| Message ID | afcb562602e54c969964a608e3b6494a@thalesgroup.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v3] drm/etnaviv: add optional reset support | expand |
Hi Philippe, Am Freitag, dem 06.12.2024 um 17:00 +0000 schrieb LECOINTRE Philippe: > Add optional reset support which is mentioned in vivante,gc.yaml to > allow the driver to work on SoCs whose reset signal is asserted by default > > Signed-off-by: Philippe Lecointre <philippe.lecointre@thalesgroup.com> > Reviewed-by: Simon Lenain <simon.lenain@thalesgroup.com> Upstream usually doesn't put much weight on such internal reviews. No harm here, as the patch is simple enough and I do review it before applying. Just as a hint for the future: if you want maintainers to take such reviews into account for speeding up the adoption of a patch, do the review on the public mailing lists. > --- > v3: > - Rework to match initial feedback > --- > drivers/gpu/drm/etnaviv/etnaviv_gpu.c | 37 +++++++++++++++++++++++++++ > drivers/gpu/drm/etnaviv/etnaviv_gpu.h | 1 + > 2 files changed, 38 insertions(+) > > diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c > index 2d4c112ce033..1961ebac315a 100644 > --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c > +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c > @@ -13,6 +13,7 @@ > #include <linux/platform_device.h> > #include <linux/pm_runtime.h> > #include <linux/regulator/consumer.h> > +#include <linux/reset.h> > #include <linux/thermal.h> > > #include "etnaviv_cmdbuf.h" > @@ -172,6 +173,25 @@ int etnaviv_gpu_get_param(struct etnaviv_gpu *gpu, u32 param, u64 *value) > return 0; > } > > +static int etnaviv_gpu_reset_deassert(struct etnaviv_gpu *gpu) > +{ > + int ret; > + > + /* 32 core clock cycles (slowest clock) required before deassertion */ > + /* 1 microsecond might match all implementations without computation */ I missed to mention this before, as I was focused on the technical side: this is not the multiline comment style used in the kernel/etnaviv. Please use the same style as already found in this file. > + usleep_range(1, 2); > + > + ret = reset_control_deassert(gpu->rst); > + if (ret) > + return ret; > + > + /* 128 core clock cycles (slowest clock) required before any activity on AHB */ > + /* 1 microsecond might match all implementations without computation */ > + usleep_range(1, 2); > + > + return 0; > +} > + > static inline bool etnaviv_is_model_rev(struct etnaviv_gpu *gpu, u32 model, u32 revision) > { > return gpu->identity.model == model && > @@ -799,6 +819,12 @@ int etnaviv_gpu_init(struct etnaviv_gpu *gpu) > goto pm_put; > } > > + ret = etnaviv_gpu_reset_deassert(gpu); > + if (ret) { > + dev_err(gpu->dev, "GPU reset deassert failed\n"); > + goto fail; > + } > + > etnaviv_hw_identify(gpu); > > if (gpu->identity.model == 0) { > @@ -1860,6 +1886,17 @@ static int etnaviv_gpu_platform_probe(struct platform_device *pdev) > if (IS_ERR(gpu->mmio)) > return PTR_ERR(gpu->mmio); > > + > + /* Get Reset: */ > + gpu->rst = devm_reset_control_get_optional_exclusive(&pdev->dev, NULL); > + if (IS_ERR(gpu->rst)) > + return dev_err_probe(dev, PTR_ERR(gpu->rst), > + "failed to get reset\n"); > + > + err = reset_control_assert(gpu->rst); > + if (err) > + return dev_err_probe(dev, err, "failed to assert reset\n"); > + > /* Get Interrupt: */ > gpu->irq = platform_get_irq(pdev, 0); > if (gpu->irq < 0) > diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.h b/drivers/gpu/drm/etnaviv/etnaviv_gpu.h > index 4d8a7d48ade3..0985ea548b82 100644 > --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.h > +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.h > @@ -158,6 +158,7 @@ struct etnaviv_gpu { > struct clk *clk_reg; > struct clk *clk_core; > struct clk *clk_shader; > + struct reset_control *rst; This needs a forward declaration of struct reset_control in the header, to avoid build failures if headers are included in a different order. Please put them right next to the existing ones for regulator and clk. Other than that, patch looks good to me. Regards, Lucas > > unsigned int freq_scale; > unsigned int fe_waitcycles;
Hi Lucas, > -----Message d'origine----- > De : Lucas Stach <l.stach@pengutronix.de> > Envoyé : lundi 9 décembre 2024 11:07 > À : LECOINTRE Philippe <philippe.lecointre@thalesgroup.com>; Russell King > <linux+etnaviv@armlinux.org.uk>; Christian Gmeiner > <christian.gmeiner@gmail.com> > Cc : David Airlie <airlied@gmail.com>; Simona Vetter <simona@ffwll.ch>; > etnaviv@lists.freedesktop.org; dri-devel@lists.freedesktop.org; linux- > kernel@vger.kernel.org; LENAIN Simon <simon.lenain@thalesgroup.com>; > BARBEAU Etienne <etienne.barbeau@thalesgroup.com>; LEJEUNE Sebastien > <sebastien.lejeune@thalesgroup.com> > Objet : Re: [PATCH v3] drm/etnaviv: add optional reset support > > Hi Philippe, > > Am Freitag, dem 06.12.2024 um 17:00 +0000 schrieb LECOINTRE Philippe: > > Add optional reset support which is mentioned in vivante,gc.yaml to > > allow the driver to work on SoCs whose reset signal is asserted by > > default > > > > Signed-off-by: Philippe Lecointre <philippe.lecointre@thalesgroup.com> > > Reviewed-by: Simon Lenain <simon.lenain@thalesgroup.com> > > Upstream usually doesn't put much weight on such internal reviews. No > harm here, as the patch is simple enough and I do review it before applying. > Just as a hint for the future: if you want maintainers to take such reviews into > account for speeding up the adoption of a patch, do the review on the public > mailing lists. > Do you think an "Acked-by:" is better for our internal review ? Regards, Simon > > --- > > v3: > > - Rework to match initial feedback > > --- > > drivers/gpu/drm/etnaviv/etnaviv_gpu.c | 37 > > +++++++++++++++++++++++++++ > drivers/gpu/drm/etnaviv/etnaviv_gpu.h | > > 1 + > > 2 files changed, 38 insertions(+) > > > > diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c > > b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c > > index 2d4c112ce033..1961ebac315a 100644 > > --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c > > +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c > > @@ -13,6 +13,7 @@ > > #include <linux/platform_device.h> > > #include <linux/pm_runtime.h> > > #include <linux/regulator/consumer.h> > > +#include <linux/reset.h> > > #include <linux/thermal.h> > > > > #include "etnaviv_cmdbuf.h" > > @@ -172,6 +173,25 @@ int etnaviv_gpu_get_param(struct etnaviv_gpu > *gpu, u32 param, u64 *value) > > return 0; > > } > > > > +static int etnaviv_gpu_reset_deassert(struct etnaviv_gpu *gpu) { > > + int ret; > > + > > + /* 32 core clock cycles (slowest clock) required before deassertion */ > > + /* 1 microsecond might match all implementations without > computation > > +*/ > > I missed to mention this before, as I was focused on the technical > side: this is not the multiline comment style used in the kernel/etnaviv. > Please use the same style as already found in this file. > > > + usleep_range(1, 2); > > + > > + ret = reset_control_deassert(gpu->rst); > > + if (ret) > > + return ret; > > + > > + /* 128 core clock cycles (slowest clock) required before any activity > on AHB */ > > + /* 1 microsecond might match all implementations without > computation */ > > + usleep_range(1, 2); > > + > > + return 0; > > +} > > + > > static inline bool etnaviv_is_model_rev(struct etnaviv_gpu *gpu, u32 > > model, u32 revision) { > > return gpu->identity.model == model && @@ -799,6 +819,12 @@ int > > etnaviv_gpu_init(struct etnaviv_gpu *gpu) > > goto pm_put; > > } > > > > + ret = etnaviv_gpu_reset_deassert(gpu); > > + if (ret) { > > + dev_err(gpu->dev, "GPU reset deassert failed\n"); > > + goto fail; > > + } > > + > > etnaviv_hw_identify(gpu); > > > > if (gpu->identity.model == 0) { > > @@ -1860,6 +1886,17 @@ static int etnaviv_gpu_platform_probe(struct > platform_device *pdev) > > if (IS_ERR(gpu->mmio)) > > return PTR_ERR(gpu->mmio); > > > > + > > + /* Get Reset: */ > > + gpu->rst = devm_reset_control_get_optional_exclusive(&pdev- > >dev, NULL); > > + if (IS_ERR(gpu->rst)) > > + return dev_err_probe(dev, PTR_ERR(gpu->rst), > > + "failed to get reset\n"); > > + > > + err = reset_control_assert(gpu->rst); > > + if (err) > > + return dev_err_probe(dev, err, "failed to assert reset\n"); > > + > > /* Get Interrupt: */ > > gpu->irq = platform_get_irq(pdev, 0); > > if (gpu->irq < 0) > > diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.h > > b/drivers/gpu/drm/etnaviv/etnaviv_gpu.h > > index 4d8a7d48ade3..0985ea548b82 100644 > > --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.h > > +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.h > > @@ -158,6 +158,7 @@ struct etnaviv_gpu { > > struct clk *clk_reg; > > struct clk *clk_core; > > struct clk *clk_shader; > > + struct reset_control *rst; > > This needs a forward declaration of struct reset_control in the header, to > avoid build failures if headers are included in a different order. > Please put them right next to the existing ones for regulator and clk. > > Other than that, patch looks good to me. > > Regards, > Lucas > > > > > unsigned int freq_scale; > > unsigned int fe_waitcycles;
diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c index 2d4c112ce033..1961ebac315a 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c @@ -13,6 +13,7 @@ #include <linux/platform_device.h> #include <linux/pm_runtime.h> #include <linux/regulator/consumer.h> +#include <linux/reset.h> #include <linux/thermal.h> #include "etnaviv_cmdbuf.h" @@ -172,6 +173,25 @@ int etnaviv_gpu_get_param(struct etnaviv_gpu *gpu, u32 param, u64 *value) return 0; } +static int etnaviv_gpu_reset_deassert(struct etnaviv_gpu *gpu) +{ + int ret; + + /* 32 core clock cycles (slowest clock) required before deassertion */ + /* 1 microsecond might match all implementations without computation */ + usleep_range(1, 2); + + ret = reset_control_deassert(gpu->rst); + if (ret) + return ret; + + /* 128 core clock cycles (slowest clock) required before any activity on AHB */ + /* 1 microsecond might match all implementations without computation */ + usleep_range(1, 2); + + return 0; +} + static inline bool etnaviv_is_model_rev(struct etnaviv_gpu *gpu, u32 model, u32 revision) { return gpu->identity.model == model && @@ -799,6 +819,12 @@ int etnaviv_gpu_init(struct etnaviv_gpu *gpu) goto pm_put; } + ret = etnaviv_gpu_reset_deassert(gpu); + if (ret) { + dev_err(gpu->dev, "GPU reset deassert failed\n"); + goto fail; + } + etnaviv_hw_identify(gpu); if (gpu->identity.model == 0) { @@ -1860,6 +1886,17 @@ static int etnaviv_gpu_platform_probe(struct platform_device *pdev) if (IS_ERR(gpu->mmio)) return PTR_ERR(gpu->mmio); + + /* Get Reset: */ + gpu->rst = devm_reset_control_get_optional_exclusive(&pdev->dev, NULL); + if (IS_ERR(gpu->rst)) + return dev_err_probe(dev, PTR_ERR(gpu->rst), + "failed to get reset\n"); + + err = reset_control_assert(gpu->rst); + if (err) + return dev_err_probe(dev, err, "failed to assert reset\n"); + /* Get Interrupt: */ gpu->irq = platform_get_irq(pdev, 0); if (gpu->irq < 0) diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.h b/drivers/gpu/drm/etnaviv/etnaviv_gpu.h index 4d8a7d48ade3..0985ea548b82 100644 --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.h +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.h @@ -158,6 +158,7 @@ struct etnaviv_gpu { struct clk *clk_reg; struct clk *clk_core; struct clk *clk_shader; + struct reset_control *rst; unsigned int freq_scale; unsigned int fe_waitcycles;