[U-Boot] Fix fsl_elbc_nand driver

[Andrei Yakimov]

On Wed, 2015-05-20 at 22:11 -0500, Scott Wood wrote:
> On Wed, 2015-05-20 at 20:03 -0700, Andrei Yakimov wrote:
> > On Wed, 2015-05-20 at 21:46 -0500, Scott Wood wrote:
> > > On Wed, 2015-05-20 at 19:42 -0700, Andrei Yakimov wrote:
> > > > For now lets stick with 1536 in u-boot.
> > > > I will send a patch.
> > > > At least it will not loosing flash over time
> > > > as nand ages.
> > > > 
> > > > I understand what you wish, and will take a look
> > > > on it inside fresh new kernel. I found one more driver  -
> > > > marvel looks like have same problem.
> > > > I will check how NAND_CMD_RNDOUT is working.
> > > > Perhaps we do not need extra read_param(),
> > > > and use only NAND_CMD_RNDOUT to get next
> > > > block inside page loop.
> > > 
> > > Again, I'm a reluctant to use RNDOUT in the default read_param() because
> > > that would change the flow for all controllers and chips, and while the
> > > chip manual I'm looking at says it's OK, it introduces risk that it
> > > doesn't work everywhere (e.g. some controller drivers that provide their
> > > own cmdfunc don't implement RNDOUT).
> 
> RNDOUT is already used by nand_flash_detect_ext_param_page(), so this
> isn't as much of a concern for ONFI, but it could be an issue with
> nand_flash_detect_jedec().
> 
> > Forget about read_param(), 
> 
> Then how will it work on controllers like eLBC/IFC which is the whole
> point?
> 
> > just like this:
 I miss this line:
      chip->cmdfunc(mtd, NAND_CMD_PARAM, 0, -1);
> >   for (i = 0; i < 3; i++) {
> >                 for (j = 0; j < sizeof(*p); j++)
> >                         ((uint8_t *)p)[j] = chip->read_byte(mtd);
> >                 if (onfi_crc16(ONFI_CRC_BASE, (uint8_t *)p, 254) ==
> >                                 le16_to_cpu(p->crc)) {
> >                         break;
> >                 }
> > 		chip->cmdfunc(mtd, NAND_CMD_RNDOUT, <offest>, -1);
> >   }
> > 
> > 
> > and this is good - will be "no op"  or "bad command" error,
> > which could be ignored - so for this drivers operation flow is
> > unchanged.
> 
> RNDOUT needs to come before read_buf() and it needs to specify the
> offset you want.
> 
column address - it is exactly offset for RNDOUT.
First param read 256/512 bytes, if it fail, we do RNDOUT to get next.
we will not do extra RNDOUT - it is for() loop.

I can test it on my board. This is really good solution if it work.
it is just 1 line, and only when ONFI mark already read from flash.

And we can leave (NAND_CMD_PARAM, 0x40) (JDEC label) handling for kernel
folks.

My problem only jesd230B do not specify PARAM command,
ONFI4.0 - do not expect column address for PARAM.

Linux kernel cleary doing (NAND_CMD_PARAM, 0x40).

This is bit annoying.

Question is elbc/ifc old controllers  - is it worth the effort?

> -Scott
> 
>