* Re: [PATCH] mtd: fsl-quadspi: Distinguish the mtd device names
2018-01-10 22:17 ` Boris Brezillon
@ 2018-01-10 22:39 ` Fabio Estevam
2018-01-10 22:42 ` Boris Brezillon
2018-01-10 22:54 ` Boris Brezillon
2 siblings, 0 replies; 6+ messages in thread
From: Fabio Estevam @ 2018-01-10 22:39 UTC (permalink / raw)
To: Boris Brezillon
Cc: Han Xu, Fabio Estevam, David Wolfe, Frank Li,
linux-mtd@lists.infradead.org, cyrille.pitchen@wedev4u.fr
Hi Boris,
On Wed, Jan 10, 2018 at 8:17 PM, Boris Brezillon
<boris.brezillon@free-electrons.com> wrote:
> With the alternative naming scheme suggested above, it gives:
>
> mtd->name = devm_kasprintf(dev, GFP_KERNEL, "%s:%d"
> dev_name(&pdev->dev),
> mtd_index);
Yes, this looks better.
I have made the suggested change. Will test it tomorrow on my board
and will send a v2.
Thanks!
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] mtd: fsl-quadspi: Distinguish the mtd device names
2018-01-10 22:17 ` Boris Brezillon
2018-01-10 22:39 ` Fabio Estevam
@ 2018-01-10 22:42 ` Boris Brezillon
2018-01-10 22:54 ` Boris Brezillon
2 siblings, 0 replies; 6+ messages in thread
From: Boris Brezillon @ 2018-01-10 22:42 UTC (permalink / raw)
To: Han Xu
Cc: Frank Li, linux-mtd@lists.infradead.org,
cyrille.pitchen@wedev4u.fr, Fabio Estevam, David Wolfe,
Fabio Estevam
On Wed, 10 Jan 2018 23:17:19 +0100
Boris Brezillon <boris.brezillon@free-electrons.com> wrote:
> On Mon, 8 Jan 2018 17:13:17 +0000
> Han Xu <han.xu@nxp.com> wrote:
>
> > ________________________________________
> > From: Fabio Estevam <festevam@gmail.com>
> > Sent: Friday, January 5, 2018 1:46 PM
> > To: Han Xu
> > Cc: cyrille.pitchen@wedev4u.fr; David Wolfe; Frank Li; linux-mtd@lists.infradead.org; Fabio Estevam
> > Subject: [PATCH] mtd: fsl-quadspi: Distinguish the mtd device names
> >
> > From: Fabio Estevam <fabio.estevam@nxp.com>
> >
> > Currently on a imx6sx-sdb board, which has two SPI NOR chips connected
> > to QSPI2 the following output from /proc/mtd is seen:
> >
> > # cat /proc/mtd
> > dev: size erasesize name
> > mtd0: 01000000 00010000 "21e4000.qspi"
> > mtd1: 01000000 00010000 "21e4000.qspi"
> >
> > Attempts to partition them on the kernel command line result in both
> > chips with identical (and identically named) partitions, which is
> > an inconvenient behavior.
> >
> > Assign a different mtd->name for each mtd device to avoid this problem.
> >
> > After this change the output from /proc/mtd becomes:
> >
> > # cat /proc/mtd
> > dev: size erasesize name
> > mtd0: 01000000 00010000 "fsl-quadspi.0"
> > mtd1: 01000000 00010000 "fsl-quadspi.1"
>
> What happens if you end up having 2 qspi IPs in the same SoC? IMHO,
> it's more future proof to have something like
> "<reg-addr>.qspi:<chipidx>" or "<reg-addr>.qspi:<chipidx>".
Hm, ':' is not a good separator because it's already used by mtdparts,
so I suggest using '-' ("<reg-addr>.qspi-<chipid>").
>
> >
> > Reported-by: David Wolfe <david.wolfe@nxp.com>
> > Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
> > ---
> > drivers/mtd/spi-nor/fsl-quadspi.c | 10 +++++++++-
> > 1 file changed, 9 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/mtd/spi-nor/fsl-quadspi.c b/drivers/mtd/spi-nor/fsl-quadspi.c
> > index f17d224..7712c63 100644
> > --- a/drivers/mtd/spi-nor/fsl-quadspi.c
> > +++ b/drivers/mtd/spi-nor/fsl-quadspi.c
> > @@ -967,7 +967,7 @@ static int fsl_qspi_probe(struct platform_device *pdev)
> > struct resource *res;
> > struct spi_nor *nor;
> > struct mtd_info *mtd;
> > - int ret, i = 0;
> > + int ret, i = 0, mtd_index = 0;
>
> It's not the MTD index if you have other MTD devices. How about
> spiflash_idx?
>
> >
> > q = devm_kzalloc(dev, sizeof(*q), GFP_KERNEL);
> > if (!q)
> > @@ -1047,6 +1047,13 @@ static int fsl_qspi_probe(struct platform_device *pdev)
> > nor = &q->nor[i];
> > mtd = &nor->mtd;
> >
> > + mtd->name = devm_kasprintf(dev, GFP_KERNEL, "fsl-quadspi.%d",
> > + mtd_index);
>
> With the alternative naming scheme suggested above, it gives:
>
> mtd->name = devm_kasprintf(dev, GFP_KERNEL, "%s:%d"
> dev_name(&pdev->dev),
> mtd_index);
I realize your index depends on the order the nodes have been defined
in the DT, which is fragile. You should find a better solution to
uniquely identify each device connected to the fsl-qspi controller
(<busid>.<devid>?) and then have a property describing that (reg?).
>
> > + if (!mtd->name) {
> > + ret = -ENOMEM;
> > + goto mutex_failed;
> > + }
> > +
> > nor->dev = dev;
> > spi_nor_set_flash_node(nor, np);
> > nor->priv = q;
> > @@ -1098,6 +1105,7 @@ static int fsl_qspi_probe(struct platform_device *pdev)
> > nor->page_size = q->devtype_data->txfifo;
> >
> > i++;
> > + mtd_index++;
> > }
> >
> > /* finish the rest init. */
> > --
> > 2.7.4
> >
> > Acked-by: Han Xu <han.xu@nxp.com>
> >
> > ______________________________________________________
> > Linux MTD discussion mailing list
> > http://lists.infradead.org/mailman/listinfo/linux-mtd/
>
>
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] mtd: fsl-quadspi: Distinguish the mtd device names
2018-01-10 22:17 ` Boris Brezillon
2018-01-10 22:39 ` Fabio Estevam
2018-01-10 22:42 ` Boris Brezillon
@ 2018-01-10 22:54 ` Boris Brezillon
2 siblings, 0 replies; 6+ messages in thread
From: Boris Brezillon @ 2018-01-10 22:54 UTC (permalink / raw)
To: Han Xu
Cc: Frank Li, linux-mtd@lists.infradead.org,
cyrille.pitchen@wedev4u.fr, Fabio Estevam, David Wolfe,
Fabio Estevam
On Wed, 10 Jan 2018 23:17:19 +0100
Boris Brezillon <boris.brezillon@free-electrons.com> wrote:
> On Mon, 8 Jan 2018 17:13:17 +0000
> Han Xu <han.xu@nxp.com> wrote:
>
> > ________________________________________
> > From: Fabio Estevam <festevam@gmail.com>
> > Sent: Friday, January 5, 2018 1:46 PM
> > To: Han Xu
> > Cc: cyrille.pitchen@wedev4u.fr; David Wolfe; Frank Li; linux-mtd@lists.infradead.org; Fabio Estevam
> > Subject: [PATCH] mtd: fsl-quadspi: Distinguish the mtd device names
> >
> > From: Fabio Estevam <fabio.estevam@nxp.com>
> >
> > Currently on a imx6sx-sdb board, which has two SPI NOR chips connected
> > to QSPI2 the following output from /proc/mtd is seen:
> >
> > # cat /proc/mtd
> > dev: size erasesize name
> > mtd0: 01000000 00010000 "21e4000.qspi"
> > mtd1: 01000000 00010000 "21e4000.qspi"
> >
> > Attempts to partition them on the kernel command line result in both
> > chips with identical (and identically named) partitions, which is
> > an inconvenient behavior.
> >
> > Assign a different mtd->name for each mtd device to avoid this problem.
> >
> > After this change the output from /proc/mtd becomes:
> >
> > # cat /proc/mtd
> > dev: size erasesize name
> > mtd0: 01000000 00010000 "fsl-quadspi.0"
> > mtd1: 01000000 00010000 "fsl-quadspi.1"
>
> What happens if you end up having 2 qspi IPs in the same SoC? IMHO,
> it's more future proof to have something like
> "<reg-addr>.qspi:<chipidx>" or "<reg-addr>.qspi:<chipidx>".
>
> >
> > Reported-by: David Wolfe <david.wolfe@nxp.com>
> > Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
> > ---
> > drivers/mtd/spi-nor/fsl-quadspi.c | 10 +++++++++-
> > 1 file changed, 9 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/mtd/spi-nor/fsl-quadspi.c b/drivers/mtd/spi-nor/fsl-quadspi.c
> > index f17d224..7712c63 100644
> > --- a/drivers/mtd/spi-nor/fsl-quadspi.c
> > +++ b/drivers/mtd/spi-nor/fsl-quadspi.c
> > @@ -967,7 +967,7 @@ static int fsl_qspi_probe(struct platform_device *pdev)
> > struct resource *res;
> > struct spi_nor *nor;
> > struct mtd_info *mtd;
> > - int ret, i = 0;
> > + int ret, i = 0, mtd_index = 0;
>
> It's not the MTD index if you have other MTD devices. How about
> spiflash_idx?
>
> >
> > q = devm_kzalloc(dev, sizeof(*q), GFP_KERNEL);
> > if (!q)
> > @@ -1047,6 +1047,13 @@ static int fsl_qspi_probe(struct platform_device *pdev)
> > nor = &q->nor[i];
> > mtd = &nor->mtd;
> >
> > + mtd->name = devm_kasprintf(dev, GFP_KERNEL, "fsl-quadspi.%d",
> > + mtd_index);
>
> With the alternative naming scheme suggested above, it gives:
>
> mtd->name = devm_kasprintf(dev, GFP_KERNEL, "%s:%d"
> dev_name(&pdev->dev),
> mtd_index);
>
> > + if (!mtd->name) {
> > + ret = -ENOMEM;
> > + goto mutex_failed;
> > + }
> > +
> > nor->dev = dev;
> > spi_nor_set_flash_node(nor, np);
One, last thing: there's a generic mechanism using the value of the
'label' property for mtd->name, which is particularly useful if one
wants to assign user-friendly names to his flashes. This is done in
spi_nor_set_flash_node() (more precisely in mtd_set_of_node() which is
called by spi_nor_set_flash_node()), but it only works if mtd->name is
NULL. So I recommend move the mtd->name assignment here, and making it
conditional:
if (!mtd->name) {
mtd->name = devm_kasprintf(...);
...
}
> > nor->priv = q;
> > @@ -1098,6 +1105,7 @@ static int fsl_qspi_probe(struct platform_device *pdev)
> > nor->page_size = q->devtype_data->txfifo;
> >
> > i++;
> > + mtd_index++;
> > }
> >
> > /* finish the rest init. */
> > --
> > 2.7.4
> >
> > Acked-by: Han Xu <han.xu@nxp.com>
> >
> > ______________________________________________________
> > Linux MTD discussion mailing list
> > http://lists.infradead.org/mailman/listinfo/linux-mtd/
>
>
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/
^ permalink raw reply [flat|nested] 6+ messages in thread