bpf: Introduce device-bound XDP programs

New flag BPF_F_XDP_DEV_BOUND_ONLY plus all the infra to have a way
to associate a netdev with a BPF program at load time.

netdevsim checks are dropped in favor of generic check in dev_xdp_attach.

Cc: John Fastabend <[e-mail]>
Cc: David Ahern <[e-mail]>
Cc: Martin KaFai Lau <[e-mail]>
Cc: Jakub Kicinski <[e-mail]>
Cc: Willem de Bruijn <[e-mail]>
Cc: Jesper Dangaard Brouer <[e-mail]>
Cc: Anatoly Burakov <[e-mail]>
Cc: Alexander Lobakin <[e-mail]>
Cc: Magnus Karlsson <[e-mail]>
Cc: Maryam Tahhan <[e-mail]>
Cc: [e-mail]
Cc: [e-mail]
Signed-off-by: Stanislav Fomichev <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]
Signed-off-by: Martin KaFai Lau <[e-mail]>

drivers/net/netdevsim/bpf.c

···
 		NSIM_EA(bpf->extack, "xdpoffload of non-bound program");
 		return -EINVAL;
 	}
-	if (!bpf_offload_dev_match(bpf->prog, ns->netdev)) {
-		NSIM_EA(bpf->extack, "program bound to different dev");
-		return -EINVAL;
-	}
 
 	state = bpf->prog->aux->offload->dev_priv;
 	if (WARN_ON(strcmp(state->state, "xlated"))) {

include/linux/bpf.h

···
 	enum bpf_prog_type saved_dst_prog_type;
 	enum bpf_attach_type saved_dst_attach_type;
 	bool verifier_zext; /* Zero extensions has been inserted by verifier. */
-	bool offload_requested;
+	bool dev_bound; /* Program is bound to the netdev. */
+	bool offload_requested; /* Program is bound and offloaded to the netdev. */
 	bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */
 	bool func_proto_unreliable;
 	bool sleepable;
···
 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
 
 int bpf_prog_offload_compile(struct bpf_prog *prog);
-void bpf_prog_offload_destroy(struct bpf_prog *prog);
+void bpf_prog_dev_bound_destroy(struct bpf_prog *prog);
 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
 			       struct bpf_prog *prog);
 
···
 void unpriv_ebpf_notify(int new_state);
 
 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
-int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
+int bpf_prog_dev_bound_init(struct bpf_prog *prog, union bpf_attr *attr);
+void bpf_dev_bound_netdev_unregister(struct net_device *dev);
+
+static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
+{
+	return aux->dev_bound;
+}
 
 static inline bool bpf_prog_is_offloaded(const struct bpf_prog_aux *aux)
 {
···
 void sock_map_destroy(struct sock *sk);
 void sock_map_close(struct sock *sk, long timeout);
 #else
-static inline int bpf_prog_offload_init(struct bpf_prog *prog,
+static inline int bpf_prog_dev_bound_init(struct bpf_prog *prog,
 					union bpf_attr *attr)
 {
 	return -EOPNOTSUPP;
+}
+
+static inline void bpf_dev_bound_netdev_unregister(struct net_device *dev)
+{
+}
+
+static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
+{
+	return false;
 }
 
 static inline bool bpf_prog_is_offloaded(struct bpf_prog_aux *aux)

include/uapi/linux/bpf.h

···
  */
 #define BPF_F_XDP_HAS_FRAGS	(1U << 5)
 
+/* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded
+ * program becomes device-bound but can access XDP metadata.
+ */
+#define BPF_F_XDP_DEV_BOUND_ONLY	(1U << 6)
+
 /* link_create.kprobe_multi.flags used in LINK_CREATE command for
  * BPF_TRACE_KPROBE_MULTI attach type to create return probe.
  */

kernel/bpf/core.c

···
 #endif
 	bpf_free_used_maps(aux);
 	bpf_free_used_btfs(aux);
-	if (bpf_prog_is_offloaded(aux))
-		bpf_prog_offload_destroy(aux->prog);
+	if (bpf_prog_is_dev_bound(aux))
+		bpf_prog_dev_bound_destroy(aux->prog);
 #ifdef CONFIG_PERF_EVENTS
 	if (aux->prog->has_callchain_buf)
 		put_callchain_buffers();

kernel/bpf/offload.c

···
 struct bpf_offload_netdev {
 	struct rhash_head l;
 	struct net_device *netdev;
-	struct bpf_offload_dev *offdev;
+	struct bpf_offload_dev *offdev; /* NULL when bound-only */
 	struct list_head progs;
 	struct list_head maps;
 	struct list_head offdev_netdevs;
···
 	INIT_LIST_HEAD(&ondev->progs);
 	INIT_LIST_HEAD(&ondev->maps);
 
-	down_write(&bpf_devs_lock);
 	err = rhashtable_insert_fast(&offdevs, &ondev->l, offdevs_params);
 	if (err) {
 		netdev_warn(netdev, "failed to register for BPF offload\n");
-		goto err_unlock_free;
+		goto err_free;
 	}
 
-	list_add(&ondev->offdev_netdevs, &offdev->netdevs);
-	up_write(&bpf_devs_lock);
+	if (offdev)
+		list_add(&ondev->offdev_netdevs, &offdev->netdevs);
 	return 0;
 
-err_unlock_free:
-	up_write(&bpf_devs_lock);
+err_free:
 	kfree(ondev);
 	return err;
 }
···
 static void __bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev,
 						struct net_device *netdev)
 {
-	struct bpf_offload_netdev *ondev, *altdev;
+	struct bpf_offload_netdev *ondev, *altdev = NULL;
 	struct bpf_offloaded_map *offmap, *mtmp;
 	struct bpf_prog_offload *offload, *ptmp;
 
 	ASSERT_RTNL();
 
-	down_write(&bpf_devs_lock);
 	ondev = rhashtable_lookup_fast(&offdevs, &netdev, offdevs_params);
 	if (WARN_ON(!ondev))
-		goto unlock;
+		return;
 
 	WARN_ON(rhashtable_remove_fast(&offdevs, &ondev->l, offdevs_params));
-	list_del(&ondev->offdev_netdevs);
 
 	/* Try to move the objects to another netdev of the device */
-	altdev = list_first_entry_or_null(&offdev->netdevs,
-					  struct bpf_offload_netdev,
-					  offdev_netdevs);
+	if (offdev) {
+		list_del(&ondev->offdev_netdevs);
+		altdev = list_first_entry_or_null(&offdev->netdevs,
+						  struct bpf_offload_netdev,
+						  offdev_netdevs);
+	}
+
 	if (altdev) {
 		list_for_each_entry(offload, &ondev->progs, offloads)
 			offload->netdev = altdev->netdev;
···
 	WARN_ON(!list_empty(&ondev->progs));
 	WARN_ON(!list_empty(&ondev->maps));
 	kfree(ondev);
-unlock:
-	up_write(&bpf_devs_lock);
 }
 
-int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr)
+int bpf_prog_dev_bound_init(struct bpf_prog *prog, union bpf_attr *attr)
 {
 	struct bpf_offload_netdev *ondev;
 	struct bpf_prog_offload *offload;
···
 	    attr->prog_type != BPF_PROG_TYPE_XDP)
 		return -EINVAL;
 
-	if (attr->prog_flags)
+	if (attr->prog_flags & ~BPF_F_XDP_DEV_BOUND_ONLY)
+		return -EINVAL;
+
+	if (attr->prog_type == BPF_PROG_TYPE_SCHED_CLS &&
+	    attr->prog_flags & BPF_F_XDP_DEV_BOUND_ONLY)
 		return -EINVAL;
 
 	offload = kzalloc(sizeof(*offload), GFP_USER);
···
 	if (err)
 		goto err_maybe_put;
 
+	prog->aux->offload_requested = !(attr->prog_flags & BPF_F_XDP_DEV_BOUND_ONLY);
+
 	down_write(&bpf_devs_lock);
 	ondev = bpf_offload_find_netdev(offload->netdev);
 	if (!ondev) {
-		err = -EINVAL;
-		goto err_unlock;
+		if (bpf_prog_is_offloaded(prog->aux)) {
+			err = -EINVAL;
+			goto err_unlock;
+		}
+
+		/* When only binding to the device, explicitly
+		 * create an entry in the hashtable.
+		 */
+		err = __bpf_offload_dev_netdev_register(NULL, offload->netdev);
+		if (err)
+			goto err_unlock;
+		ondev = bpf_offload_find_netdev(offload->netdev);
 	}
 	offload->offdev = ondev->offdev;
 	prog->aux->offload = offload;
···
 	up_read(&bpf_devs_lock);
 }
 
-void bpf_prog_offload_destroy(struct bpf_prog *prog)
+void bpf_prog_dev_bound_destroy(struct bpf_prog *prog)
 {
+	struct bpf_offload_netdev *ondev;
+	struct net_device *netdev;
+
+	rtnl_lock();
 	down_write(&bpf_devs_lock);
-	if (prog->aux->offload)
+	if (prog->aux->offload) {
+		list_del_init(&prog->aux->offload->offloads);
+
+		netdev = prog->aux->offload->netdev;
 		__bpf_prog_offload_destroy(prog);
+
+		ondev = bpf_offload_find_netdev(netdev);
+		if (!ondev->offdev && list_empty(&ondev->progs))
+			__bpf_offload_dev_netdev_unregister(NULL, netdev);
+	}
 	up_write(&bpf_devs_lock);
+	rtnl_unlock();
 }
 
 static int bpf_prog_offload_translate(struct bpf_prog *prog)
···
 	struct bpf_offload_netdev *ondev1, *ondev2;
 	struct bpf_prog_offload *offload;
 
-	if (!bpf_prog_is_offloaded(prog->aux))
+	if (!bpf_prog_is_dev_bound(prog->aux))
 		return false;
 
 	offload = prog->aux->offload;
···
 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev,
 				    struct net_device *netdev)
 {
-	return __bpf_offload_dev_netdev_register(offdev, netdev);
+	int err;
+
+	down_write(&bpf_devs_lock);
+	err = __bpf_offload_dev_netdev_register(offdev, netdev);
+	up_write(&bpf_devs_lock);
+	return err;
 }
 EXPORT_SYMBOL_GPL(bpf_offload_dev_netdev_register);
 
 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev,
 				       struct net_device *netdev)
 {
+	down_write(&bpf_devs_lock);
 	__bpf_offload_dev_netdev_unregister(offdev, netdev);
+	up_write(&bpf_devs_lock);
 }
 EXPORT_SYMBOL_GPL(bpf_offload_dev_netdev_unregister);
 
···
 	return offdev->priv;
 }
 EXPORT_SYMBOL_GPL(bpf_offload_dev_priv);
+
+void bpf_dev_bound_netdev_unregister(struct net_device *dev)
+{
+	struct bpf_offload_netdev *ondev;
+
+	ASSERT_RTNL();
+
+	down_write(&bpf_devs_lock);
+	ondev = bpf_offload_find_netdev(dev);
+	if (ondev && !ondev->offdev)
+		__bpf_offload_dev_netdev_unregister(NULL, ondev->netdev);
+	up_write(&bpf_devs_lock);
+}
 
 static int __init bpf_offload_init(void)
 {

kernel/bpf/syscall.c

···
 				 BPF_F_TEST_STATE_FREQ |
 				 BPF_F_SLEEPABLE |
 				 BPF_F_TEST_RND_HI32 |
-				 BPF_F_XDP_HAS_FRAGS))
+				 BPF_F_XDP_HAS_FRAGS |
+				 BPF_F_XDP_DEV_BOUND_ONLY))
 		return -EINVAL;
 
 	if (!IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) &&
···
 	prog->aux->attach_btf = attach_btf;
 	prog->aux->attach_btf_id = attr->attach_btf_id;
 	prog->aux->dst_prog = dst_prog;
-	prog->aux->offload_requested = !!attr->prog_ifindex;
+	prog->aux->dev_bound = !!attr->prog_ifindex;
 	prog->aux->sleepable = attr->prog_flags & BPF_F_SLEEPABLE;
 	prog->aux->xdp_has_frags = attr->prog_flags & BPF_F_XDP_HAS_FRAGS;
 
···
 	atomic64_set(&prog->aux->refcnt, 1);
 	prog->gpl_compatible = is_gpl ? 1 : 0;
 
-	if (bpf_prog_is_offloaded(prog->aux)) {
-		err = bpf_prog_offload_init(prog, attr);
+	if (bpf_prog_is_dev_bound(prog->aux)) {
+		err = bpf_prog_dev_bound_init(prog, attr);
 		if (err)
 			goto free_prog_sec;
 	}

net/core/dev.c

···
 			NL_SET_ERR_MSG(extack, "Using offloaded program without HW_MODE flag is not supported");
 			return -EINVAL;
 		}
+		if (bpf_prog_is_dev_bound(new_prog->aux) && !bpf_offload_dev_match(new_prog, dev)) {
+			NL_SET_ERR_MSG(extack, "Program bound to different device");
+			return -EINVAL;
+		}
 		if (new_prog->expected_attach_type == BPF_XDP_DEVMAP) {
 			NL_SET_ERR_MSG(extack, "BPF_XDP_DEVMAP programs can not be attached to a device");
 			return -EINVAL;
···
 		dev_shutdown(dev);
 
 		dev_xdp_uninstall(dev);
+		bpf_dev_bound_netdev_unregister(dev);
 
 		netdev_offload_xstats_disable_all(dev);
 

tools/include/uapi/linux/bpf.h

···
  */
 #define BPF_F_XDP_HAS_FRAGS	(1U << 5)
 
+/* If BPF_F_XDP_DEV_BOUND_ONLY is used in BPF_PROG_LOAD command, the loaded
+ * program becomes device-bound but can access XDP metadata.
+ */
+#define BPF_F_XDP_DEV_BOUND_ONLY	(1U << 6)
+
 /* link_create.kprobe_multi.flags used in LINK_CREATE command for
  * BPF_TRACE_KPROBE_MULTI attach type to create return probe.
  */