Re: [PATCH v4 12/16] mtd: onenand: omap2: Enable DMA by default

[Ladislav Michl <ladis@linux-mips.org>]

On Wed, Nov 15, 2017 at 12:08:05PM +0200, Roger Quadros wrote:
> Hi,
> 
> On 11/11/17 23:24, Ladislav Michl wrote:
> > DMA and R/B pin are independent on each other. Use DMA by default.
> 
> Is there a R/B pin on Onenand? It looks more like it has a RDY pin and
> an INT pin (which is used for command completion interrupt).

(See also my previous answer) Yes, there's RDY and INT pin. I tried to use
the same scheme as NAND is using, but perhaps it doesn't fit too well here.
Therefore I'm proposing to keep INT pin name and use int-gpios in DT
bindings. Suggestions?

> I think the subject and message are a bit misleading.
> 
> It should be something like
> 
> mtd: onenand: omap2: decouple DMA enabling from INT pin availability
> 
> INT pin (gpio_irq) is not really needed for DMA but only for notification
> when a command that needs wait has completed. We can still have DMA
> even without gpio_irq available.

Thanks, that's definitely better :-)

> > Signed-off-by: Ladislav Michl <ladis@linux-mips.org>
> > ---
> >  Changes:
> >  -v4: new patch
> > 
> >  drivers/mtd/onenand/omap2.c | 41 +++++++++++++++++------------------------
> >  1 file changed, 17 insertions(+), 24 deletions(-)
> > 
> > diff --git a/drivers/mtd/onenand/omap2.c b/drivers/mtd/onenand/omap2.c
> > index e4857a41760d..62e4ede918c4 100644
> > --- a/drivers/mtd/onenand/omap2.c
> > +++ b/drivers/mtd/onenand/omap2.c
> > @@ -152,17 +152,13 @@ static int omap2_onenand_wait(struct mtd_info *mtd, int state)
> >  		}
> >  
> >  		reinit_completion(&c->irq_done);
> > -		if (c->gpio_irq) {
> > -			result = gpio_get_value(c->gpio_irq);
> > -			if (result == -1) {
> > -				ctrl = read_reg(c, ONENAND_REG_CTRL_STATUS);
> > -				intr = read_reg(c, ONENAND_REG_INTERRUPT);
> > -				wait_err("gpio error", state, ctrl, intr);
> > -				return -EIO;
> > -			}
> > -		} else
> > -			result = 0;
> > -		if (result == 0) {
> > +		result = gpio_get_value(c->gpio_irq);
> > +		if (result < 0) {
> > +			ctrl = read_reg(c, ONENAND_REG_CTRL_STATUS);
> > +			intr = read_reg(c, ONENAND_REG_INTERRUPT);
> > +			wait_err("gpio error", state, ctrl, intr);
> > +			return -EIO;
> > +		} else if (result == 0) {
> >  			int retry_cnt = 0;
> >  retry:
> >  			if (!wait_for_completion_io_timeout(&c->irq_done,
> > @@ -513,13 +509,15 @@ static int omap2_onenand_probe(struct platform_device *pdev)
> 
> We need to get rid if this code from probe.
> 
>         if (pdata->dma_channel < 0) {
>                 /* if -1, don't use DMA */
>                 c->gpio_irq = 0;
>         }

Yes. I overlooked it when shuffling with patches. In ended in later patch, but it
really belongs here. Will fix.

> >  			dev_err(&pdev->dev,  "Failed to request GPIO%d for "
> >  				"OneNAND\n", c->gpio_irq);
> >  			goto err_iounmap;
> > -	}
> > -	gpio_direction_input(c->gpio_irq);
> > +		}
> > +		gpio_direction_input(c->gpio_irq);
> > +
> > +		if ((r = request_irq(gpio_to_irq(c->gpio_irq),
> > +				     omap2_onenand_interrupt, IRQF_TRIGGER_RISING,
> > +				     pdev->dev.driver->name, c)) < 0)
> > +			goto err_release_gpio;
> >  
> > -	if ((r = request_irq(gpio_to_irq(c->gpio_irq),
> > -			     omap2_onenand_interrupt, IRQF_TRIGGER_RISING,
> > -			     pdev->dev.driver->name, c)) < 0)
> > -		goto err_release_gpio;
> > +		this->wait = omap2_onenand_wait;
> >  	}
> >  
> >  	if (pdata->dma_channel >= 0) {
> > @@ -529,15 +527,11 @@ static int omap2_onenand_probe(struct platform_device *pdev)
> >  		dma_cap_set(DMA_MEMCPY, mask);
> >  
> >  		c->dma_chan = dma_request_channel(mask, NULL, NULL);
> > -		if (!c->dma_chan)
> > -			dev_info(&pdev->dev,
> > -				 "failed to allocate DMA for OneNAND, "
> > -				 "using PIO instead\n");
> 
> Why get rid of the print message? Instead we could choose to error out completely
> if a DMA channel was provided and we couldn't get it.

The point is that without pdata->dma_channel condition above, DMA is always
enabled and it seems too strict to me to fail, when driver can continue to work.

> >  	}
> >  
> >  	dev_info(&pdev->dev, "initializing on CS%d, phys base 0x%08lx, virtual "
> > -		 "base %p, freq %d MHz\n", c->gpmc_cs, c->phys_base,
> > -		 c->onenand.base, c->freq);
> > +		 "base %p, freq %d MHz, %s mode\n", c->gpmc_cs, c->phys_base,
> > +		 c->onenand.base, c->freq, c->dma_chan ? "DMA" : "PIO");

See here ^. We just merely print message driver is running in PIO mode.

> >  	c->pdev = pdev;
> >  	c->mtd.priv = &c->onenand;
> > @@ -547,7 +541,6 @@ static int omap2_onenand_probe(struct platform_device *pdev)
> >  
> >  	this = &c->onenand;
> >  	if (c->dma_chan) {
> > -		this->wait = omap2_onenand_wait;
> >  		this->read_bufferram = omap2_onenand_read_bufferram;
> >  		this->write_bufferram = omap2_onenand_write_bufferram;
> >  	}
> > 
> 
> -- 
> cheers,
> -roger
> 
> Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki