Re: [PATCH v5 5/5] PCI: mediatek: add support for Airoha AN7583 SoC

[Christian Marangi <ansuelsmth@gmail.com>]

On Sun, Oct 19, 2025 at 01:01:44PM +0530, Manivannan Sadhasivam wrote:
> On Sun, Oct 12, 2025 at 10:56:59PM +0200, Christian Marangi wrote:
> > Add support for the second PCIe Root Complex present on Airoha AN7583
> > SoC.
> > 
> > This is based on the Mediatek Gen1/2 PCIe driver and similar to Gen3
> > also require workaround for the reset signals.
> > 
> > Introduce a new flag to skip having to reset signals and also introduce
> > some additional logic to configure the PBUS registers required for
> > Airoha SoC.
> > 
> > While at it, also add additional info on the PERST# Signal delay
> > comments and use dedicated macro.
> > 
> 
> This belongs to a separate patch which should come before this one.
> 
> > Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
> > ---
> >  drivers/pci/controller/pcie-mediatek.c | 92 ++++++++++++++++++++------
> >  1 file changed, 70 insertions(+), 22 deletions(-)
> > 
> > diff --git a/drivers/pci/controller/pcie-mediatek.c b/drivers/pci/controller/pcie-mediatek.c
> > index 1678461e56d3..3340c005da4b 100644
> > --- a/drivers/pci/controller/pcie-mediatek.c
> > +++ b/drivers/pci/controller/pcie-mediatek.c
> > @@ -148,6 +148,7 @@ enum mtk_pcie_flags {
> >  	NO_MSI = BIT(2), /* Bridge has no MSI support, and relies on an
> >  			  * external block
> >  			  */
> > +	SKIP_PCIE_RSTB	= BIT(3), /* Skip calling RSTB bits on PCIe probe */
> >  };
> >  
> >  /**
> > @@ -684,28 +685,32 @@ static int mtk_pcie_startup_port_v2(struct mtk_pcie_port *port)
> >  		regmap_update_bits(pcie->cfg, PCIE_SYS_CFG_V2, val, val);
> >  	}
> >  
> > -	/* Assert all reset signals */
> > -	writel(0, port->base + PCIE_RST_CTRL);
> > -
> > -	/*
> > -	 * Enable PCIe link down reset, if link status changed from link up to
> > -	 * link down, this will reset MAC control registers and configuration
> > -	 * space.
> > -	 */
> > -	writel(PCIE_LINKDOWN_RST_EN, port->base + PCIE_RST_CTRL);
> > -
> > -	/*
> > -	 * Described in PCIe CEM specification sections 2.2 (PERST# Signal) and
> > -	 * 2.2.1 (Initial Power-Up (G3 to S0)). The deassertion of PERST# should
> > -	 * be delayed 100ms (TPVPERL) for the power and clock to become stable.
> > -	 */
> > -	msleep(100);
> > -
> > -	/* De-assert PHY, PE, PIPE, MAC and configuration reset	*/
> > -	val = readl(port->base + PCIE_RST_CTRL);
> > -	val |= PCIE_PHY_RSTB | PCIE_PERSTB | PCIE_PIPE_SRSTB |
> > -	       PCIE_MAC_SRSTB | PCIE_CRSTB;
> > -	writel(val, port->base + PCIE_RST_CTRL);
> > +	if (!(soc->flags & SKIP_PCIE_RSTB)) {
> > +		/* Assert all reset signals */
> > +		writel(0, port->base + PCIE_RST_CTRL);
> > +
> > +		/*
> > +		 * Enable PCIe link down reset, if link status changed from
> > +		 * link up to link down, this will reset MAC control registers
> > +		 * and configuration space.
> > +		 */
> > +		writel(PCIE_LINKDOWN_RST_EN, port->base + PCIE_RST_CTRL);
> > +
> > +		/*
> > +		 * Described in PCIe CEM specification revision 3.0 sections
> > +		 * 2.2 (PERST# Signal) and 2.2.1 (Initial Power-Up (G3 to S0)).
> > +		 *
> > +		 * The deassertion of PERST# should be delayed 100ms (TPVPERL)
> > +		 * for the power and clock to become stable.
> 
> You can drop the comments since PCIE_T_PVPERL_MS definition has them.
> 
> > +		 */
> > +		msleep(PCIE_T_PVPERL_MS);
> > +
> > +		/* De-assert PHY, PE, PIPE, MAC and configuration reset	*/
> > +		val = readl(port->base + PCIE_RST_CTRL);
> > +		val |= PCIE_PHY_RSTB | PCIE_PERSTB | PCIE_PIPE_SRSTB |
> > +		       PCIE_MAC_SRSTB | PCIE_CRSTB;
> > +		writel(val, port->base + PCIE_RST_CTRL);
> 
> If PCIE_LINKDOWN_RST_EN corresponds to PERST# signal, then it should be
> deasserted only after the power and REFCLK are stable. But I'm not sure what the
> above PCIE_RST_CTRL setting is doing. If it somehow affects either power or
> REFCLK, then it should come before PCIE_LINKDOWN_RST_EN.
>

Hi I checked MT7622 programming guide for this.

The order of operation for reset is exactly what you have described.

Indded, first everything is reset (assert all signals) then we set
PCIE_LINKDOWN_RST_EN that in unrelated. We wait for clock stabilization
with the msleep and only AFTER we deassert the PCIE_PERSTB.

So no, PCIE_LINKDOWN_RST_EN is not PERST# signal sw but it's actually
PCIE_PERSTB.

> 
> -- 
> மணிவண்ணன் சதாசிவம்

-- 
	Ansuel