pimd: When we are initializing a pim socket limit hellos

When we are initializing a pim socket for vrf or loopback
interfaces do not schedule a hello to go out at all.

I'm currently leaving the check on is a vrf / loopback
device on the actual send as that we have several paths
to get there.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Remove knowledge of HAVE_TCP_ZEBRA from pim

pim was the only routing daemon to have any knowledge
of how zebra connects to each daemon.  There is no
need for this.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Ensure packet received on vrf interface is configured for pim

In a vrf configuration, when we receive a pim packet we lookup
the correct incoming interface.  There exists a chance that
the correct incoming interface has not been configured to use
pim yet.  gracefully bow out and do nothing with the packet.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

lib, pimd, zebra: Allow pim to set pimregX into appropriate vrf

The pimregX devices when created by the kernel are put into
the default vrf.  When pim gets the callback that the device
exists, check to see if it is a pimregX device and if so
move it into the appropriate vrf.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Bind msdp listen socket to vrf

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Cleanup shutdown of vrf

When shutting down the vrf don't kill the default vrf
when attempting to cleanup something else.

Also move the terminate to the right spot.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix msdp show command

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: For non-default vrf's bind the msdp socket to the vrf

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix 'show ip pim state...' vrf changes

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

zebra: Allow S,G lookup to use RTNL_FAMILY_IPMR

This current implementation unfortunately must
ask the kernel for all mroutes because vrf's
do not have the ability to request a single
mroute at this time.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: debug guard error message

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

zebra: Remove strcpy from fast path

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Cleanup some join debug messages

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix help string

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Add 'debug pim trace detail' command

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Reschedule the correct interface for incoming packets

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Trust the recvfromto ifindex information

The vrf interface is receiving the pim packet
instead of the slave interface that is bound.

Lookup the ifindex ifp pointer from that.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Don't send hello's out vrf interface

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: We may receive a igmp packet in on unconfiged intf

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix dropped(?) telling of the kernel to recv cmsg data

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix mixup in addition to vrf node

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Bind pim kernel fd to appropriate vrf

Bind the pim kernel fd to the appropriate vrf, modify
the callback up into pim with the IGMP report to
retrieve the incoming interface and use that to
lookup the correct interface to use.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Push 'struct pim_instance *' a bit further

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix crash when attempting to use channel_oil->pim before it is assigned

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Cleanup some basic crashes from mistakes made converting to vrf

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix crash in 'show ip multicast vrf..'

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Convert pim_cmd.c to use correct pim instance

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix instance startup to separate data structure setup from use

Fix the instance startup to allow separation between the data structure
setup for pim instances from actual attempt to use.  More than
likely there will be other issues that need to be fixed but
this gets us 'working' again.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Add vrf help strings to some commands

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Start 'show ip pim [vrf NAME] ...' conversions

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Add some vrf knowledge to some debugging statements

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Add vrf ability to msdp show commands

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: All CONFIG_NODE commands are now available in VRF_NODE

Note -> Some commands storage needs to be moved into pim_instance

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Create macro to get appropriate pim instance

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Cleanup the rest of the global config to be vrf aware for spacing

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Add code to write spaces or not before 'ip pim rp'

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Get show run to display vrf sub mode

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix intentional pass of NULL

pim_msdp_pkt_sa_gen is intentionally passed mp as NULL
in some cases, so we need to pass through the pim instance

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix startup of pim instance a bit

We will need to refine the difference between create and enable
of a pim instance.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix crash associated with pim_ifp not actually being created yet

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Cleanup pim_msg.c to use correct pim instance

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Make pim_ssmpingd.c 'struct pim_instance' aware

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Make msdp `struct pim_instance *` aware

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Make pim MSDP pim instance aware to some extent

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Get the default pim code ready to write into vrf's or not

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Allow PIM to start writing vrf information if it has it

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Cleanup pim_pim use of pimg

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Cleanup pimg from pim_igmpv3.c

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Lookup loopback based upon vrf correctly.

The vrf loopback ip address should be associated with the
vrf device for vrf's, else use the loopback.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Cleanup pim_ssm.c to use 'struct pim_instance *'

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Finish up pimg -> pim conversion in pim_rpf.c

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Cleanup missed pimg

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Finish up pim_mroute.c 'struct pim_instance' awareness

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Make pim_hello choose correct vrf

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Cleanup pimg from pim_upstream.c

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Make zlookup 'struct pim_instance' aware

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Make XX_forward_start 'struct pim_instance *' aware

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Convert channel_oil_hash and list into 'struct pim_instance *'

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Convert vif and ifindex lookups to use 'struct pim_instance *'

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Move pim_ifchannel_list and vif array into 'struct pim_instance *'

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Remove pimg from pim_iface.c

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Remove pimg from pim_register.c

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Cleanup pim_jp_agg_upstream_verification function

Do not use pimg, figure out which instance we are in and use that.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Convert pim_rp.c to use 'struct pim_instance *'

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Remove pimg from pim_upstream.c

Move the upstream_list, hash and wheel into 'struct pim_instance'
Remove all pimg to pim in pim_upstream

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: pim_rp.c -> convert pimg to pim

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Remove pimg from pim_nht.c in totality

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: More refactor for 'struct pim_instance *' for pim_nht.c

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Add 'struct pim_instance' to find_or_track_nexthop

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Make nexthop cache pim instance aware

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix mistake in uplifting to new thread model

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Refactor some functions

A bunch of functions had return values that were never
checked for ( and not needed ) and opposite return values
for proper calling function boolean logic.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Have pim_if_find_by_vif_index iterate over all vrf's

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Refactore qpim_static_route_list into pim->static_routes

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix c_oil->pim set spot

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Refactor code to be in better spots

1) Create pim_instance.[ch] to allow us to handle the instance information there
2) Refactor some pim_rpf_ and some pim_rp commands into appropriate files and
appropriate includes.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Set c_oil->pim for static routes

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: If c_oil->pim is set use that pim instance in pim_mroute.c

As a transitory mechanism, if c_oil->pim is set, use that particular
pim instance, else use the default pimg.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Add vrf output to 'interface ...' write command

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix 'show ip multicast'

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Modify 'show ip multicast' to be VRF aware

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Create the pimreg device name correctly

The pimreg device name should be based upon the
tableid we create( as that this is what the kernel does )

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Use actual vrf table_id for pim vrf creation

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Create the pimreg per vrf

pimd: Create pim_get_pim_instance

Create function to get the pim instance from the vrf_id

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: pim_mroute_[add|del]_vif become VRF aware

Make the pim_mroute_[add|del]_vif functions be vrf aware.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Store the pim instance off of the pim interface

Storing the pim instance off of the pim interface will allow
us to easily keep track of vrf information.  This is because
pretty much is everything is interface based.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimg: Actually properly create the pim socket per vrf

Create the actual pim socket per vrf.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Start cleanup of documentation

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Convert upcalls from kernel to use appropriate pim instance

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Move thread read to struct pim_instance

When we are handling the thread read/writes for
a pim mroute socket, make it so that it can
be appropriately handled by the 'struct pim_instance'
instead of defaulting to the default VRF's

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Call code to create appropriate MRT_TABLE

When dealing with vrf's we need to create a table within
the linux kernel to handle the vrf.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Remove unnecessary NULL check

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Ensure a 0 read fails cleanly

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Prevent NULL dereference

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Error check str2prefix

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix unchecked return value

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Save pim instance on vrf pointer

This commit does these things:

1) Saves the pim instance created on the vrf pointer
2) Moves some initialization to the proper spot.
3) Remove vrf_id from 'struct pim_ssm' it is not needed
4) Removes some checks to prevent non-default vrf's from
being created.
5) When creating the pim instance, save default vrf in pimg to allow backwards compatibility

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Modify pim_mroute_socket_[enable|disable] to be vrf aware

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Use setsockopt_so_recvbuf instead of doing it ourselves

This change allows other non-linux platforms to be a bit
more forgiving if we ask for a very very large size.

Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>