openvswitch
16 years agoPrevent the secchan from dying due to SIGPIPE.
Ben Pfaff [Tue, 12 Aug 2008 23:21:45 +0000 (16:21 -0700)]
Prevent the secchan from dying due to SIGPIPE.

Fix for bug #88.

16 years agoReduce default burst limit to rate limit / 4.
Ben Pfaff [Tue, 12 Aug 2008 23:15:42 +0000 (16:15 -0700)]
Reduce default burst limit to rate limit / 4.

With the previous default burst limit of rate limit * 2, we would queue
up 2 seconds worth of packet_in messages.  This is not only much more
than actually needed, it causes an actual problem: the datapath only
retains buffered packets for up to 1 second, by default, so that flow
setups sent in response have no packet to work with.

16 years agoMake discovery happen faster.
Ben Pfaff [Tue, 12 Aug 2008 23:09:31 +0000 (16:09 -0700)]
Make discovery happen faster.

This reduces the initial delay before sending DHCPDISCOVER from
1...10 seconds to 1...3 seconds.

16 years agoImplement rate limiting in secchan.
Ben Pfaff [Tue, 12 Aug 2008 22:34:47 +0000 (15:34 -0700)]
Implement rate limiting in secchan.

16 years agoDrop rconn's responsibility for limiting the tx queue.
Ben Pfaff [Mon, 11 Aug 2008 23:24:24 +0000 (16:24 -0700)]
Drop rconn's responsibility for limiting the tx queue.

Now it helps clients track the number of in-flight messages, but lets the
clients do the limiting themselves.  This will come in handy for packet-in
rate limiting (in an upcoming commit), in which we want to track in-flight
packet-in messages separately from other in-flight messages.

16 years agoUse signal-based timekeeping functions throughout the source base.
Ben Pfaff [Mon, 11 Aug 2008 20:44:29 +0000 (13:44 -0700)]
Use signal-based timekeeping functions throughout the source base.

16 years agoInitialize the timetracking module in each userspace program.
Ben Pfaff [Tue, 12 Aug 2008 20:13:16 +0000 (13:13 -0700)]
Initialize the timetracking module in each userspace program.

Also, use time_alarm() instead of the POSIX alarm() function, because
alarm() interferes with time_alarm().

16 years agoImplement signal-driven time tracking.
Ben Pfaff [Tue, 12 Aug 2008 21:41:48 +0000 (14:41 -0700)]
Implement signal-driven time tracking.

Checking the current time only when a periodic timer expires allows us
to reduce the number of system calls to obtain the current time to 10 per
second (from an arbitrary number) plus one per poll loop.

This fixes poll_block() to properly handle being interrupted.
Otherwise, a timeout longer than the interval between signals will never
complete (because it will restart from the beginning after every
signal).

16 years agoMake fatal-signal more willing to share signals with other handlers.
Ben Pfaff [Tue, 12 Aug 2008 21:24:10 +0000 (14:24 -0700)]
Make fatal-signal more willing to share signals with other handlers.

16 years agoSet up flow to drop packets that come in on their output port.
Ben Pfaff [Fri, 8 Aug 2008 22:39:20 +0000 (15:39 -0700)]
Set up flow to drop packets that come in on their output port.

Commit 0a3d4ba3a, "Don't allow the learning switch to send packets back
out the incoming port," drops packets that come in on their incoming port,
but it does not set up a flow to drop those packets.  If we are connected
to a hub, then this can be very expensive, since we can continue to get
many packets in that flow and have to drop them in secchan one by one.  So
now we set up a flow to drop them for us.

16 years agoFix handling of IP but non-TCP, non-UDP packets in kernel flow_extract().
Ben Pfaff [Fri, 8 Aug 2008 18:43:41 +0000 (11:43 -0700)]
Fix handling of IP but non-TCP, non-UDP packets in kernel flow_extract().

In particular, ICMP was broken.

16 years agoMake vconn_open() consistently set *vconnp to NULL on failure.
Ben Pfaff [Thu, 7 Aug 2008 22:37:44 +0000 (15:37 -0700)]
Make vconn_open() consistently set *vconnp to NULL on failure.

Fixes issue #80.

16 years agoImplement OpenFlow hard timeouts.
Ben Pfaff [Thu, 7 Aug 2008 22:19:31 +0000 (15:19 -0700)]
Implement OpenFlow hard timeouts.

This updates the OpenFlow protocol version and changes the names of
structure members, so any software that uses OpenFlow will need to be
updated to match.

16 years agoFix formatting of flow matches in ofp-print:
Ben Pfaff [Wed, 6 Aug 2008 18:39:51 +0000 (11:39 -0700)]
Fix formatting of flow matches in ofp-print:

Print a comma after wildcard matches as well as exact matches.

16 years agoFix uninitialized struct member in userspace table-linear.
Ben Pfaff [Wed, 6 Aug 2008 18:40:14 +0000 (11:40 -0700)]
Fix uninitialized struct member in userspace table-linear.

16 years agoFix bug in clearing unsupported switch flags.
Ben Pfaff [Thu, 7 Aug 2008 20:19:09 +0000 (13:19 -0700)]
Fix bug in clearing unsupported switch flags.

We need to keep the flags we support, not drop them.

16 years agoDon't call kfree_skb() with interrupts disabled.
Ben Pfaff [Thu, 7 Aug 2008 18:46:22 +0000 (11:46 -0700)]
Don't call kfree_skb() with interrupts disabled.

Freeing an skb that has a destructor may require interrupts to be enabled.
This can happen when netfilter is performing NAT, for example.

Discovered by Murphy McCauley.

16 years agoDon't allow unsupported flags to be set in the switch.
Ben Pfaff [Tue, 5 Aug 2008 20:51:29 +0000 (13:51 -0700)]
Don't allow unsupported flags to be set in the switch.

16 years agoImplement OFPC_FRAG_DROP fragment handling policy.
Ben Pfaff [Mon, 4 Aug 2008 22:34:26 +0000 (15:34 -0700)]
Implement OFPC_FRAG_DROP fragment handling policy.

16 years agoReduce redundancy in datapath code.
Ben Pfaff [Tue, 5 Aug 2008 20:58:12 +0000 (13:58 -0700)]
Reduce redundancy in datapath code.

16 years agoMaintain userspace switch configuration in host byte order.
Ben Pfaff [Tue, 5 Aug 2008 21:13:54 +0000 (14:13 -0700)]
Maintain userspace switch configuration in host byte order.

This makes it easier to work with and makes the code more like the
kernel switch.

16 years agoFix memory leak when OFPP_TABLE is used for a packet that matches no flow.
Ben Pfaff [Tue, 5 Aug 2008 21:35:26 +0000 (14:35 -0700)]
Fix memory leak when OFPP_TABLE is used for a packet that matches no flow.

16 years agoDrop controller-bound traffic that arrives on the controller's port.
Ben Pfaff [Tue, 5 Aug 2008 20:05:56 +0000 (13:05 -0700)]
Drop controller-bound traffic that arrives on the controller's port.

Before, if a hub connected a number of OpenFlow switches and the controller,
then in-band control traffic from one of the OpenFlow switches would be
seen by each of the other switches and forwarded up to the controller as
an ofp_packet_in message.  That message would then be seen by all of the
other OpenFlow switches and also forwarded, and so on in an infinite loop.

This change prevents this situation by keeping secchan from forwarding
ofp_packet_in messages for a packet destined to the controller that
arrives on the port where the controller is located.

This code has at least two weaknesses.  First, if the controller's port
changes, then the flows set up to drop packets will not be deleted.  This
should not be a major problem: if this inadvertently kills a switch's
connection to the controller, then the switch will realize it after it
stops receiving data and re-connect.  Its new connection will have new
flow data and therefore its packets will not be dropped.

Second, the notion of the "controller's port" does not take into account
the possibility of loops in the network topology.  We need spanning tree
protocol for that.

16 years agoProperly track table match counts.
Justin Pettit [Wed, 6 Aug 2008 08:58:15 +0000 (01:58 -0700)]
Properly track table match counts.

16 years agoFix typo in comment.
Ben Pfaff [Mon, 4 Aug 2008 22:30:19 +0000 (15:30 -0700)]
Fix typo in comment.

16 years agoMark functions not used in other source files static.
Ben Pfaff [Mon, 4 Aug 2008 21:47:27 +0000 (14:47 -0700)]
Mark functions not used in other source files static.

16 years agoMake it easier for the compiler to optimize IP_IS_FRAGMENT.
Ben Pfaff [Mon, 4 Aug 2008 21:21:34 +0000 (14:21 -0700)]
Make it easier for the compiler to optimize IP_IS_FRAGMENT.

16 years agoFix bug in flow_extract() for extracting VLAN from short packet.
Ben Pfaff [Mon, 4 Aug 2008 21:20:26 +0000 (14:20 -0700)]
Fix bug in flow_extract() for extracting VLAN from short packet.

16 years agoDon't accept incomplete TCP headers when extracting flows in kernel.
Ben Pfaff [Mon, 4 Aug 2008 21:18:24 +0000 (14:18 -0700)]
Don't accept incomplete TCP headers when extracting flows in kernel.

This makes the kernel and the userspace implementations of flow
extraction behave the same way regarding headers.

16 years agoDon't try to extract UDP/TCP port numbers from IP fragments.
Ben Pfaff [Mon, 4 Aug 2008 21:45:59 +0000 (14:45 -0700)]
Don't try to extract UDP/TCP port numbers from IP fragments.

16 years agoFix typo in comment.
Ben Pfaff [Mon, 4 Aug 2008 21:18:36 +0000 (14:18 -0700)]
Fix typo in comment.

16 years agodhcp: Make client signal a change if renewing yields different options.
Ben Pfaff [Wed, 30 Jul 2008 22:39:03 +0000 (15:39 -0700)]
dhcp: Make client signal a change if renewing yields different options.

This allows the secchan to connect to a new controller if the DHCP server
starts announcing a new vconn without invalidating old leases.

16 years agodhcp: Make dhcp_option_to_string() act sensibly with null or empty options.
Ben Pfaff [Wed, 30 Jul 2008 22:45:50 +0000 (15:45 -0700)]
dhcp: Make dhcp_option_to_string() act sensibly with null or empty options.

Also, update ofp-switch-setup to parse the new syntax.

16 years agodhcp: New function dhcp_option_equals().
Ben Pfaff [Wed, 30 Jul 2008 22:45:38 +0000 (15:45 -0700)]
dhcp: New function dhcp_option_equals().

16 years agoClean up secchan code.
Ben Pfaff [Wed, 30 Jul 2008 22:00:01 +0000 (15:00 -0700)]
Clean up secchan code.

16 years agonetdev: Let netdev_get_in4(), netdev_get_in6() take null pointer argument.
Ben Pfaff [Wed, 30 Jul 2008 22:45:33 +0000 (15:45 -0700)]
netdev: Let netdev_get_in4(), netdev_get_in6() take null pointer argument.

Some callers want to just test whether the device has an address assigned
and don't care about the particular address, so this simplifies those
callers slightly.

16 years agoDebian: Listen on Unix domain socket for management connections.
Ben Pfaff [Wed, 30 Jul 2008 22:46:36 +0000 (15:46 -0700)]
Debian: Listen on Unix domain socket for management connections.

16 years agovconn: Implement Unix domain socket vconn.
Ben Pfaff [Wed, 30 Jul 2008 22:45:21 +0000 (15:45 -0700)]
vconn: Implement Unix domain socket vconn.

These are useful for local management connections because, unlike TCP
sockets, they are subject to regular file system permissions.

16 years agoMove unix socket helper functions into socket-util.
Ben Pfaff [Tue, 29 Jul 2008 21:38:00 +0000 (14:38 -0700)]
Move unix socket helper functions into socket-util.

This makes them available for use by the upcoming vconn_unix.

16 years agovconn: Use vconn_stream to factor out code from vconn_tcp.
Ben Pfaff [Wed, 30 Jul 2008 22:45:04 +0000 (15:45 -0700)]
vconn: Use vconn_stream to factor out code from vconn_tcp.

16 years agovconn: Introduce infrastructure for stream socket-based vconns.
Ben Pfaff [Wed, 30 Jul 2008 22:44:22 +0000 (15:44 -0700)]
vconn: Introduce infrastructure for stream socket-based vconns.

16 years agovconn: Allow vconns to delegate to underlying implementations.
Ben Pfaff [Wed, 30 Jul 2008 22:43:57 +0000 (15:43 -0700)]
vconn: Allow vconns to delegate to underlying implementations.

16 years agoFix typo in comment.
Ben Pfaff [Tue, 29 Jul 2008 22:03:22 +0000 (15:03 -0700)]
Fix typo in comment.

16 years agoIP_ARGS takes address of nw_addr
Natasha Gude [Wed, 30 Jul 2008 21:41:47 +0000 (14:41 -0700)]
IP_ARGS takes address of nw_addr

16 years agoDon't allow the learning switch to send packets back out the incoming port.
Justin Pettit [Wed, 30 Jul 2008 06:43:58 +0000 (23:43 -0700)]
Don't allow the learning switch to send packets back out the incoming port.

The learning switch functionality included in OpenFlow will send packets
out the port they arrived on if the learned destination is from that port.
This causes problems when newer switches flood packets for destinations they
don't know and OpenFlow reinjects packets back into the network.  The correct
behavior is to just drop the packet.

16 years agorconn: Reconnect reliably when underlying vconn reports error.
Ben Pfaff [Tue, 29 Jul 2008 00:29:26 +0000 (17:29 -0700)]
rconn: Reconnect reliably when underlying vconn reports error.

When a vconn reports an error, the rconn would not reliably reconnect.
In particular, if the error was reported after the call to rconn_run()
but before rconn_run_wait() was called, then the state's "run" routine
would not set min_timeout properly, leading to a potentially arbitrarily
long wait (depending on what other events were going on in) until the
state's "run" routine was called again.

The fix is to have a separate per-state "timeout" routine to compute
when the state needs to be re-entered.

This commit was tested using the following change to randomly inject
errors:

@@ -554,11 +554,16 @@
 static int
 try_send(struct rconn *rc)
 {
     int retval = 0;
     struct buffer *next = rc->txq.head->next;
-    retval = vconn_send(rc->vconn, rc->txq.head);
+    if (!random_range(1000)) {
+        fprintf(stderr, "injecting ECONNRESET\n");
+        retval = ECONNRESET;
+    } else {
+        retval = vconn_send(rc->vconn, rc->txq.head);
+    }
     if (retval) {
         if (retval != EAGAIN) {
             disconnect(rc, retval);
         }
         return retval;

16 years agoExport the "dp_mutex" symbol, since it's needed by the hardware tables.
Justin Pettit [Mon, 28 Jul 2008 23:08:37 +0000 (16:08 -0700)]
Export the "dp_mutex" symbol, since it's needed by the hardware tables.

16 years agoDrop unnecessary conditional and incorrect comment.
Ben Pfaff [Mon, 28 Jul 2008 22:07:24 +0000 (15:07 -0700)]
Drop unnecessary conditional and incorrect comment.

'controller_relay' is always non-null (but in an earlier revision of
the code that was never pushed to the repository, this was not the case.)

16 years agoBe slightly pickier about going into fail-open mode.
Ben Pfaff [Mon, 28 Jul 2008 14:08:38 +0000 (07:08 -0700)]
Be slightly pickier about going into fail-open mode.

Before, the amount of time disconnected was measured relative to the
last time we connected, that is, the last time connect(2) succeeded.
Thus, if we were connected for a long time, and then disconnected,
we would immediately go into fail-open mode.

This change make the disconnected time relative to the last OpenFlow
message received.  Thus, if we are connected for a long time and
receive plenty of packets, and then disconnect, there will still be
an opportunity to reconnect before failing open.

16 years agoAvert GCC false-positive warning.
Ben Pfaff [Mon, 28 Jul 2008 17:11:51 +0000 (10:11 -0700)]
Avert GCC false-positive warning.

16 years agoMerge branch 'locking'
Ben Pfaff [Thu, 24 Jul 2008 23:07:32 +0000 (16:07 -0700)]
Merge branch 'locking'

16 years agoDon't invoke Debconf from openflow-switch postinst.
Ben Pfaff [Thu, 24 Jul 2008 00:21:58 +0000 (17:21 -0700)]
Don't invoke Debconf from openflow-switch postinst.

This was in there because at one time it seemed desirable to configure
from the postinst, but that idea was abandoned.  Now Debconf seems to
be screwing something up somehow, so we might as well not run it at all,
since it wasn't doing anything useful.

16 years agovlog: Reduce syslog level of VLL_EMER messages to LOG_ALERT.
Ben Pfaff [Thu, 24 Jul 2008 00:04:05 +0000 (17:04 -0700)]
vlog: Reduce syslog level of VLL_EMER messages to LOG_ALERT.

The LOG_EMERG log level sprayed these messages across every xterm and
console in the system, which was excessive.

16 years agovlog: Send multi-line log messages to syslog() one line at a time.
Ben Pfaff [Thu, 24 Jul 2008 00:04:21 +0000 (17:04 -0700)]
vlog: Send multi-line log messages to syslog() one line at a time.

syslogd swallows new-lines, which makes multi-line messages otherwise
difficult to read.

16 years agoVerify OpenFlow version number in vconn_recv().
Ben Pfaff [Thu, 24 Jul 2008 00:03:03 +0000 (17:03 -0700)]
Verify OpenFlow version number in vconn_recv().

16 years agoFix kdist_clean target in debian/rules.
Ben Pfaff [Wed, 23 Jul 2008 23:42:05 +0000 (16:42 -0700)]
Fix kdist_clean target in debian/rules.

This fixes "module-assistant auto-install openflow-datapath" for
bootstrapping from a clean environment.  (Use "m-a -f purge
openflow-datapath" to test that it works starting from an un-clean
environment.)

16 years agoUpdate Debconf templates PO file.
Ben Pfaff [Wed, 23 Jul 2008 21:57:22 +0000 (14:57 -0700)]
Update Debconf templates PO file.

16 years agoDon't log messages to the console by default in Debian package.
Ben Pfaff [Wed, 23 Jul 2008 21:47:37 +0000 (14:47 -0700)]
Don't log messages to the console by default in Debian package.

16 years agoBe less picky about precise name of process in init script.
Ben Pfaff [Wed, 23 Jul 2008 21:54:19 +0000 (14:54 -0700)]
Be less picky about precise name of process in init script.

The previous version always printed ERROR because for whatever reason
we were comparing "/usr/sbin/secchan" against "secchan", which of
course failed.

16 years agoOnly generate private key and certificate request if ssl enabled.
Ben Pfaff [Wed, 23 Jul 2008 21:53:06 +0000 (14:53 -0700)]
Only generate private key and certificate request if ssl enabled.

Generating them unconditionally caused a problem with the init script:
the script required there to be a certificate for the private key if
the private key existed, but we do want to allow TCP-only configurations
and always having a private key prevented that.

Also, give the user advice on how to deal with this problem when it
arises in the init script.

16 years agoEnable secchan, ofp-discover to update /etc/resolv.conf.
Ben Pfaff [Wed, 23 Jul 2008 21:30:59 +0000 (14:30 -0700)]
Enable secchan, ofp-discover to update /etc/resolv.conf.

This way it becomes possible to more reliably refer to the controller
and the PKI server using hostnames.

16 years agoSupport controller discovery in Debian packages.
Ben Pfaff [Wed, 23 Jul 2008 20:11:00 +0000 (13:11 -0700)]
Support controller discovery in Debian packages.

16 years agoNew utility ofp-kill.
Ben Pfaff [Wed, 23 Jul 2008 20:09:25 +0000 (13:09 -0700)]
New utility ofp-kill.

Needed for controller discovery in upcoming revision of ofp-switch-setup.

16 years agoNew function make_pidfile_name().
Ben Pfaff [Wed, 23 Jul 2008 20:08:31 +0000 (13:08 -0700)]
New function make_pidfile_name().

The upcoming ofp-kill utility wants to use this.

16 years agosecchan: Improve logging when rejecting a controller vconn.
Ben Pfaff [Wed, 23 Jul 2008 20:07:40 +0000 (13:07 -0700)]
secchan: Improve logging when rejecting a controller vconn.

16 years agoLock pidfiles with fcntl and create them atomically.
Ben Pfaff [Wed, 23 Jul 2008 20:07:02 +0000 (13:07 -0700)]
Lock pidfiles with fcntl and create them atomically.

This makes it possible to verify that the program that created the
pidfile is still running.

16 years agoNew program ofp-discover.
Ben Pfaff [Tue, 22 Jul 2008 21:44:10 +0000 (14:44 -0700)]
New program ofp-discover.

16 years agoAdd SIGALRM to blockable fatal signals.
Ben Pfaff [Tue, 22 Jul 2008 20:26:03 +0000 (13:26 -0700)]
Add SIGALRM to blockable fatal signals.

We use SIGALRM to limit runtime of dpctl and ofp-discover.  There is no
reason that we should not clean up after it in the same way as any other
catchable signal.  In particular, ofp-discover wants to restore network
device flags on timeout.

16 years agodhcp: Always build test-dhcp-client.
Ben Pfaff [Tue, 22 Jul 2008 21:01:25 +0000 (14:01 -0700)]
dhcp: Always build test-dhcp-client.

Putting test-dhcp-client in check_PROGRAMS instead of noinst_PROGRAMS
makes it too easy to break test-dhcp-client without noticing it.

16 years agodhcp: Fix race condition in test-dhcp-client.
Ben Pfaff [Tue, 22 Jul 2008 21:01:17 +0000 (14:01 -0700)]
dhcp: Fix race condition in test-dhcp-client.

dhclient_wait() might also access dhcp state.

16 years agodhcp: Only print "entering" message if state actually changes.
Ben Pfaff [Tue, 22 Jul 2008 20:32:42 +0000 (13:32 -0700)]
dhcp: Only print "entering" message if state actually changes.

16 years agodhcp: Don't attempt to drain netdev if we ran out of packets.
Ben Pfaff [Tue, 22 Jul 2008 20:16:08 +0000 (13:16 -0700)]
dhcp: Don't attempt to drain netdev if we ran out of packets.

Fixes a theoretical race condition that would have caused more
retransmission than otherwise necessary.

16 years agodhcp: Make output of dhcp_option_to_string() more uniform, to ease parsing.
Ben Pfaff [Wed, 23 Jul 2008 20:12:14 +0000 (13:12 -0700)]
dhcp: Make output of dhcp_option_to_string() more uniform, to ease parsing.

(The ofp-switch-setup program will want to parse its output.)

16 years agodhcp: Make dhcp_msg_to_string() support a multiline format also.
Ben Pfaff [Tue, 22 Jul 2008 21:01:10 +0000 (14:01 -0700)]
dhcp: Make dhcp_msg_to_string() support a multiline format also.

The upcoming ofp-discover program wants to print out the binding
information, and dhcp_msg_to_string() is pretty close to what it
wants.  But one-field-per-line is easier for other programs to parse.

16 years agodhcp: Break out netdev configuration from DHCP binding.
Ben Pfaff [Tue, 22 Jul 2008 21:01:01 +0000 (14:01 -0700)]
dhcp: Break out netdev configuration from DHCP binding.

This allows us to have DHCP clients that don't actually bind or unbind
a network device's IP address.  This is useful for doing controller
discovery without actually reconfiguring a network device.

16 years agodhcp: Be more precise about updating 'changed' flag.
Ben Pfaff [Tue, 22 Jul 2008 21:00:03 +0000 (14:00 -0700)]
dhcp: Be more precise about updating 'changed' flag.

There's no need to set changed unconditionally in dhclient_init(), and
it can make extra work for the users of the dhclient, by causing them
to think that something changed when it didn't.

16 years agoMake set_pidfile() easier to use.
Ben Pfaff [Tue, 22 Jul 2008 20:53:20 +0000 (13:53 -0700)]
Make set_pidfile() easier to use.

16 years agodpctl: Get rid of effectively unused variable.
Ben Pfaff [Tue, 22 Jul 2008 20:33:02 +0000 (13:33 -0700)]
dpctl: Get rid of effectively unused variable.

16 years agoDon't include unneeded header file.
Ben Pfaff [Mon, 21 Jul 2008 23:30:53 +0000 (16:30 -0700)]
Don't include unneeded header file.

16 years agoDo the Debian package build in a _debian subdirectory.
Ben Pfaff [Mon, 21 Jul 2008 22:44:18 +0000 (15:44 -0700)]
Do the Debian package build in a _debian subdirectory.

This makes it easier to do a Debian build in the same source tree
also being used for other development.

16 years agoDon't unnecessarily link against -lresolv and -ldl.
Ben Pfaff [Mon, 21 Jul 2008 22:42:31 +0000 (15:42 -0700)]
Don't unnecessarily link against -lresolv and -ldl.

Eliminates warnings given by dpkg-shlibdeps when building Debian
packages.

16 years agoDistribute new rculist.h headers.
Ben Pfaff [Mon, 21 Jul 2008 22:41:38 +0000 (15:41 -0700)]
Distribute new rculist.h headers.

16 years agoMove Autoconf's macro definitions into config.h.
Ben Pfaff [Mon, 21 Jul 2008 22:04:54 +0000 (15:04 -0700)]
Move Autoconf's macro definitions into config.h.

This makes "make" output slightly more readable.

16 years agoDrop use of libtool for libraries.
Ben Pfaff [Mon, 21 Jul 2008 21:48:53 +0000 (14:48 -0700)]
Drop use of libtool for libraries.

libtool was needed when NOX was directly linking against libopenflow.
Since it doesn't do that anymore, we can drop its use (and that makes
OpenFlow configure and build slightly faster).

16 years agoDon't explicitly link against libdl.
Ben Pfaff [Mon, 21 Jul 2008 21:50:36 +0000 (14:50 -0700)]
Don't explicitly link against libdl.

There is no need to add -ldl explicitly to each program's LDADD, because
the configure test for dladdr does that for us.

16 years agoOnly link programs that need it against OpenSSL.
Ben Pfaff [Mon, 21 Jul 2008 21:37:21 +0000 (14:37 -0700)]
Only link programs that need it against OpenSSL.

This keeps vlogconf, in particular, from being linked against OpenSSL,
which in turn keeps the Debian package build tools from complaining that
the openflow-common package shouldn't need to depend on OpenSSL but does
because of the spurious library link.

16 years agoGet rid of now-excessive levels of indirection in hwtable_dummy.
Ben Pfaff [Mon, 21 Jul 2008 21:06:36 +0000 (14:06 -0700)]
Get rid of now-excessive levels of indirection in hwtable_dummy.

16 years agoVerify that sleeping is allowed in chain contexts where Broadcom will need to.
Ben Pfaff [Mon, 21 Jul 2008 20:39:26 +0000 (13:39 -0700)]
Verify that sleeping is allowed in chain contexts where Broadcom will need to.

16 years agoDon't need 'deleted' member of struct flow.
Ben Pfaff [Mon, 21 Jul 2008 20:51:54 +0000 (13:51 -0700)]
Don't need 'deleted' member of struct flow.

Flow deletion is already fully serialized on dp_mutex.

16 years agoDon't use atomic_t for tracking table flow counts.
Ben Pfaff [Mon, 21 Jul 2008 20:51:02 +0000 (13:51 -0700)]
Don't use atomic_t for tracking table flow counts.

Changes to n_flows is already fully serialized by dp_mutex.

16 years agoGet rid of unnecessary synchronization in tables.
Ben Pfaff [Mon, 21 Jul 2008 20:54:01 +0000 (13:54 -0700)]
Get rid of unnecessary synchronization in tables.

The table insert and delete functions are already fully serialized
on dp_mutex (and genl_mutex), so there's no need for them to use
additional spinlocks.

16 years agoProperly synchronize dp_dev destruction.
Ben Pfaff [Mon, 21 Jul 2008 18:22:28 +0000 (11:22 -0700)]
Properly synchronize dp_dev destruction.

Before, we didn't really synchronize this at all.  Now, when we want
to destroy the ofX device, we stop the transmit path and wait for
in-progress transmissions to finish.

16 years agoSimplify use of dp_mutex.
Ben Pfaff [Mon, 21 Jul 2008 20:59:10 +0000 (13:59 -0700)]
Simplify use of dp_mutex.

There was little point in taking the dp_mutex farther down in the
code than dp_genl_openflow, since that function is already completely
serialized by genl_rcv across the genl_mutex.  We could get rid of
dp_mutex completely, except that we still need it to serialize timeout.

16 years agoGet rid of useless use of rcu_read_lock.
Ben Pfaff [Mon, 21 Jul 2008 20:26:37 +0000 (13:26 -0700)]
Get rid of useless use of rcu_read_lock.

These code paths are fully protected by the Generic Netlink genl_mutex,
which is taken by genl_rcv() that calls dp_genl_openflow(), so there is
no need for them to take the rcu_read_lock.

16 years agoMake openflow.h SWIG-friendly.
Ben Pfaff [Mon, 21 Jul 2008 18:39:39 +0000 (11:39 -0700)]
Make openflow.h SWIG-friendly.

16 years agoFix build on Linux 2.6.26.
Ben Pfaff [Sat, 19 Jul 2008 23:14:18 +0000 (16:14 -0700)]
Fix build on Linux 2.6.26.

Linux 2.6.26 requires us to additionally include <linux/rculist.h>
to use RCU lists.

16 years agoUse an IP ID of 0 for transmitted DHCP packets.
Ben Pfaff [Thu, 17 Jul 2008 18:58:15 +0000 (11:58 -0700)]
Use an IP ID of 0 for transmitted DHCP packets.

We can't guarantee uniqueness of ip_id versus the host's, screwing up
fragment reassembly, so prevent fragmentation and use an all-zeros
ip_id.  RFC 791 doesn't say we can do this, but Linux does the same
thing for DF packets, so it must not screw anything up.

16 years agoDiscovery.
Ben Pfaff [Tue, 15 Jul 2008 18:20:05 +0000 (11:20 -0700)]
Discovery.

16 years agoDon't truncate flooded packets at the amount sent up by the switch.
Ben Pfaff [Mon, 14 Jul 2008 22:27:19 +0000 (15:27 -0700)]
Don't truncate flooded packets at the amount sent up by the switch.

16 years agoMake facility and level optional in -v, --verbose options.
Ben Pfaff [Mon, 14 Jul 2008 21:10:18 +0000 (14:10 -0700)]
Make facility and level optional in -v, --verbose options.

This makes it easier to set a given module to maximum verbosity,
as is often useful for debugging.