Re: [PATCH v2 20/53] mtd: nand: denali: do not set mtd->name

[Boris Brezillon]

On Wed, 29 Mar 2017 10:19:02 +0900
Masahiro Yamada <yamada.masahiro@xxxxxxxxxxxxx> wrote:

> Hi Boris,
> 
> 2017-03-29 6:40 GMT+09:00 Boris Brezillon <boris.brezillon@xxxxxxxxxxxxxxxxxx>:
> > On Wed, 29 Mar 2017 06:32:24 +0900
> > Masahiro Yamada <yamada.masahiro@xxxxxxxxxxxxx> wrote:
> >  
> >> Hi Boris,
> >>
> >> 2017-03-28 0:31 GMT+09:00 Boris Brezillon <boris.brezillon@xxxxxxxxxxxxxxxxxx>:  
> >> > On Wed, 22 Mar 2017 23:07:27 +0900
> >> > Masahiro Yamada <yamada.masahiro@xxxxxxxxxxxxx> wrote:
> >> >  
> >> >> This will be filled by nand_scan_ident() later.
> >> >>
> >> >> Signed-off-by: Masahiro Yamada <yamada.masahiro@xxxxxxxxxxxxx>
> >> >> ---
> >> >>
> >> >> Changes in v2: None
> >> >>
> >> >>  drivers/mtd/nand/denali.c | 1 -
> >> >>  1 file changed, 1 deletion(-)
> >> >>
> >> >> diff --git a/drivers/mtd/nand/denali.c b/drivers/mtd/nand/denali.c
> >> >> index 3badb1d..1706975 100644
> >> >> --- a/drivers/mtd/nand/denali.c
> >> >> +++ b/drivers/mtd/nand/denali.c
> >> >> @@ -1495,7 +1495,6 @@ int denali_init(struct denali_nand_info *denali)
> >> >>
> >> >>       /* now that our ISR is registered, we can enable interrupts */
> >> >>       denali_set_intr_modes(denali, true);
> >> >> -     mtd->name = "denali-nand";  
> >> >
> >> > Are you sure this is safe to do that? When mtd->name is NULL, the core
> >> > takes the parent name, and in the denali_dt case it's not "denali-nand",
> >> > which means you're breaking mtdparts compat.  
> >>
> >> How big impact is this?  
> >
> > Breaking boot on some platforms (those defining partitions through
> > mtdparts= cmdline parameter), which is not negligible :P.
> >  
> >>
> >> I think a bootloader could give mtdparts=denali-nand:...
> >> but, now we are able to have partitions in DT nodes.  
> >
> > Just because you have a new way to describe partitions (using DT) does
> > not mean people are not using the old one (mtdparts= parameter).  
> 
> 
> I thought DT-node derived name can identify the hardware
> even if multiple Denali controllers exist on an SoC.
> 
> Anyway, I admit this is a kind of breakage.
> 
> If this one is reject, I will drop it drop v3.
> 

Yes, please keep the existing name, changing that without extra
precautions has proven to be a bad idea [1].

Note that we now have a way to give user-friendly names to MTD devices
through DT definitions [2]. So, if you ever want to assing a specific
name to your NAND, all you have to do is add a label property to the
NAND device node, and then, in the driver:

	nand_set_flash_node(chip, denali->dev->of_node);

	/*
	 * Fallback to the default name if no label property was
	 * defined.
	 */
	if (!mtd->name)
		mtd->name = "denali-nand";
		

[1]https://patchwork.ozlabs.org/patch/707065/
[2]https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/include/linux/mtd/mtd.h?id=refs/tags/v4.11-rc4#n386