Re: [PATCH v2 4/4] PCI: j721e: Add support to build as a loadable module

[Siddharth Vadapalli]

On Sun, Apr 13, 2025 at 07:43:05PM +0530, Manivannan Sadhasivam wrote:
> On Wed, Apr 09, 2025 at 12:12:27PM +0530, Siddharth Vadapalli wrote:
> > On Wed, Apr 09, 2025 at 12:06:35PM +0530, Manivannan Sadhasivam wrote:
> > 
> > Hello Mani,
> > 
> > > On Sun, Mar 30, 2025 at 02:09:14PM +0530, Siddharth Vadapalli wrote:
> > > > The 'pci-j721e.c' driver is the application/glue/wrapper driver for the
> > > > Cadence PCIe Controllers on TI SoCs. Implement support for building it as a
> > > > loadable module.
> > > > 
> > > > Signed-off-by: Siddharth Vadapalli <s-vadapalli@xxxxxx>
> > > > ---
> > > > 
> > > > v1:
> > > > https://lore.kernel.org/r/20250307103128.3287497-5-s-vadapalli@xxxxxx/
> > > > Changes since v1:
> > > > - Based on feedback from Thomas at:
> > > >   https://lore.kernel.org/r/88b3ecbe-32b6-4310-afb9-da19a2d0506a@xxxxxxxxxxx/
> > > >   the "check" for a non-NULL "pcie-refclk" in j721e_pcie_remove() has been
> > > >   dropped.
> > > > 
> > > > Regards,
> > > > Siddharth.
> > > > 
> > > >  drivers/pci/controller/cadence/Kconfig     |  6 ++--
> > > >  drivers/pci/controller/cadence/pci-j721e.c | 33 +++++++++++++++++++++-
> > > >  2 files changed, 35 insertions(+), 4 deletions(-)
> > > > 
> > > > diff --git a/drivers/pci/controller/cadence/Kconfig b/drivers/pci/controller/cadence/Kconfig
> > > > index 82b58096eea0..72d7d264d6c3 100644
> > > > --- a/drivers/pci/controller/cadence/Kconfig
> > > > +++ b/drivers/pci/controller/cadence/Kconfig
> > > > @@ -43,10 +43,10 @@ config PCIE_CADENCE_PLAT_EP
> > > >  	  different vendors SoCs.
> > > >  
> > > >  config PCI_J721E
> > > > -	bool
> > > > +	tristate
> > > >  
> > > >  config PCI_J721E_HOST
> > > > -	bool "TI J721E PCIe controller (host mode)"
> > > > +	tristate "TI J721E PCIe controller (host mode)"
> > > >  	depends on ARCH_K3 || COMPILE_TEST
> > > >  	depends on OF
> > > >  	select PCIE_CADENCE_HOST
> > > > @@ -57,7 +57,7 @@ config PCI_J721E_HOST
> > > >  	  core.
> > > >  
> > > >  config PCI_J721E_EP
> > > > -	bool "TI J721E PCIe controller (endpoint mode)"
> > > > +	tristate "TI J721E PCIe controller (endpoint mode)"
> > > >  	depends on ARCH_K3 || COMPILE_TEST
> > > >  	depends on OF
> > > >  	depends on PCI_ENDPOINT
> > > > diff --git a/drivers/pci/controller/cadence/pci-j721e.c b/drivers/pci/controller/cadence/pci-j721e.c
> > > > index ef1cfdae33bb..8bffcd31729c 100644
> > > > --- a/drivers/pci/controller/cadence/pci-j721e.c
> > > > +++ b/drivers/pci/controller/cadence/pci-j721e.c
> > > > @@ -15,6 +15,7 @@
> > > >  #include <linux/irqchip/chained_irq.h>
> > > >  #include <linux/irqdomain.h>
> > > >  #include <linux/mfd/syscon.h>
> > > > +#include <linux/module.h>
> > > >  #include <linux/of.h>
> > > >  #include <linux/pci.h>
> > > >  #include <linux/platform_device.h>
> > > > @@ -27,6 +28,7 @@
> > > >  #define cdns_pcie_to_rc(p) container_of(p, struct cdns_pcie_rc, pcie)
> > > >  
> > > >  #define ENABLE_REG_SYS_2	0x108
> > > > +#define ENABLE_CLR_REG_SYS_2	0x308
> > > >  #define STATUS_REG_SYS_2	0x508
> > > >  #define STATUS_CLR_REG_SYS_2	0x708
> > > >  #define LINK_DOWN		BIT(1)
> > > > @@ -116,6 +118,15 @@ static irqreturn_t j721e_pcie_link_irq_handler(int irq, void *priv)
> > > >  	return IRQ_HANDLED;
> > > >  }
> > > >  
> > > > +static void j721e_pcie_disable_link_irq(struct j721e_pcie *pcie)
> > > > +{
> > > > +	u32 reg;
> > > > +
> > > > +	reg = j721e_pcie_intd_readl(pcie, ENABLE_CLR_REG_SYS_2);
> > > > +	reg |= pcie->linkdown_irq_regfield;
> > > > +	j721e_pcie_intd_writel(pcie, ENABLE_CLR_REG_SYS_2, reg);
> > > > +}
> > > > +
> > > >  static void j721e_pcie_config_link_irq(struct j721e_pcie *pcie)
> > > >  {
> > > >  	u32 reg;
> > > > @@ -633,9 +644,25 @@ static void j721e_pcie_remove(struct platform_device *pdev)
> > > >  	struct j721e_pcie *pcie = platform_get_drvdata(pdev);
> > > >  	struct cdns_pcie *cdns_pcie = pcie->cdns_pcie;
> > > >  	struct device *dev = &pdev->dev;
> > > > +	struct cdns_pcie_ep *ep;
> > > > +	struct cdns_pcie_rc *rc;
> > > > +
> > > > +	if (pcie->mode == PCI_MODE_RC) {
> > > > +		rc = container_of(cdns_pcie, struct cdns_pcie_rc, pcie);
> > > > +		cdns_pcie_host_disable(rc);
> > > > +	} else {
> > > > +		ep = container_of(cdns_pcie, struct cdns_pcie_ep, pcie);
> > > > +		cdns_pcie_ep_disable(ep);
> > > > +	}
> > > > +
> > > > +	if (pcie->reset_gpio) {
> > > > +		msleep(PCIE_T_PVPERL_MS);
> > > 
> > > There is no point in adding a delay before PERST# assertion.
> > 
> > True :)
> > 
> > > 
> > > > +		gpiod_set_value_cansleep(pcie->reset_gpio, 1);
> > > 
> > > This is not PERST# assert, isn't it? Typo?
> > 
> > It is PERST# assert.
> 
> Since the reset-gpios polarity is defined as GPIO_ACTIVE_HIGH in DT (which is
> wrong unless you have an external component that changes polarity), for PERST#
> assert, you need to set 0. If you set 1, then PERST# will be signalled as
> deassert.

I realize now that I have accidentally set it to 1 when it should have
been 0. It was meant to be PERST# assert since it is the .remove callback.
The v3 patch at:
https://patchwork.kernel.org/project/linux-pci/patch/20250410104426.463453-5-s-vadapalli@xxxxxx/
dropped the delay which is not required for PERST# assert, but the
polarity is still incorrect. I will fix it in the v4 series.

Thank you for noticing this and pointing it out.

Regards,
Siddharth.