bpf: Introduce bpf_per_cpu_ptr()

Add bpf_per_cpu_ptr() to help bpf programs access percpu vars.
bpf_per_cpu_ptr() has the same semantic as per_cpu_ptr() in the kernel
except that it may return NULL. This happens when the cpu parameter is
out of range. So the caller must check the returned value.

Signed-off-by: Hao Luo <haoluo@google.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Link: https://lore.kernel.org/bpf/20200929235049.2533242-5-haoluo@google.com

3 files changed, 79 insertions, 13 deletions

diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
index 00569afe3d0d..ed7d02e8bc93 100644
--- a/kernel/bpf/btf.c
+++ b/kernel/bpf/btf.c
@@ -188,11 +188,6 @@
 	     i < btf_type_vlen(struct_type);				\
 	     i++, member++)
 
-#define for_each_vsi(i, struct_type, member)			\
-	for (i = 0, member = btf_type_var_secinfo(struct_type);	\
-	     i < btf_type_vlen(struct_type);			\
-	     i++, member++)
-
 #define for_each_vsi_from(i, from, struct_type, member)				\
 	for (i = from, member = btf_type_var_secinfo(struct_type) + from;	\
 	     i < btf_type_vlen(struct_type);					\
@@ -598,11 +593,6 @@ static const struct btf_var *btf_type_var(const struct btf_type *t)
 	return (const struct btf_var *)(t + 1);
 }
 
-static const struct btf_var_secinfo *btf_type_var_secinfo(const struct btf_type *t)
-{
-	return (const struct btf_var_secinfo *)(t + 1);
-}
-
 static const struct btf_kind_operations *btf_type_ops(const struct btf_type *t)
 {
 	return kind_ops[BTF_INFO_KIND(t->info)];
diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
index e825441781ab..14fe3f64fd82 100644
--- a/kernel/bpf/helpers.c
+++ b/kernel/bpf/helpers.c
@@ -623,6 +623,22 @@ const struct bpf_func_proto bpf_copy_from_user_proto = {
 	.arg3_type	= ARG_ANYTHING,
 };
 
+BPF_CALL_2(bpf_per_cpu_ptr, const void *, ptr, u32, cpu)
+{
+	if (cpu >= nr_cpu_ids)
+		return (unsigned long)NULL;
+
+	return (unsigned long)per_cpu_ptr((const void __percpu *)ptr, cpu);
+}
+
+const struct bpf_func_proto bpf_per_cpu_ptr_proto = {
+	.func		= bpf_per_cpu_ptr,
+	.gpl_only	= false,
+	.ret_type	= RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL,
+	.arg1_type	= ARG_PTR_TO_PERCPU_BTF_ID,
+	.arg2_type	= ARG_ANYTHING,
+};
+
 const struct bpf_func_proto bpf_get_current_task_proto __weak;
 const struct bpf_func_proto bpf_probe_read_user_proto __weak;
 const struct bpf_func_proto bpf_probe_read_user_str_proto __weak;
@@ -689,6 +705,8 @@ bpf_base_func_proto(enum bpf_func_id func_id)
 		return &bpf_snprintf_btf_proto;
 	case BPF_FUNC_jiffies64:
 		return &bpf_jiffies64_proto;
+	case BPF_FUNC_bpf_per_cpu_ptr:
+		return &bpf_per_cpu_ptr_proto;
 	default:
 		break;
 	}
diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index fe4965079773..216b8ece23ce 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -238,6 +238,8 @@ struct bpf_call_arg_meta {
 	u64 msize_max_value;
 	int ref_obj_id;
 	int func_id;
+	u32 btf_id;
+	u32 ret_btf_id;
 };
 
 struct btf *btf_vmlinux;
@@ -517,6 +519,7 @@ static const char * const reg_type_str[] = {
 	[PTR_TO_XDP_SOCK]	= "xdp_sock",
 	[PTR_TO_BTF_ID]		= "ptr_",
 	[PTR_TO_BTF_ID_OR_NULL]	= "ptr_or_null_",
+	[PTR_TO_PERCPU_BTF_ID]	= "percpu_ptr_",
 	[PTR_TO_MEM]		= "mem",
 	[PTR_TO_MEM_OR_NULL]	= "mem_or_null",
 	[PTR_TO_RDONLY_BUF]	= "rdonly_buf",
@@ -583,7 +586,9 @@ static void print_verifier_state(struct bpf_verifier_env *env,
 			/* reg->off should be 0 for SCALAR_VALUE */
 			verbose(env, "%lld", reg->var_off.value + reg->off);
 		} else {
-			if (t == PTR_TO_BTF_ID || t == PTR_TO_BTF_ID_OR_NULL)
+			if (t == PTR_TO_BTF_ID ||
+			    t == PTR_TO_BTF_ID_OR_NULL ||
+			    t == PTR_TO_PERCPU_BTF_ID)
 				verbose(env, "%s", kernel_type_name(reg->btf_id));
 			verbose(env, "(id=%d", reg->id);
 			if (reg_type_may_be_refcounted_or_null(t))
@@ -2204,6 +2209,7 @@ static bool is_spillable_regtype(enum bpf_reg_type type)
 	case PTR_TO_RDONLY_BUF_OR_NULL:
 	case PTR_TO_RDWR_BUF:
 	case PTR_TO_RDWR_BUF_OR_NULL:
+	case PTR_TO_PERCPU_BTF_ID:
 		return true;
 	default:
 		return false;
@@ -4017,6 +4023,7 @@ static const struct bpf_reg_types alloc_mem_types = { .types = { PTR_TO_MEM } };
 static const struct bpf_reg_types const_map_ptr_types = { .types = { CONST_PTR_TO_MAP } };
 static const struct bpf_reg_types btf_ptr_types = { .types = { PTR_TO_BTF_ID } };
 static const struct bpf_reg_types spin_lock_types = { .types = { PTR_TO_MAP_VALUE } };
+static const struct bpf_reg_types percpu_btf_ptr_types = { .types = { PTR_TO_PERCPU_BTF_ID } };
 
 static const struct bpf_reg_types *compatible_reg_types[__BPF_ARG_TYPE_MAX] = {
 	[ARG_PTR_TO_MAP_KEY]		= &map_key_value_types,
@@ -4042,6 +4049,7 @@ static const struct bpf_reg_types *compatible_reg_types[__BPF_ARG_TYPE_MAX] = {
 	[ARG_PTR_TO_ALLOC_MEM_OR_NULL]	= &alloc_mem_types,
 	[ARG_PTR_TO_INT]		= &int_ptr_types,
 	[ARG_PTR_TO_LONG]		= &int_ptr_types,
+	[ARG_PTR_TO_PERCPU_BTF_ID]	= &percpu_btf_ptr_types,
 };
 
 static int check_reg_type(struct bpf_verifier_env *env, u32 regno,
@@ -4205,6 +4213,12 @@ skip_type_check:
 		err = check_helper_mem_access(env, regno,
 					      meta->map_ptr->value_size, false,
 					      meta);
+	} else if (arg_type == ARG_PTR_TO_PERCPU_BTF_ID) {
+		if (!reg->btf_id) {
+			verbose(env, "Helper has invalid btf_id in R%d\n", regno);
+			return -EACCES;
+		}
+		meta->ret_btf_id = reg->btf_id;
 	} else if (arg_type == ARG_PTR_TO_SPIN_LOCK) {
 		if (meta->func_id == BPF_FUNC_spin_lock) {
 			if (process_spin_lock(env, regno, true))
@@ -5114,6 +5128,30 @@ static int check_helper_call(struct bpf_verifier_env *env, int func_id, int insn
 		regs[BPF_REG_0].type = PTR_TO_MEM_OR_NULL;
 		regs[BPF_REG_0].id = ++env->id_gen;
 		regs[BPF_REG_0].mem_size = meta.mem_size;
+	} else if (fn->ret_type == RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL) {
+		const struct btf_type *t;
+
+		mark_reg_known_zero(env, regs, BPF_REG_0);
+		t = btf_type_skip_modifiers(btf_vmlinux, meta.ret_btf_id, NULL);
+		if (!btf_type_is_struct(t)) {
+			u32 tsize;
+			const struct btf_type *ret;
+			const char *tname;
+
+			/* resolve the type size of ksym. */
+			ret = btf_resolve_size(btf_vmlinux, t, &tsize);
+			if (IS_ERR(ret)) {
+				tname = btf_name_by_offset(btf_vmlinux, t->name_off);
+				verbose(env, "unable to resolve the size of type '%s': %ld\n",
+					tname, PTR_ERR(ret));
+				return -EINVAL;
+			}
+			regs[BPF_REG_0].type = PTR_TO_MEM_OR_NULL;
+			regs[BPF_REG_0].mem_size = tsize;
+		} else {
+			regs[BPF_REG_0].type = PTR_TO_BTF_ID_OR_NULL;
+			regs[BPF_REG_0].btf_id = meta.ret_btf_id;
+		}
 	} else if (fn->ret_type == RET_PTR_TO_BTF_ID_OR_NULL) {
 		int ret_btf_id;
 
@@ -7523,6 +7561,7 @@ static int check_ld_imm(struct bpf_verifier_env *env, struct bpf_insn *insn)
 			dst_reg->mem_size = aux->btf_var.mem_size;
 			break;
 		case PTR_TO_BTF_ID:
+		case PTR_TO_PERCPU_BTF_ID:
 			dst_reg->btf_id = aux->btf_var.btf_id;
 			break;
 		default:
@@ -9449,10 +9488,14 @@ static int check_pseudo_btf_id(struct bpf_verifier_env *env,
 			       struct bpf_insn *insn,
 			       struct bpf_insn_aux_data *aux)
 {
-	u32 type, id = insn->imm;
+	u32 datasec_id, type, id = insn->imm;
+	const struct btf_var_secinfo *vsi;
+	const struct btf_type *datasec;
 	const struct btf_type *t;
 	const char *sym_name;
+	bool percpu = false;
 	u64 addr;
+	int i;
 
 	if (!btf_vmlinux) {
 		verbose(env, "kernel is missing BTF, make sure CONFIG_DEBUG_INFO_BTF=y is specified in Kconfig.\n");
@@ -9484,12 +9527,27 @@ static int check_pseudo_btf_id(struct bpf_verifier_env *env,
 		return -ENOENT;
 	}
 
+	datasec_id = btf_find_by_name_kind(btf_vmlinux, ".data..percpu",
+					   BTF_KIND_DATASEC);
+	if (datasec_id > 0) {
+		datasec = btf_type_by_id(btf_vmlinux, datasec_id);
+		for_each_vsi(i, datasec, vsi) {
+			if (vsi->type == id) {
+				percpu = true;
+				break;
+			}
+		}
+	}
+
 	insn[0].imm = (u32)addr;
 	insn[1].imm = addr >> 32;
 
 	type = t->type;
 	t = btf_type_skip_modifiers(btf_vmlinux, type, NULL);
-	if (!btf_type_is_struct(t)) {
+	if (percpu) {
+		aux->btf_var.reg_type = PTR_TO_PERCPU_BTF_ID;
+		aux->btf_var.btf_id = type;
+	} else if (!btf_type_is_struct(t)) {
 		const struct btf_type *ret;
 		const char *tname;
 		u32 tsize;