[PATCH 06/10] drm/i915/spi: spi register with mtd

Usyskin, Alexander alexander.usyskin at intel.com
Wed Feb 14 12:16:00 UTC 2024


Hi Miquel

Intel Gfx in infinite wisdom decided to create another driver - Xe and
the spi driver part of this series should be moved to some common location.
Should it be drivers/mtd or drivers/spi, or some other place?

-- 
Thanks,
Sasha



> 
> Hi Alexander,
> 
> + Michael and Tudor
> 
> Folks, any interesting thought about the below discussion?
> 
> alexander.usyskin at intel.com wrote on Tue, 14 Nov 2023 08:47:34 +0000:
> 
> > >
> > > > > > > > +	spi->mtd.writesize = SZ_1; /* 1 byte granularity */
> > > > > > >
> > > > > > > You say writesize should be aligned with 4 in your next patch?
> > > > > >
> > > > > > We support unaligned write by reading aligned 4bytes,
> > > > > > replacing changed bytes there and writing whole 4bytes back.
> > > > > > Is there any problem with this approach?
> > > > >
> > > > > Is there a reason to do that manually rather than letting the core
> > > > > handle the complexity?
> > > > >
> > > > I was not aware that core can do this. The core implements above logic
> > > > if I put SZ_4 here and caller try to write, say, one byte?
> > > > And sync multiple writers?
> > > > If so, I can remove manual work, I think, and make the patches smaller.
> > >
> > > I haven't checked in detail but I would expect this yes. Please have a
> > > round of tests and if it works, please simplify this part.
> > >
> > > Thanks,
> > > Miquèl
> >
> > When I put SZ_4 here the "mtd_debug info /dev/mtd0" prints "mtd.writesize =
> 4",
> > but "mtd_debug write /dev/mtd0 128 1 c" passes one byte to
> > i915_spi_write (mtd._write callback).
> > I suppose that mtd subsystem do not support this.
> > Or I did something wrong?
> >
> 
> 
> Thanks,
> Miquèl


More information about the Intel-gfx mailing list