bpf: Introduce helper bpf_find_vma

In some profiler use cases, it is necessary to map an address to the
backing file, e.g., a shared library. bpf_find_vma helper provides a
flexible way to achieve this. bpf_find_vma maps an address of a task to
the vma (vm_area_struct) for this address, and feed the vma to an callback
BPF function. The callback function is necessary here, as we need to
ensure mmap_sem is unlocked.

It is necessary to lock mmap_sem for find_vma. To lock and unlock mmap_sem
safely when irqs are disable, we use the same mechanism as stackmap with
build_id. Specifically, when irqs are disabled, the unlocked is postponed
in an irq_work. Refactor stackmap.c so that the irq_work is shared among
bpf_find_vma and stackmap helpers.

Signed-off-by: Song Liu <[email protected]>
Signed-off-by: Alexei Starovoitov <[email protected]>
Tested-by: Hengqi Chen <[email protected]>
Acked-by: Yonghong Song <[email protected]>
Link: https://lore.kernel.org/bpf/[email protected]

Author: liu-song-6

include/linux/bpf.h

@@ -2157,6 +2157,7 @@ extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto;
 extern const struct bpf_func_proto bpf_sk_setsockopt_proto;
 extern const struct bpf_func_proto bpf_sk_getsockopt_proto;
 extern const struct bpf_func_proto bpf_kallsyms_lookup_name_proto;
+extern const struct bpf_func_proto bpf_find_vma_proto;
 
 const struct bpf_func_proto *tracing_prog_func_proto(
   enum bpf_func_id func_id, const struct bpf_prog *prog);

include/uapi/linux/bpf.h

@@ -4938,6 +4938,25 @@ union bpf_attr {
  *		**-ENOENT** if symbol is not found.
  *
  *		**-EPERM** if caller does not have permission to obtain kernel address.
+ *
+ * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)
+ *	Description
+ *		Find vma of *task* that contains *addr*, call *callback_fn*
+ *		function with *task*, *vma*, and *callback_ctx*.
+ *		The *callback_fn* should be a static function and
+ *		the *callback_ctx* should be a pointer to the stack.
+ *		The *flags* is used to control certain aspects of the helper.
+ *		Currently, the *flags* must be 0.
+ *
+ *		The expected callback signature is
+ *
+ *		long (\*callback_fn)(struct task_struct \*task, struct vm_area_struct \*vma, void \*callback_ctx);
+ *
+ *	Return
+ *		0 on success.
+ *		**-ENOENT** if *task->mm* is NULL, or no vma contains *addr*.
+ *		**-EBUSY** if failed to try lock mmap_lock.
+ *		**-EINVAL** for invalid **flags**.
  */
 #define __BPF_FUNC_MAPPER(FN)		\
 	FN(unspec),			\
@@ -5120,6 +5139,7 @@ union bpf_attr {
 	FN(trace_vprintk),		\
 	FN(skc_to_unix_sock),		\
 	FN(kallsyms_lookup_name),	\
+	FN(find_vma),			\
 	/* */
 
 /* integer value in 'imm' field of BPF_CALL instruction selects which helper

kernel/bpf/btf.c

@@ -6342,7 +6342,10 @@ const struct bpf_func_proto bpf_btf_find_by_name_kind_proto = {
 	.arg4_type	= ARG_ANYTHING,
 };
 
-BTF_ID_LIST_GLOBAL_SINGLE(btf_task_struct_ids, struct, task_struct)
+BTF_ID_LIST_GLOBAL(btf_task_struct_ids)
+BTF_ID(struct, task_struct)
+BTF_ID(struct, file)
+BTF_ID(struct, vm_area_struct)
 
 /* BTF ID set registration API for modules */
 

kernel/bpf/mmap_unlock_work.h

@@ -0,0 +1,65 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/* Copyright (c) 2021 Facebook
+ */
+
+#ifndef __MMAP_UNLOCK_WORK_H__
+#define __MMAP_UNLOCK_WORK_H__
+#include <linux/irq_work.h>
+
+/* irq_work to run mmap_read_unlock() in irq_work */
+struct mmap_unlock_irq_work {
+	struct irq_work irq_work;
+	struct mm_struct *mm;
+};
+
+DECLARE_PER_CPU(struct mmap_unlock_irq_work, mmap_unlock_work);
+
+/*
+ * We cannot do mmap_read_unlock() when the irq is disabled, because of
+ * risk to deadlock with rq_lock. To look up vma when the irqs are
+ * disabled, we need to run mmap_read_unlock() in irq_work. We use a
+ * percpu variable to do the irq_work. If the irq_work is already used
+ * by another lookup, we fall over.
+ */
+static inline bool bpf_mmap_unlock_get_irq_work(struct mmap_unlock_irq_work **work_ptr)
+{
+	struct mmap_unlock_irq_work *work = NULL;
+	bool irq_work_busy = false;
+
+	if (irqs_disabled()) {
+		if (!IS_ENABLED(CONFIG_PREEMPT_RT)) {
+			work = this_cpu_ptr(&mmap_unlock_work);
+			if (irq_work_is_busy(&work->irq_work)) {
+				/* cannot queue more up_read, fallback */
+				irq_work_busy = true;
+			}
+		} else {
+			/*
+			 * PREEMPT_RT does not allow to trylock mmap sem in
+			 * interrupt disabled context. Force the fallback code.
+			 */
+			irq_work_busy = true;
+		}
+	}
+
+	*work_ptr = work;
+	return irq_work_busy;
+}
+
+static inline void bpf_mmap_unlock_mm(struct mmap_unlock_irq_work *work, struct mm_struct *mm)
+{
+	if (!work) {
+		mmap_read_unlock(mm);
+	} else {
+		work->mm = mm;
+
+		/* The lock will be released once we're out of interrupt
+		 * context. Tell lockdep that we've released it now so
+		 * it doesn't complain that we forgot to release it.
+		 */
+		rwsem_release(&mm->mmap_lock.dep_map, _RET_IP_);
+		irq_work_queue(&work->irq_work);
+	}
+}
+
+#endif /* __MMAP_UNLOCK_WORK_H__ */

kernel/bpf/stackmap.c

@@ -7,10 +7,10 @@
 #include <linux/kernel.h>
 #include <linux/stacktrace.h>
 #include <linux/perf_event.h>
-#include <linux/irq_work.h>
 #include <linux/btf_ids.h>
 #include <linux/buildid.h>
 #include "percpu_freelist.h"
+#include "mmap_unlock_work.h"
 
 #define STACK_CREATE_FLAG_MASK					\
 	(BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY |	\
@@ -31,25 +31,6 @@ struct bpf_stack_map {
 	struct stack_map_bucket *buckets[];
 };
 
-/* irq_work to run up_read() for build_id lookup in nmi context */
-struct stack_map_irq_work {
-	struct irq_work irq_work;
-	struct mm_struct *mm;
-};
-
-static void do_up_read(struct irq_work *entry)
-{
-	struct stack_map_irq_work *work;
-
-	if (WARN_ON_ONCE(IS_ENABLED(CONFIG_PREEMPT_RT)))
-		return;
-
-	work = container_of(entry, struct stack_map_irq_work, irq_work);
-	mmap_read_unlock_non_owner(work->mm);
-}
-
-static DEFINE_PER_CPU(struct stack_map_irq_work, up_read_work);
-
 static inline bool stack_map_use_build_id(struct bpf_map *map)
 {
 	return (map->map_flags & BPF_F_STACK_BUILD_ID);
@@ -149,35 +130,13 @@ static void stack_map_get_build_id_offset(struct bpf_stack_build_id *id_offs,
 					  u64 *ips, u32 trace_nr, bool user)
 {
 	int i;
+	struct mmap_unlock_irq_work *work = NULL;
+	bool irq_work_busy = bpf_mmap_unlock_get_irq_work(&work);
 	struct vm_area_struct *vma;
-	bool irq_work_busy = false;
-	struct stack_map_irq_work *work = NULL;
-
-	if (irqs_disabled()) {
-		if (!IS_ENABLED(CONFIG_PREEMPT_RT)) {
-			work = this_cpu_ptr(&up_read_work);
-			if (irq_work_is_busy(&work->irq_work)) {
-				/* cannot queue more up_read, fallback */
-				irq_work_busy = true;
-			}
-		} else {
-			/*
-			 * PREEMPT_RT does not allow to trylock mmap sem in
-			 * interrupt disabled context. Force the fallback code.
-			 */
-			irq_work_busy = true;
-		}
-	}
 
-	/*
-	 * We cannot do up_read() when the irq is disabled, because of
-	 * risk to deadlock with rq_lock. To do build_id lookup when the
-	 * irqs are disabled, we need to run up_read() in irq_work. We use
-	 * a percpu variable to do the irq_work. If the irq_work is
-	 * already used by another lookup, we fall back to report ips.
-	 *
-	 * Same fallback is used for kernel stack (!user) on a stackmap
-	 * with build_id.
+	/* If the irq_work is in use, fall back to report ips. Same
+	 * fallback is used for kernel stack (!user) on a stackmap with
+	 * build_id.
 	 */
 	if (!user || !current || !current->mm || irq_work_busy ||
 	    !mmap_read_trylock(current->mm)) {
@@ -203,19 +162,7 @@ static void stack_map_get_build_id_offset(struct bpf_stack_build_id *id_offs,
 			- vma->vm_start;
 		id_offs[i].status = BPF_STACK_BUILD_ID_VALID;
 	}
-
-	if (!work) {
-		mmap_read_unlock(current->mm);
-	} else {
-		work->mm = current->mm;
-
-		/* The lock will be released once we're out of interrupt
-		 * context. Tell lockdep that we've released it now so
-		 * it doesn't complain that we forgot to release it.
-		 */
-		rwsem_release(&current->mm->mmap_lock.dep_map, _RET_IP_);
-		irq_work_queue(&work->irq_work);
-	}
+	bpf_mmap_unlock_mm(work, current->mm);
 }
 
 static struct perf_callchain_entry *
@@ -719,16 +666,3 @@ const struct bpf_map_ops stack_trace_map_ops = {
 	.map_btf_name = "bpf_stack_map",
 	.map_btf_id = &stack_trace_map_btf_id,
 };
-
-static int __init stack_map_init(void)
-{
-	int cpu;
-	struct stack_map_irq_work *work;
-
-	for_each_possible_cpu(cpu) {
-		work = per_cpu_ptr(&up_read_work, cpu);
-		init_irq_work(&work->irq_work, do_up_read);
-	}
-	return 0;
-}
-subsys_initcall(stack_map_init);

kernel/bpf/task_iter.c

@@ -8,6 +8,7 @@
 #include <linux/fdtable.h>
 #include <linux/filter.h>
 #include <linux/btf_ids.h>
+#include "mmap_unlock_work.h"
 
 struct bpf_iter_seq_task_common {
 	struct pid_namespace *ns;
@@ -524,10 +525,6 @@ static const struct seq_operations task_vma_seq_ops = {
 	.show	= task_vma_seq_show,
 };
 
-BTF_ID_LIST(btf_task_file_ids)
-BTF_ID(struct, file)
-BTF_ID(struct, vm_area_struct)
-
 static const struct bpf_iter_seq_info task_seq_info = {
 	.seq_ops		= &task_seq_ops,
 	.init_seq_private	= init_seq_pidns,
@@ -586,23 +583,88 @@ static struct bpf_iter_reg task_vma_reg_info = {
 	.seq_info		= &task_vma_seq_info,
 };
 
+BPF_CALL_5(bpf_find_vma, struct task_struct *, task, u64, start,
+	   bpf_callback_t, callback_fn, void *, callback_ctx, u64, flags)
+{
+	struct mmap_unlock_irq_work *work = NULL;
+	struct vm_area_struct *vma;
+	bool irq_work_busy = false;
+	struct mm_struct *mm;
+	int ret = -ENOENT;
+
+	if (flags)
+		return -EINVAL;
+
+	if (!task)
+		return -ENOENT;
+
+	mm = task->mm;
+	if (!mm)
+		return -ENOENT;
+
+	irq_work_busy = bpf_mmap_unlock_get_irq_work(&work);
+
+	if (irq_work_busy || !mmap_read_trylock(mm))
+		return -EBUSY;
+
+	vma = find_vma(mm, start);
+
+	if (vma && vma->vm_start <= start && vma->vm_end > start) {
+		callback_fn((u64)(long)task, (u64)(long)vma,
+			    (u64)(long)callback_ctx, 0, 0);
+		ret = 0;
+	}
+	bpf_mmap_unlock_mm(work, mm);
+	return ret;
+}
+
+const struct bpf_func_proto bpf_find_vma_proto = {
+	.func		= bpf_find_vma,
+	.ret_type	= RET_INTEGER,
+	.arg1_type	= ARG_PTR_TO_BTF_ID,
+	.arg1_btf_id	= &btf_task_struct_ids[0],
+	.arg2_type	= ARG_ANYTHING,
+	.arg3_type	= ARG_PTR_TO_FUNC,
+	.arg4_type	= ARG_PTR_TO_STACK_OR_NULL,
+	.arg5_type	= ARG_ANYTHING,
+};
+
+DEFINE_PER_CPU(struct mmap_unlock_irq_work, mmap_unlock_work);
+
+static void do_mmap_read_unlock(struct irq_work *entry)
+{
+	struct mmap_unlock_irq_work *work;
+
+	if (WARN_ON_ONCE(IS_ENABLED(CONFIG_PREEMPT_RT)))
+		return;
+
+	work = container_of(entry, struct mmap_unlock_irq_work, irq_work);
+	mmap_read_unlock_non_owner(work->mm);
+}
+
 static int __init task_iter_init(void)
 {
-	int ret;
+	struct mmap_unlock_irq_work *work;
+	int ret, cpu;
+
+	for_each_possible_cpu(cpu) {
+		work = per_cpu_ptr(&mmap_unlock_work, cpu);
+		init_irq_work(&work->irq_work, do_mmap_read_unlock);
+	}
 
 	task_reg_info.ctx_arg_info[0].btf_id = btf_task_struct_ids[0];
 	ret = bpf_iter_reg_target(&task_reg_info);
 	if (ret)
 		return ret;
 
 	task_file_reg_info.ctx_arg_info[0].btf_id = btf_task_struct_ids[0];
-	task_file_reg_info.ctx_arg_info[1].btf_id = btf_task_file_ids[0];
+	task_file_reg_info.ctx_arg_info[1].btf_id = btf_task_struct_ids[1];
 	ret =  bpf_iter_reg_target(&task_file_reg_info);
 	if (ret)
 		return ret;
 
 	task_vma_reg_info.ctx_arg_info[0].btf_id = btf_task_struct_ids[0];
-	task_vma_reg_info.ctx_arg_info[1].btf_id = btf_task_file_ids[1];
+	task_vma_reg_info.ctx_arg_info[1].btf_id = btf_task_struct_ids[2];
 	return bpf_iter_reg_target(&task_vma_reg_info);
 }
 late_initcall(task_iter_init);

kernel/bpf/verifier.c

@@ -6132,6 +6132,33 @@ static int set_timer_callback_state(struct bpf_verifier_env *env,
 	return 0;
 }
 
+static int set_find_vma_callback_state(struct bpf_verifier_env *env,
+				       struct bpf_func_state *caller,
+				       struct bpf_func_state *callee,
+				       int insn_idx)
+{
+	/* bpf_find_vma(struct task_struct *task, u64 addr,
+	 *               void *callback_fn, void *callback_ctx, u64 flags)
+	 * (callback_fn)(struct task_struct *task,
+	 *               struct vm_area_struct *vma, void *callback_ctx);
+	 */
+	callee->regs[BPF_REG_1] = caller->regs[BPF_REG_1];
+
+	callee->regs[BPF_REG_2].type = PTR_TO_BTF_ID;
+	__mark_reg_known_zero(&callee->regs[BPF_REG_2]);
+	callee->regs[BPF_REG_2].btf =  btf_vmlinux;
+	callee->regs[BPF_REG_2].btf_id = btf_task_struct_ids[2];
+
+	/* pointer to stack or null */
+	callee->regs[BPF_REG_3] = caller->regs[BPF_REG_4];
+
+	/* unused */
+	__mark_reg_not_init(env, &callee->regs[BPF_REG_4]);
+	__mark_reg_not_init(env, &callee->regs[BPF_REG_5]);
+	callee->in_callback_fn = true;
+	return 0;
+}
+
 static int prepare_func_exit(struct bpf_verifier_env *env, int *insn_idx)
 {
 	struct bpf_verifier_state *state = env->cur_state;
@@ -6489,6 +6516,13 @@ static int check_helper_call(struct bpf_verifier_env *env, struct bpf_insn *insn
 			return -EINVAL;
 	}
 
+	if (func_id == BPF_FUNC_find_vma) {
+		err = __check_func_call(env, insn, insn_idx_p, meta.subprogno,
+					set_find_vma_callback_state);
+		if (err < 0)
+			return -EINVAL;
+	}
+
 	if (func_id == BPF_FUNC_snprintf) {
 		err = check_bpf_snprintf_call(env, regs);
 		if (err < 0)