Re: [PATCH bpf-next v2 1/2] bpf: Introduce bpf_in_interrupt kfunc

[Leon Hwang <leon.hwang@linux.dev>]

On 25/8/25 23:17, Alexei Starovoitov wrote:
> On Mon, Aug 25, 2025 at 6:15 AM Leon Hwang <leon.hwang@linux.dev> wrote:
>>
>> Filtering pid_tgid is meaningless when the current task is preempted by
>> an interrupt.
>>
>> To address this, introduce the bpf_in_interrupt kfunc, which allows BPF
>> programs to determine whether they are executing in interrupt context.
>>
>> This enables programs to avoid applying pid_tgid filtering when running
>> in such contexts.
>>
>> Signed-off-by: Leon Hwang <leon.hwang@linux.dev>
>> ---
>>  kernel/bpf/helpers.c  |  9 +++++++++
>>  kernel/bpf/verifier.c | 11 +++++++++++
>>  2 files changed, 20 insertions(+)
>>
>> diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
>> index 401b4932cc49f..38991b7b4a9e9 100644
>> --- a/kernel/bpf/helpers.c
>> +++ b/kernel/bpf/helpers.c
>> @@ -3711,6 +3711,14 @@ __bpf_kfunc int bpf_strstr(const char *s1__ign, const char *s2__ign)
>>         return bpf_strnstr(s1__ign, s2__ign, XATTR_SIZE_MAX);
>>  }
>>
>> +/**
>> + * bpf_in_interrupt - Check whether it's in interrupt context
>> + */
>> +__bpf_kfunc int bpf_in_interrupt(void)
>> +{
>> +       return in_interrupt();
>> +}
> 
> It doesn't scale. Next thing people will ask for hard vs soft irq.
> 

How about adding a 'flags'?

Here are the values for 'flags':

* 0: return in_interrupt();
* 1(NMI): return in_nmi();
* 2(HARDIRQ): return in_hardirq();
* 3(SOFTIRQ): return in_softirq();

>> +
>>  __bpf_kfunc_end_defs();
>>
>>  BTF_KFUNCS_START(generic_btf_ids)
>> @@ -3751,6 +3759,7 @@ BTF_ID_FLAGS(func, bpf_throw)
>>  #ifdef CONFIG_BPF_EVENTS
>>  BTF_ID_FLAGS(func, bpf_send_signal_task, KF_TRUSTED_ARGS)
>>  #endif
>> +BTF_ID_FLAGS(func, bpf_in_interrupt, KF_FASTCALL)
>>  BTF_KFUNCS_END(generic_btf_ids)
>>
>>  static const struct btf_kfunc_id_set generic_kfunc_set = {
>> diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
>> index 5c9dd16b2c56b..e30ecbfc29dad 100644
>> --- a/kernel/bpf/verifier.c
>> +++ b/kernel/bpf/verifier.c
>> @@ -12259,6 +12259,7 @@ enum special_kfunc_type {
>>         KF_bpf_res_spin_lock_irqsave,
>>         KF_bpf_res_spin_unlock_irqrestore,
>>         KF___bpf_trap,
>> +       KF_bpf_in_interrupt,
>>  };
>>
>>  BTF_ID_LIST(special_kfunc_list)
>> @@ -12327,6 +12328,7 @@ BTF_ID(func, bpf_res_spin_unlock)
>>  BTF_ID(func, bpf_res_spin_lock_irqsave)
>>  BTF_ID(func, bpf_res_spin_unlock_irqrestore)
>>  BTF_ID(func, __bpf_trap)
>> +BTF_ID(func, bpf_in_interrupt)
>>
>>  static bool is_kfunc_ret_null(struct bpf_kfunc_call_arg_meta *meta)
>>  {
>> @@ -21977,6 +21979,15 @@ static int fixup_kfunc_call(struct bpf_verifier_env *env, struct bpf_insn *insn,
>>                    desc->func_id == special_kfunc_list[KF_bpf_rdonly_cast]) {
>>                 insn_buf[0] = BPF_MOV64_REG(BPF_REG_0, BPF_REG_1);
>>                 *cnt = 1;
>> +       } else if (desc->func_id == special_kfunc_list[KF_bpf_in_interrupt]) {
>> +#if defined(CONFIG_X86_64) && !defined(CONFIG_UML)
>> +               insn_buf[0] = BPF_MOV64_IMM(BPF_REG_0, (u32)(unsigned long)&__preempt_count);
> 
> I think bpf_per_cpu_ptr() should already be able to read that per cpu var.
> 

Correct. bpf_per_cpu_ptr() and bpf_this_cpu_ptr() are helpful to read it.

So, this patch seems useless.

>> +               insn_buf[1] = BPF_MOV64_PERCPU_REG(BPF_REG_0, BPF_REG_0);
>> +               insn_buf[2] = BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, 0);
>> +               insn_buf[3] = BPF_ALU32_IMM(BPF_AND, BPF_REG_0, NMI_MASK | HARDIRQ_MASK |
>> +                                           (IS_ENABLED(CONFIG_PREEMPT_RT) ? 0 : SOFTIRQ_MASK));
> 
> This is still wrong in PREEMPT_RT.
> 

You are right.

Double-check the following macros:

#ifdef CONFIG_PREEMPT_RT
# define softirq_count()        (current->softirq_disable_cnt &
SOFTIRQ_MASK)
# define irq_count()            ((preempt_count() & (NMI_MASK |
HARDIRQ_MASK)) | softirq_count())
#else
# define softirq_count()        (preempt_count() & SOFTIRQ_MASK)
# define irq_count()            (preempt_count() & (NMI_MASK |
HARDIRQ_MASK | SOFTIRQ_MASK))
#endif
#define in_interrupt()          (irq_count())

If PREEMPT_RT, 'softirq_count()' is missing here.

Thanks,
Leon