On Thu, Jan 30, 2025 at 04:53:14PM +0200, Eugen Hristev wrote:
>
>
> On 1/30/25 16:49, Tom Rini wrote:
> > On Thu, Jan 30, 2025 at 04:07:27PM +0200, Eugen Hristev wrote:
> >>
> >>
> >> On 1/23/25 14:12, Alexander Dahl wrote:
> >>> Hello everyone,
> >>>
> >>> this patch series is for U-Boot, although Linux might be affected.
> >>>
> >>> The quadspi controller on microchip (former atmel) sama5d2, sam9x60 and
> >>> other soc variants of the at91 family is usually used in hardware
> >>> accelerated spi-mem mode. It also supports a classic mode for
> >>> interfacing all kinds of SPI peripherals like ADCs, IO expanders,
> >>> displays, you name it. However the drivers in Linux and U-Boot did not
> >>> support that mode, yet.
> >>>
> >>> While spi-mem subsystem interface is comparable between Linux and
> >>> U-Boot, SPI subsystem interface is rather different in Linux and U-Boot.
> >>> So there's no previously written Linux support just ported here. I
> >>> tried that, but the simple implementation for U-Boot's .xfer() used here
> >>> does not work for Linux' .transfer_one() because of the OVRES flag set
> >>> all the time. I guess a more sophisticated approach has to be taken in
> >>> Linux with interrupts, completions and/or dma?
> >>>
> >>> According to commit 24c8ff4684c5 ("spi: Add Atmel QuadSPI driver")
> >>> message, the initial U-Boot variant of this driver was ported from Linux
> >>> v5.1-rc5 back then. I hope it's okay to squash the fixes added to Linux
> >>> after that into patch 2 and 3 of this series?
> >>>
> >>> Patch 4 and 5 come from my efforts to hack this feature into the Linux
> >>> driver. (Not sure if those should be added to the Linux driver
> >>> anyways, even if nobody ever implements transfer_one() over there?)
> >>>
> >>> Patch 5 is the actual work, very much inspired by the current U-Boot
> >>> driver atmel_spi for the generic SPI controllers on at91 family.
> >>>
> >>> Patch 1 is copied from my other spi series for U-Boot, just added it
> >>> here for better build results.
> >>>
> >>> Patch 6 is optional, because not portable towards Linux.
> >>>
> >>> Tested on top of v2024.10, because v2025.01 won't build as debug
> >>> for sam9x60 as described in another thread. But there are almost no
> >>> changes related to at91 socs, boards, and drivers in the last months
> >>> here, so it should be comparable.
> >>>
> >>> Greets
> >>> Alex
> >>>
> >>> (Cc-ing the contributors of the Linux fixes here.)
> >>> Cc: Tudor Ambarus <tudor.amba...@microchip.com>
> >>> Cc: Yoshitaka Ikeda <ik...@nskint.co.jp>
> >>> Cc: Csókás Bence <csokas.be...@prolan.hu>
> >>>
> >>> Alexander Dahl (7):
> >>> spi: atmel-quadspi: Depend on SPI_MEM
> >>> spi: atmel-quadspi: Port collected fixes from Linux v5.10 and v5.15
> >>> spi: atmel-quadspi: Avoid overwriting MR register settings
> >>> spi: atmel-quadspi: Remove default mode setting at probe time
> >>> spi: atmel-quadspi: Allow setting SMM to classic SPI mode
> >>> spi: atmel-quadspi: Add support for classic SPI mode
> >>> spi: atmel-quadspi: Improve probe debugging
> >>>
> >>> drivers/spi/Kconfig | 2 +-
> >>> drivers/spi/atmel-quadspi.c | 294 ++++++++++++++++++++++++++++--------
> >>> 2 files changed, 228 insertions(+), 68 deletions(-)
> >>>
> >>>
> >>> base-commit: 2eed5a1ff36217372e19f7513bd07077fc76718a
> >>
> >> Hello Alex,
> >>
> >> I had a quick look over your patches, but it would require to get a
> >> review from SPI maintainers to apply these.
> >
> > Our SPI maintainer is currently not very responsive. Would you be OK
> > taking this via the atmel tree since it's an atmel platform driver and
> > not core changes? Thanks!
> >
>
> Yes, that is fine. However the review would be to the best of my ability
> only.That's fine, thanks! -- Tom
signature.asc
Description: PGP signature
