From: Ido Schimmel <idosch@nvidia.com>
To: netdev-bot+sashiko@kernel.org
Cc: liamsmith827@grapheneos.org, netdev@vger.kernel.org,
dsahern@kernel.org, davem@davemloft.net, edumazet@kernel.org,
kuba@kernel.org, pabeni@redhat.com, horms@kernel.org,
shuah@kernel.org, linux-kselftest@vger.kernel.org,
linux-kernel@vger.kernel.org, liamsmith827@proton.me
Subject: Re: [PATCH net-next v5 2/4] ipv4: restrict on-link fallback to lookup misses
Date: Thu, 19 Nov 2026 17:20:39 +0200 [thread overview]
Message-ID: <20261119152039.GA1980872@shredder> (raw)
In-Reply-To: <179504450812.3136.16231358596487999009@kernel.org>
tl;dr - I believe the patch is fine as-is.
On Wed, Nov 18, 2026 at 11:28:28PM +0000, netdev-bot+sashiko@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found 2 potential
> issue(s) to consider.
>
> Critical: 0 · High: 1 · Medium: 0 · Low: 1
>
> If you believe none of the issues are legitimate, please reply to each
> issue of Medium severity or higher and explain why the AI is wrong.
> If any of them are legitimate and you plan to respin, please let patchwork
> know by including "pw-bot: cr" as a separate line at the end of your reply
> (one such reply per series is enough).
>
> - [High] Route-based restrictions can still be bypassed by an
> unprivileged process after this fix.
> - [Low] The commit message describes a security-relevant bug that users
> can see: unprivileged Android apps can get around a VPN kill switch…
>
> --
>
> ipv4: restrict on-link fallback to lookup misses
>
> With this patch, ip_route_output_key_hash_rcu() takes the oif-bound
> on-link fallback for non-vrf unicast only when fib_lookup() reports a
> real lookup miss (-ESRCH, via FIB_LOOKUP_REPORT_MISS). Errors from
> blackhole, prohibit and unreachable routes or rules now go back to the
> caller, and -ESRCH is mapped back to -ENETUNREACH.
>
> > diff --git a/net/ipv4/route.c b/net/ipv4/route.c
> > index b9f5397336d34..b71443e4b522c 100644
> > --- a/net/ipv4/route.c
> > +++ b/net/ipv4/route.c
> > @@ -2859,12 +2859,13 @@ struct rtable *ip_route_output_key_hash_rcu(struct net *net, struct flowi4 *fl4,
> > goto make_route;
> > }
> >
> > - err = fib_lookup(net, fl4, res, 0);
> > + err = fib_lookup(net, fl4, res, FIB_LOOKUP_REPORT_MISS);
>
> [Severity: Low]
> The commit message describes a security-relevant problem that users can
> see: unprivileged apps can bypass the Android VPN kill switch. There is
> no Fixes: tag and no Cc: stable, though.
>
> The unconditional on-link fallback in ip_route_output_key_hash_rcu()
> dates back to the initial git import. Should this carry something like
> the following?
No.
>
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
>
> A Fixes: tag would also make it clearer whether this is meant for net
> or net-next.
>
> This change also depends on the earlier patch "ipv4: add
> FIB_LOOKUP_REPORT_MISS". A backport of this patch alone would fail to
> build, because FIB_LOOKUP_REPORT_MISS would be undefined, so it would
> not misbehave silently. If this is meant for stable, could the
> dependency be stated?
>
> > if (err) {
> > res->fi = NULL;
> > res->table = NULL;
> > if (fl4->flowi4_oif &&
> > - (ipv4_is_multicast(fl4->daddr) || !fl4->flowi4_l3mdev)) {
> > + (ipv4_is_multicast(fl4->daddr) ||
> > + (!fl4->flowi4_l3mdev && err == -ESRCH))) {
>
> [Severity: High]
> Can route-based restrictions still be bypassed here when the flow asks
> for link scope?
>
> The commit message says this fixes "routes and rules that are supposed
> to restrict traffic (blackhole, prohibit, unreachable)".
>
> An unprivileged process can bind to a device (SO_BINDTODEVICE,
> IP_UNICAST_IF, IP_PKTINFO) and also set SO_DONTROUTE or MSG_DONTROUTE.
> Neither needs a capability. In that case ip_sendmsg_scope() returns
> RT_SCOPE_LINK, and so does ip_sock_rt_scope() for connect and TCP:
>
> include/net/ip.h:ip_sendmsg_scope() {
> if (sock_flag(&inet->sk, SOCK_LOCALROUTE) ||
> msg->msg_flags & MSG_DONTROUTE ||
> (ipc->opt && ipc->opt->opt.is_strictroute))
> return RT_SCOPE_LINK;
> ...
> }
>
> In fib_table_lookup(), the scope check runs before the reject error is
> read:
>
> net/ipv4/fib_trie.c:fib_table_lookup() {
> ...
> if (fa->fa_info->fib_scope < flp->flowi4_scope)
> continue;
> fib_alias_accessed(fa);
> err = fib_props[fa->fa_type].error;
> ...
> }
>
> Blackhole, prohibit and unreachable routes have universe scope by
> default. fib_create_info() keeps cfg->fc_scope for reject types, and
> iproute2 creates them with universe scope. A link-scope flow therefore
> skips the reject route as if it were not there.
>
> The table then returns -EAGAIN. Because of FIB_LOOKUP_REPORT_MISS,
> fib_lookup() turns that into -ESRCH, and so does __fib_lookup() ->
> fib_rules_lookup() when custom rules exist.
>
> The condition above then holds, and the code jumps to make_route with an
> RTN_UNICAST route on the chosen device. For example:
>
> ping -r -I eth1 <blocked dst>
>
> That packet would still seem to go out eth1, even with a prohibit route
> for that destination. Before this patch every error fell back. After it,
> the same bypass needs just one more unprivileged socket option.
>
> Rule actions (FR_ACT_PROHIBIT, FR_ACT_UNREACHABLE and FR_ACT_BLACKHOLE
> in fib4_rule_action()) are not filtered by scope. So the kill switch
> built on prohibit rules does look fixed, and only the route case
> remains open.
>
> Could the fallback tell a real miss apart from a reject route that was
> skipped only because of scope? Another option would be to not apply the
> scope filter to reject-type aliases when deciding whether to fall back.
>
> A later patch in the series, "ipv4: replace obsolete routing comment",
> adds the comment "Non-vrf unicast honors restrictions", which does not
> hold in this case either. Nothing later in the series changes
> fib_trie.c or this condition.
This was already addressed in v2:
https://lore.kernel.org/netdev/2503cfd2-5629-448f-bb9b-111ea239efb6@grapheneos.org/
next prev parent reply other threads:[~2026-11-19 15:20 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-16 20:27 [PATCH net-next v5 0/4] ipv4: restrict on-link fallback to lookup misses liamsmith827
2026-11-16 20:27 ` [PATCH net-next v5 1/4] ipv4: add FIB_LOOKUP_REPORT_MISS liamsmith827
2026-11-16 20:27 ` [PATCH net-next v5 2/4] ipv4: restrict on-link fallback to lookup misses liamsmith827
2026-11-18 23:28 ` netdev-bot+sashiko
2026-11-19 15:20 ` Ido Schimmel [this message]
2026-11-16 20:27 ` [PATCH net-next v5 3/4] selftests: net: test restricted on-link fallback liamsmith827
2026-11-18 23:28 ` netdev-bot+sashiko
2026-11-16 20:27 ` [PATCH net-next v5 4/4] ipv4: replace obsolete routing comment liamsmith827
2026-11-18 23:28 ` netdev-bot+sashiko
2026-11-19 15:20 ` Ido Schimmel
2026-11-17 10:56 ` [PATCH net-next v5 0/4] ipv4: restrict on-link fallback to lookup misses Ido Schimmel
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20261119152039.GA1980872@shredder \
--to=idosch@nvidia.com \
--cc=davem@davemloft.net \
--cc=dsahern@kernel.org \
--cc=edumazet@kernel.org \
--cc=horms@kernel.org \
--cc=kuba@kernel.org \
--cc=liamsmith827@grapheneos.org \
--cc=liamsmith827@proton.me \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-kselftest@vger.kernel.org \
--cc=netdev-bot+sashiko@kernel.org \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=shuah@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox