Setting tag sliver-openvswitch-2.1.90-2
fix packaging

mentioning files under /usr/lib/ rather than the dir itself as this conflicts with the 'filesystem' rpm

looks like the package now produces files in /usr/lib

Setting tag sliver-openvswitch-2.1.90-1
merged in mainstream 2.1.90
more robust server startup on the slivers

Merge branch 'mainstream'

netdev-tunnel: use sockaddr_storage

sliver-ovs: ignore pid files when starting servers

ovs-vswitchd and ovs-dbserver already implement their
own logic to make sure that they have only one instance
running. Our solution was inferior, since it was easily
confused by stale pid files.

socket-util: drain_rcvbuf() for Windows.

Netlink sockets are created as blocking sockets. So, we can't
afford to remove MSG_DONTWAIT for Linux.

drain_rcvbuf() is currently called from netlink-socket.c and
netdev-linux.c. As of now, I don't see it being used for Windows.

Bug #1200865.
Reported-by: Len Gao <leng@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

lib/process, socket-util: Update necessary headers

Fix the regression introduced by commit 4f57ad10.
("socket-util: Move get_max_fds() to process.c")

Signed-off-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>

signals: Remove unused functions.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

ovs-vswitchd: Don't register for SIGHUP.

We are registering an interest in SIGHUP to reopen
log files. But there is an 'ovs-appctl vlog/reopen'
command that does the same and is used in the
logrotate config for the distributions.

So remove the redundant functionality.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

ovs-thread: Use fair (but nonrecursive) rwlocks on glibc.

glibc supports two kinds of rwlocks:

    - The default kind of rwlock always allows recursive read-locks to
      succeed, but threads blocked on acquiring the write-lock are treated
      unfairly, causing them to be delayed indefinitely as long as new
      readers continue to come along.

    - An alternative "writer nonrecursive" rwlock allows recursive
      read-locks to succeed only if there are no threads waiting for the
      write-lock.  Otherwise, recursive read-lock attempts deadlock in
      the presence of blocking write-lock attempts.  However, this kind
      of rwlock is fair to writer.

POSIX allows the latter behavior, which essentially means that any portable
pthread program cannot try to take read-locks recursively.  Since that's
true, we might as well use the latter kind of rwlock with glibc and get the
benefit of fairness of writers.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Joe Stringer <joestringer@nicira.com>

ofproto-dpif-xlate: Avoid recursively taking read side of ofgroup rwlock.

With glibc, rwlocks by default allow recursive read-locking even if a
thread is blocked waiting for the write-lock.  POSIX allows such attempts
to deadlock, and it appears that the libc used in NetBSD, at least, does
deadlock.  ofproto-dpif-xlate could take the ofgroup rwlock recursively if
an ofgroup's actions caused the ofgroup to be executed again.  This commit
avoids that issue by preventing recursive translation of groups (the same
group or another group).  This is not the most user friendly solution,
but OpenFlow allows this restriction, and we can always remove the
restriction later (probably requiring more complicated code) if it
proves to be a real problem to real users.

Found by inspection.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Joe Stringer <joestringer@nicira.com>

socket-util: Move get_null_fd() to daemon.c.

get_null_fd() is only called from daemon.c.
It does not need thread safety features anymore as
it is called either through daemonize_start() or
indirectly through daemonize_complete() once.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

socket-util: poll() for Windows.

Also, Windows does not have a MSG_DONTWAIT. Get rid of it
as we always use non-blocking sockets.

Co-authored-by: Linda Sun <lsun@vmware.com>
Signed-off-by: Linda Sun <lsun@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

Replace inet_aton() with inet_pton().

Windows does not have inet_aton(), but does have a inet_pton().
inet_aton() is not defined in POSIX. But inet_pton() is.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

socket-util: Move get_max_fds() to process.c.

get_max_fds() is used only from process.c. Move it there
along with rlim_is_finite(). Since process_start() can only
be called before any additional threads are created, we
no longer need the thread safety checks in get_max_fds().

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

socket-util: Unix socket related calls for non-windows platform.

Don't try to compile Unix socket related functions for Windows.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>

dpif-netdev: Fix memory leak.

In dpif_netdev_flow_del() and dp_netdev_port_input(), the
referenced 'netdev_flow' is not un-referenced.  This causes
the leak of the struct's memory.

This commit fixes the above issue by calling dp_netdev_flow_unref()
after using the reference.

Signed-off-by: Alex Wang <alexw@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

dpif-netdev: Call ovs_refcount_destroy() before free().

This commit makes dp_netdev_flow_unref() and dp_netdev_actions_unref()
invoke the ovs_refcount_destroy() before freeing the corresponding
pointer.

Signed-off-by: Alex Wang <alexw@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

ofproto-dpif: Fix segfault removing port when STP is enabled.

Reported-by: Sridhar Samudrala <samudrala.sridhar@gmail.com>
Tested-by: Sridhar Samudrala <samudrala.sridhar@gmail.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

ovs-thread: Get rid of obsolete sparse wrappers.

These were useful back when we were trying to use the sparse lock balance
annotations, but we removed those in commit 47b52c71232c0 (sparse: Remove
support for thread-safety annotations.) and so they serve no purpose any
longer.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Joe Stringer <joestringer@nicira.com>

FAQ: Describe how to add new OpenFlow messages.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>

socket-util: af_inet_ioctl() for Windows.

There is no direct mapping for the ioctl function in
Windows.  As of now, af_inet_ioctl() is being used for Linux
and BSD. So, don't try to compile it for Windows.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

socket-util: Describe fd for Windows.

In windows there is no clear way to distinguish between a
socket fd and a file fd.

We use the function, describe_fd() mostly for debugging.
For now, return a generic statement.

Co-authored-by: Linda Sun <lsun@vmware.com>
Signed-off-by: Linda Sun <lsun@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

socket-util: getsockopt for Windows.

Windows defines the 'optval' argument as char * instead of void *.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

socket-util: fsync directory for Windows.

There is no corresponding function for Windows.
open() does not work on directories.
There is a function _commit(fd), but that is only meant
for files.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

socket-util: closesocket() for Windows.

For Windows sockets, one has to call closesocket() to
close the sockets.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

socket-util: getaddrinfo return values for Windows.

Couple of return values need changes.
* EAI_NODATA is the same as EAI_NONAME. So we prevent duplicate cases.
* Windows does not have a EAI_SYSTEM.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

socket-util: set_nonblocking for Windows.

Co-authored-by: Linda Sun <lsun@vmware.com>
Signed-off-by: Linda Sun <lsun@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

socket-util: Move sock_errno() to socket-util.

And add more users.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

ofproto: Update only OFPUTIL_PS_LINK_DOWN (not STP) from netdev state.

When a netdev indicates that its state or configuration has changed,
update_port() updates the OpenFlow port to match the changes.  However,
this was being taken too far: a netdev does not have an STP state, and a
state change was resetting the STP state of the port.  This fixes the
problem.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Reported-by: Vasu Dasari <vdasari@gmail.com>
Tested-by: Vasu Dasari <vdasari@gmail.com>

ofp-util: Avoid gcc warning on NetBSD

Avoid the following warning caused by the combination
of NetBSD's htons implementation and gcc bug.
Now --enable-Werror build succeeds on NetBSD-6.

    lib/ofp-util.c: In function 'ofputil_match_from_ofp10_match':
    lib/ofp-util.c:174:15: error: integer overflow in expression

Signed-off-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Signed-off-by: Ben Pfaff <blp@nicira.com>

tests/run-ryu: Correct logfile reporting

$logfile is already prefixed by "$sandbox/" and suffixed by ".log"
so do not duplicate this prefix and suffix combination when appending
$logfile to $logs.

Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Ben Pfaff <blp@nicira.com>

netdev: Change netdev_class_rwlock to recursive mutex, for POSIX safety.

With glibc, rwlocks by default allow recursive read-locking even if a
thread is blocked waiting for the write-lock.  POSIX allows such attempts
to deadlock, and it appears that the libc used in NetBSD, at least, does
deadlock.  The netdev_class_rwlock is in fact acquired recursively in this
way, which is a bug.  This commit fixes the problem by switching to a
recursive mutex.  This allows for less parallelism, but according to an
existing comment that doesn't matter here anyway.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Joe Stringer <joestringer@nicira.com>

ofproto-provider: Update documentation.

This wording was in ofproto.c, but missing from ofproto-provider.h.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

FAQ: Mention hairpinning under actions=in_port.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

datapath: Fix build failure.

Defines RCU_INIT_POINTER() for compatibility.

Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>

ipfix: add SCTP, ICMP, and ICMPv6 flow reporting

Signed-off-by: Romain Lenglet <rlenglet@vmware.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

process: Make changes for Windows.

As of now, we are using the process subsystem in
ovsdb-server to handle the "--run" command line
option. That particular option is not used often
and till deemed necessary, make it unsupported on
Windows platform.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

ovs-vswitchd: Get rid of call to process_init().

It is not needed as we don't use any other process_* calls.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

ipfix: fix upcall cookie size checks to support 8 byte cookies

Commit 96ed775f resizes all userspace metadata to be 8 bytes minimum.
Fix the upcall size checks accordingly.

Signed-off-by: Romain Lenglet <rlenglet@vmware.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

packets: Always set ethertype in push_mpls()

There are two different MPLS ethertypes, 0x8847 and 0x8848 and a push MPLS
action applied to an MPLS packet may cause the ethertype to change from one
to the other. To ensure that this happens update the ethertype in
push_mpls() regardless of if the packet is already MPLS or not.

Test based on a similar test by Joe Stringer.

Cc: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Ben Pfaff <blp@nicira.com>

tests: Add MPLS push on MPLS test.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Co-Authored-By: Simon Horman <horms@verge.net.au>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Ben Pfaff <blp@nicira.com>

ovs-vsctl: reconnect to the database if connection was dropped

If ovs-vsctl has to wait for ovs-vswitchd to reconfigure itself
according to the new database, then sometimes ovs-vsctl could
end up stuck in the event loop if OVSDB connection was dropped
while ovs-vsctl was still running.

This patch fixes this problem by letting ovs-vsctl to reconnect
to the OVSDB, if it has to wait cur_cfg field to be updated.

Issue: 1191997
Reported-by: Spiro Kourtessis <spiro@nicira.com>
Signed-Off-By: Ansis Atteka <aatteka@nicira.com>

configure: Fix bug report email address.

Reported-by: Arun Sharma <arun.sharma@calsoftinc.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Joe Stringer <joestringer@nicira.com>

packet: Make set_ethertype() static

Make set_ethertype() static as it is not used outside of packet.c

Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Ben Pfaff <blp@nicira.com>

ofproto: Remove 'force-miss-model' configuration.

This configuration item was introduced to assist testing of upcall
handling behaviour with and without facets. Facets were removed in
commit e79a6c833e0d7237, so this patch removes the configuration item.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

ofp-actions: Correct pop MPLS ethtype as consistency test

Correct pop MPLS ethtype consistency check to verify that
the packet has an MPLS ethtype before the pop action rather than after:
an MPLS ethtype is a pre-condition but not a post-condition of pop MPLS.

With this change the consistency check in ofpact_check__()
becomes consistent with that in ofpact_from_nxast().

This was found using Ryu tests via the new make check-ryu target.
It allows all of the "POP_MPLS"[1] tests to pass where they previously
failed.

[1] http://osrg.github.io/ryu-certification/switch/ovs

Cc: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Ben Pfaff <blp@nicira.com>

tests/run-ryu: Fix setup for Ryu tests

Tweak our configuration to match with Ryu tests' single-bridge assumption.

Signed-off-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Signed-off-by: Ben Pfaff <blp@nicira.com>
Tested-by: Simon Horman <horms@verge.net.au>

tests/run-ryu: Fix log file names

Signed-off-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Signed-off-by: Ben Pfaff <blp@nicira.com>
Tested-by: Simon Horman <horms@verge.net.au>

ofp-util: Fix a typo in ofputil_decode_ofp11_port

Signed-off-by: kmindg <kmindg@gmail.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

datapath: Use ether_addr_copy

It's slightly smaller/faster for some architectures.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>

OPENFLOW-1.1+: Update status of a few items.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Joe Stringer <joestringer@nicira.com>

OPENFLOW-1.1+: Remove "rework tag order" and MPLS BoS matching from OF1.3.

OpenFlow 1.3.4 is going to revert the tag order change made by OpenFlow
1.3.

MPLS BoS matching is implemented.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Joe Stringer <joestringer@nicira.com>

datapath: Per NUMA node flow stats.

Keep kernel flow stats for each NUMA node rather than each (logical)
CPU.  This avoids using the per-CPU allocator and removes most of the
kernel-side OVS locking overhead otherwise on the top of perf reports
and allows OVS to scale better with higher number of threads.

With 9 handlers and 4 revalidators netperf TCP_CRR test flow setup
rate doubles on a server with two hyper-threaded physical CPUs (16
logical cores each) compared to the current OVS master.  Tested with
non-trivial flow table with a TCP port match rule forcing all new
connections with unique port numbers to OVS userspace.  The IP
addresses are still wildcarded, so the kernel flows are not considered
as exact match 5-tuple flows.  This type of flows can be expected to
appear in large numbers as the result of more effective wildcarding
made possible by improvements in OVS userspace flow classifier.

Perf results for this test (master):

Events: 305K cycles
+   8.43%     ovs-vswitchd  [kernel.kallsyms]   [k] mutex_spin_on_owner
+   5.64%     ovs-vswitchd  [kernel.kallsyms]   [k] __ticket_spin_lock
+   4.75%     ovs-vswitchd  ovs-vswitchd        [.] find_match_wc
+   3.32%     ovs-vswitchd  libpthread-2.15.so  [.] pthread_mutex_lock
+   2.61%     ovs-vswitchd  [kernel.kallsyms]   [k] pcpu_alloc_area
+   2.19%     ovs-vswitchd  ovs-vswitchd        [.] flow_hash_in_minimask_range
+   2.03%          swapper  [kernel.kallsyms]   [k] intel_idle
+   1.84%     ovs-vswitchd  libpthread-2.15.so  [.] pthread_mutex_unlock
+   1.64%     ovs-vswitchd  ovs-vswitchd        [.] classifier_lookup
+   1.58%     ovs-vswitchd  libc-2.15.so        [.] 0x7f4e6
+   1.07%     ovs-vswitchd  [kernel.kallsyms]   [k] memset
+   1.03%          netperf  [kernel.kallsyms]   [k] __ticket_spin_lock
+   0.92%          swapper  [kernel.kallsyms]   [k] __ticket_spin_lock
...

And after this patch:

Events: 356K cycles
+   6.85%     ovs-vswitchd  ovs-vswitchd        [.] find_match_wc
+   4.63%     ovs-vswitchd  libpthread-2.15.so  [.] pthread_mutex_lock
+   3.06%     ovs-vswitchd  [kernel.kallsyms]   [k] __ticket_spin_lock
+   2.81%     ovs-vswitchd  ovs-vswitchd        [.] flow_hash_in_minimask_range
+   2.51%     ovs-vswitchd  libpthread-2.15.so  [.] pthread_mutex_unlock
+   2.27%     ovs-vswitchd  ovs-vswitchd        [.] classifier_lookup
+   1.84%     ovs-vswitchd  libc-2.15.so        [.] 0x15d30f
+   1.74%     ovs-vswitchd  [kernel.kallsyms]   [k] mutex_spin_on_owner
+   1.47%          swapper  [kernel.kallsyms]   [k] intel_idle
+   1.34%     ovs-vswitchd  ovs-vswitchd        [.] flow_hash_in_minimask
+   1.33%     ovs-vswitchd  ovs-vswitchd        [.] rule_actions_unref
+   1.16%     ovs-vswitchd  ovs-vswitchd        [.] hindex_node_with_hash
+   1.16%     ovs-vswitchd  ovs-vswitchd        [.] do_xlate_actions
+   1.09%     ovs-vswitchd  ovs-vswitchd        [.] ofproto_rule_ref
+   1.01%          netperf  [kernel.kallsyms]   [k] __ticket_spin_lock
...

There is a small increase in kernel spinlock overhead due to the same
spinlock being shared between multiple cores of the same physical CPU,
but that is barely visible in the netperf TCP_CRR test performance
(maybe ~1% performance drop, hard to tell exactly due to variance in
the test results), when testing for kernel module throughput (with no
userspace activity, handful of kernel flows).

On flow setup, a single stats instance is allocated (for the NUMA node
0).  As CPUs from multiple NUMA nodes start updating stats, new
NUMA-node specific stats instances are allocated.  This allocation on
the packet processing code path is made to never block or look for
emergency memory pools, minimizing the allocation latency.  If the
allocation fails, the existing preallocated stats instance is used.
Also, if only CPUs from one NUMA-node are updating the preallocated
stats instance, no additional stats instances are allocated.  This
eliminates the need to pre-allocate stats instances that will not be
used, also relieving the stats reader from the burden of reading stats
that are never used.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>

datapath: Remove 5-tuple optimization.

The 5-tuple optimization becomes unnecessary with a later per-NUMA
node stats patch.  Remove it first to make the changes easier to
grasp.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>

datapath: fix dp check in ovs_dp_reset_user_features

This fixes crash when userspace does "ovs-dpctl add-dp dev" where dev is
existing non-dp netdevice.

Introduced by:
commit 94358dcffbec33cbcfd425e925139fd7e9d6153f
"openvswitch: Drop user features if old user space attempted to create datapath"

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Jesse Gross <jesse@nicira.com>

tests: Add support for automatically running Ryu tests against OVS.

The Ryu controller comes with an extensive library of OpenFlow tests, but
it doesn't seem so easy to me to run all of them against a development
version of Open vSwitch.  This commit introduces a Makefile target so that
one can run all the Ryu tests with a simple "make check-ryu".

This commit adds documentation for the new target to INSTALL.  It also
moves the documentation for the "check-oftest" target and the
--enable-coverage configure option into INSTALL.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Reviewed-by: Simon Horman <horms@verge.net.au>

util: Pre-allocate buffer for ovs_lasterror_to_string().

This lets us call ovs_lasterror_to_string() and not having
to do an extra call of LocalFree() on the returned string.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

lockfile: Implementation for Windows platform.

Start with a simple implementation that does not support
symbolic links.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

util: follow_symlinks() for windows.

Start with not supporting symbolic links for windows.
This is useful for an upcoming commit.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

string: strerror_r() for windows.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

util: Make xreadlink a static function.

It is only used in util.c

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

ofproto-dpif: Use adaptive mutex for stats.

Stats critical sections are held for very short time, so it makes
sense to spin on them for a while rather than sleeping immediately.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

ovs-thread: Add support for pthread adaptive mutex

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

datapath: Fix race.

ovs_vport_cmd_dump() did rcu_read_lock() only after getting the
datapath, which could have been deleted in between.  Resolved by
taking rcu_read_lock() before the get_dp() call.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>

ofproto: Move 'rule->used' to the provider.

Rule's 'used' timestamp is updated at the same time with the other
stats.  So far the 'used' has been updated without proper protection,
which may lead to 'tearing' in 32-bit architectures, resulting in an
incorrect 'used' timestamp.  While in practice this is highly
improbable, it is still better to handle this correctly.

This is resolved by moving the 'used' field to the provider's stats,
so that whatever protection is used for updating packet and byte
counts, can be also used for both reading and writing of the 'used'
timestamp.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

ofproto: Optimize the case of a repeated learn action execution.

When the target flow exists already as intended, only the 'modified'
time needs to be updated.  Allow modification without taking the
'ofproto_mutex' by always using the 'rule->mutex' when accessing the
'modified' time.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

ofproto: Lock for vlan splinters only if have them.

Reading the hmap count for determining if it is empty or not is thread
safe, so avoid locking when not necessary.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

netdev-dummy: Reduce reconnect back off timeout

netdev-dummy will mostly be used for testing and debugging over fairly
reliable connection. Reduce reconnect back off timeout in case the first
connect attempt failed.

Signed-off-by: Andy Zhou <azhou@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>

netdev-dummy: Fix reconnecting.

The netdev-dummy unit test ran into the reconnect condition on Jarno's
machine. With his test environment, we found and fixed the bugs in
handling reconnect.

Co-authored-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Andy Zhou <azhou@nicira.com>
Acked-by: Jarno Rajahalme <jrajahalme@nicira.com>

meta-flow: Fix setting MFF_IP_FRAG.

mf_set_flow_value() was not setting 'flow->nw_frag' properly.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>

tunnel: Support all combinations of flow-based and specific tunnel matches.

There are 12 possible ways to specify a tunnel (2 * 2 * 3 == 12):

    - Specific in_key or flow-based (2 choices).

    - Specific ip_dst or flow-based (2 choices).

    - Specific ip_src, wildcarded, or flow-based (3 choices).

Until now, only 6 of the 12 possibilities have been supported.  We
have had a couple of requests to add another.  This commit adds all the
possibilities, so that we won't have to add the other 6 one by one.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Requested-by: Thomas Morin <thomas.morin@orange.com>
Acked-by: pritesh <pritesh.kothari@cisco.com>

bond: Change the way of assigning bond slave for unassigned bond entry.

Before this commit, ovs randomly selects a slave for unassigned
bond entry.  If the selected slave is not enabled, the active slave
is chosen instead.  In this commit, the slave is selected from the
list of all enabled slaves in a round-robin fashion.  This helps
improve the consistency of bond behavior when new flows are added.

Signed-off-by: Alex Wang <alexw@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

tests: Fix MPLS test cases.

The "userspace" MPLS test case was checking the same things as the
"drop" test case, rather than checking to see that packets were being
sent to userspace. This patch makes the testsuite consistent with itself.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

flow: Copy MPLS label on mpls_push.

OpenFlow 1.3.3 spec (and earlier) specify that the default value for an
MPLS label should be copied from the outer header.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

tests: Add basic MPLS push test.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

stream-ssl: Add support for Windows platform.

This commit creates events and through poll_fd_wait_event()
associates them with socket file descriptors to get woken up
from poll_block().

Some other changes:

* Windows does not have sys/fcntl.h but has a fcntl.h
On Linux, there is fctnl.h too.

* include <openssl/applink.c> to handle different C-Runtime linking
of OVS and openssl libraries as suggested at
https://www.openssl.org/support/faq.html#PROG2

The above include will not be needed if we compile Open vSwitch with
/MD compiler option.

* SHUT_RDWR is equivalent to SD_BOTH on Windows.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

socket-util: error number to string for sockets.

For winsock2 functions, error number has to be converted to string
using FormatMessage().

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

poll-loop: Make poll_fd_wait_event() cross-platform.

This is helpful if we want to wait either on 'fd' for POSIX or
events for Windows.

For Windows, if both 'fd' and 'wevent' is specified, we associate
that event with the 'fd' using WSAEventSelect() in poll_block().
So any 'events' on that 'fd' will wake us up from WaitForMultipleObjects().

WSAEventSelect() does not understand  POLLIN, POLLOUT etc.  Instead the
macros it understands are FD_READ, FD_ACCEPT, FD_CONNECT, FD_CLOSE etc.
So we need to make that transition.

Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

tests: Add missing banner.

Reported-by: Arun Sharma <arun.sharma@calsoftinc.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

ofproto-dpif-xlate: Make flows that match ICMP fields revalidate correctly.

ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields.  struct flow
uses the low 8 bits of the 16-bit tp_src and tp_dst members to
represent these fields.  The datapath interface, on the other hand,
represents them with just 8 bits each.  This means that if the high 8
bits of the masks for these fields somehow become set (meaning to
match on the nonexistent "high bits" of these fields) during
translation, then they will get chopped off by a round trip through
the datapath, and revalidation will spot that as an inconsistency and
delete the flow.  This commit avoids the problem by making sure that
only the low 8 bits of either field can be unwildcarded for ICMP.

This seems like the minimal fix for this problem, appropriate for
backporting to earlier branches.  The root of the issue is that these high
bits can get set in the match at all.  I have some leads on that, but they
require more invasive changes elsewhere.

Bug #23320.
Reported-by: Krishna Miriyala <miriyalak@vmware.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Andy Zhou <azhou@nicira.com>

pcap-file: Allow capturing TCP streams where the SYN is not seen.

Until now, the tcp_stream() code has ignored any TCP packets that don't
correspond to a stream that began with a TCP SYN.  This commit changes the
code so that any TCP packet that contains a payload starts a new stream.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Reported-by: Vasu Dasari <vdasari@gmail.com>

upcall: Add appctl call to set flow_limit.

This should assist testing of datapath performance, as it allows us to
skip "warming up" the flow limit value.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

datapath: Read tcp flags only then the tranport header is present.

Only the first IP fragment can have a TCP header, check for this.

Signed-off-by: Jarno Rajahalme <jrajahalme@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>

tunnel: fix warning

tunnel: removed unused function

updated to new rx_recv semantics

Merge branch 'mainstream'

tests: Don't rely on strace for IPv6 IDL testcases.

This would cause testsuite failures if someone runs the testsuite
without strace installed.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

netdev-dummy: disable periodic probing

netdev-dummy is mostly useed over reliable link for testing. Periodic
probing over those links seem overkill.

Signed-off-by: Andy Zhou <azhou@nicira.com>
Acked-by: Ben Pfaff <blp@nicira.com>

ofproto: Update documentation

'cls' is no longer a parameter to this function.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>

collectors: Improve log message to better describe errors.

Signed-off-by: Ben Pfaff <blp@nicira.com>

ovs-thread: Add a comment.

Reported-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Signed-off-by: Ben Pfaff <blp@nicira.com>

FAQ: Correct typo.

Reported-by: Pasi Kärkkäinen <pasik@iki.fi>
Signed-off-by: Ben Pfaff <blp@nicira.com>

FAQ: Describe the state of MPLS in Open vSwitch.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Simon Horman <horms@verge.net.au>

tests: Add MPLS + VLAN tests.

Originally part of "odp: Allow VLAN actions after MPLS actions"
by Joe Stringer.

Co-authored-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Ben Pfaff <blp@nicira.com>

Always insert MPLS labels after VLAN tags.

OpenFlow 1.1 and 1.2 always inserted MPLS labels after VLAN tags.
OpenFlow 1.3 and 1.4 insert MPLS labels before VLAN tags.
OpenFlow 1.3.4 and 1.5, both in preparation, recognize that the change in
1.3 was an error and revert it.  This commit implements that reversion
in Open vSwitch.

EXT-457.
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Simon Horman <horms@verge.net.au>

ofproto-dpif.at: Avoid races in "ofproto-dpif - dummy interface"

Signed-off-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Signed-off-by: Ben Pfaff <blp@nicira.com>