Re: [PATCH v3] mtd: limit OTP NVMEM Cell parse to non Nand devices

From: Christian Marangi <ansuelsmth@gmail.com>
To: "Rafał Miłecki" <rafal@milecki.pl>
Cc: Miquel Raynal <miquel.raynal@bootlin.com>,
	Richard Weinberger <richard@nod.at>,
	Vignesh Raghavendra <vigneshr@ti.com>,
	Jernej Skrabec <jernej.skrabec@gmail.com>,
	Claudiu Beznea <claudiu.beznea@tuxon.dev>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Srinivas Kandagatla <srinivas.kandagatla@linaro.org>,
	linux-mtd@lists.infradead.org, linux-kernel@vger.kernel.org,
	stable@vger.kernel.org
Subject: Re: [PATCH v3] mtd: limit OTP NVMEM Cell parse to non Nand devices
Date: Sat, 30 Mar 2024 10:13:35 +0100	[thread overview]
Message-ID: <6607d7c3.df0a0220.498b4.86f3@mx.google.com> (raw)
In-Reply-To: <a4632f1ba701f6b333c50a5366723cf4@milecki.pl>

On Thu, Mar 28, 2024 at 03:44:15PM +0100, Rafał Miłecki wrote:
> On 2024-03-28 15:19, Christian Marangi wrote:
> > On Wed, Mar 27, 2024 at 11:15:02PM +0100, Rafał Miłecki wrote:
> > > On 22.03.2024 05:09, Christian Marangi wrote:
> > > > diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
> > > > index 5887feb347a4..0de87bc63840 100644
> > > > --- a/drivers/mtd/mtdcore.c
> > > > +++ b/drivers/mtd/mtdcore.c
> > > > @@ -900,7 +900,7 @@ static struct nvmem_device *mtd_otp_nvmem_register(struct mtd_info *mtd,
> > > >   	config.name = compatible;
> > > >   	config.id = NVMEM_DEVID_AUTO;
> > > >   	config.owner = THIS_MODULE;
> > > > -	config.add_legacy_fixed_of_cells = true;
> > > > +	config.add_legacy_fixed_of_cells = !mtd_type_is_nand(mtd);
> > > >   	config.type = NVMEM_TYPE_OTP;
> > > >   	config.root_only = true;
> > > >   	config.ignore_wp = true;
> > > 
> > > I think there may be even more unwanted behaviour here. If
> > > mtd_otp_nvmem_register() fails to find node with "user-otp" /
> > > "factory-otp" compatible then it sets "config.of_node" to NULL but
> > > that
> > > means NVMEM core still looks for NVMEM cells in device's "of_node".
> > > 
> > > I believe we should not look for OTP NVMEM cells out of the
> > > "user-otp" /
> > > "factory-otp" compatible nodes.
> > > 
> > > So maybe what we need in the first place is just:
> > > config.add_legacy_fixed_of_cells = !!np;
> > > ?
> > > 
> > > Any extra limitation of .add_legacy_fixed_of_cells should probably be
> > > used only if we want to prevent new users of the legacy syntax. The
> > > problem is that mtd.yaml binding allowed "user-otp" and "factory-otp"
> > > with old syntax cells. It means every MTD device was allowed to have
> > > them.
> > > 
> > > No in-kernel DTS even used "user-otp" or "factory-otp" with NVMEM
> > > legacy
> > > cells but I'm not sure about downstream DTS files. Ideally we would do
> > > config.add_legacy_fixed_of_cells = false;
> > > but that could break compatibility with some downstream DTS files.
> > 
> > Yes the main problem is prevent regression in downstream. I feel for the
> > nand usage, this is 100% of the times broken. For SPI and other corner
> > case MTD devices it's not?
> > 
> > Anyway did you by chance have a suggestion for a better fixes tag?
> 
> My personal idea for that would be to put two Fixes with two commits and
> describe in commit body that one just exposed existing bug.
> 
> You may check my OpenWrt quick patch for an idea how I'd handle that:
> https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/generic/pending-6.6/440-mtd-don-t-look-for-OTP-legacy-NVMEM-cells-if-proper-.patch;h=d9d15a4048c144d8565c8ea38e15a79f7f4a5fe1;hb=dd78a59cd7b029560b33cb3ac0e1aa8b747bd807
>

My concern is that using !!np might pose some regression problem. Also I
feel adding the macronix commit in fixes tag might be confusing?

Think I will just use the nand check just to be extra safe and add a
kernel dependency for when the add_legacy_fixed_of_cells was introduced
since before that a different patch is needed. What do you think?

-- 
	Ansuel

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/