Merge branch 'sysctl-data-races'

Kuniyuki Iwashima says:

====================
sysctl: Fix data-races around ipv4_table.

A sysctl variable is accessed concurrently, and there is always a chance
of data-race. So, all readers and writers need some basic protection to
avoid load/store-tearing.

The first half of this series changes some proc handlers used in ipv4_table
to use READ_ONCE() and WRITE_ONCE() internally to fix data-races on the
sysctl side. Then, the second half adds READ_ONCE() to the other readers
of ipv4_table.

Changes:
v2:
* Drop some changes that makes backporting difficult
* First cleanup patch
* Lockless helpers and .proc_handler changes
* Drop the tracing part for .sysctl_mem
* Steve already posted a fix
* Drop int-to-bool change for cipso
* Should be posted to net-next later
* Drop proc_dobool() change
* Can be included in another series

v1: https://lore.kernel.org/netdev/[e-mail]/
====================

Signed-off-by: David S. Miller <[e-mail]>

Documentation/networking/ip-sysctl.rst

···
 cipso_cache_bucket_size - INTEGER
 	The CIPSO label cache consists of a fixed size hash table with each
 	hash bucket containing a number of cache entries.  This variable limits
-	the number of entries in each hash bucket; the larger the value the
+	the number of entries in each hash bucket; the larger the value is, the
 	more CIPSO label mappings that can be cached.  When the number of
 	entries in a given hash bucket reaches this limit adding new entries
 	causes the oldest entry in the bucket to be removed to make room.

include/net/sock.h

···
 /* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */
 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
 {
-	long val = sk->sk_prot->sysctl_mem[index];
+	long val = READ_ONCE(sk->sk_prot->sysctl_mem[index]);
 
 #if PAGE_SIZE > SK_MEM_QUANTUM
 	val <<= PAGE_SHIFT - SK_MEM_QUANTUM_SHIFT;

kernel/sysctl.c

···
 		if (*negp) {
 			if (*lvalp > (unsigned long) INT_MAX + 1)
 				return -EINVAL;
-			*valp = -*lvalp;
+			WRITE_ONCE(*valp, -*lvalp);
 		} else {
 			if (*lvalp > (unsigned long) INT_MAX)
 				return -EINVAL;
-			*valp = *lvalp;
+			WRITE_ONCE(*valp, *lvalp);
 		}
 	} else {
-		int val = *valp;
+		int val = READ_ONCE(*valp);
 		if (val < 0) {
 			*negp = true;
 			*lvalp = -(unsigned long)val;
···
 	if (write) {
 		if (*lvalp > UINT_MAX)
 			return -EINVAL;
-		*valp = *lvalp;
+		WRITE_ONCE(*valp, *lvalp);
 	} else {
-		unsigned int val = *valp;
+		unsigned int val = READ_ONCE(*valp);
 		*lvalp = (unsigned long)val;
 	}
 	return 0;
···
 		if ((param->min && *param->min > tmp) ||
 		    (param->max && *param->max < tmp))
 			return -EINVAL;
-		*valp = tmp;
+		WRITE_ONCE(*valp, tmp);
 	}
 
 	return 0;
···
 		    (param->max && *param->max < tmp))
 			return -ERANGE;
 
-		*valp = tmp;
+		WRITE_ONCE(*valp, tmp);
 	}
 
 	return 0;
···
 				err = -EINVAL;
 				break;
 			}
-			*i = val;
+			WRITE_ONCE(*i, val);
 		} else {
-			val = convdiv * (*i) / convmul;
+			val = convdiv * READ_ONCE(*i) / convmul;
 			if (!first)
 				proc_put_char(&buffer, &left, '\t');
 			proc_put_long(&buffer, &left, val, false);
···
 	if (write) {
 		if (*lvalp > INT_MAX / HZ)
 			return 1;
-		*valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
+		if (*negp)
+			WRITE_ONCE(*valp, -*lvalp * HZ);
+		else
+			WRITE_ONCE(*valp, *lvalp * HZ);
 	} else {
-		int val = *valp;
+		int val = READ_ONCE(*valp);
 		unsigned long lval;
 		if (val < 0) {
 			*negp = true;

net/ipv4/cipso_ipv4.c

···
 	struct cipso_v4_map_cache_entry *prev_entry = NULL;
 	u32 hash;
 
-	if (!cipso_v4_cache_enabled)
+	if (!READ_ONCE(cipso_v4_cache_enabled))
 		return -ENOENT;
 
 	hash = cipso_v4_map_cache_hash(key, key_len);
···
 int cipso_v4_cache_add(const unsigned char *cipso_ptr,
 		       const struct netlbl_lsm_secattr *secattr)
 {
+	int bkt_size = READ_ONCE(cipso_v4_cache_bucketsize);
 	int ret_val = -EPERM;
 	u32 bkt;
 	struct cipso_v4_map_cache_entry *entry = NULL;
 	struct cipso_v4_map_cache_entry *old_entry = NULL;
 	u32 cipso_ptr_len;
 
-	if (!cipso_v4_cache_enabled || cipso_v4_cache_bucketsize <= 0)
+	if (!READ_ONCE(cipso_v4_cache_enabled) || bkt_size <= 0)
 		return 0;
 
 	cipso_ptr_len = cipso_ptr[1];
···
 
 	bkt = entry->hash & (CIPSO_V4_CACHE_BUCKETS - 1);
 	spin_lock_bh(&cipso_v4_cache[bkt].lock);
-	if (cipso_v4_cache[bkt].size < cipso_v4_cache_bucketsize) {
+	if (cipso_v4_cache[bkt].size < bkt_size) {
 		list_add(&entry->list, &cipso_v4_cache[bkt].list);
 		cipso_v4_cache[bkt].size += 1;
 	} else {
···
 		/* This will send packets using the "optimized" format when
 		 * possible as specified in  section 3.4.2.6 of the
 		 * CIPSO draft. */
-		if (cipso_v4_rbm_optfmt && ret_val > 0 && ret_val <= 10)
+		if (READ_ONCE(cipso_v4_rbm_optfmt) && ret_val > 0 &&
+		    ret_val <= 10)
 			tag_len = 14;
 		else
 			tag_len = 4 + ret_val;
···
 			 * all the CIPSO validations here but it doesn't
 			 * really specify _exactly_ what we need to validate
 			 * ... so, just make it a sysctl tunable. */
-			if (cipso_v4_rbm_strictvalid) {
+			if (READ_ONCE(cipso_v4_rbm_strictvalid)) {
 				if (cipso_v4_map_lvl_valid(doi_def,
 							   tag[3]) < 0) {
 					err_offset = opt_iter + 3;

net/ipv4/fib_trie.c

···
 		tn = container_of(head, struct tnode, rcu)->kv;
 	}
 
-	if (tnode_free_size >= sysctl_fib_sync_mem) {
+	if (tnode_free_size >= READ_ONCE(sysctl_fib_sync_mem)) {
 		tnode_free_size = 0;
 		synchronize_rcu();
 	}

net/ipv4/icmp.c

···
 	spin_lock(&icmp_global.lock);
 	delta = min_t(u32, now - icmp_global.stamp, HZ);
 	if (delta >= HZ / 50) {
-		incr = sysctl_icmp_msgs_per_sec * delta / HZ ;
+		incr = READ_ONCE(sysctl_icmp_msgs_per_sec) * delta / HZ;
 		if (incr)
 			WRITE_ONCE(icmp_global.stamp, now);
 	}
-	credit = min_t(u32, icmp_global.credit + incr, sysctl_icmp_msgs_burst);
+	credit = min_t(u32, icmp_global.credit + incr,
+		       READ_ONCE(sysctl_icmp_msgs_burst));
 	if (credit) {
 		/* We want to use a credit of one in average, but need to randomize
 		 * it for security reasons.

net/ipv4/inetpeer.c

···
 			 struct inet_peer *gc_stack[],
 			 unsigned int gc_cnt)
 {
+	int peer_threshold, peer_maxttl, peer_minttl;
 	struct inet_peer *p;
 	__u32 delta, ttl;
 	int i;
 
-	if (base->total >= inet_peer_threshold)
+	peer_threshold = READ_ONCE(inet_peer_threshold);
+	peer_maxttl = READ_ONCE(inet_peer_maxttl);
+	peer_minttl = READ_ONCE(inet_peer_minttl);
+
+	if (base->total >= peer_threshold)
 		ttl = 0; /* be aggressive */
 	else
-		ttl = inet_peer_maxttl
-				- (inet_peer_maxttl - inet_peer_minttl) / HZ *
-					base->total / inet_peer_threshold * HZ;
+		ttl = peer_maxttl - (peer_maxttl - peer_minttl) / HZ *
+			base->total / peer_threshold * HZ;
 	for (i = 0; i < gc_cnt; i++) {
 		p = gc_stack[i];
 

net/ipv4/tcp.c

···
 
 static bool tcp_too_many_orphans(int shift)
 {
-	return READ_ONCE(tcp_orphan_cache) << shift > sysctl_tcp_max_orphans;
+	return READ_ONCE(tcp_orphan_cache) << shift >
+		READ_ONCE(sysctl_tcp_max_orphans);
 }
 
 bool tcp_check_oom(struct sock *sk, int shift)