This path introduces several kfuncs to help BPF programs determine their current execution context. When hooking functions for statistics, we often need to use current->comm to get the process name.
However, these hooked functions can be called from either process context or interrupt context. When called from interrupt context, the current we obtain may refer to the process that was interrupted, which may not be what we need.
These new kfuncs expose APIs that allow users to determine the actual execution context.
Jiayuan Chen (2): bpf: Add kfuncs for detecting execution context selftests/bpf: Add selftests for context detection kfuncs
kernel/bpf/helpers.c | 45 +++++++++++++++++++ .../selftests/bpf/prog_tests/context.c | 32 +++++++++++++ .../selftests/bpf/progs/context_prog.c | 33 ++++++++++++++ 3 files changed, 110 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/context.c create mode 100644 tools/testing/selftests/bpf/progs/context_prog.c
This path introduces several kfuncs to help BPF programs determine their current execution context. When hooking functions for statistics, we often need to use current->comm to get the process name.
However, these hooked functions can be called from either process context or interrupt context. When called from interrupt context, the current we obtain may refer to the process that was interrupted, which may not be what we need.
These new kfuncs expose APIs that allow users to determine the actual execution context.
Signed-off-by: Jiayuan Chen jiayuan.chen@linux.dev --- kernel/bpf/helpers.c | 45 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 45 insertions(+)
diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c index b9ec6ee21c94..e09c70b4eaea 100644 --- a/kernel/bpf/helpers.c +++ b/kernel/bpf/helpers.c @@ -4252,6 +4252,47 @@ __bpf_kfunc int bpf_task_work_schedule_resume(struct task_struct *task, struct b return bpf_task_work_schedule(task, tw, map__map, callback, aux__prog, TWA_RESUME); }
+/** + * bpf_in_nmi_context - Check whether we are serving NMI + * + * Return: true if we are serving NMI + */ +__bpf_kfunc bool bpf_in_nmi_context(void) +{ + return in_nmi(); +} + +/** + * bpf_in_hardirq_context - Check whether we are serving hard irq + * + * Return: true if we are serving hard irq + */ +__bpf_kfunc bool bpf_in_hardirq_context(void) +{ + return in_hardirq(); +} + +/** + * bpf_in_softirq_context - Check whether we are serving soft irq + * + * Return: true if we are serving soft irq + */ +__bpf_kfunc bool bpf_in_softirq_context(void) +{ + /* in_softirq() has been deprecated */ + return in_serving_softirq(); +} + +/** + * bpf_in_task_context - Check whether we are in task context + * + * Return: true if we are in task context + */ +__bpf_kfunc bool bpf_in_task_context(void) +{ + return in_task(); +} + __bpf_kfunc_end_defs();
static void bpf_task_work_cancel_scheduled(struct irq_work *irq_work) @@ -4429,6 +4470,10 @@ BTF_ID_FLAGS(func, bpf_cgroup_read_xattr, KF_RCU) BTF_ID_FLAGS(func, bpf_stream_vprintk, KF_TRUSTED_ARGS) BTF_ID_FLAGS(func, bpf_task_work_schedule_signal, KF_TRUSTED_ARGS) BTF_ID_FLAGS(func, bpf_task_work_schedule_resume, KF_TRUSTED_ARGS) +BTF_ID_FLAGS(func, bpf_in_softirq_context) +BTF_ID_FLAGS(func, bpf_in_hardirq_context) +BTF_ID_FLAGS(func, bpf_in_task_context) +BTF_ID_FLAGS(func, bpf_in_nmi_context) BTF_KFUNCS_END(common_btf_ids)
static const struct btf_kfunc_id_set common_kfunc_set = {
On 2025/10/22 19:33, Jiayuan Chen wrote:
This path introduces several kfuncs to help BPF programs determine their current execution context. When hooking functions for statistics, we often need to use current->comm to get the process name.
However, these hooked functions can be called from either process context or interrupt context. When called from interrupt context, the current we obtain may refer to the process that was interrupted, which may not be what we need.
These new kfuncs expose APIs that allow users to determine the actual execution context.
Hi Jiayuan,
Rather than introducing multiple kfuncs to determine the current execution context, this can already be achieved by using the 'bpf_this_cpu_ptr()' helper to read the underlying preemption count.
Please refer to my earlier patch, "selftests/bpf: Introduce experimental bpf_in_interrupt()"[1], which demonstrates this approach.
Links: [1] https://lore.kernel.org/bpf/20250903140438.59517-1-leon.hwang@linux.dev/
Thanks, Leon
[...]
Hi Jiayuan,
kernel test robot noticed the following build warnings:
[auto build test WARNING on bpf-next/master]
url: https://github.com/intel-lab-lkp/linux/commits/Jiayuan-Chen/bpf-Add-kfuncs-f... base: https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git master patch link: https://lore.kernel.org/r/20251022113412.352307-2-jiayuan.chen%40linux.dev patch subject: [PATCH bpf-next v1 1/2] bpf: Add kfuncs for detecting execution context config: alpha-randconfig-r123-20251023 (https://download.01.org/0day-ci/archive/20251024/202510242353.hCaLzqVt-lkp@i...) compiler: alpha-linux-gcc (GCC) 11.5.0 reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20251024/202510242353.hCaLzqVt-lkp@i...)
If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags | Reported-by: kernel test robot lkp@intel.com | Closes: https://lore.kernel.org/oe-kbuild-all/202510242353.hCaLzqVt-lkp@intel.com/
sparse warnings: (new ones prefixed by >>) kernel/bpf/helpers.c:1199:21: sparse: sparse: incorrect type in assignment (different address spaces) @@ expected unsigned long long ( *[usertype] callback_fn )( ... ) @@ got void [noderef] __rcu * @@ kernel/bpf/helpers.c:1199:21: sparse: expected unsigned long long ( *[usertype] callback_fn )( ... ) kernel/bpf/helpers.c:1199:21: sparse: got void [noderef] __rcu * kernel/bpf/helpers.c:2480:18: sparse: sparse: symbol 'bpf_task_release_dtor' was not declared. Should it be static? kernel/bpf/helpers.c:2510:18: sparse: sparse: symbol 'bpf_cgroup_release_dtor' was not declared. Should it be static?
kernel/bpf/helpers.c:4260:18: sparse: sparse: symbol 'bpf_in_nmi_context' was not declared. Should it be static?
kernel/bpf/helpers.c:2976:18: sparse: sparse: context imbalance in 'bpf_rcu_read_lock' - wrong count at exit kernel/bpf/helpers.c:2981:18: sparse: sparse: context imbalance in 'bpf_rcu_read_unlock' - unexpected unlock
vim +/bpf_in_nmi_context +4260 kernel/bpf/helpers.c
4254 4255 /** 4256 * bpf_in_nmi_context - Check whether we are serving NMI 4257 * 4258 * Return: true if we are serving NMI 4259 */
4260 __bpf_kfunc bool bpf_in_nmi_context(void)
4261 { 4262 return in_nmi(); 4263 } 4264
Add selftests for the newly introduced context detection kfuncs.
The tests verify that each kfunc correctly identifies its respective execution context by triggering different contexts
Signed-off-by: Jiayuan Chen jiayuan.chen@linux.dev --- .../selftests/bpf/prog_tests/context.c | 32 ++++++++++++++++++ .../selftests/bpf/progs/context_prog.c | 33 +++++++++++++++++++ 2 files changed, 65 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/context.c create mode 100644 tools/testing/selftests/bpf/progs/context_prog.c
diff --git a/tools/testing/selftests/bpf/prog_tests/context.c b/tools/testing/selftests/bpf/prog_tests/context.c new file mode 100644 index 000000000000..f09d24069941 --- /dev/null +++ b/tools/testing/selftests/bpf/prog_tests/context.c @@ -0,0 +1,32 @@ +// SPDX-License-Identifier: GPL-2.0 +#include <error.h> +#include <test_progs.h> +#include "context_prog.skel.h" + +void test_context(void) +{ + struct context_prog *skel = NULL; + + skel = context_prog__open_and_load(); + if (!ASSERT_OK_PTR(skel, "loading prog fail")) + return; + + context_prog__attach(skel); + getuid(); + sleep(5); + + if (!ASSERT_EQ(1, skel->bss->in_hardirq, "hardirq not triggered")) + goto out; + if (!ASSERT_EQ(1, skel->bss->in_softriq, "softirq not triggered")) + goto out; + if (!ASSERT_EQ(1, skel->bss->in_task, "task context not triggered")) + goto out; +out: + context_prog__destroy(skel); +} + +void test_bpf_context(void) +{ + if (test__start_subtest("context")) + test_context(); +} diff --git a/tools/testing/selftests/bpf/progs/context_prog.c b/tools/testing/selftests/bpf/progs/context_prog.c new file mode 100644 index 000000000000..81e21f684a84 --- /dev/null +++ b/tools/testing/selftests/bpf/progs/context_prog.c @@ -0,0 +1,33 @@ +// SPDX-License-Identifier: GPL-2.0 +#include <vmlinux.h> +#include <bpf/bpf_helpers.h> +#include <bpf/bpf_tracing.h> +#include "bpf_misc.h" +#include "bpf_experimental.h" + +int in_hardirq = 0; +int in_softriq = 0; +int in_task = 0; + +SEC("tp/irq/irq_handler_entry") +int trace_irq_handler_entry(const void *ctx) +{ + in_hardirq = bpf_in_hardirq_context(); + return 0; +} + +SEC("tp/irq/softirq_entry") +int trace_softirq_entry(const void *ctx) +{ + in_softriq = bpf_in_softirq_context(); + return 0; +} + +SEC("tp/syscalls/sys_enter_getuid") +int trace_syscall(const void *ctx) +{ + in_task = bpf_in_task_context(); + return 0; +} + +char _license[] SEC("license") = "GPL";
linux-kselftest-mirror@lists.linaro.org