In the medium term, we plan to migrate the datapath to use Netlink as its
communication channel. In the short term, we need to be able to have
actions with 64-bit arguments but "struct odp_action" only has room for
48 bits. So this patch shifts to variable-length arguments using Netlink
attributes, which starts in on the Netlink transition and makes 64-bit
arguments possible at the same time.
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Jesse Gross <jesse@nicira.com>
For some time now, Open vSwitch datapaths have internally made a
distinction between adding a vport and attaching it to a datapath. Adding
a vport just means to create it, as an entity detached from any datapath.
Attaching it gives it a port number and a datapath. Similarly, a vport
could be detached and deleted separately.
After some study, I think I understand why this distinction exists. It is
because ovs-vswitchd tries to open all the datapath ports before it tries
to create them. However, changing it to create them before it tries to
open them is not difficult, so this commit does this.
The bulk of this commit, however, changes the datapath interface to one
that always creates a vport and attaches it to a datapath in a single step,
and similarly detaches a vport and deletes it in a single step.
Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Jesse Gross <jesse@nicira.com>
Until now, the collection of coverage counters supported by a given OVS
program was not specific to that program. That means that, for example,
even though ovs-dpctl does not have anything to do with mac_learning, it
still has a coverage counter for it. This is confusing, at best.
This commit fixes the problem on some systems, in particular on ones that
use GCC and the GNU linker. It uses the feature of the GNU linker
described in its manual as:
If an orphaned section's name is representable as a C identifier then
the linker will automatically see PROVIDE two symbols: __start_SECNAME
and __end_SECNAME, where SECNAME is the name of the section. These
indicate the start address and end address of the orphaned section
respectively.
Systems that don't support these features retain the earlier behavior.
This commit also fixes the annoyance that files that include coverage
counters must be listed on COVERAGE_FILES in lib/automake.mk.
This commit also fixes the annoyance that modifying any source file that
includes a coverage counter caused all programs that link against
libopenvswitch.a to relink, even programs that the source file was not
linked into. For example, modifying ofproto/ofproto.c (which includes
coverage counters) caused tests/test-aes128 to relink, even though
test-aes128 does not link again ofproto.o.
This makes it easier for dpif_provider implementations to share code but
distinguish the class actually in use, because comparing a pointer is
easier than comparing a string.
I don't see a reason here to copy the dpif_class instead of using the
pointer provided by the caller. Using the caller's pointer allows the
caller to compare a dpif's 'class' member against the address of its
class structure, which seems like a reasonable thing to do.
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.
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.
Signed-off-by: Ben Pfaff <blp@nicira.com>
Adding a macro to define the vlog module in use adds a level of
indirection, which makes it easier to change how the vlog module must be
defined. A followup commit needs to do that, so getting these widespread
changes out of the way first should make that commit easier to review.
When the QoS code was integrated, I didn't yet know how to abstract the
translation from a queue ID in an OpenFlow OFPAT_ENQUEUE action into a
priority value for an ODP ODPAT_SET_PRIORITY action. This commit is a
first attempt that works OK for Linux, so far. It's possible that in fact
this translation needs the 'netdev' as an argument too, but it's not needed
yet.
When a dpif passes an odp_msg down to ofproto, and ofproto transforms it
into an ofp_packet_in to send to the controller, until now this always
involved a full copy of the packet inside ofproto. This commit eliminates
this copy by ensuring that there is always enough headroom in the ofpbuf
that holds the odp_msg to replace it by an ofp_packet_in in-place.
From Jean Tourrilhes <jt@hpl.hp.com>, with some revisions.
If dpif_flow_get()'s caller is less cautious than it should be, then it
will get surprising results when it looks at the returned flow on error.
This commit at least gives it plausible results.
The comment here was out of touch with the actual definition. Limiting a
log message to 9999 per minute is not much of a limit!
I suspect that I turned this way up during initial development at some
point and forgot to turn it down to a reasonable level.
This brings over some features that were added to the netdev interface,
most notably the separation between the name and the type. In addition
to being cleaner, this also avoids problems where it is expected that
the local port has the same name as the datapath.
No conflicts, but lib/dpif.c needed a few changes since struct dpif's
member "class" was renamed to "dpif_class" in master since sflow was
branched off.
This function combines what dpif_create() and dpif_open() do. It allows
us to factor a tiny amount of code out of the vswitch, but more importantly
this function is also useful in the following commit.
Originally, the function dp_enumerate() initialized the 'all_dps'
argument. This is inconsistent with most other functions that take an
svec argument, which would only clear the contents. Further, if someone
were not careful when reusing the svec, it could lead to memory leaks.
With this change, the caller is expected to first call svec_init() on
the argument.
This new abstraction layer allows multiple implementations of network
devices in a single running process. This will be useful, for example, to
support network devices that are simulated entirely in the running process
or that communicate with other processes over Unix domain sockets, etc.
The reimplemented tap device support in this commit has not been tested.
At startup, the vswitch needs to delete datapaths that are not configured
by the administrator. Until now this was done by knowing the possible
names of Linux datapaths. This commit cleans up by allowing each
datapath class to enumerate its existing datapaths and their names.
The dpif and netdev code has had various ways to check for changes to
dpifs and netdevs over the course of Open vSwitch development. All of
these have been thus far fairly specific to the Linux implementation. This
commit is the start of a more general API for watching for such changes.
The dpif-related parts seem fairly mature and so they are documented,
the netdev parts will probably need to change somewhat and so they are
not documented yet.
This commit initially introduces only a single datapath implementation,
which is the same as the original one, but it paves the way for
additional implementations, such as the upcoming userspace datapath.
Until now, ODP_PORT_LIST has reported the number of ports actually copied
out. It's better for the caller, however, if it reports the number of
ports that were available to be copied out.
Soon we will allow for multiple datapath implementations. By allowing
the datapath to choose the port numbers, we possibly simplify some datapath
implementations, and the datapath's clients don't have to guess (or to
check) what port numbers are free, so this seems like a better way to go.
This function is easier for callers to use if they do not have to guess
how many ports are in the group. Since it's not performance critical at
all, introduce these easier semantics.
One potential user of dpif_port_query_by_name() actually open-coded the
function because it didn't want logging on failure. There was only one
actual caller of the function, which didn't want logging on failure either.
So, clean up by reducing the failure log level to DBG and making the
open-coded version an actual caller.
dpif_id() is often used in error messages, e.g. "dp%u: screwed up". But
soon we will be generalizing the concept of a datapath, so it is better
to have a function that returns a full name, e.g. "%s: screwed up".
Accordingly, this commit replaces dpif_id() by a new function dpif_name()
that does so.
The 'minor' member of struct dpif is used for two different purposes:
for printing in log messages and for encapsulating in NetFlow messages.
The needs in each case are different, so we should break up these uses.
This commit does half of that, by introducing a new function to retrieve
NetFlow ids and using it where appropriate.
With multiple kinds of datapaths, code should not just use
"dp%u" along with dpif_minor() to print a datapath name, because not all
datapaths can sensibly be named that way. We want to use a function
with a name like dpif_get_name() to retrieve a datapath name for printing
to the user, in which case the existing dpif_get_name() function would be
confusing. So rename the existing one to something more explicit.
