Re: [PATCH] PCI: dwc: Fix writing wrong value if snps,enable-cdm-check

[Bjorn Helgaas <helgaas@kernel.org>]

On Thu, Feb 16, 2023 at 06:20:12PM +0900, Yoshihiro Shimoda wrote:
> The "val" of PCIE_PORT_LINK_CONTROL will be reused on the
> "Set the number of lanes". But, if snps,enable-cdm-check" exists,
> the "val" will be set to PCIE_PL_CHK_REG_CONTROL_STATUS.
> Therefore, unexpected register value is possible to be used
> to PCIE_PORT_LINK_CONTROL register if snps,enable-cdm-check" exists.
> So, read PCIE_PORT_LINK_CONTROL register again to fix the issue.
> 
> Fixes: ec7b952f453c ("PCI: dwc: Always enable CDM check if "snps,enable-cdm-check" exists")
> Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
> ---
>  drivers/pci/controller/dwc/pcie-designware.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/pci/controller/dwc/pcie-designware.c b/drivers/pci/controller/dwc/pcie-designware.c
> index 6d5d619ab2e9..3bb9ca14fb9c 100644
> --- a/drivers/pci/controller/dwc/pcie-designware.c
> +++ b/drivers/pci/controller/dwc/pcie-designware.c
> @@ -824,6 +824,7 @@ void dw_pcie_setup(struct dw_pcie *pci)
>  	}
>  
>  	/* Set the number of lanes */
> +	val = dw_pcie_readl_dbi(pci, PCIE_PORT_LINK_CONTROL);

Definitely a bug, thanks for the fix and the Fixes: tag.

But I would like the whole function better if it could be structured
so we read PCIE_PORT_LINK_CONTROL once and wrote it once.  And the
same for PCIE_LINK_WIDTH_SPEED_CONTROL.

Maybe there's a reason PCIE_PL_CHK_REG_CONTROL_STATUS must be written
between the two PCIE_PORT_LINK_CONTROL writes or the two
PCIE_LINK_WIDTH_SPEED_CONTROL writes, I dunno.  If so, a comment there
about why that is would be helpful.

>  	val &= ~PORT_LINK_FAST_LINK_MODE;
>  	val &= ~PORT_LINK_MODE_MASK;
>  	switch (pci->num_lanes) {
> -- 
> 2.25.1
>