Ben Pfaff [Mon, 18 Oct 2010 23:15:50 +0000 (16:15 -0700)]
ofproto: Simplify send_flow_removed().
I have no evidence that the optimization in this function is valuable.
An upcoming commit will introduce a new form of flow expiration message
that is sent to controllers that ask for it, while the standard OpenFlow
1.0 message is sent to other controllers. Since retaining this
optimization with that logic would complicate the function, this commit
drops it.
Jesse Gross [Thu, 4 Nov 2010 23:20:32 +0000 (16:20 -0700)]
datapath: Directly use free_netdev for internal devices.
Currently internal devices register a destructor function which
simply calls free_netdev. Instead we can simply set the destructor
to free_netdev. In addition to being cleaner, it is also a bug fix
because the module could be unloaded before the destructor is called,
making a call into our code illegal.
Signed-off-by: Jesse Gross <jesse@nicira.com> Acked-by: Ben Pfaff <blp@nicira.com>
Ben Pfaff [Thu, 4 Nov 2010 17:53:57 +0000 (10:53 -0700)]
ovs-vsctl: Create bridge interface as type "internal".
The bridge device is always an internal interface, so it seems best to
actually mark it as one. Jesse would like to eventually make the "type"
column mandatory for Interface records, so here's a step in the right
direction.
Ben Pfaff [Tue, 2 Nov 2010 21:50:48 +0000 (14:50 -0700)]
ovsdb-client: Refactor table code to save original json during formatting.
This refactoring should not change user-visible behavior, but saving the
JSON used to format tables will make it possible to print the raw JSON in
the following commit.
Ben Pfaff [Wed, 3 Nov 2010 18:07:16 +0000 (11:07 -0700)]
ofproto: Don't show separate wildcard and exact-match tables via OpenFlow.
Formerly the classifier used in Open vSwitch was very sensitive to the
number of exact-match versus wildcarded rules, so it made sense to show
them as separate tables over OpenFlow. The current classifier doesn't
have the same performance sensitivity, so we might as well just show one
table.
Ben Pfaff [Wed, 3 Nov 2010 18:00:58 +0000 (11:00 -0700)]
classifier: Rewrite.
The old classifier was not adaptive: it required knowing the structure of
the flows that were likely to be in use to get good performance. It is
likely that it degenerated to linear search in any real-world case.
This new classifier is adaptive and should perform better in the real
world.
Jesse Gross [Wed, 3 Nov 2010 17:14:06 +0000 (10:14 -0700)]
autoconf: Compat checks no longer support multiple files.
We now test that a file exists before grepping but that doesn't
work if there are multiple possible files. For the one case that
needs this (kmemdup) break the test into two: check the first file
and if the pattern isn't found check the second file.
Jesse Gross [Tue, 2 Nov 2010 22:43:32 +0000 (15:43 -0700)]
datapath: Work around debugging for csum_and_copy_to_user().
Certain versions of XenServer add debugging to csum_and_copy_to_user()
in such a way that it changes the function signature. This adds a
check and a workaround to allow us to build on these versions.
Signed-off-by: Jesse Gross <jesse@nicira.com> Acked-by: Ben Pfaff <blp@nicira.com>
Jesse Gross [Tue, 2 Nov 2010 23:00:16 +0000 (16:00 -0700)]
autoconf: Tolerate missing file when grepping.
Currently we die when grepping for compatibility strings if the
file does not exist. Since this can be a valid situation when
files are added in later versions, we shouldn't kill the build.
Instead, note that the file doesn't exist but otherwise treat it
as if the string was not found.
Ben Pfaff [Tue, 2 Nov 2010 20:13:12 +0000 (13:13 -0700)]
ovsdb: Fix formatting of ovs.db.Error on Python 2.6.
All of the negative Python OVSDB tests were failing on Python 2.6 because
"%s\n" % e yielded the empty string on that version of Python. In turn,
that was because ovs.db.error.Error.__unicode__ was being called instead of
ovs.db.error.Error.__str__. I'm puzzled why that was happening, but this
commit fixes it and also seems like a small code cleanup.
Peter Balland helped me gain some insight on this problem.
CC: Peter Balland <peter@nicira.com> CC: Reid Price <reid@nicira.com>
Simon Horman [Tue, 2 Nov 2010 12:35:14 +0000 (21:35 +0900)]
datapath: The definition of struct tbl is needed by rcu_dereference()
Due to the use of typeof(*p) by __rcu_dereference_check(),
which was introduced by the Linux changeset ca5ecddfa8fcbd948c95530e7e817cee9fb43a3d
(rcu: define __rcu address space modifier for sparse) the definition
of the type of a the parameter to rcu_dereference() is needed.
In terms of the datapath, this means that the definition of
struct tbl needs to be available wherever rcu_dereference(dp->table)
is called.
Signed-off-by: Simon Horman <horms@verge.net.au> Signed-off-by: Jesse Gross <jesse@nicira.com>
Ben Pfaff [Thu, 14 Oct 2010 17:30:07 +0000 (10:30 -0700)]
classifier: Remove classifier_insert_exact().
This function doesn't provide any extra useful functionality. It is
amenable to a slightly optimized implementation in the current classifier,
but not in the one that will soon replace it, so get rid of it.
Ben Pfaff [Thu, 28 Oct 2010 04:17:27 +0000 (21:17 -0700)]
ofproto: Use ovs_be<N> types in appropriate places.
There are many more places in OVS where using these types would be an
improvement, but the ofproto code is extra confusing because it uses
a mix of byte orders.
Ben Pfaff [Tue, 26 Oct 2010 22:24:26 +0000 (15:24 -0700)]
flow: Use ovs_be<N> types in appropriate places.
There are many more places in OVS where using these types would be an
improvement, but the flow code is particularly confusing because it uses
a mix of byte orders.
Jesse Gross [Wed, 27 Oct 2010 22:29:16 +0000 (15:29 -0700)]
netdev: Fix carrier status for down interfaces.
Currently netdev_get_carrier() returns both a carrier status and
an error code. However, usage of the error code was inconsistent:
most callers either ignored it or didn't perform their task if an
error occured, which prevented bond rebalancing. This makes the
handling consistent by translating an error into a down status in
the netdev library.
Ben Pfaff [Wed, 27 Oct 2010 16:29:08 +0000 (09:29 -0700)]
daemon: Don't call a normal exit from the monitor a "crash".
When the monitored child is killed with SIGTERM, the monitoring process
currently logs a message like "1 crashes: pid 12345 died, killed by
signal 15 (Terminated), exiting". This counts the SIGTERM as a crash, even
though it's intentional.
This commit changes the log message to omit the "%d crashes" part on normal
termination.
Justin Pettit [Tue, 26 Oct 2010 21:54:13 +0000 (14:54 -0700)]
ofproto: Add coverage counter for reinit_ports()
The reinit_ports() function is an interesting enough case that it would
be good to know if we're hitting it often. This adds a
"ofproto_reinit_ports" coverage counter.
Justin Pettit [Tue, 26 Oct 2010 00:26:44 +0000 (17:26 -0700)]
ovs-external-ids: Better handle VIFs from recently resumed VMs
XAPI doesn't provide a way to look up a VIF entry based on the name, so
we have to locate it by other methods. Previously, we were breaking up
the name into the domid and device number. Unfortunately, it can take
XAPI a few seconds to update the domid of the VM, when resuming from a
suspend. Since we have the VIF UUID, we can just look up the needed
information directly based on that.
Jesse Gross [Mon, 18 Oct 2010 22:30:20 +0000 (15:30 -0700)]
datapath: Add loop detection for RT kernels.
Our normal loop detection requires disabling preemption while
packet processing takes place. On RT kernels this isn't acceptable
and interacts badly with spinlocks, so we can't use it. This
takes advantage of some extra space that is added to struct
task_struct on RT kernels (and the knowledge that we will always
have a valid task_struct) to store the loop counter for a given
thread. Since we can't make these assumptions on non-RT kernels,
we continue to use the previous method of loop detection there.
Signed-off-by: Jesse Gross <jesse@nicira.com> Acked-by: Ben Pfaff <blp@nicira.com>
Jesse Gross [Mon, 18 Oct 2010 22:36:59 +0000 (15:36 -0700)]
datapath: #define local_bh_disable to preempt_disable on RT.
We use local_bh_disable() to mean that we don't want to get
preempted or interrupted and normally it covers both situations.
However, on RT kernels local_bh_disable() becomes a no-op because
bottom halves aren't interrupts any more. We still want the
behavior of disabling preemption so #define them to be the same
on these kernels.
Signed-off-by: Jesse Gross <jesse@nicira.com> Acked-by: Ben Pfaff <blp@nicira.com>
Ben Pfaff [Fri, 22 Oct 2010 20:42:40 +0000 (13:42 -0700)]
vswitchd: Reset ofport columns of invalid interfaces to -1.
Until now, when ovs-vswitchd dropped interfaces because of duplicate names,
it would not write -1 to the ofport column in the OVSDB record for the
interface, which made it appear that the interface was still active (if
it had been before). This commit fixes the problem.
This commit also sets ofport to -1 for interfaces that cannot be created
for other reasons.
Ben Pfaff [Fri, 22 Oct 2010 21:46:30 +0000 (14:46 -0700)]
ovs-vsctl: Check for dirty reads within transactions.
OVSDB is transactional but it does not implement any form of locking. This
means that read-modify-write operations must verify that the values that
they read are still in place before writing. This commit adds such
checking.
Ben Pfaff [Fri, 22 Oct 2010 17:19:13 +0000 (10:19 -0700)]
vswitch: Make unique "name" columns immutable, to simplify transactions.
OVSDB has transactions but not locking, so a transaction that must
maintain an invariant must check that the starting state is what it
expects. For example, to add a bridge a client must verify that the new
bridge's name does not conflict with any existing bridge's name, given
that the set of bridges might have changed. One way to do that is for
the client to verify that that the set of bridges is the same and that none
of the bridges has been renamed to the new bridge's name. By making
bridge names immutable, the latter part of the check can be omitted.
Mirror names are not required to be unique so this commit does not make
them immutable.
Ethan Jackson [Thu, 21 Oct 2010 22:28:36 +0000 (22:28 +0000)]
netdev-linux: Make queue 0 the default QOS policy
This patch defines, by convention, queue 0 as the default queue in
a particular QOS. Thus, if queue 0 is defined, all traffic going
through the relevant interface will be enqueued in it. If queue 0
is not defined then ovs will send the traffic directly through the
interface without applying any policy to it.
Ben Pfaff [Thu, 21 Oct 2010 18:28:00 +0000 (11:28 -0700)]
tests: Update expected test result to match "ovsdb-client dump" changes.
Commit 06036898849 "ovsdb-client: Make "dump" print table names in default
output format too" changed the output format of "ovsdb-client dump" without
updating the tests that depended on that output format. This commit
updates the expected test results to match.
Ben Pfaff [Thu, 21 Oct 2010 17:48:56 +0000 (10:48 -0700)]
ovsdb-client: Make "dump" print table names in default output format too.
The "dump" command printed table names in CSV and HTML output formats but
they were omitted in the default tabular form. This commit corrects the
discrepancy.
Justin Pettit [Tue, 19 Oct 2010 22:58:35 +0000 (15:58 -0700)]
ovs-vswitchd: Add ability to disable in-band on a bridge
In-band control is always enabled when a manager is configured. Some
applications wish to control all the flows within a bridge, and in-band
control can override that wish. Depending on how the network is
configured, this can lead to loops as the in-band control rules try to
learn where to send traffic.
This commit adds a "disable-in-band" key to the "other_config" column of
bridge tables. If set to "true", in-band will be disabled regardless of
manager or controller configuration.
Justin Pettit [Mon, 18 Oct 2010 05:43:14 +0000 (22:43 -0700)]
ovs-monitor-ipsec: Add ability to traverse NATs
Stable versions of ipsec-tools have a bug that prevents our using
transport mode through a NAT box. Even though the bug has been fixed
for years, it is only available in the 0.8 alpha release of ipsec-tools.
This commit modifies our configuration to allow NAT traversal with that
version of ipsec-tools.
NB: An official package for this version of ipsec-tools is not yet
available on Debian, so we're requiring a custom version number.
Sajjad Lateef [Fri, 15 Oct 2010 20:41:46 +0000 (13:41 -0700)]
Create /etc/openvswitch dir when package is installed.
Earlier, the /etc/openvswitch directory would get installed as part of
the init script's start sequence. However, as newer packages are added,
including some that create files in this directory during their init
script sequences, a race condition may occur.
This change ensures that the directory /etc/openvswitch is created even
before the init scripts execute. Thus, the race condition is avoided.
Ethan Jackson [Wed, 13 Oct 2010 22:12:12 +0000 (22:12 +0000)]
vswitchd: Bubble no-flood configuration up to bridge
When bridge.c decides to flood a packet as the result of a "normal"
flow action, it now checks whether each port is configured to
receive flood packets.
Ben Pfaff [Thu, 14 Oct 2010 00:11:56 +0000 (17:11 -0700)]
ofproto: Fix implementation of OFPP_FLOOD.
Commit f1588b1fa1 "datapath: Remove implementation of port groups" removed
a "break" statement accidentally, which caused OFPP_FLOOD to send duplicate
packets and to ignore OFPPC_NO_FLOOD.
Justin Pettit [Wed, 13 Oct 2010 07:14:31 +0000 (00:14 -0700)]
netdev-vport: Actually destroy the vport when "destroy" is called
Commit 2b9d65 (netdev-vport: Merge in netdev-patch and netdev-tunnel.)
refactored the common parts of the netdev-patch and netdev-tunnel
sources into netdev-vport. During the refactoring, the "destroy" method
didn't inherit the netdev_vport_do_ioctl(ODP_VPORT_DEL, ...) call, which
is needed to actually destroy the device in the kernel. This commit
fixes that.
Justin Pettit [Tue, 12 Oct 2010 18:24:04 +0000 (11:24 -0700)]
netflow: Back-out optimization that could lead to infinite loop
Commit 924282 (netflow: Do 64-bit division less often.) attempted to
remove the 64-bit division used to break flow records with large byte
counts into multiple NetFlow records. The calculation to determine the
number of records was incorrect and should have shifted "byte_delta" by
31 instead of 32. This commit reverts the change (while keeping commit
f22a24 (netflow: Avoid (theoretically) looping 2**32 times.) ), since
the logic is more straight-forward than the optimized version.
Justin Pettit [Tue, 12 Oct 2010 07:01:37 +0000 (00:01 -0700)]
ofproto: Update flow context for resubmit on VLAN actions
Any modifications that happen to the packet are supposed to be reflected
in the flow context for a resubmitted lookup. Commit 26233b (datapath:
Combine dl_vlan and dl_vlan_pcp.) broke this for OFPAT_SET_VLAN_VID and
OFPAT_SET_VLAN_PCP. This corrects that behavior.
Justin Pettit [Tue, 12 Oct 2010 06:50:36 +0000 (23:50 -0700)]
odp-util: Remove extraneous "set_tci(" from format_odp_action()
The function format_odp_action(), which is used to format the output of
"ovs-dpctl dump-flows", printed "set_tci(" before printing a correct
"set_tci" with actual parameters.
Ben Pfaff [Mon, 11 Oct 2010 21:05:37 +0000 (14:05 -0700)]
datapath: Fix build break in compat_query_flows().
Commit a01ef04ce "datapath: Drop padding from struct odp_flow_key" removed
the "reserved" member from struct odp_flow_key but overlooked uses of that
member from 64-bit compatibility code. This commit fixes up the problem.
Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
Ben Pfaff [Mon, 11 Oct 2010 21:04:10 +0000 (14:04 -0700)]
datapath: Fix typo in 64-bit compat header.
Commit f1588b1f "datapath: Remove implementation of port groups" added a
spurious "V" in the odp-compat.h header. I failed to test-build that
commit on a 64-bit system and thus this broke the build on x86-64 and other
64-bit systems.
Signed-off-by: Ben Pfaff <blp@nicira.com> Acked-by: Jesse Gross <jesse@nicira.com>
Hao Zheng [Fri, 8 Oct 2010 23:26:15 +0000 (16:26 -0700)]
datapath: fix flow used time stats
This commit fixes a problem where datapath flow used stats can report
incorrent value, thus may result in flow expirations being incorrect.
This happens when a kernel jiffies rollover occurs between the last time
a flow is hit and the flow stats is queried.
Ben Pfaff [Fri, 10 Sep 2010 18:16:31 +0000 (11:16 -0700)]
datapath: Always use GFP_ATOMIC to execute actions.
These functions run 99% of the time in atomic context and the benefit of
passing along the 'gfp' argument for the other 1% doesn't seem to outweigh
the cost.
Suggested-by: Stephen Hemminger <shemminger@vyatta.com> Acked-by: Jesse Gross <jesse@nicira.com> Signed-off-by: Ben Pfaff <blp@nicira.com>
Ben Pfaff [Fri, 3 Sep 2010 18:30:02 +0000 (11:30 -0700)]
flow: Get rid of flow_t typedef.
When userspace and the kernel were using the same structure for flows,
flow_t was a useful way to indicate that a structure was really a userspace
flow instead of a kernel one, but now it's better to just write "struct
flow" for consistency, since OVS doesn't use typedefs for structs
elsewhere.
Ben Pfaff [Mon, 11 Oct 2010 20:31:35 +0000 (13:31 -0700)]
flow: Separate "flow_t" from "struct odp_flow_key".
The "struct odp_flow_key" used in the kernel datapath is conceptually
separate from the "flow_t" used in userspace, but until now we have
used the latter as a typedef for the former for convenience. This commit
separates them. This makes it possible in upcoming commits to change
them independently.
This is cross-ported from the "wdp" branch, which has had it for months.
Ben Pfaff [Fri, 8 Oct 2010 23:36:13 +0000 (16:36 -0700)]
datapath: Remove implementation of port groups.
The "port group" concept seems like a good one, but it has not been
used very much in userspace so far, so before we commit ourselves to
a frozen API that we must maintain forever, remove it. We can always
add it back in later as a new kind of vport.
Ben Pfaff [Wed, 6 Oct 2010 22:35:53 +0000 (15:35 -0700)]
netdev-vport: Merge in netdev-patch and netdev-tunnel.
The only real difference between netdev-patch and netdev-tunnel is in their
parse_config() implementation. That's a lot of extra code to maintain, for
questionable benefit. This commit merges them into the netdev-vport code,
which was heretofore merely a collection of helper functions.
Ben Pfaff [Fri, 1 Oct 2010 22:59:56 +0000 (15:59 -0700)]
ofproto: Get rid of port_changed_cb.
Jesse pointed out that port_changed_cb isn't a great interface. It's only
around because, earlier, we had a lousy interface for monitoring netdev
status, so that we needed to pass along information obtained by ofproto
into the bridge. But netdev_monitor is now sufficiently sophisticated that
the bridge can set up an independent netdev_monitor without any important
loss of efficiency. Since this makes the code cleaner, this commit does
so.
Ben Pfaff [Tue, 28 Sep 2010 18:57:40 +0000 (11:57 -0700)]
vswitchd: Better tolerate changes in datapath ports.
Until now, a command that removed and added ports in a single change to
the database, e.g.:
ovs-vsctl del-port br0 vif1.0 -- add-port br0 vif2.0
typically failed, because of this sequence of events:
1. Bridge code removes vif1.0 from br0.
2. Bridge code adds vif2.0 to br0.
3. ofproto_run() receives kernel notification that vif1.0 was deleted, so
it notifies the bridge by calling back to bridge_port_changed_ofhook_cb,
which sees that it has an interface with the specified port number, and
deletes it. Oops--this is where the problem occurs. For completeness:
4. ofproto_run() receives kernel notification that vif2.0 was added, so
it notifies the bridge by calling back to ,
which sees that it has no interface with the specified port number, and
does nothing.
This commit fixes the problem by making bridge_port_changed_ofhook_cb() not
care about ports being dropped. This is a corner case that we shouldn't
work too hard to care about, since it can only happen if an administrator
is meddling with datapaths using ovs-dpctl, and the consequences are simply
that packets directed to that device will take longer to be rerouted to
another device (it will take a while for the MAC learning table to time out
the entry). Basically, the admin gets what he deserves.
Thanks to Jesse Gross for identifying the problem.