forked from ~ljy/RK356X_SDK_RELEASE
blame | history | patch | commit | commitdiff
hc
2024-10-09 244b2c5ca8b14627e4a17755e5922221e121c771 
 kernel/net/ipv6/ip6_flowlabel.c
....@@ -1,10 +1,6 @@
1
+// SPDX-License-Identifier: GPL-2.0-or-later
12 /*
23  *	ip6_flowlabel.c		IPv6 flowlabel manager.
3
- *
4
- *	This program is free software; you can redistribute it and/or
5
- *      modify it under the terms of the GNU General Public License
6
- *      as published by the Free Software Foundation; either version
7
- *      2 of the License, or (at your option) any later version.
84  *
95  *	Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
106  */
....@@ -21,6 +17,7 @@
2117 #include <linux/slab.h>
2218 #include <linux/export.h>
2319 #include <linux/pid_namespace.h>
20
+#include <linux/jump_label_ratelimit.h>
2421 
2522 #include <net/net_namespace.h>
2623 #include <net/sock.h>
....@@ -56,6 +53,9 @@
5653 /* Big socket sock */
5754 
5855 static DEFINE_SPINLOCK(ip6_sk_fl_lock);
56
+
57
+DEFINE_STATIC_KEY_DEFERRED_FALSE(ipv6_flowlabel_exclusive, HZ);
58
+EXPORT_SYMBOL(ipv6_flowlabel_exclusive);
5959 
6060 #define for_each_fl_rcu(hash, fl)				\
6161 	for (fl = rcu_dereference_bh(fl_ht[(hash)]);		\
....@@ -94,6 +94,13 @@
9494 	return fl;
9595 }
9696 
97
+static bool fl_shared_exclusive(struct ip6_flowlabel *fl)
98
+{
99
+	return fl->share == IPV6_FL_S_EXCL ||
100
+	       fl->share == IPV6_FL_S_PROCESS ||
101
+	       fl->share == IPV6_FL_S_USER;
102
+}
103
+
97104 static void fl_free_rcu(struct rcu_head *head)
98105 {
99106 	struct ip6_flowlabel *fl = container_of(head, struct ip6_flowlabel, rcu);
....@@ -107,8 +114,13 @@
107114 
108115 static void fl_free(struct ip6_flowlabel *fl)
109116 {
110
-	if (fl)
111
-		call_rcu(&fl->rcu, fl_free_rcu);
117
+	if (!fl)
118
+		return;
119
+
120
+	if (fl_shared_exclusive(fl) || fl->opt)
121
+		static_branch_slow_dec_deferred(&ipv6_flowlabel_exclusive);
122
+
123
+	call_rcu(&fl->rcu, fl_free_rcu);
112124 }
113125 
114126 static void fl_release(struct ip6_flowlabel *fl)
....@@ -244,7 +256,7 @@
244256 
245257 /* Socket flowlabel lists */
246258 
247
-struct ip6_flowlabel *fl6_sock_lookup(struct sock *sk, __be32 label)
259
+struct ip6_flowlabel *__fl6_sock_lookup(struct sock *sk, __be32 label)
248260 {
249261 	struct ipv6_fl_socklist *sfl;
250262 	struct ipv6_pinfo *np = inet6_sk(sk);
....@@ -264,7 +276,7 @@
264276 	rcu_read_unlock_bh();
265277 	return NULL;
266278 }
267
-EXPORT_SYMBOL_GPL(fl6_sock_lookup);
279
+EXPORT_SYMBOL_GPL(__fl6_sock_lookup);
268280 
269281 void fl6_free_socklist(struct sock *sk)
270282 {
....@@ -359,7 +371,7 @@
359371 
360372 static struct ip6_flowlabel *
361373 fl_create(struct net *net, struct sock *sk, struct in6_flowlabel_req *freq,
362
-	  char __user *optval, int optlen, int *err_p)
374
+	  sockptr_t optval, int optlen, int *err_p)
363375 {
364376 	struct ip6_flowlabel *fl = NULL;
365377 	int olen;
....@@ -389,7 +401,8 @@
389401 		memset(fl->opt, 0, sizeof(*fl->opt));
390402 		fl->opt->tot_len = sizeof(*fl->opt) + olen;
391403 		err = -EFAULT;
392
-		if (copy_from_user(fl->opt+1, optval+CMSG_ALIGN(sizeof(*freq)), olen))
404
+		if (copy_from_sockptr_offset(fl->opt + 1, optval,
405
+				CMSG_ALIGN(sizeof(*freq)), olen))
393406 			goto done;
394407 
395408 		msg.msg_controllen = olen;
....@@ -437,10 +450,15 @@
437450 		err = -EINVAL;
438451 		goto done;
439452 	}
453
+	if (fl_shared_exclusive(fl) || fl->opt)
454
+		static_branch_deferred_inc(&ipv6_flowlabel_exclusive);
440455 	return fl;
441456 
442457 done:
443
-	fl_free(fl);
458
+	if (fl) {
459
+		kfree(fl->opt);
460
+		kfree(fl);
461
+	}
444462 	*err_p = err;
445463 	return NULL;
446464 }
....@@ -516,185 +534,210 @@
516534 	return -ENOENT;
517535 }
518536 
519
-int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen)
537
+#define socklist_dereference(__sflp) \
538
+	rcu_dereference_protected(__sflp, lockdep_is_held(&ip6_sk_fl_lock))
539
+
540
+static int ipv6_flowlabel_put(struct sock *sk, struct in6_flowlabel_req *freq)
520541 {
521
-	int uninitialized_var(err);
522
-	struct net *net = sock_net(sk);
523542 	struct ipv6_pinfo *np = inet6_sk(sk);
524
-	struct in6_flowlabel_req freq;
525
-	struct ipv6_fl_socklist *sfl1 = NULL;
526
-	struct ipv6_fl_socklist *sfl;
527543 	struct ipv6_fl_socklist __rcu **sflp;
544
+	struct ipv6_fl_socklist *sfl;
545
+
546
+	if (freq->flr_flags & IPV6_FL_F_REFLECT) {
547
+		if (sk->sk_protocol != IPPROTO_TCP)
548
+			return -ENOPROTOOPT;
549
+		if (!np->repflow)
550
+			return -ESRCH;
551
+		np->flow_label = 0;
552
+		np->repflow = 0;
553
+		return 0;
554
+	}
555
+
556
+	spin_lock_bh(&ip6_sk_fl_lock);
557
+	for (sflp = &np->ipv6_fl_list;
558
+	     (sfl = socklist_dereference(*sflp)) != NULL;
559
+	     sflp = &sfl->next) {
560
+		if (sfl->fl->label == freq->flr_label)
561
+			goto found;
562
+	}
563
+	spin_unlock_bh(&ip6_sk_fl_lock);
564
+	return -ESRCH;
565
+found:
566
+	if (freq->flr_label == (np->flow_label & IPV6_FLOWLABEL_MASK))
567
+		np->flow_label &= ~IPV6_FLOWLABEL_MASK;
568
+	*sflp = sfl->next;
569
+	spin_unlock_bh(&ip6_sk_fl_lock);
570
+	fl_release(sfl->fl);
571
+	kfree_rcu(sfl, rcu);
572
+	return 0;
573
+}
574
+
575
+static int ipv6_flowlabel_renew(struct sock *sk, struct in6_flowlabel_req *freq)
576
+{
577
+	struct ipv6_pinfo *np = inet6_sk(sk);
578
+	struct net *net = sock_net(sk);
579
+	struct ipv6_fl_socklist *sfl;
580
+	int err;
581
+
582
+	rcu_read_lock_bh();
583
+	for_each_sk_fl_rcu(np, sfl) {
584
+		if (sfl->fl->label == freq->flr_label) {
585
+			err = fl6_renew(sfl->fl, freq->flr_linger,
586
+					freq->flr_expires);
587
+			rcu_read_unlock_bh();
588
+			return err;
589
+		}
590
+	}
591
+	rcu_read_unlock_bh();
592
+
593
+	if (freq->flr_share == IPV6_FL_S_NONE &&
594
+	    ns_capable(net->user_ns, CAP_NET_ADMIN)) {
595
+		struct ip6_flowlabel *fl = fl_lookup(net, freq->flr_label);
596
+
597
+		if (fl) {
598
+			err = fl6_renew(fl, freq->flr_linger,
599
+					freq->flr_expires);
600
+			fl_release(fl);
601
+			return err;
602
+		}
603
+	}
604
+	return -ESRCH;
605
+}
606
+
607
+static int ipv6_flowlabel_get(struct sock *sk, struct in6_flowlabel_req *freq,
608
+		sockptr_t optval, int optlen)
609
+{
610
+	struct ipv6_fl_socklist *sfl, *sfl1 = NULL;
528611 	struct ip6_flowlabel *fl, *fl1 = NULL;
612
+	struct ipv6_pinfo *np = inet6_sk(sk);
613
+	struct net *net = sock_net(sk);
614
+	int err;
529615 
616
+	if (freq->flr_flags & IPV6_FL_F_REFLECT) {
617
+		if (net->ipv6.sysctl.flowlabel_consistency) {
618
+			net_info_ratelimited("Can not set IPV6_FL_F_REFLECT if flowlabel_consistency sysctl is enable\n");
619
+			return -EPERM;
620
+		}
530621 
531
-	if (optlen < sizeof(freq))
622
+		if (sk->sk_protocol != IPPROTO_TCP)
623
+			return -ENOPROTOOPT;
624
+		np->repflow = 1;
625
+		return 0;
626
+	}
627
+
628
+	if (freq->flr_label & ~IPV6_FLOWLABEL_MASK)
532629 		return -EINVAL;
630
+	if (net->ipv6.sysctl.flowlabel_state_ranges &&
631
+	    (freq->flr_label & IPV6_FLOWLABEL_STATELESS_FLAG))
632
+		return -ERANGE;
533633 
534
-	if (copy_from_user(&freq, optval, sizeof(freq)))
535
-		return -EFAULT;
634
+	fl = fl_create(net, sk, freq, optval, optlen, &err);
635
+	if (!fl)
636
+		return err;
536637 
537
-	switch (freq.flr_action) {
538
-	case IPV6_FL_A_PUT:
539
-		if (freq.flr_flags & IPV6_FL_F_REFLECT) {
540
-			if (sk->sk_protocol != IPPROTO_TCP)
541
-				return -ENOPROTOOPT;
542
-			if (!np->repflow)
543
-				return -ESRCH;
544
-			np->flow_label = 0;
545
-			np->repflow = 0;
546
-			return 0;
547
-		}
548
-		spin_lock_bh(&ip6_sk_fl_lock);
549
-		for (sflp = &np->ipv6_fl_list;
550
-		     (sfl = rcu_dereference_protected(*sflp,
551
-						      lockdep_is_held(&ip6_sk_fl_lock))) != NULL;
552
-		     sflp = &sfl->next) {
553
-			if (sfl->fl->label == freq.flr_label) {
554
-				if (freq.flr_label == (np->flow_label&IPV6_FLOWLABEL_MASK))
555
-					np->flow_label &= ~IPV6_FLOWLABEL_MASK;
556
-				*sflp = sfl->next;
557
-				spin_unlock_bh(&ip6_sk_fl_lock);
558
-				fl_release(sfl->fl);
559
-				kfree_rcu(sfl, rcu);
560
-				return 0;
561
-			}
562
-		}
563
-		spin_unlock_bh(&ip6_sk_fl_lock);
564
-		return -ESRCH;
638
+	sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL);
565639 
566
-	case IPV6_FL_A_RENEW:
640
+	if (freq->flr_label) {
641
+		err = -EEXIST;
567642 		rcu_read_lock_bh();
568643 		for_each_sk_fl_rcu(np, sfl) {
569
-			if (sfl->fl->label == freq.flr_label) {
570
-				err = fl6_renew(sfl->fl, freq.flr_linger, freq.flr_expires);
571
-				rcu_read_unlock_bh();
572
-				return err;
644
+			if (sfl->fl->label == freq->flr_label) {
645
+				if (freq->flr_flags & IPV6_FL_F_EXCL) {
646
+					rcu_read_unlock_bh();
647
+					goto done;
648
+				}
649
+				fl1 = sfl->fl;
650
+				if (!atomic_inc_not_zero(&fl1->users))
651
+					fl1 = NULL;
652
+				break;
573653 			}
574654 		}
575655 		rcu_read_unlock_bh();
576656 
577
-		if (freq.flr_share == IPV6_FL_S_NONE &&
578
-		    ns_capable(net->user_ns, CAP_NET_ADMIN)) {
579
-			fl = fl_lookup(net, freq.flr_label);
580
-			if (fl) {
581
-				err = fl6_renew(fl, freq.flr_linger, freq.flr_expires);
582
-				fl_release(fl);
583
-				return err;
584
-			}
585
-		}
586
-		return -ESRCH;
587
-
588
-	case IPV6_FL_A_GET:
589
-		if (freq.flr_flags & IPV6_FL_F_REFLECT) {
590
-			struct net *net = sock_net(sk);
591
-			if (net->ipv6.sysctl.flowlabel_consistency) {
592
-				net_info_ratelimited("Can not set IPV6_FL_F_REFLECT if flowlabel_consistency sysctl is enable\n");
593
-				return -EPERM;
594
-			}
595
-
596
-			if (sk->sk_protocol != IPPROTO_TCP)
597
-				return -ENOPROTOOPT;
598
-
599
-			np->repflow = 1;
600
-			return 0;
601
-		}
602
-
603
-		if (freq.flr_label & ~IPV6_FLOWLABEL_MASK)
604
-			return -EINVAL;
605
-
606
-		if (net->ipv6.sysctl.flowlabel_state_ranges &&
607
-		    (freq.flr_label & IPV6_FLOWLABEL_STATELESS_FLAG))
608
-			return -ERANGE;
609
-
610
-		fl = fl_create(net, sk, &freq, optval, optlen, &err);
611
-		if (!fl)
612
-			return err;
613
-		sfl1 = kmalloc(sizeof(*sfl1), GFP_KERNEL);
614
-
615
-		if (freq.flr_label) {
616
-			err = -EEXIST;
617
-			rcu_read_lock_bh();
618
-			for_each_sk_fl_rcu(np, sfl) {
619
-				if (sfl->fl->label == freq.flr_label) {
620
-					if (freq.flr_flags&IPV6_FL_F_EXCL) {
621
-						rcu_read_unlock_bh();
622
-						goto done;
623
-					}
624
-					fl1 = sfl->fl;
625
-					if (!atomic_inc_not_zero(&fl1->users))
626
-						fl1 = NULL;
627
-					break;
628
-				}
629
-			}
630
-			rcu_read_unlock_bh();
631
-
632
-			if (!fl1)
633
-				fl1 = fl_lookup(net, freq.flr_label);
634
-			if (fl1) {
657
+		if (!fl1)
658
+			fl1 = fl_lookup(net, freq->flr_label);
659
+		if (fl1) {
635660 recheck:
636
-				err = -EEXIST;
637
-				if (freq.flr_flags&IPV6_FL_F_EXCL)
638
-					goto release;
639
-				err = -EPERM;
640
-				if (fl1->share == IPV6_FL_S_EXCL ||
641
-				    fl1->share != fl->share ||
642
-				    ((fl1->share == IPV6_FL_S_PROCESS) &&
643
-				     (fl1->owner.pid != fl->owner.pid)) ||
644
-				    ((fl1->share == IPV6_FL_S_USER) &&
645
-				     !uid_eq(fl1->owner.uid, fl->owner.uid)))
646
-					goto release;
661
+			err = -EEXIST;
662
+			if (freq->flr_flags&IPV6_FL_F_EXCL)
663
+				goto release;
664
+			err = -EPERM;
665
+			if (fl1->share == IPV6_FL_S_EXCL ||
666
+			    fl1->share != fl->share ||
667
+			    ((fl1->share == IPV6_FL_S_PROCESS) &&
668
+			     (fl1->owner.pid != fl->owner.pid)) ||
669
+			    ((fl1->share == IPV6_FL_S_USER) &&
670
+			     !uid_eq(fl1->owner.uid, fl->owner.uid)))
671
+				goto release;
647672 
648
-				err = -ENOMEM;
649
-				if (!sfl1)
650
-					goto release;
651
-				if (fl->linger > fl1->linger)
652
-					fl1->linger = fl->linger;
653
-				if ((long)(fl->expires - fl1->expires) > 0)
654
-					fl1->expires = fl->expires;
655
-				fl_link(np, sfl1, fl1);
656
-				fl_free(fl);
657
-				return 0;
673
+			err = -ENOMEM;
674
+			if (!sfl1)
675
+				goto release;
676
+			if (fl->linger > fl1->linger)
677
+				fl1->linger = fl->linger;
678
+			if ((long)(fl->expires - fl1->expires) > 0)
679
+				fl1->expires = fl->expires;
680
+			fl_link(np, sfl1, fl1);
681
+			fl_free(fl);
682
+			return 0;
658683 
659684 release:
660
-				fl_release(fl1);
661
-				goto done;
662
-			}
685
+			fl_release(fl1);
686
+			goto done;
663687 		}
664
-		err = -ENOENT;
665
-		if (!(freq.flr_flags&IPV6_FL_F_CREATE))
666
-			goto done;
688
+	}
689
+	err = -ENOENT;
690
+	if (!(freq->flr_flags & IPV6_FL_F_CREATE))
691
+		goto done;
667692 
668
-		err = -ENOMEM;
669
-		if (!sfl1)
670
-			goto done;
693
+	err = -ENOMEM;
694
+	if (!sfl1)
695
+		goto done;
671696 
672
-		err = mem_check(sk);
673
-		if (err != 0)
674
-			goto done;
697
+	err = mem_check(sk);
698
+	if (err != 0)
699
+		goto done;
675700 
676
-		fl1 = fl_intern(net, fl, freq.flr_label);
677
-		if (fl1)
678
-			goto recheck;
701
+	fl1 = fl_intern(net, fl, freq->flr_label);
702
+	if (fl1)
703
+		goto recheck;
679704 
680
-		if (!freq.flr_label) {
681
-			if (copy_to_user(&((struct in6_flowlabel_req __user *) optval)->flr_label,
682
-					 &fl->label, sizeof(fl->label))) {
683
-				/* Intentionally ignore fault. */
684
-			}
705
+	if (!freq->flr_label) {
706
+		size_t offset = offsetof(struct in6_flowlabel_req, flr_label);
707
+
708
+		if (copy_to_sockptr_offset(optval, offset, &fl->label,
709
+				sizeof(fl->label))) {
710
+			/* Intentionally ignore fault. */
685711 		}
686
-
687
-		fl_link(np, sfl1, fl);
688
-		return 0;
689
-
690
-	default:
691
-		return -EINVAL;
692712 	}
693713 
714
+	fl_link(np, sfl1, fl);
715
+	return 0;
694716 done:
695717 	fl_free(fl);
696718 	kfree(sfl1);
697719 	return err;
720
+}
721
+
722
+int ipv6_flowlabel_opt(struct sock *sk, sockptr_t optval, int optlen)
723
+{
724
+	struct in6_flowlabel_req freq;
725
+
726
+	if (optlen < sizeof(freq))
727
+		return -EINVAL;
728
+	if (copy_from_sockptr(&freq, optval, sizeof(freq)))
729
+		return -EFAULT;
730
+
731
+	switch (freq.flr_action) {
732
+	case IPV6_FL_A_PUT:
733
+		return ipv6_flowlabel_put(sk, &freq);
734
+	case IPV6_FL_A_RENEW:
735
+		return ipv6_flowlabel_renew(sk, &freq);
736
+	case IPV6_FL_A_GET:
737
+		return ipv6_flowlabel_get(sk, &freq, optval, optlen);
738
+	default:
739
+		return -EINVAL;
740
+	}
698741 }
699742 
700743 #ifdef CONFIG_PROC_FS
....@@ -762,7 +805,7 @@
762805 {
763806 	struct ip6fl_iter_state *state = ip6fl_seq_private(seq);
764807 
765
-	state->pid_ns = proc_pid_ns(file_inode(seq->file));
808
+	state->pid_ns = proc_pid_ns(file_inode(seq->file)->i_sb);
766809 
767810 	rcu_read_lock_bh();
768811 	return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
....@@ -858,6 +901,7 @@
858901 
859902 void ip6_flowlabel_cleanup(void)
860903 {
904
+	static_key_deferred_flush(&ipv6_flowlabel_exclusive);
861905 	del_timer(&ip6_fl_gc_timer);
862906 	unregister_pernet_subsys(&ip6_flowlabel_net_ops);
863907 }