| Message ID | 20220826115607.1148489-1-michael@walle.cc |
|---|---|
| Headers | show |
| Series | reset: microchip-sparx5: fix the broken switch reset | expand |
Hi Michael, On Fri, 2022-08-26 at 13:56 +0200, Michael Walle wrote: > EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe > > Originally this was used in by the switch core driver to issue a reset. > But it turns out, this isn't just a switch core reset but instead it > will reset almost the complete SoC. > > Instead of adding almost all devices of the SoC a shared reset line, > issue the reset once early on startup. Keep the reset controller for > backwards compatibility, but make the actual reset a noop. > > Suggested-by: Philipp Zabel <p.zabel@pengutronix.de> > Signed-off-by: Michael Walle <michael@walle.cc> > --- > drivers/reset/reset-microchip-sparx5.c | 22 +++++++++++++++++----- > 1 file changed, 17 insertions(+), 5 deletions(-) > > diff --git a/drivers/reset/reset-microchip-sparx5.c b/drivers/reset/reset-microchip-sparx5.c > index 00b612a0effa..f3528dd1d084 100644 > --- a/drivers/reset/reset-microchip-sparx5.c > +++ b/drivers/reset/reset-microchip-sparx5.c > @@ -33,11 +33,8 @@ static struct regmap_config sparx5_reset_regmap_config = { > .reg_stride = 4, > }; > > -static int sparx5_switch_reset(struct reset_controller_dev *rcdev, > - unsigned long id) > +static int sparx5_switch_reset(struct mchp_reset_context *ctx) > { > - struct mchp_reset_context *ctx = > - container_of(rcdev, struct mchp_reset_context, rcdev); > u32 val; > > /* Make sure the core is PROTECTED from reset */ > @@ -54,8 +51,14 @@ static int sparx5_switch_reset(struct reset_controller_dev *rcdev, > 1, 100); > } > > +static int sparx5_reset_noop(struct reset_controller_dev *rcdev, > + unsigned long id) > +{ > + return 0; > +} > + > static const struct reset_control_ops sparx5_reset_ops = { > - .reset = sparx5_switch_reset, > + .reset = sparx5_reset_noop, > }; > > static int mchp_sparx5_map_syscon(struct platform_device *pdev, char *name, > @@ -122,6 +125,11 @@ static int mchp_sparx5_reset_probe(struct platform_device *pdev) > ctx->rcdev.of_node = dn; > ctx->props = device_get_match_data(&pdev->dev); > > + /* Issue the reset very early, our actual reset callback is a noop. */ > + err = sparx5_switch_reset(ctx); > + if (err) > + return err; > + > return devm_reset_controller_register(&pdev->dev, &ctx->rcdev); > } > > @@ -163,6 +171,10 @@ static int __init mchp_sparx5_reset_init(void) > return platform_driver_register(&mchp_sparx5_reset_driver); > } > > +/* > + * Because this is a global reset, keep this postcore_initcall() to issue the > + * reset as early as possible during the kernel startup. > + */ > postcore_initcall(mchp_sparx5_reset_init); > > MODULE_DESCRIPTION("Microchip Sparx5 switch reset driver"); > -- > 2.30.2 > Tested-by: Steen Hegelund <Steen.Hegelund@microchip.com> on Sparx5 BR Steen
Hi Steen, Am 2022-08-29 11:14, schrieb Steen Hegelund: > On Fri, 2022-08-26 at 13:56 +0200, Michael Walle wrote: >> EXTERNAL EMAIL: Do not click links or open attachments unless you know >> the content is safe >> >> Originally this was used in by the switch core driver to issue a >> reset. >> But it turns out, this isn't just a switch core reset but instead it >> will reset almost the complete SoC. >> >> Instead of adding almost all devices of the SoC a shared reset line, >> issue the reset once early on startup. Keep the reset controller for >> backwards compatibility, but make the actual reset a noop. >> >> Suggested-by: Philipp Zabel <p.zabel@pengutronix.de> >> Signed-off-by: Michael Walle <michael@walle.cc> .. > Tested-by: Steen Hegelund <Steen.Hegelund@microchip.com> on Sparx5 Thanks for testing! -michael
Am 2022-08-26 13:56, schrieb Michael Walle: > The reset which is used by the switch to reset the switch core has many > different side effects. It is not just a switch reset. Thus don't treat > it > as one, but just issue the reset early during boot. > > Michael Walle (3): > reset: microchip-sparx5: issue a reset on startup > dt-bindings: net: sparx5: don't require a reset line > net: lan966x: make reset optional > > .../bindings/net/microchip,sparx5-switch.yaml | 2 -- > .../ethernet/microchip/lan966x/lan966x_main.c | 3 ++- > drivers/reset/reset-microchip-sparx5.c | 22 ++++++++++++++----- > 3 files changed, 19 insertions(+), 8 deletions(-) Philipp, you could just patch #1, I guess. I'd then resend patches #2 and #3 to the netdev ML targetting net-next. As long as the device tree itself isn't changed, there should be no dependency between these two. -michael
On Fr, 2022-08-26 at 13:56 +0200, Michael Walle wrote: > Originally this was used in by the switch core driver to issue a reset. > But it turns out, this isn't just a switch core reset but instead it > will reset almost the complete SoC. > > Instead of adding almost all devices of the SoC a shared reset line, > issue the reset once early on startup. Keep the reset controller for > backwards compatibility, but make the actual reset a noop. > > Suggested-by: Philipp Zabel <p.zabel@pengutronix.de> > Signed-off-by: Michael Walle <michael@walle.cc> I've applied this patch to the reset/fixes branch. regards Philipp