[RFC PATCH 6.1.y] net/sched: act_mirred: don't override retval if we already lost the skb

[RFC PATCH 6.1.y] net/sched: act_mirred: don't override retval if we already lost the skb

[jianqi.ren.cn]

From: Jakub Kicinski <kuba@kernel.org>

[ Upstream commit 166c2c8a6a4dc2e4ceba9e10cfe81c3e469e3210 ]

If we're redirecting the skb, and haven't called tcf_mirred_forward(),
yet, we need to tell the core to drop the skb by setting the retcode
to SHOT. If we have called tcf_mirred_forward(), however, the skb
is out of our hands and returning SHOT will lead to UaF.

Move the retval override to the error path which actually need it.

Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Fixes: e5cf1baf92cb ("act_mirred: use TC_ACT_REINSERT when possible")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Jianqi Ren <jianqi.ren.cn@windriver.com>
Signed-off-by: He Zhe <zhe.he@windriver.com>
---
Verified the build test
---
 net/sched/act_mirred.c | 20 ++++++++++++--------
 1 file changed, 12 insertions(+), 8 deletions(-)

diff --git a/net/sched/act_mirred.c b/net/sched/act_mirred.c
index 36395e5db3b4..24c70ba6eebc 100644
--- a/net/sched/act_mirred.c
+++ b/net/sched/act_mirred.c
@@ -259,13 +259,13 @@ static int tcf_mirred_act(struct sk_buff *skb, const struct tc_action *a,
 	dev = rcu_dereference_bh(m->tcfm_dev);
 	if (unlikely(!dev)) {
 		pr_notice_once("tc mirred: target device is gone\n");
-		goto out;
+		goto err_cant_do;
 	}
 
 	if (unlikely(!(dev->flags & IFF_UP)) || !netif_carrier_ok(dev)) {
 		net_notice_ratelimited("tc mirred to Houston: device %s is down\n",
 				       dev->name);
-		goto out;
+		goto err_cant_do;
 	}
 
 	/* we could easily avoid the clone only if called by ingress and clsact;
@@ -279,7 +279,7 @@ static int tcf_mirred_act(struct sk_buff *skb, const struct tc_action *a,
 	if (!use_reinsert) {
 		skb2 = skb_clone(skb, GFP_ATOMIC);
 		if (!skb2)
-			goto out;
+			goto err_cant_do;
 	}
 
 	want_ingress = tcf_mirred_act_wants_ingress(m_eaction);
@@ -321,12 +321,16 @@ static int tcf_mirred_act(struct sk_buff *skb, const struct tc_action *a,
 	}
 
 	err = tcf_mirred_forward(want_ingress, skb2);
-	if (err) {
-out:
+	if (err)
 		tcf_action_inc_overlimit_qstats(&m->common);
-		if (tcf_mirred_is_act_redirect(m_eaction))
-			retval = TC_ACT_SHOT;
-	}
+	__this_cpu_dec(mirred_nest_level);
+
+	return retval;
+
+err_cant_do:
+	if (tcf_mirred_is_act_redirect(m_eaction))
+		retval = TC_ACT_SHOT;
+	tcf_action_inc_overlimit_qstats(&m->common);
 	__this_cpu_dec(mirred_nest_level);
 
 	return retval;
-- 
2.25.1

Re: [RFC PATCH 6.1.y] net/sched: act_mirred: don't override retval if we already lost the skb

[Simon Horman]

On Wed, Mar 19, 2025 at 09:22:25AM +0800, jianqi.ren.cn@windriver.com wrote:
> From: Jakub Kicinski <kuba@kernel.org>
> 
> [ Upstream commit 166c2c8a6a4dc2e4ceba9e10cfe81c3e469e3210 ]
> 
> If we're redirecting the skb, and haven't called tcf_mirred_forward(),
> yet, we need to tell the core to drop the skb by setting the retcode
> to SHOT. If we have called tcf_mirred_forward(), however, the skb
> is out of our hands and returning SHOT will lead to UaF.
> 
> Move the retval override to the error path which actually need it.
> 
> Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
> Fixes: e5cf1baf92cb ("act_mirred: use TC_ACT_REINSERT when possible")
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
> Signed-off-by: David S. Miller <davem@davemloft.net>
> Signed-off-by: Jianqi Ren <jianqi.ren.cn@windriver.com>
> Signed-off-by: He Zhe <zhe.he@windriver.com>
> ---
> Verified the build test

Sorry if it is obvious, but I'm confused by the intention of posting
an RFC for stable. Are you asking for buy-in regarding backporting
this patch to 6.1.y because for some reason it hasn't already propagated
there?

RE: [RFC PATCH 6.1.y] net/sched: act_mirred: don't override retval if we already lost the skb

[Ren, Jianqi (Jacky) (CN)]

The context of this patch is changed compared with the original fix.  Adding RFC means that I want to let the author or other experts to make a possible review to make sure the logic is right.

-----Original Message-----
From: Simon Horman <horms@kernel.org> 
Sent: Saturday, March 22, 2025 02:01
To: Ren, Jianqi (Jacky) (CN) <Jianqi.Ren.CN@windriver.com>
Cc: stable@vger.kernel.org; patches@lists.linux.dev; gregkh@linuxfoundation.org; linux-kernel@vger.kernel.org; jhs@mojatatu.com; xiyou.wangcong@gmail.com; jiri@resnulli.us; davem@davemloft.net; edumazet@google.com; kuba@kernel.org; pabeni@redhat.com; netdev@vger.kernel.org; michal.swiatkowski@linux.intel.com
Subject: Re: [RFC PATCH 6.1.y] net/sched: act_mirred: don't override retval if we already lost the skb

CAUTION: This email comes from a non Wind River email account!
Do not click links or open attachments unless you recognize the sender and know the content is safe.

On Wed, Mar 19, 2025 at 09:22:25AM +0800, jianqi.ren.cn@windriver.com wrote:
> From: Jakub Kicinski <kuba@kernel.org>
>
> [ Upstream commit 166c2c8a6a4dc2e4ceba9e10cfe81c3e469e3210 ]
>
> If we're redirecting the skb, and haven't called tcf_mirred_forward(), 
> yet, we need to tell the core to drop the skb by setting the retcode 
> to SHOT. If we have called tcf_mirred_forward(), however, the skb is 
> out of our hands and returning SHOT will lead to UaF.
>
> Move the retval override to the error path which actually need it.
>
> Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
> Fixes: e5cf1baf92cb ("act_mirred: use TC_ACT_REINSERT when possible")
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
> Signed-off-by: David S. Miller <davem@davemloft.net>
> Signed-off-by: Jianqi Ren <jianqi.ren.cn@windriver.com>
> Signed-off-by: He Zhe <zhe.he@windriver.com>
> ---
> Verified the build test

Sorry if it is obvious, but I'm confused by the intention of posting an RFC for stable. Are you asking for buy-in regarding backporting this patch to 6.1.y because for some reason it hasn't already propagated there?