Re: [PATCH v5] PCI: j721e: Delay T_PVPERL+TPERST_CLK before PERST# inactive

From: Siddharth Vadapalli
Date: Fri Dec 01 2023 - 05:53:12 EST


Hello Lorenzo, Bjorn,

On 17/08/23 15:06, Achal Verma wrote:
> As per the PCIe Card Electromechanical specification REV. 5.0, PERST#
> signal should be de-asserted after minimum 100ms from the time power-rails
> achieve specified operating limits and 100us after reference clock gets
> stable.
>
> From PCIe Card Electromechanical specification REV. 5.0 section 2.9.2:
> TPVPERL: Power stable to PERST# inactive - 100ms
> TPERST_CLK: REFCLK stable before PERST# inactive - 100us
>
> Fixes: f3e25911a430 ("PCI: j721e: Add TI J721E PCIe driver")
> Signed-off-by: Achal Verma <a-verma1@xxxxxx>
> ---
> drivers/pci/controller/cadence/pci-j721e.c | 30 +++++++++++-----------
> drivers/pci/pci.h | 3 +++
> 2 files changed, 18 insertions(+), 15 deletions(-)
>
> diff --git a/drivers/pci/controller/cadence/pci-j721e.c b/drivers/pci/controller/cadence/pci-j721e.c
> index e70213c9060a..b09924b010ab 100644
> --- a/drivers/pci/controller/cadence/pci-j721e.c
> +++ b/drivers/pci/controller/cadence/pci-j721e.c
> @@ -34,6 +34,8 @@
> #define J721E_PCIE_USER_LINKSTATUS 0x14
> #define LINK_STATUS GENMASK(1, 0)
>
> +#define PERST_INACTIVE_US (PCIE_TPVPERL_MS*USEC_PER_MSEC + PCIE_TPERST_CLK_US)

This implementation appears incorrect to me since T_PVPERL already accounts for
T_PERST-CLK according to Figure 2-3 of PCI Express Card Electromechanical
Specification Revision 5.1:
https://members.pcisig.com/wg/PCI-SIG/document/19922

Could you please share your opinion since I wish to post a v6 for this patch,
rebasing it to the latest tree which has the commit:
164f66be0c25 PCI: Add T_PVPERL macro
The macro can be used in the current patch, instead of the PERST_INACTIVE_US macro.

> +
> enum link_status {
> NO_RECEIVERS_DETECTED,
> LINK_TRAINING_IN_PROGRESS,
> @@ -359,7 +361,7 @@ static int j721e_pcie_probe(struct platform_device *pdev)
> struct j721e_pcie *pcie;
> struct cdns_pcie_rc *rc = NULL;
> struct cdns_pcie_ep *ep = NULL;
> - struct gpio_desc *gpiod;
> + struct gpio_desc *perst_gpiod;
> void __iomem *base;
> struct clk *clk;
> u32 num_lanes;
> @@ -468,11 +470,10 @@ static int j721e_pcie_probe(struct platform_device *pdev)
>
> switch (mode) {
> case PCI_MODE_RC:
> - gpiod = devm_gpiod_get_optional(dev, "reset", GPIOD_OUT_LOW);
> - if (IS_ERR(gpiod)) {
> - ret = PTR_ERR(gpiod);
> - if (ret != -EPROBE_DEFER)
> - dev_err(dev, "Failed to get reset GPIO\n");
> + perst_gpiod = devm_gpiod_get(dev, "reset", GPIOD_OUT_LOW);
> + if (IS_ERR(perst_gpiod)) {
> + ret = PTR_ERR(perst_gpiod);
> + dev_err(dev, "Failed to get reset GPIO\n");
> goto err_get_sync;
> }
>
> @@ -498,16 +499,15 @@ static int j721e_pcie_probe(struct platform_device *pdev)
>
> /*
> * "Power Sequencing and Reset Signal Timings" table in
> - * PCI EXPRESS CARD ELECTROMECHANICAL SPECIFICATION, REV. 3.0
> - * indicates PERST# should be deasserted after minimum of 100us
> - * once REFCLK is stable. The REFCLK to the connector in RC
> - * mode is selected while enabling the PHY. So deassert PERST#
> - * after 100 us.
> + * PCI EXPRESS CARD ELECTROMECHANICAL SPECIFICATION, REV. 5.0
> + * indicates PERST# should be deasserted after minimum of 100ms
> + * after power rails achieve specified operating limits and
> + * 100us after reference clock gets stable.
> + * PERST_INACTIVE_US accounts for both delays.
> */
> - if (gpiod) {
> - usleep_range(100, 200);
> - gpiod_set_value_cansleep(gpiod, 1);
> - }
> +
> + fsleep(PERST_INACTIVE_US);
> + gpiod_set_value_cansleep(perst_gpiod, 1);
>
> ret = cdns_pcie_host_setup(rc);
> if (ret < 0) {
> diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
> index a4c397434057..80d520be34e6 100644
> --- a/drivers/pci/pci.h
> +++ b/drivers/pci/pci.h
> @@ -13,6 +13,9 @@
>
> #define PCIE_LINK_RETRAIN_TIMEOUT_MS 1000
>
> +#define PCIE_TPVPERL_MS 100 /* see PCIe CEM r5.0, sec 2.9.2 */
> +#define PCIE_TPERST_CLK_US 100
> +
> extern const unsigned char pcie_link_speed[];
> extern bool pci_early_dump;
>

--
Regards,
Siddharth.