Re: [PATCH] context_tracking: Use arch_atomic_*() in __ct_user_enter and __ct_user_exit

[genjian zhang]

On Fri, Jan 6, 2023 at 10:53 PM Paul E. McKenney <paulmck@xxxxxxxxxx> wrote:
>
> On Thu, Nov 10, 2022 at 04:24:43PM +0800, Genjian wrote:
> > From: Genjian Zhang <zhanggenjian@xxxxxxxxxx>
> >
> > vmlinux.o: warning: objtool: __ct_user_enter+0x45: call to
> > __kasan_check_write() leaves .noinstr.text section
> > vmlinux.o: warning: objtool: __ct_user_exit+0x3f: call to
> > __kasan_check_write() leaves .noinstr.text section
> >
> > noinstr cannot have atomic_*() functions.because they have explicit
> > instrumentation.Switch to arch_ prefixed atomic operation functions to
> > avoid the explicit instrumentation.
> >
> > Reported-by: k2ci <kernel-bot@xxxxxxxxxx>
> > Reported-by: kernel test robot <lkp@xxxxxxxxx>
> > Signed-off-by: Genjian Zhang <zhanggenjian@xxxxxxxxxx>
>
> Apologies for the delay, but finally queued, thank you!  Frederic gave
> his ack off-list, which I have included in the wordsmithed version below.
> Could you please check to make sure that I did not mess anything up?
>
>                                                         Thanx, Paul
>
> ------------------------------------------------------------------------
>
> commit 936acd859f4a7b2b0f9900e26bc972385286df6e
> Author: Genjian Zhang <zhanggenjian@xxxxxxxxxx>
> Date:   Thu Nov 10 16:24:43 2022 +0800
>
>     context_tracking: Use arch_atomic_*() in __ct_user_enter and __ct_user_exit
>
>     The following diagnostics are issued by objtool:
>
>     vmlinux.o: warning: objtool: __ct_user_enter+0x45: call to
>     __kasan_check_write() leaves .noinstr.text section
>     vmlinux.o: warning: objtool: __ct_user_exit+0x3f: call to
>     __kasan_check_write() leaves .noinstr.text section
>
>     The reason for these diagnostics is that code marked noinstr if prohibited
>     from using atomic_*() functions, which have explicit instrumentation.
>     Therefore, switch to arch_ prefixed atomic operation functions to avoid
>     the explicit instrumentation.
>
>     Reported-by: k2ci <kernel-bot@xxxxxxxxxx>
>     Reported-by: kernel test robot <lkp@xxxxxxxxx>
>     Signed-off-by: Genjian Zhang <zhanggenjian@xxxxxxxxxx>
>     Acked-by: Frederic Weisbecker <frederic@xxxxxxxxxx>
>     Signed-off-by: Paul E. McKenney <paulmck@xxxxxxxxxx>
>
> diff --git a/kernel/context_tracking.c b/kernel/context_tracking.c
> index 77978e3723771..a09f1c19336ae 100644
> --- a/kernel/context_tracking.c
> +++ b/kernel/context_tracking.c
> @@ -510,7 +510,7 @@ void noinstr __ct_user_enter(enum ctx_state state)
>                          * In this we case we don't care about any concurrency/ordering.
>                          */
>                         if (!IS_ENABLED(CONFIG_CONTEXT_TRACKING_IDLE))
> -                               atomic_set(&ct->state, state);
> +                               arch_atomic_set(&ct->state, state);
>                 } else {
>                         /*
>                          * Even if context tracking is disabled on this CPU, because it's outside
> @@ -527,7 +527,7 @@ void noinstr __ct_user_enter(enum ctx_state state)
>                          */
>                         if (!IS_ENABLED(CONFIG_CONTEXT_TRACKING_IDLE)) {
>                                 /* Tracking for vtime only, no concurrent RCU EQS accounting */
> -                               atomic_set(&ct->state, state);
> +                               arch_atomic_set(&ct->state, state);
>                         } else {
>                                 /*
>                                  * Tracking for vtime and RCU EQS. Make sure we don't race
> @@ -535,7 +535,7 @@ void noinstr __ct_user_enter(enum ctx_state state)
>                                  * RCU only requires RCU_DYNTICKS_IDX increments to be fully
>                                  * ordered.
>                                  */
> -                               atomic_add(state, &ct->state);
> +                               arch_atomic_add(state, &ct->state);
>                         }
>                 }
>         }
> @@ -630,12 +630,12 @@ void noinstr __ct_user_exit(enum ctx_state state)
>                          * In this we case we don't care about any concurrency/ordering.
>                          */
>                         if (!IS_ENABLED(CONFIG_CONTEXT_TRACKING_IDLE))
> -                               atomic_set(&ct->state, CONTEXT_KERNEL);
> +                               arch_atomic_set(&ct->state, CONTEXT_KERNEL);
>
>                 } else {
>                         if (!IS_ENABLED(CONFIG_CONTEXT_TRACKING_IDLE)) {
>                                 /* Tracking for vtime only, no concurrent RCU EQS accounting */
> -                               atomic_set(&ct->state, CONTEXT_KERNEL);
> +                               arch_atomic_set(&ct->state, CONTEXT_KERNEL);
>                         } else {
>                                 /*
>                                  * Tracking for vtime and RCU EQS. Make sure we don't race
> @@ -643,7 +643,7 @@ void noinstr __ct_user_exit(enum ctx_state state)
>                                  * RCU only requires RCU_DYNTICKS_IDX increments to be fully
>                                  * ordered.
>                                  */
> -                               atomic_sub(state, &ct->state);
> +                               arch_atomic_sub(state, &ct->state);
>                         }
>                 }
>         }

Seems good to me.

Thanks,

Genjian.