Re: [PATCH net-next v4 01/10] ipv6: add ip6_del_rt_reason()

[Ido Schimmel <idosch@nvidia.com>]

On Wed, Jul 29, 2026 at 09:07:30PM +0900, Yuyang Huang wrote:
> Add RTA_DEL_REASON and enum rta_del_reason to the rtnetlink uAPI, and
> add ip6_del_rt_reason(), which takes the reason a route is being
> deleted. ip6_del_rt() becomes a wrapper that passes
> RTA_DEL_REASON_UNSPEC, so its callers do not change.
> 
> The reason is unused for now. Subsequent patches propagate it to the
> deletion path and report it on RTM_DELROUTE.
> 
> Signed-off-by: Yuyang Huang <sigefriedhyy@gmail.com>
> ---
>  include/net/ip6_route.h        |  9 +++++++++
>  include/uapi/linux/rtnetlink.h | 17 +++++++++++++++++
>  net/ipv6/route.c               |  8 +++++++-
>  3 files changed, 33 insertions(+), 1 deletion(-)
> 
> diff --git a/include/net/ip6_route.h b/include/net/ip6_route.h
> index 09ffe0f13ce7..92ad5a0d03a2 100644
> --- a/include/net/ip6_route.h
> +++ b/include/net/ip6_route.h
> @@ -128,12 +128,21 @@ int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
>  int ip6_ins_rt(struct net *net, struct fib6_info *f6i);
>  #if IS_ENABLED(CONFIG_IPV6)
>  int ip6_del_rt(struct net *net, struct fib6_info *f6i, bool skip_notify);
> +int ip6_del_rt_reason(struct net *net, struct fib6_info *f6i, bool skip_notify,
> +		      enum rta_del_reason del_reason);

From Sashiko:

"
This isn't a bug, but was carrying the reason in struct nl_info considered
instead of adding a parallel argument?
"

I did consider it, but it doesn't look like a good fit to carry a route
deletion reason in a generic netlink structure (also used by nexthops).

It later says:

"
One counter-point worth noting: rt6_fill_node() is also reached from dump,
GET and hw-flags paths that have no struct nl_info at all, so an explicit
parameter there is unavoidable either way, and nl_info is shared with
IPv4, MPLS and nexthop code.
"

https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20260729120739.47880-1-sigefriedhyy%40gmail.com


>  #else
>  static inline int ip6_del_rt(struct net *net, struct fib6_info *f6i,
>  			     bool skip_notify)
>  {
>  	return -EAFNOSUPPORT;
>  }
> +
> +static inline int ip6_del_rt_reason(struct net *net, struct fib6_info *f6i,
> +				    bool skip_notify,
> +				    enum rta_del_reason del_reason)
> +{
> +	return -EAFNOSUPPORT;
> +}
>  #endif
>  
>  void rt6_flush_exceptions(struct fib6_info *f6i);
> diff --git a/include/uapi/linux/rtnetlink.h b/include/uapi/linux/rtnetlink.h
> index 27265fd31e5f..fe00e624f3c3 100644
> --- a/include/uapi/linux/rtnetlink.h
> +++ b/include/uapi/linux/rtnetlink.h
> @@ -399,6 +399,7 @@ enum rtattr_type_t {
>  	RTA_DPORT,
>  	RTA_NH_ID,
>  	RTA_FLOWLABEL,
> +	RTA_DEL_REASON,
>  	__RTA_MAX
>  };
>  
> @@ -407,6 +408,22 @@ enum rtattr_type_t {
>  #define RTM_RTA(r)  ((struct rtattr*)(((char*)(r)) + NLMSG_ALIGN(sizeof(struct rtmsg))))
>  #define RTM_PAYLOAD(n) NLMSG_PAYLOAD(n,sizeof(struct rtmsg))
>  
> +/* RTA_DEL_REASON: why the kernel deleted the route. u32.
> + * Emitted only on RTM_DELROUTE notifications, and only when the deletion
> + * path records a cause. Absence means either an older kernel or a
> + * deletion path that does not (yet) record its cause - consumers must
> + * treat "absent" and "unspec" identically. New causes may be appended.
> + * Currently only IPv6 deletion paths record a cause.
> + */
> +enum rta_del_reason {
> +	RTA_DEL_REASON_UNSPEC,		/* cause not recorded */
> +	RTA_DEL_REASON_EXPIRED,		/* RTF_EXPIRES lifetime ran out (GC) */
> +	RTA_DEL_REASON_RA_WITHDRAWN,	/* zero-lifetime RA / PIO / RIO */
> +	__RTA_DEL_REASON_MAX
> +};
> +
> +#define RTA_DEL_REASON_MAX (__RTA_DEL_REASON_MAX - 1)

Valid comments from Sashiko:

"
This isn't a bug, but since these names become uAPI on first release, is
the RTA_ prefix the right choice for the payload values?

RTA_ is the established prefix for route attribute ids in enum
rtattr_type_t, and RTA_DEL_REASON is added to that very enum a few lines
above. So RTA_DEL_REASON_EXPIRED reads like an attribute id, and
RTA_DEL_REASON_MAX visually parallels RTA_MAX while meaning something
completely different. Elsewhere the payload value space of an attribute
gets its own prefix, for instance LWTUNNEL_ENCAP_* for RTA_ENCAP_TYPE.

[....]

RTA_DEL_REASON_RA_WITHDRAWN is an IPv6-specific cause placed in the
family-agnostic route attribute value space, and RTA_DEL_REASON_MAX has no
in-tree user by the end of the series (rtm_ipv6_policy has no
RTA_DEL_REASON entry). Would it be worth stating in the comment that the
value space must never be reinterpreted per family, so a later family
cannot reuse the numbers for its own causes?
"

And be aware of:

https://lore.kernel.org/all/83360de7addb13a3b5f4d5e722148f248fdb2ae0.1784884817.git.pabeni@redhat.com/

> +
>  /* RTM_MULTIPATH --- array of struct rtnexthop.
>   *
>   * "struct rtnexthop" describes all necessary nexthop information,
> diff --git a/net/ipv6/route.c b/net/ipv6/route.c
> index a1301334da48..9f82829923ab 100644
> --- a/net/ipv6/route.c
> +++ b/net/ipv6/route.c
> @@ -3994,7 +3994,8 @@ static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
>  	return err;
>  }
>  
> -int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
> +int ip6_del_rt_reason(struct net *net, struct fib6_info *rt, bool skip_notify,
> +		      enum rta_del_reason del_reason)

It's weird to have 'skip_notify' in a function that is supposed to
notify the deletion reason to user space. All the callers except
ip6_del_rt() pass 'false'. Can ip6_del_rt() call __ip6_del_rt() with
'RTA_DEL_REASON_UNSPEC' ?

>  {
>  	struct nl_info info = {
>  		.nl_net = net,
> @@ -4004,6 +4005,11 @@ int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
>  	return __ip6_del_rt(rt, &info);
>  }
>  
> +int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify)
> +{
> +	return ip6_del_rt_reason(net, rt, skip_notify, RTA_DEL_REASON_UNSPEC);
> +}
> +
>  static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
>  {
>  	struct nl_info *info = &cfg->fc_nlinfo;
> -- 
> 2.43.0
>