During vrf delete, the vxlan_info.work_list
linked list was deleted which is a global list
containing the SGs for all the VRFs.
If two vrfs are configured, vrf a and vrf b and
both has SGs assocaited with them which are
inserted in the vxlan_info.work_list. Now if
vrf a is deleted, it deletes the work_list also.
Due to this when any SG add or del comes for vrf b
it tries to access the work_list and crashes.
Fix
Delete the vxlan_info.work_list only when all the
VRFs are terminated and unset the vxlan_info.flags
so if new add cmd comes it re-allocates the work_list.
Signed-off-by: usrivastava-nvidia <usrivastava@nvidia.com>
Bind the autorp socket to the vrf device.
Also fixed mroute socket to use vrf_bind instead of directly
setting the socket option.
Signed-off-by: Nathan Bahr <nbahr@atcorp.com>
In 9461953914 and
8ebcc02328, transmission of PIM register and
register stop messages was changed to use a separate socket. However, that
socket is not bound to a possible VRF, so the messages were sent in the
default VRF instead. Call vrf_bind() once after socket creation and when the
VRF is ready to ensure transmission in the correct VRF. vrf_bind() handles
the non-VRF case (i.e. VRF_DEFAULT) automatically, so it may be called
unconditionally.
Signed-off-by: Martin Buck <mb-tmp-tvguho.pbz@gromit.dyndns.org>
Refactor the next hop tracking in PIM to fully support the configured RPF lookup mode.
Moved many NHT related functions to pim_nht.h/c
NHT now tracks both MRIB and URIB tables and makes nexthop decisions based on the configured lookup mode.
Signed-off-by: Nathan Bahr <nbahr@atcorp.com>
Reorganize the MSDP initialization code and configuration writing code
to its appropriated place.
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
Signalize termination to functions so they can avoid accessing pointers
that might be no longer available.
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
Perform AutoRP discovery and candidate RP announcements using the
AutoRP protocol.
Mapping agent is not yet implemented, but this feature is not
necessary for FRR to support AutoRP as we only need one AutoRP
mapping agent in the network.
Signed-off-by: Nathan Bahr <nbahr@atcorp.com>
Moved all existing global/vrf PIM config to the new subnode.
Existing configuration updated to be hidden and deprecated.
Both versions of configuration still work together.
Signed-off-by: Nathan Bahr <nbahr@atcorp.com>
`pim_if_create_pimreg()` need use the `mroute_socket`, so adjust the order.
First call `pim_mroute_socket_enable()`, then call `pim_if_create_pimreg()`.
Signed-off-by: anlan_cs <vic.lan@pica8.com>
Add a hash_clean_and_free() function as well as convert
the code to use it. This function also takes a double
pointer to the hash to set it NULL. Also it cleanly
does nothing if the pointer is NULL( as a bunch of
code tested for ).
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
The problem here is when the same node is FHR as well as RP,
then the node keeps on sending the register packet.
Register-stop is not sent as well.
This problem has occurred because the RP is the same node
and there is no socket created on loopback interface, so the
packet is never send out and never received back on the same
node, so register recv could not be processed on the node and
hence no register-stop is sent.
Since register packets are unicast packets, its better to handle
the send of register packet via a separate register socket.
This fixes the problem mentioned above as well.
Fixes: #11331
Signed-off-by: Mobashshera Rasool <mrasool@vmware.com>
Issue:
==16837== Invalid read of size 8
==16837== at 0x17971C: pim_neighbor_find (pim_neighbor.c:431)
==16837== by 0x186439: join_timer_stop (pim_upstream.c:348)
==16837== by 0x186794: pim_upstream_del (pim_upstream.c:231)
==16837== by 0x189A66: pim_upstream_terminate (pim_upstream.c:1951)
==16837== by 0x17111B: pim_instance_terminate (pim_instance.c:54)
==16837== by 0x17111B: pim_vrf_delete (pim_instance.c:172)
==16837== by 0x4F1D6C8: vrf_delete (vrf.c:264)
==16837== by 0x19006F: pim_terminate (pimd.c:160)
==16837== by 0x1B2E4D: pim_sigterm (pim_signals.c:51)
==16837== by 0x4F08FA2: frr_sigevent_process (sigevent.c:130)
==16837== by 0x4F1A2CC: thread_fetch (thread.c:1771)
==16837== by 0x4ED4F92: frr_run (libfrr.c:1197)
==16837== by 0x15D81A: main (pim_main.c:176)
Root Cause:
In the pim_terminate flow, the interface is deleted
before the pim_interface clean up. Because of this,
the pim_interface is having garbage value.
Fix:
Release the pim interface memory and then delete the
interface.
Signed-off-by: Sarita Patra <saritap@vmware.com>
pim_mroute_socket_disable api is present but nowhere called.
This should be called when pim instance is terminated.
Fixed it.
Signed-off-by: Mobashshera Rasool <mrasool@vmware.com>
This is causing build issues on BSD by including (transitively)
`linux/mroute6.h` - try to address by disentangling the headers a bunch.
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>
Currently, it is possible to rename the default VRF either by passing
`-o` option to zebra or by creating a file in `/var/run/netns` and
binding it to `/proc/self/ns/net`.
In both cases, only zebra knows about the rename and other daemons learn
about it only after they connect to zebra. This is a problem, because
daemons may read their config before they connect to zebra. To handle
this rename after the config is read, we have some special code in every
single daemon, which is not very bad but not desirable in my opinion.
But things are getting worse when we need to handle this in northbound
layer as we have to manually rewrite the config nodes. This approach is
already hacky, but still works as every daemon handles its own NB
structures. But it is completely incompatible with the central
management daemon architecture we are aiming for, as mgmtd doesn't even
have a connection with zebra to learn from it. And it shouldn't have it,
because operational state changes should never affect configuration.
To solve the problem and simplify the code, I propose to expand the `-o`
option to all daemons. By using the startup option, we let daemons know
about the rename before they read their configs so we don't need any
special code to deal with it. There's an easy way to pass the option to
all daemons by using `frr_global_options` variable.
Unfortunately, the second way of renaming by creating a file in
`/var/run/netns` is incompatible with the new mgmtd architecture.
Theoretically, we could force daemons to read their configs only after
they connect to zebra, but it means adding even more code to handle a
very specific use-case. And anyway this won't work for mgmtd as it
doesn't have a connection with zebra. So I had to remove this option.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
There is a possibility that the same line can be matched as a command in
some node and its parent node. In this case, when reading the config,
this line is always executed as a command of the child node.
For example, with the following config:
```
router ospf
network 193.168.0.0/16 area 0
!
mpls ldp
discovery hello interval 111
!
```
Line `mpls ldp` is processed as command `mpls ldp-sync` inside the
`router ospf` node. This leads to a complete loss of `mpls ldp` node
configuration.
To eliminate this issue and all possible similar issues, let's print an
explicit "exit" at the end of every node config.
This commit also changes indentation for a couple of existing exit
commands so that all existing commands are on the same level as their
corresponding node-entering commands.
Fixes#9206.
Signed-off-by: Igor Ryzhov <iryzhov@nfware.com>
VRF creation can happen from either cli or from
knowledged about the vrf learned from zebra.
In the case where we learn about the vrf from
the cli, the vrf id is UNKNOWN. Upon actual
creation of the vrf, lib/vrf.c touches up the vrf_id
and calls pim_vrf_enable to turn it on properly.
At this point in time we have a pim->vrf_id of
UNKNOWN and the vrf->vrf_id of the right value.
There is no point in duplicating this data. So just
remove all pim->vrf_id and use the vrf->vrf_id instead
since we keep a copy of the pim->vrf pointer.
This will remove some crashes where we expect the
pim->vrf_id to be usable and it's not.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
When creating configuration for a vrf *Before* the vrf has been
created, pim will not properly create the pimreg device and
we will promptly crash when we try to pass data.
Put some code checks in place to ensure that the pimreg is
created for vrf's.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
Valgrind is reporting this:
==22220== Invalid read of size 4
==22220== at 0x11DC2B: pim_if_delete (pim_iface.c:215)
==22220== by 0x11DD71: pim_if_terminate (pim_iface.c:76)
==22220== by 0x128E03: pim_instance_terminate (pim_instance.c:66)
==22220== by 0x128E03: pim_vrf_delete (pim_instance.c:159)
==22220== by 0x48E0010: vrf_delete (vrf.c:251)
==22220== by 0x48E0010: vrf_delete (vrf.c:225)
==22220== by 0x48E02FE: vrf_terminate (vrf.c:551)
==22220== by 0x149495: pim_terminate (pimd.c:142)
==22220== by 0x13C61B: pim_sigint (pim_signals.c:44)
==22220== by 0x48CF862: quagga_sigevent_process (sigevent.c:103)
==22220== by 0x48DD324: thread_fetch (thread.c:1404)
==22220== by 0x48A926A: frr_run (libfrr.c:1122)
==22220== by 0x11B85E: main (pim_main.c:167)
==22220== Address 0x5912160 is 1,200 bytes inside a block of size 1,624 free'd
==22220== at 0x48369AB: free (in /usr/lib/x86_64-linux-gnu/valgrind/vgpreload_memcheck-amd64-linux.so)
==22220== by 0x128E52: pim_instance_terminate (pim_instance.c:74)
==22220== by 0x128E52: pim_vrf_delete (pim_instance.c:159)
==22220== by 0x48E0010: vrf_delete (vrf.c:251)
==22220== by 0x48E0010: vrf_delete (vrf.c:225)
==22220== by 0x48F1353: zclient_vrf_delete (zclient.c:1896)
==22220== by 0x48F1353: zclient_read (zclient.c:3511)
==22220== by 0x48DD826: thread_call (thread.c:1585)
==22220== by 0x48A925F: frr_run (libfrr.c:1123)
==22220== by 0x11B85E: main (pim_main.c:167)
==22220== Block was alloc'd at
==22220== at 0x4837B65: calloc (in /usr/lib/x86_64-linux-gnu/valgrind/vgpreload_memcheck-amd64-linux.so)
==22220== by 0x48ADA4F: qcalloc (memory.c:110)
==22220== by 0x128B9B: pim_instance_init (pim_instance.c:82)
==22220== by 0x128B9B: pim_vrf_new (pim_instance.c:142)
==22220== by 0x48E0C5A: vrf_get (vrf.c:217)
==22220== by 0x48F13C9: zclient_vrf_add (zclient.c:1863)
==22220== by 0x48F13C9: zclient_read (zclient.c:3508)
==22220== by 0x48DD826: thread_call (thread.c:1585)
==22220== by 0x48A925F: frr_run (libfrr.c:1123)
==22220== by 0x11B85E: main (pim_main.c:167)
On pim vrf deletion, ensure that the vrf->info pointers are NULL as well
as the free'd pim pointer for ->vrf is NULL as well.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
Replace sprintf with snprintf where straightforward to do so.
- sprintf's into local scope buffers of known size are replaced with the
equivalent snprintf call
- snprintf's into local scope buffers of known size that use the buffer
size expression now use sizeof(buffer)
- sprintf(buf + strlen(buf), ...) replaced with snprintf() into temp
buffer followed by strlcat
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
RCA: When configured more than 32(MAXVIS), the inerfaces that are configured
after 32nd interfaces have the value of MAXVIF.
This is used as index to access the free vif tracker of array size 32(MAXVIFS).
So the channel oil list pointer which is present as the next field in pim structure get corrupt, when updating free vif.
This gets accessed during rpf update resulting in crash.
Fix: Refrain from allocating mcast interface structure and throw config error when more than MAXVIFS are attempted to configure.
Max vif checks are exempted for vrf device and pimreg as vrf device will be the first interface and not expected to fail and pimreg has reserved vif.
vxlan tunnel termination device creation has this check and throw warning on max vif.
All other creation are through CLI.
Signed-off-by: Saravanan K <saravanank@vmware.com>
When pim receives a register packet, we will apply the
received source to the prefix list. If accepted normal
processing continues. If denied we will send a register
stop message to the source.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The memory type PIM_SPT_PLIST_NAME is specific to
SPT but we are going to store more prefix-list names
in pim, make it generic to allow for less confusion
in the future.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
when ever a FRR Client wants to send any data to another node
using MLAG Channel, uses below mechanisam.
1. sends a MLAG Registration to zebra with interested messages that
it is intended to receive from peer.
2. In response to this request, Zebra opens communication channel with
MLAG. and also in Rx. diretion zebra forwards only those messages which
client shown interest during registration
3. when client is no-longer interested in communicating with MLAG, client
posts De-register to Zebra
4. if this is the last client which is interested for MLAG Communication,
zebra closes the channel.
why PIM Needs MLAG Communication
================================
1. In general on LAN Networks elecetd DR will send the Join towards
Multicast RP in case of a LHR and Register in case of FHR.
2. But in case DR Goes down, traffic will be re-converged only after
the New DR is elected, but this can take time based on Hold Timer to
detect the DR down.
3. this can be optimised by using MLAG Mecganisam.
4. and also Traffic can be forwarded more efficiently by knowing the cost
towards RP using MLAG
Signed-off-by: Satheesh Kumar K <sathk@cumulusnetworks.com>
when ever a FRR Client wants to send any data to another node
using MLAG Channel, uses below mechanisam.
1. sends a MLAG Registration to zebra with interested messages that
it is intended to receive from peer.
2. In response to this request, Zebra opens communication channel with
MLAG. and also in Rx. diretion zebra forwards only those messages which
client shown interest during registration
3. when client is no-longer interested in communicating with MLAG, client
posts De-register to Zebra
4. if this is the last client which is interested for MLAG Communication,
zebra closes the channel.
why PIM Needs MLAG Communication
================================
1. In general on LAN Networks elecetd DR will send the Join towards
Multicast RP in case of a LHR and Register in case of FHR.
2. But in case DR Goes down, traffic will be re-converged only after
the New DR is elected, but this can take time based on Hold Timer to
detect the DR down.
3. this can be optimised by using MLAG Mecganisam.
4. and also Traffic can be forwarded more efficiently by knowing the cost
towards RP using MLAG
Signed-off-by: Satheesh Kumar K <sathk@cumulusnetworks.com>
Apart from datastructure, bsm scope initialization and deinitialiation
routines called during pim instance init and deinit. Also makefile changes.
Signed-off-by: Saravanan K <saravanank@vmware.com>
1. vxlan instance cleanup needs to be done before the upstream entries are
force-flushed.
2. also vxlan callbacks need to be ignored post instance-cleanup.
Signed-off-by: Anuradha Karuppiah <anuradhak@cumulusnetworks.com>
These entries will be used over the subsequent commits for
1. vxlan-tunnel-termination handling - setup MDT to rx VxLAN encapsulated
BUM traffic.
2. vxlan-tunnel-origination handling - register local-vtep-ip as a
multicast source.
Signed-off-by: Anuradha Karuppiah <anuradhak@cumulusnetworks.com>
The pimg data structure is only used in one spot to send the default
vrf id to zebra upon startup. Add the default vrf id to the struct pim_router
data structure and remove the pimg pointer.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Create a `struct pim_router` and move the thread master into it.
Future commits will further move global varaibles into the pim_router
structure.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Resolves issue with exit-vrf being placed at the end of zebra's portion
of a vrf block, but before other daemons' portions of the same config
block.
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
This commit fixes two issues during pim shutdown.
1) The rp_info structure was being freed before the
outgoing notifications that depended on it's information
was sent out as part of shutdown.
2) The pim->upstream_list shutdown involved iterating
over the list via ALL_LIST_ELEMENTS. This typically
is enough but pim will auto delete child nodes as well
as itself when it goes away and they depend on it. As such
the node and nnode could possibly already have been freed.
So change the way we look at all the data in the upstream_list
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The Vrf aliases can be known with a specific hook. That hook will then,
from zebra propagate the information to the relevant zapi clients.
The registration hook function is the same for all daemons.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
