Skip to contentSkip to conversation

arp: Random clean up and RCU conversion for ioctl(SIOCGARP).

PATCH Applied Kuniyuki Iwashima 2 participants 9 messages Apr 30, 2024 – May 2, 2024

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v3 cover arp: Random clean up and RCU conversion for ioctl(SIOCGARP).

arp_ioctl() holds rtnl_lock() regardless of cmd (SIOCDARP, SIOCSARP, and SIOCGARP) to get net_device by __dev_get_by_name() and copy dev->name safely.

In the SIOCGARP path, arp_req_get() calls neigh_lookup(), which looks up a neighbour entry under RCU.

This series cleans up ioctl() code a bit and extends the RCU section not to take rtnl_lock() and instead use dev_get_by_name_rcu() and netdev_copy_name() for SIOCGARP.

Changes:
v3:

    Add Patch 6 to read dev->name safely under seqlock.

v2: https://lore.kernel.org/netdev/[email protected]/

    Patch 5: s/!IS_ERR/IS_ERR/ in arp_req_delete().

v1: https://lore.kernel.org/netdev/[email protected]/

Kuniyuki Iwashima (7):
arp: Move ATF_COM setting in arp_req_set().
arp: Validate netmask earlier for SIOCDARP and SIOCSARP in

    arp_ioctl().
  arp: Factorise ip_route_output() call in arp_req_set() and
    arp_req_delete().

arp: Remove a nest in arp_req_get().
arp: Get dev after calling arp_req_(delete|set|get)().
net: Protect dev->name by seqlock.
arp: Convert ioctl(SIOCGARP) to RCU.

 include/linux/netdevice.h |   1 +
 net/core/dev.c            |  27 ++++-
 net/ipv4/arp.c            | 203 +++++++++++++++++++++++---------------
 3 files changed, 147 insertions(+), 84 deletions(-)
Signature · 1 line
2.30.2
#1 lore raw

PATCH v3 1/7 arp: Move ATF_COM setting in arp_req_set().

In arp_req_set(), if ATF_PERM is set in arpreq.arp_flags, ATF_COM is set automatically.

The flag will be used later for neigh_update() only when a neighbour entry is found.

Let's set ATF_COM just before calling neigh_update().

---
 net/ipv4/arp.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
net/ipv4/arp.c +6 -3
index ab82ca104496..3093374165fa 100644
@@ -1054,8 +1054,7 @@ static int arp_req_set(struct net *net, struct arpreq *r,
 		return arp_req_set_public(net, r, dev);
 
 	ip = ((struct sockaddr_in *)&r->arp_pa)->sin_addr.s_addr;
-	if (r->arp_flags & ATF_PERM)
-		r->arp_flags |= ATF_COM;
+
 	if (!dev) {
 		struct rtable *rt = ip_route_output(net, ip, 0, 0, 0,
 						    RT_SCOPE_LINK);
@@ -1092,8 +1091,12 @@ static int arp_req_set(struct net *net, struct arpreq *r,
 	err = PTR_ERR(neigh);
 	if (!IS_ERR(neigh)) {
 		unsigned int state = NUD_STALE;
-		if (r->arp_flags & ATF_PERM)
+
+		if (r->arp_flags & ATF_PERM) {
+			r->arp_flags |= ATF_COM;
 			state = NUD_PERMANENT;
+		}
+
 		err = neigh_update(neigh, (r->arp_flags & ATF_COM) ?
 				   r->arp_ha.sa_data : NULL, state,
 				   NEIGH_UPDATE_F_OVERRIDE |
Signature · 1 line
2.30.2
#2 lore raw

PATCH v3 2/7 arp: Validate netmask earlier for SIOCDARP and SIOCSARP in arp_ioctl().

In reply to Kuniyuki Iwashima

When ioctl(SIOCDARP/SIOCSARP) is issued with ATF_PUBL, r.arp_netmask must be 0.0.0.0 or 255.255.255.255.

Currently, the netmask is validated in arp_req_delete_public() or arp_req_set_public() under rtnl_lock().

We have ATF_NETMASK test in arp_ioctl() before holding rtnl_lock(), so let's move the netmask validation there.

---
 net/ipv4/arp.c | 27 +++++++++++++++------------
 1 file changed, 15 insertions(+), 12 deletions(-)
net/ipv4/arp.c +15 -12
index 3093374165fa..b20a5771d069 100644
@@ -1023,11 +1023,8 @@ static int arp_req_set_proxy(struct net *net, struct net_device *dev, int on)
 static int arp_req_set_public(struct net *net, struct arpreq *r,
 		struct net_device *dev)
 {
-	__be32 ip = ((struct sockaddr_in *)&r->arp_pa)->sin_addr.s_addr;
 	__be32 mask = ((struct sockaddr_in *)&r->arp_netmask)->sin_addr.s_addr;
 
-	if (mask && mask != htonl(0xFFFFFFFF))
-		return -EINVAL;
 	if (!dev && (r->arp_flags & ATF_COM)) {
 		dev = dev_getbyhwaddr_rcu(net, r->arp_ha.sa_family,
 				      r->arp_ha.sa_data);
@@ -1035,6 +1032,8 @@ static int arp_req_set_public(struct net *net, struct arpreq *r,
 			return -ENODEV;
 	}
 	if (mask) {
+		__be32 ip = ((struct sockaddr_in *)&r->arp_pa)->sin_addr.s_addr;
+
 		if (!pneigh_lookup(&arp_tbl, net, &ip, dev, 1))
 			return -ENOBUFS;
 		return 0;
@@ -1171,14 +1170,13 @@ int arp_invalidate(struct net_device *dev, __be32 ip, bool force)
 static int arp_req_delete_public(struct net *net, struct arpreq *r,
 		struct net_device *dev)
 {
-	__be32 ip = ((struct sockaddr_in *) &r->arp_pa)->sin_addr.s_addr;
 	__be32 mask = ((struct sockaddr_in *)&r->arp_netmask)->sin_addr.s_addr;
 
-	if (mask == htonl(0xFFFFFFFF))
-		return pneigh_delete(&arp_tbl, net, &ip, dev);
+	if (mask) {
+		__be32 ip = ((struct sockaddr_in *)&r->arp_pa)->sin_addr.s_addr;
 
-	if (mask)
-		return -EINVAL;
+		return pneigh_delete(&arp_tbl, net, &ip, dev);
+	}
 
 	return arp_req_set_proxy(net, dev, 0);
 }
@@ -1211,9 +1209,10 @@ static int arp_req_delete(struct net *net, struct arpreq *r,
 
 int arp_ioctl(struct net *net, unsigned int cmd, void __user *arg)
 {
-	int err;
-	struct arpreq r;
 	struct net_device *dev = NULL;
+	struct arpreq r;
+	__be32 *netmask;
+	int err;
 
 	switch (cmd) {
 	case SIOCDARP:
@@ -1236,9 +1235,13 @@ int arp_ioctl(struct net *net, unsigned int cmd, void __user *arg)
 	if (!(r.arp_flags & ATF_PUBL) &&
 	    (r.arp_flags & (ATF_NETMASK | ATF_DONTPUB)))
 		return -EINVAL;
+
+	netmask = &((struct sockaddr_in *)&r.arp_netmask)->sin_addr.s_addr;
 	if (!(r.arp_flags & ATF_NETMASK))
-		((struct sockaddr_in *)&r.arp_netmask)->sin_addr.s_addr =
-							   htonl(0xFFFFFFFFUL);
+		*netmask = htonl(0xFFFFFFFFUL);
+	else if (*netmask && *netmask != htonl(0xFFFFFFFFUL))
+		return -EINVAL;
+
 	rtnl_lock();
 	if (r.arp_dev[0]) {
 		err = -ENODEV;
Signature · 1 line
2.30.2
#3 lore raw

PATCH v3 3/7 arp: Factorise ip_route_output() call in arp_req_set() and arp_req_delete().

In reply to Kuniyuki Iwashima

When ioctl(SIOCDARP/SIOCSARP) is issued for non-proxy entry (no ATF_COM) without arpreq.arp_dev[] set, arp_req_set() and arp_req_delete() looks up dev based on IPv4 address by ip_route_output().

Let's factorise the same code as arp_req_dev().

---
 net/ipv4/arp.c | 50 ++++++++++++++++++++++++++++++--------------------
 1 file changed, 30 insertions(+), 20 deletions(-)
net/ipv4/arp.c +30 -20
index b20a5771d069..ac3e15799c2f 100644
@@ -1003,6 +1003,27 @@ static int arp_rcv(struct sk_buff *skb, struct net_device *dev,
  *	User level interface (ioctl)
  */
 
+static struct net_device *arp_req_dev(struct net *net, struct arpreq *r)
+{
+	struct net_device *dev;
+	struct rtable *rt;
+	__be32 ip;
+
+	ip = ((struct sockaddr_in *)&r->arp_pa)->sin_addr.s_addr;
+
+	rt = ip_route_output(net, ip, 0, 0, 0, RT_SCOPE_LINK);
+	if (IS_ERR(rt))
+		return ERR_CAST(rt);
+
+	dev = rt->dst.dev;
+	ip_rt_put(rt);
+
+	if (!dev)
+		return ERR_PTR(-EINVAL);
+
+	return dev;
+}
+
 /*
  *	Set (create) an ARP cache entry.
  */
@@ -1045,25 +1066,17 @@ static int arp_req_set_public(struct net *net, struct arpreq *r,
 static int arp_req_set(struct net *net, struct arpreq *r,
 		       struct net_device *dev)
 {
-	__be32 ip;
 	struct neighbour *neigh;
+	__be32 ip;
 	int err;
 
 	if (r->arp_flags & ATF_PUBL)
 		return arp_req_set_public(net, r, dev);
 
-	ip = ((struct sockaddr_in *)&r->arp_pa)->sin_addr.s_addr;
-
 	if (!dev) {
-		struct rtable *rt = ip_route_output(net, ip, 0, 0, 0,
-						    RT_SCOPE_LINK);
-
-		if (IS_ERR(rt))
-			return PTR_ERR(rt);
-		dev = rt->dst.dev;
-		ip_rt_put(rt);
-		if (!dev)
-			return -EINVAL;
+		dev = arp_req_dev(net, r);
+		if (IS_ERR(dev))
+			return PTR_ERR(dev);
 	}
 	switch (dev->type) {
 #if IS_ENABLED(CONFIG_FDDI)
@@ -1086,6 +1099,8 @@ static int arp_req_set(struct net *net, struct arpreq *r,
 		break;
 	}
 
+	ip = ((struct sockaddr_in *)&r->arp_pa)->sin_addr.s_addr;
+
 	neigh = __neigh_lookup_errno(&arp_tbl, &ip, dev);
 	err = PTR_ERR(neigh);
 	if (!IS_ERR(neigh)) {
@@ -1191,14 +1206,9 @@ static int arp_req_delete(struct net *net, struct arpreq *r,
 
 	ip = ((struct sockaddr_in *)&r->arp_pa)->sin_addr.s_addr;
 	if (!dev) {
-		struct rtable *rt = ip_route_output(net, ip, 0, 0, 0,
-						    RT_SCOPE_LINK);
-		if (IS_ERR(rt))
-			return PTR_ERR(rt);
-		dev = rt->dst.dev;
-		ip_rt_put(rt);
-		if (!dev)
-			return -EINVAL;
+		dev = arp_req_dev(net, r);
+		if (IS_ERR(dev))
+			return PTR_ERR(dev);
 	}
 	return arp_invalidate(dev, ip, true);
 }
Signature · 1 line
2.30.2
#4 lore raw

PATCH v3 4/7 arp: Remove a nest in arp_req_get().

In reply to Kuniyuki Iwashima

This is a prep patch to make the following changes tidy.

No functional change intended.

---
 net/ipv4/arp.c | 31 ++++++++++++++++++-------------
 1 file changed, 18 insertions(+), 13 deletions(-)
net/ipv4/arp.c +18 -13
index ac3e15799c2f..60f633b24ec8 100644
@@ -1138,23 +1138,28 @@ static int arp_req_get(struct arpreq *r, struct net_device *dev)
 {
 	__be32 ip = ((struct sockaddr_in *) &r->arp_pa)->sin_addr.s_addr;
 	struct neighbour *neigh;
-	int err = -ENXIO;
 
 	neigh = neigh_lookup(&arp_tbl, &ip, dev);
-	if (neigh) {
-		if (!(READ_ONCE(neigh->nud_state) & NUD_NOARP)) {
-			read_lock_bh(&neigh->lock);
-			memcpy(r->arp_ha.sa_data, neigh->ha,
-			       min(dev->addr_len, sizeof(r->arp_ha.sa_data_min)));
-			r->arp_flags = arp_state_to_flags(neigh);
-			read_unlock_bh(&neigh->lock);
-			r->arp_ha.sa_family = dev->type;
-			strscpy(r->arp_dev, dev->name, sizeof(r->arp_dev));
-			err = 0;
-		}
+	if (!neigh)
+		return -ENXIO;
+
+	if (READ_ONCE(neigh->nud_state) & NUD_NOARP) {
 		neigh_release(neigh);
+		return -ENXIO;
 	}
-	return err;
+
+	read_lock_bh(&neigh->lock);
+	memcpy(r->arp_ha.sa_data, neigh->ha,
+	       min(dev->addr_len, sizeof(r->arp_ha.sa_data_min)));
+	r->arp_flags = arp_state_to_flags(neigh);
+	read_unlock_bh(&neigh->lock);
+
+	neigh_release(neigh);
+
+	r->arp_ha.sa_family = dev->type;
+	strscpy(r->arp_dev, dev->name, sizeof(r->arp_dev));
+
+	return 0;
 }
 
 int arp_invalidate(struct net_device *dev, __be32 ip, bool force)
Signature · 1 line
2.30.2
#5 lore raw

PATCH v3 5/7 arp: Get dev after calling arp_req_(delete|set|get)().

In reply to Kuniyuki Iwashima

arp_ioctl() holds rtnl_lock() first regardless of cmd (SIOCDARP, SIOCSARP, and SIOCGARP) to get net_device by __dev_get_by_name() and copy dev->name safely.

In the SIOCGARP path, arp_req_get() calls neigh_lookup(), which looks up a neighbour entry under RCU.

We will extend the RCU section not to take rtnl_lock() and instead use dev_get_by_name_rcu() for SIOCGARP.

As a preparation, let's move __dev_get_by_name() into another function and call it from arp_req_delete(), arp_req_set(), and arp_req_get().

---
 net/ipv4/arp.c | 86 +++++++++++++++++++++++++++++---------------------
 1 file changed, 50 insertions(+), 36 deletions(-)
net/ipv4/arp.c +50 -36
index 60f633b24ec8..5034920be85a 100644
@@ -1003,12 +1003,36 @@ static int arp_rcv(struct sk_buff *skb, struct net_device *dev,
  *	User level interface (ioctl)
  */
 
+static struct net_device *arp_req_dev_by_name(struct net *net, struct arpreq *r)
+{
+	struct net_device *dev;
+
+	dev = __dev_get_by_name(net, r->arp_dev);
+	if (!dev)
+		return ERR_PTR(-ENODEV);
+
+	/* Mmmm... It is wrong... ARPHRD_NETROM == 0 */
+	if (!r->arp_ha.sa_family)
+		r->arp_ha.sa_family = dev->type;
+
+	if ((r->arp_flags & ATF_COM) && r->arp_ha.sa_family != dev->type)
+		return ERR_PTR(-EINVAL);
+
+	return dev;
+}
+
 static struct net_device *arp_req_dev(struct net *net, struct arpreq *r)
 {
 	struct net_device *dev;
 	struct rtable *rt;
 	__be32 ip;
 
+	if (r->arp_dev[0])
+		return arp_req_dev_by_name(net, r);
+
+	if (r->arp_flags & ATF_PUBL)
+		return NULL;
+
 	ip = ((struct sockaddr_in *)&r->arp_pa)->sin_addr.s_addr;
 
 	rt = ip_route_output(net, ip, 0, 0, 0, RT_SCOPE_LINK);
@@ -1063,21 +1087,20 @@ static int arp_req_set_public(struct net *net, struct arpreq *r,
 	return arp_req_set_proxy(net, dev, 1);
 }
 
-static int arp_req_set(struct net *net, struct arpreq *r,
-		       struct net_device *dev)
+static int arp_req_set(struct net *net, struct arpreq *r)
 {
 	struct neighbour *neigh;
+	struct net_device *dev;
 	__be32 ip;
 	int err;
 
+	dev = arp_req_dev(net, r);
+	if (IS_ERR(dev))
+		return PTR_ERR(dev);
+
 	if (r->arp_flags & ATF_PUBL)
 		return arp_req_set_public(net, r, dev);
 
-	if (!dev) {
-		dev = arp_req_dev(net, r);
-		if (IS_ERR(dev))
-			return PTR_ERR(dev);
-	}
 	switch (dev->type) {
 #if IS_ENABLED(CONFIG_FDDI)
 	case ARPHRD_FDDI:
@@ -1134,10 +1157,18 @@ static unsigned int arp_state_to_flags(struct neighbour *neigh)
  *	Get an ARP cache entry.
  */
 
-static int arp_req_get(struct arpreq *r, struct net_device *dev)
+static int arp_req_get(struct net *net, struct arpreq *r)
 {
 	__be32 ip = ((struct sockaddr_in *) &r->arp_pa)->sin_addr.s_addr;
 	struct neighbour *neigh;
+	struct net_device *dev;
+
+	if (!r->arp_dev[0])
+		return -ENODEV;
+
+	dev = arp_req_dev_by_name(net, r);
+	if (IS_ERR(dev))
+		return PTR_ERR(dev);
 
 	neigh = neigh_lookup(&arp_tbl, &ip, dev);
 	if (!neigh)
@@ -1201,20 +1232,20 @@ static int arp_req_delete_public(struct net *net, struct arpreq *r,
 	return arp_req_set_proxy(net, dev, 0);
 }
 
-static int arp_req_delete(struct net *net, struct arpreq *r,
-			  struct net_device *dev)
+static int arp_req_delete(struct net *net, struct arpreq *r)
 {
+	struct net_device *dev;
 	__be32 ip;
 
+	dev = arp_req_dev(net, r);
+	if (IS_ERR(dev))
+		return PTR_ERR(dev);
+
 	if (r->arp_flags & ATF_PUBL)
 		return arp_req_delete_public(net, r, dev);
 
 	ip = ((struct sockaddr_in *)&r->arp_pa)->sin_addr.s_addr;
-	if (!dev) {
-		dev = arp_req_dev(net, r);
-		if (IS_ERR(dev))
-			return PTR_ERR(dev);
-	}
+
 	return arp_invalidate(dev, ip, true);
 }
 
@@ -1224,7 +1255,6 @@ static int arp_req_delete(struct net *net, struct arpreq *r,
 
 int arp_ioctl(struct net *net, unsigned int cmd, void __user *arg)
 {
-	struct net_device *dev = NULL;
 	struct arpreq r;
 	__be32 *netmask;
 	int err;
@@ -1258,35 +1288,19 @@ int arp_ioctl(struct net *net, unsigned int cmd, void __user *arg)
 		return -EINVAL;
 
 	rtnl_lock();
-	if (r.arp_dev[0]) {
-		err = -ENODEV;
-		dev = __dev_get_by_name(net, r.arp_dev);
-		if (!dev)
-			goto out;
-
-		/* Mmmm... It is wrong... ARPHRD_NETROM==0 */
-		if (!r.arp_ha.sa_family)
-			r.arp_ha.sa_family = dev->type;
-		err = -EINVAL;
-		if ((r.arp_flags & ATF_COM) && r.arp_ha.sa_family != dev->type)
-			goto out;
-	} else if (cmd == SIOCGARP) {
-		err = -ENODEV;
-		goto out;
-	}
 
 	switch (cmd) {
 	case SIOCDARP:
-		err = arp_req_delete(net, &r, dev);
+		err = arp_req_delete(net, &r);
 		break;
 	case SIOCSARP:
-		err = arp_req_set(net, &r, dev);
+		err = arp_req_set(net, &r);
 		break;
 	case SIOCGARP:
-		err = arp_req_get(&r, dev);
+		err = arp_req_get(net, &r);
 		break;
 	}
-out:
+
 	rtnl_unlock();
 	if (cmd == SIOCGARP && !err && copy_to_user(arg, &r, sizeof(r)))
 		err = -EFAULT;
Signature · 1 line
2.30.2
#6 lore raw

PATCH v3 6/7 net: Protect dev->name by seqlock.

In reply to Kuniyuki Iwashima

We will convert ioctl(SIOCGARP) to RCU, and then we need to copy dev->name which is currently protected by rtnl_lock().

This patch does the following:

1) Add seqlock netdev_rename_lock to protect dev->name

2) Add netdev_copy_name() that copies dev->name to buffer under netdev_rename_lock

3) Use netdev_copy_name() in netdev_get_name() and drop devnet_rename_sem

---
 include/linux/netdevice.h |  1 +
 net/core/dev.c            | 27 +++++++++++++++++++++++----
 2 files changed, 24 insertions(+), 4 deletions(-)
include/linux/netdevice.h +1 -0
index f849e7d110ed..41853424b41d 100644
@@ -3136,6 +3136,7 @@ struct net_device *netdev_get_by_name(struct net *net, const char *name,
 				      netdevice_tracker *tracker, gfp_t gfp);
 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
+void netdev_copy_name(struct net_device *dev, char *name);
 
 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
 				  unsigned short type,
net/core/dev.c +23 -4
index c9e59eff8ec8..01a9d6e451e2 100644
@@ -940,6 +940,18 @@ struct net_device *dev_get_by_napi_id(unsigned int napi_id)
 }
 EXPORT_SYMBOL(dev_get_by_napi_id);
 
+static DEFINE_SEQLOCK(netdev_rename_lock);
+
+void netdev_copy_name(struct net_device *dev, char *name)
+{
+	unsigned int seq;
+
+	do {
+		seq = read_seqbegin(&netdev_rename_lock);
+		strscpy(name, dev->name, IFNAMSIZ);
+	} while (read_seqretry(&netdev_rename_lock, seq));
+}
+
 /**
  *	netdev_get_name - get a netdevice name, knowing its ifindex.
  *	@net: network namespace
@@ -951,7 +963,6 @@ int netdev_get_name(struct net *net, char *name, int ifindex)
 	struct net_device *dev;
 	int ret;
 
-	down_read(&devnet_rename_sem);
 	rcu_read_lock();
 
 	dev = dev_get_by_index_rcu(net, ifindex);
@@ -960,12 +971,11 @@ int netdev_get_name(struct net *net, char *name, int ifindex)
 		goto out;
 	}
 
-	strcpy(name, dev->name);
+	netdev_copy_name(dev, name);
 
 	ret = 0;
 out:
 	rcu_read_unlock();
-	up_read(&devnet_rename_sem);
 	return ret;
 }
 
@@ -1217,7 +1227,10 @@ int dev_change_name(struct net_device *dev, const char *newname)
 
 	memcpy(oldname, dev->name, IFNAMSIZ);
 
+	write_seqlock(&netdev_rename_lock);
 	err = dev_get_valid_name(net, dev, newname);
+	write_sequnlock(&netdev_rename_lock);
+
 	if (err < 0) {
 		up_write(&devnet_rename_sem);
 		return err;
@@ -1257,7 +1270,9 @@ int dev_change_name(struct net_device *dev, const char *newname)
 		if (err >= 0) {
 			err = ret;
 			down_write(&devnet_rename_sem);
+			write_seqlock(&netdev_rename_lock);
 			memcpy(dev->name, oldname, IFNAMSIZ);
+			write_sequnlock(&netdev_rename_lock);
 			memcpy(oldname, newname, IFNAMSIZ);
 			WRITE_ONCE(dev->name_assign_type, old_assign_type);
 			old_assign_type = NET_NAME_RENAMED;
@@ -11404,8 +11419,12 @@ int __dev_change_net_namespace(struct net_device *dev, struct net *net,
 	dev_net_set(dev, net);
 	dev->ifindex = new_ifindex;
 
-	if (new_name[0]) /* Rename the netdev to prepared name */
+	if (new_name[0]) {
+		/* Rename the netdev to prepared name */
+		write_seqlock(&netdev_rename_lock);
 		strscpy(dev->name, new_name, IFNAMSIZ);
+		write_sequnlock(&netdev_rename_lock);
+	}
 
 	/* Fixup kobjects */
 	dev_set_uevent_suppress(&dev->dev, 1);
Signature · 1 line
2.30.2
#7 lore raw

PATCH v3 7/7 arp: Convert ioctl(SIOCGARP) to RCU.

In reply to Kuniyuki Iwashima

ioctl(SIOCGARP) holds rtnl_lock() to get netdev by __dev_get_by_name() and copy dev->name safely and calls neigh_lookup() later, which looks up a neighbour entry under RCU.

Let's replace __dev_get_by_name() with dev_get_by_name_rcu() and strscpy() with netdev_copy_name() to avoid locking rtnl_lock().

---
 net/ipv4/arp.c | 28 ++++++++++++++++++----------
 1 file changed, 18 insertions(+), 10 deletions(-)
net/ipv4/arp.c +18 -10
index 5034920be85a..11c1519b3699 100644
@@ -1003,11 +1003,15 @@ static int arp_rcv(struct sk_buff *skb, struct net_device *dev,
  *	User level interface (ioctl)
  */
 
-static struct net_device *arp_req_dev_by_name(struct net *net, struct arpreq *r)
+static struct net_device *arp_req_dev_by_name(struct net *net, struct arpreq *r,
+					      bool getarp)
 {
 	struct net_device *dev;
 
-	dev = __dev_get_by_name(net, r->arp_dev);
+	if (getarp)
+		dev = dev_get_by_name_rcu(net, r->arp_dev);
+	else
+		dev = __dev_get_by_name(net, r->arp_dev);
 	if (!dev)
 		return ERR_PTR(-ENODEV);
 
@@ -1028,7 +1032,7 @@ static struct net_device *arp_req_dev(struct net *net, struct arpreq *r)
 	__be32 ip;
 
 	if (r->arp_dev[0])
-		return arp_req_dev_by_name(net, r);
+		return arp_req_dev_by_name(net, r, false);
 
 	if (r->arp_flags & ATF_PUBL)
 		return NULL;
@@ -1166,7 +1170,7 @@ static int arp_req_get(struct net *net, struct arpreq *r)
 	if (!r->arp_dev[0])
 		return -ENODEV;
 
-	dev = arp_req_dev_by_name(net, r);
+	dev = arp_req_dev_by_name(net, r, true);
 	if (IS_ERR(dev))
 		return PTR_ERR(dev);
 
@@ -1188,7 +1192,7 @@ static int arp_req_get(struct net *net, struct arpreq *r)
 	neigh_release(neigh);
 
 	r->arp_ha.sa_family = dev->type;
-	strscpy(r->arp_dev, dev->name, sizeof(r->arp_dev));
+	netdev_copy_name(dev, r->arp_dev);
 
 	return 0;
 }
@@ -1287,23 +1291,27 @@ int arp_ioctl(struct net *net, unsigned int cmd, void __user *arg)
 	else if (*netmask && *netmask != htonl(0xFFFFFFFFUL))
 		return -EINVAL;
 
-	rtnl_lock();
-
 	switch (cmd) {
 	case SIOCDARP:
+		rtnl_lock();
 		err = arp_req_delete(net, &r);
+		rtnl_unlock();
 		break;
 	case SIOCSARP:
+		rtnl_lock();
 		err = arp_req_set(net, &r);
+		rtnl_unlock();
 		break;
 	case SIOCGARP:
+		rcu_read_lock();
 		err = arp_req_get(net, &r);
+		rcu_read_unlock();
+
+		if (!err && copy_to_user(arg, &r, sizeof(r)))
+			err = -EFAULT;
 		break;
 	}
 
-	rtnl_unlock();
-	if (cmd == SIOCGARP && !err && copy_to_user(arg, &r, sizeof(r)))
-		err = -EFAULT;
 	return err;
 }
 
Signature · 1 line
2.30.2
#8 lore raw
In reply to Kuniyuki Iwashima
This series was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This series was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:

Quoted 8 lines · On Mon, 29 Apr 2024 18:58:06 -0700 you wrote:

arp_ioctl() holds rtnl_lock() regardless of cmd (SIOCDARP, SIOCSARP, and SIOCGARP) to get net_device by __dev_get_by_name() and copy dev->name safely.

In the SIOCGARP path, arp_req_get() calls neigh_lookup(), which looks up a neighbour entry under RCU.

[...]

Here is the summary with links:
- [v3,net-next,1/7] arp: Move ATF_COM setting in arp_req_set().

    https://git.kernel.org/netdev/net-next/c/42033d0cfc86
  - [v3,net-next,2/7] arp: Validate netmask earlier for SIOCDARP and SIOCSARP in arp_ioctl().
    https://git.kernel.org/netdev/net-next/c/0592367424bb
  - [v3,net-next,3/7] arp: Factorise ip_route_output() call in arp_req_set() and arp_req_delete().
    https://git.kernel.org/netdev/net-next/c/f8696133f6aa
  - [v3,net-next,4/7] arp: Remove a nest in arp_req_get().
    https://git.kernel.org/netdev/net-next/c/51e9ba48d487
  - [v3,net-next,5/7] arp: Get dev after calling arp_req_(delete|set|get)().
    https://git.kernel.org/netdev/net-next/c/a428bfc77a4d
  - [v3,net-next,6/7] net: Protect dev->name by seqlock.
    https://git.kernel.org/netdev/net-next/c/0840556e5a3a
  - [v3,net-next,7/7] arp: Convert ioctl(SIOCGARP) to RCU.
    https://git.kernel.org/netdev/net-next/c/bf4ea58874df

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html