Re: [PATCH 1/1] I2C: designware: fix IO timeout issue for AMD controller

[Mika Westerberg <mika.westerberg@linux.intel.com>]

On Fri, Nov 06, 2015 at 04:34:19AM +0000, Yu, Xiangliang wrote:
> > -----Original Message-----
> > From: Mika Westerberg [mailto:mika.westerberg@linux.intel.com]
> > Sent: Thursday, November 05, 2015 9:52 PM
> > To: Yu, Xiangliang
> > Cc: andriy.shevchenko@linux.intel.com; jarkko.nikula@linux.intel.com;
> > wsa@the-dreams.de; linux-i2c@vger.kernel.org; linux-
> > kernel@vger.kernel.org; Xue, Ken; Wan, Vincent; Huang, Ray; Wang, Annie;
> > Li, Tony
> > Subject: Re: [PATCH 1/1] I2C: designware: fix IO timeout issue for AMD
> > controller
> > 
> > On Thu, Nov 05, 2015 at 08:34:44PM +0800, Xiangliang Yu wrote:
> > > Because of some hardware limitation, AMD I2C controller can't trigger
> > > next interrupt if interrupt status has been changed after clearing
> > > interrupt status bits. Then, I2C will lost interrupt and IO timeout.
> > >
> > > According to hardware design, this patch implements a workaround to
> > > disable i2c controller interrupt after clearing interrupt bits when
> > > entering ISR and to enable i2c interrupt before exiting ISR.
> > >
> > > To reduce the performance impacts on other vendors, use unlikely
> > > function to check flag in ISR.
> > >
> > > Signed-off-by: Xiangliang Yu <Xiangliang.Yu@amd.com>
> > > ---
> > >  drivers/i2c/busses/i2c-designware-core.c    | 6 ++++++
> > >  drivers/i2c/busses/i2c-designware-core.h    | 1 +
> > >  drivers/i2c/busses/i2c-designware-platdrv.c | 4 ++++
> > >  3 files changed, 11 insertions(+)
> > >
> > > diff --git a/drivers/i2c/busses/i2c-designware-core.c
> > > b/drivers/i2c/busses/i2c-designware-core.c
> > > index 7441cdc..04e7b65 100644
> > > --- a/drivers/i2c/busses/i2c-designware-core.c
> > > +++ b/drivers/i2c/busses/i2c-designware-core.c
> > > @@ -783,6 +783,9 @@ irqreturn_t i2c_dw_isr(int this_irq, void *dev_id)
> > >
> > >  	stat = i2c_dw_read_clear_intrbits(dev);
> > 
> > What if the status changes right here, before you go and mask the interrupt?
> Have no effect, because i2c controller can't trigger next interrupt.
> 
> > 
> > >
> > > +	if (unlikely(dev->accessor_flags & ACCESS_INTR_MASK))
> > > +		i2c_dw_disable_int(dev);
> > > +
> > >  	if (stat & DW_IC_INTR_TX_ABRT) {
> > >  		dev->cmd_err |= DW_IC_ERR_TX_ABRT;
> > >  		dev->status = STATUS_IDLE;
> > > @@ -811,6 +814,9 @@ tx_aborted:
> > >  	if ((stat & (DW_IC_INTR_TX_ABRT | DW_IC_INTR_STOP_DET)) ||
> > dev->msg_err)
> > >  		complete(&dev->cmd_complete);
> > >
> > > +	if (unlikely(dev->accessor_flags & ACCESS_INTR_MASK))
> > > +		dw_writel(dev, DW_IC_INTR_DEFAULT_MASK,
> > DW_IC_INTR_MASK);
> > 
> > The driver disables TX interrupt when it is not needed anymore or when TX
> > gets aborted but the above will re-enable all interrupts regardless.
> > Is that the intention?
> No, i2c controller can trigger next interrupt only after re-enable all interrupt.

If you get an error the function masks all interrupts and jumps to
tx_aborted label. With this patch you unmask all interrupts again before
exiting the function.

> > 
> > > +
> > >  	return IRQ_HANDLED;
> > >  }
> > >  EXPORT_SYMBOL_GPL(i2c_dw_isr);
> > > diff --git a/drivers/i2c/busses/i2c-designware-core.h
> > > b/drivers/i2c/busses/i2c-designware-core.h
> > > index 9630222..808ef6a 100644
> > > --- a/drivers/i2c/busses/i2c-designware-core.h
> > > +++ b/drivers/i2c/busses/i2c-designware-core.h
> > > @@ -111,6 +111,7 @@ struct dw_i2c_dev {
> > >
> > >  #define ACCESS_SWAP		0x00000001
> > >  #define ACCESS_16BIT		0x00000002
> > > +#define ACCESS_INTR_MASK	0x00000004
> > >
> > >  extern u32 dw_readl(struct dw_i2c_dev *dev, int offset);  extern void
> > > dw_writel(struct dw_i2c_dev *dev, u32 b, int offset); diff --git
> > > a/drivers/i2c/busses/i2c-designware-platdrv.c
> > > b/drivers/i2c/busses/i2c-designware-platdrv.c
> > > index 472b882..0c59357 100644
> > > --- a/drivers/i2c/busses/i2c-designware-platdrv.c
> > > +++ b/drivers/i2c/busses/i2c-designware-platdrv.c
> > > @@ -251,6 +251,10 @@ static int dw_i2c_probe(struct platform_device
> > *pdev)
> > >  		dev->rx_fifo_depth = ((param1 >> 8)  & 0xff) + 1;
> > >  		dev->adapter.nr = pdev->id;
> > >  	}
> > > +
> > > +	if (!strncmp(pdev->name, "AMD0010", 7))
> > > +		dev->accessor_flags |= ACCESS_INTR_MASK;
> > > +
> > 
> > Can't you put this to ->driver_data? For example it could refer to a
> > configuration structure that then contains quirk flags.
> I think it will need to add a function to support it, but the function
> Is  rarely used. It will easy to add if i2c driver has quirk mechanisms.
> Added code is very simple and have no influence on others.

What if the next AMD I2C controller needs another quirk, and then
another?

I would rather pass flags or similar (reference to config structure
including flags) with ->driver_data.

> 
> > 
> > >  	r = i2c_dw_init(dev);
> > >  	if (r)
> > >  		return r;
> > > --
> > > 1.9.1