[v2,0/2] RK3588 PCIe3 support

Message

Sebastian Reichel July 17, 2023, 5:35 p.m. UTC

Hi,

This adds PCIe v3 support for RK3588. The series depends on the PCIe
v2 series [0], since the the same binding is used. It has been tested
on Rockchip EVB1 and Radxa Rock 5B.

Note, that the PCIe3 PHY driver is currently missing bifurcation
support for RK3588. Thus after this series only PCIe3x4 is usable
(in aggregated x4 mode) without adding support for the PHY's
"rockchip,pcie30-phymode" DT property, which allows configuring
how the lanes are distributed. Aggregated 3x4 mode seems to be the
most common configuration. Both EVB1 and Rock 5B use it, so I
cannot test anything else anyways.

[0] https://lore.kernel.org/all/[e-mail]/

Changes since v1:
 * https://lore.kernel.org/all/[e-mail]/
 * Collected Acked-by from Conor Dooley
 * Dropped deprecated and useless num-ib-windows/num-ob-windows/num-viewport
   properties from DT (Serge Semin)

-- Sebastian

Sebastian Reichel (2):
  dt-bindings: phy: rockchip: add RK3588 PCIe v3 phy
  arm64: dts: rockchip: rk3588: add PCIe3 support

 .../bindings/phy/rockchip,pcie3-phy.yaml      |  33 ++++-
 arch/arm64/boot/dts/rockchip/rk3588.dtsi      | 120 ++++++++++++++++++
 2 files changed, 148 insertions(+), 5 deletions(-)

Comments

Vinod Koul July 18, 2023, 6:15 a.m. UTC | #1

On Mon, 17 Jul 2023 19:35:10 +0200, Sebastian Reichel wrote:
> This adds PCIe v3 support for RK3588. The series depends on the PCIe
> v2 series [0], since the the same binding is used. It has been tested
> on Rockchip EVB1 and Radxa Rock 5B.
> 
> Note, that the PCIe3 PHY driver is currently missing bifurcation
> support for RK3588. Thus after this series only PCIe3x4 is usable
> (in aggregated x4 mode) without adding support for the PHY's
> "rockchip,pcie30-phymode" DT property, which allows configuring
> how the lanes are distributed. Aggregated 3x4 mode seems to be the
> most common configuration. Both EVB1 and Rock 5B use it, so I
> cannot test anything else anyways.
> 
> [...]

Applied, thanks!

[1/2] dt-bindings: phy: rockchip: add RK3588 PCIe v3 phy
      commit: f66782cff479807ad7e98f0cf6a0c0babfe0159b

Best regards,

Heiko Stuebner July 24, 2023, 7:12 p.m. UTC | #2

On Mon, 17 Jul 2023 19:35:10 +0200, Sebastian Reichel wrote:
> This adds PCIe v3 support for RK3588. The series depends on the PCIe
> v2 series [0], since the the same binding is used. It has been tested
> on Rockchip EVB1 and Radxa Rock 5B.
> 
> Note, that the PCIe3 PHY driver is currently missing bifurcation
> support for RK3588. Thus after this series only PCIe3x4 is usable
> (in aggregated x4 mode) without adding support for the PHY's
> "rockchip,pcie30-phymode" DT property, which allows configuring
> how the lanes are distributed. Aggregated 3x4 mode seems to be the
> most common configuration. Both EVB1 and Rock 5B use it, so I
> cannot test anything else anyways.
> 
> [...]

Applied, thanks!

[2/2] arm64: dts: rockchip: rk3588: add PCIe3 support
      commit: 0acf4fa7f187cd7e3dad93f1ee14e9509687621e

Best regards,

Sebastian Reichel July 24, 2023, 7:22 p.m. UTC | #3

Hi Heiko,

On Mon, Jul 24, 2023 at 09:12:35PM +0200, Heiko Stuebner wrote:
> On Mon, 17 Jul 2023 19:35:10 +0200, Sebastian Reichel wrote:
> > This adds PCIe v3 support for RK3588. The series depends on the PCIe
> > v2 series [0], since the the same binding is used. It has been tested
> > on Rockchip EVB1 and Radxa Rock 5B.
> > 
> > Note, that the PCIe3 PHY driver is currently missing bifurcation
> > support for RK3588. Thus after this series only PCIe3x4 is usable
> > (in aggregated x4 mode) without adding support for the PHY's
> > "rockchip,pcie30-phymode" DT property, which allows configuring
> > how the lanes are distributed. Aggregated 3x4 mode seems to be the
> > most common configuration. Both EVB1 and Rock 5B use it, so I
> > cannot test anything else anyways.
> > 
> > [...]
> 
> Applied, thanks!
> 
> [2/2] arm64: dts: rockchip: rk3588: add PCIe3 support
>       commit: 0acf4fa7f187cd7e3dad93f1ee14e9509687621e

Applying this without the PCIe 2 series being applied introduces
warnings for the RK3588 dtbs_check, as mentioned in the second
setence of the text you quoted :)

(FWIW RK356x already has the same warnings, though)

-- Sebastian

Heiko Stuebner July 25, 2023, 6:06 p.m. UTC | #4

Am Montag, 24. Juli 2023, 21:22:23 CEST schrieb Sebastian Reichel:
> Hi Heiko,
> 
> On Mon, Jul 24, 2023 at 09:12:35PM +0200, Heiko Stuebner wrote:
> > On Mon, 17 Jul 2023 19:35:10 +0200, Sebastian Reichel wrote:
> > > This adds PCIe v3 support for RK3588. The series depends on the PCIe
> > > v2 series [0], since the the same binding is used. It has been tested
> > > on Rockchip EVB1 and Radxa Rock 5B.
> > > 
> > > Note, that the PCIe3 PHY driver is currently missing bifurcation
> > > support for RK3588. Thus after this series only PCIe3x4 is usable
> > > (in aggregated x4 mode) without adding support for the PHY's
> > > "rockchip,pcie30-phymode" DT property, which allows configuring
> > > how the lanes are distributed. Aggregated 3x4 mode seems to be the
> > > most common configuration. Both EVB1 and Rock 5B use it, so I
> > > cannot test anything else anyways.
> > > 
> > > [...]
> > 
> > Applied, thanks!
> > 
> > [2/2] arm64: dts: rockchip: rk3588: add PCIe3 support
> >       commit: 0acf4fa7f187cd7e3dad93f1ee14e9509687621e
> 
> Applying this without the PCIe 2 series being applied introduces
> warnings for the RK3588 dtbs_check, as mentioned in the second
> setence of the text you quoted :)
> 
> (FWIW RK356x already has the same warnings, though)

darn, I didn't do a full binding check, only a "compiles" check
after applying. We're at rc3 now, so I would hope that the pcie2
binding changes get applied soon'ish so I can then also pick up the
dts change from there.


Heiko