Quentin Young [Tue, 6 Mar 2018 23:08:37 +0000 (18:08 -0500)]
zebra: reorganize zserv, batch i/o
Group send and receive functions together, change handlers to take a
message instead of looking at ->ibuf and ->obuf, allow zebra to read
multiple packets off the wire at a time.
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
Quentin Young [Tue, 6 Mar 2018 22:57:33 +0000 (17:57 -0500)]
zebra: standardize ZAPI message handler args
A lot of the handler functions that are called directly from the ZAPI
input processing code take different argument sets where they don't need
to. These functions are called from only one place and all have the same
fundamental information available to them to do their work. There is no
need to specialize what information is passed to them; it is cleaner and
easier to understand when they all accept the same base set of
information and extract what they need inline.
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
Quentin Young [Tue, 6 Mar 2018 22:01:42 +0000 (17:01 -0500)]
zebra: dont return a status code in zapi handlers
All of the ZAPI message handlers return an integer that means different
things to each of them, but nobody ever reads these integers, so this is
technical debt that we can just eliminate outright.
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
Quentin Young [Fri, 9 Mar 2018 17:07:25 +0000 (12:07 -0500)]
lib: add convenience debugging macros
* Add DEBUG*() macros
This set of macros allows you to write printf-like debugging lines that
automatically check whether a debug is on before printing. This should
eliminate the need for explicit checks in simple cases. For example:
if (SUCH_AND_SUCH_DEBUG_IS_ON) {
zlog_warn(...);
}
Becomes:
DEBUG(warn, such_and_such, ...);
Or, equivalently,
DEBUGE(such_and_such, ...);
The levels passed to DEBUG are expanded into the names of zlog_*
functions, so the same zlog levels are available. There's also a set of
macros that have the level built into them; DEBUGE for errors, DEBUGW
for warnings, etc. Good for brevity.
* Add singular setting macros
Change the 'SET' macros to accept a boolean indicating whether the
provided bits should be set or unset, and map on/off macros to them.
Helps condense code where you already have a boolean condition that
tells you what you want to do as you can avoid writing the branch.
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
Arthur Jones [Sun, 11 Mar 2018 17:32:24 +0000 (10:32 -0700)]
alpine: minimal alpine packaging
For building dev packages for alpine, we provide a minimal APKBUILD
file and add a configure option for only numeric versions in the
VERSION variable as alpine does not allow non-numeric characters
in the version string.
These changes allow alpine to be built, but don't yet provide a
mechanism to build. Changes to do the build in docker are coming
soon...
Testing done:
Built alpine packages in local docker environment, packages
showed no "dev" in the package name. Also built CentOS packages
with numeric version disabled and the "dev" is still in the package
name.
Issue: https://github.com/FRRouting/frr/issues/1859 Signed-off-by: Arthur Jones <arthur.jones@riverbed.com>
Arthur Jones [Fri, 9 Mar 2018 23:08:24 +0000 (15:08 -0800)]
zebra/if_netlink: compile under musl-libc
musl-libc is a lightweight libc used by alpine linux:
https://www.musl-libc.org/
AFAICT, this is the only change to the source needed to get
basic frr support compiling on musl.
Two changes in one patch, get ethhdr from netinet/if_ether.h
and replace the only __caddr_t I could find in the source base
with caddr_t.
Testing done:
Compiled apk packages using a docker environment (patches
coming soon) also compiled redhat and debian using a similar
docker environment (RFC patches for those changes are queued
up too)...
Issue: https://github.com/FRRouting/frr/issues/1859 Signed-off-by: Arthur Jones <arthur.jones@riverbed.com>
Donald Sharp [Fri, 23 Feb 2018 18:45:36 +0000 (13:45 -0500)]
zebra: Make the ifp part of the rule structure
Every place we need to pass around the rule structure
we need to pass around the ifp as well. Move it into
the structure. This will also allow us to notify up
to higher level protocols that this worked properly
or not better too.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Donald Sharp [Fri, 16 Feb 2018 01:11:12 +0000 (20:11 -0500)]
zebra: Cleanup api
Allow the add/delete to go through a intermediary function in
zebra_pbr.c instead of directly to the underlying os call. This
will allow future refinements to track the data a bit better
so that on shutdown we can delete the rules.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
vivek [Sun, 4 Feb 2018 12:33:33 +0000 (12:33 +0000)]
*: PBR - netlink interaction and basic definitions
Implement netlink interactions for Policy Based Routing. This includes
APIs to install and uninstall rules and handle notifications from the
kernel related to rule addition or deletion. Various definitions are
added to facilitate this.
Signed-off-by: Vivek Venkatraman <vivek@cumulusnetworks.com> Reviewed-by: Donald Sharp <sharpd@cumulusnetworks.com>
Donald Sharp [Mon, 5 Feb 2018 15:40:09 +0000 (10:40 -0500)]
bgpd, lib, zebra: Switch to work_queue_free_and_null
The work_queue_free function free'd up the wq pointer but
did not set it too NULL. This of course causes situations
where we may use the work_queue after it is freed. Let's
modify the work_queue to set the pointer for you.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Donald Sharp [Tue, 6 Mar 2018 21:07:14 +0000 (16:07 -0500)]
lib: Increase zapi buffer size
The buffer size is currently 4k. Increase x4 times to allow for bigger
messages to be sent over the zapi.
The current size sufficient for most cases, but there are a couple
of cases with installing data to the kernel ip rules where we will
quickly hit this 4k size limit. I forsee flowspec getting close
to this limit as well.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Quentin Young [Tue, 6 Mar 2018 18:42:53 +0000 (13:42 -0500)]
ospfd: cancel SR thread at shutdown
Otherwise if it is scheduled the thread pointer will be accessed after
the shutdown task finishes accessing, having deleted the structure that
owns said pointer, which causes a heap UAF.
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
The NS_DEFAULT value returns UNKNOWN in the case the vrf lite backend is
used, whereas this is wrong. This commit fixes the default value.
Also, it fixes the default value in the case NETNS support from system
is not ok, or some error can occur when reading default NS at startup.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
bgpd: permit bgp vrf socket creation on some cases
When VRF is not yet available at startup, the check for main socket
presence must be done. As the main socket creation is made in a separate
place from vrf socket for netns, ths main socket creation must not be
prevented when a BGP VRF relies on vrf lite mechanism.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
Quentin Young [Mon, 5 Mar 2018 18:20:22 +0000 (13:20 -0500)]
tools: add LeakSanitizer suppressions list
Building FRR with AddressSanitizer is kind of annoying since
libpython3.5 leaks memory, clippy links libpython3.5 and clippy runs as
part of the build process. LeakSanitizer has a way to suppress leaks at
runtime by setting the LSAN_OPTIONS environment variable to contain a
file path to a suppression list:
LSAN_OPTIONS=suppressions=path/to/suppr.txt
This commit provides the file. Setting this environment variable to
before building should allow a clean build with ASAN enabled. The
relative path is there because LeakSanitizer looks at paths relative to
the binary it is sanitizing; clippy is in lib/ so the path is set
relative to lib/.
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
This capability, when used, is mapped over linux sys_admin capability.
This is necessary from the daemon perspective, in order to handle NETNS
based VRFs, because calling setns() requires sys admin capability.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
Chirag Shah [Tue, 27 Feb 2018 19:24:16 +0000 (11:24 -0800)]
ospf6d: assign zebra router-id to ospf6 instance
Store zebra router-id in global structure.
Before router ospf6 instance created,
zebra router-id callback called.
During ospf6 main execution zebra init happens,
but default instance does not execute until
cli replay 'router ospf6'.
Call ospf6_router_id_change during 'router ospf6'
to assign zebra router id to ospf6 instance.
Quentin Young [Wed, 28 Feb 2018 21:14:45 +0000 (16:14 -0500)]
lib: add mt-safe debugging facilities
The current strategy for fine-grained debugging across FRR is to use
static long int bitfields, in combination with helper macros that are
copy-pasted between daemons, to hold state on what debugging information
should be collected at any given time. This has a couple of problems:
* These bitfields are generally extern'd and accessed everywhere, so
they are not MT-safe or easy to make MT-safe
* Lots of code duplication from copy-pasting the DEBUG_* macros...
* Code duplication because of the "term" vs "conf" debugging concept
This patch aims to remedy that by providing some infrastructure to work
with debugs. The core concept of using bitfields has been retained, but
the number of these for each debug has been reduced to 1. This allows
easy use of lock-free methods for synchronizing access to debugging
info.
The helper macros have also been retained but they are now collected in
one place and perform exclusively atomic operations.
Finally there is a bit of code that allows daemons to register
callbacks, which I used to implement a command that will toggle all
debugging for any daemons that use these facilities.
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
Chirag Shah [Thu, 15 Feb 2018 03:02:11 +0000 (19:02 -0800)]
ospf6d: ECMP for external routes
The route being added check its origin matches,
with any of the existing path (list of paths).
Remove the existing path, add if its cost is
eqaual or less than any of the existing path.
For a given route and of existing path cost is lower
(better) than one being added, discard new route update.
The existing path cost is higher (lower) than one being
added, ospf6_route_add replaces existing with new route
info.
Compare cost between delete request and with existing
route.
Ticket:CM-16139
Signed-off-by: Chirag Shah <chirag@cumulusnetworks.com>
Donald Sharp [Thu, 1 Mar 2018 00:22:47 +0000 (19:22 -0500)]
lib, zebra: Add type and instance to nexthop update message
Add the originating routes type and instance to the nexthop
update message. This is necessary because there exist
scenarios where BGP needs to make a decision about the
originating route type and instance to know if it is
going to be doing a route replace to a route that would
resolve to itself.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Donald Sharp [Thu, 1 Mar 2018 00:16:51 +0000 (19:16 -0500)]
ospfd: Add vrf to nhlfe route add
So when a ospf SR is sending down routes to the kernel
ensure that the nexthop vrf_id is set appropriately.
Yes SR is in the default VRF. But for people who
run across this code in the future, they will know
to do the right thing from it.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
projects / mirror / frr.git / log