On Fri, Apr 04, 2025 at 06:59:23AM +0000, Musham, Sai Krishna wrote:
> [AMD Official Use Only - AMD Internal Distribution Only]
>
> Hi Manivannan,
>
> Thanks for the review.
>
> > -----Original Message-----
> > From: Manivannan Sadhasivam <manivannan.sadhasivam@xxxxxxxxxx>
> > Sent: Thursday, March 27, 2025 10:56 PM
> > To: Musham, Sai Krishna <sai.krishna.musham@xxxxxxx>
> > Cc: bhelgaas@xxxxxxxxxx; lpieralisi@xxxxxxxxxx; kw@xxxxxxxxx; robh@xxxxxxxxxx;
> > krzk+dt@xxxxxxxxxx; conor+dt@xxxxxxxxxx; cassel@xxxxxxxxxx; linux-
> > pci@xxxxxxxxxxxxxxx; devicetree@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> > Simek, Michal <michal.simek@xxxxxxx>; Gogada, Bharat Kumar
> > <bharat.kumar.gogada@xxxxxxx>; Havalige, Thippeswamy
> > <thippeswamy.havalige@xxxxxxx>
> > Subject: Re: [PATCH v6 2/2] PCI: xilinx-cpm: Add support for PCIe RP PERST#
> > signal
> >
> > Caution: This message originated from an External Source. Use proper caution
> > when opening attachments, clicking links, or responding.
> >
> >
> > On Wed, Mar 26, 2025 at 07:58:11AM +0530, Sai Krishna Musham wrote:
> > > Add PCIe IP reset along with GPIO-based control for the PCIe Root
> > > Port PERST# signal. Synchronizing the PCIe IP reset with the PERST#
> > > signal's assertion and deassertion avoids Link Training failures.
> > >
> > > Adapt to use GPIO framework and make reset optional to maintain
> > > backward compatibility with existing DTBs.
> > >
> > > Add clear firewall after Link reset for CPM5NC.
> > >
> > > Signed-off-by: Sai Krishna Musham <sai.krishna.musham@xxxxxxx>
> > > ---
> > > Changes for v6:
> > > - Correct version check condition of CPM5NC_HOST.
> > >
> > > Changes for v5:
> > > - Handle probe defer for reset_gpio.
> > > - Resolve ABI break.
> > >
> > > Changes for v4:
> > > - Add PCIe PERST# support for CPM5NC.
> > > - Add PCIe IP reset along with PERST# to avoid Link Training Errors.
> > > - Remove PCIE_T_PVPERL_MS define and PCIE_T_RRS_READY_MS after
> > > PERST# deassert.
> > > - Move PCIe PERST# assert and deassert logic to
> > > xilinx_cpm_pcie_init_port() before cpm_pcie_link_up(), since
> > > Interrupts enable and PCIe RP bridge enable should be done after
> > > Link up.
> > > - Update commit message.
> > >
> > > Changes for v3:
> > > - Use PCIE_T_PVPERL_MS define.
> > >
> > > Changes for v2:
> > > - Make the request GPIO optional.
> > > - Correct the reset sequence as per PERST#
> > > - Update commit message
> > > ---
> > > drivers/pci/controller/pcie-xilinx-cpm.c | 86 ++++++++++++++++++++++--
> > > 1 file changed, 82 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/drivers/pci/controller/pcie-xilinx-cpm.c b/drivers/pci/controller/pcie-xilinx-
> > cpm.c
> > > index d0ab187d917f..b10c0752a94f 100644
> > > --- a/drivers/pci/controller/pcie-xilinx-cpm.c
> > > +++ b/drivers/pci/controller/pcie-xilinx-cpm.c
> > > @@ -6,6 +6,8 @@
> > > */
> > >
> > > #include <linux/bitfield.h>
> > > +#include <linux/delay.h>
> > > +#include <linux/gpio/consumer.h>
> > > #include <linux/interrupt.h>
> > > #include <linux/irq.h>
> > > #include <linux/irqchip.h>
> > > @@ -21,6 +23,13 @@
> > > #include "pcie-xilinx-common.h"
> > >
> > > /* Register definitions */
> > > +#define XILINX_CPM_PCIE0_RST 0x00000308
> > > +#define XILINX_CPM5_PCIE0_RST 0x00000318
> > > +#define XILINX_CPM5_PCIE1_RST 0x0000031C
> > > +#define XILINX_CPM5NC_PCIE0_RST 0x00000324
> > > +
> > > +#define XILINX_CPM5NC_PCIE0_FRWALL 0x00001140
> > > +
> > > #define XILINX_CPM_PCIE_REG_IDR 0x00000E10
> > > #define XILINX_CPM_PCIE_REG_IMR 0x00000E14
> > > #define XILINX_CPM_PCIE_REG_PSCR 0x00000E1C
> > > @@ -99,6 +108,7 @@ struct xilinx_cpm_variant {
> > > u32 ir_status;
> > > u32 ir_enable;
> > > u32 ir_misc_value;
> > > + u32 cpm_pcie_rst;
> > > };
> > >
> > > /**
> > > @@ -106,6 +116,8 @@ struct xilinx_cpm_variant {
> > > * @dev: Device pointer
> > > * @reg_base: Bridge Register Base
> > > * @cpm_base: CPM System Level Control and Status Register(SLCR) Base
> > > + * @crx_base: CPM Clock and Reset Control Registers Base
> > > + * @cpm5nc_attr_base: CPM5NC Control and Status Registers Base
> > > * @intx_domain: Legacy IRQ domain pointer
> > > * @cpm_domain: CPM IRQ domain pointer
> > > * @cfg: Holds mappings of config space window
> > > @@ -118,6 +130,8 @@ struct xilinx_cpm_pcie {
> > > struct device *dev;
> > > void __iomem *reg_base;
> > > void __iomem *cpm_base;
> > > + void __iomem *crx_base;
> > > + void __iomem *cpm5nc_attr_base;
> > > struct irq_domain *intx_domain;
> > > struct irq_domain *cpm_domain;
> > > struct pci_config_window *cfg;
> > > @@ -475,12 +489,45 @@ static int xilinx_cpm_setup_irq(struct xilinx_cpm_pcie
> > *port)
> > > * xilinx_cpm_pcie_init_port - Initialize hardware
> > > * @port: PCIe port information
> > > */
> > > -static void xilinx_cpm_pcie_init_port(struct xilinx_cpm_pcie *port)
> > > +static int xilinx_cpm_pcie_init_port(struct xilinx_cpm_pcie *port)
> > > {
> > > const struct xilinx_cpm_variant *variant = port->variant;
> > > + struct device *dev = port->dev;
> > > + struct gpio_desc *reset_gpio;
> > > +
> > > + /* Request the GPIO for PCIe reset signal */
> > > + reset_gpio = devm_gpiod_get_optional(dev, "reset", GPIOD_OUT_HIGH);
> > > + if (IS_ERR(reset_gpio)) {
> > > + if (PTR_ERR(reset_gpio) != -EPROBE_DEFER)
> > > + dev_err(dev, "Failed to request reset GPIO\n");
> > > + return PTR_ERR(reset_gpio);
> > > + }
> > >
> > > - if (variant->version == CPM5NC_HOST)
> > > - return;
> > > + if (reset_gpio && port->crx_base) {
> > > + /* Assert the PCIe IP reset */
> > > + writel_relaxed(0x1, port->crx_base + variant->cpm_pcie_rst);
> > > +
> > > + /* Controller specific delay */
> > > + udelay(50);
> > > +
> >
> > There should be atleast 100ms delay before PERST# deassert as per the spec. So
> > use PCIE_T_PVPERL_MS. I know that you had it before, but removed in v4. I don't
> > see a valid reason for that.
>
> For CPM/CPM5/CPM5NC, the "Power Up" sequence mentioned in section 2.2.1
> of PCIe Electromechanical Spec is handled in the design. The PERST# we are
> using here is applied after the Power Up sequence and will be used for warm reset,
> where power of the system is already stable.
>
I don't quite understand what you mean by 'warm reset' here. Even if the power
was already stable, what is the guarantee that the 100ms time is elapsed before
deasserting the PERST#? Does the hardware logic ensure 100ms time is elapsed
before the driver is probed?
> So, we changed the delay after PERST# and IP reset assertion to 50us controller
> specific delay, similar to TPERST(PERST# active time 100us) delay in "Power
> sequencing and Reset Signal Timings" of PCIe Electromechanical Spec. After
> deassertion of PERST# signal and IP reset, a delay of PCIE_T_RRS_READY_MS
> is required before checking the Link. Please let me know if you have further queries.
>
This part is fine.
> Thanks, I will update this information in commit message.
> >
> > > + /* Deassert the PCIe IP reset */
> > > + writel_relaxed(0x0, port->crx_base + variant->cpm_pcie_rst);
> > > +
> > > + /* Deassert the reset signal */
> > > + gpiod_set_value(reset_gpio, 0);
> > > + mdelay(PCIE_T_RRS_READY_MS);
> > > +
> > > + if (variant->version == CPM5NC_HOST && port->cpm5nc_attr_base) {
> > > + /* Clear Firewall */
> > > + writel_relaxed(0x00, port->cpm5nc_attr_base +
> > > + XILINX_CPM5NC_PCIE0_FRWALL);
> > > + writel_relaxed(0x01, port->cpm5nc_attr_base +
> > > + XILINX_CPM5NC_PCIE0_FRWALL);
> > > + writel_relaxed(0x00, port->cpm5nc_attr_base +
> > > + XILINX_CPM5NC_PCIE0_FRWALL);
> > > + return 0;
> > > + }
> > > + }
> > >
> > > if (cpm_pcie_link_up(port))
> > > dev_info(port->dev, "PCIe Link is UP\n");
> > > @@ -512,6 +559,8 @@ static void xilinx_cpm_pcie_init_port(struct
> > xilinx_cpm_pcie *port)
> > > pcie_write(port, pcie_read(port, XILINX_CPM_PCIE_REG_RPSC) |
> > > XILINX_CPM_PCIE_REG_RPSC_BEN,
> > > XILINX_CPM_PCIE_REG_RPSC);
> > > +
> > > + return 0;
> > > }
> > >
> > > /**
> > > @@ -551,6 +600,27 @@ static int xilinx_cpm_pcie_parse_dt(struct
> > xilinx_cpm_pcie *port,
> > > port->reg_base = port->cfg->win;
> > > }
> > >
> > > + port->crx_base = devm_platform_ioremap_resource_byname(pdev,
> > > + "cpm_crx");
> > > + if (IS_ERR(port->crx_base)) {
> > > + if (PTR_ERR(port->crx_base) == -EINVAL)
> > > + port->crx_base = NULL;
> > > + else
> > > + return PTR_ERR(port->crx_base);
> > > + }
> > > +
> > > + if (port->variant->version == CPM5NC_HOST) {
> > > + port->cpm5nc_attr_base =
> > > + devm_platform_ioremap_resource_byname(pdev,
> > > + "cpm5nc_attr");
> >
> > Where is this resource defined in the binding?
>
> This patch is tested for mentioned CPM versions, I apologize that
> I missed adding the cpm5nc_attr resource in DT binding. I will not
> repeat this again. I will add the resource in the next patch.
> Thanks for your understanding.
> >
> > > + if (IS_ERR(port->cpm5nc_attr_base)) {
> > > + if (PTR_ERR(port->cpm5nc_attr_base) == -EINVAL)
> >
> > Why?
>
> This condition check is added to make cpm5nc_attr_base optional,
> once I add missing resource in DT this condition will be applicable.
Why are you checking for -EINVAL? What does it correspond to?
If your intention is to make the resource_get optional, you should use
platform_get_resource_byname() first. If it returns NULL, then it means the
resource is not defined in DT.
- Mani
--
மணிவண்ணன் சதாசிவம்
