lib: Fix various matching bugs

Missed a copy, disambigs forgot to walkback

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Reorganize grammar sandbox

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Add testing shim for graph deletion

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Add nullcheck before graph node deletion

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Fix dangling pointer in matcher

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Fix uninitialized pointer segfault

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Update grammar sandbox driver

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Add <cr> text for end nodes

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Remove automatic node deletion

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Fix deduplication bug, reinstate graph print

Comparing the wrong nodes led to duplication during
graph construction, fixed. Also update graph pretty
print to use new graph struct.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Continue matching system refactor

Most things back to working, all CLI units refactored
to use improved graph implementation.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Refactor command_parse.y for graph ds

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Generalize graph to work for any data type

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Refactor CLI interface function names

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Remove debug fprintf for cmd_element copy

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Add completion string convenience functions

And set END_GN default text to <cr>, and scrub
a coupld of extraneous files

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

debian: Add bison and flex to Build-Depends

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Fix broken disambiguation on leader tokens

Fix incorrect assumption that the set of first tokens
of all commands are perfectly unambiguous

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Handle vararg in graph pretty-print

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Remove unnecessary node pointer

Selector and option heads had pointers to their
ends, but this turned out to be unnecessary

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Remove unnecessary XSTRDUP

yy_scan_string duplicates its input

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Revert breaking change to cmp_node

Bad conditional refactor broke graph deduplication
and thus disambiguation precedence

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Fix use after free in lexer

Fix occasional bug where deleting flex's input
buffer happens earlier than flex expects

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Fix various minor bugs

- cmd_make_strvec returns null pointer if str
  begins with a '#'
- disallow options nested options
- NULL out state variable in parser
- flip backwards comparison
- fix memory leak in lexer

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Code cleanup, formatting, & headers

Gnu-style code, add copyright headers, cleanup some
random style issues, shuffle around code into
relevant units, add docs.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Major parser refactor

Lots of cleanup, code consolidation, organizational
improvements.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Add docstring support

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Improve matching disambiguation

Disambiguation logic now compares full paths
instead of defaulting to an ambiguous match
if the heads cannot be disambiguated

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Unrefactor to signed long long for ranges

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Allow optional whitespace in ranges

Makes ranges where both endpoints are negative
somewhat more readable. Also validate ranges.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Add support for negative ranges

And convert range delimiters to signed int

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Refactor format parser

Also fix a bug where selector elements beginning
with the same word are only parsed once

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Fix use after free in matcher

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Implement status variable

Matcher now keeps track of why it failed

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Implement full command disambiguation logic

Disambiguation is smarter and somewhat more compact.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Add vararg support

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Reorganize command matcher

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Add partial completion support

Completions now include nodes that the input
partially matches as well as the children of
nodes those that the input exactly matches.

Also some minor cleanup and bugfixes.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Fix OOB range parses, variable matches

Variables now allow strings beginning with numbers
to match, ranges and numbers are now long long to
fix OOB parses resulting in integer wraparounds.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Fixed bad node copy, modified token regex

When building argv for matched command, only the
last node was being copied to argv; the rest were
added by reference. Additionally the regex for
certain tokens was too restrictive and disallowed
characters allowed by the old parser; these have
been reinstated.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Add partial matching support

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Retab command_match.c

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Re-add partial IPV4 matching support

Also removed debugging output for matcher.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Cleanup some memory issues in CLI

Various memory leaks have been fixed and the quagga
memory macros are in use. Also consolidated the argv
and matching code into one graph traversal.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Improve argv construction

Fixed a variety of failure cases and wrote a nice
doc string. Reverted add_node doc string to correct
explanation of procedure.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Cleanup parser memory management

Free as appropriate. Additionally add new type
of node to demark graph head (START_GN).

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Add matching and argv support

Queries may be run against DFA's to find matching
cmd_element, and argument lists may be constructed.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Incremental matching improvement

Shotgun changes to matching system

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Reorganize some matching stuff

Introduce new node type, END_GN, and remove
is_leaf flags. Reorganize command_match.c & remove
internal functions from command_match.h. Start
rewriting command.h in command_new.h with changes
for new backend.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Mostly complete matcher

Input matching and completions works. Still some
rough edges.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Break up functions, begin matcher

Moved test hook out of command.c into vtysh.c,
renamed graph modules, added matching code

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Implement node comparison function

Implement comparator for nodes, some miscellaneous cleanup.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Fix some DFA construction bugs

Options get null paths, parser state is properly
cleaned up, caller passes start node

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Finish implementing grammar, DFA construction

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Add generated parser files to .gitignore

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Change some includes

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Update build config for new parser

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Start implementing DFA

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Add parser, lexer, and command tree skeleton

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Cleanup cmd_execute_command

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

Add grammar sandbox framework

pimd: Start naive implementation of anysource_forward_stop

Naive implementation of igmp_anysource_foward_stop.

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

pimd: IGMPv3 leave not deleting group entry

After sending a IGMPv3 exclude report for a multicast address
with 0 sources, send an include report for the same group and also 0
sources.  This should cause IGMP to GS query and age/delete
the entry.

This fix addresses this issue.

Ticket: CM-11685
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

lib: Rewrite ipv4 address and prefix validator

Simplify ipv4 prefix and address matcher / validator to use standard
Linux networking functions instead of a state machine.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>

lib: Disallow % in IPv6 addresses and prefixes

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

lib: Rewrite ipv6 prefix matcher

Simplify ipv6 prefix matcher / validator to use standard Linux
networking functions instead of a state machine.

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Reviewed-by: Don Slice <dslice@cumulusnetworks.com>

all: add default log file if none are defined

Added a default log file named /var/log/quagga/Quagga.log to every daemon
to capture log entries if no log file is defined.  This also allows the
capture of logged information prior to reading each daemon's config file.
If a log file is defined manually, it will override this default file name.

Ticket: CM-10987
Signed-off-by: Don Slice
Reviewed By: Donald Sharp
Testing Done: Manual testing

pimd: static joins no longer worked

Static joins were killed by a previous commit, which
has been backed out.  I've recoded the igmp join
code to ignore 224.0.0.0/24 from ourselves a bit
differently now.

Ticket: CM-11751
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

Revert "pimd: Prevent igmp packet loopback."

This reverts commit 5bbcd1f0fd82e0e982700dbc12dd959629e91b58.

This commit badly breaks local 'ip igmp join XXX YYY' commands.

Revert this and fix the issue a slightly different way

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

bgpd: BGP should accept "router-id IFNAME"
ospfd: OSPF should accept "router-id IFNAME"

Added commands in BGP and OSPF where user can specify interface for router-id.

Ticket: CM-5040
Reviewed By: CCR-4908
Testing Done: Manual

bgpd: "neigbor <interface> ttl-security hops" should reject a hops value greater than 1
      "neighbor <interface> disable-connected-check" should not be allowed by the parser

Made changes to not allow hops greater than 1 and disable-connected check for neighbor <interface>

Ticket: CM-5536 CM-5537
Reviewed By: CCR-4865
Testing Done: Manual

pimd: Set default logfile to Quagga.log

Until such time a log file line is generated,
default to using /var/log/quagga/Quagga.log

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

pim: Show Iif for 'show ip pim upstream' command

Modify the 'show ip pim upstream' command to
show the expected incoming interface for
the source.

Ticket: CM-11220
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Fix igmp isex_excl

When we receive an igmpv3 group report that has
an implied *,g make sure that we look at
the source for the g and handle it appropriately

Ticket: CM-11396
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Pass the appropriate data structure around

Several static functions were passing a list around
when the reality is we are going to need to
look at the group information in order to make an
informated decision.

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

pimd: Ensure we should accept a kernel upcall

When we receive a multicast packet from a source
that is not connected to us, silently ignore it.

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

pimd: Prevent igmp packet loopback.

pim is joining the 224.0.0.13 and 224.0.0.22 groups
This is causing the creation of (*, 224.0.0.13) and
(*, 224.0.0.22) multicast routes which are immediately
sent to the pim network.

If we are the originator of the igmp report than
there is no need to accept the packet for
processing.

Ticket: CM-11397
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

pimd: Make pim_channel_add_oif failures debugs

Make failure messages of adding a mroute
a debug and wrapper them in PIM_DEBUG_MROUTE

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

pimd: Make received packet zlog_err a zlog_debug

When receiving packets and the parse fails
a zlog_err is generated.  This should be
protected by a debug.

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

pimd: Bind pim sockets to interface they are associated with

When pim is receiving packets, each interface's fd is receiving
packets for all interfaces.  Modify the code to bind the
pim interface sockets to the interface they were created for.

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

Merge branch 'cmaster-next' of ssh://stash.cumulusnetworks.com:7999/quag/quagga into cmaster-next

zebra/ospf/ospf6: Fix several memory leaks on if up/down

Resolved several memory leaks caused by ifdown/ifup the vrf device or
a swp port.  For bgp/zebra/ospf/ospf6, bouncing the vrf device would cause
a linked list, Interface, and route-table to get leaked.  For ospf6,
bouncing the swp device also caused leaks of Connected and Prefix entries.

Ticket: CM-10841
Signed-off-by: Don Slice
Reviewed-By: Donald Sharp
Testing Done: Manual testing, bgp and ospf mins passed, smokes had fewer failures than base

pimd: When receiving the register packet, create the mroute

When a RP receives the register packet, move the (*,G) to a
(S,G) route and send the pim message upstream to the source.

With this change the multicast packets are sent from the
source to the interested parties.

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

pimd: Refactor pim_scan_oil

Allow the workings of pim_scan_oil to be called
on a per channel basis.

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

pimd: Allow (*,G) to work properly.

When the kernel looks up a *,G route it
expects the incoming interface to be part
of the outgoing interface ttl list.

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

Provide example on how to run sudo vtysh show commands without password

Ticket: CM-4709
Reviewed By: dsharp
Testing Done: Uncommented, ran show commands

See the comments added, and in the bug.  Set up for users in group
quagga, with NOEXEC to not open security holes.

zebra: resolved problem with show ip route vrf

Repaired damage done by commit upstream, which changed the way show_ip_route
is called to allow for multicast rpf table display.  Matched the technique of
the other callers to the new function.

Ticket: CM-11345
Signed-off-by: Don Slice
Reviewed By: Donald Sharp
Testing Done: Manual testing and vrf-min

pimd: Fix register receive pointer arithmetic

When receiving the register packet from another pim
neighbor at the RP, we were adding an incorrect
amount of bytes to find the start of the ip_hdr
of the encapsulated data.  This commit fixes
this issue.

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

pimd: Removing extra asserts

in the igmp_group_free path, the normal deletion path
was causing asserts to happen.  There is no need
to have asserts here for this reason.

Further testing done w/ valgrind.

Ticket: CM-11401
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>

lib: Add documentation to internal command format parser functions

Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>

pimd: Fix igmp Trace logging

In general we don't need to know
when a read is scheduled.

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

pimd: Move timer information to TRACE_DETAIL

Move another timer informational debug to TRACE_DETAIL

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

pimd: Handle pimreg a bit better

The pimreg device is vif_index == 0.  Allow it
to be used.

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

pimd: Turn some trace information to trace detail

Turn some PIM_TRACE to PIM_TRACE_DETAIL.  Log file
was getting spammed with unneeded information.

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

pimd: Allow pimreg device to be output vif

With the change to moving the pimreg to vifindex 0,
when a packet is kicked up from the kernel to us
we need to install the route into the system.

This change allows us to do this now.

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

pimd: Get proper group on register receive

When we receive a encapsulated packet from another
pim neighbor, the code needs to grab the group
from the ip_hdr->src.

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

pimd: Fix checking to see if we are a RP or not

When we create the pim_ifp we need to see if this
interface allows us to elect our selves the RP.

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

pimd: Log Rotate is not working properly

The log line was causing confusion when we
were rotating quagga.  None of the other
protocols were logging lines on rotate
so just remove it.

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

vtysh: auto-generated vtysh_cmd.c file should depend on its creator

bgpd: Fix another clang warning

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