Re: [PATCH V5 3/3] i2c: i2c-qcom-geni: Add shutdown callback for i2c

[rojay]

Hi Stephen,

On 2020-10-03 07:09, Stephen Boyd wrote:
> Quoting Roja Rani Yarubandi (2020-10-01 01:44:25)
> > diff --git a/drivers/i2c/busses/i2c-qcom-geni.c 
> > b/drivers/i2c/busses/i2c-qcom-geni.c
> > index aee2a1dd2c62..56d3fbfe7eb6 100644
> > --- a/drivers/i2c/busses/i2c-qcom-geni.c
> > +++ b/drivers/i2c/busses/i2c-qcom-geni.c
> > @@ -380,6 +380,36 @@ static void geni_i2c_tx_msg_cleanup(struct 
> > geni_i2c_dev *gi2c,
> >         }
> >  }
> >
> > +static void geni_i2c_stop_xfer(struct geni_i2c_dev *gi2c)
> > +{
> > +       int ret;
> > +       u32 geni_status;
> > +       unsigned long flags;
> > +       struct i2c_msg *cur;
> > +
> > +       /* Resume device, runtime suspend can happen anytime during 
> > transfer */
> > +       ret = pm_runtime_get_sync(gi2c->se.dev);
> > +       if (ret < 0) {
> > +               dev_err(gi2c->se.dev, "Failed to resume device: %d\n", 
> > ret);
> > +               return;
> > +       }
> > +
> > +       spin_lock_irqsave(&gi2c->lock, flags);
>
> We grab the lock here.
>
> > +       geni_status = readl_relaxed(gi2c->se.base + SE_GENI_STATUS);
> > +       if (!(geni_status & M_GENI_CMD_ACTIVE))
> > +               goto out;
> > +
> > +       cur = gi2c->cur;
> > +       geni_i2c_abort_xfer(gi2c);
>
> But it looks like this function takes the lock again? Did you test this
> with lockdep enabled? It should hang even without lockdep, so it seems
> like this path of code has not been tested.

Tested with lockdep enabled, and fixed the unsafe lock order issue here.
And to be more proper moved spin_lock/unlock to cleanup functions.

> > +       if (cur->flags & I2C_M_RD)
> > +               geni_i2c_rx_msg_cleanup(gi2c, cur);
> > +       else
> > +               geni_i2c_tx_msg_cleanup(gi2c, cur);
> > +       spin_unlock_irqrestore(&gi2c->lock, flags);
> > +out:
> > +       pm_runtime_put_sync_suspend(gi2c->se.dev);
> > +}
> > +
> >  static int geni_i2c_rx_one_msg(struct geni_i2c_dev *gi2c, struct 
> > i2c_msg *msg,
> >                                 u32 m_param)
> >  {