[PATCH bpf v2 0/2] bpf: Avoid RCU context warning when unpinning htab with internal structs

[PATCH bpf v2 0/2] bpf: Avoid RCU context warning when unpinning htab with internal structs

[KaFai Wan]

This small patchset is about avoid RCU context warning when unpinning
htab with internal structs (timer, workqueue, or task_work).

v2:
  - rename bpf_free_inode() to bpf_destroy_inode() (Andrii)

v1:
 https://lore.kernel.org/all/20251003084528.502518-1-kafai.wan@linux.dev/

---
KaFai Wan (2):
  bpf: Avoid RCU context warning when unpinning htab with internal
    structs
  selftests/bpf: Add test for unpinning htab with internal timer struct

 kernel/bpf/inode.c                            |  4 +-
 .../selftests/bpf/prog_tests/pinning_htab.c   | 37 +++++++++++++++++++
 .../selftests/bpf/progs/test_pinning_htab.c   | 25 +++++++++++++
 3 files changed, 64 insertions(+), 2 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/pinning_htab.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_pinning_htab.c

-- 
2.43.0

[PATCH bpf v2 1/2] bpf: Avoid RCU context warning when unpinning htab with internal structs

[KaFai Wan]

When unpinning a BPF hash table (htab or htab_lru) that contains internal
structures (timer, workqueue, or task_work) in its values, a BUG warning
is triggered:
 BUG: sleeping function called from invalid context at kernel/bpf/hashtab.c:244
 in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 14, name: ksoftirqd/0
 ...

The issue arises from the interaction between BPF object unpinning and
RCU callback mechanisms:
1. BPF object unpinning uses ->free_inode() which schedules cleanup via
   call_rcu(), deferring the actual freeing to an RCU callback that
   executes within the RCU_SOFTIRQ context.
2. During cleanup of hash tables containing internal structures,
   htab_map_free_internal_structs() is invoked, which includes
   cond_resched() or cond_resched_rcu() calls to yield the CPU during
   potentially long operations.

However, cond_resched() or cond_resched_rcu() cannot be safely called from
atomic RCU softirq context, leading to the BUG warning when attempting
to reschedule.

Fix this by changing from ->free_inode() to ->destroy_inode() and rename
bpf_free_inode() to bpf_destroy_inode() for BPF objects (prog, map, link).
This allows direct inode freeing without RCU callback scheduling,
avoiding the invalid context warning.

Reported-by: Le Chen <tom2cat@sjtu.edu.cn>
Closes: https://lore.kernel.org/all/1444123482.1827743.1750996347470.JavaMail.zimbra@sjtu.edu.cn/
Fixes: 68134668c17f ("bpf: Add map side support for bpf timers.")
Suggested-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: KaFai Wan <kafai.wan@linux.dev>
---
 kernel/bpf/inode.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/bpf/inode.c b/kernel/bpf/inode.c
index f90bdcc0a047..81780bcf8d25 100644
--- a/kernel/bpf/inode.c
+++ b/kernel/bpf/inode.c
@@ -775,7 +775,7 @@ static int bpf_show_options(struct seq_file *m, struct dentry *root)
 	return 0;
 }
 
-static void bpf_free_inode(struct inode *inode)
+static void bpf_destroy_inode(struct inode *inode)
 {
 	enum bpf_type type;
 
@@ -790,7 +790,7 @@ const struct super_operations bpf_super_ops = {
 	.statfs		= simple_statfs,
 	.drop_inode	= inode_just_drop,
 	.show_options	= bpf_show_options,
-	.free_inode	= bpf_free_inode,
+	.destroy_inode	= bpf_destroy_inode,
 };
 
 enum {
-- 
2.43.0

[PATCH bpf v2 2/2] selftests/bpf: Add test for unpinning htab with internal timer struct

[KaFai Wan]

Add test to verify that unpinning hash tables containing internal timer
structures does not trigger context warnings.

Each subtest (timer_prealloc and timer_no_prealloc) can trigger the
context warning when unpinning, but the warning cannot be triggered
twice within a short time interval (a HZ), which is expected behavior.

Signed-off-by: KaFai Wan <kafai.wan@linux.dev>
---
 .../selftests/bpf/prog_tests/pinning_htab.c   | 37 +++++++++++++++++++
 .../selftests/bpf/progs/test_pinning_htab.c   | 25 +++++++++++++
 2 files changed, 62 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/pinning_htab.c
 create mode 100644 tools/testing/selftests/bpf/progs/test_pinning_htab.c

diff --git a/tools/testing/selftests/bpf/prog_tests/pinning_htab.c b/tools/testing/selftests/bpf/prog_tests/pinning_htab.c
new file mode 100644
index 000000000000..fc804bb87b26
--- /dev/null
+++ b/tools/testing/selftests/bpf/prog_tests/pinning_htab.c
@@ -0,0 +1,37 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <test_progs.h>
+#include "test_pinning_htab.skel.h"
+
+static void unpin_map(const char *map_name, const char *pin_path)
+{
+	struct test_pinning_htab *skel;
+	struct bpf_map *map;
+	int err;
+
+	skel = test_pinning_htab__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "skel open_and_load"))
+		return;
+
+	map = bpf_object__find_map_by_name(skel->obj, map_name);
+	if (!ASSERT_OK_PTR(map, "bpf_object__find_map_by_name"))
+		goto out;
+
+	err = bpf_map__pin(map, pin_path);
+	if (!ASSERT_OK(err, "bpf_map__pin"))
+		goto out;
+
+	err = bpf_map__unpin(map, pin_path);
+	if (!ASSERT_OK(err, "bpf_map__unpin"))
+		goto out;
+out:
+	test_pinning_htab__destroy(skel);
+}
+
+void test_pinning_htab(void)
+{
+	if (test__start_subtest("timer_prealloc"))
+		unpin_map("timer_prealloc", "/sys/fs/bpf/timer_prealloc");
+	if (test__start_subtest("timer_no_prealloc"))
+		unpin_map("timer_no_prealloc", "/sys/fs/bpf/timer_no_prealloc");
+}
diff --git a/tools/testing/selftests/bpf/progs/test_pinning_htab.c b/tools/testing/selftests/bpf/progs/test_pinning_htab.c
new file mode 100644
index 000000000000..ae227930c73c
--- /dev/null
+++ b/tools/testing/selftests/bpf/progs/test_pinning_htab.c
@@ -0,0 +1,25 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include "vmlinux.h"
+#include <bpf/bpf_helpers.h>
+
+char _license[] SEC("license") = "GPL";
+
+struct timer_val {
+	struct bpf_timer timer;
+};
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__type(key, __u32);
+	__type(value, struct timer_val);
+	__uint(max_entries, 1);
+} timer_prealloc SEC(".maps");
+
+struct {
+	__uint(type, BPF_MAP_TYPE_HASH);
+	__type(key, __u32);
+	__type(value, struct timer_val);
+	__uint(max_entries, 1);
+	__uint(map_flags, BPF_F_NO_PREALLOC);
+} timer_no_prealloc SEC(".maps");
-- 
2.43.0

Re: [PATCH bpf v2 2/2] selftests/bpf: Add test for unpinning htab with internal timer struct

[Yonghong Song]

On 10/6/25 6:22 PM, KaFai Wan wrote:
> Add test to verify that unpinning hash tables containing internal timer
> structures does not trigger context warnings.
>
> Each subtest (timer_prealloc and timer_no_prealloc) can trigger the
> context warning when unpinning, but the warning cannot be triggered
> twice within a short time interval (a HZ), which is expected behavior.
>
> Signed-off-by: KaFai Wan <kafai.wan@linux.dev>

Ack with a nit below.

Acked-by: Yonghong Song <yonghong.song@linux.dev>

> ---
>   .../selftests/bpf/prog_tests/pinning_htab.c   | 37 +++++++++++++++++++
>   .../selftests/bpf/progs/test_pinning_htab.c   | 25 +++++++++++++
>   2 files changed, 62 insertions(+)
>   create mode 100644 tools/testing/selftests/bpf/prog_tests/pinning_htab.c
>   create mode 100644 tools/testing/selftests/bpf/progs/test_pinning_htab.c
>
> diff --git a/tools/testing/selftests/bpf/prog_tests/pinning_htab.c b/tools/testing/selftests/bpf/prog_tests/pinning_htab.c
> new file mode 100644
> index 000000000000..fc804bb87b26
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/prog_tests/pinning_htab.c
> @@ -0,0 +1,37 @@
> +// SPDX-License-Identifier: GPL-2.0
> +
> +#include <test_progs.h>
> +#include "test_pinning_htab.skel.h"
> +
> +static void unpin_map(const char *map_name, const char *pin_path)
> +{
> +	struct test_pinning_htab *skel;
> +	struct bpf_map *map;
> +	int err;
> +
> +	skel = test_pinning_htab__open_and_load();
> +	if (!ASSERT_OK_PTR(skel, "skel open_and_load"))
> +		return;
> +
> +	map = bpf_object__find_map_by_name(skel->obj, map_name);
> +	if (!ASSERT_OK_PTR(map, "bpf_object__find_map_by_name"))
> +		goto out;
> +
> +	err = bpf_map__pin(map, pin_path);
> +	if (!ASSERT_OK(err, "bpf_map__pin"))
> +		goto out;
> +
> +	err = bpf_map__unpin(map, pin_path);
> +	if (!ASSERT_OK(err, "bpf_map__unpin"))
> +		goto out;

No need to have 'goto out'. Just do
	ASSERT_OK(err, "bpf_map__unpin"

> +out:
> +	test_pinning_htab__destroy(skel);
> +}
> +
> +void test_pinning_htab(void)
> +{
> +	if (test__start_subtest("timer_prealloc"))
> +		unpin_map("timer_prealloc", "/sys/fs/bpf/timer_prealloc");
> +	if (test__start_subtest("timer_no_prealloc"))
> +		unpin_map("timer_no_prealloc", "/sys/fs/bpf/timer_no_prealloc");
> +}
> diff --git a/tools/testing/selftests/bpf/progs/test_pinning_htab.c b/tools/testing/selftests/bpf/progs/test_pinning_htab.c
> new file mode 100644
> index 000000000000..ae227930c73c
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/progs/test_pinning_htab.c
> @@ -0,0 +1,25 @@
> +// SPDX-License-Identifier: GPL-2.0
> +
> +#include "vmlinux.h"
> +#include <bpf/bpf_helpers.h>
> +
> +char _license[] SEC("license") = "GPL";
> +
> +struct timer_val {
> +	struct bpf_timer timer;
> +};
> +
> +struct {
> +	__uint(type, BPF_MAP_TYPE_HASH);
> +	__type(key, __u32);
> +	__type(value, struct timer_val);
> +	__uint(max_entries, 1);
> +} timer_prealloc SEC(".maps");
> +
> +struct {
> +	__uint(type, BPF_MAP_TYPE_HASH);
> +	__type(key, __u32);
> +	__type(value, struct timer_val);
> +	__uint(max_entries, 1);
> +	__uint(map_flags, BPF_F_NO_PREALLOC);
> +} timer_no_prealloc SEC(".maps");

Re: [PATCH bpf v2 1/2] bpf: Avoid RCU context warning when unpinning htab with internal structs

[Yonghong Song]

On 10/6/25 6:22 PM, KaFai Wan wrote:
> When unpinning a BPF hash table (htab or htab_lru) that contains internal
> structures (timer, workqueue, or task_work) in its values, a BUG warning
> is triggered:
>   BUG: sleeping function called from invalid context at kernel/bpf/hashtab.c:244
>   in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 14, name: ksoftirqd/0
>   ...
>
> The issue arises from the interaction between BPF object unpinning and
> RCU callback mechanisms:
> 1. BPF object unpinning uses ->free_inode() which schedules cleanup via
>     call_rcu(), deferring the actual freeing to an RCU callback that
>     executes within the RCU_SOFTIRQ context.
> 2. During cleanup of hash tables containing internal structures,
>     htab_map_free_internal_structs() is invoked, which includes
>     cond_resched() or cond_resched_rcu() calls to yield the CPU during
>     potentially long operations.
>
> However, cond_resched() or cond_resched_rcu() cannot be safely called from
> atomic RCU softirq context, leading to the BUG warning when attempting
> to reschedule.
>
> Fix this by changing from ->free_inode() to ->destroy_inode() and rename
> bpf_free_inode() to bpf_destroy_inode() for BPF objects (prog, map, link).
> This allows direct inode freeing without RCU callback scheduling,
> avoiding the invalid context warning.
>
> Reported-by: Le Chen <tom2cat@sjtu.edu.cn>
> Closes: https://lore.kernel.org/all/1444123482.1827743.1750996347470.JavaMail.zimbra@sjtu.edu.cn/
> Fixes: 68134668c17f ("bpf: Add map side support for bpf timers.")
> Suggested-by: Alexei Starovoitov <ast@kernel.org>
> Signed-off-by: KaFai Wan <kafai.wan@linux.dev>

Acked-by: Yonghong Song <yonghong.song@linux.dev>