[bpf-next,1/2] bpf: introduce helper bpf_find_vma

Commit Message

Song Liu Oct. 27, 2021, 10 p.m. UTC

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.

Signed-off-by: Song Liu <[e-mail]>
---
 include/linux/bpf.h            |   1 +
 include/uapi/linux/bpf.h       |  20 +++++++
 kernel/bpf/task_iter.c         | 102 ++++++++++++++++++++++++++++++++-
 kernel/bpf/verifier.c          |  36 ++++++++++++
 kernel/trace/bpf_trace.c       |   2 +
 tools/include/uapi/linux/bpf.h |  19 ++++++
 6 files changed, 179 insertions(+), 1 deletion(-)

Comments

Daniel Borkmann Nov. 1, 2021, 10:23 p.m. UTC | #1

On 10/28/21 12:00 AM, Song Liu wrote:
[...]
>   /* integer value in 'imm' field of BPF_CALL instruction selects which helper
> diff --git a/kernel/bpf/task_iter.c b/kernel/bpf/task_iter.c
> index b48750bfba5aa..ad30f2e885356 100644
> --- a/kernel/bpf/task_iter.c
> +++ b/kernel/bpf/task_iter.c
> @@ -8,6 +8,7 @@
>   #include <linux/fdtable.h>
>   #include <linux/filter.h>
>   #include <linux/btf_ids.h>
> +#include <linux/irq_work.h>
>   
>   struct bpf_iter_seq_task_common {
>   	struct pid_namespace *ns;
> @@ -21,6 +22,25 @@ struct bpf_iter_seq_task_info {
>   	u32 tid;
>   };
>   
> +/* irq_work to run mmap_read_unlock() */
> +struct task_iter_irq_work {
> +	struct irq_work irq_work;
> +	struct mm_struct *mm;
> +};
> +
> +static DEFINE_PER_CPU(struct task_iter_irq_work, mmap_unlock_work);
> +
> +static void do_mmap_read_unlock(struct irq_work *entry)
> +{
> +	struct task_iter_irq_work *work;
> +
> +	if (WARN_ON_ONCE(IS_ENABLED(CONFIG_PREEMPT_RT)))
> +		return;
> +
> +	work = container_of(entry, struct task_iter_irq_work, irq_work);
> +	mmap_read_unlock_non_owner(work->mm);
> +}
> +
>   static struct task_struct *task_seq_get_next(struct pid_namespace *ns,
>   					     u32 *tid,
>   					     bool skip_if_dup_files)
> @@ -586,9 +606,89 @@ 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 task_iter_irq_work *work = NULL;
> +	struct mm_struct *mm = task->mm;

Won't this NULL deref if called with task argument as NULL?

> +	struct vm_area_struct *vma;
> +	bool irq_work_busy = false;
> +	int ret = -ENOENT;
> +
> +	if (flags)
> +		return -EINVAL;
> +
> +	if (!mm)
> +		return -ENOENT;
> +
> +	/*
> +	 * Similar to stackmap with build_id support, we cannot simply do
> +	 * mmap_read_unlock when the irq is disabled. Instead, we need do
> +	 * the unlock in the irq_work.
> +	 */
> +	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 mmap_unlock, abort. */
> +				irq_work_busy = true;
> +			}
> +		} else {
> +			/*
> +			 * PREEMPT_RT does not allow to trylock mmap sem in
> +			 * interrupt disabled context, abort.
> +			 */
> +			irq_work_busy = true;
> +		}
> +	}
> +
> +	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;
> +	}
> +	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);
> +	}

Given this is pretty much the same logic around the vma retrieval, could this be
refactored/consolidated with stack map build id retrieval into a common function?

> +	return ret;
> +}
> +
> +BTF_ID_LIST_SINGLE(btf_find_vma_ids, struct, task_struct)
> +
> +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_find_vma_ids[0],
> +	.arg2_type	= ARG_ANYTHING,
> +	.arg3_type	= ARG_PTR_TO_FUNC,
> +	.arg4_type	= ARG_PTR_TO_STACK_OR_NULL,
> +	.arg5_type	= ARG_ANYTHING,
> +};
[...]
> diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h
> index c108200378834..056c00da1b5d6 100644
> --- a/tools/include/uapi/linux/bpf.h
> +++ b/tools/include/uapi/linux/bpf.h
> @@ -4915,6 +4915,24 @@ union bpf_attr {
>    *		Dynamically cast a *sk* pointer to a *unix_sock* pointer.
>    *	Return
>    *		*sk* if casting is valid, or **NULL** otherwise.
> + * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)

nit: Wrongly copied uapi header over to tooling?

> + *	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 \*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),			\
> @@ -5096,6 +5114,7 @@ union bpf_attr {
>   	FN(get_branch_snapshot),	\
>   	FN(trace_vprintk),		\
>   	FN(skc_to_unix_sock),		\
> +	FN(find_vma),			\
>   	/* */
>   
>   /* integer value in 'imm' field of BPF_CALL instruction selects which helper
>

Song Liu Nov. 3, 2021, 11:54 p.m. UTC | #2

> On Nov 1, 2021, at 3:23 PM, Daniel Borkmann <[e-mail]> wrote:
> 
> On 10/28/21 12:00 AM, Song Liu wrote:
> [...]
>>  /* integer value in 'imm' field of BPF_CALL instruction selects which helper
>> diff --git a/kernel/bpf/task_iter.c b/kernel/bpf/task_iter.c
>> index b48750bfba5aa..ad30f2e885356 100644
>> --- a/kernel/bpf/task_iter.c
>> +++ b/kernel/bpf/task_iter.c
>> @@ -8,6 +8,7 @@
>>  #include <linux/fdtable.h>
>>  #include <linux/filter.h>
>>  #include <linux/btf_ids.h>
>> +#include <linux/irq_work.h>
>>    struct bpf_iter_seq_task_common {
>>  	struct pid_namespace *ns;
>> @@ -21,6 +22,25 @@ struct bpf_iter_seq_task_info {
>>  	u32 tid;
>>  };
>>  +/* irq_work to run mmap_read_unlock() */
>> +struct task_iter_irq_work {
>> +	struct irq_work irq_work;
>> +	struct mm_struct *mm;
>> +};
>> +
>> +static DEFINE_PER_CPU(struct task_iter_irq_work, mmap_unlock_work);
>> +
>> +static void do_mmap_read_unlock(struct irq_work *entry)
>> +{
>> +	struct task_iter_irq_work *work;
>> +
>> +	if (WARN_ON_ONCE(IS_ENABLED(CONFIG_PREEMPT_RT)))
>> +		return;
>> +
>> +	work = container_of(entry, struct task_iter_irq_work, irq_work);
>> +	mmap_read_unlock_non_owner(work->mm);
>> +}
>> +
>>  static struct task_struct *task_seq_get_next(struct pid_namespace *ns,
>>  					     u32 *tid,
>>  					     bool skip_if_dup_files)
>> @@ -586,9 +606,89 @@ 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 task_iter_irq_work *work = NULL;
>> +	struct mm_struct *mm = task->mm;
> 
> Won't this NULL deref if called with task argument as NULL?

Will fix. 

> 
>> +	struct vm_area_struct *vma;
>> +	bool irq_work_busy = false;
>> +	int ret = -ENOENT;
>> +
>> +	if (flags)
>> +		return -EINVAL;
>> +
>> +	if (!mm)
>> +		return -ENOENT;
>> +
>> +	/*
>> +	 * Similar to stackmap with build_id support, we cannot simply do
>> +	 * mmap_read_unlock when the irq is disabled. Instead, we need do
>> +	 * the unlock in the irq_work.
>> +	 */
>> +	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 mmap_unlock, abort. */
>> +				irq_work_busy = true;
>> +			}
>> +		} else {
>> +			/*
>> +			 * PREEMPT_RT does not allow to trylock mmap sem in
>> +			 * interrupt disabled context, abort.
>> +			 */
>> +			irq_work_busy = true;
>> +		}
>> +	}
>> +
>> +	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;
>> +	}
>> +	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);
>> +	}
> 
> Given this is pretty much the same logic around the vma retrieval, could this be
> refactored/consolidated with stack map build id retrieval into a common function?

I thought about sharing the irq_work code amount the two. The problem was we need
to include irq_work.h in bpf.h. But on a second thought, maybe we should just 
move bpf_find_vma to stackmap.c? This will avoid including irq_work.h. I guess it 
is not too weird to have bpf_find_vma in stackmap.c.  

> 
>> +	return ret;
>> +}
>> +
>> +BTF_ID_LIST_SINGLE(btf_find_vma_ids, struct, task_struct)
>> +
>> +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_find_vma_ids[0],
>> +	.arg2_type	= ARG_ANYTHING,
>> +	.arg3_type	= ARG_PTR_TO_FUNC,
>> +	.arg4_type	= ARG_PTR_TO_STACK_OR_NULL,
>> +	.arg5_type	= ARG_ANYTHING,
>> +};
> [...]
>> diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h
>> index c108200378834..056c00da1b5d6 100644
>> --- a/tools/include/uapi/linux/bpf.h
>> +++ b/tools/include/uapi/linux/bpf.h
>> @@ -4915,6 +4915,24 @@ union bpf_attr {
>>   *		Dynamically cast a *sk* pointer to a *unix_sock* pointer.
>>   *	Return
>>   *		*sk* if casting is valid, or **NULL** otherwise.
>> + * long bpf_find_vma(struct task_struct *task, u64 addr, void *callback_fn, void *callback_ctx, u64 flags)
> 
> nit: Wrongly copied uapi header over to tooling?

Right... You get really good eyes. :-)

[...]

Song Liu Nov. 4, 2021, 5:56 a.m. UTC | #3

> On Nov 3, 2021, at 4:54 PM, Song Liu <[e-mail]> wrote:
> 
> 
> 
>> On Nov 1, 2021, at 3:23 PM, Daniel Borkmann <[e-mail]> wrote:
>> 
>> On 10/28/21 12:00 AM, Song Liu wrote:
>> [...]
>>> /* integer value in 'imm' field of BPF_CALL instruction selects which helper
>>> diff --git a/kernel/bpf/task_iter.c b/kernel/bpf/task_iter.c
>>> index b48750bfba5aa..ad30f2e885356 100644
>>> --- a/kernel/bpf/task_iter.c
>>> +++ b/kernel/bpf/task_iter.c
>>> @@ -8,6 +8,7 @@
>>> #include <linux/fdtable.h>
>>> #include <linux/filter.h>
>>> #include <linux/btf_ids.h>
>>> +#include <linux/irq_work.h>
>>>   struct bpf_iter_seq_task_common {
>>> 	struct pid_namespace *ns;
>>> @@ -21,6 +22,25 @@ struct bpf_iter_seq_task_info {
>>> 	u32 tid;
>>> };
>>> +/* irq_work to run mmap_read_unlock() */
>>> +struct task_iter_irq_work {
>>> +	struct irq_work irq_work;
>>> +	struct mm_struct *mm;
>>> +};
>>> +
>>> +static DEFINE_PER_CPU(struct task_iter_irq_work, mmap_unlock_work);
>>> +
>>> +static void do_mmap_read_unlock(struct irq_work *entry)
>>> +{
>>> +	struct task_iter_irq_work *work;
>>> +
>>> +	if (WARN_ON_ONCE(IS_ENABLED(CONFIG_PREEMPT_RT)))
>>> +		return;
>>> +
>>> +	work = container_of(entry, struct task_iter_irq_work, irq_work);
>>> +	mmap_read_unlock_non_owner(work->mm);
>>> +}
>>> +
>>> static struct task_struct *task_seq_get_next(struct pid_namespace *ns,
>>> 					     u32 *tid,
>>> 					     bool skip_if_dup_files)
>>> @@ -586,9 +606,89 @@ 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 task_iter_irq_work *work = NULL;
>>> +	struct mm_struct *mm = task->mm;
>> 
>> Won't this NULL deref if called with task argument as NULL?
> 
> Will fix. 
> 
>> 
>>> +	struct vm_area_struct *vma;
>>> +	bool irq_work_busy = false;
>>> +	int ret = -ENOENT;
>>> +
>>> +	if (flags)
>>> +		return -EINVAL;
>>> +
>>> +	if (!mm)
>>> +		return -ENOENT;
>>> +
>>> +	/*
>>> +	 * Similar to stackmap with build_id support, we cannot simply do
>>> +	 * mmap_read_unlock when the irq is disabled. Instead, we need do
>>> +	 * the unlock in the irq_work.
>>> +	 */
>>> +	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 mmap_unlock, abort. */
>>> +				irq_work_busy = true;
>>> +			}
>>> +		} else {
>>> +			/*
>>> +			 * PREEMPT_RT does not allow to trylock mmap sem in
>>> +			 * interrupt disabled context, abort.
>>> +			 */
>>> +			irq_work_busy = true;
>>> +		}
>>> +	}
>>> +
>>> +	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;
>>> +	}
>>> +	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);
>>> +	}
>> 
>> Given this is pretty much the same logic around the vma retrieval, could this be
>> refactored/consolidated with stack map build id retrieval into a common function?
> 
> I thought about sharing the irq_work code amount the two. The problem was we need
> to include irq_work.h in bpf.h. But on a second thought, maybe we should just 
> move bpf_find_vma to stackmap.c? This will avoid including irq_work.h. I guess it 
> is not too weird to have bpf_find_vma in stackmap.c.  

Actually, we can just add a local header for it in kernel/bpf. Adding bpf_find_vma
to stackmap.c means bpf_find_vma requires CONFIG_PERF_EVENTS. It is not a real 
issue in most systems, but may break some build tests. 

Song

Patch

diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index 31421c74ba081..65def8467b609 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -2109,6 +2109,7 @@  extern const struct bpf_func_proto bpf_for_each_map_elem_proto;
 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_find_vma_proto;
 
 const struct bpf_func_proto *tracing_prog_func_proto(
   enum bpf_func_id func_id, const struct bpf_prog *prog);
diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index c108200378834..e1401ae4decc9 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -4915,6 +4915,25 @@  union bpf_attr {
  *		Dynamically cast a *sk* pointer to a *unix_sock* pointer.
  *	Return
  *		*sk* if casting is valid, or **NULL** otherwise.
+ *
+ * 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 \*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),			\
@@ -5096,6 +5115,7 @@  union bpf_attr {
 	FN(get_branch_snapshot),	\
 	FN(trace_vprintk),		\
 	FN(skc_to_unix_sock),		\
+	FN(find_vma),			\
 	/* */
 
 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
diff --git a/kernel/bpf/task_iter.c b/kernel/bpf/task_iter.c
index b48750bfba5aa..ad30f2e885356 100644
--- a/kernel/bpf/task_iter.c
+++ b/kernel/bpf/task_iter.c
@@ -8,6 +8,7 @@ 
 #include <linux/fdtable.h>
 #include <linux/filter.h>
 #include <linux/btf_ids.h>
+#include <linux/irq_work.h>
 
 struct bpf_iter_seq_task_common {
 	struct pid_namespace *ns;
@@ -21,6 +22,25 @@  struct bpf_iter_seq_task_info {
 	u32 tid;
 };
 
+/* irq_work to run mmap_read_unlock() */
+struct task_iter_irq_work {
+	struct irq_work irq_work;
+	struct mm_struct *mm;
+};
+
+static DEFINE_PER_CPU(struct task_iter_irq_work, mmap_unlock_work);
+
+static void do_mmap_read_unlock(struct irq_work *entry)
+{
+	struct task_iter_irq_work *work;
+
+	if (WARN_ON_ONCE(IS_ENABLED(CONFIG_PREEMPT_RT)))
+		return;
+
+	work = container_of(entry, struct task_iter_irq_work, irq_work);
+	mmap_read_unlock_non_owner(work->mm);
+}
+
 static struct task_struct *task_seq_get_next(struct pid_namespace *ns,
 					     u32 *tid,
 					     bool skip_if_dup_files)
@@ -586,9 +606,89 @@  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 task_iter_irq_work *work = NULL;
+	struct mm_struct *mm = task->mm;
+	struct vm_area_struct *vma;
+	bool irq_work_busy = false;
+	int ret = -ENOENT;
+
+	if (flags)
+		return -EINVAL;
+
+	if (!mm)
+		return -ENOENT;
+
+	/*
+	 * Similar to stackmap with build_id support, we cannot simply do
+	 * mmap_read_unlock when the irq is disabled. Instead, we need do
+	 * the unlock in the irq_work.
+	 */
+	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 mmap_unlock, abort. */
+				irq_work_busy = true;
+			}
+		} else {
+			/*
+			 * PREEMPT_RT does not allow to trylock mmap sem in
+			 * interrupt disabled context, abort.
+			 */
+			irq_work_busy = true;
+		}
+	}
+
+	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;
+	}
+	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);
+	}
+	return ret;
+}
+
+BTF_ID_LIST_SINGLE(btf_find_vma_ids, struct, task_struct)
+
+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_find_vma_ids[0],
+	.arg2_type	= ARG_ANYTHING,
+	.arg3_type	= ARG_PTR_TO_FUNC,
+	.arg4_type	= ARG_PTR_TO_STACK_OR_NULL,
+	.arg5_type	= ARG_ANYTHING,
+};
+
 static int __init task_iter_init(void)
 {
-	int ret;
+	struct task_iter_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);
diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index c6616e3258038..393ab21529b08 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -6119,6 +6119,35 @@  static int set_timer_callback_state(struct bpf_verifier_env *env,
 	return 0;
 }
 
+BTF_ID_LIST_SINGLE(btf_set_find_vma_ids, struct, vm_area_struct)
+
+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 start,
+	 *               void *callback_fn, void *callback_ctx, u64 flags)
+	 * (callback_fn)(struct task_struct *task,
+	 *               struct vm_area_struct *vma, void *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_set_find_vma_ids[0];
+
+	/* 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;
@@ -6476,6 +6505,13 @@  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)
diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index cbcd0d6fca7c7..c95397c55d0e8 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -1208,6 +1208,8 @@  bpf_tracing_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 		return &bpf_get_func_ip_proto_tracing;
 	case BPF_FUNC_get_branch_snapshot:
 		return &bpf_get_branch_snapshot_proto;
+	case BPF_FUNC_find_vma:
+		return &bpf_find_vma_proto;
 	case BPF_FUNC_trace_vprintk:
 		return bpf_get_trace_vprintk_proto();
 	default:
diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h
index c108200378834..056c00da1b5d6 100644
--- a/tools/include/uapi/linux/bpf.h
+++ b/tools/include/uapi/linux/bpf.h
@@ -4915,6 +4915,24 @@  union bpf_attr {
  *		Dynamically cast a *sk* pointer to a *unix_sock* pointer.
  *	Return
  *		*sk* if casting is valid, or **NULL** otherwise.
+ * 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 \*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),			\
@@ -5096,6 +5114,7 @@  union bpf_attr {
 	FN(get_branch_snapshot),	\
 	FN(trace_vprintk),		\
 	FN(skc_to_unix_sock),		\
+	FN(find_vma),			\
 	/* */
 
 /* integer value in 'imm' field of BPF_CALL instruction selects which helper