Re: [net-next PATCH 1/3] net: introduce napi_is_scheduled helper

[Christian Marangi]

On Sat, Sep 30, 2023 at 03:42:20PM +0200, Eric Dumazet wrote:
> On Sat, Sep 30, 2023 at 2:11 PM Christian Marangi <ansuelsmth@xxxxxxxxx> wrote:
> >
> > On Sat, Sep 30, 2023 at 01:59:53PM +0200, Eric Dumazet wrote:
> > > On Fri, Sep 22, 2023 at 1:13 PM Christian Marangi <ansuelsmth@xxxxxxxxx> wrote:
> > > >
> > > > We currently have napi_if_scheduled_mark_missed that can be used to
> > > > check if napi is scheduled but that does more thing than simply checking
> > > > it and return a bool. Some driver already implement custom function to
> > > > check if napi is scheduled.
> > > >
> > > > Drop these custom function and introduce napi_is_scheduled that simply
> > > > check if napi is scheduled atomically.
> > > >
> > > > Update any driver and code that implement a similar check and instead
> > > > use this new helper.
> > > >
> > > > Signed-off-by: Christian Marangi <ansuelsmth@xxxxxxxxx>
> > > > ---
> > > >  drivers/net/ethernet/chelsio/cxgb3/sge.c  | 8 --------
> > > >  drivers/net/wireless/realtek/rtw89/core.c | 2 +-
> > > >  include/linux/netdevice.h                 | 5 +++++
> > > >  net/core/dev.c                            | 2 +-
> > > >  4 files changed, 7 insertions(+), 10 deletions(-)
> > > >
> > > > diff --git a/drivers/net/ethernet/chelsio/cxgb3/sge.c b/drivers/net/ethernet/chelsio/cxgb3/sge.c
> > > > index 2e9a74fe0970..71fa2dc19034 100644
> > > > --- a/drivers/net/ethernet/chelsio/cxgb3/sge.c
> > > > +++ b/drivers/net/ethernet/chelsio/cxgb3/sge.c
> > > > @@ -2501,14 +2501,6 @@ static int napi_rx_handler(struct napi_struct *napi, int budget)
> > > >         return work_done;
> > > >  }
> > > >
> > > > -/*
> > > > - * Returns true if the device is already scheduled for polling.
> > > > - */
> > > > -static inline int napi_is_scheduled(struct napi_struct *napi)
> > > > -{
> > > > -       return test_bit(NAPI_STATE_SCHED, &napi->state);
> > > > -}
> > > > -
> > > >  /**
> > > >   *     process_pure_responses - process pure responses from a response queue
> > > >   *     @adap: the adapter
> > > > diff --git a/drivers/net/wireless/realtek/rtw89/core.c b/drivers/net/wireless/realtek/rtw89/core.c
> > > > index 133bf289bacb..bbf4ea3639d4 100644
> > > > --- a/drivers/net/wireless/realtek/rtw89/core.c
> > > > +++ b/drivers/net/wireless/realtek/rtw89/core.c
> > > > @@ -1744,7 +1744,7 @@ static void rtw89_core_rx_to_mac80211(struct rtw89_dev *rtwdev,
> > > >         struct napi_struct *napi = &rtwdev->napi;
> > > >
> > > >         /* In low power mode, napi isn't scheduled. Receive it to netif. */
> > > > -       if (unlikely(!test_bit(NAPI_STATE_SCHED, &napi->state)))
> > > > +       if (unlikely(!napi_is_scheduled(napi)))
> > > >                 napi = NULL;
> > > >
> > > >         rtw89_core_hw_to_sband_rate(rx_status);
> > > > diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
> > > > index db3d8429d50d..8eac00cd3b92 100644
> > > > --- a/include/linux/netdevice.h
> > > > +++ b/include/linux/netdevice.h
> > > > @@ -482,6 +482,11 @@ static inline bool napi_prefer_busy_poll(struct napi_struct *n)
> > > >         return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
> > > >  }
> > > >
> > >
> > >
> > > In which context is it safe to call this helper ?
> > >
> >
> > test_bit is atomic so it should be always safe. Also the idea of this
> > check (and from what I can see this apply also to the other 2 user) is
> > somehow best effort, we check if in the current istant there is a napi
> > scheduled and we act.
> 
> I think testing a bit here is not enough to take any kind of useful decision,
> unless used in a particular context.
>

Ehhh the idea here was to reduce code duplication since the very same
test will be done in stmmac. So I guess this code cleanup is a NACK and
I have to duplicate the test in the stmmac driver.

> >
> > > I fear that making this available will add more bugs.
> > >
> > > For instance rspq_check_napi() seems buggy to me.
> > >
> >
> > Mhhh why? Am I opening a can of worms?
> 
> Yes I think :/
> 
> Because only the thread that has grabbed the bit can make any sense of it.
> 
> Another thread reading it would not really know if the value is not going to
> change immediately. So what would be the point ?
> 
> It seems rspq_check_napi() real intent was maybe the following,
> but really this is hard to know if the current race in this code is okay or not.
> 
> diff --git a/drivers/net/ethernet/chelsio/cxgb3/sge.c
> b/drivers/net/ethernet/chelsio/cxgb3/sge.c
> index 2e9a74fe0970df333226b80af8716f30865c01b7..e153c9590b36b38e430bc93660146b428e9b3347
> 100644
> --- a/drivers/net/ethernet/chelsio/cxgb3/sge.c
> +++ b/drivers/net/ethernet/chelsio/cxgb3/sge.c
> @@ -2676,8 +2676,10 @@ static int rspq_check_napi(struct sge_qset *qs)
> 
>         if (!napi_is_scheduled(&qs->napi) &&
>             is_new_response(&q->desc[q->cidx], q)) {
> -               napi_schedule(&qs->napi);
> -               return 1;
> +               if (napi_schedule_prep(&qs->napi)) {
> +                       __napi_schedule(&qs->napi);
> +                       return 1;
> +               }
>         }
>         return 0;
>  }

-- 
	Ansuel