Both PCIe0 and PCIe1 controllers are capable of signalling the MSIs
received from endpoint devices to the CPU using GIC-ITS MSI controller.
Add support for it.
The GIC-ITS MSI implementation provides an advantage over internal MSI
implementation using Locality-specific Peripheral Interrupts (LPI) that
would allow MSIs to be targeted for each CPU core.
Like SM8450 & SM8550, the IDs are swapped, but works fine on PCIe0 and PCIe1.
WiFi PCIe Device on SM8650-QRD using GIC-ITS:
159: 0 0 0 0 0 0 0 0 ITS-MSI 0 Edge PCIe PME, aerdrv
167: 0 4 0 0 0 0 0 0 ITS-MSI 524288 Edge bhi
168: 0 0 4 0 0 0 0 0 ITS-MSI 524289 Edge mhi
169: 0 0 0 34 0 0 0 0 ITS-MSI 524290 Edge mhi
170: 0 0 0 0 3 0 0 0 ITS-MSI 524291 Edge ce0
171: 0 0 0 0 0 2 0 0 ITS-MSI 524292 Edge ce1
172: 0 0 0 0 0 0 806 0 ITS-MSI 524293 Edge ce2
173: 0 0 0 0 0 0 0 76 ITS-MSI 524294 Edge ce3
174: 0 0 0 0 0 0 0 0 ITS-MSI 524295 Edge ce5
175: 0 13 0 0 0 0 0 0 ITS-MSI 524296 Edge DP_EXT_IRQ
176: 0 0 0 0 0 0 0 0 ITS-MSI 524297 Edge DP_EXT_IRQ
177: 0 0 0 5493 0 0 0 0 ITS-MSI 524298 Edge DP_EXT_IRQ
178: 0 0 0 0 82 0 0 0 ITS-MSI 524299 Edge DP_EXT_IRQ
179: 0 0 0 0 0 7204 0 0 ITS-MSI 524300 Edge DP_EXT_IRQ
180: 0 0 0 0 0 0 672 0 ITS-MSI 524301 Edge DP_EXT_IRQ
181: 0 0 0 0 0 0 0 30 ITS-MSI 524302 Edge DP_EXT_IRQ
Signed-off-by: Neil Armstrong <[email protected]>
---
arch/arm64/boot/dts/qcom/sm8650.dtsi | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/arch/arm64/boot/dts/qcom/sm8650.dtsi b/arch/arm64/boot/dts/qcom/sm8650.dtsi
index 2df77123a8c7..7b3dfcb9a57b 100644
--- a/arch/arm64/boot/dts/qcom/sm8650.dtsi
+++ b/arch/arm64/boot/dts/qcom/sm8650.dtsi
@@ -2255,6 +2255,10 @@ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>,
interrupt-map-mask = <0 0 0 0x7>;
#interrupt-cells = <1>;
+ /* Entries are reversed due to the unusual ITS DeviceID encoding */
+ msi-map = <0x0 &gic_its 0x1401 0x1>,
+ <0x100 &gic_its 0x1400 0x1>;
+
linux,pci-domain = <0>;
num-lanes = <2>;
bus-range = <0 0xff>;
@@ -2364,6 +2368,10 @@ &mc_virt SLAVE_EBI1 QCOM_ICC_TAG_ALWAYS>,
interrupt-map-mask = <0 0 0 0x7>;
#interrupt-cells = <1>;
+ /* Entries are reversed due to the unusual ITS DeviceID encoding */
+ msi-map = <0x0 &gic_its 0x1481 0x1>,
+ <0x100 &gic_its 0x1480 0x1>;
+
linux,pci-domain = <1>;
num-lanes = <2>;
bus-range = <0 0xff>;
---
base-commit: 6613476e225e090cc9aad49be7fa504e290dd33d
change-id: 20240125-topic-sm8650-upstream-pcie-its-94572c7f1a73
Best regards,
--
Neil Armstrong <[email protected]>
On 25.01.2024 17:55, Neil Armstrong wrote:
> Both PCIe0 and PCIe1 controllers are capable of signalling the MSIs
> received from endpoint devices to the CPU using GIC-ITS MSI controller.
> Add support for it.
>
> The GIC-ITS MSI implementation provides an advantage over internal MSI
> implementation using Locality-specific Peripheral Interrupts (LPI) that
> would allow MSIs to be targeted for each CPU core.
>
> Like SM8450 & SM8550, the IDs are swapped, but works fine on PCIe0 and PCIe1.
>
> WiFi PCIe Device on SM8650-QRD using GIC-ITS:
> 159: 0 0 0 0 0 0 0 0 ITS-MSI 0 Edge PCIe PME, aerdrv
> 167: 0 4 0 0 0 0 0 0 ITS-MSI 524288 Edge bhi
> 168: 0 0 4 0 0 0 0 0 ITS-MSI 524289 Edge mhi
> 169: 0 0 0 34 0 0 0 0 ITS-MSI 524290 Edge mhi
> 170: 0 0 0 0 3 0 0 0 ITS-MSI 524291 Edge ce0
> 171: 0 0 0 0 0 2 0 0 ITS-MSI 524292 Edge ce1
> 172: 0 0 0 0 0 0 806 0 ITS-MSI 524293 Edge ce2
> 173: 0 0 0 0 0 0 0 76 ITS-MSI 524294 Edge ce3
> 174: 0 0 0 0 0 0 0 0 ITS-MSI 524295 Edge ce5
> 175: 0 13 0 0 0 0 0 0 ITS-MSI 524296 Edge DP_EXT_IRQ
> 176: 0 0 0 0 0 0 0 0 ITS-MSI 524297 Edge DP_EXT_IRQ
Is it by chance that this one never fired?
(lgtm otherwise)
Konrad
On 01/02/2024 20:59, Konrad Dybcio wrote:
> On 25.01.2024 17:55, Neil Armstrong wrote:
>> Both PCIe0 and PCIe1 controllers are capable of signalling the MSIs
>> received from endpoint devices to the CPU using GIC-ITS MSI controller.
>> Add support for it.
>>
>> The GIC-ITS MSI implementation provides an advantage over internal MSI
>> implementation using Locality-specific Peripheral Interrupts (LPI) that
>> would allow MSIs to be targeted for each CPU core.
>>
>> Like SM8450 & SM8550, the IDs are swapped, but works fine on PCIe0 and PCIe1.
>>
>> WiFi PCIe Device on SM8650-QRD using GIC-ITS:
>> 159: 0 0 0 0 0 0 0 0 ITS-MSI 0 Edge PCIe PME, aerdrv
>> 167: 0 4 0 0 0 0 0 0 ITS-MSI 524288 Edge bhi
>> 168: 0 0 4 0 0 0 0 0 ITS-MSI 524289 Edge mhi
>> 169: 0 0 0 34 0 0 0 0 ITS-MSI 524290 Edge mhi
>> 170: 0 0 0 0 3 0 0 0 ITS-MSI 524291 Edge ce0
>> 171: 0 0 0 0 0 2 0 0 ITS-MSI 524292 Edge ce1
>> 172: 0 0 0 0 0 0 806 0 ITS-MSI 524293 Edge ce2
>> 173: 0 0 0 0 0 0 0 76 ITS-MSI 524294 Edge ce3
>> 174: 0 0 0 0 0 0 0 0 ITS-MSI 524295 Edge ce5
>> 175: 0 13 0 0 0 0 0 0 ITS-MSI 524296 Edge DP_EXT_IRQ
>> 176: 0 0 0 0 0 0 0 0 ITS-MSI 524297 Edge DP_EXT_IRQ
>
> Is it by chance that this one never fired?
Yeah I only associated to an SSID and did a simple iperf, not enough to trigger all MSIs
Neil
>
> (lgtm otherwise)
>
> Konrad
On 2.02.2024 09:03, Neil Armstrong wrote:
> On 01/02/2024 20:59, Konrad Dybcio wrote:
>> On 25.01.2024 17:55, Neil Armstrong wrote:
>>> Both PCIe0 and PCIe1 controllers are capable of signalling the MSIs
>>> received from endpoint devices to the CPU using GIC-ITS MSI controller.
>>> Add support for it.
>>>
>>> The GIC-ITS MSI implementation provides an advantage over internal MSI
>>> implementation using Locality-specific Peripheral Interrupts (LPI) that
>>> would allow MSIs to be targeted for each CPU core.
>>>
>>> Like SM8450 & SM8550, the IDs are swapped, but works fine on PCIe0 and PCIe1.
>>>
>>> WiFi PCIe Device on SM8650-QRD using GIC-ITS:
>>> 159: 0 0 0 0 0 0 0 0 ITS-MSI 0 Edge PCIe PME, aerdrv
>>> 167: 0 4 0 0 0 0 0 0 ITS-MSI 524288 Edge bhi
>>> 168: 0 0 4 0 0 0 0 0 ITS-MSI 524289 Edge mhi
>>> 169: 0 0 0 34 0 0 0 0 ITS-MSI 524290 Edge mhi
>>> 170: 0 0 0 0 3 0 0 0 ITS-MSI 524291 Edge ce0
>>> 171: 0 0 0 0 0 2 0 0 ITS-MSI 524292 Edge ce1
>>> 172: 0 0 0 0 0 0 806 0 ITS-MSI 524293 Edge ce2
>>> 173: 0 0 0 0 0 0 0 76 ITS-MSI 524294 Edge ce3
>>> 174: 0 0 0 0 0 0 0 0 ITS-MSI 524295 Edge ce5
>>> 175: 0 13 0 0 0 0 0 0 ITS-MSI 524296 Edge DP_EXT_IRQ
>>> 176: 0 0 0 0 0 0 0 0 ITS-MSI 524297 Edge DP_EXT_IRQ
>>
>> Is it by chance that this one never fired?
>
> Yeah I only associated to an SSID and did a simple iperf, not enough to trigger all MSIs
Reviewed-by: Konrad Dybcio <[email protected]>
Konrad
On Thu, 25 Jan 2024 17:55:04 +0100, Neil Armstrong wrote:
> Both PCIe0 and PCIe1 controllers are capable of signalling the MSIs
> received from endpoint devices to the CPU using GIC-ITS MSI controller.
> Add support for it.
>
> The GIC-ITS MSI implementation provides an advantage over internal MSI
> implementation using Locality-specific Peripheral Interrupts (LPI) that
> would allow MSIs to be targeted for each CPU core.
>
> [...]
Applied, thanks!
[1/1] arm64: dts: qcom: sm8650: Use GIC-ITS for PCIe0 and PCIe1
commit: a33a532b3b1ecd6a64f6280d29d19f3ed6e31a92
Best regards,
--
Bjorn Andersson <[email protected]>