* [PATCH RESEND v4 01/20] dt-bindings: iommu: arm,smmu: Document interconnects property
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 02/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (18 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro, Krzysztof Kozlowski
On Qualcomm platforms using the ARM MMU-500 (qcom,smmu-500) as the
adreno SMMU, starting from SM8150 (Snapdragon 855), an active
bandwidth vote is required on the interconnect path before the SMMU
register space is accessible.
On these platforms the GPU TCU is a separate hardware block with its
own AXI master port (alm_gpu_tcu) on GEM_NOC, independent of the GPU
rendering data ports. This means the TCU can be active during wakeup
while the GPU data path is idle, leaving GEM_NOC without any other
voter and causing intermittent SMMU register access failures.
On earlier platforms using QSMMU (SMMU-v2) the TCU is integrated
inside the SMMU block and shares the GPU data AXI port, so no
independent vote is required.
Add the optional 'interconnects' property and allow it for all
qcom,smmu-500 + qcom,adreno-smmu combinations.
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@oss.qualcomm.com>
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
.../devicetree/bindings/iommu/arm,smmu.yaml | 25 ++++++++++++++++++++++
1 file changed, 25 insertions(+)
diff --git a/Documentation/devicetree/bindings/iommu/arm,smmu.yaml b/Documentation/devicetree/bindings/iommu/arm,smmu.yaml
index 6748351c7812..ec0fc5f7d4fd 100644
--- a/Documentation/devicetree/bindings/iommu/arm,smmu.yaml
+++ b/Documentation/devicetree/bindings/iommu/arm,smmu.yaml
@@ -252,6 +252,13 @@ properties:
minItems: 1
maxItems: 3
+ interconnects:
+ maxItems: 1
+ description:
+ Interconnect path to the SMMU register space. Required on SoCs
+ where the SMMU registers are only accessible after a bandwidth
+ vote has been placed on the interconnect fabric.
+
nvidia,memory-controller:
description: |
A phandle to the memory controller on NVIDIA Tegra186 and later SoCs.
@@ -656,6 +663,24 @@ allOf:
clock-names: false
clocks: false
+ - if:
+ allOf:
+ - properties:
+ compatible:
+ contains:
+ const: qcom,smmu-500
+ - properties:
+ compatible:
+ contains:
+ const: qcom,adreno-smmu
+ then:
+ properties:
+ interconnects:
+ maxItems: 1
+ else:
+ properties:
+ interconnects: false
+
- if:
properties:
compatible:
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 02/20] iommu/arm-smmu: Add interconnect bandwidth voting support
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 01/20] dt-bindings: iommu: arm,smmu: Document interconnects property Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:59 ` sashiko-bot
2026-11-20 13:41 ` [PATCH RESEND v4 03/20] arm64: dts: qcom: kodiak: Add GEM_NOC interconnect for adreno SMMU Bibek Kumar Patro
` (17 subsequent siblings)
19 siblings, 1 reply; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro, Val Packett
On Qualcomm SoCs using the ARM MMU-500 as the adreno SMMU, the GPU TCU
sits on GEM_NOC as an independent AXI master and requires an active
interconnect bandwidth vote for its registers to be accessible. While
other clients typically satisfy this requirement implicitly, certain
corner cases (e.g. during sleep/wakeup transitions) can leave the SMMU
without a vote, causing intermittent register access failures.
Add support for an optional interconnect path to the arm-smmu
driver and vote for bandwidth while the SMMU is active. The path
is acquired from DT if present and ignored otherwise.
The bandwidth vote is enabled before accessing SMMU registers
during probe and runtime resume, and released during runtime
suspend and on error paths.
Generally, from an architectural perspective, GEM_NOC and DDR are
expected to have an active vote whenever the adreno_smmu block is
powered on. In most common use cases, this requirement is implicitly
satisfied because other GPU-related clients (for example, the GMU
device) already hold a GEM_NOC vote when adreno_smmu is enabled.
However, there are certain corner cases, such as during sleep/wakeup
transitions, where the GEM_NOC vote can be removed before adreno_smmu
is powered down. If adreno_smmu is then accessed while the interconnect
vote is missing, it can lead to the observed failures. Because of the
precise ordering involved, this scenario is difficult to reproduce
consistently. Additionally, GDSC power state transitions in adreno use
cases can independently drop the interconnect vote, further contributing
to this race.
The design difference between QSMMU (SMMU-v2) and SMMU-500 explains
why an explicit ICC vote is needed only for the latter:
QSMMU (SMMU-v2) — SM6250 / SDM845 era: The TCU is NOT a separate
hardware master. It is integrated INSIDE the SMMU-v2 block itself.
GPU SS
+----------------------------------+
| QSMMU (SMMU-v2 @ 0x5040000) |
| +----------------------------+ |
| | TCU (integrated, no | |
| | separate AXI master port) | |
| +------------+---------------+ |
| | page table walk |
| | issued on same |
| | AXI port as GPU |
+---------------+------------------+
|
| qxm_gpu0 / qxm_gpu1
| (GPU data + TCU walks share same NOC master)
v
GEM_NOC -> LLCC -> MC -> DDR
On SMMU-v2 platforms the TCU never needs an independent vote because
it physically cannot be active without the GPU data path also being
active.
SMMU-500 — SM8150+ / SA8775P / SC7280 etc.: The TCU is a SEPARATE
hardware block with its OWN AXI master.
GPU SS
+----------------------------------+
| +-------------+ +-----------+ |
| | GPU Core | | GPU TCU | | <- distinct HW block
| | (rendering)| | + DVM | |
| +------+------+ +-----+-----+ |
+---------+---------------+--------+
| |
qnm_gpu0/1 alm_gpu_tcu <- separate NOC master port
| |
+-------+-------+
v
GEM_NOC -> LLCC -> MC -> DDR
On SMMU-500 the TCU has its own master port — it can be active (e.g.
page table walks during wakeup) while the GPU data path is idle,
leaving GEM_NOC without any other voter. Hence the explicit
MASTER_GPU_TCU -> SLAVE_EBI1 vote is required.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
Tested-by: Val Packett <val@packett.cool> # sm7325-motorola-dubai
---
drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c | 50 ++++++++++++++++++++++++++++++
drivers/iommu/arm/arm-smmu/arm-smmu-qcom.h | 4 +++
drivers/iommu/arm/arm-smmu/arm-smmu.c | 29 +++++++++++++++--
drivers/iommu/arm/arm-smmu/arm-smmu.h | 2 ++
4 files changed, 82 insertions(+), 3 deletions(-)
diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c b/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
index 795e8ec7e8fa..eee20811eede 100644
--- a/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
+++ b/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.c
@@ -6,6 +6,7 @@
#include <linux/acpi.h>
#include <linux/adreno-smmu-priv.h>
#include <linux/delay.h>
+#include <linux/interconnect.h>
#include <linux/of_device.h>
#include <linux/firmware/qcom/qcom_scm.h>
#include <linux/platform_device.h>
@@ -608,6 +609,45 @@ static int qcom_sdm845_smmu500_reset(struct arm_smmu_device *smmu)
return ret;
}
+static int qcom_adreno_smmu_icc_init(struct arm_smmu_device *smmu)
+{
+ struct qcom_smmu *qsmmu = container_of(smmu, struct qcom_smmu, smmu);
+ int err;
+
+ qsmmu->icc_path = devm_of_icc_get(smmu->dev, NULL);
+ if (!IS_ERR(qsmmu->icc_path))
+ return 0;
+
+ err = PTR_ERR(qsmmu->icc_path);
+
+ if (err == -ENODEV) {
+ qsmmu->icc_path = NULL;
+ return 0;
+ }
+ return dev_err_probe(smmu->dev, err,
+ "failed to get interconnect path\n");
+}
+
+static int qcom_adreno_smmu_runtime_resume(struct arm_smmu_device *smmu)
+{
+ struct qcom_smmu *qsmmu = container_of(smmu, struct qcom_smmu, smmu);
+ int ret;
+
+ ret = icc_set_bw(qsmmu->icc_path, 0, 1);
+ WARN_ON_ONCE(ret);
+ return ret;
+}
+
+static int qcom_adreno_smmu_runtime_suspend(struct arm_smmu_device *smmu)
+{
+ struct qcom_smmu *qsmmu = container_of(smmu, struct qcom_smmu, smmu);
+ int ret;
+
+ ret = icc_set_bw(qsmmu->icc_path, 0, 0);
+ WARN_ON_ONCE(ret);
+ return ret;
+}
+
static const struct arm_smmu_impl qcom_smmu_v2_impl = {
.init_context = qcom_smmu_init_context,
.cfg_probe = qcom_smmu_cfg_probe,
@@ -659,6 +699,8 @@ static const struct arm_smmu_impl qcom_adreno_smmu_500_impl = {
.alloc_context_bank = qcom_adreno_smmu_alloc_context_bank,
.write_sctlr = qcom_adreno_smmu_write_sctlr,
.tlb_sync = qcom_smmu_tlb_sync,
+ .runtime_resume = qcom_adreno_smmu_runtime_resume,
+ .runtime_suspend = qcom_adreno_smmu_runtime_suspend,
.context_fault_needs_threaded_irq = true,
};
@@ -668,6 +710,7 @@ static struct arm_smmu_device *qcom_smmu_create(struct arm_smmu_device *smmu,
const struct device_node *np = smmu->dev->of_node;
const struct arm_smmu_impl *impl;
struct qcom_smmu *qsmmu;
+ int ret;
if (!data)
return ERR_PTR(-EINVAL);
@@ -692,6 +735,12 @@ static struct arm_smmu_device *qcom_smmu_create(struct arm_smmu_device *smmu,
qsmmu->smmu.impl = impl;
qsmmu->data = data;
+ if (data->icc_init) {
+ ret = data->icc_init(&qsmmu->smmu);
+ if (ret)
+ return ERR_PTR(ret);
+ }
+
return &qsmmu->smmu;
}
@@ -734,6 +783,7 @@ static const struct qcom_smmu_match_data qcom_smmu_500_impl0_data = {
.adreno_impl = &qcom_adreno_smmu_500_impl,
.cfg = &qcom_smmu_impl0_cfg,
.client_match = qcom_smmu_actlr_client_of_match,
+ .icc_init = qcom_adreno_smmu_icc_init,
};
/*
diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.h b/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.h
index 8addd453f5f1..12c3eda780dc 100644
--- a/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.h
+++ b/drivers/iommu/arm/arm-smmu/arm-smmu-qcom.h
@@ -6,12 +6,15 @@
#ifndef _ARM_SMMU_QCOM_H
#define _ARM_SMMU_QCOM_H
+struct icc_path;
+
struct qcom_smmu {
struct arm_smmu_device smmu;
const struct qcom_smmu_match_data *data;
bool bypass_quirk;
u8 bypass_cbndx;
u32 stall_enabled;
+ struct icc_path *icc_path;
};
enum qcom_smmu_impl_reg_offset {
@@ -29,6 +32,7 @@ struct qcom_smmu_match_data {
const struct arm_smmu_impl *impl;
const struct arm_smmu_impl *adreno_impl;
const struct of_device_id * const client_match;
+ int (*icc_init)(struct arm_smmu_device *smmu);
};
irqreturn_t qcom_smmu_context_fault(int irq, void *dev);
diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu.c b/drivers/iommu/arm/arm-smmu/arm-smmu.c
index 8d50a4ad221d..f015985cbb8e 100644
--- a/drivers/iommu/arm/arm-smmu/arm-smmu.c
+++ b/drivers/iommu/arm/arm-smmu/arm-smmu.c
@@ -2183,6 +2183,14 @@ static int arm_smmu_device_probe(struct platform_device *pdev)
if (err)
return err;
+ if (smmu->impl && smmu->impl->runtime_resume) {
+ err = smmu->impl->runtime_resume(smmu);
+ if (err) {
+ clk_bulk_disable_unprepare(smmu->num_clks, smmu->clks);
+ return err;
+ }
+ }
+
err = arm_smmu_device_cfg_probe(smmu);
if (err)
return err;
@@ -2263,10 +2271,13 @@ static void arm_smmu_device_shutdown(struct platform_device *pdev)
arm_smmu_gr0_write(smmu, ARM_SMMU_GR0_sCR0, ARM_SMMU_sCR0_CLIENTPD);
arm_smmu_rpm_put(smmu);
- if (pm_runtime_enabled(smmu->dev))
+ if (pm_runtime_enabled(smmu->dev)) {
pm_runtime_force_suspend(smmu->dev);
- else
+ } else {
clk_bulk_disable(smmu->num_clks, smmu->clks);
+ if (smmu->impl && smmu->impl->runtime_suspend)
+ smmu->impl->runtime_suspend(smmu);
+ }
clk_bulk_unprepare(smmu->num_clks, smmu->clks);
}
@@ -2286,9 +2297,18 @@ static int __maybe_unused arm_smmu_runtime_resume(struct device *dev)
struct arm_smmu_device *smmu = dev_get_drvdata(dev);
int ret;
+ if (smmu->impl && smmu->impl->runtime_resume) {
+ ret = smmu->impl->runtime_resume(smmu);
+ if (ret)
+ return ret;
+ }
+
ret = clk_bulk_enable(smmu->num_clks, smmu->clks);
- if (ret)
+ if (ret) {
+ if (smmu->impl && smmu->impl->runtime_suspend)
+ smmu->impl->runtime_suspend(smmu);
return ret;
+ }
arm_smmu_device_reset(smmu);
@@ -2301,6 +2321,9 @@ static int __maybe_unused arm_smmu_runtime_suspend(struct device *dev)
clk_bulk_disable(smmu->num_clks, smmu->clks);
+ if (smmu->impl && smmu->impl->runtime_suspend)
+ return smmu->impl->runtime_suspend(smmu);
+
return 0;
}
diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu.h b/drivers/iommu/arm/arm-smmu/arm-smmu.h
index 26d2e33cd328..ed08f86cf99d 100644
--- a/drivers/iommu/arm/arm-smmu/arm-smmu.h
+++ b/drivers/iommu/arm/arm-smmu/arm-smmu.h
@@ -455,6 +455,8 @@ struct arm_smmu_impl {
void (*write_s2cr)(struct arm_smmu_device *smmu, int idx);
void (*write_sctlr)(struct arm_smmu_device *smmu, int idx, u32 reg);
void (*probe_finalize)(struct arm_smmu_device *smmu, struct device *dev);
+ int (*runtime_resume)(struct arm_smmu_device *smmu);
+ int (*runtime_suspend)(struct arm_smmu_device *smmu);
};
#define INVALID_SMENDX -1
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH RESEND v4 02/20] iommu/arm-smmu: Add interconnect bandwidth voting support
2026-11-20 13:41 ` [PATCH RESEND v4 02/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
@ 2026-11-20 13:59 ` sashiko-bot
0 siblings, 0 replies; 23+ messages in thread
From: sashiko-bot @ 2026-11-20 13:59 UTC (permalink / raw)
To: Bibek Kumar Patro; +Cc: devicetree, robh, conor+dt
Thank you for your contribution! Sashiko AI review found 2 potential
issues to consider:
- [High] iommu/arm-smmu: clocks left disabled on
arm_smmu_runtime_suspend() error
- [Medium] iommu/arm-smmu: missing runtime_suspend and clk cleanup on
probe error
--
commit 8c75f1c42f79338d6f97167ba97ee3051a422cca
Author: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
iommu/arm-smmu: Add interconnect bandwidth voting support
This commit adds optional interconnect bandwidth voting support to the
arm-smmu driver and wires it up for Qualcomm Adreno SMMU-500 instances
across probe and runtime PM transitions.
> diff --git a/drivers/iommu/arm/arm-smmu/arm-smmu.c b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> index 8d50a4ad221dd..f015985cbb8ef 100644
> --- a/drivers/iommu/arm/arm-smmu/arm-smmu.c
> +++ b/drivers/iommu/arm/arm-smmu/arm-smmu.c
> @@ -2183,6 +2183,14 @@ static int arm_smmu_device_probe(struct platform_device *pdev)
> if (err)
> return err;
>
> + if (smmu->impl && smmu->impl->runtime_resume) {
> + err = smmu->impl->runtime_resume(smmu);
> + if (err) {
> + clk_bulk_disable_unprepare(smmu->num_clks, smmu->clks);
> + return err;
> + }
> + }
> +
> err = arm_smmu_device_cfg_probe(smmu);
> if (err)
> return err;
[Severity: Medium]
If arm_smmu_device_cfg_probe() or any later step in arm_smmu_device_probe()
fails, does this return without calling smmu->impl->runtime_suspend() and
clk_bulk_disable_unprepare()?
Because devm_clk_bulk_get_all() only releases the clock handles on probe
failure without disabling or unpreparing them, a later failure or probe
deferral leaves the clocks enabled and prepared while devres teardown of
qsmmu->icc_path drops the interconnect bandwidth vote.
[ ... ]
> @@ -2301,6 +2321,9 @@ static int __maybe_unused arm_smmu_runtime_suspend(struct device *dev)
>
> clk_bulk_disable(smmu->num_clks, smmu->clks);
>
> + if (smmu->impl && smmu->impl->runtime_suspend)
> + return smmu->impl->runtime_suspend(smmu);
> +
> return 0;
> }
[Severity: High]
Does arm_smmu_runtime_suspend() leave smmu->clks disabled when
smmu->impl->runtime_suspend() returns an error?
When runtime_suspend fails (for example, if icc_set_bw() fails in
qcom_adreno_smmu_runtime_suspend()), the PM core aborts the suspend and
keeps the device runtime-active. With the clocks already disabled, later
SMMU register accesses will fault or hang on unclocked hardware.
--
Sashiko AI review · https://sashiko.dev/#/patchset/20261120-smmu_interconnect_addition-v4-0-be30526391d9@oss.qualcomm.com?part=2
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH RESEND v4 03/20] arm64: dts: qcom: kodiak: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 01/20] dt-bindings: iommu: arm,smmu: Document interconnects property Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 02/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 04/20] arm64: dts: qcom: lemans: " Bibek Kumar Patro
` (16 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro, Val Packett
On Kodiak platforms, the Adreno SMMU requires a bandwidth vote on
the GEM_NOC path (MASTER_GPU_TCU -> SLAVE_EBI1) before its registers
are accessible. Without this vote, the SMMU may become unreachable,
leading to intermittent probe failures and runtime issues.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
Tested-by: Val Packett <val@packett.cool> # sm7325-motorola-dubai
---
arch/arm64/boot/dts/qcom/kodiak.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/kodiak.dtsi b/arch/arm64/boot/dts/qcom/kodiak.dtsi
index 10e9a5279ed6..c142c7cb509f 100644
--- a/arch/arm64/boot/dts/qcom/kodiak.dtsi
+++ b/arch/arm64/boot/dts/qcom/kodiak.dtsi
@@ -3413,6 +3413,8 @@ adreno_smmu: iommu@3da0000 {
power-domains = <&gpucc GPU_CC_CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
gfx_0_tbu: tbu@3dd9000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 04/20] arm64: dts: qcom: lemans: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (2 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 03/20] arm64: dts: qcom: kodiak: Add GEM_NOC interconnect for adreno SMMU Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:56 ` sashiko-bot
2026-11-20 13:41 ` [PATCH RESEND v4 05/20] arm64: dts: qcom: monaco: " Bibek Kumar Patro
` (15 subsequent siblings)
19 siblings, 1 reply; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On Lemans platforms, the Adreno SMMU requires a bandwidth vote on
the GEM_NOC path (MASTER_GPU_TCU -> SLAVE_EBI1) before its registers
are accessible. Without this vote, the SMMU may become unreachable,
leading to intermittent probe failures and runtime issues.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/lemans.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/lemans.dtsi b/arch/arm64/boot/dts/qcom/lemans.dtsi
index 20f329c0df75..be71eb30f997 100644
--- a/arch/arm64/boot/dts/qcom/lemans.dtsi
+++ b/arch/arm64/boot/dts/qcom/lemans.dtsi
@@ -3127,6 +3127,8 @@ adreno_smmu: iommu@3da0000 {
<GIC_SPI 685 IRQ_TYPE_LEVEL_HIGH>,
<GIC_SPI 686 IRQ_TYPE_LEVEL_HIGH>,
<GIC_SPI 687 IRQ_TYPE_LEVEL_HIGH>;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
llcc_bwmon: pmu@9091000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* Re: [PATCH RESEND v4 04/20] arm64: dts: qcom: lemans: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 ` [PATCH RESEND v4 04/20] arm64: dts: qcom: lemans: " Bibek Kumar Patro
@ 2026-11-20 13:56 ` sashiko-bot
0 siblings, 0 replies; 23+ messages in thread
From: sashiko-bot @ 2026-11-20 13:56 UTC (permalink / raw)
To: Bibek Kumar Patro; +Cc: devicetree, robh, conor+dt
Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [High] arm64: dts: qcom: lemans: adreno_smmu interconnects placed in
lemans.dtsi
--
commit 69c1b90a142a42fba13058e11d6ad8fe981b319c
Author: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
arm64: dts: qcom: lemans: Add GEM_NOC interconnect for adreno SMMU
This commit adds the GEM_NOC interconnect property to the adreno_smmu node
in lemans.dtsi.
> diff --git a/arch/arm64/boot/dts/qcom/lemans.dtsi b/arch/arm64/boot/dts/qcom/lemans.dtsi
> index 20f329c0df751..be71eb30f9979 100644
> --- a/arch/arm64/boot/dts/qcom/lemans.dtsi
> +++ b/arch/arm64/boot/dts/qcom/lemans.dtsi
> @@ -3127,6 +3127,8 @@ adreno_smmu: iommu@3da0000 {
> <GIC_SPI 685 IRQ_TYPE_LEVEL_HIGH>,
> <GIC_SPI 686 IRQ_TYPE_LEVEL_HIGH>,
> <GIC_SPI 687 IRQ_TYPE_LEVEL_HIGH>;
> + interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
> + &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
> };
[Severity: High]
Should interconnects be placed under &adreno_smmu in lemans-embedded.dtsi
instead of lemans.dtsi?
Commit 1f90a2bab7f6 ("arm64: dts: qcom: lemans: Move platform resources
into lemans-embedded.dtsi") moved the interconnect dt-binding headers, the
gem_noc and mc_virt provider nodes, and &adreno_smmu's power-domains and
clocks out of lemans.dtsi into lemans-embedded.dtsi. Adding this property
to lemans.dtsi breaks building any target that includes lemans-gearvm.dtsi
(added in commit c1ef028799fc ("arm64: dts: qcom: Add device tree for
Lemans GearVM variant")), because neither lemans.dtsi nor
lemans-gearvm.dtsi includes the interconnect headers or defines &gem_noc
and &mc_virt.
--
Sashiko AI review · https://sashiko.dev/#/patchset/20261120-smmu_interconnect_addition-v4-0-be30526391d9@oss.qualcomm.com?part=4
^ permalink raw reply [flat|nested] 23+ messages in thread
* [PATCH RESEND v4 05/20] arm64: dts: qcom: monaco: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (3 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 04/20] arm64: dts: qcom: lemans: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 06/20] arm64: dts: qcom: talos: " Bibek Kumar Patro
` (14 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On Monaco platforms, the Adreno SMMU requires a bandwidth vote on
the GEM_NOC path (MASTER_GPU_TCU -> SLAVE_EBI1) before its registers
are accessible. Without this vote, the SMMU may become unreachable,
leading to intermittent probe failures and runtime issues.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/monaco.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/monaco.dtsi b/arch/arm64/boot/dts/qcom/monaco.dtsi
index 3541313a79bf..13214aa77279 100644
--- a/arch/arm64/boot/dts/qcom/monaco.dtsi
+++ b/arch/arm64/boot/dts/qcom/monaco.dtsi
@@ -5113,6 +5113,8 @@ adreno_smmu: iommu@3da0000 {
"gpu_cc_hub_aon_clk";
power-domains = <&gpucc GPU_CC_CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
pmu@9091000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 06/20] arm64: dts: qcom: talos: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (4 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 05/20] arm64: dts: qcom: monaco: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 07/20] arm64: dts: qcom: sm8150: " Bibek Kumar Patro
` (13 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On Talos platforms, the Adreno SMMU requires a bandwidth vote on
the GEM_NOC path (MASTER_GPU_TCU -> SLAVE_EBI1) before its registers
are accessible. Without this vote, the SMMU may become unreachable,
leading to intermittent probe failures and runtime issues.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/talos.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/talos.dtsi b/arch/arm64/boot/dts/qcom/talos.dtsi
index 1d74e459f892..f022a44ae5c8 100644
--- a/arch/arm64/boot/dts/qcom/talos.dtsi
+++ b/arch/arm64/boot/dts/qcom/talos.dtsi
@@ -2102,6 +2102,8 @@ adreno_smmu: iommu@50a0000 {
"iface";
power-domains = <&gpucc CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
stm@6002000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 07/20] arm64: dts: qcom: sm8150: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (5 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 06/20] arm64: dts: qcom: talos: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 08/20] arm64: dts: qcom: sm8250: " Bibek Kumar Patro
` (12 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On SM8150 the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on GEM_NOC. Without an explicit
interconnect bandwidth vote, the GEM_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/sm8150.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/sm8150.dtsi b/arch/arm64/boot/dts/qcom/sm8150.dtsi
index b9846b424a93..4b2ff316db72 100644
--- a/arch/arm64/boot/dts/qcom/sm8150.dtsi
+++ b/arch/arm64/boot/dts/qcom/sm8150.dtsi
@@ -2396,6 +2396,8 @@ adreno_smmu: iommu@2ca0000 {
clock-names = "ahb", "bus", "iface";
power-domains = <&gpucc GPU_CX_GDSC>;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI_CH0 QCOM_ICC_TAG_ALWAYS>;
};
tlmm: pinctrl@3100000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 08/20] arm64: dts: qcom: sm8250: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (6 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 07/20] arm64: dts: qcom: sm8150: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 09/20] arm64: dts: qcom: sm8350: " Bibek Kumar Patro
` (11 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On SM8250 the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on GEM_NOC. Without an explicit
interconnect bandwidth vote, the GEM_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/sm8250.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/sm8250.dtsi b/arch/arm64/boot/dts/qcom/sm8250.dtsi
index 992284e64d38..3a368ef7c22a 100644
--- a/arch/arm64/boot/dts/qcom/sm8250.dtsi
+++ b/arch/arm64/boot/dts/qcom/sm8250.dtsi
@@ -3089,6 +3089,8 @@ adreno_smmu: iommu@3da0000 {
power-domains = <&gpucc GPU_CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI_CH0 QCOM_ICC_TAG_ALWAYS>;
};
slpi: remoteproc@5c00000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 09/20] arm64: dts: qcom: sm8350: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (7 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 08/20] arm64: dts: qcom: sm8250: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 10/20] arm64: dts: qcom: sc8180x: " Bibek Kumar Patro
` (10 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On SM8350 the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on GEM_NOC. Without an explicit
interconnect bandwidth vote, the GEM_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/sm8350.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/sm8350.dtsi b/arch/arm64/boot/dts/qcom/sm8350.dtsi
index 7b9966e8fd4d..4edcb06dbae8 100644
--- a/arch/arm64/boot/dts/qcom/sm8350.dtsi
+++ b/arch/arm64/boot/dts/qcom/sm8350.dtsi
@@ -2566,6 +2566,8 @@ adreno_smmu: iommu@3da0000 {
power-domains = <&gpucc GPU_CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
lpass_ag_noc: interconnect@3c40000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 10/20] arm64: dts: qcom: sc8180x: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (8 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 09/20] arm64: dts: qcom: sm8350: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 11/20] arm64: dts: qcom: sc8280xp: " Bibek Kumar Patro
` (9 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On SC8180x the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on GEM_NOC. Without an explicit
interconnect bandwidth vote, the GEM_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/sc8180x.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/sc8180x.dtsi b/arch/arm64/boot/dts/qcom/sc8180x.dtsi
index be79608b7161..acd05eac700c 100644
--- a/arch/arm64/boot/dts/qcom/sc8180x.dtsi
+++ b/arch/arm64/boot/dts/qcom/sc8180x.dtsi
@@ -2410,6 +2410,8 @@ adreno_smmu: iommu@2ca0000 {
clock-names = "ahb", "bus", "iface";
power-domains = <&gpucc GPU_CX_GDSC>;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI_CH0 QCOM_ICC_TAG_ALWAYS>;
};
tlmm: pinctrl@3100000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 11/20] arm64: dts: qcom: sc8280xp: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (9 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 10/20] arm64: dts: qcom: sc8180x: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 12/20] arm64: dts: qcom: sm8450: " Bibek Kumar Patro
` (8 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On SC8280xp the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on GEM_NOC. Without an explicit
interconnect bandwidth vote, the GEM_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/sc8280xp.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/sc8280xp.dtsi b/arch/arm64/boot/dts/qcom/sc8280xp.dtsi
index 703b4f4e2059..74c53e70f0ba 100644
--- a/arch/arm64/boot/dts/qcom/sc8280xp.dtsi
+++ b/arch/arm64/boot/dts/qcom/sc8280xp.dtsi
@@ -3564,6 +3564,8 @@ gpu_smmu: iommu@3da0000 {
power-domains = <&gpucc GPU_CC_CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
sdc2: mmc@8804000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 12/20] arm64: dts: qcom: sm8450: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (10 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 11/20] arm64: dts: qcom: sc8280xp: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 13/20] arm64: dts: qcom: sm8550: " Bibek Kumar Patro
` (7 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On SM8450 the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on GEM_NOC. Without an explicit
interconnect bandwidth vote, the GEM_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/sm8450.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/sm8450.dtsi b/arch/arm64/boot/dts/qcom/sm8450.dtsi
index 8f84668169fd..f7dab346cc70 100644
--- a/arch/arm64/boot/dts/qcom/sm8450.dtsi
+++ b/arch/arm64/boot/dts/qcom/sm8450.dtsi
@@ -2619,6 +2619,8 @@ adreno_smmu: iommu@3da0000 {
"ahb";
power-domains = <&gpucc GPU_CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
usb_1_hsphy: phy@88e3000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 13/20] arm64: dts: qcom: sm8550: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (11 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 12/20] arm64: dts: qcom: sm8450: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 14/20] arm64: dts: qcom: sm8650: " Bibek Kumar Patro
` (6 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On SM8550 the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on GEM_NOC. Without an explicit
interconnect bandwidth vote, the GEM_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/sm8550.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/sm8550.dtsi b/arch/arm64/boot/dts/qcom/sm8550.dtsi
index cf7d67faef92..1604d590d7f5 100644
--- a/arch/arm64/boot/dts/qcom/sm8550.dtsi
+++ b/arch/arm64/boot/dts/qcom/sm8550.dtsi
@@ -3073,6 +3073,8 @@ adreno_smmu: iommu@3da0000 {
"ahb";
power-domains = <&gpucc GPU_CC_CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
ipa: ipa@3f40000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 14/20] arm64: dts: qcom: sm8650: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (12 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 13/20] arm64: dts: qcom: sm8550: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 15/20] arm64: dts: qcom: hamoa: " Bibek Kumar Patro
` (5 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On SM8650 the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on GEM_NOC. Without an explicit
interconnect bandwidth vote, the GEM_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/sm8650.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/sm8650.dtsi b/arch/arm64/boot/dts/qcom/sm8650.dtsi
index 5a030f9646c8..ecd051ecd680 100644
--- a/arch/arm64/boot/dts/qcom/sm8650.dtsi
+++ b/arch/arm64/boot/dts/qcom/sm8650.dtsi
@@ -4413,6 +4413,8 @@ adreno_smmu: iommu@3da0000 {
"ahb";
power-domains = <&gpucc GPU_CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
ipa: ipa@3f40000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 15/20] arm64: dts: qcom: hamoa: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (13 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 14/20] arm64: dts: qcom: sm8650: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 16/20] arm64: dts: qcom: kaanapali: " Bibek Kumar Patro
` (4 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On X1E80100 the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on GEM_NOC. Without an explicit
interconnect bandwidth vote, the GEM_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/hamoa.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/hamoa.dtsi b/arch/arm64/boot/dts/qcom/hamoa.dtsi
index 443c1a5c212d..f5b720d73a4e 100644
--- a/arch/arm64/boot/dts/qcom/hamoa.dtsi
+++ b/arch/arm64/boot/dts/qcom/hamoa.dtsi
@@ -4289,6 +4289,8 @@ adreno_smmu: iommu@3da0000 {
"ahb";
power-domains = <&gpucc GPU_CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
gem_noc: interconnect@26400000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 16/20] arm64: dts: qcom: kaanapali: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (14 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 15/20] arm64: dts: qcom: hamoa: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 17/20] arm64: dts: qcom: glymur: Add HSC_NOC " Bibek Kumar Patro
` (3 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On Kaanapali the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on GEM_NOC. Without an explicit
interconnect bandwidth vote, the GEM_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/kaanapali.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/kaanapali.dtsi b/arch/arm64/boot/dts/qcom/kaanapali.dtsi
index 9f04a5ad75f2..eccd5451e875 100644
--- a/arch/arm64/boot/dts/qcom/kaanapali.dtsi
+++ b/arch/arm64/boot/dts/qcom/kaanapali.dtsi
@@ -2997,6 +2997,8 @@ adreno_smmu: iommu@3da0000 {
clock-names = "hlos";
power-domains = <&gpucc GPU_CC_CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
remoteproc_adsp: remoteproc@6800000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 17/20] arm64: dts: qcom: glymur: Add HSC_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (15 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 16/20] arm64: dts: qcom: kaanapali: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 18/20] arm64: dts: qcom: milos: Add GEM_NOC " Bibek Kumar Patro
` (2 subsequent siblings)
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On Glymur the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on HSC_NOC. Without an explicit
interconnect bandwidth vote, the HSC_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/glymur.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/glymur.dtsi b/arch/arm64/boot/dts/qcom/glymur.dtsi
index cdaf842814fb..ae3cedcaf9d8 100644
--- a/arch/arm64/boot/dts/qcom/glymur.dtsi
+++ b/arch/arm64/boot/dts/qcom/glymur.dtsi
@@ -4378,6 +4378,8 @@ adreno_smmu: iommu@3da0000 {
clock-names = "hlos";
power-domains = <&gpucc GPU_CC_CX_GDSC>;
dma-coherent;
+ interconnects = <&hsc_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
ipcc: mailbox@3e04000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 18/20] arm64: dts: qcom: milos: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (16 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 17/20] arm64: dts: qcom: glymur: Add HSC_NOC " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 19/20] arm64: dts: qcom: hawi: " Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 20/20] arm64: dts: qcom: sar2130p: " Bibek Kumar Patro
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On Milos the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on GEM_NOC. Without an explicit
interconnect bandwidth vote, the GEM_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/milos.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/milos.dtsi b/arch/arm64/boot/dts/qcom/milos.dtsi
index 9cae93b6a610..49ea0a8759f3 100644
--- a/arch/arm64/boot/dts/qcom/milos.dtsi
+++ b/arch/arm64/boot/dts/qcom/milos.dtsi
@@ -1988,6 +1988,8 @@ adreno_smmu: iommu@3da0000 {
"ahb";
power-domains = <&gpucc GPU_CC_CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
ipa: ipa@3f40000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 19/20] arm64: dts: qcom: hawi: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (17 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 18/20] arm64: dts: qcom: milos: Add GEM_NOC " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
2026-11-20 13:41 ` [PATCH RESEND v4 20/20] arm64: dts: qcom: sar2130p: " Bibek Kumar Patro
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On Hawi the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on GEM_NOC. Without an explicit
interconnect bandwidth vote, the GEM_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/hawi.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/hawi.dtsi b/arch/arm64/boot/dts/qcom/hawi.dtsi
index 77d704fa6cf7..bb60ed9384c6 100644
--- a/arch/arm64/boot/dts/qcom/hawi.dtsi
+++ b/arch/arm64/boot/dts/qcom/hawi.dtsi
@@ -3685,6 +3685,8 @@ adreno_smmu: iommu@3da0000 {
clock-names = "hlos";
power-domains = <&gpucc GPU_CC_CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
remoteproc_mpss: remoteproc@4000000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread* [PATCH RESEND v4 20/20] arm64: dts: qcom: sar2130p: Add GEM_NOC interconnect for adreno SMMU
2026-11-20 13:41 [PATCH RESEND v4 00/20] iommu/arm-smmu: Add interconnect bandwidth voting support Bibek Kumar Patro
` (18 preceding siblings ...)
2026-11-20 13:41 ` [PATCH RESEND v4 19/20] arm64: dts: qcom: hawi: " Bibek Kumar Patro
@ 2026-11-20 13:41 ` Bibek Kumar Patro
19 siblings, 0 replies; 23+ messages in thread
From: Bibek Kumar Patro @ 2026-11-20 13:41 UTC (permalink / raw)
To: Will Deacon, Robin Murphy, Joerg Roedel (AMD), Rob Herring,
Krzysztof Kozlowski, Conor Dooley, Rob Clark, Bjorn Andersson,
Konrad Dybcio, Abel Vesa
Cc: linux-arm-kernel, iommu, devicetree, linux-kernel, linux-arm-msm,
Bibek Kumar Patro
On SAR2130P the Adreno SMMU GPU TCU is a separate hardware block with a
dedicated alm_gpu_tcu master port on GEM_NOC. Without an explicit
interconnect bandwidth vote, the GEM_NOC path to DDR can be left
unvoted during sleep/wakeup transitions, causing intermittent SMMU
register access failures.
Add the required interconnect to ensure reliable register access.
Signed-off-by: Bibek Kumar Patro <bibek.patro@oss.qualcomm.com>
---
arch/arm64/boot/dts/qcom/sar2130p.dtsi | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/sar2130p.dtsi b/arch/arm64/boot/dts/qcom/sar2130p.dtsi
index cc4077005e66..950fb83f26db 100644
--- a/arch/arm64/boot/dts/qcom/sar2130p.dtsi
+++ b/arch/arm64/boot/dts/qcom/sar2130p.dtsi
@@ -1857,6 +1857,8 @@ adreno_smmu: iommu@3da0000 {
"ahb";
power-domains = <&gpucc GPU_CX_GDSC>;
dma-coherent;
+ interconnects = <&gem_noc MASTER_GPU_TCU QCOM_ICC_TAG_ALWAYS
+ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>;
};
usb_1_hsphy: phy@88e3000 {
--
2.34.1
^ permalink raw reply related [flat|nested] 23+ messages in thread