Re: [RFC/RFT PATCH v1 0/9] mtd: fsl: quadspi: Fixes for fsl-quadspi.c driver (vybrid HW)

From: Lukasz Majewski <lukma@denx.de>
To: Boris Brezillon <boris.brezillon@bootlin.com>
Cc: Fabio Estevam <festevam@gmail.com>,
	prabhakar.kushwaha@nxp.com, han.xu@nxp.com,
	Mark Rutland <mark.rutland@arm.com>,
	yogeshnarayan.gaur@nxp.com, richard@nod.at,
	linux-kernel@vger.kernel.org, Stefan Agner <stefan@agner.ch>,
	linux-spi@vger.kernel.org, albert.aribaud@3adev.fr,
	Frieder Schrempf <frieder.schrempf@exceet.de>,
	broonie@kernel.org, linux-mtd@lists.infradead.org,
	Fabio Estevam <fabio.estevam@nxp.com>,
	david.wolfe@nxp.com, computersforpeace@gmail.com,
	dwmw2@infradead.org
Subject: Re: [RFC/RFT PATCH v1 0/9] mtd: fsl: quadspi: Fixes for fsl-quadspi.c driver (vybrid HW)
Date: Sun, 30 Sep 2018 18:22:20 +0200	[thread overview]
Message-ID: <20180930182220.30d0ccf3@jawa> (raw)
In-Reply-To: <20180930073927.001c16d2@bbrezillon>

[-- Attachment #1: Type: text/plain, Size: 2387 bytes --]

Hi Boris,

> Hi Lukasz,
> 
> On Sat, 29 Sep 2018 23:02:40 +0200
> Lukasz Majewski <lukma@denx.de> wrote:
> 
> > > Talking about that, can you try to port your fixes on top of
> > > Frieder's patchset? I'm pretty sure some bug fixes are irrelevant
> > > after the migration to spi-mem (patch 1, 3, 4, 5, 6, 7 and 9
> > > should be dropped I think).    
> > 
> > The problem is that Frieder's patch is using IP command mode for
> > transfers smaller than AHB RX fifo size.
> > This according to the comment in the current driver is broken in HW
> > for Vybrid (so this is a regression).  
> 
> Why not fixing that in the new driver?

I did not say that I will not fix it in the new driver :-)

The problem is that I'm waiting for NXP's community reply - to get some
more info regarding the bug.

> 
> > 
> > I'm also wondering if other users of vf610 based boards experience
> > issues with QSPI?  
> 
> Are you sure the 4 I/O lines are wired on your design? 

This is a somewhat special case.

There are two identical SPI-NOR memories: one with QUAD lines connected
and second only for DUAL due to HW design decision.

> Anyway, if it's
> a bug that only hurts vf610, you should mask quad modes in ->hwcaps
> (or patch fsl_qspi_supports_op() in the new driver), not change the
> SPI NOR definition.

Yes, this was also pointed out by Cyrille - and yes, I do agree that I
shouldn't mask it.

> 
> > 
> > In my case, after running the UBI/UBIFS tests (on the original and
> > new driver without those "fixes") I cannot mount the volume after
> > creation as the header is wrongly read.  
> 
> I'm not denying this fact, I'm just saying, now that you've found
> where the issue comes from, you can also port the fix to the new
> driver.

Yes, as the old driver is now in a "good enough" shape (though I don't
know the exact HW bug reason) - the code can be ported to the new
driver.

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

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd@denx.de

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]