Donald Sharp [Wed, 16 Oct 2019 17:19:09 +0000 (13:19 -0400)]
*: Allow disabling of `getrusage` calls
getrusage, in a heavily stressed system, can account for
signficant running time due to process switching to the kernel.
Allow the end-operator to specify `--disable-cpu-time` to
avoid this call. Additionally we cause `show thread cpu` to
not show up if this is selected.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Mark Stapp [Wed, 16 Oct 2019 12:51:43 +0000 (08:51 -0400)]
nhrpd: be more careful with linked lists
NHRPD has its own linked-list implementation, and one of the
apis is a little free and easy with pointers. Also be safer
with one list iteration operation.
GalaxyGorilla [Tue, 15 Oct 2019 11:15:22 +0000 (11:15 +0000)]
lib: Let libyang log everything possible
Currently libyang logs errors only (LY_LLERR by default), independent of
FRR's log level. This commit lets libyang log everything including all
sorts of debug logs (when libyang is built in 'Debug' mode). FRR's
logging infrastructure filters logs out according to the configured log
level.
There is a very small performance overhead involved, even when libyang
is build in 'Release' mode. This overhead is mainly affecting config
processing and barely measurable being around 0-3% of the processing
time without this change.
Stephen Worley [Tue, 15 Oct 2019 19:39:49 +0000 (15:39 -0400)]
zebra: Don't bother ref'ing ifp in zebra_pbr_rule
If we only really use the ifp for the name, then
don't bother referencing the ifp. If that ifp is
freed, we don't expect zebra to handle the rules that
use it (that's pbrd's job), so it is going to be
pointing to unintialized memory when we decide to remove
that rule later. Thus, just keep the name in the data
and dont mess with pointer refs.
Signed-off-by: Stephen Worley <sworley@cumulusnetworks.com>
Stephen Worley [Tue, 15 Oct 2019 18:50:10 +0000 (14:50 -0400)]
zebra: Use the rule ifindex as a hash key, not ifp
Use the ifindex value as a primary hash key/identifier, not
the ifp pointer. It is possible for that data to be freed
and then we would not be able to hash and find the rule entry
anymore. Using the ifindex, we can still find the rule even
if the interface is removed.
Signed-off-by: Stephen Worley <sworley@cumulusnetworks.com>
Stephen Worley [Tue, 15 Oct 2019 17:42:15 +0000 (13:42 -0400)]
zebra: Only free if rule was found/release in table
We were seeing a double free on shutdown if the
hash release fails here due to the interface state
changing. We probably shouldn't free the data if its
still being handled in the table so adding a check there
and a debug message.
Signed-off-by: Stephen Worley <sworley@cumulusnetworks.com>
Zebra tracked the last command sent to it's peer via peeking
into the data it was sending to each client ( since we had
lost the idea of what the command was when it was time to track
the data ).
Add a define to track this and add a bit of verbiage
to the code to allow us to notice when we screw with
the header again so that this is just fixed correctly
when it happens again.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Donald Sharp [Tue, 15 Oct 2019 12:27:22 +0000 (08:27 -0400)]
bgpd: Soft reconfig-in should find the right bgp_path_info
When using soft reconfiguration inbound we are storing packet
data on the side for replaying when necessary. The problem here
is that we are just grabbing the first bgp_path_info and using
that as the base. What happens when we have soft-reconfig turned
on with multiple bgp_path_info's for a path? This was introduced
in commit 8692c506520f6b268525b80890702432c95f13c4, yes back
in 2012! I would argue, though, that it was just broken
in a different way before this.
Choose the correct bgp_path_info that corresponds to the peer
we received the data from for rethinking.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Spantik [Mon, 14 Oct 2019 12:17:29 +0000 (05:17 -0700)]
lib: Clean up the index that is being used for storing FD and events.
When POLLNVAL is received for a FD then that FD is removed from the
pfd array and also array is rearranged using memmove. When memmove
is used then unused index are not cleanedup. When a new FD takes
up that index then it ends up using stale events without any handler
set for the same.
Renato Westphal [Tue, 15 Oct 2019 01:20:24 +0000 (22:20 -0300)]
lib: fix harmless lyd_schema_sort() warning
The dnode member of the nb_config structure can be null on
daemons that don't implement any YANG module. As such, update
the nb_cli_show_config_prepare() function to always check if the
libyang data node that is going to be displayed is null or not
before operating on it.
This fixes the following warning (introduced by commit 5e6a9350c1):
libyang: Invalid arguments (lyd_schema_sort())
Reported-by: Donald Sharp <sharpd@cumulusnetworks.com> Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Donald Sharp [Tue, 15 Oct 2019 01:09:55 +0000 (21:09 -0400)]
bgpd: Be careful about displaying vni's as labels.
When a type 2/3 or 5 route is received, verified and the
resulting route generated is pushed into the appropriate vrf
the vni's associated with the route are also passed in.
This is showing up as a Remote label when you dump
the route in bgp:
BGP routing table entry for 0.0.0.0/0^M
Paths: (1 available, best #1, table third)
Advertised to non peer-group peers:
10.10.120.22
42001 42005 42006 42055
10.10.120.22 from 10.10.120.22 (10.10.255.193)
Origin IGP, valid, external, bestpath-from-AS 42001, best
Remote label: 62750
AddPath ID: RX 0, TX 2
Last update: Fri Oct 11 12:59:56 2019
The `Remote label: 62750` is the mpls label version of the
vni passed in. This is meaningless and confusing to the end
user. Do not display this information.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Donald Sharp [Tue, 15 Oct 2019 00:44:56 +0000 (20:44 -0400)]
bgpd: Properly lock parent node for type4 routes
When creating a bgp_path_info for a type 4 route the pi->extra->parent
and the route node for the originating table were not being locked
properly. This will prevent BGP from not properly cleaning up
the data structures on cleanup.
Possibly every one of the functions that we use to create the
new bgp_path_info's should use an abstracted version of this code,
but I am unsure at this point in time if a type 4 should use the same
or not.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Quentin Young [Mon, 14 Oct 2019 16:09:36 +0000 (16:09 +0000)]
bgpd: move assert out of error case
bgp_process_packets has an assert to make sure an appropriate amount of
working space in the input buffer has been freed up for future reads.
However, this assert shouldn't be made when we have encountered an error
that's going to tear down the session, because in this case we may not
be able to process the full contents of the input buffer.
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
Quentin Young [Mon, 14 Oct 2019 16:09:36 +0000 (16:09 +0000)]
bgpd: move assert out of error case
bgp_process_packets has an assert to make sure an appropriate amount of
working space in the input buffer has been freed up for future reads.
However, this assert shouldn't be made when we have encountered an error
that's going to tear down the session, because in this case we may not
be able to process the full contents of the input buffer.
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
Rafael Zalamena [Fri, 11 Oct 2019 23:15:46 +0000 (20:15 -0300)]
lib: use `prefix` for yang get prefix wrapper
This change fixes a static analyzer warning and should also make us
safer when using this function. At the moment the code that triggered
the warning is the only one that uses this function.
Passing anything other than `struct prefix` to `str2prefix` function is
dangerous, because the structure might be smaller than expected and we
might have an buffer overflow.
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
Renato Westphal [Thu, 10 Oct 2019 01:20:26 +0000 (22:20 -0300)]
lib: reduce memory allocation when processing large config transactions
Remove the xpath field from the nb_config_cb structure in order
to reduce its size. This allows the northbound to spend less time
allocating memory during the processing of large configuration
transactions.
To make this work, use yang_dnode_get_path() to obtain the xpath
from the dnode field whenever necessary.
Renato Westphal [Thu, 10 Oct 2019 02:24:04 +0000 (23:24 -0300)]
lib: fix processing of the 'apply_finish' callbacks
Commit 6b5d6e2dbc88 changed how we order configuration callbacks
and introduced a regression in the processing of the 'apply_finish'
callbacks. Fix this.
Renato Westphal [Thu, 10 Oct 2019 01:17:08 +0000 (22:17 -0300)]
lib: optimize loading of the startup configuration
Load the startup configuration directly into the CLI shared candidate
configuration instead of loading it into a private candidate
configuration. This way we don't need to initialize the shared
candidate separately later as a copy of the running configuration,
which is a potentially expensive operation.
Also, make the northbound process SIGHUP correctly even when --tcli
is not used.
Renato Westphal [Sun, 6 Oct 2019 04:16:47 +0000 (01:16 -0300)]
lib: avoid expensive operations when editing a candidate config
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.
Renato Westphal [Sun, 6 Oct 2019 00:20:28 +0000 (21:20 -0300)]
lib: remove expensive error handling in the northbound CLI client
The nb_cli_apply_changes() function was creating a full copy of
the candidate configuration before editing it. This excerpt from
the northbond documentation explains why this was being done:
"NOTE: the nb_cli_cfg_change() function clones the candidate
configuration before actually editing it. This way, if any error
happens during the editing, the original candidate is restored to
avoid inconsistencies. Either all changes from the configuration
command are performed successfully or none are. It's like a
mini-transaction but happening on the candidate configuration
(thus the northbound callbacks are not involved)".
The problem is that this kind of error handling is just too
expensive. A command should never fail to edit the candidate
configuration unless there's a bug in the code (e.g. when the
CLI wrapper command passes an integer value that YANG rejects due
to a "range" statement). In such cases, a command might fail to
be applied or applied only partially if it edits multiple YANG
nodes. When that happens, just log an error to make the operator
aware of the problem, but otherwise ignore it instead of rejecting
the command and restoring the candidate to its previous state. We
shouldn't add an extreme overhead to the northbound CLI client only
to handle errors that should never happen in practice.
Renato Westphal [Sat, 5 Oct 2019 23:52:36 +0000 (20:52 -0300)]
lib: optimize VTY_CHECK_XPATH
There's no need to check for removed configuration objects in the
following two cases:
* A private candidate configuration is being edited;
* The startup configuration is being read.
Rafael Zalamena [Fri, 11 Oct 2019 19:13:24 +0000 (16:13 -0300)]
bfdd: simplify session observers code
Don't be selective about what to observe, always observe all possible
aspects of the session that may change on run-time (i.e. bind address,
interface and VRF existence).
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
Rafael Zalamena [Fri, 11 Oct 2019 16:12:26 +0000 (13:12 -0300)]
bfdd: set session down after disabling it
If a session is no longer able to send/receive packets, it is very
likely it will be down in a few milliseconds so lets speed up the
process and correctly mark it as down.
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
sudhanshukumar22 [Thu, 10 Oct 2019 06:13:09 +0000 (23:13 -0700)]
lib: changes for making snmp socket non-blocking
Description: The changes have been done to make the snmp socket
non-blocking before calling snmp_read()
Problem Description/Summary :
vtysh hangs on first try to enter after a reboot with BGP dynamic peers
Expected Behavior :
VTYSH should not hang.
When we debug more into bgpd docker by doing gdb on its threads, we find the below thread of bgpd, which is causing the issue.
Thread 1 (Thread 0x7f1e1ec46d40 (LWP 47)):
0x00007f1e1d762593 in recvfrom () from /lib/x86_64-linux-gnu/libpthread.so.0
0x00007f1e1aadd09b in netsnmp_tcpbase_recv () from /usr/lib/x86_64-linux-gnu/libnetsnmp.so.30
0x00007f1e1aad9617 in netsnmp_transport_recv () from /usr/lib/x86_64-linux-gnu/libnetsnmp.so.30
0x00007f1e1aab2c07 in _sess_read () from /usr/lib/x86_64-linux-gnu/libnetsnmp.so.30
0x00007f1e1aab3a29 in snmp_sess_read2 () from /usr/lib/x86_64-linux-gnu/libnetsnmp.so.30
0x00007f1e1aab3a7b in snmp_read2 () from /usr/lib/x86_64-linux-gnu/libnetsnmp.so.30
0x00007f1e1aab3acf in snmp_read () from /usr/lib/x86_64-linux-gnu/libnetsnmp.so.30
0x00007f1e1b44d7ec in agentx_read (t=0x7fffa75f0080) at lib/agentx.c:63
0x00007f1e1e7d6451 in thread_call (thread=0x7fffa75f0080) at lib/thread.c:1620
0x00007f1e1e770699 in frr_run (master=0x559396ea60f0) at lib/libfrr.c:1011
0x0000559395b4d953 in main (argc=5, argv=0x7fffa75f02b8) at bgpd/bgp_main.c:492
(gdb) bt
0x00007f830c89d210 in __read_nocancel () from /lib/x86_64-linux-gnu/libpthread.so.0
0x000056450e1e8238 in vtysh_client_run (vclient=0x56450e4a8b40 <vtysh_client+24768>, line=0x56450e21add0 enable, callback=0x0, cbarg=0x0) at vtysh/vtysh.c:216
0x000056450e1e8c6b in vtysh_client_run_all (head_client=0x56450e4a8b40 <vtysh_client+24768>, line=0x56450e21add0 enable, continue_on_err=0, callback=0x0, cbarg=0x0) at vtysh/vtysh.c:356
0x000056450e1e8ddb in vtysh_client_execute (head_client=0x56450e4a8b40 <vtysh_client+24768>, line=0x56450e21add0 enable) at vtysh/vtysh.c:393
0x000056450e1e9c82 in vtysh_execute_func (line=0x56450e21add0 enable, pager=0) at vtysh/vtysh.c:598
0x000056450e1e9dee in vtysh_execute_no_pager (line=0x56450e21add0 enable) at vtysh/vtysh.c:619
0x000056450e1f7d48 in vtysh_read_file (confp=0x56451000a9d0, top_cfg=1) at vtysh/vtysh_config.c:494
0x000056450e1f7ef2 in vtysh_read_config (config_default_dir=0x56450e4edc20 <frr_config> /etc/frr/frr.conf, top_cfg=1) at vtysh/vtysh_config.c:522
0x000056450e1e5de4 in vtysh_apply_top_level_config () at vtysh/vtysh_main.c:301
0x000056450e1e7842 in main (argc=2, argv=0x7ffc81e6f598, env=0x7ffc81e6f5b0) at vtysh/vtysh_main.c:692
The fix has been taken from the following link.
https://sourceforge.net/p/net-snmp/patches/1348/
zebra: entries can be pushed in tables, under vrf netns backend
initially, vrf backend if vrf-lite, and a specific table identifier is
associated to a vrf. here, with netns vrf backend, there is no specific
table assigned to except default routing table. use the passed table_id
parameter in zapi api, and apply it to the entry to be pushed in, or to
be removed.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
projects / mirror_frr.git / log