From: Lee Jones lee@kernel.org
Upstream commit 04c55383fa5689357bcdd2c8036725a55ed632bc.
In the event of a failure in tcf_change_indev(), u32_set_parms() will immediately return without decrementing the recently incremented reference counter. If this happens enough times, the counter will rollover and the reference freed, leading to a double free which can be used to do 'bad things'.
In order to prevent this, move the point of possible failure above the point where the reference counter is incremented. Also save any meaningful return values to be applied to the return data at the appropriate point in time.
This issue was caught with KASAN.
Fixes: 705c7091262d ("net: sched: cls_u32: no need to call tcf_exts_change for newly allocated struct") Suggested-by: Eric Dumazet edumazet@google.com Signed-off-by: Lee Jones lee@kernel.org Reviewed-by: Eric Dumazet edumazet@google.com Acked-by: Jamal Hadi Salim jhs@mojatatu.com Signed-off-by: David S. Miller davem@davemloft.net Signed-off-by: Rishabh Bhatnagar risbhat@amazon.com --- net/sched/cls_u32.c | 21 ++++++++++++++------- 1 file changed, 14 insertions(+), 7 deletions(-)
diff --git a/net/sched/cls_u32.c b/net/sched/cls_u32.c index fdbdcba44917..a4e01220a53a 100644 --- a/net/sched/cls_u32.c +++ b/net/sched/cls_u32.c @@ -774,11 +774,22 @@ static int u32_set_parms(struct net *net, struct tcf_proto *tp, struct nlattr *est, bool ovr) { int err; +#ifdef CONFIG_NET_CLS_IND + int ifindex = -1; +#endif
err = tcf_exts_validate(net, tp, tb, est, &n->exts, ovr); if (err < 0) return err;
+#ifdef CONFIG_NET_CLS_IND + if (tb[TCA_U32_INDEV]) { + ifindex = tcf_change_indev(net, tb[TCA_U32_INDEV]); + if (ifindex < 0) + return -EINVAL; + } +#endif + if (tb[TCA_U32_LINK]) { u32 handle = nla_get_u32(tb[TCA_U32_LINK]); struct tc_u_hnode *ht_down = NULL, *ht_old; @@ -806,14 +817,10 @@ static int u32_set_parms(struct net *net, struct tcf_proto *tp, }
#ifdef CONFIG_NET_CLS_IND - if (tb[TCA_U32_INDEV]) { - int ret; - ret = tcf_change_indev(net, tb[TCA_U32_INDEV]); - if (ret < 0) - return -EINVAL; - n->ifindex = ret; - } + if (ifindex >= 0) + n->ifindex = ifindex; #endif + return 0; }
On Thu, 27 Jul 2023 19:15:54 +0000 Rishabh Bhatnagar risbhat@amazon.com wrote:
From: Lee Jones lee@kernel.org
Upstream commit 04c55383fa5689357bcdd2c8036725a55ed632bc.
In the event of a failure in tcf_change_indev(), u32_set_parms() will immediately return without decrementing the recently incremented reference counter. If this happens enough times, the counter will rollover and the reference freed, leading to a double free which can be used to do 'bad things'.
In order to prevent this, move the point of possible failure above the point where the reference counter is incremented. Also save any meaningful return values to be applied to the return data at the appropriate point in time.
This issue was caught with KASAN.
Fixes: 705c7091262d ("net: sched: cls_u32: no need to call tcf_exts_change for newly allocated struct") Suggested-by: Eric Dumazet edumazet@google.com Signed-off-by: Lee Jones lee@kernel.org Reviewed-by: Eric Dumazet edumazet@google.com Acked-by: Jamal Hadi Salim jhs@mojatatu.com Signed-off-by: David S. Miller davem@davemloft.net Signed-off-by: Rishabh Bhatnagar risbhat@amazon.com
Acked-by: SeongJae Park sj@kernel.org
net/sched/cls_u32.c | 21 ++++++++++++++------- 1 file changed, 14 insertions(+), 7 deletions(-)
diff --git a/net/sched/cls_u32.c b/net/sched/cls_u32.c index fdbdcba44917..a4e01220a53a 100644 --- a/net/sched/cls_u32.c +++ b/net/sched/cls_u32.c @@ -774,11 +774,22 @@ static int u32_set_parms(struct net *net, struct tcf_proto *tp, struct nlattr *est, bool ovr) { int err; +#ifdef CONFIG_NET_CLS_IND
- int ifindex = -1;
+#endif err = tcf_exts_validate(net, tp, tb, est, &n->exts, ovr); if (err < 0) return err; +#ifdef CONFIG_NET_CLS_IND
- if (tb[TCA_U32_INDEV]) {
ifindex = tcf_change_indev(net, tb[TCA_U32_INDEV]);if (ifindex < 0)return -EINVAL;- }
+#endif
- if (tb[TCA_U32_LINK]) { u32 handle = nla_get_u32(tb[TCA_U32_LINK]); struct tc_u_hnode *ht_down = NULL, *ht_old;
@@ -806,14 +817,10 @@ static int u32_set_parms(struct net *net, struct tcf_proto *tp, } #ifdef CONFIG_NET_CLS_IND
- if (tb[TCA_U32_INDEV]) {
int ret;ret = tcf_change_indev(net, tb[TCA_U32_INDEV]);if (ret < 0)return -EINVAL;n->ifindex = ret;- }
- if (ifindex >= 0)
n->ifindex = ifindex;#endif
- return 0;
Very trivial nit: Someone might think the above new line is better not to be added? I don't really care, though.
} -- 2.40.1
Thanks, SJ
On Thu, Jul 27, 2023 at 07:15:54PM +0000, Rishabh Bhatnagar wrote:
From: Lee Jones lee@kernel.org
Upstream commit 04c55383fa5689357bcdd2c8036725a55ed632bc.
In the event of a failure in tcf_change_indev(), u32_set_parms() will immediately return without decrementing the recently incremented reference counter. If this happens enough times, the counter will rollover and the reference freed, leading to a double free which can be used to do 'bad things'.
In order to prevent this, move the point of possible failure above the point where the reference counter is incremented. Also save any meaningful return values to be applied to the return data at the appropriate point in time.
This issue was caught with KASAN.
Fixes: 705c7091262d ("net: sched: cls_u32: no need to call tcf_exts_change for newly allocated struct") Suggested-by: Eric Dumazet edumazet@google.com Signed-off-by: Lee Jones lee@kernel.org Reviewed-by: Eric Dumazet edumazet@google.com Acked-by: Jamal Hadi Salim jhs@mojatatu.com Signed-off-by: David S. Miller davem@davemloft.net Signed-off-by: Rishabh Bhatnagar risbhat@amazon.com
net/sched/cls_u32.c | 21 ++++++++++++++------- 1 file changed, 14 insertions(+), 7 deletions(-)
We need a 4.19.y backport before we can apply a 4.14.y version, as you do not want to upgrade and have a regression.
thanks,
greg k-h
linux-stable-mirror@lists.linaro.org
-
Greg KH -
Rishabh Bhatnagar -
SeongJae Park