Merge branch 'bpf-type-format'

Martin KaFai Lau says:

====================
This patch introduces BPF Type Format (BTF).

BTF (BPF Type Format) is the meta data format which describes
the data types of BPF program/map.  Hence, it basically focus
on the C programming language which the modern BPF is primary
using.  The first use case is to provide a generic pretty print
capability for a BPF map.

A modified pahole that can convert dwarf to BTF is here:

  https://github.com/iamkafai/pahole/tree/btf

Please see individual patch for details.

v5:
- Remove BTF_KIND_FLOAT and BTF_KIND_FUNC which are not
  currently used.  They can be added in the future.
  Some bpf_df_xxx() are removed together.
- Add comment in patch 7 to clarify that the new bpffs_map_fops
  should not be extended further.

v4:
- Fix warning (remove unneeded semicolon)
- Remove a redundant variable (nr_bytes) from btf_int_check_meta() in
  patch 1.  Caught by W=1.

v3:
- Rebase to bpf-next
- Fix sparse warning (by adding static)
- Add BTF header logging: btf_verifier_log_hdr()
- Fix the alignment test on btf->type_off
- Add tests for the BTF header
- Lower the max BTF size to 16MB.  It should be enough
  for some time.  We could raise it later if it would
  be needed.

v2:
- Use kvfree where needed in patch 1 and 2
- Also consider BTF_INT_OFFSET() in the btf_int_check_meta()
  in patch 1
- Fix an incorrect goto target in map_create() during
  the btf-error-path in patch 7
- re-org some local vars to keep the rev xmas tree in btf.c
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

include/linux/bpf.h

@@ -22,6 +22,8 @@ struct perf_event;
 struct bpf_prog;
 struct bpf_map;
 struct sock;
+struct seq_file;
+struct btf;
 
 /* map is generic key/value storage optionally accesible by eBPF programs */
 struct bpf_map_ops {
@@ -43,10 +45,14 @@ struct bpf_map_ops {
 	void (*map_fd_put_ptr)(void *ptr);
 	u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
 	u32 (*map_fd_sys_lookup_elem)(void *ptr);
+	void (*map_seq_show_elem)(struct bpf_map *map, void *key,
+				  struct seq_file *m);
+	int (*map_check_btf)(const struct bpf_map *map, const struct btf *btf,
+			     u32 key_type_id, u32 value_type_id);
 };
 
 struct bpf_map {
-	/* 1st cacheline with read-mostly members of which some
+	/* The first two cachelines with read-mostly members of which some
 	 * are also accessed in fast-path (e.g. ops, max_entries).
 	 */
 	const struct bpf_map_ops *ops ____cacheline_aligned;
@@ -62,10 +68,13 @@ struct bpf_map {
 	u32 pages;
 	u32 id;
 	int numa_node;
+	u32 btf_key_id;
+	u32 btf_value_id;
+	struct btf *btf;
 	bool unpriv_array;
-	/* 7 bytes hole */
+	/* 55 bytes hole */
 
-	/* 2nd cacheline with misc members to avoid false sharing
+	/* The 3rd and 4th cacheline with misc members to avoid false sharing
 	 * particularly with refcounting.
 	 */
 	struct user_struct *user ____cacheline_aligned;
@@ -100,6 +109,11 @@ static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
 	return container_of(map, struct bpf_offloaded_map, map);
 }
 
+static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
+{
+	return map->ops->map_seq_show_elem && map->ops->map_check_btf;
+}
+
 extern const struct bpf_map_ops bpf_map_offload_ops;
 
 /* function argument constraints */

include/linux/btf.h

+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2018 Facebook */
+
+#ifndef _LINUX_BTF_H
+#define _LINUX_BTF_H 1
+
+#include <linux/types.h>
+
+struct btf;
+struct btf_type;
+union bpf_attr;
+
+extern const struct file_operations btf_fops;
+
+void btf_put(struct btf *btf);
+int btf_new_fd(const union bpf_attr *attr);
+struct btf *btf_get_by_fd(int fd);
+int btf_get_info_by_fd(const struct btf *btf,
+		       const union bpf_attr *attr,
+		       union bpf_attr __user *uattr);
+/* Figure out the size of a type_id.  If type_id is a modifier
+ * (e.g. const), it will be resolved to find out the type with size.
+ *
+ * For example:
+ * In describing "const void *",  type_id is "const" and "const"
+ * refers to "void *".  The return type will be "void *".
+ *
+ * If type_id is a simple "int", then return type will be "int".
+ *
+ * @btf: struct btf object
+ * @type_id: Find out the size of type_id. The type_id of the return
+ *           type is set to *type_id.
+ * @ret_size: It can be NULL.  If not NULL, the size of the return
+ *            type is set to *ret_size.
+ * Return: The btf_type (resolved to another type with size info if needed).
+ *         NULL is returned if type_id itself does not have size info
+ *         (e.g. void) or it cannot be resolved to another type that
+ *         has size info.
+ *         *type_id and *ret_size will not be changed in the
+ *         NULL return case.
+ */
+const struct btf_type *btf_type_id_size(const struct btf *btf,
+					u32 *type_id,
+					u32 *ret_size);
+void btf_type_seq_show(const struct btf *btf, u32 type_id, void *obj,
+		       struct seq_file *m);
+
+#endif

include/uapi/linux/bpf.h

@@ -95,6 +95,7 @@ enum bpf_cmd {
 	BPF_OBJ_GET_INFO_BY_FD,
 	BPF_PROG_QUERY,
 	BPF_RAW_TRACEPOINT_OPEN,
+	BPF_BTF_LOAD,
 };
 
 enum bpf_map_type {
@@ -279,6 +280,9 @@ union bpf_attr {
 					 */
 		char	map_name[BPF_OBJ_NAME_LEN];
 		__u32	map_ifindex;	/* ifindex of netdev to create on */
+		__u32	btf_fd;		/* fd pointing to a BTF type data */
+		__u32	btf_key_id;	/* BTF type_id of the key */
+		__u32	btf_value_id;	/* BTF type_id of the value */
 	};
 
 	struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
@@ -363,6 +367,14 @@ union bpf_attr {
 		__u64 name;
 		__u32 prog_fd;
 	} raw_tracepoint;
+
+	struct { /* anonymous struct for BPF_BTF_LOAD */
+		__aligned_u64	btf;
+		__aligned_u64	btf_log_buf;
+		__u32		btf_size;
+		__u32		btf_log_size;
+		__u32		btf_log_level;
+	};
 } __attribute__((aligned(8)));
 
 /* BPF helper function descriptions:

include/uapi/linux/btf.h

+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+/* Copyright (c) 2018 Facebook */
+#ifndef _UAPI__LINUX_BTF_H__
+#define _UAPI__LINUX_BTF_H__
+
+#include <linux/types.h>
+
+#define BTF_MAGIC	0xeB9F
+#define BTF_MAGIC_SWAP	0x9FeB
+#define BTF_VERSION	1
+#define BTF_FLAGS_COMPR	0x01
+
+struct btf_header {
+	__u16	magic;
+	__u8	version;
+	__u8	flags;
+
+	__u32	parent_label;
+	__u32	parent_name;
+
+	/* All offsets are in bytes relative to the end of this header */
+	__u32	label_off;	/* offset of label section	*/
+	__u32	object_off;	/* offset of data object section*/
+	__u32	func_off;	/* offset of function section	*/
+	__u32	type_off;	/* offset of type section	*/
+	__u32	str_off;	/* offset of string section	*/
+	__u32	str_len;	/* length of string section	*/
+};
+
+/* Max # of type identifier */
+#define BTF_MAX_TYPE	0x7fffffff
+/* Max offset into the string section */
+#define BTF_MAX_NAME_OFFSET	0x7fffffff
+/* Max # of struct/union/enum members or func args */
+#define BTF_MAX_VLEN	0xffff
+
+/* The type id is referring to a parent BTF */
+#define BTF_TYPE_PARENT(id)	(((id) >> 31) & 0x1)
+#define BTF_TYPE_ID(id)		((id) & BTF_MAX_TYPE)
+
+/* String is in the ELF string section */
+#define BTF_STR_TBL_ELF_ID(ref)	(((ref) >> 31) & 0x1)
+#define BTF_STR_OFFSET(ref)	((ref) & BTF_MAX_NAME_OFFSET)
+
+struct btf_type {
+	__u32 name;
+	/* "info" bits arrangement
+	 * bits  0-15: vlen (e.g. # of struct's members)
+	 * bits 16-23: unused
+	 * bits 24-28: kind (e.g. int, ptr, array...etc)
+	 * bits 29-30: unused
+	 * bits    31: root
+	 */
+	__u32 info;
+	/* "size" is used by INT, ENUM, STRUCT and UNION.
+	 * "size" tells the size of the type it is describing.
+	 *
+	 * "type" is used by PTR, TYPEDEF, VOLATILE, CONST and RESTRICT.
+	 * "type" is a type_id referring to another type.
+	 */
+	union {
+		__u32 size;
+		__u32 type;
+	};
+};
+
+#define BTF_INFO_KIND(info)	(((info) >> 24) & 0x1f)
+#define BTF_INFO_ISROOT(info)	(!!(((info) >> 24) & 0x80))
+#define BTF_INFO_VLEN(info)	((info) & 0xffff)
+
+#define BTF_KIND_UNKN		0	/* Unknown	*/
+#define BTF_KIND_INT		1	/* Integer	*/
+#define BTF_KIND_PTR		2	/* Pointer	*/
+#define BTF_KIND_ARRAY		3	/* Array	*/
+#define BTF_KIND_STRUCT		4	/* Struct	*/
+#define BTF_KIND_UNION		5	/* Union	*/
+#define BTF_KIND_ENUM		6	/* Enumeration	*/
+#define BTF_KIND_FWD		7	/* Forward	*/
+#define BTF_KIND_TYPEDEF	8	/* Typedef	*/
+#define BTF_KIND_VOLATILE	9	/* Volatile	*/
+#define BTF_KIND_CONST		10	/* Const	*/
+#define BTF_KIND_RESTRICT	11	/* Restrict	*/
+#define BTF_KIND_MAX		11
+#define NR_BTF_KINDS		12
+
+/* For some specific BTF_KIND, "struct btf_type" is immediately
+ * followed by extra data.
+ */
+
+/* BTF_KIND_INT is followed by a u32 and the following
+ * is the 32 bits arrangement:
+ */
+#define BTF_INT_ENCODING(VAL)	(((VAL) & 0xff000000) >> 24)
+#define BTF_INT_OFFSET(VAL)	(((VAL  & 0x00ff0000)) >> 16)
+#define BTF_INT_BITS(VAL)	((VAL)  & 0x0000ffff)
+
+/* Attributes stored in the BTF_INT_ENCODING */
+#define BTF_INT_SIGNED	0x1
+#define BTF_INT_CHAR	0x2
+#define BTF_INT_BOOL	0x4
+#define BTF_INT_VARARGS	0x8
+
+/* BTF_KIND_ENUM is followed by multiple "struct btf_enum".
+ * The exact number of btf_enum is stored in the vlen (of the
+ * info in "struct btf_type").
+ */
+struct btf_enum {
+	__u32	name;
+	__s32	val;
+};
+
+/* BTF_KIND_ARRAY is followed by one "struct btf_array" */
+struct btf_array {
+	__u32	type;
+	__u32	index_type;
+	__u32	nelems;
+};
+
+/* BTF_KIND_STRUCT and BTF_KIND_UNION are followed
+ * by multiple "struct btf_member".  The exact number
+ * of btf_member is stored in the vlen (of the info in
+ * "struct btf_type").
+ */
+struct btf_member {
+	__u32	name;
+	__u32	type;
+	__u32	offset;	/* offset in bits */
+};
+
+#endif /* _UAPI__LINUX_BTF_H__ */

kernel/bpf/Makefile

@@ -4,6 +4,7 @@ obj-y := core.o
 obj-$(CONFIG_BPF_SYSCALL) += syscall.o verifier.o inode.o helpers.o tnum.o
 obj-$(CONFIG_BPF_SYSCALL) += hashtab.o arraymap.o percpu_freelist.o bpf_lru_list.o lpm_trie.o map_in_map.o
 obj-$(CONFIG_BPF_SYSCALL) += disasm.o
+obj-$(CONFIG_BPF_SYSCALL) += btf.o
 ifeq ($(CONFIG_NET),y)
 obj-$(CONFIG_BPF_SYSCALL) += devmap.o
 obj-$(CONFIG_BPF_SYSCALL) += cpumap.o

kernel/bpf/arraymap.c

@@ -11,11 +11,13 @@
  * General Public License for more details.
  */
 #include <linux/bpf.h>
+#include <linux/btf.h>
 #include <linux/err.h>
 #include <linux/slab.h>
 #include <linux/mm.h>
 #include <linux/filter.h>
 #include <linux/perf_event.h>
+#include <uapi/linux/btf.h>
 
 #include "map_in_map.h"
 
@@ -336,6 +338,52 @@ static void array_map_free(struct bpf_map *map)
 	bpf_map_area_free(array);
 }
 
+static void array_map_seq_show_elem(struct bpf_map *map, void *key,
+				    struct seq_file *m)
+{
+	void *value;
+
+	rcu_read_lock();
+
+	value = array_map_lookup_elem(map, key);
+	if (!value) {
+		rcu_read_unlock();
+		return;
+	}
+
+	seq_printf(m, "%u: ", *(u32 *)key);
+	btf_type_seq_show(map->btf, map->btf_value_id, value, m);
+	seq_puts(m, "\n");
+
+	rcu_read_unlock();
+}
+
+static int array_map_check_btf(const struct bpf_map *map, const struct btf *btf,
+			       u32 btf_key_id, u32 btf_value_id)
+{
+	const struct btf_type *key_type, *value_type;
+	u32 key_size, value_size;
+	u32 int_data;
+
+	key_type = btf_type_id_size(btf, &btf_key_id, &key_size);
+	if (!key_type || BTF_INFO_KIND(key_type->info) != BTF_KIND_INT)
+		return -EINVAL;
+
+	int_data = *(u32 *)(key_type + 1);
+	/* bpf array can only take a u32 key.  This check makes
+	 * sure that the btf matches the attr used during map_create.
+	 */
+	if (BTF_INT_BITS(int_data) != 32 || key_size != 4 ||
+	    BTF_INT_OFFSET(int_data))
+		return -EINVAL;
+
+	value_type = btf_type_id_size(btf, &btf_value_id, &value_size);
+	if (!value_type || value_size > map->value_size)
+		return -EINVAL;
+
+	return 0;
+}
+
 const struct bpf_map_ops array_map_ops = {
 	.map_alloc_check = array_map_alloc_check,
 	.map_alloc = array_map_alloc,
@@ -345,6 +393,8 @@ const struct bpf_map_ops array_map_ops = {
 	.map_update_elem = array_map_update_elem,
 	.map_delete_elem = array_map_delete_elem,
 	.map_gen_lookup = array_map_gen_lookup,
+	.map_seq_show_elem = array_map_seq_show_elem,
+	.map_check_btf = array_map_check_btf,
 };
 
 const struct bpf_map_ops percpu_array_map_ops = {

kernel/bpf/inode.c

@@ -150,8 +150,154 @@ static int bpf_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
 	return 0;
 }
 
+struct map_iter {
+	void *key;
+	bool done;
+};
+
+static struct map_iter *map_iter(struct seq_file *m)
+{
+	return m->private;
+}
+
+static struct bpf_map *seq_file_to_map(struct seq_file *m)
+{
+	return file_inode(m->file)->i_private;
+}
+
+static void map_iter_free(struct map_iter *iter)
+{
+	if (iter) {
+		kfree(iter->key);
+		kfree(iter);
+	}
+}
+
+static struct map_iter *map_iter_alloc(struct bpf_map *map)
+{
+	struct map_iter *iter;
+
+	iter = kzalloc(sizeof(*iter), GFP_KERNEL | __GFP_NOWARN);
+	if (!iter)
+		goto error;
+
+	iter->key = kzalloc(map->key_size, GFP_KERNEL | __GFP_NOWARN);
+	if (!iter->key)
+		goto error;
+
+	return iter;
+
+error:
+	map_iter_free(iter);
+	return NULL;
+}
+
+static void *map_seq_next(struct seq_file *m, void *v, loff_t *pos)
+{
+	struct bpf_map *map = seq_file_to_map(m);
+	void *key = map_iter(m)->key;
+
+	if (map_iter(m)->done)
+		return NULL;
+
+	if (unlikely(v == SEQ_START_TOKEN))
+		goto done;
+
+	if (map->ops->map_get_next_key(map, key, key)) {
+		map_iter(m)->done = true;
+		return NULL;
+	}
+
+done:
+	++(*pos);
+	return key;
+}
+
+static void *map_seq_start(struct seq_file *m, loff_t *pos)
+{
+	if (map_iter(m)->done)
+		return NULL;
+
+	return *pos ? map_iter(m)->key : SEQ_START_TOKEN;
+}
+
+static void map_seq_stop(struct seq_file *m, void *v)
+{
+}
+
+static int map_seq_show(struct seq_file *m, void *v)
+{
+	struct bpf_map *map = seq_file_to_map(m);
+	void *key = map_iter(m)->key;
+
+	if (unlikely(v == SEQ_START_TOKEN)) {
+		seq_puts(m, "# WARNING!! The output is for debug purpose only\n");
+		seq_puts(m, "# WARNING!! The output format will change\n");
+	} else {
+		map->ops->map_seq_show_elem(map, key, m);
+	}
+
+	return 0;
+}
+
+static const struct seq_operations bpffs_map_seq_ops = {
+	.start	= map_seq_start,
+	.next	= map_seq_next,
+	.show	= map_seq_show,
+	.stop	= map_seq_stop,
+};
+
+static int bpffs_map_open(struct inode *inode, struct file *file)
+{
+	struct bpf_map *map = inode->i_private;
+	struct map_iter *iter;
+	struct seq_file *m;
+	int err;
+
+	iter = map_iter_alloc(map);
+	if (!iter)
+		return -ENOMEM;
+
+	err = seq_open(file, &bpffs_map_seq_ops);
+	if (err) {
+		map_iter_free(iter);
+		return err;
+	}
+
+	m = file->private_data;
+	m->private = iter;
+
+	return 0;
+}
+
+static int bpffs_map_release(struct inode *inode, struct file *file)
+{
+	struct seq_file *m = file->private_data;
+
+	map_iter_free(map_iter(m));
+
+	return seq_release(inode, file);
+}
+
+/* bpffs_map_fops should only implement the basic
+ * read operation for a BPF map.  The purpose is to
+ * provide a simple user intuitive way to do
+ * "cat bpffs/pathto/a-pinned-map".
+ *
+ * Other operations (e.g. write, lookup...) should be realized by
+ * the userspace tools (e.g. bpftool) through the
+ * BPF_OBJ_GET_INFO_BY_FD and the map's lookup/update
+ * interface.
+ */
+static const struct file_operations bpffs_map_fops = {
+	.open		= bpffs_map_open,
+	.read		= seq_read,
+	.release	= bpffs_map_release,
+};
+
 static int bpf_mkobj_ops(struct dentry *dentry, umode_t mode, void *raw,
-			 const struct inode_operations *iops)
+			 const struct inode_operations *iops,
+			 const struct file_operations *fops)
 {
 	struct inode *dir = dentry->d_parent->d_inode;
 	struct inode *inode = bpf_get_inode(dir->i_sb, dir, mode);
@@ -159,6 +305,7 @@ static int bpf_mkobj_ops(struct dentry *dentry, umode_t mode, void *raw,
 		return PTR_ERR(inode);
 
 	inode->i_op = iops;
+	inode->i_fop = fops;
 	inode->i_private = raw;
 
 	bpf_dentry_finalize(dentry, inode, dir);
@@ -167,12 +314,15 @@ static int bpf_mkobj_ops(struct dentry *dentry, umode_t mode, void *raw,
 
 static int bpf_mkprog(struct dentry *dentry, umode_t mode, void *arg)
 {
-	return bpf_mkobj_ops(dentry, mode, arg, &bpf_prog_iops);
+	return bpf_mkobj_ops(dentry, mode, arg, &bpf_prog_iops, NULL);
 }
 
 static int bpf_mkmap(struct dentry *dentry, umode_t mode, void *arg)
 {
-	return bpf_mkobj_ops(dentry, mode, arg, &bpf_map_iops);
+	struct bpf_map *map = arg;
+
+	return bpf_mkobj_ops(dentry, mode, arg, &bpf_map_iops,
+			     map->btf ? &bpffs_map_fops : NULL);
 }
 
 static struct dentry *

kernel/bpf/syscall.c

@@ -11,6 +11,7 @@
  */
 #include <linux/bpf.h>
 #include <linux/bpf_trace.h>
+#include <linux/btf.h>
 #include <linux/syscalls.h>
 #include <linux/slab.h>
 #include <linux/sched/signal.h>
@@ -26,6 +27,7 @@
 #include <linux/cred.h>
 #include <linux/timekeeping.h>
 #include <linux/ctype.h>
+#include <linux/btf.h>
 
 #define IS_FD_ARRAY(map) ((map)->map_type == BPF_MAP_TYPE_PROG_ARRAY || \
 			   (map)->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY || \
@@ -250,6 +252,7 @@ static void bpf_map_free_deferred(struct work_struct *work)
 
 	bpf_map_uncharge_memlock(map);
 	security_bpf_map_free(map);
+	btf_put(map->btf);
 	/* implementation dependent freeing */
 	map->ops->map_free(map);
 }
@@ -415,7 +418,7 @@ static int bpf_obj_name_cpy(char *dst, const char *src)
 	return 0;
 }
 
-#define BPF_MAP_CREATE_LAST_FIELD map_ifindex
+#define BPF_MAP_CREATE_LAST_FIELD btf_value_id
 /* called via syscall */
 static int map_create(union bpf_attr *attr)
 {
@@ -449,6 +452,33 @@ static int map_create(union bpf_attr *attr)
 	atomic_set(&map->refcnt, 1);
 	atomic_set(&map->usercnt, 1);
 
+	if (bpf_map_support_seq_show(map) &&
+	    (attr->btf_key_id || attr->btf_value_id)) {
+		struct btf *btf;
+
+		if (!attr->btf_key_id || !attr->btf_value_id) {
+			err = -EINVAL;
+			goto free_map_nouncharge;
+		}
+
+		btf = btf_get_by_fd(attr->btf_fd);
+		if (IS_ERR(btf)) {
+			err = PTR_ERR(btf);
+			goto free_map_nouncharge;
+		}
+
+		err = map->ops->map_check_btf(map, btf, attr->btf_key_id,
+					      attr->btf_value_id);
+		if (err) {
+			btf_put(btf);
+			goto free_map_nouncharge;
+		}
+
+		map->btf = btf;
+		map->btf_key_id = attr->btf_key_id;
+		map->btf_value_id = attr->btf_value_id;
+	}
+
 	err = security_bpf_map_alloc(map);
 	if (err)
 		goto free_map_nouncharge;
@@ -481,6 +511,7 @@ static int map_create(union bpf_attr *attr)
 free_map_sec:
 	security_bpf_map_free(map);
 free_map_nouncharge:
+	btf_put(map->btf);
 	map->ops->map_free(map);
 	return err;
 }
@@ -2016,6 +2047,8 @@ static int bpf_obj_get_info_by_fd(const union bpf_attr *attr,
 	else if (f.file->f_op == &bpf_map_fops)
 		err = bpf_map_get_info_by_fd(f.file->private_data, attr,
 					     uattr);
+	else if (f.file->f_op == &btf_fops)
+		err = btf_get_info_by_fd(f.file->private_data, attr, uattr);
 	else
 		err = -EINVAL;
 
@@ -2023,6 +2056,19 @@ static int bpf_obj_get_info_by_fd(const union bpf_attr *attr,
 	return err;
 }
 
+#define BPF_BTF_LOAD_LAST_FIELD btf_log_level
+
+static int bpf_btf_load(const union bpf_attr *attr)
+{
+	if (CHECK_ATTR(BPF_BTF_LOAD))
+		return -EINVAL;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	return btf_new_fd(attr);
+}
+
 SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, size)
 {
 	union bpf_attr attr = {};
@@ -2103,6 +2149,9 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, siz
 	case BPF_RAW_TRACEPOINT_OPEN:
 		err = bpf_raw_tracepoint_open(&attr);
 		break;
+	case BPF_BTF_LOAD:
+		err = bpf_btf_load(&attr);
+		break;
 	default:
 		err = -EINVAL;
 		break;

tools/include/uapi/linux/bpf.h

@@ -95,6 +95,7 @@ enum bpf_cmd {
 	BPF_OBJ_GET_INFO_BY_FD,
 	BPF_PROG_QUERY,
 	BPF_RAW_TRACEPOINT_OPEN,
+	BPF_BTF_LOAD,
 };
 
 enum bpf_map_type {
@@ -279,6 +280,9 @@ union bpf_attr {
 					 */
 		char	map_name[BPF_OBJ_NAME_LEN];
 		__u32	map_ifindex;	/* ifindex of netdev to create on */
+		__u32	btf_fd;		/* fd pointing to a BTF type data */
+		__u32	btf_key_id;	/* BTF type_id of the key */
+		__u32	btf_value_id;	/* BTF type_id of the value */
 	};
 
 	struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
@@ -363,6 +367,14 @@ union bpf_attr {
 		__u64 name;
 		__u32 prog_fd;
 	} raw_tracepoint;
+
+	struct { /* anonymous struct for BPF_BTF_LOAD */
+		__aligned_u64	btf;
+		__aligned_u64	btf_log_buf;
+		__u32		btf_size;
+		__u32		btf_log_size;
+		__u32		btf_log_level;
+	};
 } __attribute__((aligned(8)));
 
 /* BPF helper function descriptions:

tools/include/uapi/linux/btf.h

+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+/* Copyright (c) 2018 Facebook */
+#ifndef _UAPI__LINUX_BTF_H__
+#define _UAPI__LINUX_BTF_H__
+
+#include <linux/types.h>
+
+#define BTF_MAGIC	0xeB9F
+#define BTF_MAGIC_SWAP	0x9FeB
+#define BTF_VERSION	1
+#define BTF_FLAGS_COMPR	0x01
+
+struct btf_header {
+	__u16	magic;
+	__u8	version;
+	__u8	flags;
+
+	__u32	parent_label;
+	__u32	parent_name;
+
+	/* All offsets are in bytes relative to the end of this header */
+	__u32	label_off;	/* offset of label section	*/
+	__u32	object_off;	/* offset of data object section*/
+	__u32	func_off;	/* offset of function section	*/
+	__u32	type_off;	/* offset of type section	*/
+	__u32	str_off;	/* offset of string section	*/
+	__u32	str_len;	/* length of string section	*/
+};
+
+/* Max # of type identifier */
+#define BTF_MAX_TYPE	0x7fffffff
+/* Max offset into the string section */
+#define BTF_MAX_NAME_OFFSET	0x7fffffff
+/* Max # of struct/union/enum members or func args */
+#define BTF_MAX_VLEN	0xffff
+
+/* The type id is referring to a parent BTF */
+#define BTF_TYPE_PARENT(id)	(((id) >> 31) & 0x1)
+#define BTF_TYPE_ID(id)		((id) & BTF_MAX_TYPE)
+
+/* String is in the ELF string section */
+#define BTF_STR_TBL_ELF_ID(ref)	(((ref) >> 31) & 0x1)
+#define BTF_STR_OFFSET(ref)	((ref) & BTF_MAX_NAME_OFFSET)
+
+struct btf_type {
+	__u32 name;
+	/* "info" bits arrangement
+	 * bits  0-15: vlen (e.g. # of struct's members)
+	 * bits 16-23: unused
+	 * bits 24-28: kind (e.g. int, ptr, array...etc)
+	 * bits 29-30: unused
+	 * bits    31: root
+	 */
+	__u32 info;
+	/* "size" is used by INT, ENUM, STRUCT and UNION.
+	 * "size" tells the size of the type it is describing.
+	 *
+	 * "type" is used by PTR, TYPEDEF, VOLATILE, CONST and RESTRICT.
+	 * "type" is a type_id referring to another type.
+	 */
+	union {
+		__u32 size;
+		__u32 type;
+	};
+};
+
+#define BTF_INFO_KIND(info)	(((info) >> 24) & 0x1f)
+#define BTF_INFO_ISROOT(info)	(!!(((info) >> 24) & 0x80))
+#define BTF_INFO_VLEN(info)	((info) & 0xffff)
+
+#define BTF_KIND_UNKN		0	/* Unknown	*/
+#define BTF_KIND_INT		1	/* Integer	*/
+#define BTF_KIND_PTR		2	/* Pointer	*/
+#define BTF_KIND_ARRAY		3	/* Array	*/
+#define BTF_KIND_STRUCT		4	/* Struct	*/
+#define BTF_KIND_UNION		5	/* Union	*/
+#define BTF_KIND_ENUM		6	/* Enumeration	*/
+#define BTF_KIND_FWD		7	/* Forward	*/
+#define BTF_KIND_TYPEDEF	8	/* Typedef	*/
+#define BTF_KIND_VOLATILE	9	/* Volatile	*/
+#define BTF_KIND_CONST		10	/* Const	*/
+#define BTF_KIND_RESTRICT	11	/* Restrict	*/
+#define BTF_KIND_MAX		11
+#define NR_BTF_KINDS		12
+
+/* For some specific BTF_KIND, "struct btf_type" is immediately
+ * followed by extra data.
+ */
+
+/* BTF_KIND_INT is followed by a u32 and the following
+ * is the 32 bits arrangement:
+ */
+#define BTF_INT_ENCODING(VAL)	(((VAL) & 0xff000000) >> 24)
+#define BTF_INT_OFFSET(VAL)	(((VAL  & 0x00ff0000)) >> 16)
+#define BTF_INT_BITS(VAL)	((VAL)  & 0x0000ffff)
+
+/* Attributes stored in the BTF_INT_ENCODING */
+#define BTF_INT_SIGNED	0x1
+#define BTF_INT_CHAR	0x2
+#define BTF_INT_BOOL	0x4
+#define BTF_INT_VARARGS	0x8
+
+/* BTF_KIND_ENUM is followed by multiple "struct btf_enum".
+ * The exact number of btf_enum is stored in the vlen (of the
+ * info in "struct btf_type").
+ */
+struct btf_enum {
+	__u32	name;
+	__s32	val;
+};
+
+/* BTF_KIND_ARRAY is followed by one "struct btf_array" */
+struct btf_array {
+	__u32	type;
+	__u32	index_type;
+	__u32	nelems;
+};
+
+/* BTF_KIND_STRUCT and BTF_KIND_UNION are followed
+ * by multiple "struct btf_member".  The exact number
+ * of btf_member is stored in the vlen (of the info in
+ * "struct btf_type").
+ */
+struct btf_member {
+	__u32	name;
+	__u32	type;
+	__u32	offset;	/* offset in bits */
+};
+
+#endif /* _UAPI__LINUX_BTF_H__ */

tools/lib/bpf/Build

-libbpf-y := libbpf.o bpf.o nlattr.o
+libbpf-y := libbpf.o bpf.o nlattr.o btf.o

tools/lib/bpf/bpf.c

@@ -73,43 +73,76 @@ static inline int sys_bpf(enum bpf_cmd cmd, union bpf_attr *attr,
 	return syscall(__NR_bpf, cmd, attr, size);
 }
 
-int bpf_create_map_node(enum bpf_map_type map_type, const char *name,
-			int key_size, int value_size, int max_entries,
-			__u32 map_flags, int node)
+int bpf_create_map_xattr(const struct bpf_create_map_attr *create_attr)
 {
-	__u32 name_len = name ? strlen(name) : 0;
+	__u32 name_len = create_attr->name ? strlen(create_attr->name) : 0;
 	union bpf_attr attr;
 
 	memset(&attr, '\0', sizeof(attr));
 
-	attr.map_type = map_type;
-	attr.key_size = key_size;
-	attr.value_size = value_size;
-	attr.max_entries = max_entries;
-	attr.map_flags = map_flags;
-	memcpy(attr.map_name, name, min(name_len, BPF_OBJ_NAME_LEN - 1));
+	attr.map_type = create_attr->map_type;
+	attr.key_size = create_attr->key_size;
+	attr.value_size = create_attr->value_size;
+	attr.max_entries = create_attr->max_entries;
+	attr.map_flags = create_attr->map_flags;
+	memcpy(attr.map_name, create_attr->name,
+	       min(name_len, BPF_OBJ_NAME_LEN - 1));
+	attr.numa_node = create_attr->numa_node;
+	attr.btf_fd = create_attr->btf_fd;
+	attr.btf_key_id = create_attr->btf_key_id;
+	attr.btf_value_id = create_attr->btf_value_id;
 
+	return sys_bpf(BPF_MAP_CREATE, &attr, sizeof(attr));
+}
+
+int bpf_create_map_node(enum bpf_map_type map_type, const char *name,
+			int key_size, int value_size, int max_entries,
+			__u32 map_flags, int node)
+{
+	struct bpf_create_map_attr map_attr = {};
+
+	map_attr.name = name;
+	map_attr.map_type = map_type;
+	map_attr.map_flags = map_flags;
+	map_attr.key_size = key_size;
+	map_attr.value_size = value_size;
+	map_attr.max_entries = max_entries;
 	if (node >= 0) {
-		attr.map_flags |= BPF_F_NUMA_NODE;
-		attr.numa_node = node;
+		map_attr.numa_node = node;
+		map_attr.map_flags |= BPF_F_NUMA_NODE;
 	}
 
-	return sys_bpf(BPF_MAP_CREATE, &attr, sizeof(attr));
+	return bpf_create_map_xattr(&map_attr);
 }
 
 int bpf_create_map(enum bpf_map_type map_type, int key_size,
 		   int value_size, int max_entries, __u32 map_flags)
 {
-	return bpf_create_map_node(map_type, NULL, key_size, value_size,
-				   max_entries, map_flags, -1);
+	struct bpf_create_map_attr map_attr = {};
+
+	map_attr.map_type = map_type;
+	map_attr.map_flags = map_flags;
+	map_attr.key_size = key_size;
+	map_attr.value_size = value_size;
+	map_attr.max_entries = max_entries;
+
+	return bpf_create_map_xattr(&map_attr);
 }
 
 int bpf_create_map_name(enum bpf_map_type map_type, const char *name,
 			int key_size, int value_size, int max_entries,
 			__u32 map_flags)
 {
-	return bpf_create_map_node(map_type, name, key_size, value_size,
-				   max_entries, map_flags, -1);
+	struct bpf_create_map_attr map_attr = {};
+
+	map_attr.name = name;
+	map_attr.map_type = map_type;
+	map_attr.map_flags = map_flags;
+	map_attr.key_size = key_size;
+	map_attr.value_size = value_size;
+	map_attr.max_entries = max_entries;
+
+	return bpf_create_map_xattr(&map_attr);
 }
 
 int bpf_create_map_in_map_node(enum bpf_map_type map_type, const char *name,
@@ -573,3 +606,28 @@ int bpf_set_link_xdp_fd(int ifindex, int fd, __u32 flags)
 	close(sock);
 	return ret;
 }
+
+int bpf_load_btf(void *btf, __u32 btf_size, char *log_buf, __u32 log_buf_size,
+		 bool do_log)
+{
+	union bpf_attr attr = {};
+	int fd;
+
+	attr.btf = ptr_to_u64(btf);
+	attr.btf_size = btf_size;
+
+retry:
+	if (do_log && log_buf && log_buf_size) {
+		attr.btf_log_level = 1;
+		attr.btf_log_size = log_buf_size;
+		attr.btf_log_buf = ptr_to_u64(log_buf);
+	}
+
+	fd = sys_bpf(BPF_BTF_LOAD, &attr, sizeof(attr));
+	if (fd == -1 && !do_log && log_buf && log_buf_size) {
+		do_log = true;
+		goto retry;
+	}
+
+	return fd;
+}

tools/lib/bpf/bpf.h

@@ -26,6 +26,20 @@
 #include <linux/bpf.h>
 #include <stddef.h>
 
+struct bpf_create_map_attr {
+	const char *name;
+	enum bpf_map_type map_type;
+	__u32 map_flags;
+	__u32 key_size;
+	__u32 value_size;
+	__u32 max_entries;
+	__u32 numa_node;
+	__u32 btf_fd;
+	__u32 btf_key_id;
+	__u32 btf_value_id;
+};
+
+int bpf_create_map_xattr(const struct bpf_create_map_attr *create_attr);
 int bpf_create_map_node(enum bpf_map_type map_type, const char *name,
 			int key_size, int value_size, int max_entries,
 			__u32 map_flags, int node);
@@ -87,4 +101,6 @@ int bpf_obj_get_info_by_fd(int prog_fd, void *info, __u32 *info_len);
 int bpf_prog_query(int target_fd, enum bpf_attach_type type, __u32 query_flags,
 		   __u32 *attach_flags, __u32 *prog_ids, __u32 *prog_cnt);
 int bpf_raw_tracepoint_open(const char *name, int prog_fd);
+int bpf_load_btf(void *btf, __u32 btf_size, char *log_buf, __u32 log_buf_size,
+		 bool do_log);
 #endif

tools/lib/bpf/btf.c

+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2018 Facebook */
+
+#include <stdlib.h>
+#include <stdint.h>
+#include <string.h>
+#include <unistd.h>
+#include <errno.h>
+#include <linux/err.h>
+#include <linux/btf.h>
+#include "btf.h"
+#include "bpf.h"
+
+#define elog(fmt, ...) { if (err_log) err_log(fmt, ##__VA_ARGS__); }
+#define max(a, b) ((a) > (b) ? (a) : (b))
+#define min(a, b) ((a) < (b) ? (a) : (b))
+
+#define BTF_MAX_NR_TYPES 65535
+
+static struct btf_type btf_void;
+
+struct btf {
+	union {
+		struct btf_header *hdr;
+		void *data;
+	};
+	struct btf_type **types;
+	const char *strings;
+	void *nohdr_data;
+	uint32_t nr_types;
+	uint32_t types_size;
+	uint32_t data_size;
+	int fd;
+};
+
+static const char *btf_name_by_offset(const struct btf *btf, uint32_t offset)
+{
+	if (!BTF_STR_TBL_ELF_ID(offset) &&
+	    BTF_STR_OFFSET(offset) < btf->hdr->str_len)
+		return &btf->strings[BTF_STR_OFFSET(offset)];
+	else
+		return NULL;
+}
+
+static int btf_add_type(struct btf *btf, struct btf_type *t)
+{
+	if (btf->types_size - btf->nr_types < 2) {
+		struct btf_type **new_types;
+		u32 expand_by, new_size;
+
+		if (btf->types_size == BTF_MAX_NR_TYPES)
+			return -E2BIG;
+
+		expand_by = max(btf->types_size >> 2, 16);
+		new_size = min(BTF_MAX_NR_TYPES, btf->types_size + expand_by);
+
+		new_types = realloc(btf->types, sizeof(*new_types) * new_size);
+		if (!new_types)
+			return -ENOMEM;
+
+		if (btf->nr_types == 0)
+			new_types[0] = &btf_void;
+
+		btf->types = new_types;
+		btf->types_size = new_size;
+	}
+
+	btf->types[++(btf->nr_types)] = t;
+
+	return 0;
+}
+
+static int btf_parse_hdr(struct btf *btf, btf_print_fn_t err_log)
+{
+	const struct btf_header *hdr = btf->hdr;
+	u32 meta_left;
+
+	if (btf->data_size < sizeof(struct btf_header)) {
+		elog("BTF header not found\n");
+		return -EINVAL;
+	}
+
+	if (hdr->magic != BTF_MAGIC) {
+		elog("Invalid BTF magic:%x\n", hdr->magic);
+		return -EINVAL;
+	}
+
+	if (hdr->version != BTF_VERSION) {
+		elog("Unsupported BTF version:%u\n", hdr->version);
+		return -ENOTSUP;
+	}
+
+	if (hdr->flags) {
+		elog("Unsupported BTF flags:%x\n", hdr->flags);
+		return -ENOTSUP;
+	}
+
+	meta_left = btf->data_size - sizeof(*hdr);
+	if (!meta_left) {
+		elog("BTF has no data\n");
+		return -EINVAL;
+	}
+
+	if (meta_left < hdr->type_off) {
+		elog("Invalid BTF type section offset:%u\n", hdr->type_off);
+		return -EINVAL;
+	}
+
+	if (meta_left < hdr->str_off) {
+		elog("Invalid BTF string section offset:%u\n", hdr->str_off);
+		return -EINVAL;
+	}
+
+	if (hdr->type_off >= hdr->str_off) {
+		elog("BTF type section offset >= string section offset. No type?\n");
+		return -EINVAL;
+	}
+
+	if (hdr->type_off & 0x02) {
+		elog("BTF type section is not aligned to 4 bytes\n");
+		return -EINVAL;
+	}
+
+	btf->nohdr_data = btf->hdr + 1;
+
+	return 0;
+}
+
+static int btf_parse_str_sec(struct btf *btf, btf_print_fn_t err_log)
+{
+	const struct btf_header *hdr = btf->hdr;
+	const char *start = btf->nohdr_data + hdr->str_off;
+	const char *end = start + btf->hdr->str_len;
+
+	if (!hdr->str_len || hdr->str_len - 1 > BTF_MAX_NAME_OFFSET ||
+	    start[0] || end[-1]) {
+		elog("Invalid BTF string section\n");
+		return -EINVAL;
+	}
+
+	btf->strings = start;
+
+	return 0;
+}
+
+static int btf_parse_type_sec(struct btf *btf, btf_print_fn_t err_log)
+{
+	struct btf_header *hdr = btf->hdr;
+	void *nohdr_data = btf->nohdr_data;
+	void *next_type = nohdr_data + hdr->type_off;
+	void *end_type = nohdr_data + hdr->str_off;
+
+	while (next_type < end_type) {
+		struct btf_type *t = next_type;
+		uint16_t vlen = BTF_INFO_VLEN(t->info);
+		int err;
+
+		next_type += sizeof(*t);
+		switch (BTF_INFO_KIND(t->info)) {
+		case BTF_KIND_INT:
+			next_type += sizeof(int);
+			break;
+		case BTF_KIND_ARRAY:
+			next_type += sizeof(struct btf_array);
+			break;
+		case BTF_KIND_STRUCT:
+		case BTF_KIND_UNION:
+			next_type += vlen * sizeof(struct btf_member);
+			break;
+		case BTF_KIND_ENUM:
+			next_type += vlen * sizeof(struct btf_enum);
+			break;
+		case BTF_KIND_TYPEDEF:
+		case BTF_KIND_PTR:
+		case BTF_KIND_FWD:
+		case BTF_KIND_VOLATILE:
+		case BTF_KIND_CONST:
+		case BTF_KIND_RESTRICT:
+			break;
+		default:
+			elog("Unsupported BTF_KIND:%u\n",
+			     BTF_INFO_KIND(t->info));
+			return -EINVAL;
+		}
+
+		err = btf_add_type(btf, t);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+static const struct btf_type *btf_type_by_id(const struct btf *btf,
+					     uint32_t type_id)
+{
+	if (type_id > btf->nr_types)
+		return NULL;
+
+	return btf->types[type_id];
+}
+
+static bool btf_type_is_void(const struct btf_type *t)
+{
+	return t == &btf_void || BTF_INFO_KIND(t->info) == BTF_KIND_FWD;
+}
+
+static bool btf_type_is_void_or_null(const struct btf_type *t)
+{
+	return !t || btf_type_is_void(t);
+}
+
+static int64_t btf_type_size(const struct btf_type *t)
+{
+	switch (BTF_INFO_KIND(t->info)) {
+	case BTF_KIND_INT:
+	case BTF_KIND_STRUCT:
+	case BTF_KIND_UNION:
+	case BTF_KIND_ENUM:
+		return t->size;
+	case BTF_KIND_PTR:
+		return sizeof(void *);
+	default:
+		return -EINVAL;
+	}
+}
+
+#define MAX_RESOLVE_DEPTH 32
+
+int64_t btf__resolve_size(const struct btf *btf, uint32_t type_id)
+{
+	const struct btf_array *array;
+	const struct btf_type *t;
+	uint32_t nelems = 1;
+	int64_t size = -1;
+	int i;
+
+	t = btf_type_by_id(btf, type_id);
+	for (i = 0; i < MAX_RESOLVE_DEPTH && !btf_type_is_void_or_null(t);
+	     i++) {
+		size = btf_type_size(t);
+		if (size >= 0)
+			break;
+
+		switch (BTF_INFO_KIND(t->info)) {
+		case BTF_KIND_TYPEDEF:
+		case BTF_KIND_VOLATILE:
+		case BTF_KIND_CONST:
+		case BTF_KIND_RESTRICT:
+			type_id = t->type;
+			break;
+		case BTF_KIND_ARRAY:
+			array = (const struct btf_array *)(t + 1);
+			if (nelems && array->nelems > UINT32_MAX / nelems)
+				return -E2BIG;
+			nelems *= array->nelems;
+			type_id = array->type;
+			break;
+		default:
+			return -EINVAL;
+		}
+
+		t = btf_type_by_id(btf, type_id);
+	}
+
+	if (size < 0)
+		return -EINVAL;
+
+	if (nelems && size > UINT32_MAX / nelems)
+		return -E2BIG;
+
+	return nelems * size;
+}
+
+int32_t btf__find_by_name(const struct btf *btf, const char *type_name)
+{
+	uint32_t i;
+
+	if (!strcmp(type_name, "void"))
+		return 0;
+
+	for (i = 1; i <= btf->nr_types; i++) {
+		const struct btf_type *t = btf->types[i];
+		const char *name = btf_name_by_offset(btf, t->name);
+
+		if (name && !strcmp(type_name, name))
+			return i;
+	}
+
+	return -ENOENT;
+}
+
+void btf__free(struct btf *btf)
+{
+	if (!btf)
+		return;
+
+	if (btf->fd != -1)
+		close(btf->fd);
+
+	free(btf->data);
+	free(btf->types);
+	free(btf);
+}
+
+struct btf *btf__new(uint8_t *data, uint32_t size,
+		     btf_print_fn_t err_log)
+{
+	uint32_t log_buf_size = 0;
+	char *log_buf = NULL;
+	struct btf *btf;
+	int err;
+
+	btf = calloc(1, sizeof(struct btf));
+	if (!btf)
+		return ERR_PTR(-ENOMEM);
+
+	btf->fd = -1;
+
+	if (err_log) {
+		log_buf = malloc(BPF_LOG_BUF_SIZE);
+		if (!log_buf) {
+			err = -ENOMEM;
+			goto done;
+		}
+		*log_buf = 0;
+		log_buf_size = BPF_LOG_BUF_SIZE;
+	}
+
+	btf->data = malloc(size);
+	if (!btf->data) {
+		err = -ENOMEM;
+		goto done;
+	}
+
+	memcpy(btf->data, data, size);
+	btf->data_size = size;
+
+	btf->fd = bpf_load_btf(btf->data, btf->data_size,
+			       log_buf, log_buf_size, false);
+
+	if (btf->fd == -1) {
+		err = -errno;
+		elog("Error loading BTF: %s(%d)\n", strerror(errno), errno);
+		if (log_buf && *log_buf)
+			elog("%s\n", log_buf);
+		goto done;
+	}
+
+	err = btf_parse_hdr(btf, err_log);
+	if (err)
+		goto done;
+
+	err = btf_parse_str_sec(btf, err_log);
+	if (err)
+		goto done;
+
+	err = btf_parse_type_sec(btf, err_log);
+
+done:
+	free(log_buf);
+
+	if (err) {
+		btf__free(btf);
+		return ERR_PTR(err);
+	}
+
+	return btf;
+}
+
+int btf__fd(const struct btf *btf)
+{
+	return btf->fd;
+}

tools/lib/bpf/btf.h

+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2018 Facebook */
+
+#ifndef __BPF_BTF_H
+#define __BPF_BTF_H
+
+#include <stdint.h>
+
+#define BTF_ELF_SEC ".BTF"
+
+struct btf;
+
+typedef int (*btf_print_fn_t)(const char *, ...)
+	__attribute__((format(printf, 1, 2)));
+
+void btf__free(struct btf *btf);
+struct btf *btf__new(uint8_t *data, uint32_t size, btf_print_fn_t err_log);
+int32_t btf__find_by_name(const struct btf *btf, const char *type_name);
+int64_t btf__resolve_size(const struct btf *btf, uint32_t type_id);
+int btf__fd(const struct btf *btf);
+
+#endif

tools/lib/bpf/libbpf.c

@@ -45,6 +45,7 @@
 
 #include "libbpf.h"
 #include "bpf.h"
+#include "btf.h"
 
 #ifndef EM_BPF
 #define EM_BPF 247
@@ -212,6 +213,8 @@ struct bpf_map {
 	char *name;
 	size_t offset;
 	struct bpf_map_def def;
+	uint32_t btf_key_id;
+	uint32_t btf_value_id;
 	void *priv;
 	bpf_map_clear_priv_t clear_priv;
 };
@@ -256,6 +259,8 @@ struct bpf_object {
 	 */
 	struct list_head list;
 
+	struct btf *btf;
+
 	void *priv;
 	bpf_object_clear_priv_t clear_priv;
 
@@ -819,7 +824,15 @@ static int bpf_object__elf_collect(struct bpf_object *obj)
 							data->d_size);
 		else if (strcmp(name, "maps") == 0)
 			obj->efile.maps_shndx = idx;
-		else if (sh.sh_type == SHT_SYMTAB) {
+		else if (strcmp(name, BTF_ELF_SEC) == 0) {
+			obj->btf = btf__new(data->d_buf, data->d_size,
+					    __pr_debug);
+			if (IS_ERR(obj->btf)) {
+				pr_warning("Error loading ELF section %s: %ld. Ignored and continue.\n",
+					   BTF_ELF_SEC, PTR_ERR(obj->btf));
+				obj->btf = NULL;
+			}
+		} else if (sh.sh_type == SHT_SYMTAB) {
 			if (obj->efile.symbols) {
 				pr_warning("bpf: multiple SYMTAB in %s\n",
 					   obj->path);
@@ -996,33 +1009,126 @@ bpf_program__collect_reloc(struct bpf_program *prog, GElf_Shdr *shdr,
 	return 0;
 }
 
+static int bpf_map_find_btf_info(struct bpf_map *map, const struct btf *btf)
+{
+	struct bpf_map_def *def = &map->def;
+	const size_t max_name = 256;
+	int64_t key_size, value_size;
+	int32_t key_id, value_id;
+	char name[max_name];
+
+	/* Find key type by name from BTF */
+	if (snprintf(name, max_name, "%s_key", map->name) == max_name) {
+		pr_warning("map:%s length of BTF key_type:%s_key is too long\n",
+			   map->name, map->name);
+		return -EINVAL;
+	}
+
+	key_id = btf__find_by_name(btf, name);
+	if (key_id < 0) {
+		pr_debug("map:%s key_type:%s cannot be found in BTF\n",
+			 map->name, name);
+		return key_id;
+	}
+
+	key_size = btf__resolve_size(btf, key_id);
+	if (key_size < 0) {
+		pr_warning("map:%s key_type:%s cannot get the BTF type_size\n",
+			   map->name, name);
+		return key_size;
+	}
+
+	if (def->key_size != key_size) {
+		pr_warning("map:%s key_type:%s has BTF type_size:%ld != key_size:%u\n",
+			   map->name, name, key_size, def->key_size);
+		return -EINVAL;
+	}
+
+	/* Find value type from BTF */
+	if (snprintf(name, max_name, "%s_value", map->name) == max_name) {
+		pr_warning("map:%s length of BTF value_type:%s_value is too long\n",
+			  map->name, map->name);
+		return -EINVAL;
+	}
+
+	value_id = btf__find_by_name(btf, name);
+	if (value_id < 0) {
+		pr_debug("map:%s value_type:%s cannot be found in BTF\n",
+			 map->name, name);
+		return value_id;
+	}
+
+	value_size = btf__resolve_size(btf, value_id);
+	if (value_size < 0) {
+		pr_warning("map:%s value_type:%s cannot get the BTF type_size\n",
+			   map->name, name);
+		return value_size;
+	}
+
+	if (def->value_size != value_size) {
+		pr_warning("map:%s value_type:%s has BTF type_size:%ld != value_size:%u\n",
+			   map->name, name, value_size, def->value_size);
+		return -EINVAL;
+	}
+
+	map->btf_key_id = key_id;
+	map->btf_value_id = value_id;
+
+	return 0;
+}
+
 static int
 bpf_object__create_maps(struct bpf_object *obj)
 {
+	struct bpf_create_map_attr create_attr = {};
 	unsigned int i;
+	int err;
 
 	for (i = 0; i < obj->nr_maps; i++) {
-		struct bpf_map_def *def = &obj->maps[i].def;
-		int *pfd = &obj->maps[i].fd;
-
-		*pfd = bpf_create_map_name(def->type,
-					   obj->maps[i].name,
-					   def->key_size,
-					   def->value_size,
-					   def->max_entries,
-					   def->map_flags);
+		struct bpf_map *map = &obj->maps[i];
+		struct bpf_map_def *def = &map->def;
+		int *pfd = &map->fd;
+
+		create_attr.name = map->name;
+		create_attr.map_type = def->type;
+		create_attr.map_flags = def->map_flags;
+		create_attr.key_size = def->key_size;
+		create_attr.value_size = def->value_size;
+		create_attr.max_entries = def->max_entries;
+		create_attr.btf_fd = 0;
+		create_attr.btf_key_id = 0;
+		create_attr.btf_value_id = 0;
+
+		if (obj->btf && !bpf_map_find_btf_info(map, obj->btf)) {
+			create_attr.btf_fd = btf__fd(obj->btf);
+			create_attr.btf_key_id = map->btf_key_id;
+			create_attr.btf_value_id = map->btf_value_id;
+		}
+
+		*pfd = bpf_create_map_xattr(&create_attr);
+		if (*pfd < 0 && create_attr.btf_key_id) {
+			pr_warning("Error in bpf_create_map_xattr(%s):%s(%d). Retrying without BTF.\n",
+				   map->name, strerror(errno), errno);
+			create_attr.btf_fd = 0;
+			create_attr.btf_key_id = 0;
+			create_attr.btf_value_id = 0;
+			map->btf_key_id = 0;
+			map->btf_value_id = 0;
+			*pfd = bpf_create_map_xattr(&create_attr);
+		}
+
 		if (*pfd < 0) {
 			size_t j;
-			int err = *pfd;
 
+			err = *pfd;
 			pr_warning("failed to create map (name: '%s'): %s\n",
-				   obj->maps[i].name,
+				   map->name,
 				   strerror(errno));
 			for (j = 0; j < i; j++)
 				zclose(obj->maps[j].fd);
 			return err;
 		}
-		pr_debug("create map %s: fd=%d\n", obj->maps[i].name, *pfd);
+		pr_debug("create map %s: fd=%d\n", map->name, *pfd);
 	}
 
 	return 0;
@@ -1641,6 +1747,7 @@ void bpf_object__close(struct bpf_object *obj)
 
 	bpf_object__elf_finish(obj);
 	bpf_object__unload(obj);
+	btf__free(obj->btf);
 
 	for (i = 0; i < obj->nr_maps; i++) {
 		zfree(&obj->maps[i].name);
@@ -1692,6 +1799,11 @@ unsigned int bpf_object__kversion(struct bpf_object *obj)
 	return obj ? obj->kern_version : 0;
 }
 
+int bpf_object__btf_fd(const struct bpf_object *obj)
+{
+	return obj->btf ? btf__fd(obj->btf) : -1;
+}
+
 int bpf_object__set_priv(struct bpf_object *obj, void *priv,
 			 bpf_object_clear_priv_t clear_priv)
 {
@@ -1937,6 +2049,16 @@ const char *bpf_map__name(struct bpf_map *map)
 	return map ? map->name : NULL;
 }
 
+uint32_t bpf_map__btf_key_id(const struct bpf_map *map)
+{
+	return map ? map->btf_key_id : 0;
+}
+
+uint32_t bpf_map__btf_value_id(const struct bpf_map *map)
+{
+	return map ? map->btf_value_id : 0;
+}
+
 int bpf_map__set_priv(struct bpf_map *map, void *priv,
 		     bpf_map_clear_priv_t clear_priv)
 {

tools/lib/bpf/libbpf.h

@@ -78,6 +78,7 @@ int bpf_object__load(struct bpf_object *obj);
 int bpf_object__unload(struct bpf_object *obj);
 const char *bpf_object__name(struct bpf_object *obj);
 unsigned int bpf_object__kversion(struct bpf_object *obj);
+int bpf_object__btf_fd(const struct bpf_object *obj);
 
 struct bpf_object *bpf_object__next(struct bpf_object *prev);
 #define bpf_object__for_each_safe(pos, tmp)			\
@@ -241,6 +242,8 @@ bpf_map__next(struct bpf_map *map, struct bpf_object *obj);
 int bpf_map__fd(struct bpf_map *map);
 const struct bpf_map_def *bpf_map__def(struct bpf_map *map);
 const char *bpf_map__name(struct bpf_map *map);
+uint32_t bpf_map__btf_key_id(const struct bpf_map *map);
+uint32_t bpf_map__btf_value_id(const struct bpf_map *map);
 
 typedef void (*bpf_map_clear_priv_t)(struct bpf_map *, void *);
 int bpf_map__set_priv(struct bpf_map *map, void *priv,

tools/testing/selftests/bpf/Makefile

@@ -24,14 +24,15 @@ urandom_read: urandom_read.c
 # Order correspond to 'make run_tests' order
 TEST_GEN_PROGS = test_verifier test_tag test_maps test_lru_map test_lpm_map test_progs \
 	test_align test_verifier_log test_dev_cgroup test_tcpbpf_user \
-	test_sock test_sock_addr
+	test_sock test_sock_addr test_btf
 
 TEST_GEN_FILES = test_pkt_access.o test_xdp.o test_l4lb.o test_tcp_estats.o test_obj_id.o \
 	test_pkt_md_access.o test_xdp_redirect.o test_xdp_meta.o sockmap_parse_prog.o     \
 	sockmap_verdict_prog.o dev_cgroup.o sample_ret0.o test_tracepoint.o \
 	test_l4lb_noinline.o test_xdp_noinline.o test_stacktrace_map.o \
 	sample_map_ret0.o test_tcpbpf_kern.o test_stacktrace_build_id.o \
-	sockmap_tcp_msg_prog.o connect4_prog.o connect6_prog.o test_adjust_tail.o
+	sockmap_tcp_msg_prog.o connect4_prog.o connect6_prog.o test_adjust_tail.o \
+	test_btf_haskv.o test_btf_nokv.o
 
 # Order correspond to 'make run_tests' order
 TEST_PROGS := test_kmod.sh \
@@ -66,6 +67,8 @@ $(BPFOBJ): force
 
 CLANG ?= clang
 LLC   ?= llc
+LLVM_OBJCOPY ?= llvm-objcopy
+BTF_PAHOLE ?= pahole
 
 PROBE := $(shell $(LLC) -march=bpf -mcpu=probe -filetype=null /dev/null 2>&1)
 
@@ -83,9 +86,26 @@ CLANG_FLAGS = -I. -I./include/uapi -I../../../include/uapi \
 $(OUTPUT)/test_l4lb_noinline.o: CLANG_FLAGS += -fno-inline
 $(OUTPUT)/test_xdp_noinline.o: CLANG_FLAGS += -fno-inline
 
+BTF_LLC_PROBE := $(shell $(LLC) -march=bpf -mattr=help |& grep dwarfris)
+BTF_PAHOLE_PROBE := $(shell $(BTF_PAHOLE) --help |& grep BTF)
+BTF_OBJCOPY_PROBE := $(shell $(LLVM_OBJCOPY) --version |& grep LLVM)
+
+ifneq ($(BTF_LLC_PROBE),)
+ifneq ($(BTF_PAHOLE_PROBE),)
+ifneq ($(BTF_OBJCOPY_PROBE),)
+	CLANG_FLAGS += -g
+	LLC_FLAGS += -mattr=dwarfris
+	DWARF2BTF = y
+endif
+endif
+endif
+
 $(OUTPUT)/%.o: %.c
 	$(CLANG) $(CLANG_FLAGS) \
 		 -O2 -target bpf -emit-llvm -c $< -o - |      \
-	$(LLC) -march=bpf -mcpu=$(CPU) -filetype=obj -o $@
+	$(LLC) -march=bpf -mcpu=$(CPU) $(LLC_FLAGS) -filetype=obj -o $@
+ifeq ($(DWARF2BTF),y)
+	$(BTF_PAHOLE) -J $@
+endif
 
 EXTRA_CLEAN := $(TEST_CUSTOM_PROGS)

tools/testing/selftests/bpf/test_btf_haskv.c

+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2018 Facebook */
+#include <linux/bpf.h>
+#include "bpf_helpers.h"
+
+int _version SEC("version") = 1;
+
+struct ipv_counts {
+	unsigned int v4;
+	unsigned int v6;
+};
+
+typedef int btf_map_key;
+typedef struct ipv_counts btf_map_value;
+btf_map_key dumm_key;
+btf_map_value dummy_value;
+
+struct bpf_map_def SEC("maps") btf_map = {
+	.type = BPF_MAP_TYPE_ARRAY,
+	.key_size = sizeof(int),
+	.value_size = sizeof(struct ipv_counts),
+	.max_entries = 4,
+};
+
+struct dummy_tracepoint_args {
+	unsigned long long pad;
+	struct sock *sock;
+};
+
+SEC("dummy_tracepoint")
+int _dummy_tracepoint(struct dummy_tracepoint_args *arg)
+{
+	struct ipv_counts *counts;
+	int key = 0;
+
+	if (!arg->sock)
+		return 0;
+
+	counts = bpf_map_lookup_elem(&btf_map, &key);
+	if (!counts)
+		return 0;
+
+	counts->v6++;
+
+	return 0;
+}
+
+char _license[] SEC("license") = "GPL";

tools/testing/selftests/bpf/test_btf_nokv.c

+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2018 Facebook */
+#include <linux/bpf.h>
+#include "bpf_helpers.h"
+
+int _version SEC("version") = 1;
+
+struct ipv_counts {
+	unsigned int v4;
+	unsigned int v6;
+};
+
+struct bpf_map_def SEC("maps") btf_map = {
+	.type = BPF_MAP_TYPE_ARRAY,
+	.key_size = sizeof(int),
+	.value_size = sizeof(struct ipv_counts),
+	.max_entries = 4,
+};
+
+struct dummy_tracepoint_args {
+	unsigned long long pad;
+	struct sock *sock;
+};
+
+SEC("dummy_tracepoint")
+int _dummy_tracepoint(struct dummy_tracepoint_args *arg)
+{
+	struct ipv_counts *counts;
+	int key = 0;
+
+	if (!arg->sock)
+		return 0;
+
+	counts = bpf_map_lookup_elem(&btf_map, &key);
+	if (!counts)
+		return 0;
+
+	counts->v6++;
+
+	return 0;
+}
+
+char _license[] SEC("license") = "GPL";