lkml.org 
[lkml]   [2020]   [Aug]   [22]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] spi: spi-fsl-dspi: Add ACPI support
On Sat, Aug 22, 2020 at 07:37:25PM +0530, Kuldip Dwivedi wrote:
> > -----Original Message-----
> > From: Mark Brown <broonie@kernel.org>
> > Sent: Friday, August 21, 2020 7:37 PM
> > To: kuldip dwivedi <kuldip.dwivedi@puresoftware.com>
> > Cc: linux-spi@vger.kernel.org; linux-kernel@vger.kernel.org; Qiang Zhao
> > <qiang.zhao@nxp.com>; Pankaj Bansal <pankaj.bansal@nxp.com>; Varun Sethi
> > <V.Sethi@nxp.com>; tanveer <tanveer.alam@puresoftware.com>
> > Subject: Re: [PATCH] spi: spi-fsl-dspi: Add ACPI support
> >
> > On Fri, Aug 21, 2020 at 06:40:29PM +0530, kuldip dwivedi wrote:
> >
> > > +static const struct acpi_device_id fsl_dspi_acpi_ids[] = {
> > > + { "NXP0005", .driver_data =
> (kernel_ulong_t)&devtype_data[LS2085A], },
> > > + {},
> > > +};
> > > +MODULE_DEVICE_TABLE(acpi, fsl_dspi_acpi_ids);
> >
> > Does NXP know about this ID assignment from their namespace? ACPI
> > IDs should be namespaced by whoever's assigning the ID to avoid
> > collisions.
> Yes, I got HID from NXP only.
> >
> > > - ret = of_property_read_u32(np, "spi-num-chipselects",
> &cs_num);
> > > + if (is_acpi_node(pdev->dev.fwnode))
> > > + ret = device_property_read_u32(&pdev->dev,
> > > + "spi-num-chipselects", &cs_num);
> > > + else
> > > + ret = of_property_read_u32(np,
> > > + "spi-num-chipselects", &cs_num);
> >
> > The whole point with the device property API is that it works with
> > both DT and ACPI without needing separate parsing, though in this
> > case I'm wondering why we'd need to specify this in an ACPI system
> > at all?
> Understood. Will take care in v2 PATCH
> >

IMO there is zero reason for the existence of the "spi-num-chipselects"
property even for DT. We should deprecate it (start ignoring it in
existing device tree deployments) and populate struct
fsl_dspi_devtype_data with that info based on SoC compatible string.

> > > - of_property_read_u32(np, "bus-num", &bus_num);
> > > + if (is_acpi_node(pdev->dev.fwnode)) {
> > > + ret = device_property_read_u32(&pdev->dev,
> > > + "bus-num",
> &bus_num);
> >
> > This is a bad idea for DT and I can't understand why you'd carry it
> > over for ACPI - why would an ACPI system ever care about this? It's
> > Linux internal at the best of times.
> Will take care in v2 PATCH

Yes, definitely bloatware from the old days. I think this driver needs
the existing device tree bindings rethought a little bit before
mindlessly porting them to ACPI.

> >
> > It looks like you've done this by simply adding these device
> > property alternatives for every DT property. This isn't how that
> > API is intended to be used and suggests that this isn't a thought
> > through, idiomatic ACPI binding. I'd suggest looking at the
> > Synquacer driver for an example of how this would normally be done,
> > I'd expect your ACPI code to look very much like theirs.

Speaking of which, on what SPI peripherals was this tested?
I am not sure how other controllers deal with this, but DSPI has, by
default, no CS-to-SCK and a SCK-to-CS delays. Those must be explicitly
requested through the custom "fsl,spi-cs-sck-delay" and
"fsl,spi-sck-cs-delay" DT bindings for each individual SPI peripheral.
Some peripherals just don't work when the CS-to-SCK and SCK-to-CS delays
are zero, and I don't see the ACPI variant of the driver attempting to
read those properties, hence the question.

Thanks,
-Vladimir

\
 
 \ /
  Last update: 2020-08-22 17:22    [W:0.088 / U:25.080 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site