struct list is a common name and can't be used in public headers.
Signed-off-by: Thomas Graf <tgraf@noironetworks.com>
Acked-by: Ben Pfaff <blp@nicira.com>
This is a thin wrapper around an atomic_uint. It is useful anyhow because
each ovs_refcount_ref() or ovs_refcount_unref() call saves a few lines of
code.
This commit also changes all the potential direct users over to use the new
data structure.
Signed-off-by: Ben Pfaff <blp@nicira.com>
This patch transitions mac learning away from using tags as required
by future patches.
Signed-off-by: Ethan Jackson <ethan@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>
Until now, datapath ports and openflow ports were both represented by
unsigned integers of various sizes. With implicit conversions, etc., it is
easy to mix them up and use one where the other is expected. This commit
creates two typedefs, ofp_port_t and odp_port_t. Both of these two types
are marked by "__attribute__((bitwise))" so that sparse can be used to
detect any misuse.
Signed-off-by: Alex Wang <alexw@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
With this patch, the mac_learning module takes responsibility for
remembering tags which need revalidation after a
mac_learning_changed() call. This removes one of
ofproto-dpif-xlate's dpif_backer uses.
Signed-off-by: Ethan Jackson <ethan@nicira.com>
This commit changes variable "int i" of "union port" of "struct mac_entry"
to "uint16_t ofp_port", since it is used to store the OpenFlow port number.
Signed-off-by: Alex Wang <alexw@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
We've had a couple of requests for this over the years. It's easy to do,
so let's implement it.
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Kyle Mestery <kmestery@cisco.com>
Replaced all instances of Nicira Networks(, Inc) to Nicira, Inc.
Feature #10593
Signed-off-by: Raju Subramanian <rsubramanian@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
When Cisco (and other?) routers are configured in high-availability modes,
they use two different MAC addresses. The router uses MAC 1 only for ARP
replies. The router uses MAC 2 for forwarding IP packets to end hosts.
When a MAC learning switch is attached to the router, therefore, it will
only learn the location of MAC 1 from ARP replies. If the end host's ARP
cache refresh timer is longer than the switch's MAC learning timeout, then
packets to the router will be flooded from the MAC learning timeout until
the next ARP reply.
This commit fixes the problem by increasing the MAC learning timeout from
60 seconds to 300 seconds. According to research by Sanjay Sane, this is
always sufficient, even with operating systems that use ARP timeouts
longer than 300 seconds (such as FreeBSD and Mac OS, which have 1200
seconds ARP timeouts) because the routers that cause this problem send
unsolicited ARP replies every 180 seconds.
This issue arises in any situation where traffic between two hosts flows
only in one direction. The explanation above describes only one special
case.
NICS-11.
Signed-off-by: Ben Pfaff <blp@nicira.com>
Most callers of mac_learning_flush() need to revalidate all flows
anyway, but this update forces new callers of mac_learning_flush()
to think about whether or not they need to explicitly handle
revalidation. Also, it's theoretically more efficient when there
are lots of flows but the learning table is empty, (e.g. bridges
that don't use the NORMAL action).
Signed-off-by: Ethan Jackson <ethan@nicira.com>
The mac-learning implementation used a free list to keep track of
statically allocated table entries. This made the code slightly
more difficult to understand than the more straightforward heap
based strategy implemented by this patch.
The mac-learning code predates the hmap data structure in OVS. For
this reason, it rolled it's own hmap-like bucket concept. This
patch converts it to use an hmap which makes the code slightly
simpler and easier to understand.
These new semantics are less efficient in the case where the flood_vlans
actually changed, but that should be very rare.
There are no advantages to this change on its own, but upcoming commits
will add multiple layers between the code supplying the flood_vlans and
actually calling mac_learning_set_flood_vlans(). Consistency in this
multilayered interface seems valuable, and the rest of it does not transfer
ownership from the caller to the callee.
In an upcoming commit I want to store a pointer in MAC learning entries
in the bridge, instead of an integer port number. The MAC learning library
has other clients, and the others do not gracefully fit this new model, so
in fact the data will have to become a union. However, this does not fit
well with the current mac_learning API, since mac_learning_learn()
currently initializes and compares the data. It seems better to break up
the API so that only the client has to know the data's format and how to
initialize it or compare it. This commit makes this possible.
This commit doesn't change the type of the data stored in a MAC learning
entry yet.
As a side effect this commit has the benefit that clients that don't need
gratuitous ARP locking don't have to specify any policy for it at all.
Normally we filter out packets received on a bond if we have
learned the source MAC as belonging to another port to avoid packets
sent on one slave and reflected back on another. The exception to
this is gratuitous ARPs because they indicate that the host
has moved to another port. However, this can result in an additional
problem on the switch that the host moved to if the gratuitous ARP is
reflected back on a bond slave. In this case, we incorrectly relearn
the slave as the source of the MAC address. To solve this, we lock the
learning entry for 5 seconds after receiving a gratuitous ARP against
further updates caused by gratuitous ARPs on bond slaves.
Bug #2516
Reported-by: Ian Campbell <ian.campbell@citrix.com>
RSPAN does not work properly unless MAC learning for the VLAN is
disabled on all switches between the origin and monitoring point.
This allows learning to be disabled on a given VLAN so vSwitch can
acts as an intermediate switch.
Feature #2136
Given a possible 1,024 ports on a bridge the previous limit of 2,048
entries seems low.
If we want to increase this further we should introduce dynamic allocation
of table entries to avoid wasting memory in the common case.
CC: Keith Amidon <keith@nicira.com>
The vswitchd bonding code needs to iterate through the table entries to
be able to send out gratuitous learning packets when bond slaves go down.
It might be best to create an abstract interface to the MAC learning table,
but this commit does the simpler thing and exposes the data structures in
the header file.
