diff mbox series

[v3,1/3] PCI: apple: Follow the PCIe specifications when resetting the port

Message ID 20211123180636.80558-2-maz@kernel.org (mailing list archive)
State Accepted
Delegated to: Lorenzo Pieralisi
Headers show
Series PCI: apple: Assorted #PERST fixes | expand

Commit Message

Marc Zyngier Nov. 23, 2021, 6:06 p.m. UTC
While the Apple PCIe driver works correctly when directly booted
from the firmware, it fails to initialise when the kernel is booted
from a bootloader using PCIe such as u-boot.

That's beacuse we're missing a proper reset of the port (we only
clear the reset, but never assert it).

The PCIe spec requirements are two-fold:

- #PERST must be asserted before setting up the clocks, and
  stay asserted for at least 100us (Tperst-clk).

- Once #PERST is deasserted, the OS must wait for at least 100ms
  "from the end of a Conventional Reset" before we can start talking
  to the devices

Implementing this results in a booting system.

Fixes: 1e33888fbe44 ("PCI: apple: Add initial hardware bring-up")
Acked-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Cc: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
---
 drivers/pci/controller/pcie-apple.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

Comments

Luca Ceresoli Nov. 24, 2021, 9:22 a.m. UTC | #1
Hi,

On 23/11/21 19:06, Marc Zyngier wrote:
> While the Apple PCIe driver works correctly when directly booted
> from the firmware, it fails to initialise when the kernel is booted
> from a bootloader using PCIe such as u-boot.
> 
> That's beacuse we're missing a proper reset of the port (we only
> clear the reset, but never assert it).
> 
> The PCIe spec requirements are two-fold:
> 
> - #PERST must be asserted before setting up the clocks, and
>   stay asserted for at least 100us (Tperst-clk).
> 
> - Once #PERST is deasserted, the OS must wait for at least 100ms
>   "from the end of a Conventional Reset" before we can start talking
>   to the devices
> 
> Implementing this results in a booting system.
> 
> Fixes: 1e33888fbe44 ("PCI: apple: Add initial hardware bring-up")
> Acked-by: Pali Rohár <pali@kernel.org>
> Signed-off-by: Marc Zyngier <maz@kernel.org>
> Cc: Alyssa Rosenzweig <alyssa@rosenzweig.io>
> Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> Cc: Bjorn Helgaas <bhelgaas@google.com>>
> ---
>  drivers/pci/controller/pcie-apple.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
> 
> diff --git a/drivers/pci/controller/pcie-apple.c b/drivers/pci/controller/pcie-apple.c
> index 1bf4d75b61be..957960a733c4 100644
> --- a/drivers/pci/controller/pcie-apple.c
> +++ b/drivers/pci/controller/pcie-apple.c
> @@ -539,13 +539,23 @@ static int apple_pcie_setup_port(struct apple_pcie *pcie,
>  
>  	rmw_set(PORT_APPCLK_EN, port->base + PORT_APPCLK);
>  
> +	/* Engage #PERST before setting up the clock */

"Assert" is the verb typically used instead of "Engage".

With this fixed, or even without as this is a pretty urgent fix:

Reviewed-by: Luca Ceresoli <luca@lucaceresoli.net>
Robin Murphy Nov. 24, 2021, 12:56 p.m. UTC | #2
On 2021-11-23 18:06, Marc Zyngier wrote:
> While the Apple PCIe driver works correctly when directly booted
> from the firmware, it fails to initialise when the kernel is booted
> from a bootloader using PCIe such as u-boot.
> 
> That's beacuse we're missing a proper reset of the port (we only
> clear the reset, but never assert it).
> 
> The PCIe spec requirements are two-fold:
> 
> - #PERST must be asserted before setting up the clocks, and
>    stay asserted for at least 100us (Tperst-clk).
> 
> - Once #PERST is deasserted, the OS must wait for at least 100ms
>    "from the end of a Conventional Reset" before we can start talking
>    to the devices
> 
> Implementing this results in a booting system.
> 
> Fixes: 1e33888fbe44 ("PCI: apple: Add initial hardware bring-up")
> Acked-by: Pali Rohár <pali@kernel.org>
> Signed-off-by: Marc Zyngier <maz@kernel.org>
> Cc: Alyssa Rosenzweig <alyssa@rosenzweig.io>
> Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> Cc: Bjorn Helgaas <bhelgaas@google.com>
> ---
>   drivers/pci/controller/pcie-apple.c | 10 ++++++++++
>   1 file changed, 10 insertions(+)
> 
> diff --git a/drivers/pci/controller/pcie-apple.c b/drivers/pci/controller/pcie-apple.c
> index 1bf4d75b61be..957960a733c4 100644
> --- a/drivers/pci/controller/pcie-apple.c
> +++ b/drivers/pci/controller/pcie-apple.c
> @@ -539,13 +539,23 @@ static int apple_pcie_setup_port(struct apple_pcie *pcie,
>   
>   	rmw_set(PORT_APPCLK_EN, port->base + PORT_APPCLK);
>   
> +	/* Engage #PERST before setting up the clock */
> +	gpiod_set_value(reset, 0);

FWIW, given that getting the GPIO with GPIOD_OUT_LOW should have had 
this effect in the first place, if this isn't a no-op at this point then 
it would hint at something being more significantly wrong down at the 
GPIO/pinctrl end :/

Once you fix the polarity in the later patch, though, adding the 
explicit reset assertion here does seem a far nicer option than fiddling 
the flags to preserve the implicit assertion earlier.

Cheers,
Robin.

> +
>   	ret = apple_pcie_setup_refclk(pcie, port);
>   	if (ret < 0)
>   		return ret;
>   
> +	/* The minimal Tperst-clk value is 100us (PCIe CMS r2.0, 2.6.2) */
> +	usleep_range(100, 200);
> +
> +	/* Deassert #PERST */
>   	rmw_set(PORT_PERST_OFF, port->base + PORT_PERST);
>   	gpiod_set_value(reset, 1);
>   
> +	/* Wait for 100ms after #PERST deassertion (PCIe r2.0, 6.6.1) */
> +	msleep(100);
> +
>   	ret = readl_relaxed_poll_timeout(port->base + PORT_STATUS, stat,
>   					 stat & PORT_STATUS_READY, 100, 250000);
>   	if (ret < 0) {
>
Bjorn Helgaas Dec. 7, 2021, 4:22 p.m. UTC | #3
On Tue, Nov 23, 2021 at 06:06:34PM +0000, Marc Zyngier wrote:
> While the Apple PCIe driver works correctly when directly booted
> from the firmware, it fails to initialise when the kernel is booted
> from a bootloader using PCIe such as u-boot.
> 
> That's beacuse we're missing a proper reset of the port (we only
> clear the reset, but never assert it).
> 
> The PCIe spec requirements are two-fold:
> 
> - #PERST must be asserted before setting up the clocks, and
>   stay asserted for at least 100us (Tperst-clk).
> 
> - Once #PERST is deasserted, the OS must wait for at least 100ms
>   "from the end of a Conventional Reset" before we can start talking
>   to the devices

Unless somebody objects, I'll s/#PERST/PERST#/ to match the spec
usage, both here and in the comments below.

I also notice gpiod_get_from_of_node(..., "#PERST") earlier in
apple_pcie_setup_port().  If it wouldn't break anything, I'd like to
change that, too.

> Implementing this results in a booting system.
> 
> Fixes: 1e33888fbe44 ("PCI: apple: Add initial hardware bring-up")
> Acked-by: Pali Rohár <pali@kernel.org>
> Signed-off-by: Marc Zyngier <maz@kernel.org>
> Cc: Alyssa Rosenzweig <alyssa@rosenzweig.io>
> Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> Cc: Bjorn Helgaas <bhelgaas@google.com>
> ---
>  drivers/pci/controller/pcie-apple.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
> 
> diff --git a/drivers/pci/controller/pcie-apple.c b/drivers/pci/controller/pcie-apple.c
> index 1bf4d75b61be..957960a733c4 100644
> --- a/drivers/pci/controller/pcie-apple.c
> +++ b/drivers/pci/controller/pcie-apple.c
> @@ -539,13 +539,23 @@ static int apple_pcie_setup_port(struct apple_pcie *pcie,
>  
>  	rmw_set(PORT_APPCLK_EN, port->base + PORT_APPCLK);
>  
> +	/* Engage #PERST before setting up the clock */
> +	gpiod_set_value(reset, 0);
> +
>  	ret = apple_pcie_setup_refclk(pcie, port);
>  	if (ret < 0)
>  		return ret;
>  
> +	/* The minimal Tperst-clk value is 100us (PCIe CMS r2.0, 2.6.2) */
> +	usleep_range(100, 200);

Spec says "min 100us from REFCLK stable before PERST# inactive".  So I
guess when apple_pcie_setup_refclk() returns, we know REFCLK is
already stable?

> +	/* Deassert #PERST */
>  	rmw_set(PORT_PERST_OFF, port->base + PORT_PERST);
>  	gpiod_set_value(reset, 1);
>  
> +	/* Wait for 100ms after #PERST deassertion (PCIe r2.0, 6.6.1) */
> +	msleep(100);

Does this port support speeds greater than 5 GT/s?  If so, 6.6.1 says
we need "100ms after Link training completes," not just after
deasserting PERST#.

I'll update this citation to "PCIe r5.0, 6.6.1" to reference the
current spec.

>  	ret = readl_relaxed_poll_timeout(port->base + PORT_STATUS, stat,
>  					 stat & PORT_STATUS_READY, 100, 250000);
>  	if (ret < 0) {
> -- 
> 2.30.2
>
diff mbox series

Patch

diff --git a/drivers/pci/controller/pcie-apple.c b/drivers/pci/controller/pcie-apple.c
index 1bf4d75b61be..957960a733c4 100644
--- a/drivers/pci/controller/pcie-apple.c
+++ b/drivers/pci/controller/pcie-apple.c
@@ -539,13 +539,23 @@  static int apple_pcie_setup_port(struct apple_pcie *pcie,
 
 	rmw_set(PORT_APPCLK_EN, port->base + PORT_APPCLK);
 
+	/* Engage #PERST before setting up the clock */
+	gpiod_set_value(reset, 0);
+
 	ret = apple_pcie_setup_refclk(pcie, port);
 	if (ret < 0)
 		return ret;
 
+	/* The minimal Tperst-clk value is 100us (PCIe CMS r2.0, 2.6.2) */
+	usleep_range(100, 200);
+
+	/* Deassert #PERST */
 	rmw_set(PORT_PERST_OFF, port->base + PORT_PERST);
 	gpiod_set_value(reset, 1);
 
+	/* Wait for 100ms after #PERST deassertion (PCIe r2.0, 6.6.1) */
+	msleep(100);
+
 	ret = readl_relaxed_poll_timeout(port->base + PORT_STATUS, stat,
 					 stat & PORT_STATUS_READY, 100, 250000);
 	if (ret < 0) {