firmware: xilinx: Clear IOCTL_SET_SD_TAPDELAY using PM_MMIO_WRITE
authorMarek Vasut <marex@denx.de>
Thu, 15 Dec 2022 15:20:23 +0000 (16:20 +0100)
committerMichal Simek <michal.simek@amd.com>
Thu, 5 Jan 2023 09:18:07 +0000 (10:18 +0100)
commita4b2e6063cfeaab1160501acfb27e8618a7c693f
treefa849d688d301834a7e2410b87efba1ef64b8ae9
parentfcc2f972f948d8bd5f8f149617d821b0ae3ad5fd
firmware: xilinx: Clear IOCTL_SET_SD_TAPDELAY using PM_MMIO_WRITE

In case the tap delay required by Arasan SDHCI is set to 0, the current
embeddedsw firmware unconditionally writes IOU_SLCR SD_ITAPDLY to 0x100
(SD0_ITAPDLYENA=1, SD0_ITAPDLYSEL=0). Previous behavior was to keep the
IOU_SLCR SD_ITAPDLY set to 0x0. There is some sort of difference in the
behavior between SD0_ITAPDLYENA=1/0 with the same SD0_ITAPDLYSEL=0, even
though the behavior should be identical -- zero delay added to rxclk_in
line. The former breaks HS200 training in low temperature conditions.

Write IOU_SLCR SD_ITAPDLY register to 0 using PM_MMIO_WRITE which seem
to allow unrestricted WRITE access (and PM_MMIO_READ which allows read
access) to the entire address space. This way, it is possible to work
around the defect in IOCTL_SET_SD_TAPDELAY design which does not permit
clearing SDx_ITAPDLYENA bit.

Note that the embeddedsw firmware does not permit clearing the SD_ITAPDLY
SD0_ITAPDLYENA bit, this bit can only ever be set by the firmware and it
is often impossible to update the possibly broken firmware.

Signed-off-by: Marek Vasut <marex@denx.de>
Link: https://lore.kernel.org/r/20221215152023.8387-1-marex@denx.de
Signed-off-by: Michal Simek <michal.simek@amd.com>
drivers/firmware/xilinx/zynqmp.c
include/linux/firmware/xlnx-zynqmp.h