| Message ID | 20210107134101.195426-1-paul.kocialkowski@bootlin.com |
|---|---|
| Headers | show |
| Series | Rockchip PX30 RGA and VPU support | expand |
Hi Paul, On Thu, 2021-01-07 at 14:40 +0100, Paul Kocialkowski wrote: > The Rockchip PX30 SoC has a Hantro VPU that features a decoder (VDPU2) > and an encoder (VEPU2). It is similar to the RK3399's VPU but takes an > extra clock (SCLK). > > Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com> > --- > .../bindings/media/rockchip-vpu.yaml | 25 +++++++++++++------ > 1 file changed, 17 insertions(+), 8 deletions(-) > > diff --git a/Documentation/devicetree/bindings/media/rockchip-vpu.yaml b/Documentation/devicetree/bindings/media/rockchip-vpu.yaml > index c81dbc3e8960..c446b9ead21b 100644 > --- a/Documentation/devicetree/bindings/media/rockchip-vpu.yaml > +++ b/Documentation/devicetree/bindings/media/rockchip-vpu.yaml > @@ -15,10 +15,13 @@ description: > > properties: > compatible: > - enum: > - - rockchip,rk3288-vpu > - - rockchip,rk3328-vpu > - - rockchip,rk3399-vpu > + oneOf: > + - const: rockchip,rk3288-vpu > + - const: rockchip,rk3328-vpu > + - const: rockchip,rk3399-vpu > + - items: > + - const: rockchip,px30-vpu > + - const: rockchip,rk3399-vpu > > reg: > maxItems: 1 > @@ -35,12 +38,18 @@ properties: > - const: vdpu > > clocks: > - maxItems: 2 > + minItems: 2 > + maxItems: 3 > > clock-names: > - items: > - - const: aclk > - - const: hclk > + oneOf: > + - items: > + - const: aclk > + - const: hclk > + - items: > + - const: aclk > + - const: hclk > + - const: sclk You could make this: clock-names: minItems: 2 items: - const: aclk - const: hclk - const: sclk And then: allOf: - if: properties: compatible: contains: const: rockchip,px30-vpu then: properties: clock-names: minItems: 3 to make sure each variant has the correct clocks set. regards Philipp
On Thu, 07 Jan 2021 14:40:57 +0100, Paul Kocialkowski wrote: > Add a new compatible for the PX30 Rockchip SoC, which also features > a RGA block. It is compatible with the RK3288 RGA block. > > Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com> > --- > Documentation/devicetree/bindings/media/rockchip-rga.yaml | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > My bot found errors running 'make dt_binding_check' on your patch: yamllint warnings/errors: ./Documentation/devicetree/bindings/media/rockchip-rga.yaml:25:13: [warning] wrong indentation: expected 14 but found 12 (indentation) dtschema/dtc warnings/errors: See https://patchwork.ozlabs.org/patch/1423299 This check can fail if there are any dependencies. The base for a patch series is generally the most recent rc1. If you already ran 'make dt_binding_check' and didn't see the above error(s), then make sure 'yamllint' is installed and dt-schema is up to date: pip3 install dtschema --upgrade Please check and re-submit.
Hi Paul, Happy to see this patch. It was on my TODO list, but I hadn't had time to bringup my rk3326 device. A few comments. On Thu, 2021-01-07 at 14:41 +0100, Paul Kocialkowski wrote: > The PX30 SoC includes both the VDPU2 and VEPU2 blocks which are similar > to the RK3399 (Hantro G1/H1 with shuffled registers). > > Besides taking an extra clock, it also shares an interrupt with the IOMMU > so it's necessary to request the interrupt shared. > Could you clarify on the commit description which iommu device interrupt line is being shared? > Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com> > --- > drivers/staging/media/hantro/hantro_drv.c | 5 +++-- > drivers/staging/media/hantro/hantro_hw.h | 1 + > drivers/staging/media/hantro/rk3399_vpu_hw.c | 21 ++++++++++++++++++++ > 3 files changed, 25 insertions(+), 2 deletions(-) > > diff --git a/drivers/staging/media/hantro/hantro_drv.c b/drivers/staging/media/hantro/hantro_drv.c > index e5f200e64993..076a7782b476 100644 > --- a/drivers/staging/media/hantro/hantro_drv.c > +++ b/drivers/staging/media/hantro/hantro_drv.c > @@ -472,6 +472,7 @@ static const struct v4l2_file_operations hantro_fops = { > > static const struct of_device_id of_hantro_match[] = { > #ifdef CONFIG_VIDEO_HANTRO_ROCKCHIP > + { .compatible = "rockchip,px30-vpu", .data = &px30_vpu_variant, }, > { .compatible = "rockchip,rk3399-vpu", .data = &rk3399_vpu_variant, }, > { .compatible = "rockchip,rk3328-vpu", .data = &rk3328_vpu_variant, }, > { .compatible = "rockchip,rk3288-vpu", .data = &rk3288_vpu_variant, }, > @@ -796,8 +797,8 @@ static int hantro_probe(struct platform_device *pdev) > return -ENXIO; > > ret = devm_request_irq(vpu->dev, irq, > - vpu->variant->irqs[i].handler, 0, > - dev_name(vpu->dev), vpu); > + vpu->variant->irqs[i].handler, > + IRQF_SHARED, dev_name(vpu->dev), vpu); Maybe this irq flag should be part of vpu->variant? It sounds like an IP block integration specific thing. Also, you will need a px30-specific interrupt handler now, since the rk3399 one is not shared-friendly. > if (ret) { > dev_err(vpu->dev, "Could not request %s IRQ.\n", > irq_name); > diff --git a/drivers/staging/media/hantro/hantro_hw.h b/drivers/staging/media/hantro/hantro_hw.h > index 34c9e4649a25..07f516fd7a2e 100644 > --- a/drivers/staging/media/hantro/hantro_hw.h > +++ b/drivers/staging/media/hantro/hantro_hw.h > @@ -148,6 +148,7 @@ enum hantro_enc_fmt { > RK3288_VPU_ENC_FMT_UYVY422 = 3, > }; > > +extern const struct hantro_variant px30_vpu_variant; > extern const struct hantro_variant rk3399_vpu_variant; > extern const struct hantro_variant rk3328_vpu_variant; > extern const struct hantro_variant rk3288_vpu_variant; > diff --git a/drivers/staging/media/hantro/rk3399_vpu_hw.c b/drivers/staging/media/hantro/rk3399_vpu_hw.c > index 7a7962cf771e..4112f98baa60 100644 > --- a/drivers/staging/media/hantro/rk3399_vpu_hw.c > +++ b/drivers/staging/media/hantro/rk3399_vpu_hw.c Perhaps it's time to rename this to rockchip_vpu_hw.c, and merge rk3288 and rk3399? It's a nitpick, though. > @@ -220,3 +220,24 @@ const struct hantro_variant rk3328_vpu_variant = { > .clk_names = rk3399_clk_names, > .num_clocks = ARRAY_SIZE(rk3399_clk_names), > }; > + > +static const char * const px30_clk_names[] = { > + "aclk", "hclk", "sclk" > +}; > + > +const struct hantro_variant px30_vpu_variant = { > + .enc_offset = 0x0, > + .enc_fmts = rk3399_vpu_enc_fmts, > + .num_enc_fmts = ARRAY_SIZE(rk3399_vpu_enc_fmts), > + .dec_offset = 0x400, > + .dec_fmts = rk3399_vpu_dec_fmts, > + .num_dec_fmts = ARRAY_SIZE(rk3399_vpu_dec_fmts), > + .codec = HANTRO_JPEG_ENCODER | HANTRO_MPEG2_DECODER | > + HANTRO_VP8_DECODER, > + .codec_ops = rk3399_vpu_codec_ops, > + .irqs = rk3399_irqs, > + .num_irqs = ARRAY_SIZE(rk3399_irqs), > + .init = rk3399_vpu_hw_init, > + .clk_names = px30_clk_names, > + .num_clocks = ARRAY_SIZE(px30_clk_names) > +}; Thanks, Ezequiel
Hi Ezequiel, On Thu 07 Jan 21, 16:08, Ezequiel Garcia wrote: > Happy to see this patch. It was on my TODO list, > but I hadn't had time to bringup my rk3326 device. Same here, I just had an occasion to use it again these days so I jumped on it! > A few comments. > > On Thu, 2021-01-07 at 14:41 +0100, Paul Kocialkowski wrote: > > The PX30 SoC includes both the VDPU2 and VEPU2 blocks which are similar > > to the RK3399 (Hantro G1/H1 with shuffled registers). > > > > Besides taking an extra clock, it also shares an interrupt with the IOMMU > > so it's necessary to request the interrupt shared. > > > > Could you clarify on the commit description which iommu device interrupt > line is being shared? Sure! It's IRQ 79 of the GIC that's shared with vopl_mmu. It's not very obvious in the dt commit. > > Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com> > > --- > > drivers/staging/media/hantro/hantro_drv.c | 5 +++-- > > drivers/staging/media/hantro/hantro_hw.h | 1 + > > drivers/staging/media/hantro/rk3399_vpu_hw.c | 21 ++++++++++++++++++++ > > 3 files changed, 25 insertions(+), 2 deletions(-) > > > > diff --git a/drivers/staging/media/hantro/hantro_drv.c b/drivers/staging/media/hantro/hantro_drv.c > > index e5f200e64993..076a7782b476 100644 > > --- a/drivers/staging/media/hantro/hantro_drv.c > > +++ b/drivers/staging/media/hantro/hantro_drv.c > > @@ -472,6 +472,7 @@ static const struct v4l2_file_operations hantro_fops = { > > > > static const struct of_device_id of_hantro_match[] = { > > #ifdef CONFIG_VIDEO_HANTRO_ROCKCHIP > > + { .compatible = "rockchip,px30-vpu", .data = &px30_vpu_variant, }, > > { .compatible = "rockchip,rk3399-vpu", .data = &rk3399_vpu_variant, }, > > { .compatible = "rockchip,rk3328-vpu", .data = &rk3328_vpu_variant, }, > > { .compatible = "rockchip,rk3288-vpu", .data = &rk3288_vpu_variant, }, > > @@ -796,8 +797,8 @@ static int hantro_probe(struct platform_device *pdev) > > return -ENXIO; > > > > ret = devm_request_irq(vpu->dev, irq, > > - vpu->variant->irqs[i].handler, 0, > > - dev_name(vpu->dev), vpu); > > + vpu->variant->irqs[i].handler, > > + IRQF_SHARED, dev_name(vpu->dev), vpu); > > Maybe this irq flag should be part of vpu->variant? It sounds like an IP block > integration specific thing. Ah right, I agree that it would be justified. But it would also be simple to just fix the irq handlers and assume this can generally be the case, because it feels like a bit of a detail to justify a flag. Do you think this could be a safe/workable assumption? > Also, you will need a px30-specific interrupt handler now, > since the rk3399 one is not shared-friendly. Yeah I realize I haven't been very careful there and didn't really check that the IOMMU driver is really safe to handle shared interrupts either. I'll take a look a that when crafting v2. > > if (ret) { > > dev_err(vpu->dev, "Could not request %s IRQ.\n", > > irq_name); > > diff --git a/drivers/staging/media/hantro/hantro_hw.h b/drivers/staging/media/hantro/hantro_hw.h > > index 34c9e4649a25..07f516fd7a2e 100644 > > --- a/drivers/staging/media/hantro/hantro_hw.h > > +++ b/drivers/staging/media/hantro/hantro_hw.h > > @@ -148,6 +148,7 @@ enum hantro_enc_fmt { > > RK3288_VPU_ENC_FMT_UYVY422 = 3, > > }; > > > > +extern const struct hantro_variant px30_vpu_variant; > > extern const struct hantro_variant rk3399_vpu_variant; > > extern const struct hantro_variant rk3328_vpu_variant; > > extern const struct hantro_variant rk3288_vpu_variant; > > diff --git a/drivers/staging/media/hantro/rk3399_vpu_hw.c b/drivers/staging/media/hantro/rk3399_vpu_hw.c > > index 7a7962cf771e..4112f98baa60 100644 > > --- a/drivers/staging/media/hantro/rk3399_vpu_hw.c > > +++ b/drivers/staging/media/hantro/rk3399_vpu_hw.c > > Perhaps it's time to rename this to rockchip_vpu_hw.c, > and merge rk3288 and rk3399? It's a nitpick, though. Haha, I was thinking the exact same thing but wasn't sure it would be welcome! I was thinking of rockchip_vpu2_hw.c or rockchip_vdpu2_hw.c since that's apparently how it's called in Rockchip terminology: VDPU2 and VEPU2 for the Hantro G1 and H1 with the shuffled register layout. The rk3288 stuff is probably VDPU1/VEPU1 and we might want to rename it accordingly as well. Cheers and thanks for the review! Paul > > @@ -220,3 +220,24 @@ const struct hantro_variant rk3328_vpu_variant = { > > .clk_names = rk3399_clk_names, > > .num_clocks = ARRAY_SIZE(rk3399_clk_names), > > }; > > + > > +static const char * const px30_clk_names[] = { > > + "aclk", "hclk", "sclk" > > +}; > > + > > +const struct hantro_variant px30_vpu_variant = { > > + .enc_offset = 0x0, > > + .enc_fmts = rk3399_vpu_enc_fmts, > > + .num_enc_fmts = ARRAY_SIZE(rk3399_vpu_enc_fmts), > > + .dec_offset = 0x400, > > + .dec_fmts = rk3399_vpu_dec_fmts, > > + .num_dec_fmts = ARRAY_SIZE(rk3399_vpu_dec_fmts), > > + .codec = HANTRO_JPEG_ENCODER | HANTRO_MPEG2_DECODER | > > + HANTRO_VP8_DECODER, > > + .codec_ops = rk3399_vpu_codec_ops, > > + .irqs = rk3399_irqs, > > + .num_irqs = ARRAY_SIZE(rk3399_irqs), > > + .init = rk3399_vpu_hw_init, > > + .clk_names = px30_clk_names, > > + .num_clocks = ARRAY_SIZE(px30_clk_names) > > +}; > > Thanks, > Ezequiel > -- Paul Kocialkowski, Bootlin Embedded Linux and kernel engineering https://bootlin.com
Am Freitag, 8. Januar 2021, 10:05:16 CET schrieb Paul Kocialkowski: > Hi Ezequiel, > > On Thu 07 Jan 21, 16:08, Ezequiel Garcia wrote: > > Happy to see this patch. It was on my TODO list, > > but I hadn't had time to bringup my rk3326 device. > > Same here, I just had an occasion to use it again these days so I jumped > on it! > > > A few comments. > > > > On Thu, 2021-01-07 at 14:41 +0100, Paul Kocialkowski wrote: > > > The PX30 SoC includes both the VDPU2 and VEPU2 blocks which are similar > > > to the RK3399 (Hantro G1/H1 with shuffled registers). > > > > > > Besides taking an extra clock, it also shares an interrupt with the IOMMU > > > so it's necessary to request the interrupt shared. > > > > > > > Could you clarify on the commit description which iommu device interrupt > > line is being shared? > > Sure! It's IRQ 79 of the GIC that's shared with vopl_mmu. > It's not very obvious in the dt commit. Having looked through the docs again, I think that the vopl_mmu using irq 79 is just a mistake: (1) in general vop and vop-mmu use the same irq (78 in that case) (2) Rockchip does seem to have fixed that in their 4.19 tree as well: https://github.com/rockchip-linux/kernel/blob/develop-4.19/arch/arm64/boot/dts/rockchip/px30.dtsi#L1598 So to me it looks like this doesn't need to be shared and instead "simply" the px30 dtsi fixed ;-) Heiko > > > Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com> > > > --- > > > drivers/staging/media/hantro/hantro_drv.c | 5 +++-- > > > drivers/staging/media/hantro/hantro_hw.h | 1 + > > > drivers/staging/media/hantro/rk3399_vpu_hw.c | 21 ++++++++++++++++++++ > > > 3 files changed, 25 insertions(+), 2 deletions(-) > > > > > > diff --git a/drivers/staging/media/hantro/hantro_drv.c b/drivers/staging/media/hantro/hantro_drv.c > > > index e5f200e64993..076a7782b476 100644 > > > --- a/drivers/staging/media/hantro/hantro_drv.c > > > +++ b/drivers/staging/media/hantro/hantro_drv.c > > > @@ -472,6 +472,7 @@ static const struct v4l2_file_operations hantro_fops = { > > > > > > static const struct of_device_id of_hantro_match[] = { > > > #ifdef CONFIG_VIDEO_HANTRO_ROCKCHIP > > > + { .compatible = "rockchip,px30-vpu", .data = &px30_vpu_variant, }, > > > { .compatible = "rockchip,rk3399-vpu", .data = &rk3399_vpu_variant, }, > > > { .compatible = "rockchip,rk3328-vpu", .data = &rk3328_vpu_variant, }, > > > { .compatible = "rockchip,rk3288-vpu", .data = &rk3288_vpu_variant, }, > > > @@ -796,8 +797,8 @@ static int hantro_probe(struct platform_device *pdev) > > > return -ENXIO; > > > > > > ret = devm_request_irq(vpu->dev, irq, > > > - vpu->variant->irqs[i].handler, 0, > > > - dev_name(vpu->dev), vpu); > > > + vpu->variant->irqs[i].handler, > > > + IRQF_SHARED, dev_name(vpu->dev), vpu); > > > > Maybe this irq flag should be part of vpu->variant? It sounds like an IP block > > integration specific thing. > > Ah right, I agree that it would be justified. But it would also be simple to > just fix the irq handlers and assume this can generally be the case, because it > feels like a bit of a detail to justify a flag. > > Do you think this could be a safe/workable assumption? > > > Also, you will need a px30-specific interrupt handler now, > > since the rk3399 one is not shared-friendly. > > Yeah I realize I haven't been very careful there and didn't really check that > the IOMMU driver is really safe to handle shared interrupts either. I'll take > a look a that when crafting v2. > > > > if (ret) { > > > dev_err(vpu->dev, "Could not request %s IRQ.\n", > > > irq_name); > > > diff --git a/drivers/staging/media/hantro/hantro_hw.h b/drivers/staging/media/hantro/hantro_hw.h > > > index 34c9e4649a25..07f516fd7a2e 100644 > > > --- a/drivers/staging/media/hantro/hantro_hw.h > > > +++ b/drivers/staging/media/hantro/hantro_hw.h > > > @@ -148,6 +148,7 @@ enum hantro_enc_fmt { > > > RK3288_VPU_ENC_FMT_UYVY422 = 3, > > > }; > > > > > > +extern const struct hantro_variant px30_vpu_variant; > > > extern const struct hantro_variant rk3399_vpu_variant; > > > extern const struct hantro_variant rk3328_vpu_variant; > > > extern const struct hantro_variant rk3288_vpu_variant; > > > diff --git a/drivers/staging/media/hantro/rk3399_vpu_hw.c b/drivers/staging/media/hantro/rk3399_vpu_hw.c > > > index 7a7962cf771e..4112f98baa60 100644 > > > --- a/drivers/staging/media/hantro/rk3399_vpu_hw.c > > > +++ b/drivers/staging/media/hantro/rk3399_vpu_hw.c > > > > Perhaps it's time to rename this to rockchip_vpu_hw.c, > > and merge rk3288 and rk3399? It's a nitpick, though. > > Haha, I was thinking the exact same thing but wasn't sure it would be welcome! > > I was thinking of rockchip_vpu2_hw.c or rockchip_vdpu2_hw.c since that's > apparently how it's called in Rockchip terminology: VDPU2 and VEPU2 for the > Hantro G1 and H1 with the shuffled register layout. The rk3288 stuff is > probably VDPU1/VEPU1 and we might want to rename it accordingly as well. > > Cheers and thanks for the review! > > Paul > > > > @@ -220,3 +220,24 @@ const struct hantro_variant rk3328_vpu_variant = { > > > .clk_names = rk3399_clk_names, > > > .num_clocks = ARRAY_SIZE(rk3399_clk_names), > > > }; > > > + > > > +static const char * const px30_clk_names[] = { > > > + "aclk", "hclk", "sclk" > > > +}; > > > + > > > +const struct hantro_variant px30_vpu_variant = { > > > + .enc_offset = 0x0, > > > + .enc_fmts = rk3399_vpu_enc_fmts, > > > + .num_enc_fmts = ARRAY_SIZE(rk3399_vpu_enc_fmts), > > > + .dec_offset = 0x400, > > > + .dec_fmts = rk3399_vpu_dec_fmts, > > > + .num_dec_fmts = ARRAY_SIZE(rk3399_vpu_dec_fmts), > > > + .codec = HANTRO_JPEG_ENCODER | HANTRO_MPEG2_DECODER | > > > + HANTRO_VP8_DECODER, > > > + .codec_ops = rk3399_vpu_codec_ops, > > > + .irqs = rk3399_irqs, > > > + .num_irqs = ARRAY_SIZE(rk3399_irqs), > > > + .init = rk3399_vpu_hw_init, > > > + .clk_names = px30_clk_names, > > > + .num_clocks = ARRAY_SIZE(px30_clk_names) > > > +}; > > > > Thanks, > > Ezequiel > > > >
Am Freitag, 8. Januar 2021, 11:48:26 CET schrieb Heiko Stübner: > Am Freitag, 8. Januar 2021, 10:05:16 CET schrieb Paul Kocialkowski: > > Hi Ezequiel, > > > > On Thu 07 Jan 21, 16:08, Ezequiel Garcia wrote: > > > Happy to see this patch. It was on my TODO list, > > > but I hadn't had time to bringup my rk3326 device. > > > > Same here, I just had an occasion to use it again these days so I jumped > > on it! > > > > > A few comments. > > > > > > On Thu, 2021-01-07 at 14:41 +0100, Paul Kocialkowski wrote: > > > > The PX30 SoC includes both the VDPU2 and VEPU2 blocks which are similar > > > > to the RK3399 (Hantro G1/H1 with shuffled registers). > > > > > > > > Besides taking an extra clock, it also shares an interrupt with the IOMMU > > > > so it's necessary to request the interrupt shared. > > > > > > > > > > Could you clarify on the commit description which iommu device interrupt > > > line is being shared? > > > > Sure! It's IRQ 79 of the GIC that's shared with vopl_mmu. > > It's not very obvious in the dt commit. > > Having looked through the docs again, I think that the vopl_mmu using > irq 79 is just a mistake: > > (1) in general vop and vop-mmu use the same irq (78 in that case) > (2) Rockchip does seem to have fixed that in their 4.19 tree as well: > https://github.com/rockchip-linux/kernel/blob/develop-4.19/arch/arm64/boot/dts/rockchip/px30.dtsi#L1598 (3) https://github.com/rockchip-linux/kernel/commit/391a5c5f96d177896f9fe92ca1c83e00f4352191 ;-) > > So to me it looks like this doesn't need to be shared and instead > "simply" the px30 dtsi fixed ;-) > > > Heiko > > > > > > Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com> > > > > --- > > > > drivers/staging/media/hantro/hantro_drv.c | 5 +++-- > > > > drivers/staging/media/hantro/hantro_hw.h | 1 + > > > > drivers/staging/media/hantro/rk3399_vpu_hw.c | 21 ++++++++++++++++++++ > > > > 3 files changed, 25 insertions(+), 2 deletions(-) > > > > > > > > diff --git a/drivers/staging/media/hantro/hantro_drv.c b/drivers/staging/media/hantro/hantro_drv.c > > > > index e5f200e64993..076a7782b476 100644 > > > > --- a/drivers/staging/media/hantro/hantro_drv.c > > > > +++ b/drivers/staging/media/hantro/hantro_drv.c > > > > @@ -472,6 +472,7 @@ static const struct v4l2_file_operations hantro_fops = { > > > > > > > > static const struct of_device_id of_hantro_match[] = { > > > > #ifdef CONFIG_VIDEO_HANTRO_ROCKCHIP > > > > + { .compatible = "rockchip,px30-vpu", .data = &px30_vpu_variant, }, > > > > { .compatible = "rockchip,rk3399-vpu", .data = &rk3399_vpu_variant, }, > > > > { .compatible = "rockchip,rk3328-vpu", .data = &rk3328_vpu_variant, }, > > > > { .compatible = "rockchip,rk3288-vpu", .data = &rk3288_vpu_variant, }, > > > > @@ -796,8 +797,8 @@ static int hantro_probe(struct platform_device *pdev) > > > > return -ENXIO; > > > > > > > > ret = devm_request_irq(vpu->dev, irq, > > > > - vpu->variant->irqs[i].handler, 0, > > > > - dev_name(vpu->dev), vpu); > > > > + vpu->variant->irqs[i].handler, > > > > + IRQF_SHARED, dev_name(vpu->dev), vpu); > > > > > > Maybe this irq flag should be part of vpu->variant? It sounds like an IP block > > > integration specific thing. > > > > Ah right, I agree that it would be justified. But it would also be simple to > > just fix the irq handlers and assume this can generally be the case, because it > > feels like a bit of a detail to justify a flag. > > > > Do you think this could be a safe/workable assumption? > > > > > Also, you will need a px30-specific interrupt handler now, > > > since the rk3399 one is not shared-friendly. > > > > Yeah I realize I haven't been very careful there and didn't really check that > > the IOMMU driver is really safe to handle shared interrupts either. I'll take > > a look a that when crafting v2. > > > > > > if (ret) { > > > > dev_err(vpu->dev, "Could not request %s IRQ.\n", > > > > irq_name); > > > > diff --git a/drivers/staging/media/hantro/hantro_hw.h b/drivers/staging/media/hantro/hantro_hw.h > > > > index 34c9e4649a25..07f516fd7a2e 100644 > > > > --- a/drivers/staging/media/hantro/hantro_hw.h > > > > +++ b/drivers/staging/media/hantro/hantro_hw.h > > > > @@ -148,6 +148,7 @@ enum hantro_enc_fmt { > > > > RK3288_VPU_ENC_FMT_UYVY422 = 3, > > > > }; > > > > > > > > +extern const struct hantro_variant px30_vpu_variant; > > > > extern const struct hantro_variant rk3399_vpu_variant; > > > > extern const struct hantro_variant rk3328_vpu_variant; > > > > extern const struct hantro_variant rk3288_vpu_variant; > > > > diff --git a/drivers/staging/media/hantro/rk3399_vpu_hw.c b/drivers/staging/media/hantro/rk3399_vpu_hw.c > > > > index 7a7962cf771e..4112f98baa60 100644 > > > > --- a/drivers/staging/media/hantro/rk3399_vpu_hw.c > > > > +++ b/drivers/staging/media/hantro/rk3399_vpu_hw.c > > > > > > Perhaps it's time to rename this to rockchip_vpu_hw.c, > > > and merge rk3288 and rk3399? It's a nitpick, though. > > > > Haha, I was thinking the exact same thing but wasn't sure it would be welcome! > > > > I was thinking of rockchip_vpu2_hw.c or rockchip_vdpu2_hw.c since that's > > apparently how it's called in Rockchip terminology: VDPU2 and VEPU2 for the > > Hantro G1 and H1 with the shuffled register layout. The rk3288 stuff is > > probably VDPU1/VEPU1 and we might want to rename it accordingly as well. > > > > Cheers and thanks for the review! > > > > Paul > > > > > > @@ -220,3 +220,24 @@ const struct hantro_variant rk3328_vpu_variant = { > > > > .clk_names = rk3399_clk_names, > > > > .num_clocks = ARRAY_SIZE(rk3399_clk_names), > > > > }; > > > > + > > > > +static const char * const px30_clk_names[] = { > > > > + "aclk", "hclk", "sclk" > > > > +}; > > > > + > > > > +const struct hantro_variant px30_vpu_variant = { > > > > + .enc_offset = 0x0, > > > > + .enc_fmts = rk3399_vpu_enc_fmts, > > > > + .num_enc_fmts = ARRAY_SIZE(rk3399_vpu_enc_fmts), > > > > + .dec_offset = 0x400, > > > > + .dec_fmts = rk3399_vpu_dec_fmts, > > > > + .num_dec_fmts = ARRAY_SIZE(rk3399_vpu_dec_fmts), > > > > + .codec = HANTRO_JPEG_ENCODER | HANTRO_MPEG2_DECODER | > > > > + HANTRO_VP8_DECODER, > > > > + .codec_ops = rk3399_vpu_codec_ops, > > > > + .irqs = rk3399_irqs, > > > > + .num_irqs = ARRAY_SIZE(rk3399_irqs), > > > > + .init = rk3399_vpu_hw_init, > > > > + .clk_names = px30_clk_names, > > > > + .num_clocks = ARRAY_SIZE(px30_clk_names) > > > > +}; > > > > > > Thanks, > > > Ezequiel > > > > > > > > >
On Fri, 2021-01-08 at 10:05 +0100, Paul Kocialkowski wrote: > Hi Ezequiel, > > On Thu 07 Jan 21, 16:08, Ezequiel Garcia wrote: > > Happy to see this patch. It was on my TODO list, > > but I hadn't had time to bringup my rk3326 device. > > Same here, I just had an occasion to use it again these days so I jumped > on it! > > > A few comments. > > > > On Thu, 2021-01-07 at 14:41 +0100, Paul Kocialkowski wrote: > > > The PX30 SoC includes both the VDPU2 and VEPU2 blocks which are similar > > > to the RK3399 (Hantro G1/H1 with shuffled registers). > > > > > > Besides taking an extra clock, it also shares an interrupt with the IOMMU > > > so it's necessary to request the interrupt shared. > > > > > [..] > > > diff --git a/drivers/staging/media/hantro/hantro_hw.h b/drivers/staging/media/hantro/hantro_hw.h > > > index 34c9e4649a25..07f516fd7a2e 100644 > > > --- a/drivers/staging/media/hantro/hantro_hw.h > > > +++ b/drivers/staging/media/hantro/hantro_hw.h > > > @@ -148,6 +148,7 @@ enum hantro_enc_fmt { > > > RK3288_VPU_ENC_FMT_UYVY422 = 3, > > > }; > > > > > > +extern const struct hantro_variant px30_vpu_variant; > > > extern const struct hantro_variant rk3399_vpu_variant; > > > extern const struct hantro_variant rk3328_vpu_variant; > > > extern const struct hantro_variant rk3288_vpu_variant; > > > diff --git a/drivers/staging/media/hantro/rk3399_vpu_hw.c b/drivers/staging/media/hantro/rk3399_vpu_hw.c > > > index 7a7962cf771e..4112f98baa60 100644 > > > --- a/drivers/staging/media/hantro/rk3399_vpu_hw.c > > > +++ b/drivers/staging/media/hantro/rk3399_vpu_hw.c > > > > Perhaps it's time to rename this to rockchip_vpu_hw.c, > > and merge rk3288 and rk3399? It's a nitpick, though. > > Haha, I was thinking the exact same thing but wasn't sure it would be welcome! > > I was thinking of rockchip_vpu2_hw.c or rockchip_vdpu2_hw.c since that's > apparently how it's called in Rockchip terminology: VDPU2 and VEPU2 for the > Hantro G1 and H1 with the shuffled register layout. The rk3288 stuff is > probably VDPU1/VEPU1 and we might want to rename it accordingly as well. > I'd rather keep it simple as rockchip_vpu_hw.c and just throw in there all the rockchip stuff. Thanks, Ezequiel
Hi Philipp, On Thu 07 Jan 21, 15:55, Philipp Zabel wrote: > Hi Paul, > > On Thu, 2021-01-07 at 14:40 +0100, Paul Kocialkowski wrote: > > The Rockchip PX30 SoC has a Hantro VPU that features a decoder (VDPU2) > > and an encoder (VEPU2). It is similar to the RK3399's VPU but takes an > > extra clock (SCLK). > > > > Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com> > > --- > > .../bindings/media/rockchip-vpu.yaml | 25 +++++++++++++------ > > 1 file changed, 17 insertions(+), 8 deletions(-) > > > > diff --git a/Documentation/devicetree/bindings/media/rockchip-vpu.yaml b/Documentation/devicetree/bindings/media/rockchip-vpu.yaml > > index c81dbc3e8960..c446b9ead21b 100644 > > --- a/Documentation/devicetree/bindings/media/rockchip-vpu.yaml > > +++ b/Documentation/devicetree/bindings/media/rockchip-vpu.yaml > > @@ -15,10 +15,13 @@ description: > > > > properties: > > compatible: > > - enum: > > - - rockchip,rk3288-vpu > > - - rockchip,rk3328-vpu > > - - rockchip,rk3399-vpu > > + oneOf: > > + - const: rockchip,rk3288-vpu > > + - const: rockchip,rk3328-vpu > > + - const: rockchip,rk3399-vpu > > + - items: > > + - const: rockchip,px30-vpu > > + - const: rockchip,rk3399-vpu > > > > reg: > > maxItems: 1 > > @@ -35,12 +38,18 @@ properties: > > - const: vdpu > > > > clocks: > > - maxItems: 2 > > + minItems: 2 > > + maxItems: 3 > > > > clock-names: > > - items: > > - - const: aclk > > - - const: hclk > > + oneOf: > > + - items: > > + - const: aclk > > + - const: hclk > > + - items: > > + - const: aclk > > + - const: hclk > > + - const: sclk > > You could make this: > > clock-names: > minItems: 2 > items: > - const: aclk > - const: hclk > - const: sclk > > And then: > > allOf: > - if: > properties: > compatible: > contains: > const: rockchip,px30-vpu > then: > properties: > clock-names: > minItems: 3 > > to make sure each variant has the correct clocks set. Thanks for the suggestion! That's a finer grain that my proposal. Rob, is there a preference for how this case should be handled? Here, we want to specify an extra clock for the PX30 case. Cheers, Paul -- Paul Kocialkowski, Bootlin Embedded Linux and kernel engineering https://bootlin.com
Hi Paul, On Thu, 2021-01-07 at 14:40 +0100, Paul Kocialkowski wrote: > This series adds the required bits for RGA and VPU support on the > Rockchip PX30 SoC. > Do you plan to resend this series? Alex recently renamed [1] things so some tweaking will be needed, but it shouldn't be complicated. [1] https://lore.kernel.org/linux-media/20210614213215.99389-1-knaerzche@gmail.com/ It would be great to have support RK3326 and PX30 :) Kindly, Ezequiel