[PATCH bpf-next 1/2] libbpf: add bpf_map__reuse_fd_from_loader_ctx() to allow shared maps

[John Fastabend <john.fastabend@gmail.com>]

This lets the gen_loader use a map that may already be loaded/pinned on
the target system. When we are loading many signed programs on the target
we often need them to share maps to communicate/pass data between the
programs.

Note it is slightly complicated by needing to ensure on error/close that
we only emit_sys_close_blob() for map ids that are not these shared
type otherwise we could close a shared map unintentionally. To track this
we added a borrowed_fd_array.

Signed-off-by: John Fastabend <john.fastabend@gmail.com>
---
 tools/lib/bpf/bpf_gen_internal.h |  2 ++
 tools/lib/bpf/gen_loader.c       | 50 ++++++++++++++++++++++++++++++--
 tools/lib/bpf/libbpf.c           | 11 +++++++
 tools/lib/bpf/libbpf.h           |  1 +
 tools/lib/bpf/libbpf.map         |  1 +
 5 files changed, 63 insertions(+), 2 deletions(-)

diff --git a/tools/lib/bpf/bpf_gen_internal.h b/tools/lib/bpf/bpf_gen_internal.h
index 49af4260b8e6..e555c3165105 100644
--- a/tools/lib/bpf/bpf_gen_internal.h
+++ b/tools/lib/bpf/bpf_gen_internal.h
@@ -50,6 +50,7 @@ struct bpf_gen {
 	struct ksym_desc *ksyms;
 	__u32 nr_ksyms;
 	int fd_array;
+	int borrowed_fd_array;
 	int nr_fd_array;
 	int hash_insn_offset[SHA256_DWORD_SIZE];
 };
@@ -62,6 +63,7 @@ void bpf_gen__map_create(struct bpf_gen *gen,
 			 enum bpf_map_type map_type, const char *map_name,
 			 __u32 key_size, __u32 value_size, __u32 max_entries,
 			 struct bpf_map_create_opts *map_attr, int map_idx);
+void bpf_gen__map_reuse_fd(struct bpf_gen *gen, int map_idx);
 void bpf_gen__prog_load(struct bpf_gen *gen,
 			enum bpf_prog_type prog_type, const char *prog_name,
 			const char *license, struct bpf_insn *insns, size_t insn_cnt,
diff --git a/tools/lib/bpf/gen_loader.c b/tools/lib/bpf/gen_loader.c
index c7f2d2ac7bb3..1cb00e0244b5 100644
--- a/tools/lib/bpf/gen_loader.c
+++ b/tools/lib/bpf/gen_loader.c
@@ -47,6 +47,11 @@ static int blob_fd_array_off(struct bpf_gen *gen, int index)
 	return gen->fd_array + index * sizeof(int);
 }
 
+static int blob_borrowed_fd_array_off(struct bpf_gen *gen, int index)
+{
+	return gen->borrowed_fd_array + index * sizeof(int);
+}
+
 static int realloc_insn_buf(struct bpf_gen *gen, __u32 size)
 {
 	size_t off = gen->insn_cur - gen->insn_start;
@@ -111,6 +116,7 @@ static void emit2(struct bpf_gen *gen, struct bpf_insn insn1, struct bpf_insn in
 
 static int add_data(struct bpf_gen *gen, const void *data, __u32 size);
 static void emit_sys_close_blob(struct bpf_gen *gen, int blob_off);
+static void emit_sys_close_owned_blob(struct bpf_gen *gen, int map_idx);
 static void emit_signature_match(struct bpf_gen *gen);
 
 void bpf_gen__init(struct bpf_gen *gen, int log_level, int nr_progs, int nr_maps)
@@ -119,6 +125,7 @@ void bpf_gen__init(struct bpf_gen *gen, int log_level, int nr_progs, int nr_maps
 	int i;
 
 	gen->fd_array = add_data(gen, NULL, MAX_FD_ARRAY_SZ * sizeof(int));
+	gen->borrowed_fd_array = add_data(gen, NULL, nr_maps * sizeof(int));
 	gen->log_level = log_level;
 	/* save ctx pointer into R6 */
 	emit(gen, BPF_MOV64_REG(BPF_REG_6, BPF_REG_1));
@@ -137,9 +144,10 @@ void bpf_gen__init(struct bpf_gen *gen, int log_level, int nr_progs, int nr_maps
 			      /* size of cleanup code below (including map fd cleanup) */
 			      (nr_progs_sz / 4) * 3 + 2 +
 			      /* 6 insns for emit_sys_close_blob,
+			       * 4 insns for borrowed fd check,
 			       * 6 insns for debug_regs in emit_sys_close_blob
 			       */
-			      nr_maps * (6 + (gen->log_level ? 6 : 0))));
+			      nr_maps * (10 + (gen->log_level ? 6 : 0))));
 
 	/* remember the label where all error branches will jump to */
 	gen->cleanup_label = gen->insn_cur - gen->insn_start;
@@ -150,7 +158,7 @@ void bpf_gen__init(struct bpf_gen *gen, int log_level, int nr_progs, int nr_maps
 		emit(gen, BPF_EMIT_CALL(BPF_FUNC_sys_close));
 	}
 	for (i = 0; i < nr_maps; i++)
-		emit_sys_close_blob(gen, blob_fd_array_off(gen, i));
+		emit_sys_close_owned_blob(gen, i);
 	/* R7 contains the error code from sys_bpf. Copy it into R0 and exit. */
 	emit(gen, BPF_MOV64_REG(BPF_REG_0, BPF_REG_7));
 	emit(gen, BPF_EXIT_INSN());
@@ -377,6 +385,17 @@ static void emit_sys_close_blob(struct bpf_gen *gen, int blob_off)
 	__emit_sys_close(gen);
 }
 
+static void emit_sys_close_owned_blob(struct bpf_gen *gen, int map_idx)
+{
+	int close_insn_cnt = 6 + (gen->log_level ? 6 : 0);
+
+	emit2(gen, BPF_LD_IMM64_RAW_FULL(BPF_REG_0, BPF_PSEUDO_MAP_IDX_VALUE,
+					 0, 0, 0, blob_borrowed_fd_array_off(gen, map_idx)));
+	emit(gen, BPF_LDX_MEM(BPF_W, BPF_REG_1, BPF_REG_0, 0));
+	emit(gen, BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, close_insn_cnt));
+	emit_sys_close_blob(gen, blob_fd_array_off(gen, map_idx));
+}
+
 static void compute_sha_update_offsets(struct bpf_gen *gen);
 
 int bpf_gen__finish(struct bpf_gen *gen, int nr_progs, int nr_maps)
@@ -460,6 +479,11 @@ void bpf_gen__free(struct bpf_gen *gen)
 	_val;							\
 })
 
+static void set_blob_borrowed_fd_array_off(struct bpf_gen *gen, int index)
+{
+	*(__u32 *)(gen->data_start + blob_borrowed_fd_array_off(gen, index)) = tgt_endian(1);
+}
+
 static void compute_sha_update_offsets(struct bpf_gen *gen)
 {
 	__u64 sha[SHA256_DWORD_SIZE];
@@ -596,6 +620,28 @@ void bpf_gen__map_create(struct bpf_gen *gen,
 		emit_sys_close_stack(gen, stack_off(inner_map_fd));
 }
 
+void bpf_gen__map_reuse_fd(struct bpf_gen *gen, int map_idx)
+{
+	int idx;
+
+	if (map_idx != gen->nr_maps) {
+		gen->error = -EDOM;
+		return;
+	}
+
+	idx = add_map_fd(gen);
+	if (gen->error)
+		return;
+
+	set_blob_borrowed_fd_array_off(gen, idx);
+	move_ctx2blob(gen, blob_fd_array_off(gen, idx), 4,
+		      sizeof(struct bpf_loader_ctx) +
+		      sizeof(struct bpf_map_desc) * map_idx +
+		      offsetof(struct bpf_map_desc, map_fd),
+		      false);
+	pr_debug("gen: map_reuse_fd idx %d\n", map_idx);
+}
+
 static void emit_signature_match(struct bpf_gen *gen)
 {
 	__s64 off;
diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index 7162146280a8..af1553e9d4df 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -5085,6 +5085,15 @@ int bpf_map__reuse_fd(struct bpf_map *map, int fd)
 	return libbpf_err(err);
 }
 
+int bpf_map__reuse_fd_from_loader_ctx(struct bpf_map *map)
+{
+	if (map_is_created(map))
+		return libbpf_err(-EBUSY);
+
+	map->reused = true;
+	return 0;
+}
+
 __u32 bpf_map__max_entries(const struct bpf_map *map)
 {
 	return map->def.max_entries;
@@ -5659,6 +5668,8 @@ bpf_object__create_maps(struct bpf_object *obj)
 		}
 
 		if (map->reused) {
+			if (obj->gen_loader)
+				bpf_gen__map_reuse_fd(obj->gen_loader, map - obj->maps);
 			pr_debug("map '%s': skipping creation (preset fd=%d)\n",
 				 map->name, map->fd);
 		} else {
diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
index b965ad571540..4145a16c8963 100644
--- a/tools/lib/bpf/libbpf.h
+++ b/tools/lib/bpf/libbpf.h
@@ -1135,6 +1135,7 @@ LIBBPF_API bool bpf_map__autoattach(const struct bpf_map *map);
  */
 LIBBPF_API int bpf_map__fd(const struct bpf_map *map);
 LIBBPF_API int bpf_map__reuse_fd(struct bpf_map *map, int fd);
+LIBBPF_API int bpf_map__reuse_fd_from_loader_ctx(struct bpf_map *map);
 /* get map name */
 LIBBPF_API const char *bpf_map__name(const struct bpf_map *map);
 /* get/set map type */
diff --git a/tools/lib/bpf/libbpf.map b/tools/lib/bpf/libbpf.map
index b731df19ae69..d6c2f1d75cda 100644
--- a/tools/lib/bpf/libbpf.map
+++ b/tools/lib/bpf/libbpf.map
@@ -458,6 +458,7 @@ LIBBPF_1.7.0 {
 
 LIBBPF_1.8.0 {
 	global:
+		bpf_map__reuse_fd_from_loader_ctx;
 		bpf_program__attach_tracing_multi;
 		bpf_program__clone;
 		btf__new_empty_opts;
-- 
2.53.0