Set of modifications and additions to establish a platform
for implementing a hardware table in userspace. Modules in
skeleton directory is used as a template for creating
hardware dependent libraries.
Fixes:
- use char instead of uint_8 for string (ofp_phy_port)
- use [0] when declaring variable-length array
(ofp_queue_get_config_reply)
Thanks to KK Yap (yapkke@stanford.edu) for reporting these issues.
This flag was added on March 3, 2008 to comply with a specific vendor's
hardware limitations. This vendor is out of business, and with the
assumption of a software datapath backup, it is unlikely that a switch
would be unable to support sending to a physical port and the controller
simultaneously.
This commmit removes the corresponding spec text and replaces the
capability bit with a reserved one.
Thanks to Justin Pettit for the history.
Enable the port stats for a particular port to be requested by
specifying a port number. Can request all port stats by specifying
OFPP_NONE for the port number.
added a human readable string to ofp_stats_desc to describe
the datapath, e.g., "switch4 in closet 3b of Gates"
* updated openflow.h
* updated udatapath: defaults to "$HOSTNAME pid=$$"
* updated kernel datapath: defaults to "$HOSTNAME idx=$i"
* updated wireshark plugin
* added regression test for ofp_desc_stats
* addressed KK's review comments from:
https://mailman.stanford.edu/pipermail/openflow-dev/2009-December/000545.html
Improve precision of flow duration (time a flow has
been installed in a switch) from seconds to nanoseconds
to improve throughput measurement accuracy when using
flow stats messages. This implementation splits the
existing duration fields into duration_sec and
duration_nsec for both flow expiration and flow stats
messages. Passes included unit tests on Ubuntu 8.04.
Private vendor extension file had separate handling for big/little
endian. All places in the code where these values are put on/removed
from the wire are wrapped by hton* and ntoh* function calls, eliminating
the need for separate constants.
Commit 238c5065 (Vlan pcp match: Reorder VLAN priority field...)
changed ofp_match from 36B to 40B, breaking the 8B alignment of a
number of structs. This commit modifies those structs to at least
pass the assertion tests and regression tests.
Enable additional basic QoS support, without requiring a full QoS
framework. This commit defines a flow action to rewrite the DiffServ
CodePoint bits of the IP ToS field in the IP header.
Before this spec change, flow expiration messages included a duration
field for the time the flow was active. To find out the amount of time
the flow received traffic would have been much harder - the controller
would need to store a record of all flow mods it has sent, to know the
idle_timeout of this flow. Additionally, both phrases, 'time flow was
active' and 'time traffic was received' were used in the spec for the
duration field.
Now we return the idle timeout in the flow expiration message, and ensure
duration is the time the entry was present in the flow table. The hard
timeout will be obvious if it is the reason for the deletion, since the
reason field will be OFPER_HARD_TIMEOUT, and the duration will be the hard
timeout value. From the duration and the idle_timeout, we can also
compute the time the flow received traffic.
A number of protocols such as SNMP and STP start counting ports at one. To
increase compatibility, this commit adds a requirement that OpenFlow
ingress port values start at one.
An OpenFlow hypervisor might choose to reject an OpenFlow request, but
currently has no message type defined for this error. It is not a
vendor-specific error, thus we define an _EPERM error code for each error
type.
Also remove unused, commented-out text.
When a switch receives a barrier message it must first complete all
commands sent before the barrier message before executing any commands
after it. When all commands before the Barrier message have completed,
it must send a barrier reply message back to the controller.
detection handler for table maintanance
Add a CHECK_OVERLAP flag to flow mods which requires the switch to do
the (potentially more costly) check that there doesn't already exist a
conflicting flow with the same priority. If there is one, the mod
fails and an error code is returned.
Flow expiry: Record OFPFF_SEND_FLOW_EXP flag in user data path
Previously, a switch would only notify the controller when flows expired.
Now, when a switch is told to delete flows, it sends a flow expiration
message, enabling the controller to get packet and byte counts from
deleted flows. Without this, the controller must retrieve flow statistics
and then delete the flow, during which time the counts could change. This
could also aid in debugging if multiple applications or programs
controlling the same switch are deleting each others' flows.
This commit adds a OFPER_DELETE as a choice for the reason field of a flow
expiration.
Make flow expirations per-flow-entry, rather than per-switch.
This commit should give controllers extra flexibility to decide when they
want expirations, will simplify the job of an OpenFlow hypervisor, and may
reduce CPU load at the switch.