As part of normal processing we allow bgp commands to walk
up the command node chain. We are experiencing this crash:
Thread 1 "bgpd" received signal SIGABRT, Aborted.
__GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:50
50 ../sysdeps/unix/sysv/linux/raise.c: No such file or directory.
(gdb) bt
assertion=0x7ffff7f3ba4f "set", file=0x7ffff7f3ba44 "lib/yang.c", line=413, function=<optimized out>)
at assert.c:92
line=413, function=0x7ffff7f3bc50 <__PRETTY_FUNCTION__.9> "yang_dnode_get") at assert.c:101
vty=0x5555561715a0, argc=3, argv=0x555558601620) at bgpd/bgp_vty.c:9568
cmd=0x0) at lib/command.c:937
at lib/command.c:997
matched=0x0, vtysh=0) at lib/command.c:1161
at lib/vty.c:517
(gdb)
9582 bgp_glb_dnode = yang_dnode_get(vty->candidate_config->dnode,
(gdb) p vty->xpath
$8 = {
"/frr-routing:routing/control-plane-protocols/control-plane-protocol[type='frr-bgp:bgp'][name='bgp'][vrf='default']/frr-bgp:bgp", '\000' <repeats 897 times>, '\000' <repeats 1023 times>, '\000' <repeats 1023 times>,
'\000' <repeats 1023 times>, '\000' <repeats 1023 times>, '\000' <repeats 1023 times>, '\000' <repeats 1023 times>,
'\000' <repeats 1023 times>}
(gdb) p vty->xpath_index
$9 = 0
(gdb)
We are effectively sending in an array index based upon vty->xpath_index( which is zero) but
the VTY_CURR_XPATH macro subtracts 1 from that value to find the appropriate xpath to use.
This of course subtracts 1 from 0 and we underflow the array.
The relevant section in a config file is this:
address-family ipv6 flowspec
bgp maxim...
Effectively we were trying to walk up the command chain for flowspec to see
if the command is entered correctly. There is a function vty_check_node_for_xpath_decrement
that was looking at bgp sub-modes to make the decision to allow us to decrement
the vty->xpath_index which did not have the v4 or v6 flowspec bgp sub modes in the
check.
Adding them in fixes the problem.
Signed-off-by: Donald Sharp <sharpd@nvidia.com>
In transactional cli mode, bgp address-family <afi> <afi>
node builds xpath on top of `router bgp` node's xpath.
When `exit` is applied under afi-safi commands, retain
xpath_index to 1 to keep using bgp global xpath.
Signed-off-by: Chirag Shah <chirag@nvidia.com>
When using the default CLI mode, the northbound layer needs to create
a separate transaction to process each YANG-modeled command since
they are supposed to be applied immediately (there's no candidate
configuration nor the "commit" command like in the transactional
CLI). The problem is that configuration transactions have an overhead
associated to them, in big part because of the use of some heavy
libyang functions like `lyd_validate()` and `lyd_diff()`. As of
now this overhead is substantial and doesn't scale well when large
numbers of transactions need to be performed in sequence.
As an example, loading 50k prefix-lists using a single transaction
takes about 2 seconds on a modern CPU. Loading the same 50k
prefix-lists using 50k transactions can take more than an hour
to complete (which is unacceptable by any standard). To fix this
problem, some heavy optimization work needs to be done on libyang and
on the FRR northbound itself too (e.g. perform partial configuration
diffs whenever possible). This, however, should be a long term
effort since these optimizations shouldn't be trivial to implement
and we're far from having the performance numbers we need.
In the meanwhile, this commit introduces a simple but efficient
workaround to alleviate the issue. In short, a new back-off timer
was introduced in the CLI to monitor and detect when too many
YANG-modeled commands are being received at the same time. When
a certain threshold is reached (100 YANG-modeled commands within
one second), the northbound starts to group all subsequent commands
into a single large transaction, which allows them to be processed
much faster (e.g. seconds and not hours). It's essentially a
protection mechanism that creates dynamically-sized transactions
when necessary to prevent performance issues from happening. This
mechanism is enabled both when parsing configuration files and when
reading commands from a terminal.
The downside of this optimization is that, if several YANG-modeled
commands are grouped into the same transaction and at least one of
them fails, the whole transaction is rejected. This is undesirable
since users don't expect transactional behavior when that's not
enabled explicitly. To minimize this issue, the CLI will log all
commands that were rejected whenever that happens, to make the
user aware of what happened and have enough information to fix
the problem. Commands that fail due to parsing errors or CLI-level
validations in general are rejected separately.
Again, this proposed workaround is intended to be temporary. The
goal is to provided a quick fix to issues like #6658 while we work
on better long-term solutions.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
rpki vrf subnode is instantiated under the vrf subnode.
It it to be noted that this commit contains a change in vtysh.
Actually, the output of bgp daemon from show running-config is extracted
in vtysh, and reengineered ( hence the vtysh_config.c change done). This
permits having a subnode under vrf sub node.
Also, add vrf node support to bgpd, as rpki command can not be found
under vrf node.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
A new config option `--disable-version-build-config`
allows you to show short version string by dropping
"configured with:" and all of its build configs
Signed-off-by: Jafar Al-Gharaibeh <jafar@atcorp.com>
This reverts commit d741915ecd.
This is because it breaks this behavior:
router ospf6
<commands>
!
int enp39s0
<more commands>
!
This is a very legal set of commands and completely destroys the
ability to do this.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Replace all `random()` calls with a function called `frr_weak_random()`
and make it clear that it is only supposed to be used for weak random
applications.
Use the annotation described by the Coverity Scan documentation to
ignore `random()` call warnings.
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
Rather than doing a f*gly hack for the RPKI code, let's do an on-exit
hook in cmd_node. Also allows replacing some special-casing in the vty
code.
Signed-off-by: David Lamparter <equinox@diac24.net>
And again for the name. Why on earth would we centralize this, just so
people can forget to update it?
Signed-off-by: David Lamparter <equinox@diac24.net>
Same as before, instead of shoving this into a big central list we can
just put the parent node in cmd_node.
Signed-off-by: David Lamparter <equinox@diac24.net>
There is really no reason to not put this in the cmd_node.
And while we're add it, rename from pointless ".func" to ".config_write".
[v2: fix forgotten ldpd config_write]
Signed-off-by: David Lamparter <equinox@diac24.net>
The only nodes that have this as 0 don't have a "->func" anyway, so the
entire thing is really just pointless.
Signed-off-by: David Lamparter <equinox@diac24.net>
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>
Since we've been writing out "frr version" and "frr defaults" for about
a year and a half now, we can now actually use them to manage defaults.
Signed-off-by: David Lamparter <equinox@diac24.net>
This allows to set motd from an input instead of creating a file.
Example:
root@exit2-debian-9:~/frr# telnet 127.0.0.1 2605
Trying 127.0.0.1...
Connected to 127.0.0.1.
Escape character is '^]'.
Hello, this is bgpd
User Access Verification
Password:
exit2-debian-9> enable
exit2-debian-9# sh run
Current configuration:
!
frr version 7.3-dev-MyOwnFRRVersion
frr defaults traditional
!
hostname exit2-debian-9
password belekas
log file /var/log/frr/labas.log
log syslog informational
banner motd line Hello, this is bgpd
!
!
!
line vty
!
end
exit2-debian-9#
Signed-off-by: Donatas Abraitis <donatas.abraitis@gmail.com>
nb_candidate_edit() was calling both the lyd_schema_sort() and
lyd_validate() functions whenever a new node was added to the
candidate configuration. This was done to ensure the candidate
is always ready to be displayed correctly (libyang only creates
default child nodes during the validation process, and data nodes
aren't guaranteed to be ordered by default).
The problem is that the two aforementioned functions are too
expensive to be called in the northbound hot path. Instead, it makes
more sense to call them only before displaying the configuration
(in which case a recursive sort needs to be done). Introduce the
nb_cli_show_config_prepare() to achieve that purpose.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Adding a lock to protect the global running configuration doesn't
help much since the FRR daemons are not prepared to process
configuration changes in a pthread that is not the main one (a
whole lot of new protections would be necessary to prevent race
conditions).
This means the lock added by commit 83981138 only adds more
complexity for no benefit. Remove it now to simplify the code.
All northbound clients, including the gRPC one, should either run
in the main pthread or use synchronization primitives to process
configuration transactions in the main pthread.
This reverts commit 83981138fe.
Commit eaf6705d7a fixed a problem caused by configuration changes
coming from the kernel. The fix consisted of regenerating the
candidate configuration before every configuration command (when
using the non-transactional CLI mode). There's no need, however,
to regenerate the candidate when it's identical to the running
configuration. Since the northbound keeps track of the version
of each configuration, we can use that information to prevent
regenerating the candidate configuration when that is not necessary.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
This implements BMP. There's no fine-grained history here, the non-BMP
preparations are already split out from here so all that remains is BMP
proper.
Signed-off-by: David Lamparter <equinox@diac24.net>
The correct cast for these is (unsigned char), because "char" could be
signed and thus have some negative value. isalpha & co. expect an int
arg that is positive, i.e. 0-255. So we need to cast to (unsigned char)
when calling any of these.
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>
Even when using the classic CLI mode (i.e. when --tcli is not
used), the northbound code still uses vty->candidate_config
to perform configuration changes. From the perspective of the
user, the running configuration is being edited directly, but
under the hood the northbound layer does a full configuration
transaction for each command. When the running configuration is
edited by a northbound client other than the CLI (e.g. kernel,
gRPC), vty->candidate_config might become outdated, and this can
lead to lots of weird problems. To fix this, always regenerate
vty->candidate_config before each configuration command when
using the classic CLI mode. When using the transactional CLI,
the user needs to update the candidate manually using the "update"
command, otherwise the "commit" command will fail with this error:
"% Candidate configuration needs to be updated before commit".
Fixes some problems reported by Don after moving an interface from
one VRF to another one while zebra is running.
Reported-by: Don Slice <dslice@cumulusnetworks.com>
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Added a CLI "debug route-map" to enble route-map debugs
Added debugs for following triggers
1. Add/delete a route-map
2. Add/delete a sequence in route-map
3. Add/delete a match statement(dependency)
4. Update a dependency
5. Apply a route-map
Signed-off-by: Ameya Dharkar <adharkar@vmware.com>
Although the RFC states hostname length should be < 255 chars,
FRR allows infinite length technically. However, when you try
to set a hostname > 80 chars, you would immediately notice a crash.
RCA: Crash due to buffer overflow. Large buffer sprintf'd into smaller
buffer. Usage of sprintf function instead of snprintf which is safer.
Signed-off-by: Lakshman Krishnamoorthy <lkrishnamoor@vmware.com>
It doesn't make much sense for a hash function to modify its argument,
so const the hash input.
BGP does it in a couple places, those cast away the const. Not great but
not any worse than it was.
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
The upcoming gRPC-based northbound plugin will run on a separate
pthread, and it will need to have access to the running configuration
global variable. Introduce a rw-lock to control concurrent access
to the running configuration. Add the lock inside the "nb_config"
structure so that it can be used to protect candidate configurations
as well (this might be necessary depending on the threading scheme
of future northbound plugins).
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
