bpf: pass attached BTF to the bpf_struct_ops subsystem

Pass the fd of a btf from the userspace to the bpf() syscall, and then convert the fd into a btf. The btf is generated from the module that defines the target BPF struct_ops type. In order to inform the kernel about the module that defines the target struct_ops type, the userspace program needs to provide a btf fd for the respective module's btf. This btf contains essential information on the types defined within the module, including the target struct_ops type. A btf fd must be provided to the kernel for struct_ops maps and for the bpf programs attached to those maps. In the case of the bpf programs, the attach_btf_obj_fd parameter is passed as part of the bpf_attr and is converted into a btf. This btf is then stored in the prog->aux->attach_btf field. Here, it just let the verifier access attach_btf directly. In the case of struct_ops maps, a btf fd is passed as value_type_btf_obj_fd of bpf_attr. The bpf_struct_ops_map_alloc() function converts the fd to a btf and stores it as st_map->btf. A flag BPF_F_VTYPE_BTF_OBJ_FD is added for map_flags to indicate that the value of value_type_btf_obj_fd is set.

Kui-Feng Lee committed 2 years ago

Commit fcc2c1f1 parent 689423d5 files · +70 −22Browse files

Unified · Split

5 files

include/uapi/linux/bpf.h +8 −0
kernel/bpf/bpf_struct_ops.c +47 −18
kernel/bpf/syscall.c +1 −1
kernel/bpf/verifier.c +6 −3
tools/include/uapi/linux/bpf.h +8 −0

include/uapi/linux/bpf.h+8−0

@@ -1330,6 +1330,9 @@ enum { · 1329 unmodified lines
1330	1330
1331	1331	/* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */
1332	1332	BPF_F_PATH_FD = (1U << 14),
	1333	+
	1334	+/* Flag for value_type_btf_obj_fd, the fd is available */
	1335	+ BPF_F_VTYPE_BTF_OBJ_FD = (1U << 15),
1333	1336	};
1334	1337
1335	1338	/* Flags for BPF_PROG_QUERY. */
@@ -1403,6 +1406,11 @@ union bpf_attr { · 67 unmodified lines
1403	1406	* to using 5 hash functions).
1404	1407	*/
1405	1408	__u64 map_extra;
	1409	+
	1410	+ __s32 value_type_btf_obj_fd; /* fd pointing to a BTF
	1411	+ * type data for
	1412	+ * btf_vmlinux_value_type_id.
	1413	+ */
1406	1414	};
1407	1415
1408	1416	struct { /* anonymous struct used by BPF_MAP__ELEM commands /

kernel/bpf/bpf_struct_ops.c+47−18

@@ -641,6 +641,7 @@ static void __bpf_struct_ops_map_free(struct bpf_map *map) · 640 unmodified lines
641	641	bpf_jit_uncharge_modmem(PAGE_SIZE);
642	642	}
643	643	bpf_map_area_free(st_map->uvalue);
	644	+ btf_put(st_map->btf);
644	645	bpf_map_area_free(st_map);
645	646	}
646	647
@@ -669,7 +670,8 @@ static void bpf_struct_ops_map_free(struct bpf_map *map) · 22 unmodified lines
669	670	static int bpf_struct_ops_map_alloc_check(union bpf_attr *attr)
670	671	{
671	672	if (attr->key_size != sizeof(unsigned int) \|\| attr->max_entries != 1 \|\|
672		− (attr->map_flags & ~BPF_F_LINK) \|\| !attr->btf_vmlinux_value_type_id)
	673	+ (attr->map_flags & ~(BPF_F_LINK \| BPF_F_VTYPE_BTF_OBJ_FD)) \|\|
	674	+ !attr->btf_vmlinux_value_type_id)
673	675	return -EINVAL;
674	676	return 0;
675	677	}
@@ -681,15 +683,36 @@ static struct bpf_map bpf_struct_ops_map_alloc(union bpf_attr attr) · 5 unmodified lines
681	683	struct bpf_struct_ops_map *st_map;
682	684	const struct btf_type t, vt;
683	685	struct bpf_map *map;
	686	+ struct btf *btf;
684	687	int ret;
685	688
686		− st_ops_desc = bpf_struct_ops_find_value(btf_vmlinux, attr->btf_vmlinux_value_type_id);
687		− if (!st_ops_desc)
688		− return ERR_PTR(-ENOTSUPP);
	689	+ if (attr->map_flags & BPF_F_VTYPE_BTF_OBJ_FD) {
	690	+ /* The map holds btf for its whole life time. */
	691	+ btf = btf_get_by_fd(attr->value_type_btf_obj_fd);
	692	+ if (IS_ERR(btf))
	693	+ return ERR_CAST(btf);
	694	+ if (!btf_is_module(btf)) {
	695	+ btf_put(btf);
	696	+ return ERR_PTR(-EINVAL);
	697	+ }
	698	+ } else {
	699	+ btf = bpf_get_btf_vmlinux();
	700	+ if (IS_ERR(btf))
	701	+ return ERR_CAST(btf);
	702	+ btf_get(btf);
	703	+ }
	704	+
	705	+ st_ops_desc = bpf_struct_ops_find_value(btf, attr->btf_vmlinux_value_type_id);
	706	+ if (!st_ops_desc) {
	707	+ ret = -ENOTSUPP;
	708	+ goto errout;
	709	+ }
689	710
690	711	vt = st_ops_desc->value_type;
691		− if (attr->value_size != vt->size)
692		− return ERR_PTR(-EINVAL);
	712	+ if (attr->value_size != vt->size) {
	713	+ ret = -EINVAL;
	714	+ goto errout;
	715	+ }
693	716
694	717	t = st_ops_desc->type;
695	718
@@ -700,17 +723,17 @@ static struct bpf_map bpf_struct_ops_map_alloc(union bpf_attr attr) · 4 unmodified lines
700	723	(vt->size - sizeof(struct bpf_struct_ops_value));
701	724
702	725	st_map = bpf_map_area_alloc(st_map_size, NUMA_NO_NODE);
703		− if (!st_map)
704		− return ERR_PTR(-ENOMEM);
	726	+ if (!st_map) {
	727	+ ret = -ENOMEM;
	728	+ goto errout;
	729	+ }
705	730
706	731	st_map->st_ops_desc = st_ops_desc;
707	732	map = &st_map->map;
708	733
709	734	ret = bpf_jit_charge_modmem(PAGE_SIZE);
710		− if (ret) {
711		− __bpf_struct_ops_map_free(map);
712		− return ERR_PTR(ret);
713		− }
	735	+ if (ret)
	736	+ goto errout_free;
714	737
715	738	st_map->image = arch_alloc_bpf_trampoline(PAGE_SIZE);
716	739	if (!st_map->image) {
@@ -719,24 +742,30 @@ static struct bpf_map bpf_struct_ops_map_alloc(union bpf_attr attr) · 2 unmodified lines
719	742	* here.
720	743	*/
721	744	bpf_jit_uncharge_modmem(PAGE_SIZE);
722		− __bpf_struct_ops_map_free(map);
723		− return ERR_PTR(-ENOMEM);
	745	+ ret = -ENOMEM;
	746	+ goto errout_free;
724	747	}
725	748	st_map->uvalue = bpf_map_area_alloc(vt->size, NUMA_NO_NODE);
726	749	st_map->links =
727	750	bpf_map_area_alloc(btf_type_vlen(t) * sizeof(struct bpf_links *),
728	751	NUMA_NO_NODE);
729	752	if (!st_map->uvalue \|\| !st_map->links) {
730		− __bpf_struct_ops_map_free(map);
731		− return ERR_PTR(-ENOMEM);
	753	+ ret = -ENOMEM;
	754	+ goto errout_free;
732	755	}
733		−
734		− st_map->btf = btf_vmlinux;
	756	+ st_map->btf = btf;
735	757
736	758	mutex_init(&st_map->lock);
737	759	bpf_map_init_from_attr(map, attr);
738	760
739	761	return map;
	762	+
	763	+errout_free:
	764	+ __bpf_struct_ops_map_free(map);
	765	+errout:
	766	+ btf_put(btf);
	767	+
	768	+ return ERR_PTR(ret);
740	769	}
741	770
742	771	static u64 bpf_struct_ops_map_mem_usage(const struct bpf_map *map)

kernel/bpf/syscall.c+1−1

@@ -1123,7 +1123,7 @@ free_map_tab: · 1122 unmodified lines
1123	1123	return ret;
1124	1124	}
1125	1125
1126		−#define BPF_MAP_CREATE_LAST_FIELD map_extra
	1126	+#define BPF_MAP_CREATE_LAST_FIELD value_type_btf_obj_fd
1127	1127	/* called via syscall */
1128	1128	static int map_create(union bpf_attr *attr)
1129	1129	{

kernel/bpf/verifier.c+6−3

@@ -20290,6 +20290,7 @@ static int check_struct_ops_btf_id(struct bpf_verifier_env *env) · 20289 unmodified lines
20290	20290	const struct btf_member *member;
20291	20291	struct bpf_prog *prog = env->prog;
20292	20292	u32 btf_id, member_idx;
	20293	+ struct btf *btf;
20293	20294	const char *mname;
20294	20295
20295	20296	if (!prog->gpl_compatible) {
@@ -20297,8 +20298,10 @@ static int check_struct_ops_btf_id(struct bpf_verifier_env *env) · 1 unmodified line
20297	20298	return -EINVAL;
20298	20299	}
20299	20300
	20301	+ btf = prog->aux->attach_btf ?: bpf_get_btf_vmlinux();
	20302	+
20300	20303	btf_id = prog->aux->attach_btf_id;
20301		− st_ops_desc = bpf_struct_ops_find(btf_vmlinux, btf_id);
	20304	+ st_ops_desc = bpf_struct_ops_find(btf, btf_id);
20302	20305	if (!st_ops_desc) {
20303	20306	verbose(env, "attach_btf_id %u is not a supported struct\n",
20304	20307	btf_id);
@@ -20315,8 +20318,8 @@ static int check_struct_ops_btf_id(struct bpf_verifier_env *env) · 10 unmodified lines
20315	20318	}
20316	20319
20317	20320	member = &btf_type_member(t)[member_idx];
20318		− mname = btf_name_by_offset(btf_vmlinux, member->name_off);
20319		− func_proto = btf_type_resolve_func_ptr(btf_vmlinux, member->type,
	20321	+ mname = btf_name_by_offset(btf, member->name_off);
	20322	+ func_proto = btf_type_resolve_func_ptr(btf, member->type,
20320	20323	NULL);
20321	20324	if (!func_proto) {
20322	20325	verbose(env, "attach to invalid member %s(@idx %u) of struct %s\n",

tools/include/uapi/linux/bpf.h+8−0

@@ -1330,6 +1330,9 @@ enum { · 1329 unmodified lines
1330	1330
1331	1331	/* Get path from provided FD in BPF_OBJ_PIN/BPF_OBJ_GET commands */
1332	1332	BPF_F_PATH_FD = (1U << 14),
	1333	+
	1334	+/* Flag for value_type_btf_obj_fd, the fd is available */
	1335	+ BPF_F_VTYPE_BTF_OBJ_FD = (1U << 15),
1333	1336	};
1334	1337
1335	1338	/* Flags for BPF_PROG_QUERY. */
@@ -1403,6 +1406,11 @@ union bpf_attr { · 67 unmodified lines
1403	1406	* to using 5 hash functions).
1404	1407	*/
1405	1408	__u64 map_extra;
	1409	+
	1410	+ __s32 value_type_btf_obj_fd; /* fd pointing to a BTF
	1411	+ * type data for
	1412	+ * btf_vmlinux_value_type_id.
	1413	+ */
1406	1414	};
1407	1415
1408	1416	struct { /* anonymous struct used by BPF_MAP__ELEM commands /

bpf: pass attached BTF to the bpf_struct_ops subsystem

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code