After commit f1ad5338a4d5 ("of: Fix "dma-ranges" handling for bus
controllers"), the dma-ranges is not allowed for dts leaf node.
but we still would like to separate the different masters into
different iova regions. Thus we adjust the internal flow, separate
the 16GB iova range by the master HW larbid/portid and add the
dma-ranges property in the parent "soc" node. This also could avoid
the users forget/abuse the iova regions.
The commit f1ad5338a4d5 doesn't introduce the regression for us in
upstream, currently mt8195 vcodec/jpeg use the parent/child node.
thus I don't add "Fixes:" tag.
In this series, I add functions for mt8192/mt8195/mt8186, mt8188 will
be in its special patchset. and the previous mt8173/mt8183...support
0-4GB only, no need this function.
Base on v6.2-rc3.
Yong Wu (10):
dt-bindings: media: mediatek,vcodec: Remove dma-ranges property
dt-bindings: media: mediatek,jpeg: Remove dma-ranges property
iommu/mediatek: Get regionid from larb/port id
iommu/mediatek: mt8195: Add larb_region_msk
iommu/mediatek: mt8186: add larb_region_msk
iommu/mediatek: mt8192: add larb_region_msk
iommu/mediatek: Add a gap for the iova regions
arm64: dts: mt8195: Add dma-ranges for the parent "soc" node
arm64: dts: mt8195: Remove the unnecessary dma-ranges
arm64: dts: mt8186: Add dma-ranges for the parent "soc" node
.../media/mediatek,mt8195-jpegdec.yaml | 7 --
.../media/mediatek,mt8195-jpegenc.yaml | 7 --
.../media/mediatek,vcodec-decoder.yaml | 5 -
.../media/mediatek,vcodec-encoder.yaml | 5 -
.../media/mediatek,vcodec-subdev-decoder.yaml | 7 --
.../bindings/media/mediatek-jpeg-encoder.yaml | 5 -
arch/arm64/boot/dts/mediatek/mt8186.dtsi | 1 +
arch/arm64/boot/dts/mediatek/mt8195.dtsi | 2 +-
drivers/iommu/mtk_iommu.c | 95 ++++++++++++++-----
9 files changed, 72 insertions(+), 62 deletions(-)
--
2.18.0
Mediatek IOMMU has already controlled the masters' iova ranges by the
master's larb/port id. then the dma-ranges property is unnecessary for
the master's node. the master is jpeg here.
Cc: Mauro Carvalho Chehab <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Krzysztof Kozlowski <[email protected]>
Cc: Matthias Brugger <[email protected]>
Cc: Bin Liu <[email protected]>
Cc: kyrie wu <[email protected]>
Cc: Xia Jiang <[email protected]>
Signed-off-by: Yong Wu <[email protected]>
---
.../devicetree/bindings/media/mediatek,mt8195-jpegdec.yaml | 7 -------
.../devicetree/bindings/media/mediatek,mt8195-jpegenc.yaml | 7 -------
.../devicetree/bindings/media/mediatek-jpeg-encoder.yaml | 5 -----
3 files changed, 19 deletions(-)
diff --git a/Documentation/devicetree/bindings/media/mediatek,mt8195-jpegdec.yaml b/Documentation/devicetree/bindings/media/mediatek,mt8195-jpegdec.yaml
index 71595c013dbb..e5448c60e3eb 100644
--- a/Documentation/devicetree/bindings/media/mediatek,mt8195-jpegdec.yaml
+++ b/Documentation/devicetree/bindings/media/mediatek,mt8195-jpegdec.yaml
@@ -26,11 +26,6 @@ properties:
Documentation/devicetree/bindings/iommu/mediatek,iommu.yaml for details.
Ports are according to the HW.
- dma-ranges:
- maxItems: 1
- description: |
- Describes the physical address space of IOMMU maps to memory.
-
"#address-cells":
const: 2
@@ -89,7 +84,6 @@ required:
- compatible
- power-domains
- iommus
- - dma-ranges
- ranges
additionalProperties: false
@@ -115,7 +109,6 @@ examples:
<&iommu_vpp M4U_PORT_L19_JPGDEC_BSDMA1>,
<&iommu_vpp M4U_PORT_L19_JPGDEC_BUFF_OFFSET1>,
<&iommu_vpp M4U_PORT_L19_JPGDEC_BUFF_OFFSET0>;
- dma-ranges = <0x1 0x0 0x0 0x40000000 0x0 0xfff00000>;
#address-cells = <2>;
#size-cells = <2>;
ranges;
diff --git a/Documentation/devicetree/bindings/media/mediatek,mt8195-jpegenc.yaml b/Documentation/devicetree/bindings/media/mediatek,mt8195-jpegenc.yaml
index 95990539f7c0..596186497b68 100644
--- a/Documentation/devicetree/bindings/media/mediatek,mt8195-jpegenc.yaml
+++ b/Documentation/devicetree/bindings/media/mediatek,mt8195-jpegenc.yaml
@@ -26,11 +26,6 @@ properties:
Documentation/devicetree/bindings/iommu/mediatek,iommu.yaml for details.
Ports are according to the HW.
- dma-ranges:
- maxItems: 1
- description: |
- Describes the physical address space of IOMMU maps to memory.
-
"#address-cells":
const: 2
@@ -89,7 +84,6 @@ required:
- compatible
- power-domains
- iommus
- - dma-ranges
- ranges
additionalProperties: false
@@ -113,7 +107,6 @@ examples:
<&iommu_vpp M4U_PORT_L20_JPGENC_C_RDMA>,
<&iommu_vpp M4U_PORT_L20_JPGENC_Q_TABLE>,
<&iommu_vpp M4U_PORT_L20_JPGENC_BSDMA>;
- dma-ranges = <0x1 0x0 0x0 0x40000000 0x0 0xfff00000>;
#address-cells = <2>;
#size-cells = <2>;
ranges;
diff --git a/Documentation/devicetree/bindings/media/mediatek-jpeg-encoder.yaml b/Documentation/devicetree/bindings/media/mediatek-jpeg-encoder.yaml
index c8412e8ab353..37800e1908cc 100644
--- a/Documentation/devicetree/bindings/media/mediatek-jpeg-encoder.yaml
+++ b/Documentation/devicetree/bindings/media/mediatek-jpeg-encoder.yaml
@@ -44,11 +44,6 @@ properties:
Documentation/devicetree/bindings/iommu/mediatek,iommu.yaml for details.
Ports are according to the HW.
- dma-ranges:
- maxItems: 1
- description: |
- Describes the physical address space of IOMMU maps to memory.
-
required:
- compatible
- reg
--
2.18.0
Add larb_region_msk for mt8195. we separate the 16GB iova regions
by each device's larbid/portid.
Refer to include/dt-bindings/memory/mt8195-memory-port.h
Signed-off-by: Yong Wu <[email protected]>
---
drivers/iommu/mtk_iommu.c | 17 +++++++++++++++++
1 file changed, 17 insertions(+)
diff --git a/drivers/iommu/mtk_iommu.c b/drivers/iommu/mtk_iommu.c
index d30020d0a190..d8be7584702f 100644
--- a/drivers/iommu/mtk_iommu.c
+++ b/drivers/iommu/mtk_iommu.c
@@ -1546,6 +1546,21 @@ static const struct mtk_iommu_plat_data mt8195_data_infra = {
.iova_region_nr = ARRAY_SIZE(single_domain),
};
+static const unsigned int mt8195_larb_region_msk[][32] = {
+ [0] = {~0, ~0, ~0, ~0}, /* Region0: all ports for larb0/1/2/3 */
+ [1] = {0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, ~0, ~0, ~0, ~0, ~0, /* Region1: larb19/20/21/22/23/24 */
+ ~0},
+ [2] = {0, 0, 0, 0, ~0, ~0, ~0, ~0, /* Region2: the other larbs. */
+ ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0,
+ ~0, ~0, 0, 0, 0, 0, 0, 0,
+ 0, ~0, ~0, ~0, ~0},
+ [3] = {0},
+ [4] = {[18] = BIT(0) | BIT(1)}, /* Only larb18 port0/1 */
+ [5] = {[18] = BIT(2) | BIT(3)}, /* Only larb18 port2/3 */
+};
+
static const struct mtk_iommu_plat_data mt8195_data_vdo = {
.m4u_plat = M4U_MT8195,
.flags = HAS_BCLK | HAS_SUB_COMM_2BITS | OUT_ORDER_WR_EN |
@@ -1556,6 +1571,7 @@ static const struct mtk_iommu_plat_data mt8195_data_vdo = {
.banks_enable = {true},
.iova_region = mt8192_multi_dom,
.iova_region_nr = ARRAY_SIZE(mt8192_multi_dom),
+ .larb_region_msk = mt8195_larb_region_msk,
.larbid_remap = {{2, 0}, {21}, {24}, {7}, {19}, {9, 10, 11},
{13, 17, 15/* 17b */, 25}, {5}},
};
@@ -1570,6 +1586,7 @@ static const struct mtk_iommu_plat_data mt8195_data_vpp = {
.banks_enable = {true},
.iova_region = mt8192_multi_dom,
.iova_region_nr = ARRAY_SIZE(mt8192_multi_dom),
+ .larb_region_msk = mt8195_larb_region_msk,
.larbid_remap = {{1}, {3},
{22, MTK_INVALID_LARBID, MTK_INVALID_LARBID, MTK_INVALID_LARBID, 23},
{8}, {20}, {12},
--
2.18.0
Add larb_region_msk for mt8186. we separate the 16GB iova regions
by each device's larbid/portid.
Note: larb5/6/10/12/14/15/18 connect nothing in this SoC.
Refer to include/dt-bindings/memory/mt8186-memory-port.h
Signed-off-by: Yong Wu <[email protected]>
---
drivers/iommu/mtk_iommu.c | 13 +++++++++++++
1 file changed, 13 insertions(+)
diff --git a/drivers/iommu/mtk_iommu.c b/drivers/iommu/mtk_iommu.c
index d8be7584702f..ffd0203f515e 100644
--- a/drivers/iommu/mtk_iommu.c
+++ b/drivers/iommu/mtk_iommu.c
@@ -1504,6 +1504,18 @@ static const struct mtk_iommu_plat_data mt8183_data = {
.larbid_remap = {{0}, {4}, {5}, {6}, {7}, {2}, {3}, {1}},
};
+static const unsigned int mt8186_larb_region_msk[][32] = {
+ [0] = {~0, ~0, ~0}, /* Region0: all ports for larb0/1/2 */
+ [1] = {0, 0, 0, 0, ~0, 0, 0, ~0}, /* Region1: larb4/7 */
+ [2] = {0, 0, 0, 0, 0, 0, 0, 0, /* Region2: larb8/9/11/13/16/17/19/20 */
+ ~0, ~0, 0, ~0, 0, ~(u32)(BIT(9) | BIT(10)), 0, 0,
+ /* larb13: the other ports except port9/10 */
+ ~0, ~0, 0, ~0, ~0},
+ [3] = {0},
+ [4] = {[13] = BIT(9) | BIT(10)}, /* larb13 port9/10 */
+ [5] = {[14] = ~0}, /* larb14 */
+};
+
static const struct mtk_iommu_plat_data mt8186_data_mm = {
.m4u_plat = M4U_MT8186,
.flags = HAS_BCLK | HAS_SUB_COMM_2BITS | OUT_ORDER_WR_EN |
@@ -1516,6 +1528,7 @@ static const struct mtk_iommu_plat_data mt8186_data_mm = {
.banks_enable = {true},
.iova_region = mt8192_multi_dom,
.iova_region_nr = ARRAY_SIZE(mt8192_multi_dom),
+ .larb_region_msk = mt8186_larb_region_msk,
};
static const struct mtk_iommu_plat_data mt8192_data = {
--
2.18.0
After commit f1ad5338a4d5 ("of: Fix "dma-ranges" handling for bus
controllers"), the dma-ranges is not allowed for dts leaf node.
but we still would like to separate to different masters
into different iova regions.
Thus we have to separate it by the HW larbid and portid. For example,
larb1/2 are in region2 and larb3 is in region3. The problem is that
some ports inside a larb are in region4 while some ports inside this
larb are in region5. Therefore I define a "larb_region_msk" to help
record the information for each a port. Take a example for a larb:
[1] = ~0: means all ports in this larb are in region1;
[2] = BIT(3) | BIT(4): means port3/4 in this larb are region2;
[3] = ~(BIT(3) | BIT(4)): means all the other ports except port3/4
in this larb are region3.
This method also avoid the users forget/abuse the iova regions.
Signed-off-by: Yong Wu <[email protected]>
---
drivers/iommu/mtk_iommu.c | 43 +++++++++++++++++++++------------------
1 file changed, 23 insertions(+), 20 deletions(-)
diff --git a/drivers/iommu/mtk_iommu.c b/drivers/iommu/mtk_iommu.c
index 2badd6acfb23..d30020d0a190 100644
--- a/drivers/iommu/mtk_iommu.c
+++ b/drivers/iommu/mtk_iommu.c
@@ -8,7 +8,6 @@
#include <linux/clk.h>
#include <linux/component.h>
#include <linux/device.h>
-#include <linux/dma-direct.h>
#include <linux/err.h>
#include <linux/interrupt.h>
#include <linux/io.h>
@@ -194,6 +193,7 @@ struct mtk_iommu_plat_data {
enum mtk_iommu_plat m4u_plat;
u32 flags;
u32 inv_sel_reg;
+ const u32 (*larb_region_msk)[32];
char *pericfg_comp_str;
struct list_head *hw_list;
@@ -508,30 +508,33 @@ static unsigned int mtk_iommu_get_bank_id(struct device *dev,
static int mtk_iommu_get_iova_region_id(struct device *dev,
const struct mtk_iommu_plat_data *plat_data)
{
- const struct mtk_iommu_iova_region *rgn = plat_data->iova_region;
- const struct bus_dma_region *dma_rgn = dev->dma_range_map;
- int i, candidate = -1;
- dma_addr_t dma_end;
+ struct iommu_fwspec *fwspec = dev_iommu_fwspec_get(dev);
+ unsigned int portidmsk = 0, larbid;
+ const u32 *rgn_larb_msk;
+ int i, region_id = -1;
- if (!dma_rgn || plat_data->iova_region_nr == 1)
+ if (plat_data->iova_region_nr == 1)
return 0;
- dma_end = dma_rgn->dma_start + dma_rgn->size - 1;
- for (i = 0; i < plat_data->iova_region_nr; i++, rgn++) {
- /* Best fit. */
- if (dma_rgn->dma_start == rgn->iova_base &&
- dma_end == rgn->iova_base + rgn->size - 1)
- return i;
- /* ok if it is inside this region. */
- if (dma_rgn->dma_start >= rgn->iova_base &&
- dma_end < rgn->iova_base + rgn->size)
- candidate = i;
+ larbid = MTK_M4U_TO_LARB(fwspec->ids[0]);
+ for (i = 0; i < fwspec->num_ids; i++)
+ portidmsk |= BIT(MTK_M4U_TO_PORT(fwspec->ids[i]));
+
+ for (i = 0; i < plat_data->iova_region_nr; i++) {
+ rgn_larb_msk = plat_data->larb_region_msk[i];
+ if (!rgn_larb_msk)
+ continue;
+
+ if ((rgn_larb_msk[larbid] & portidmsk) == portidmsk) {
+ region_id = i;
+ break;
+ }
}
- if (candidate >= 0)
- return candidate;
- dev_err(dev, "Can NOT find the iommu domain id(%pad 0x%llx).\n",
- &dma_rgn->dma_start, dma_rgn->size);
+ if (region_id >= 0)
+ return region_id;
+ dev_err(dev, "Can NOT find the region for larb(%d-%x).\n",
+ larbid, portidmsk);
return -EINVAL;
}
--
2.18.0
After we add the dma-ranges in the parent "soc" node,
this property is unnecessary for the leaf node.
Signed-off-by: Yong Wu <[email protected]>
---
arch/arm64/boot/dts/mediatek/mt8195.dtsi | 1 -
1 file changed, 1 deletion(-)
diff --git a/arch/arm64/boot/dts/mediatek/mt8195.dtsi b/arch/arm64/boot/dts/mediatek/mt8195.dtsi
index 175eb7a25307..8f745442eef8 100644
--- a/arch/arm64/boot/dts/mediatek/mt8195.dtsi
+++ b/arch/arm64/boot/dts/mediatek/mt8195.dtsi
@@ -2137,7 +2137,6 @@
power-domains = <&spm MT8195_POWER_DOMAIN_VENC>;
#address-cells = <2>;
#size-cells = <2>;
- dma-ranges = <0x1 0x0 0x0 0x40000000 0x0 0xfff00000>;
};
vencsys_core1: clock-controller@1b000000 {
--
2.18.0
Prepare for the MM nodes whose dma-ranges(iova range) is 16GB.
Signed-off-by: Yong Wu <[email protected]>
---
arch/arm64/boot/dts/mediatek/mt8186.dtsi | 1 +
1 file changed, 1 insertion(+)
diff --git a/arch/arm64/boot/dts/mediatek/mt8186.dtsi b/arch/arm64/boot/dts/mediatek/mt8186.dtsi
index c326aeb33a10..ea90d2d08316 100644
--- a/arch/arm64/boot/dts/mediatek/mt8186.dtsi
+++ b/arch/arm64/boot/dts/mediatek/mt8186.dtsi
@@ -265,6 +265,7 @@
#address-cells = <2>;
#size-cells = <2>;
compatible = "simple-bus";
+ dma-ranges = <0x0 0x0 0x0 0x0 0x4 0x0>;
ranges;
gic: interrupt-controller@c000000 {
--
2.18.0
Il 13/01/23 07:01, Yong Wu ha scritto:
> After we add the dma-ranges in the parent "soc" node,
> this property is unnecessary for the leaf node.
>
> Signed-off-by: Yong Wu <[email protected]>
Reviewed-by: AngeloGioacchino Del Regno <[email protected]>
Il 13/01/23 07:01, Yong Wu ha scritto:
> Mediatek IOMMU has already controlled the masters' iova ranges by the
> master's larb/port id. then the dma-ranges property is unnecessary for
> the master's node. the master is jpeg here.
>
> Cc: Mauro Carvalho Chehab <[email protected]>
> Cc: Rob Herring <[email protected]>
> Cc: Krzysztof Kozlowski <[email protected]>
> Cc: Matthias Brugger <[email protected]>
> Cc: Bin Liu <[email protected]>
> Cc: kyrie wu <[email protected]>
> Cc: Xia Jiang <[email protected]>
> Signed-off-by: Yong Wu <[email protected]>
Reviewed-by: AngeloGioacchino Del Regno <[email protected]>
Il 13/01/23 07:01, Yong Wu ha scritto:
> Prepare for the MM nodes whose dma-ranges(iova range) is 16GB.
>
> Signed-off-by: Yong Wu <[email protected]>
Reviewed-by: AngeloGioacchino Del Regno <[email protected]>