[AMD Official Use Only - AMD Internal Distribution Only]
> -----Original Message-----
> From: Marek Vasut <[email protected]>
> Sent: Friday, February 21, 2025 11:28 PM
> To: Abbarapu, Venkatesh <[email protected]>; u-
> [email protected]; [email protected]; [email protected]
> Cc: Simek, Michal <[email protected]>; [email protected];
> [email protected]; [email protected]; [email protected];
> [email protected]; [email protected]; [email protected];
> [email protected]; [email protected];
> [email protected]; [email protected]; p-
> [email protected]; git (AMD-Xilinx) <[email protected]>; Kummari, Prasad
> <[email protected]>
> Subject: Re: [UBOOT PATCH] spi: spi-uclass: Fix reduced read speed when
> parallel config is enabled
>
> Caution: This message originated from an External Source. Use proper
> caution when opening attachments, clicking links, or responding.
>
>
> On 2/21/25 10:57 AM, Venkatesh Yadav Abbarapu wrote:
> > From: Prasad Kummari <[email protected]>
> >
> > When SPI_STACKED_PARALLEL is enabled, the chip selects properties are
> > read from the device tree using the reg property and stored in the
> > plat->cs[] array. However, the function dev_read_u32_array() returns
> > zero, causing the execution to enter the else block, print an error
> > message, and return. As a result, the spi-max-frequency property is
> > not read from the device tree, resulting in the use of the default
> > frequency SPI_DEFAULT_SPEED_HZ (100,000). This leads to a lower
> > spi->speed, reducing QSPI DMA read performance.
> >
> > To address this issue, an additional condition has been added to check
> > the return value of reg. If it is nonzero, the function prints an
> > error message and returns. Otherwise, it proceeds to read the
> > spi-max-frequency property from the device tree.
> >
> > Signed-off-by: Prasad Kummari <[email protected]>
> > Signed-off-by: Venkatesh Yadav Abbarapu
> <[email protected]>
> > ---
> > drivers/spi/spi-uclass.c | 6 ++++--
> > 1 file changed, 4 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/spi/spi-uclass.c b/drivers/spi/spi-uclass.c index
> > d6049753740..a7783b6b6dd 100644
> > --- a/drivers/spi/spi-uclass.c
> > +++ b/drivers/spi/spi-uclass.c
> > @@ -523,8 +523,10 @@ int spi_slave_of_to_plat(struct udevice *dev, struct
> dm_spi_slave_plat *plat)
> > if (ret == -EOVERFLOW || ret == -FDT_ERR_BADLAYOUT) {
> > dev_read_u32(dev, "reg", &plat->cs[0]);
> > } else {
> > - dev_err(dev, "has no valid 'reg' property (%d)\n", ret);
> > - return ret;
> > + if (ret) {
> > + dev_err(dev, "has no valid 'reg' property (%d)\n",
> > ret);
> > + return ret;
> > + }
> > }
> > #else
> > plat->cs[0] = dev_read_u32_default(dev, "reg", -1);
>
> Why not simply do this?
Sure, will update!
>
> diff --git a/drivers/spi/spi-uclass.c b/drivers/spi/spi-uclass.c index
> d6049753740..4d9376dbdca 100644
> --- a/drivers/spi/spi-uclass.c
> +++ b/drivers/spi/spi-uclass.c
> @@ -522,7 +522,7 @@ int spi_slave_of_to_plat(struct udevice *dev, struct
> dm_spi_slave_plat *plat)
>
> if (ret == -EOVERFLOW || ret == -FDT_ERR_BADLAYOUT) {
> dev_read_u32(dev, "reg", &plat->cs[0]);
> - } else {
> + } else if (ret) {
> dev_err(dev, "has no valid 'reg' property (%d)\n", ret);
> return ret;
> }
