Re: [PATCH 07/12] mtd: nand: return failure when failing to get ECC from ONFI

[Ezequiel Garcia <ezequiel.garcia@free-electrons.com>]

On Wed, Sep 11, 2013 at 02:43:19PM -0700, Brian Norris wrote:
> On Thu, Sep 05, 2013 at 10:34:21AM +0800, Huang Shijie wrote:
> > 于 2013年09月05日 02:25, Brian Norris 写道:
> > > If ONFI fails to provide ECC information, we should not use it. Print
> > > appropriate error messages and return.
> > >
> > > Signed-off-by: Brian Norris <computersforpeace@gmail.com>
> > > Cc: Huang Shijie <b32955@freescale.com>
> > > ---
> > >  drivers/mtd/nand/nand_base.c | 9 +++++++--
> > >  1 file changed, 7 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
> > > index 7ed4841..38b8dd4 100644
> > > --- a/drivers/mtd/nand/nand_base.c
> > > +++ b/drivers/mtd/nand/nand_base.c
> > > @@ -3010,8 +3010,13 @@ static int nand_flash_detect_onfi(struct mtd_info *mtd, struct nand_chip *chip,
> > >  			chip->cmdfunc = nand_command_lp;
> > >  
> > >  		/* The Extended Parameter Page is supported since ONFI 2.1. */
> > > -		if (nand_flash_detect_ext_param_page(mtd, chip, p))
> > > -			pr_info("Failed to detect the extended param page.\n");
> > > +		if (nand_flash_detect_ext_param_page(mtd, chip, p)) {
> > > +			pr_info("Failed to detect ONFI extended param page\n");
> > > +			return 0;
> > > +		}
> > > +	} else {
> > > +		pr_err("Could not retrieve ONFI ECC requirements\n");
> > > +		return 0;
> > This return will skips the message "ONFI flash detected", could we
> > remove the 'return' from this patch?
> 
> Well, the "ONFI flash detected" message is not significant at all. In
> fact, it is pretty useless and should just be dropped, I think. The real
> question is whether we signal ONFI detected (return 1) or failed (return
> 0) when ONFI doesn't report ECC properly.
> 
> There are two "return"s in my patch. Are you suggesting removing both of
> them, and just warning the user of the failure?
> 
> > most of the drivers will not use the ECC info, so i think even we fail
> > to detect the ECC info
> > we do not need to skip the message "ONFI flash detected".
> 
> You do have a point here, that most drivers don't (yet) use the ECC
> info.
> 
> > My suggestion is to use the pr_warn() to just print out the warning,
> > such as:
> > pr_warn("Could not retrieve ONFI ECC requirements\n");
> 
> I may turn both prints into warnings, drop both of the 'return 0' (at
> least until ONFI ECC detection is more heavily relied on), and drop all
> the prints that are done for the success case ("ONFI flash detected" and
> "ONFI param page X valid").
> 

FWIW, I second the "drop all prints for the success case".
We're being too verbose in MTD, printing all sorts of useless information.

Or, if some 'notification' message is considered important, then let's make
it a pr_debug and be available for the only use case that will need it:
debugging/development.
-- 
Ezequiel García, Free Electrons
Embedded Linux, Kernel and Android Engineering
http://free-electrons.com