Try IPPROTO_IPV6/IPV6_TCLASS socket option as well as IPPROTO_IP/IP_TOS
so that this can work for IPv6 sockets.
IPPROTO_IP/IP_TOS socket option is, as it's SOL indicates, for IPv4.
What happens when it's used for IPv6 sockets? On Linux, it seems to
be forwarded to IPv4 code and affects IPv4 part of the socket.
(e.g. non-V6ONLY case) But it doesn't seem to be the intention of
this function. On other platforms including NetBSD, it just fails
with ENOPROTOOPT.
Probably this function should take the address family but passing
it around lib/*stream*.c would be a bigger change.
Cc: Arun Sharma <arun.sharma@calsoftinc.com>
Signed-off-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Signed-off-by: Ben Pfaff <blp@nicira.com>
revalidator_sweep__() does not provide udumps, so push_dump_ops() can't
look them up.
This is my fault: I introduced it during review.
Signed-off-by: Ben Pfaff <blp@nicira.com>
Previously, we would delete all ukeys when changing the number of
threads, but leave all flows in the datapath. This would cause
double-counting of stats for any flows that remain in the datapath. This
patch fixes the issue by ensuring that all flows are deleted from the
datapath before changing the number of threads.
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
Previously, we would clean up the ukeys whose flow was not seen in the
most recent dump, while leaving the flow in the datapath. In the
unlikely case that the datapath fails to dump a flow that still exists
in the datapath, this would cause double-counting of those flow stats.
This is currently very rare to see due to batching of datapath flow
deletion, but is more easily observable with upcoming patches which
modify the batch size based on dpif implementation.
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
This splits out functions for re-use by later patches, and compacts the
udump revalidation code.
Co-authored-by: Ethan Jackson <ethan@nicira.com>
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
It is possible for a datapath to dump the same flow twice, for instance
if the flow is the last in a batch of flows to be dumped, then a new
flow is inserted into the same bucket before the flow dumper fetches
another batch.
In this case, datapath flow stats may be duplicated: The revalidator
records the stats from the first flow, using the ukey to get the stats
delta. The ukey is deleted, then the revalidator reads the second
(duplicate) flow and cannot lookup the ukey for the delta. As such, it
will push the stats as-is.
This patch reduces the likelihood of such stats duplications by
deferring ukey deletion until after stats are pushed for deleted flows.
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
Basic test to check that the datapath reports the correct number of
packets seen, after a delay of 1 second.
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ben Pfaff <blp@nicira.com>
Use send/recv for socket stream instead of read/write.
Use event handle for polling on socket stream.
Check windows specific return code.
Signed-off-by: Linda Sun <lsun@vmware.com>
Signed-off-by: Gurucharan Shetty <gshetty@nicira.com>
The following scenario can occur:
1. Handler thread grabs a pointer to an ofproto in handle_upcalls().
2. Main thread removes ofproto and destroys it in destruct().
3. Handler thread uses pointer to ofproto and accesses freed memory.
BOOM!
Each individual step above happens under the xlate_rwlock, but the ofproto
pointer is retained from step 1 to step 3, hence the problem. This commit
fixes the problem by ensuring that after an ofproto is removed but before
it is destroyed, all packet translations get pushed all the way through
the upcall handler pipeline. (No new packet translations can get a pointer
to the removed ofproto.)
Bug #1200351.
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Alex Wang <alexw@nicira.com>
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>
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>
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>
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>
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>
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>
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>
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>
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>
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>
These were useful back when we were trying to use the sparse lock balance
annotations, but we removed those in commit 47b52c7123 (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>
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>
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>
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>
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>
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>
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>
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>
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>
$logfile is already prefixed by "$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>
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>
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>
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>
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>
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>
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>
