Re: [PATCH v2] axs_nand - add driver for NAND controller used on Synopsys AXS dev boards

[ezequiel.garcia@xxxxxxxxxxxxxxxxxx]

On Apr 11, Alexey Brodkin wrote:
> Hi Ezequiel,
> 
> On Mon, 2014-04-07 at 10:13 +0400, Alexey Brodkin wrote:
> > Hi Ezequiel,
> > 
> > On Fri, 2014-04-04 at 11:09 -0300, Ezequiel Garcia wrote:
> > > On Apr 04, Alexey Brodkin wrote:
> > > > Signed-off-by: Alexey Brodkin <abrodkin@xxxxxxxxxxxx>
> > > > 
> > > 
> > > Maybe it would be nice adding some driver description here, so the commit
> > > log actually says something useful about the commit.
> > 
> > Well, I thought about it but frankly I'm not sure if there's anything
> > else to add to commit title "driver for NAND controller used on Synopsys
> > AXS dev boards".
> > 
> > Do you think more info may make sense?
> > 
> > > > +/**
> > > > + * axs_flag_wait_and_reset - Waits until requested flag in INT_STATUS register
> > > > + *              is set by HW and resets it by writing "1" in INT_CLR_STATUS.
> > > > + * @host:	Pointer to private data structure.
> > > > + * @flag:	Bit/flag offset in INT_STATUS register
> > > > + */
> > > > +static void axs_flag_wait_and_reset(struct axs_nand_host *host, int flag)
> > > > +{
> > > > +	unsigned int i;
> > > > +
> > > > +	for (i = 0; i < AXS_FLAG_WAIT_DELAY * 100; i++) {
> > > > +		unsigned int status = reg_get(host, INT_STATUS);
> > > > +
> > > > +		if (status & (1 << flag)) {
> > > > +			reg_set(host, INT_CLR_STATUS, 1 << flag);
> > > > +			return;
> > > > +		}
> > > > +
> > > > +		udelay(10);
> > > > +	}
> > > > +
> > > > +	/*
> > > > +	 * Since we cannot report this problem any further than
> > > > +	 * axs_nand_{write|read}_buf() letting user know there's a problem.
> > > > +	 */
> > > > +	dev_err(host->dev, "Waited too long (%d s.) for flag/bit %d\n",
> > > > +		AXS_FLAG_WAIT_DELAY, flag);
> > > > +}
> > > 
> > > Hm... I'm not sure the above is really true.
> > > 
> > > The NAND core uses the replaceable chip->waitfunc callback to check the
> > > status of issued commands. See for instance:
> > > 
> > > static int nand_write_oob_std(struct mtd_info *mtd, struct nand_chip *chip,
> > >                               int page)
> > > {
> > >         int status = 0;
> > >         const uint8_t *buf = chip->oob_poi;
> > >         int length = mtd->oobsize;
> > > 
> > >         chip->cmdfunc(mtd, NAND_CMD_SEQIN, mtd->writesize, page);
> > >         chip->write_buf(mtd, buf, length);
> > >         /* Send command to program the OOB data */
> > >         chip->cmdfunc(mtd, NAND_CMD_PAGEPROG, -1, -1);
> > > 
> > >         status = chip->waitfunc(mtd, chip);
> > > 
> > >         return status & NAND_STATUS_FAIL ? -EIO : 0;
> > > }
> > > 
> > > On the other side, if you are clearing the flags in axs_flag_wait_and_reset()
> > > it might be a bit hard for you to get this right.
> > > 
> > > IOW, I'm not saying you *must* do this, but instead suggesting that you take
> > > a look at waitfunc() and see if it helps report a proper error in the
> > > read/write path.
> > 
> > As I may understand from generic implementation of "waitfunc" in
> > "nand_base.c" it checks status of NAND chip itself - which IMHO makes
> > sense. And this is done via NAND_CMD_STATUS command sent to chip through
> > NAND controller.
> > 
> > In AXS NAND driver you may see 2 types of wait checks:
> > 1. NAND_ISR_CMDDONE - this flag means that NAND controller has just
> > executed provided command (i.e. set proper signal on control lines of
> > NAND chip etc), but it doesn't mean NAND chip itself has completed
> > command execution. That's why "waitfunc" is not direct replacement here.
> > 
> > 2. NAND_ISR_{RX|TX}DMACOMPLETE - this flag indicates completion of DMA
> > transfer to/from NAND chip to RAM. "waitfunc" won't work here as well.
> > 
> > I hope this explanation makes sense.
> 
> Please treat this message as a gentle reminder.
> I'm wondering if my explanation in the previous email makes any sense or
> I still need to fix stuff.

Well, I was merely suggesting to look into using waitfunc(), so you definitely
don't need to fix anything (at least from my side).

Just as a comment, regarding your explanation above, I think you can override
the waitfunc() and check the NAND_CMD_STATUS, together with the check for your
ISR flags. However, you know more about your hardware than me, so this is
*just* a suggestion.

The driver looks fine broadly speaking. I guess it's up to Brian now to provide
further feedback.
-- 
Ezequiel García, Free Electrons
Embedded Linux, Kernel and Android Engineering
http://free-electrons.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/