[PATCH bpf-next 1/2] bpf: Add lookup_and_delete_elem for BPF_MAP_STACK_TRACE

[PATCH bpf-next 1/2] bpf: Add lookup_and_delete_elem for BPF_MAP_STACK_TRACE

[Tao Chen]

The stacktrace map can be easily full, which will lead to failure in
obtaining the stack. In addition to increasing the size of the map,
another solution is to delete the stack_id after looking it up from
the user, so extend the existing bpf_map_lookup_and_delete_elem()
functionality to stacktrace map types.

Signed-off-by: Tao Chen <chen.dylane@linux.dev>
---
 include/linux/bpf.h   |  2 +-
 kernel/bpf/stackmap.c | 16 ++++++++++++++--
 kernel/bpf/syscall.c  |  8 +++++---
 3 files changed, 20 insertions(+), 6 deletions(-)

diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index 8f6e87f0f3a..84a41c42495 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -2703,7 +2703,7 @@ int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
 			    u64 flags);
 
-int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
+int bpf_stackmap_copy_and_delete(struct bpf_map *map, void *key, void *value, bool delete);
 
 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
 				 void *key, void *value, u64 map_flags);
diff --git a/kernel/bpf/stackmap.c b/kernel/bpf/stackmap.c
index 3615c06b7df..bb63a74db7b 100644
--- a/kernel/bpf/stackmap.c
+++ b/kernel/bpf/stackmap.c
@@ -646,7 +646,15 @@ static void *stack_map_lookup_elem(struct bpf_map *map, void *key)
 }
 
 /* Called from syscall */
-int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value)
+static int stack_map_lookup_and_delete_elem(struct bpf_map *map, void *key,
+					    void *value, u64 flags)
+{
+	return bpf_stackmap_copy_and_delete(map, key, value, true);
+}
+
+/* Called from syscall */
+int bpf_stackmap_copy_and_delete(struct bpf_map *map, void *key, void *value,
+				 bool delete)
 {
 	struct bpf_stack_map *smap = container_of(map, struct bpf_stack_map, map);
 	struct stack_map_bucket *bucket, *old_bucket;
@@ -663,7 +671,10 @@ int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value)
 	memcpy(value, bucket->data, trace_len);
 	memset(value + trace_len, 0, map->value_size - trace_len);
 
-	old_bucket = xchg(&smap->buckets[id], bucket);
+	if (delete)
+		old_bucket = bucket;
+	else
+		old_bucket = xchg(&smap->buckets[id], bucket);
 	if (old_bucket)
 		pcpu_freelist_push(&smap->freelist, &old_bucket->fnode);
 	return 0;
@@ -754,6 +765,7 @@ const struct bpf_map_ops stack_trace_map_ops = {
 	.map_free = stack_map_free,
 	.map_get_next_key = stack_map_get_next_key,
 	.map_lookup_elem = stack_map_lookup_elem,
+	.map_lookup_and_delete_elem = stack_map_lookup_and_delete_elem,
 	.map_update_elem = stack_map_update_elem,
 	.map_delete_elem = stack_map_delete_elem,
 	.map_check_btf = map_check_no_btf,
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 0fbfa8532c3..d606d156c62 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -318,7 +318,7 @@ static int bpf_map_copy_value(struct bpf_map *map, void *key, void *value,
 	} else if (map->map_type == BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE) {
 		err = bpf_percpu_cgroup_storage_copy(map, key, value);
 	} else if (map->map_type == BPF_MAP_TYPE_STACK_TRACE) {
-		err = bpf_stackmap_copy(map, key, value);
+		err = bpf_stackmap_copy_and_delete(map, key, value, false);
 	} else if (IS_FD_ARRAY(map) || IS_FD_PROG_ARRAY(map)) {
 		err = bpf_fd_array_map_lookup_elem(map, key, value);
 	} else if (IS_FD_HASH(map)) {
@@ -1627,7 +1627,8 @@ struct bpf_map *bpf_map_inc_not_zero(struct bpf_map *map)
 }
 EXPORT_SYMBOL_GPL(bpf_map_inc_not_zero);
 
-int __weak bpf_stackmap_copy(struct bpf_map *map, void *key, void *value)
+int __weak bpf_stackmap_copy_and_delete(struct bpf_map *map, void *key, void *value,
+					bool delete)
 {
 	return -ENOTSUPP;
 }
@@ -2158,7 +2159,8 @@ static int map_lookup_and_delete_elem(union bpf_attr *attr)
 	} else if (map->map_type == BPF_MAP_TYPE_HASH ||
 		   map->map_type == BPF_MAP_TYPE_PERCPU_HASH ||
 		   map->map_type == BPF_MAP_TYPE_LRU_HASH ||
-		   map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) {
+		   map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH ||
+		   map->map_type == BPF_MAP_TYPE_STACK_TRACE) {
 		if (!bpf_map_is_offloaded(map)) {
 			bpf_disable_instrumentation();
 			rcu_read_lock();
-- 
2.48.1

[PATCH bpf-next 2/2] selftests/bpf: Add stacktrace map lookup_and_delete_elem test case

[Tao Chen]

...
test_stacktrace_map:PASS:compare_stack_ips stackmap vs. stack_amap 0 nsec
test_stacktrace_map:PASS:stack_key_map lookup 0 nsec
test_stacktrace_map:PASS:stackmap lookup and detele 0 nsec
 #397     stacktrace_map:OK
...

Signed-off-by: Tao Chen <chen.dylane@linux.dev>
---
 .../selftests/bpf/prog_tests/stacktrace_map.c  | 18 +++++++++++++++++-
 .../selftests/bpf/progs/test_stacktrace_map.c  | 12 +++++++++++-
 2 files changed, 28 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/stacktrace_map.c b/tools/testing/selftests/bpf/prog_tests/stacktrace_map.c
index 84a7e405e91..496c4dcf4ea 100644
--- a/tools/testing/selftests/bpf/prog_tests/stacktrace_map.c
+++ b/tools/testing/selftests/bpf/prog_tests/stacktrace_map.c
@@ -3,7 +3,7 @@
 
 void test_stacktrace_map(void)
 {
-	int control_map_fd, stackid_hmap_fd, stackmap_fd, stack_amap_fd;
+	int control_map_fd, stackid_hmap_fd, stackmap_fd, stack_amap_fd, stack_key_map_fd;
 	const char *prog_name = "oncpu";
 	int err, prog_fd, stack_trace_len;
 	const char *file = "./test_stacktrace_map.bpf.o";
@@ -11,6 +11,9 @@ void test_stacktrace_map(void)
 	struct bpf_program *prog;
 	struct bpf_object *obj;
 	struct bpf_link *link;
+	__u32 stackmap_key;
+	char val_buf[PERF_MAX_STACK_DEPTH *
+		sizeof(struct bpf_stack_build_id)];
 
 	err = bpf_prog_test_load(file, BPF_PROG_TYPE_TRACEPOINT, &obj, &prog_fd);
 	if (CHECK(err, "prog_load", "err %d errno %d\n", err, errno))
@@ -41,6 +44,10 @@ void test_stacktrace_map(void)
 	if (CHECK_FAIL(stack_amap_fd < 0))
 		goto disable_pmu;
 
+	stack_key_map_fd = bpf_find_map(__func__, obj, "stack_key_map");
+	if (CHECK_FAIL(stack_key_map_fd < 0))
+		goto disable_pmu;
+
 	/* give some time for bpf program run */
 	sleep(1);
 
@@ -68,6 +75,15 @@ void test_stacktrace_map(void)
 		  "err %d errno %d\n", err, errno))
 		goto disable_pmu;
 
+	err = bpf_map_lookup_elem(stack_key_map_fd, &key, &stackmap_key);
+	if (CHECK(err, "stack_key_map lookup", "err %d errno %d\n", err, errno))
+		goto disable_pmu;
+
+	err = bpf_map_lookup_and_delete_elem(stackmap_fd, &stackmap_key, &val_buf);
+	if (CHECK(err, "stackmap lookup and detele",
+		  "err %d errno %d\n", err, errno))
+		goto disable_pmu;
+
 disable_pmu:
 	bpf_link__destroy(link);
 close_prog:
diff --git a/tools/testing/selftests/bpf/progs/test_stacktrace_map.c b/tools/testing/selftests/bpf/progs/test_stacktrace_map.c
index 47568007b66..d036e8e9c83 100644
--- a/tools/testing/selftests/bpf/progs/test_stacktrace_map.c
+++ b/tools/testing/selftests/bpf/progs/test_stacktrace_map.c
@@ -38,6 +38,13 @@ struct {
 	__type(value, stack_trace_t);
 } stack_amap SEC(".maps");
 
+struct {
+	__uint(type, BPF_MAP_TYPE_ARRAY);
+	__uint(max_entries, 1);
+	__type(key, __u32);
+	__type(value, __u32);
+} stack_key_map SEC(".maps");
+
 /* taken from /sys/kernel/tracing/events/sched/sched_switch/format */
 struct sched_switch_args {
 	unsigned long long pad;
@@ -54,7 +61,7 @@ SEC("tracepoint/sched/sched_switch")
 int oncpu(struct sched_switch_args *ctx)
 {
 	__u32 max_len = PERF_MAX_STACK_DEPTH * sizeof(__u64);
-	__u32 key = 0, val = 0, *value_p;
+	__u32 key = 0, val = 0, *value_p, stackmap_key = 0;
 	void *stack_p;
 
 	value_p = bpf_map_lookup_elem(&control_map, &key);
@@ -64,6 +71,9 @@ int oncpu(struct sched_switch_args *ctx)
 	/* The size of stackmap and stackid_hmap should be the same */
 	key = bpf_get_stackid(ctx, &stackmap, 0);
 	if ((int)key >= 0) {
+		val = key;
+		bpf_map_update_elem(&stack_key_map, &stackmap_key, &val, 0);
+		val = 0;
 		bpf_map_update_elem(&stackid_hmap, &key, &val, 0);
 		stack_p = bpf_map_lookup_elem(&stack_amap, &key);
 		if (stack_p)
-- 
2.48.1

Re: [PATCH bpf-next 2/2] selftests/bpf: Add stacktrace map lookup_and_delete_elem test case

[Jiri Olsa]

On Mon, Sep 08, 2025 at 07:36:22PM +0800, Tao Chen wrote:
> ...
> test_stacktrace_map:PASS:compare_stack_ips stackmap vs. stack_amap 0 nsec
> test_stacktrace_map:PASS:stack_key_map lookup 0 nsec
> test_stacktrace_map:PASS:stackmap lookup and detele 0 nsec
>  #397     stacktrace_map:OK
> ...
> 
> Signed-off-by: Tao Chen <chen.dylane@linux.dev>
> ---
>  .../selftests/bpf/prog_tests/stacktrace_map.c  | 18 +++++++++++++++++-
>  .../selftests/bpf/progs/test_stacktrace_map.c  | 12 +++++++++++-
>  2 files changed, 28 insertions(+), 2 deletions(-)
> 
> diff --git a/tools/testing/selftests/bpf/prog_tests/stacktrace_map.c b/tools/testing/selftests/bpf/prog_tests/stacktrace_map.c
> index 84a7e405e91..496c4dcf4ea 100644
> --- a/tools/testing/selftests/bpf/prog_tests/stacktrace_map.c
> +++ b/tools/testing/selftests/bpf/prog_tests/stacktrace_map.c
> @@ -3,7 +3,7 @@
>  
>  void test_stacktrace_map(void)
>  {
> -	int control_map_fd, stackid_hmap_fd, stackmap_fd, stack_amap_fd;
> +	int control_map_fd, stackid_hmap_fd, stackmap_fd, stack_amap_fd, stack_key_map_fd;
>  	const char *prog_name = "oncpu";
>  	int err, prog_fd, stack_trace_len;
>  	const char *file = "./test_stacktrace_map.bpf.o";
> @@ -11,6 +11,9 @@ void test_stacktrace_map(void)
>  	struct bpf_program *prog;
>  	struct bpf_object *obj;
>  	struct bpf_link *link;
> +	__u32 stackmap_key;
> +	char val_buf[PERF_MAX_STACK_DEPTH *
> +		sizeof(struct bpf_stack_build_id)];
>  
>  	err = bpf_prog_test_load(file, BPF_PROG_TYPE_TRACEPOINT, &obj, &prog_fd);
>  	if (CHECK(err, "prog_load", "err %d errno %d\n", err, errno))
> @@ -41,6 +44,10 @@ void test_stacktrace_map(void)
>  	if (CHECK_FAIL(stack_amap_fd < 0))
>  		goto disable_pmu;
>  
> +	stack_key_map_fd = bpf_find_map(__func__, obj, "stack_key_map");
> +	if (CHECK_FAIL(stack_key_map_fd < 0))
> +		goto disable_pmu;
> +
>  	/* give some time for bpf program run */
>  	sleep(1);
>  
> @@ -68,6 +75,15 @@ void test_stacktrace_map(void)
>  		  "err %d errno %d\n", err, errno))
>  		goto disable_pmu;
>  
> +	err = bpf_map_lookup_elem(stack_key_map_fd, &key, &stackmap_key);
> +	if (CHECK(err, "stack_key_map lookup", "err %d errno %d\n", err, errno))
> +		goto disable_pmu;
> +
> +	err = bpf_map_lookup_and_delete_elem(stackmap_fd, &stackmap_key, &val_buf);
> +	if (CHECK(err, "stackmap lookup and detele",

nit typo 's/detele/delete/'

> +		  "err %d errno %d\n", err, errno))
> +		goto disable_pmu;

should we also check the record got deleted? like make sure following
lookup fails with NOENT:

  bpf_map_lookup_elem(stackmap_fd, &stackmap_key, &val_buf)

> +
>  disable_pmu:
>  	bpf_link__destroy(link);
>  close_prog:
> diff --git a/tools/testing/selftests/bpf/progs/test_stacktrace_map.c b/tools/testing/selftests/bpf/progs/test_stacktrace_map.c
> index 47568007b66..d036e8e9c83 100644
> --- a/tools/testing/selftests/bpf/progs/test_stacktrace_map.c
> +++ b/tools/testing/selftests/bpf/progs/test_stacktrace_map.c
> @@ -38,6 +38,13 @@ struct {
>  	__type(value, stack_trace_t);
>  } stack_amap SEC(".maps");
>  
> +struct {
> +	__uint(type, BPF_MAP_TYPE_ARRAY);
> +	__uint(max_entries, 1);
> +	__type(key, __u32);
> +	__type(value, __u32);
> +} stack_key_map SEC(".maps");
> +
>  /* taken from /sys/kernel/tracing/events/sched/sched_switch/format */
>  struct sched_switch_args {
>  	unsigned long long pad;
> @@ -54,7 +61,7 @@ SEC("tracepoint/sched/sched_switch")
>  int oncpu(struct sched_switch_args *ctx)
>  {
>  	__u32 max_len = PERF_MAX_STACK_DEPTH * sizeof(__u64);
> -	__u32 key = 0, val = 0, *value_p;
> +	__u32 key = 0, val = 0, *value_p, stackmap_key = 0;
>  	void *stack_p;
>  
>  	value_p = bpf_map_lookup_elem(&control_map, &key);
> @@ -64,6 +71,9 @@ int oncpu(struct sched_switch_args *ctx)
>  	/* The size of stackmap and stackid_hmap should be the same */
>  	key = bpf_get_stackid(ctx, &stackmap, 0);
>  	if ((int)key >= 0) {
> +		val = key;
> +		bpf_map_update_elem(&stack_key_map, &stackmap_key, &val, 0);

why not use '&key' directly as the update value?

jirka


> +		val = 0;
>  		bpf_map_update_elem(&stackid_hmap, &key, &val, 0);
>  		stack_p = bpf_map_lookup_elem(&stack_amap, &key);
>  		if (stack_p)
> -- 
> 2.48.1
> 
> 

Re: [PATCH bpf-next 2/2] selftests/bpf: Add stacktrace map lookup_and_delete_elem test case

[Tao Chen]

在 2025/9/9 20:35, Jiri Olsa 写道:
> On Mon, Sep 08, 2025 at 07:36:22PM +0800, Tao Chen wrote:
>> ...
>> test_stacktrace_map:PASS:compare_stack_ips stackmap vs. stack_amap 0 nsec
>> test_stacktrace_map:PASS:stack_key_map lookup 0 nsec
>> test_stacktrace_map:PASS:stackmap lookup and detele 0 nsec
>>   #397     stacktrace_map:OK
>> ...
>>
>> Signed-off-by: Tao Chen <chen.dylane@linux.dev>
>> ---
>>   .../selftests/bpf/prog_tests/stacktrace_map.c  | 18 +++++++++++++++++-
>>   .../selftests/bpf/progs/test_stacktrace_map.c  | 12 +++++++++++-
>>   2 files changed, 28 insertions(+), 2 deletions(-)
>>
>> diff --git a/tools/testing/selftests/bpf/prog_tests/stacktrace_map.c b/tools/testing/selftests/bpf/prog_tests/stacktrace_map.c
>> index 84a7e405e91..496c4dcf4ea 100644
>> --- a/tools/testing/selftests/bpf/prog_tests/stacktrace_map.c
>> +++ b/tools/testing/selftests/bpf/prog_tests/stacktrace_map.c
>> @@ -3,7 +3,7 @@
>>   
>>   void test_stacktrace_map(void)
>>   {
>> -	int control_map_fd, stackid_hmap_fd, stackmap_fd, stack_amap_fd;
>> +	int control_map_fd, stackid_hmap_fd, stackmap_fd, stack_amap_fd, stack_key_map_fd;
>>   	const char *prog_name = "oncpu";
>>   	int err, prog_fd, stack_trace_len;
>>   	const char *file = "./test_stacktrace_map.bpf.o";
>> @@ -11,6 +11,9 @@ void test_stacktrace_map(void)
>>   	struct bpf_program *prog;
>>   	struct bpf_object *obj;
>>   	struct bpf_link *link;
>> +	__u32 stackmap_key;
>> +	char val_buf[PERF_MAX_STACK_DEPTH *
>> +		sizeof(struct bpf_stack_build_id)];
>>   
>>   	err = bpf_prog_test_load(file, BPF_PROG_TYPE_TRACEPOINT, &obj, &prog_fd);
>>   	if (CHECK(err, "prog_load", "err %d errno %d\n", err, errno))
>> @@ -41,6 +44,10 @@ void test_stacktrace_map(void)
>>   	if (CHECK_FAIL(stack_amap_fd < 0))
>>   		goto disable_pmu;
>>   
>> +	stack_key_map_fd = bpf_find_map(__func__, obj, "stack_key_map");
>> +	if (CHECK_FAIL(stack_key_map_fd < 0))
>> +		goto disable_pmu;
>> +
>>   	/* give some time for bpf program run */
>>   	sleep(1);
>>   
>> @@ -68,6 +75,15 @@ void test_stacktrace_map(void)
>>   		  "err %d errno %d\n", err, errno))
>>   		goto disable_pmu;
>>   
>> +	err = bpf_map_lookup_elem(stack_key_map_fd, &key, &stackmap_key);
>> +	if (CHECK(err, "stack_key_map lookup", "err %d errno %d\n", err, errno))
>> +		goto disable_pmu;
>> +
>> +	err = bpf_map_lookup_and_delete_elem(stackmap_fd, &stackmap_key, &val_buf);
>> +	if (CHECK(err, "stackmap lookup and detele",
> 
> nit typo 's/detele/delete/'

my fault, will fix it in v2.

> 
>> +		  "err %d errno %d\n", err, errno))
>> +		goto disable_pmu;
> 
> should we also check the record got deleted? like make sure following
> lookup fails with NOENT:
> 

yes we can, will add it in v2. Thanks.

>    bpf_map_lookup_elem(stackmap_fd, &stackmap_key, &val_buf)
> 
>> +
>>   disable_pmu:
>>   	bpf_link__destroy(link);
>>   close_prog:
>> diff --git a/tools/testing/selftests/bpf/progs/test_stacktrace_map.c b/tools/testing/selftests/bpf/progs/test_stacktrace_map.c
>> index 47568007b66..d036e8e9c83 100644
>> --- a/tools/testing/selftests/bpf/progs/test_stacktrace_map.c
>> +++ b/tools/testing/selftests/bpf/progs/test_stacktrace_map.c
>> @@ -38,6 +38,13 @@ struct {
>>   	__type(value, stack_trace_t);
>>   } stack_amap SEC(".maps");
>>   
>> +struct {
>> +	__uint(type, BPF_MAP_TYPE_ARRAY);
>> +	__uint(max_entries, 1);
>> +	__type(key, __u32);
>> +	__type(value, __u32);
>> +} stack_key_map SEC(".maps");
>> +
>>   /* taken from /sys/kernel/tracing/events/sched/sched_switch/format */
>>   struct sched_switch_args {
>>   	unsigned long long pad;
>> @@ -54,7 +61,7 @@ SEC("tracepoint/sched/sched_switch")
>>   int oncpu(struct sched_switch_args *ctx)
>>   {
>>   	__u32 max_len = PERF_MAX_STACK_DEPTH * sizeof(__u64);
>> -	__u32 key = 0, val = 0, *value_p;
>> +	__u32 key = 0, val = 0, *value_p, stackmap_key = 0;
>>   	void *stack_p;
>>   
>>   	value_p = bpf_map_lookup_elem(&control_map, &key);
>> @@ -64,6 +71,9 @@ int oncpu(struct sched_switch_args *ctx)
>>   	/* The size of stackmap and stackid_hmap should be the same */
>>   	key = bpf_get_stackid(ctx, &stackmap, 0);
>>   	if ((int)key >= 0) {
>> +		val = key;
>> +		bpf_map_update_elem(&stack_key_map, &stackmap_key, &val, 0);
> 
> why not use '&key' directly as the update value?
> 
will clean it in v2.

> jirka
> 
> 
>> +		val = 0;
>>   		bpf_map_update_elem(&stackid_hmap, &key, &val, 0);
>>   		stack_p = bpf_map_lookup_elem(&stack_amap, &key);
>>   		if (stack_p)
>> -- 
>> 2.48.1
>>
>>
-- 
Best Regards
Tao Chen