On 1/5/26 11:02 AM, Ales Musil via dev wrote:
> The nexthop of Logical Router Policy can be NULL or empty, make sure
> the northd doesn't crash when that happens:

Hi, Ales.  Not a full review, but see a few comments below.

The patch subject, IMO, is too generic and can't be used to find the
issue in the git history.  Should be something like
"northd: Fix crash on explicit output-port reroute without nexthop".
Or something like this.

> 
> northd/northd.c:14771:39: runtime error: null pointer passed as argument 1, 
> which is declared to never be null

bit: This line can be wrapped.

> string.h:247:33: note: nonnull attribute specified here
>     build_route_policies northd/northd.c:14771:32
>     en_route_policies_run northd/en-northd.c:324:9
>     engine_recompute lib/inc-proc-eng.c:448:33
>     engine_compute lib/inc-proc-eng.c:491:17
>     engine_run_node lib/inc-proc-eng.c:550:14
>     engine_run ovn/lib/inc-proc-eng.c:579:9
>     inc_proc_northd_run northd/inc-proc-northd.c:608:5
>     main northd/ovn-northd.c:1134:36
> 
> Reported-at: https://issues.redhat.com/browse/FDP-2919
> Signed-off-by: Ales Musil <[email protected]>
> ---
>  northd/northd.c     |  5 +++++
>  tests/ovn-northd.at | 28 ++++++++++++++++++++++++++++
>  2 files changed, 33 insertions(+)
> 
> diff --git a/northd/northd.c b/northd/northd.c
> index c3c0780a3..2faf5d09a 100644
> --- a/northd/northd.c
> +++ b/northd/northd.c
> @@ -14762,6 +14762,11 @@ build_route_policies(struct ovn_datapath *od, const 
> struct hmap *lr_ports,
>              for (size_t j = 0; j < n_nexthops; j++) {
>                  char *nexthop = rule->n_nexthops
>                      ? rule->nexthops[j] : rule->nexthop;
> +
> +                if (!nexthop || !nexthop[0]) {

I'm a little confused, how the nexthop can be NULL here?  The default
database value for a string is an empty string, which is not supposed
to be NULL.  And if n_nexthops is not zero, then we must have the
rule->nexthops[j] non-NULL.  What am I missing here?

> +                    continue;
> +                }
> +
>                  struct ovn_port *out_port = NULL;
>                  bool is_ipv4 = strchr(nexthop, '.') ? true : false;
>  
> diff --git a/tests/ovn-northd.at b/tests/ovn-northd.at
> index 5b1a8b6f8..79075465b 100644
> --- a/tests/ovn-northd.at
> +++ b/tests/ovn-northd.at
> @@ -19030,3 +19030,31 @@ check ovn-nbctl --wait=sb sync
>  OVN_CLEANUP_NORTHD
>  AT_CLEANUP
>  ])
> +
> +OVN_FOR_EACH_NORTHD_NO_HV([
> +AT_SETUP([Router policies - crash])

This test name also doesn't provide any useful information about the test.

> +ovn_start
> +
> +dnl Test with a simple reroute policy with an explicit output-port set.

This comment should be converted into the test name, IMO.  E.g.:

AT_SETUP([Router policies - explicit output-port without nexthop])

> +check ovn-nbctl                                                          \
> +    -- lr-add lr0                                                        \
> +    -- lrp-add lr0 lrp1 00:00:00:00:00:01 1.1.1.1/24                     \
> +    -- lrp-add lr0 lrp2 00:00:00:00:00:02 2.2.2.1/24                     \
> +    -- --output-port=lrp2 lr-policy-add lr0 100 "ip4.src == 42.42.42.42" \
> +                                        reroute 3.3.3.3
> +
> +dnl Clear the nexthops. This shouldn't crash and not produce any lflows.

s/and not/or/ ?

> +uuid=$(fetch_column nb:logical_router_policy _uuid priority=100)
> +check ovn-nbctl clear logical_router_policy $uuid nexthops
> +check ovn-nbctl --wait=sb sync
> +
> +ovn-sbctl dump-flows lr0 > lr0flows

nit: Can be wrapped in AT_CHECK.

> +AT_CAPTURE_FILE([lr0flows])
> +
> +AT_CHECK([grep "lr_in_policy[[^_]]" lr0flows | ovn_strip_lflows | sort], 
> [0], [dnl
> +  table=??(lr_in_policy       ), priority=0    , match=(1), 
> action=(reg8[[0..15]] = 0; next;)
> +])
> +
> +OVN_CLEANUP_NORTHD
> +AT_CLEANUP
> +])

_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to