linux-kernel - Re: [PATCH 2/2] PCI: xilinx-cpm: Add support for PCIe RP PERST# signal

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <Z7xx8ZxpD011oCCj@ryzen>
Date: Mon, 24 Feb 2025 14:19:45 +0100
From: Niklas Cassel <cassel@...nel.org>
To: Manivannan Sadhasivam <manivannan.sadhasivam@...aro.org>
Cc: Sai Krishna Musham <sai.krishna.musham@....com>, bhelgaas@...gle.com,
	lpieralisi@...nel.org, kw@...ux.com, robh@...nel.org,
	krzk+dt@...nel.org, conor+dt@...nel.org, linux-pci@...r.kernel.org,
	devicetree@...r.kernel.org, linux-kernel@...r.kernel.org,
	michal.simek@....com, bharat.kumar.gogada@....com,
	thippeswamy.havalige@....com
Subject: Re: [PATCH 2/2] PCI: xilinx-cpm: Add support for PCIe RP PERST#
 signal

On Mon, Feb 24, 2025 at 12:32:36PM +0530, Manivannan Sadhasivam wrote:
> On Mon, Feb 24, 2025 at 12:00:46PM +0530, Sai Krishna Musham wrote:
> > Add GPIO-based control for the PCIe Root Port PERST# signal.
> > 
> > According to section 2.2 of the PCIe Electromechanical Specification
> > (Revision 6.0), PERST# signal has to be deasserted after a delay of
> > 100 ms (TPVPERL) to ensure proper reset sequencing during PCIe
> > initialization.
> > 
> > Signed-off-by: Sai Krishna Musham <sai.krishna.musham@....com>
> > ---
> > This patch depends on the following patch series.
> > https://lore.kernel.org/all/20250217072713.635643-3-thippeswamy.havalige@amd.com/
> > ---
> >  drivers/pci/controller/pcie-xilinx-cpm.c | 23 +++++++++++++++++++++++
> >  1 file changed, 23 insertions(+)
> > 
> > diff --git a/drivers/pci/controller/pcie-xilinx-cpm.c b/drivers/pci/controller/pcie-xilinx-cpm.c
> > index 81e8bfae53d0..0e31b85658e6 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>
> > @@ -568,8 +570,29 @@ static int xilinx_cpm_pcie_probe(struct platform_device *pdev)
> >  	struct device *dev = &pdev->dev;
> >  	struct pci_host_bridge *bridge;
> >  	struct resource_entry *bus;
> > +	struct gpio_desc *reset_gpio;
> >  	int err;
> >  
> > +	/* Request the GPIO for PCIe reset signal */
> > +	reset_gpio = devm_gpiod_get(dev, "reset", GPIOD_OUT_LOW);
> 
> You've defined the polarity as 0x1 in the binding. Which corresponds to
> GPIO_ACTIVE_LOW. So if you request the GPIO as GPIOD_OUT_LOW, it means the host
> is going to drive the PERST# line as 'high', which corresponds to PERST#
> deassert. I don't think you'd want that and if that is what is really happening,
> the endpoint state machine would be broken. So I suspect that the polarity of
> your PERST# line is wrong.
> 
> - Mani
> 
> > +	if (IS_ERR(reset_gpio)) {
> > +		dev_err(dev, "Failed to request reset GPIO\n");
> > +		return PTR_ERR(reset_gpio);
> > +	}
> > +
> > +	/* Assert the reset signal */
> > +	gpiod_set_value(reset_gpio, 0);
> > +
> > +	/*
> > +	 * As per section 2.2 of the PCI Express Card Electromechanical
> > +	 * Specification (Revision 6.0), the deassertion of the PERST# signal
> > +	 * should be delayed by 100 ms (TPVPERL).
> > +	 */
> > +	msleep(100);
> > +
> > +	/* Deassert the reset signal */
> > +	gpiod_set_value(reset_gpio, 1);
> > +

Since you are adding PERST# to your DT binding, please make sure that you
define it properly (as active low) in the DT binding.

Then in the driver things should look like this and nothing else:
devm_gpiod_get(..., GPIOD_OUT_HIGH); // will assert PERST#
gpiod_set_value(reset_gpio, 1); // assert PERST# (will be a no-op since PERST# is already asserted)
msleep(100);
gpiod_set_value(reset_gpio, 0); // deassert PERST#



Kind regards,
Niklas