This is a full rewrite of the "back end" logging code. It now uses a
lock-free list to iterate over logging targets, and the targets
themselves are as lock-free as possible. (syslog() may have a hidden
internal mutex in the C library; the file/fd targets use a single
write() call which should ensure atomicity kernel-side.)
Note that some functionality is lost in this patch:
- Solaris printstack() backtraces are ditched (unlikely to come back)
- the `log-filter` machinery is gone (re-added in followup commit)
- `terminal monitor` is temporarily stubbed out. The old code had a
race condition with VTYs going away. It'll likely come back rewritten
and with vtysh support.
- The `zebra_ext_log` hook is gone. Instead, it's now much easier to
add a "proper" logging target.
v2: TLS buffer to get some actual performance
Signed-off-by: David Lamparter <equinox@diac24.net>
Zebra is currently sending messages on interface add/delete/update,
VRF add/delete, and interface address change - regardless of whether
its clients had requested them. This is problematic for lde and isis,
which only listens to label chunk messages, and only when it is
waiting for one (synchronous client). The effect is the that messages
accumulate on the lde synchronous message queue.
With this change:
- Zebra does not send unsolicited messages to synchronous clients.
- Synchronous clients send a ZEBRA_HELLO to zebra.
The ZEBRA_HELLO contains a new boolean field: sychronous.
- LDP and PIM have been updated to send a ZEBRA_HELLO for their
synchronous clients.
Signed-off-by: Karen Schoener <karen@voltanet.io>
LDP ordered label distribution control only binds a label to
a FEC if it is the egress LSR, or the router received a label
binding for a FEC from the next hop router. In this mode,
an MPLS router will create a label binding for each FEC and
distribute it to its neighbors so long as he has a entry in
the RIB for the destination.
Signed-off-by: Lynne Morrison <lynne@voltanet.io>
Signed-off-by: Karen Schoener <karen@voltanet.io>
As I understand it ldpd was originally developed as a standalone
daemon for *BSD land. Then ported to FRR. FRR uses ifindex_t
as the base type for the ifindex. Mixing `unsigned short` and
`int` and `unsigned int` is going to lead to fun somewhere
along the way. Especially when we get to run on a system
with ifindex churn( I'm looking at you docker ).
Attempt to convert all of ldpd to think of the ifindex as a
`ifindex_t`.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
GCC 10 switched to -fno-common by default, see
https://gcc.gnu.org/gcc-10/porting_to.html#common for details.
Fixes:
CCLD ldpd/ldpd
/usr/bin/ld: ldpd/libldp.a(adjacency.o):/home/ruben/src/frr/ldpd/ldpe.h:294: multiple definition of `pkt_ptr'; ldpd/ldpd.o:/home/ruben/src/frr/ldpd/ldpe.h:294: first defined here
Signed-off-by: Ruben Kerkhof <ruben@rubenkerkhof.com>
The lif variable was being set in the if statement and
immediately copied into from xf. No need to do this
twice.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Use the zapi_nexthop struct with the mpls_labels
zapi messages instead of the special-purpose (and
more limited) nexthop struct that was being used.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
The vrrpd one conflicts with the standalone vrrpd package; also we're
installing daemons to /usr/lib/frr on some systems so they're not on
PATH.
Signed-off-by: David Lamparter <equinox@diac24.net>
Validate that the FEC prefix length is within the allowed limit
(depending on the FEC address family) in order to prevent possible
buffer overflows.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
For all the places we have a zclient->interface_up convert
them to use the interface ifp_up callback instead.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Switch the zclient->interface_add functionality to have everyone
use the interface create callback in lib/if.c
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Start the conversion to allow zapi interface callbacks to be
controlled like vrf creation/destruction/change callbacks.
This will allow us to consolidate control into the interface.c
instead of having each daemon read the stream and react accordingly.
This will hopefully reduce a bunch of cut-n-paste stuff
Create 4 new callback functions that will be controlled by
lib/if.c
create -> A upper level protocol receives an interface creation event
The ifp is brand spanking newly created in the system.
up -> A upper level protocol receives a interface up event
This means the interface is up and ready to go.
down -> A upper level protocol receives a interface down
destroy -> A upper level protocol receives a destroy event
This means to delete the pointers associated with it.
At this point this is just boilerplate setup for future commits.
There is no new functionality.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
This new message makes it possible to install/reinstall LSPs with
multiple nexthops using a single ZAPI message.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
* Add ability to specify the nexthop type;
* Add ability to install or not a FTN (in addition to an LSP).
These two additions will be useful to install local SR Prefix-SIDs
configured with the no-PHP option.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Use the route type and instance instead of the route distance
to identify MPLS FTNs. This is a more robust approach since the
routing daemons can modify the distance of their announced routes
via configuration, which can cause inconsistencies.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Do this for the following reasons:
* Improve modularity of the code by separating the decoding of the
ZAPI messages from their processing;
* Create an API that is easier to use by the client daemons.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
LDP opens two sockets to zebra, one through ldpd (always using
instance 0) and another through lde (using whatever instance
was set through the -n command line parameter). If no instance
was set, both connections would use the same protocol and instance,
making it impossible to distinguish them through zserv_find_client.
This meant that a response to a lm connect would erroneously go to
the wrong process. Fix this by having a default instance value of 1,
in case the user does not specify a different one.
Signed-off-by: Emanuele Di Pascale <emanuele@voltanet.io>
For SRGB, we need to support chunk requests starting at a
specific point in the label space, rather than just asking
for any sufficiently large chunk. To this purpose, we extend
the label manager api to request a chunk with a base value;
if the base is set to 0, the label manager will behave as it
currently does, i.e. fetching the first free chunk big enough
to satisfy the request.
update all the existing calls to get chunks from the label
manager so that they use MPLS_LABEL_BASE_ANY as the base
for the requested chunk
Signed-off-by: Emanuele Di Pascale <emanuele@voltanet.io>
If --ctl_socket is used this will override any other option and will
be used
If -N <namespace> is used, then we will setup the LDPD_SOCKET
in $frr_statedir/<namespace>/ldpd.sock
If neither option is used, then we will use $frr_statedir/ldpd.sock
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
This macro:
- Marks ZAPI callbacks for readability
- Standardizes argument names
- Makes it simple to add ZAPI arguments in the future
- Ensures proper types
- Looks better
- Shortens function declarations
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
- some target_CFLAGS that needed to include AM_CFLAGS didn't do so
- libyang/sysrepo/sqlite3/confd CFLAGS + LIBS weren't used at all
- consistently use $(FOO_CFLAGS) instead of @FOO_CFLAGS@
- 2 dependencies were missing for clippy
Signed-off-by: David Lamparter <equinox@diac24.net>
The calculation to know when an LDP PDU went past the maximum
negotiated PDU length was wrong because it wasn't taking the
"Version" and "PDU Length" fields into account (total of four
bytes). Fix this.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
We must open the PF_KEY socket before dropping privileges, otherwise the
socket creation will fail with permission problems.
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
A while ago all FRR configuration commands were converted to use the
QOBJ infrastructure to keep track of configuration objects. This
means the configuration lock isn't necessary anymore because the
QOBJ code detects when someones tries to edit a configuration object
that was deleted and react accordingly (log an error and abort the
command). The possibility of accessing dangling pointers doesn't
exist anymore since vty->index was removed.
Summary of the changes:
* remove the configuration lock and the vty_config_lockless() function.
* rename vty_config_unlock() to vty_config_exit() since we need to
clean up a few things when exiting from the configuration mode.
* rename vty_config_lock() to vty_config_enter() to remove code
duplication that existed between the three different "configuration"
commands (terminal, private and exclusive).
Configuration commands converted to the new northbound model don't
need the configuration lock either since the northbound API also
detects when someone tries to edit a configuration object that
doesn't exist anymore.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
It's been a year since we added the new optional parameters
to instantiation. Let's switch over to the new name.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
FRR_DAEMON_INFO should now contain an array of 'frr_yang_module_info'
structures describing the YANG modules implemented by the daemon.
This array will be used by frr_init() function to load all YANG modules
and initialize the northbound callbacks during the daemon initialization.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Corrections so that the BGP daemon can work with the label manager properly
through a label-manager proxy. Details:
- Correction so the BGP daemon behind a proxy label manager gets the range
correctly (-I added to the BGP daemon, to set the daemon instance id)
- For the BGP case, added an asynchronous label manager connect command so
the labels get recycled in case of a BGP daemon reconnection. With this,
BGPd and LDPd would behave similarly.
Signed-off-by: F. Aragon <paco@voltanet.io>
config.h (or, transitively, zebra.h) must be the first include file
listed for autoconf things like _GNU_SOURCE and _POSIX_C_SOURCE to work
correctly.
Signed-off-by: David Lamparter <equinox@diac24.net>
Since we're now building through one large Makefile, we can easily put
things with their daemons and crossreference nicely.
Signed-off-by: David Lamparter <equinox@diac24.net>
Add a TAILQ_POP_FIRST so Clang understands it's the same item that is
getting removed from the list.
Signed-off-by: David Lamparter <equinox@diac24.net>
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>
All ldpd commands were written using DEFPY except the ones that needed
to be ignored by vtysh, in which case we were using DEFUN_NOSH. Now that
DEFPY_NOSH is available, convert the remaining commands for consistency
and to simplify the code a little bit.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
This commit fixes two issues:
- memcpy() using containers of different sizes when using addr2sa(), mixing
'struct sockaddr_storage' and 'union sockunion'.
- addr2sa() function not being thread safe (using a local static variable as
container.
Signed-off-by: F. Aragon <paco@voltanet.io>
With commit e94b38d94b5 we are now scheduling the read
of vty config until after the startup of main thread
processing. It now becomes necessary to move the
application of the config until after the read
in of the config from a file if we are using a
non-integrated config.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The following types are nonstandard:
- u_char
- u_short
- u_int
- u_long
- u_int8_t
- u_int16_t
- u_int32_t
Replace them with the C99 standard types:
- uint8_t
- unsigned short
- unsigned int
- unsigned long
- uint8_t
- uint16_t
- uint32_t
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
Turns out we had 3 different ways to define labels
all of them overlapping with the same meanings.
Consolidate to 1. This one choosen is consistent
naming wise with what the *bsd and linux kernels
use.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When compiling ldpd on a mac, there exists a #define MSG_SEND
which conflicts with a define in ldp_debug.h.
During discussion about this we decided that it would be
better to remove the macro massaging that was going on and
to just call our own #define for it.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
ldpd should ignore blackhole routes and any other route that doesn't
have a nexthop address (connected routes being an exception).
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
In some circumstances zebra and ldpd would display a pseudowire as UP
when in reality it's not (example: MTU mismatch between the two ends). Fix
this to avoid confusion.
Reported-by: ßingen <bingen@voltanet.io>
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Allow the higher level protocol to specify if it would
like to receive notifications about it's routes that
it has installed.
I've purposely made it part of zclient_new_notify because
we need to track the routes on a per daemon basis only.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
With non-targeted LDP receiving a PW label mapping before configuring
the PW was causing the SET message to be sent before the ADD one, so
Zebra PW manager wouldn't find the PW on first message reception.
Signed-off-by: ßingen <bingen@voltanet.io>
The original clippy implementation used if VTYSH_EXTRACT_PL
was defined or not to know to include as part of the creation
of vtysh.
Follow their lead here.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
This problems happens because, in this port, whenever the child
processes want to log something they send a message to the parent. But
in the shutdown functions the first thing we do is to close the pipes
to the parent process. With that said, add some protections to prevent
the child processes from trying to use a closed pipe and just ignore
their log messages during shutdown. In the future we need to share
the logging configuration with the child processes so they can send
log messages on their own.
While here, remove some unnecessary calls to msgbuf_write() in
ldpe_shutdown().
Fixes#1253.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
This improves code readability and also future-proofs our codebase
against new changes in the data structure used to store interfaces.
The FOR_ALL_INTERFACES_ADDRESSES macro was also moved to lib/ but
for now only babeld is using it.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Performance tests showed that, when running on a system with a large
number of interfaces, some daemons would spend a considerable amount
of time in the if_lookup_by_index() function. Introduce a new rb-tree
to solve this problem.
With this change, we need to use the if_set_index() function whenever
we want to change the ifindex of an interface. This is necessary to
ensure that the 'ifaces_by_index' rb-tree is updated accordingly. The
return value of all insert/remove operations in the interface rb-trees
is checked to ensure that an error is logged if a corruption is
detected.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
IFINDEX_DELETED is not necessary anymore as we moved from a global
list of interfaces to a list of interfaces per VRF.
This reverts commit 84361d615.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
This is an important optimization for users running FRR on systems with
a large number of interfaces (e.g. thousands of tunnels). Red-black
trees scale much better than sorted linked-lists and also store the
elements in an ordered way (contrary to hash tables).
This is a big patch but the interesting bits are all in lib/if.[ch].
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
size is not used for further parsing. Keep it updated but tell
to the compiler that we know it is not used just in case one
needs to extend the parsing somedays.
Signed-off-by: Vincent Jardin <vincent.jardin@6wind.com>
Certain compilers cannot recognize that rt is
actually being init'ed, but let's set it to
NULL 'till we get them updated.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
