From e49d8c22f1261c43a986a7fdbf677ac309682a07 Mon Sep 17 00:00:00 2001 From: Cong Wang Date: Tue, 22 Sep 2020 20:56:23 -0700 Subject: net_sched: defer tcf_idr_insert() in tcf_action_init_1() All TC actions call tcf_idr_insert() for new action at the end of their ->init(), so we can actually move it to a central place in tcf_action_init_1(). And once the action is inserted into the global IDR, other parallel process could free it immediately as its refcnt is still 1, so we can not fail after this, we need to move it after the goto action validation to avoid handling the failure case after insertion. This is found during code review, is not directly triggered by syzbot. And this prepares for the next patch. Cc: Vlad Buslov Cc: Jamal Hadi Salim Cc: Jiri Pirko Signed-off-by: Cong Wang Signed-off-by: David S. Miller --- net/sched/act_api.c | 38 ++++++++++++++++++++------------------ 1 file changed, 20 insertions(+), 18 deletions(-) (limited to 'net/sched/act_api.c') diff --git a/net/sched/act_api.c b/net/sched/act_api.c index 063d8aaf2900..0030f00234ee 100644 --- a/net/sched/act_api.c +++ b/net/sched/act_api.c @@ -467,17 +467,6 @@ int tcf_idr_create_from_flags(struct tc_action_net *tn, u32 index, } EXPORT_SYMBOL(tcf_idr_create_from_flags); -void tcf_idr_insert(struct tc_action_net *tn, struct tc_action *a) -{ - struct tcf_idrinfo *idrinfo = tn->idrinfo; - - mutex_lock(&idrinfo->lock); - /* Replace ERR_PTR(-EBUSY) allocated by tcf_idr_check_alloc */ - WARN_ON(!IS_ERR(idr_replace(&idrinfo->action_idr, a, a->tcfa_index))); - mutex_unlock(&idrinfo->lock); -} -EXPORT_SYMBOL(tcf_idr_insert); - /* Cleanup idr index that was allocated but not initialized. */ void tcf_idr_cleanup(struct tc_action_net *tn, u32 index) @@ -902,6 +891,16 @@ static const struct nla_policy tcf_action_policy[TCA_ACT_MAX + 1] = { [TCA_ACT_HW_STATS] = NLA_POLICY_BITFIELD32(TCA_ACT_HW_STATS_ANY), }; +static void tcf_idr_insert(struct tc_action *a) +{ + struct tcf_idrinfo *idrinfo = a->idrinfo; + + mutex_lock(&idrinfo->lock); + /* Replace ERR_PTR(-EBUSY) allocated by tcf_idr_check_alloc */ + WARN_ON(!IS_ERR(idr_replace(&idrinfo->action_idr, a, a->tcfa_index))); + mutex_unlock(&idrinfo->lock); +} + struct tc_action *tcf_action_init_1(struct net *net, struct tcf_proto *tp, struct nlattr *nla, struct nlattr *est, char *name, int ovr, int bind, @@ -989,6 +988,16 @@ struct tc_action *tcf_action_init_1(struct net *net, struct tcf_proto *tp, if (err < 0) goto err_mod; + if (TC_ACT_EXT_CMP(a->tcfa_action, TC_ACT_GOTO_CHAIN) && + !rcu_access_pointer(a->goto_chain)) { + tcf_action_destroy_1(a, bind); + NL_SET_ERR_MSG(extack, "can't use goto chain with NULL chain"); + return ERR_PTR(-EINVAL); + } + + if (err == ACT_P_CREATED) + tcf_idr_insert(a); + if (!name && tb[TCA_ACT_COOKIE]) tcf_set_action_cookie(&a->act_cookie, cookie); @@ -1002,13 +1011,6 @@ struct tc_action *tcf_action_init_1(struct net *net, struct tcf_proto *tp, if (err != ACT_P_CREATED) module_put(a_o->owner); - if (TC_ACT_EXT_CMP(a->tcfa_action, TC_ACT_GOTO_CHAIN) && - !rcu_access_pointer(a->goto_chain)) { - tcf_action_destroy_1(a, bind); - NL_SET_ERR_MSG(extack, "can't use goto chain with NULL chain"); - return ERR_PTR(-EINVAL); - } - return a; err_mod: -- cgit v1.2.3 From 0fedc63fadf0404a729e73a35349481c8009c02f Mon Sep 17 00:00:00 2001 From: Cong Wang Date: Tue, 22 Sep 2020 20:56:24 -0700 Subject: net_sched: commit action insertions together syzbot is able to trigger a failure case inside the loop in tcf_action_init(), and when this happens we clean up with tcf_action_destroy(). But, as these actions are already inserted into the global IDR, other parallel process could free them before tcf_action_destroy(), then we will trigger a use-after-free. Fix this by deferring the insertions even later, after the loop, and committing all the insertions in a separate loop, so we will never fail in the middle of the insertions any more. One side effect is that the window between alloction and final insertion becomes larger, now it is more likely that the loop in tcf_del_walker() sees the placeholder -EBUSY pointer. So we have to check for error pointer in tcf_del_walker(). Reported-and-tested-by: syzbot+2287853d392e4b42374a@syzkaller.appspotmail.com Fixes: 0190c1d452a9 ("net: sched: atomically check-allocate action") Cc: Vlad Buslov Cc: Jamal Hadi Salim Cc: Jiri Pirko Signed-off-by: Cong Wang Signed-off-by: David S. Miller --- net/sched/act_api.c | 32 +++++++++++++++++++++++--------- 1 file changed, 23 insertions(+), 9 deletions(-) (limited to 'net/sched/act_api.c') diff --git a/net/sched/act_api.c b/net/sched/act_api.c index 0030f00234ee..104b47f5184f 100644 --- a/net/sched/act_api.c +++ b/net/sched/act_api.c @@ -307,6 +307,8 @@ static int tcf_del_walker(struct tcf_idrinfo *idrinfo, struct sk_buff *skb, mutex_lock(&idrinfo->lock); idr_for_each_entry_ul(idr, p, tmp, id) { + if (IS_ERR(p)) + continue; ret = tcf_idr_release_unsafe(p); if (ret == ACT_P_DELETED) { module_put(ops->owner); @@ -891,14 +893,24 @@ static const struct nla_policy tcf_action_policy[TCA_ACT_MAX + 1] = { [TCA_ACT_HW_STATS] = NLA_POLICY_BITFIELD32(TCA_ACT_HW_STATS_ANY), }; -static void tcf_idr_insert(struct tc_action *a) +static void tcf_idr_insert_many(struct tc_action *actions[]) { - struct tcf_idrinfo *idrinfo = a->idrinfo; + int i; - mutex_lock(&idrinfo->lock); - /* Replace ERR_PTR(-EBUSY) allocated by tcf_idr_check_alloc */ - WARN_ON(!IS_ERR(idr_replace(&idrinfo->action_idr, a, a->tcfa_index))); - mutex_unlock(&idrinfo->lock); + for (i = 0; i < TCA_ACT_MAX_PRIO; i++) { + struct tc_action *a = actions[i]; + struct tcf_idrinfo *idrinfo; + + if (!a) + continue; + idrinfo = a->idrinfo; + mutex_lock(&idrinfo->lock); + /* Replace ERR_PTR(-EBUSY) allocated by tcf_idr_check_alloc if + * it is just created, otherwise this is just a nop. + */ + idr_replace(&idrinfo->action_idr, a, a->tcfa_index); + mutex_unlock(&idrinfo->lock); + } } struct tc_action *tcf_action_init_1(struct net *net, struct tcf_proto *tp, @@ -995,9 +1007,6 @@ struct tc_action *tcf_action_init_1(struct net *net, struct tcf_proto *tp, return ERR_PTR(-EINVAL); } - if (err == ACT_P_CREATED) - tcf_idr_insert(a); - if (!name && tb[TCA_ACT_COOKIE]) tcf_set_action_cookie(&a->act_cookie, cookie); @@ -1053,6 +1062,11 @@ int tcf_action_init(struct net *net, struct tcf_proto *tp, struct nlattr *nla, actions[i - 1] = act; } + /* We have to commit them all together, because if any error happened in + * between, we could not handle the failure gracefully. + */ + tcf_idr_insert_many(actions); + *attr_size = tcf_action_full_attrs_size(sz); return i - 1; -- cgit v1.2.3 From 1aad8049909a6d3379175ef2824a68ac35c0b564 Mon Sep 17 00:00:00 2001 From: Cong Wang Date: Mon, 28 Sep 2020 11:31:03 -0700 Subject: net_sched: remove a redundant goto chain check All TC actions call tcf_action_check_ctrlact() to validate goto chain, so this check in tcf_action_init_1() is actually redundant. Remove it to save troubles of leaking memory. Fixes: e49d8c22f126 ("net_sched: defer tcf_idr_insert() in tcf_action_init_1()") Reported-by: Vlad Buslov Suggested-by: Davide Caratti Cc: Jamal Hadi Salim Cc: Jiri Pirko Signed-off-by: Cong Wang Reviewed-by: Davide Caratti Signed-off-by: David S. Miller --- net/sched/act_api.c | 14 -------------- 1 file changed, 14 deletions(-) (limited to 'net/sched/act_api.c') diff --git a/net/sched/act_api.c b/net/sched/act_api.c index 104b47f5184f..5612b336e18e 100644 --- a/net/sched/act_api.c +++ b/net/sched/act_api.c @@ -722,13 +722,6 @@ int tcf_action_destroy(struct tc_action *actions[], int bind) return ret; } -static int tcf_action_destroy_1(struct tc_action *a, int bind) -{ - struct tc_action *actions[] = { a, NULL }; - - return tcf_action_destroy(actions, bind); -} - static int tcf_action_put(struct tc_action *p) { return __tcf_action_put(p, false); @@ -1000,13 +993,6 @@ struct tc_action *tcf_action_init_1(struct net *net, struct tcf_proto *tp, if (err < 0) goto err_mod; - if (TC_ACT_EXT_CMP(a->tcfa_action, TC_ACT_GOTO_CHAIN) && - !rcu_access_pointer(a->goto_chain)) { - tcf_action_destroy_1(a, bind); - NL_SET_ERR_MSG(extack, "can't use goto chain with NULL chain"); - return ERR_PTR(-EINVAL); - } - if (!name && tb[TCA_ACT_COOKIE]) tcf_set_action_cookie(&a->act_cookie, cookie); -- cgit v1.2.3 From 580e4273d7a883ececfefa692c1f96bdbacb99b5 Mon Sep 17 00:00:00 2001 From: Cong Wang Date: Fri, 2 Oct 2020 12:13:34 -0700 Subject: net_sched: check error pointer in tcf_dump_walker() Although we take RTNL on dump path, it is possible to skip RTNL on insertion path. So the following race condition is possible: rtnl_lock() // no rtnl lock mutex_lock(&idrinfo->lock); // insert ERR_PTR(-EBUSY) mutex_unlock(&idrinfo->lock); tc_dump_action() rtnl_unlock() So we have to skip those temporary -EBUSY entries on dump path too. Reported-and-tested-by: syzbot+b47bc4f247856fb4d9e1@syzkaller.appspotmail.com Fixes: 0fedc63fadf0 ("net_sched: commit action insertions together") Cc: Vlad Buslov Cc: Jamal Hadi Salim Cc: Jiri Pirko Signed-off-by: Cong Wang Signed-off-by: David S. Miller --- net/sched/act_api.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'net/sched/act_api.c') diff --git a/net/sched/act_api.c b/net/sched/act_api.c index 5612b336e18e..798430e1a79f 100644 --- a/net/sched/act_api.c +++ b/net/sched/act_api.c @@ -235,6 +235,8 @@ static int tcf_dump_walker(struct tcf_idrinfo *idrinfo, struct sk_buff *skb, index++; if (index < s_i) continue; + if (IS_ERR(p)) + continue; if (jiffy_since && time_after(jiffy_since, -- cgit v1.2.3