Sarita Patra [Tue, 3 Mar 2020 11:35:43 +0000 (03:35 -0800)]
pimd: fix crash seen while executing igmp related configuration
Issue 1:
1. Enable pim on an interface.
2. Configure query-interval or query max response time,
which results in pimd crash.
Root cause:
1. When pim is enabled on an interface, it creates a igmp socket
with querier_timer and other_querier time as NULL.
2. When query-interval/max_response_time is configured, it call the
function igmp_sock_query_reschedule() to reshedule the query. This
function check either of querier_timer or other_querier timer should
be running. Since in this case both are NULL, it results in crash.
Issue 2:
1. Enable pim on an interface.
2. Execute no ip igmp query-interval or query max response time,
which results in pimd crash.
Root cause:
1. When pim is enabled on an interface, it creates a pim interface
with querier_timer and other_querier time as NULL.
2. When no ip igmp query-interval/max_response_time is executed, it will
check either of querier_timer or other_querier timer should be running.
Since in this case both are NULL, it results in crash.
Fix:
When pim is enabled on an interface, it creates a igmp socket with
mtrace_only as true. So add a check if mtrace_only is true, then don't
reshedule the query.
Sarita Patra [Tue, 3 Mar 2020 11:31:22 +0000 (03:31 -0800)]
pimd: neighbor needed true in pim nexthop lookup
Issue:
Client---LHR---RP
1. Add kernel route for RP on LHR. Client send join
2. (*,G) will be get created in LHR and RP.
3. Kill the FRR on all the nodes
4. Start FRR only on LHR node
5. In LHR, (*, G) will be created with iif as unknown.
Root cause:
In the step 4, When LHR will receive igmp join, it will call
the function pim_ecmp_fib_lookup_if_vif_index which will look
for nexthop to RP with neighbor needed as false. So RPF lookup will
be true as the route is present in the kernel. It will create a
(*, G) channel_oil with incoming interface as the RPF interface
towards RP and install the (*,G) mroute in kernel.
Along with this (*,G) upstream gets craeted, which call the function
pim_rpf_update, which will look for the nexthop to RP with neighbor
needed as true. As the frr is not running in RP, no neighbor is present
on the nexthop interface. Due to which this will fail and will update
the channel_oil incoming interface as MAXVIFS(32).
Fix:
pim_ecmp_fib_lookup_if_vif_index() call the function pim_ecmp_nexthop_lookup
with neighbor_needed as true.
Mark Stapp [Mon, 2 Mar 2020 20:51:22 +0000 (15:51 -0500)]
zebra: allow multiple labels in LSP zapi message
The handlers for a couple of the main LSP-oriented zapi
messages explicitly limited themselves to a single out-label.
Allow multiple labels if the sender ... sends them.
So, we should trust directly connected routes the same way, assuming
the interface exists.
Old Behavior:
K 2.2.2.1/32 [0/0] is directly connected, unknown inactive, 00:00:39
New Behavior:
K>* 2.2.2.1/32 [0/0] is directly connected, test1, 00:00:03
As a bonus, this fixes the issues we were seeing with not removing
directly connected routes of certain interface types when
those interfaces go down/are deleted.
Signed-off-by: Stephen Worley <sworley@cumulusnetworks.com>
Donald Sharp [Mon, 24 Feb 2020 14:08:54 +0000 (09:08 -0500)]
bgpd: Cleanup indentation in bgp_route_refresh_receive
Some code in bgp_route_refresh_receive was spread across several
lines because of an end of line commit. Move comment to a place
to allow better formating.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
rbauduin [Mon, 17 Feb 2020 12:13:11 +0000 (13:13 +0100)]
doc: document attribute-unchanged directive
This directive was undocumented, but is the way to configure the
next-hop to be left unchanged for ipv6 setups (in ipv4, a route-map with
`set next-hop unchanged` works fine, but this is unavailable for ipv6) .
Tomas Korbar [Thu, 20 Feb 2020 06:37:55 +0000 (07:37 +0100)]
pimd: Make frr able to be built by gcc 10
- Add extern modifier to some declarations in header file and move
qpim_all_pim_routers_addr definition to pimd/pimd.c
`GCC now defaults to -fno-common. As a result, global variable accesses
are more efficient on various targets. In C, global variables with
multiple tentative definitions now result in linker errors.`
Taken from https://gcc.gnu.org/gcc-10/changes.html
Quentin Young [Wed, 19 Feb 2020 21:43:57 +0000 (16:43 -0500)]
zebra: reduce scope of XCALLOC for gr processing
Somewhat gnarly code flow here that might be leaking memory - can't tell
if it's a test artifact or not, but in any case this reduces the
situations in which we need to alloc a block.
And we don't need to check XCALLOC for success...
And we don't need to null check before XFREE...
Or set XFREE'd pointers to NULL...
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
Donald Sharp [Wed, 19 Feb 2020 15:52:14 +0000 (10:52 -0500)]
bgpd: Add a better breadcrumb for when bgp is missconfiged
Currently During bgp open collision resolution if both
the router-id's are the same, we correctly follow
the RFC and close the connection. The problem is of course
that there is no notification of the error in configuration
to the end user other than a subtle open debug message.
Explicitly call out the miss-configuration as an error message
as that this miss-config took several hours of debugging to notice.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Mark Stapp [Tue, 18 Feb 2020 19:08:21 +0000 (14:08 -0500)]
doc: correct ip prefix token mapping table
Fix the table that shows the mapping between ip network cli
tokens and the data struct that the cli handler will see.
Optional network/subnet types appear as a pointer to an
all-zeroes struct, not a NULL pointer.
Donald Sharp [Tue, 18 Feb 2020 00:43:22 +0000 (19:43 -0500)]
zebra: Fix checksum calculation to not include old checksum
As part of checksum calculation for a received packet we were
comparing the checksum returned from in_cksum. Typically
when we calculate the checksum the value stored in the checksum
must be all 0's. Store the received checksum and then set
the checksum to 0 and then compare.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
* This commit implements the code style suggestions from Polychaeta.
* This commit also introduces a CLI to toggle the optimization and, a hidden
CLI to display the contents of the constructed prefix tree.
* This commit introduces the building blocks.
A per-route-map prefix tree is introduced.
This tree will consist of the prefixes defined within the prefix-lists
that are added to the match clause of that route-map.
projects / matthieu / frr.git / log