Message ID | 20231218214451.2345691-1-cristian.ciocaltea@collabora.com |
---|---|
Headers | show |
Series | Enable networking support for StarFive JH7100 SoC | expand |
On 12/18/2023 1:44 PM, Cristian Ciocaltea wrote: > This patch series adds ethernet support for the StarFive JH7100 SoC and makes it > available for the StarFive VisionFive V1 and BeagleV Starlight boards, although > I could only validate on the former SBC. Thank you Emil and Geert for helping > with tests on BeagleV! > > The work is heavily based on the reference implementation [1] and depends on the > SiFive Composable Cache controller and non-coherent DMA support provided by Emil > via [2] and [3]. > > *Update*: as of next-20231214, all dependencies have been merged. > > [1] https://github.com/starfive-tech/linux/commits/visionfive > [2] https://lore.kernel.org/all/CAJM55Z_pdoGxRXbmBgJ5GbVWyeM1N6+LHihbNdT26Oo_qA5VYA@mail.gmail.com/ > [3] https://lore.kernel.org/all/20231130151932.729708-1-emil.renner.berthing@canonical.com/ > I'm not super familiar with how the various pieces fit together, so I'm not sure how valuable my read through is.. but I didn't see anything obviously wrong. For the series: Reviewed-by: Jacob Keller <jacob.e.keller@intel.com> > Changes in v4: > - Restricted double usage of 'ahb' reset name in PATCH 2 (Jessica, Samuel) > - Moved phy reference from PATCH 5 to both PATCH 6 & 7 where the node is > actually defined (Emil, Conor) > - Drop unnecessary gpio include in PATCH 6; also added a DTS comment describing > the rational behind RX internal delay adjustment (Andrew) > - v3: > https://lore.kernel.org/lkml/20231215204050.2296404-1-cristian.ciocaltea@collabora.com/ > > Changes in v3: > - Rebased series onto next-20231214 and dropped the ccache & DMA coherency > related patches (v2 06-08/12) handled by Emil via [3] > - Squashed PATCH v2 01/12 into PATCH v3 2/9, per Krzysztof's review > - Dropped incorrect PATCH v2 02/12 > - Incorporated Emil's feedback; also added his Co-developed-by on all dts > patches > - Documented the need of adjusting RX internal delay in PATCH v3 8/9, per > Andrew's request > - Added clock fixes from Emil (PATCH v3 8-9/9) required to support 10/100Mb > link speeds > - v2: > https://lore.kernel.org/lkml/20231029042712.520010-1-cristian.ciocaltea@collabora.com/ > > Changes in v2: > - Dropped ccache PATCH 01-05 reworked by Emil via [2] > - Dropped already applied PATCH 06/12 > - Added PATCH v2 01 to prepare snps-dwmac binding for JH7100 support > - Added PATCH v2 02-03 to provide some jh7110-dwmac binding optimizations > - Handled JH7110 conflicting work in PATCH 07 via PATCH v2 04 > - Reworked PATCH 8 via PATCH v2 05, adding JH7100 quirk and dropped > starfive,gtxclk-dlychain DT property; also fixed register naming > - Added PATCH v2 08 providing DMA coherency related DT changes > - Updated PATCH 9 commit msg: > s/OF_DMA_DEFAULT_COHERENT/ARCH_DMA_DEFAULT_COHERENT/ > - Replaced 'uncached-offset' property with 'sifive,cache-ops' in PATCH 10/12 > and dropped 'sideband' reg > - Add new patch providing coherent DMA memory pool (PATCH v2 10) > - Updated PATCH 11/12 according to the stmmac glue layer changes in upstream > - Split PATCH 12/12 into PATCH v2 10-12 to handle individual gmac setup of > VisionFive v1 and BeagleV boards as they use different PHYs; also switched > phy-mode from "rgmii-tx" to "rgmii-id" (requires a reduction of > rx-internal-delay-ps by ~50%) > - Rebased series onto next-20231024 > - v1: > https://lore.kernel.org/lkml/20230211031821.976408-1-cristian.ciocaltea@collabora.com/ > > Cristian Ciocaltea (7): > dt-bindings: net: starfive,jh7110-dwmac: Drop redundant reset > description > dt-bindings: net: starfive,jh7110-dwmac: Add JH7100 SoC compatible > net: stmmac: dwmac-starfive: Add support for JH7100 SoC > riscv: dts: starfive: jh7100: Add sysmain and gmac DT nodes > riscv: dts: starfive: jh7100-common: Setup pinmux and enable gmac > riscv: dts: starfive: visionfive-v1: Setup ethernet phy > riscv: dts: starfive: beaglev-starlight: Setup phy reset gpio > > Emil Renner Berthing (2): > clk: starfive: Add flags argument to JH71X0__MUX macro > clk: starfive: jh7100: Add CLK_SET_RATE_PARENT to gmac_tx > > .../devicetree/bindings/net/snps,dwmac.yaml | 11 ++- > .../bindings/net/starfive,jh7110-dwmac.yaml | 75 ++++++++++++----- > .../dts/starfive/jh7100-beaglev-starlight.dts | 11 +++ > .../boot/dts/starfive/jh7100-common.dtsi | 84 +++++++++++++++++++ > .../jh7100-starfive-visionfive-v1.dts | 22 ++++- > arch/riscv/boot/dts/starfive/jh7100.dtsi | 36 ++++++++ > .../clk/starfive/clk-starfive-jh7100-audio.c | 2 +- > drivers/clk/starfive/clk-starfive-jh7100.c | 32 +++---- > .../clk/starfive/clk-starfive-jh7110-aon.c | 6 +- > .../clk/starfive/clk-starfive-jh7110-isp.c | 2 +- > .../clk/starfive/clk-starfive-jh7110-sys.c | 26 +++--- > drivers/clk/starfive/clk-starfive-jh71x0.h | 4 +- > drivers/net/ethernet/stmicro/stmmac/Kconfig | 6 +- > .../ethernet/stmicro/stmmac/dwmac-starfive.c | 32 ++++++- > 14 files changed, 280 insertions(+), 69 deletions(-) >
On 12/19/23 00:53, Jacob Keller wrote: > > > On 12/18/2023 1:44 PM, Cristian Ciocaltea wrote: >> This patch series adds ethernet support for the StarFive JH7100 SoC and makes it >> available for the StarFive VisionFive V1 and BeagleV Starlight boards, although >> I could only validate on the former SBC. Thank you Emil and Geert for helping >> with tests on BeagleV! [...] > I'm not super familiar with how the various pieces fit together, so I'm > not sure how valuable my read through is.. but I didn't see anything > obviously wrong. > > For the series: > > Reviewed-by: Jacob Keller <jacob.e.keller@intel.com> Thanks for the review, Cristian
On Mon, Dec 18, 2023 at 11:44:46PM +0200, Cristian Ciocaltea wrote: > The StarFive VisionFive V1 SBC uses a Motorcomm YT8521 PHY supporting > RGMII-ID, but requires manual adjustment of the RX internal delay to > work properly. > > The default RX delay provided by the driver is 1.95 ns, which proves to > be too high. Applying a 50% reduction seems to mitigate the issue. > > Also note this adjustment is not necessary on BeagleV Starlight SBC, > which uses a Microchip PHY. Hence, there is no indication of a > misbehaviour on the GMAC side, but most likely the issue stems from > the Motorcomm PHY. > > While at it, drop the redundant gpio include, which is already provided > by jh7100-common.dtsi. > > Co-developed-by: Emil Renner Berthing <emil.renner.berthing@canonical.com> > Signed-off-by: Emil Renner Berthing <emil.renner.berthing@canonical.com> > Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Andrew
On Mon, Dec 18, 2023 at 11:44:40PM +0200, Cristian Ciocaltea wrote: > This patch series adds ethernet support for the StarFive JH7100 SoC and makes it > available for the StarFive VisionFive V1 and BeagleV Starlight boards, although > I could only validate on the former SBC. Thank you Emil and Geert for helping > with tests on BeagleV! You will need to split this into patch sets per subsystem. The changes to the stmmac driver can then go via netdev, and the rest via each subsystem maintainer. It should then all meet in linux-next and work there. Andrew
On Mon, Dec 18, 2023 at 11:44:43PM +0200, Cristian Ciocaltea wrote: > Add a missing quirk to enable support for the StarFive JH7100 SoC. > > Additionally, for greater flexibility in operation, allow using the > rgmii-rxid and rgmii-txid phy modes. > > Co-developed-by: Emil Renner Berthing <kernel@esmil.dk> > Signed-off-by: Emil Renner Berthing <kernel@esmil.dk> > Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Andrew
On 12/19/23 11:03, Andrew Lunn wrote: > On Mon, Dec 18, 2023 at 11:44:40PM +0200, Cristian Ciocaltea wrote: >> This patch series adds ethernet support for the StarFive JH7100 SoC and makes it >> available for the StarFive VisionFive V1 and BeagleV Starlight boards, although >> I could only validate on the former SBC. Thank you Emil and Geert for helping >> with tests on BeagleV! > > You will need to split this into patch sets per subsystem. The changes > to the stmmac driver can then go via netdev, and the rest via each > subsystem maintainer. It should then all meet in linux-next and work > there. Thanks for the reviews and your support to get those networking issues properly handled! Just to confirm the split will be done correctly, I'm going to keep just the bindings and dts patches in v5, while the networking driver and clock related patches will be submitted as part of two additional sets. Regards, Cristian