev_port.
That changes the output of dpctl when showing the local port, and also when
trying to change its type. So, corresponding tests are fixed.
Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@redhat.com>
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
New 'other_config:pmd-rxq-affinity' field for Interface table to
perform manual pinning of RX queues to desired cores.
This functionality is required to achieve maximum performance because
all kinds of ports have different cost of rx/tx operations and
only user can know about expected workload on different ports.
Example:
# ./bin/ovs-vsctl set interface dpdk0 options:n_rxq=4 \
other_config:pmd-rxq-affinity="0:3,1:7,3:8"
Queue #0 pinned to core 3;
Queue #1 pinned to core 7;
Queue #2 not pinned.
Queue #3 pinned to core 8;
It's decided to automatically isolate cores that have rxq explicitly
assigned to them because it's useful to keep constant polling rate on
some performance critical ports while adding/deleting other ports
without explicit pinning of all ports.
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Next patches will add new conditions when reconfiguration will be
required. It'll be simpler to have common way to request reconfiguration.
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Daniele Di Proietto <diproiettod@vmware.com>
This commit adds functionality to pass value of 'other_config' column
of 'Interface' table to datapath.
This may be used to pass not directly connected with netdev options and
configure behaviour of the datapath for different ports.
For example: pinning of rx queues to polling threads in dpif-netdev.
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Daniele Di Proietto <diproiettod@vmware.com>
If CPU number in pmd-cpu-mask is not divisible by the number of queues and
in a few more complex situations there may be unfair distribution of TX
queue-ids between PMD threads.
For example, if we have 2 ports with 4 queues and 6 CPUs in pmd-cpu-mask
such distribution is possible:
<------------------------------------------------------------------------>
pmd thread numa_id 0 core_id 13:
port: vhost-user1 queue-id: 1
port: dpdk0 queue-id: 3
pmd thread numa_id 0 core_id 14:
port: vhost-user1 queue-id: 2
pmd thread numa_id 0 core_id 16:
port: dpdk0 queue-id: 0
pmd thread numa_id 0 core_id 17:
port: dpdk0 queue-id: 1
pmd thread numa_id 0 core_id 12:
port: vhost-user1 queue-id: 0
port: dpdk0 queue-id: 2
pmd thread numa_id 0 core_id 15:
port: vhost-user1 queue-id: 3
<------------------------------------------------------------------------>
As we can see above dpdk0 port polled by threads on cores:
12, 13, 16 and 17.
By design of dpif-netdev, there is only one TX queue-id assigned to each
pmd thread. This queue-id's are sequential similar to core-id's. And
thread will send packets to queue with exact this queue-id regardless
of port.
In previous example:
pmd thread on core 12 will send packets to tx queue 0
pmd thread on core 13 will send packets to tx queue 1
...
pmd thread on core 17 will send packets to tx queue 5
So, for dpdk0 port after truncating in netdev-dpdk:
core 12 --> TX queue-id 0 % 4 == 0
core 13 --> TX queue-id 1 % 4 == 1
core 16 --> TX queue-id 4 % 4 == 0
core 17 --> TX queue-id 5 % 4 == 1
As a result only 2 of 4 queues used.
To fix this issue some kind of XPS implemented in following way:
* TX queue-ids are allocated dynamically.
* When PMD thread first time tries to send packets to new port
it allocates less used TX queue for this port.
* PMD threads periodically performes revalidation of
allocated TX queue-ids. If queue wasn't used in last
XPS_TIMEOUT_MS milliseconds it will be freed while revalidation.
* XPS is not working if we have enough TX queues.
Reported-by: Zhihong Wang <zhihong.wang@intel.com>
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
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>
The PMD thread needs to keep processing RX queues in order
to achieve maximum throughput. It also needs to sweep emc
cache and quiesce which use seq_mutex. That mutex can
eventually block the PMD thread causing latency spikes and
affecting the throughput.
Since there is no requirement for running those tasks at a
specific time, this patch extend seq API to allow tentative
locking instead.
Reported-by: Karl Rister <krister@redhat.com>
Co-authored-by: Karl Rister <krister@redhat.com>
Signed-off-by: Flavio Leitner <fbl@redhat.com>
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Currently, there are few inconsistencies in ways to configure number of
queues for netdev device:
* dpif-netdev can't know about exact number of queues
allocated inside netdev.
This leads to constant mapping of queue-ids to 'real' ones.
* We are able to configure 'n_rxq' for vhost-user devices, but
there is only one sane number of rx queues which must be used
and configured manually (number of queues that allocated
in QEMU).
This patch disables configuration of 'n_rxq' for DPDK vHost devices.
Configuration of rx and tx queues now automatically applied from
connected virtio device. Standard reconfiguration mechanism was used to
apply this changes.
Also, now 'n_txq' and 'n_rxq' are always the real numbers of queues
in the device.
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
The patch adds a new action to support packet truncation. The new action
is formatted as 'output(port=n,max_len=m)', as output to port n, with
packet size being MIN(original_size, m).
One use case is to enable port mirroring to send smaller packets to the
destination port so that only useful packet information is mirrored/copied,
saving some performance overhead of copying entire packet payload. Example
use case is below as well as shown in the testcases:
- Output to port 1 with max_len 100 bytes.
- The output packet size on port 1 will be MIN(original_packet_size, 100).
# ovs-ofctl add-flow br0 'actions=output(port=1,max_len=100)'
- The scope of max_len is limited to output action itself. The following
packet size of output:1 and output:2 will be intact.
# ovs-ofctl add-flow br0 \
'actions=output(port=1,max_len=100),output:1,output:2'
- The Datapath actions shows:
# Datapath actions: trunc(100),1,1,2
Tested-at: https://travis-ci.org/williamtu/ovs-travis/builds/140037134
Signed-off-by: William Tu <u9012063@gmail.com>
Acked-by: Pravin B Shelar <pshelar@ovn.org>
When debug logging is enabled, dpif-netdev can print each flow as it is
installed, which it currently does using OpenFlow match formatting. Compared
to ODP formatting, there generally isn't too much difference since the
fields are largely the same but it is inconsistent with other logging in
dpif-netdev as well as the analogous functions that deal with the kernel.
However, in some cases there is a difference between the two formats, such
as in the cases of input port or tunnel metadata. For input port, datapath
format helped detect that the generated masks were incorrect. As for tunnels,
at the moment, it's possible to convert between the two formats on demand as
we have a global metadata table. In the future, though this won't be possible
as the metadata table becomes per-bridge which the datapath won't have access
to.
Signed-off-by: Jesse Gross <jesse@kernel.org>
Acked-by: Daniele Di Proietto <diproiettod@vmware.com>
When calling odp_flow_key_from_flow (or _mask), the in_port included
as part of the flow is ignored and must be explicitly passed as a
separate parameter. This is because the assumption was that the flow's
version would often be in OFP format, rather than ODP.
However, at this point all flows that are ready for serialization in
netlink format already have their in_port properly set to ODP format.
As a result, every caller needs to explicitly initialize the extra
paramter to the value that is in the flow. This switches to just use
the value in the flow to simply things and avoid the possibility of
forgetting to initialize the extra parameter.
Signed-off-by: Jesse Gross <jesse@kernel.org>
Acked-by: Daniele Di Proietto <diproiettod@vmware.com>
This commit moves the code that sets the pmd threads affinity from
netdev-dpdk to ovs-numa. There's one small part left in netdev-dpdk, to
set the lcore_id.
Now dpif-netdev will call both modules (ovs-numa and netdev-dpdk) when
starting a pmd thread.
This change will allow having a dummy implementation of the set affinity
call, for testing purposes.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
Currently 'dpctl/flow-get' doesn't work for flows installed by
PMD threads.
Fix that by implementing search across all PMD threads. Will be returned
flow from first PMD thread with match.
Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
This introduces in dpif-netdev and netdev-dpdk the first use for the
newly introduce reconfigure netdev call.
When a request to change the number of queues comes, netdev-dpdk will
remember this and notify the upper layer via
netdev_request_reconfigure().
The datapath, instead of periodically calling netdev_set_multiq(), can
detect this and call reconfigure().
This mechanism can also be used to:
* Automatically match the number of rxq with the one provided by qemu
via the new_device callback.
* Provide a way to change the MTU of dpdk devices at runtime.
* Move a DPDK vhost device to the proper NUMA socket.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Tested-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
Errors returned by netdev_set_multiq() and netdev_rxq_open() weren't
handled properly in reconfigure_pmd_threads(). In case of error now we
remove the port from the datapath.
Also, part of the code is moved in a new function, port_reconfigure().
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Tested-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
netdev objects are hard to use with RCU, because it's not possible to
split removal and reclamation. Postponing the removal means that the
port is not removed and cannot be readded immediately. Waiting for
reclamation means introducing a quiescent state, and that may introduce
subtle bugs, due to the RCU model we use in userspace.
This commit changes the port container from cmap to hmap. 'port_mutex'
must be held by readers and writers. This shouldn't have performance
impact, as readers in the fast path use a thread local cache.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Tested-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
A future commit will stop using RCU for 'dp->ports' and use a mutex for
reading/writing them. To avoid taking a mutex in dp_execute_cb(), which
is called in the fast path, this commit introduces a pmd thread local
cache of ports.
The downside is that every port add/remove now needs to synchronize with
every pmd thread.
Among the advantages, keeping a per thread port mapping could allow
greater control over the txq assigment.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Tested-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
The pmds and the main threads are synchronized using a condition
variable. The main thread writes a new configuration, then it waits on
the condition variable. A pmd thread reads the new configuration, then
it calls signal() on the condition variable. To make sure that the pmds
and the main thread have a consistent view, each signal() should be
backed by a wait().
Currently the first signal() doesn't have a corresponding wait(). If
the pmd thread takes a long time to start and the signal() is received
by a later wait, the threads will have an inconsistent view.
The commit fixes the problem by removing the first signal() from the
pmd thread.
This is hardly a problem on current master, because the main thread
will call the first wait() a long time after the creation of a pmd
thread. It becomes a problem with the next commits.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Tested-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
This commit introduces some functions to add/remove rxqs from pmd
threads without reloading them. They will be used by next commits.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Tested-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
Instead of performing every operation inside do_port_add() it seems
clearer to introduce port_create(), since we already have
port_destroy().
No functional change.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Tested-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
The datapath code expects the RSS hash to always be initialized. This
is enforced by checking in emc_processing() that the hash is valid, and
eventually by computing a new one.
Unfortunately, there is another entry point to the datapath,
dpif_netdev_execute(). A packet generated by OVS (BFD frame,
packet-out from controller) doesn't have a valid RSS hash and so is
allowed to enter the datapath with an uninitialized hash value.
This commit recomputes the hash (if not valid) in dpif_netdev_execute().
The only place where we would use an invalid hash is netdev-vport, in
push_udp_header(). This caused an uninitialized memory read, and a
random value to be assigned to the outer tunnel header source port.
Reported-by: William Tu <u9012063@gmail.com>
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Acked-by: William Tu <u9012063@gmail.com>
Acked-by: Ben Pfaff <blp@ovn.org>
Large segment support need to use this refactored function to
send individual segments.
Signed-off-by: Pravin B Shelar <pshelar@ovn.org>
Acked-by: Jesse Gross <jesse@kernel.org>
Once datapath support large packets, we need to segment packet before
sending it to upcall. Refactoring this code make it bit cleaner.
Signed-off-by: Pravin B Shelar <pshelar@ovn.org>
Acked-by: Jesse Gross <jesse@kernel.org>
in case of error from netdev_push_header() batch of packets was not
freed. Following patch fixes this issue.
Signed-off-by: Pravin B Shelar <pshelar@ovn.org>
Acked-by: Jesse Gross <jesse@kernel.org>
The tunnel header pop action can leak batch of packet
in case of error. Following patch fixex the error code path.
Signed-off-by: Pravin B Shelar <pshelar@ovn.org>
Acked-by: Jesse Gross <jesse@kernel.org>
DPDK datapath operate on batch of packets. To pass the batch of
packets around we use packets array and count. Next patch needs
to associate meta-data with each batch of packets. So Introducing
a batch structure to make handling the metadata easier.
Signed-off-by: Pravin B Shelar <pshelar@ovn.org>
Acked-by: Jesse Gross <jesse@kernel.org>
Next patch introduces new structure named packet_batch. So
I am renaming it to packet_batch_per_flow.
This does not change any functionality.
Signed-off-by: Pravin B Shelar <pshelar@ovn.org>
Acked-by: Jesse Gross <jesse@kernel.org>
Current tunnel-pop API does not allow the netdev implementation
retain a packet but STT can keep a packet from batch of packets
during TCP reassembly processing. To return exact count of
valid packet STT need to pass this number of packet parameter
as a reference.
Signed-off-by: Pravin B Shelar <pshelar@ovn.org>
Acked-by: Jesse Gross <jesse@kernel.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>
After removing a flow from the dpcls classifier there might still be
readers who have access to the flow, until the next grace period.
Setting flow->cr.mask to NULL can cause concurrent readers to crash,
so this commit avoids doing it.
The crash can be reproduced, for example, by invoking an operation
that cause datapath flows to be deleted (such as `ovs-appctl
upcall/enable-megaflows`) while traffic is running.
I think the assignment was intended just as a safety measure to catch
race conditions, and it should be safe to remove.
Here's a stack trace of a possible crash:
Program terminated with signal SIGSEGV, Segmentation fault.
rule=0x7f3ae8006190) at ../lib/dpif-netdev.c:4156
4156 if (OVS_UNLIKELY((value & *maskp++) != *keyp++)) {
(gdb) bt
rule=0x7f3ae8006190) at ../lib/dpif-netdev.c:4156
rules=0x7f3afa3f2e40, cnt=<optimized out>) at ../lib/dpif-netdev.c:4225
(pmd=pmd@entry=0x7f3afa3fc010, packets=packets@entry=0x7f3afa3fa420,
cnt=cnt@entry=32, keys=keys@entry=0x7f3afa3f6428,
batches=batches@entry=0x7f3afa3f4118,
n_batches=n_batches@entry=0x7f3afa3fa3b0)
at ../lib/dpif-netdev.c:3483
(pmd=pmd@entry=0x7f3afa3fc010, packets=packets@entry=0x7f3afa3fa420,
cnt=<optimized out>, md_is_valid=md_is_valid@entry=false,
port_no=<optimized out>) at ../lib/dpif-netdev.c:3625
cnt=<optimized out>, packets=0x7f3afa3fa420, pmd=0x7f3afa3fc010) at
../lib/dpif-netdev.c:3642
rxq=<optimized out>, port=<optimized out>, port=<optimized out>) at
../lib/dpif-netdev.c:2574
../lib/dpif-netdev.c:2693
../lib/ovs-thread.c:340
pthread_create.c:312
../sysdeps/unix/sysv/linux/x86_64/clone.S:111
Fixes: 361d808dd9e4("flow: Split miniflow's map.")
CC: Jarno Rajahalme <jarno@ovn.org>
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>
Only the main thread will delete ports after pausing every other
thread. There's no need to keep count.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Tested-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Mark Kavanagh <mark.b.kavanagh@intel.com>
It is only used in the testsuite and it can be replaced by a dpctl
command.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Tested-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Mark Kavanagh <mark.b.kavanagh@intel.com>
This will ease deleting a port with no open rxqs.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Tested-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Mark Kavanagh <mark.b.kavanagh@intel.com>
This fixes multiple error path mistakes in do_add_port, none of which
has been a problem in practice so far. This change will make it easier
for a following commit to return in case of error.
Also, this removes an unneeded special case for tunnel ports.
Signed-off-by: Daniele Di Proietto <diproiettod@vmware.com>
Tested-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Ilya Maximets <i.maximets@samsung.com>
Acked-by: Mark Kavanagh <mark.b.kavanagh@intel.com>
