Mathieu Desnoyers <mathieu.desnoyers@xxxxxxxxxxxx> writes:
> On 14-Feb-2020 02:39:24 PM, Thomas Gleixner wrote:
> [...]
>> +#define BPF_PROG_RUN_PIN_ON_CPU(prog, ctx) ({ \
>> + u32 ret; \
>> + migrate_disable(); \
>> + ret = __BPF_PROG_RUN(prog, ctx, bpf_dispatcher_nopfunc); \
>> + migrate_enable(); \
>> + ret; })
>
> Does it really have to be a statement expression with a local variable ?
>
> If so, we should consider renaming "ret" to "__ret" to minimize the
> chances of a caller issuing BPF_PROG_RUN_PIN_ON_CPU with "ret" as
> prog or ctx argument, which would lead to unexpected results.
Indeed. That really can be an inline.
Thanks,
tglx
