]> git.ipfire.org Git - thirdparty/kernel/stable.git/commitdiff
Merge branch 'selftests-bpf-introduce-experimental-bpf_in_interrupt'
authorAlexei Starovoitov <ast@kernel.org>
Thu, 4 Sep 2025 16:05:58 +0000 (09:05 -0700)
committerAlexei Starovoitov <ast@kernel.org>
Thu, 4 Sep 2025 16:05:58 +0000 (09:05 -0700)
Leon Hwang says:

====================
selftests/bpf: Introduce experimental bpf_in_interrupt()

Filtering pid_tgid is meanlingless when the current task is preempted by
an interrupt.

To address this, introduce 'bpf_in_interrupt()' helper function, which
allows BPF programs to determine whether they are executing in interrupt
context.

'get_preempt_count()':

* On x86, '*(int *) bpf_this_cpu_ptr(&__preempt_count)'.
* On arm64, 'bpf_get_current_task_btf()->thread_info.preempt.count'.

Then 'bpf_in_interrupt()' will be:

* If !PREEMPT_RT, 'get_preempt_count() & (NMI_MASK | HARDIRQ_MASK
  | SOFTIRQ_MASK)'.
* If PREEMPT_RT, '(get_preempt_count() & (NMI_MASK | HARDIRQ_MASK))
  | (bpf_get_current_task_btf()->softirq_disable_cnt & SOFTIRQ_MASK)'.

'bpf_in_interrupt()' runs well when PREEMPT_RT is enabled. But it's
difficult for me to test it well because I'm not familiar with
PREEMPT_RT.

Changes:
v2 -> v3:
* Address comments from Alexei:
  * Move bpf_in_interrupt() to bpf_experimental.h.
  * Add support for arm64.
v2: https://lore.kernel.org/bpf/20250825131502.54269-1-leon.hwang@linux.dev/

v1 -> v2:
* Fix a build error reported by test bot.
====================

Link: https://patch.msgid.link/20250903140438.59517-1-leon.hwang@linux.dev
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Trivial merge