| Message ID | 20230621175046.61521-1-alexeymin@postmarketos.org |
|---|---|
| State | Not Applicable, archived |
| Headers | show |
| Series | [v2,1/3] dt-bindings: remoteproc: qcom,msm8996-mss-pil: Add SDM660 compatible | expand |
| Context | Check | Description |
|---|---|---|
| robh/checkpatch | success | |
| robh/patch-applied | success | |
| robh/dtbs-check | warning | build log |
| robh/dt-meta-schema | success |
On 21/06/2023 19:50, Alexey Minnekhanov wrote: > Modem subsystem in SDM630/660 is similar to MSM8998 and > device tree node for it is based on the one from msm8998.dtsi. > > Signed-off-by: Alexey Minnekhanov <alexeymin@postmarketos.org> > --- > arch/arm64/boot/dts/qcom/sdm630.dtsi | 65 ++++++++++++++++++++++++++++ > 1 file changed, 65 insertions(+) > ... > + > + qcom,halt-regs = <&tcsr_regs_1 0x3000 0x5000 0x4000>; > + > + power-domains = <&rpmpd SDM660_VDDCX>, > + <&rpmpd SDM660_VDDMX>; > + power-domain-names = "cx", "mx"; > + > + status = "disabled"; > + > + mba { > + memory-region = <&mba_region>; > + }; > + > + mpss { > + memory-region = <&mpss_region>; > + }; mba and mpss are deprecated. I think this was changed to memory-regions in remoteproc node. Best regards, Krzysztof
On 22.06.2023 09:22, Krzysztof Kozlowski wrote: > On 21/06/2023 19:50, Alexey Minnekhanov wrote: >> Modem subsystem in SDM630/660 is similar to MSM8998 and >> device tree node for it is based on the one from msm8998.dtsi. >> >> Signed-off-by: Alexey Minnekhanov <alexeymin@postmarketos.org> >> --- >> arch/arm64/boot/dts/qcom/sdm630.dtsi | 65 ++++++++++++++++++++++++++++ >> 1 file changed, 65 insertions(+) >> > > ... >> + >> + mba { >> + memory-region = <&mba_region>; >> + }; >> + >> + mpss { >> + memory-region = <&mpss_region>; >> + }; > > mba and mpss are deprecated. I think this was changed to memory-regions > in remoteproc node. > That's what I thought after reading DT bindings document, but none of closest siblings (msm8996.dtsi, msm8998.dtsi) use the new way. Will change in V3.
On 21.06.2023 19:50, Alexey Minnekhanov wrote: > Snapdragon 630/660 modem subsystem is similar to one in MSM8998 > and can almost reuse it's reset sequence. > > Downstream sources call this q6v5 version "qdsp6v62-1-5" and its > code path has additional checks for QDSP6v55_BHS_EN_REST_ACK > status [2]. > > Inspiration is taken from Konrad Dybcio's work in [1], but reworked > to use common code path with MSM8996/8998, instead of completely > separate "if" block for SDM660. > > [1] https://github.com/SoMainline/linux/commit/7dd6dd9b936dc8d6c1f1abe299e5b065c33741e8 > [2] https://github.com/MiCode/Xiaomi_Kernel_OpenSource/blob/lavender-q-oss/drivers/soc/qcom/pil-q6v5.c#L393 > > Co-developed-by: Konrad Dybcio <konradybcio@gmail.com> > Signed-off-by: Konrad Dybcio <konradybcio@gmail.com> > Signed-off-by: Alexey Minnekhanov <alexeymin@postmarketos.org> > > --- > > V2: use readl_relaxed_poll_timeout instead of hand-coded for loop. > > In his commit Konrad mentions that modem was unstable, but I don't > observe such behaviour on my device. Modem does not restart by itself, > and I could successfully enable ath10k wcn3990 Wi-Fi with this. > > Also worth saying that in my initial tests just using qcom,msm8998-mss-pil > as-is, without separate resource struct and separate code paths for > SDM660, was also working fine. > --- Looks like pre-SMxxxx Sony devices have some touchy watchdog.. Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org> Konrad > drivers/remoteproc/qcom_q6v5_mss.c | 51 ++++++++++++++++++++++++++++-- > 1 file changed, 49 insertions(+), 2 deletions(-) > > diff --git a/drivers/remoteproc/qcom_q6v5_mss.c b/drivers/remoteproc/qcom_q6v5_mss.c > index 8e15e4f85de1..dcbecae675e2 100644 > --- a/drivers/remoteproc/qcom_q6v5_mss.c > +++ b/drivers/remoteproc/qcom_q6v5_mss.c > @@ -71,6 +71,7 @@ > #define QDSP6SS_MEM_PWR_CTL 0x0B0 > #define QDSP6V6SS_MEM_PWR_CTL 0x034 > #define QDSP6SS_STRAP_ACC 0x110 > +#define QDSP6V62SS_BHS_STATUS 0x0C4 > > /* AXI Halt Register Offsets */ > #define AXI_HALTREQ_REG 0x0 > @@ -123,6 +124,7 @@ > #define QDSP6v56_CLAMP_QMC_MEM BIT(22) > #define QDSP6SS_XO_CBCR 0x0038 > #define QDSP6SS_ACC_OVERRIDE_VAL 0x20 > +#define QDSP6v55_BHS_EN_REST_ACK BIT(0) > > /* QDSP6v65 parameters */ > #define QDSP6SS_CORE_CBCR 0x20 > @@ -130,6 +132,7 @@ > #define QDSP6SS_BOOT_CORE_START 0x400 > #define QDSP6SS_BOOT_CMD 0x404 > #define BOOT_FSM_TIMEOUT 10000 > +#define BHS_CHECK_MAX_LOOPS 200 > > struct reg_info { > struct regulator *reg; > @@ -250,6 +253,7 @@ enum { > MSS_MSM8998, > MSS_SC7180, > MSS_SC7280, > + MSS_SDM660, > MSS_SDM845, > }; > > @@ -700,7 +704,8 @@ static int q6v5proc_reset(struct q6v5 *qproc) > } else if (qproc->version == MSS_MSM8909 || > qproc->version == MSS_MSM8953 || > qproc->version == MSS_MSM8996 || > - qproc->version == MSS_MSM8998) { > + qproc->version == MSS_MSM8998 || > + qproc->version == MSS_SDM660) { > > if (qproc->version != MSS_MSM8909 && > qproc->version != MSS_MSM8953) > @@ -734,6 +739,16 @@ static int q6v5proc_reset(struct q6v5 *qproc) > val |= readl(qproc->reg_base + QDSP6SS_PWR_CTL_REG); > udelay(1); > > + if (qproc->version == MSS_SDM660) { > + ret = readl_relaxed_poll_timeout(qproc->reg_base + QDSP6V62SS_BHS_STATUS, > + i, (i & QDSP6v55_BHS_EN_REST_ACK), > + 1, BHS_CHECK_MAX_LOOPS); > + if (ret == -ETIMEDOUT) { > + dev_err(qproc->dev, "BHS_EN_REST_ACK not set!\n"); > + return -ETIMEDOUT; > + } > + } > + > /* Put LDO in bypass mode */ > val |= QDSP6v56_LDO_BYP; > writel(val, qproc->reg_base + QDSP6SS_PWR_CTL_REG); > @@ -756,7 +771,7 @@ static int q6v5proc_reset(struct q6v5 *qproc) > mem_pwr_ctl = QDSP6SS_MEM_PWR_CTL; > i = 19; > } else { > - /* MSS_MSM8998 */ > + /* MSS_MSM8998, MSS_SDM660 */ > mem_pwr_ctl = QDSP6V6SS_MEM_PWR_CTL; > i = 28; > } > @@ -2193,6 +2208,37 @@ static const struct rproc_hexagon_res sc7280_mss = { > .version = MSS_SC7280, > }; > > +static const struct rproc_hexagon_res sdm660_mss = { > + .hexagon_mba_image = "mba.mbn", > + .proxy_clk_names = (char*[]){ > + "xo", > + "qdss", > + "mem", > + NULL > + }, > + .active_clk_names = (char*[]){ > + "iface", > + "bus", > + "gpll0_mss", > + "mnoc_axi", > + "snoc_axi", > + NULL > + }, > + .proxy_pd_names = (char*[]){ > + "cx", > + "mx", > + NULL > + }, > + .need_mem_protection = true, > + .has_alt_reset = false, > + .has_mba_logs = false, > + .has_spare_reg = false, > + .has_qaccept_regs = false, > + .has_ext_cntl_regs = false, > + .has_vq6 = false, > + .version = MSS_SDM660, > +}; > + > static const struct rproc_hexagon_res sdm845_mss = { > .hexagon_mba_image = "mba.mbn", > .proxy_clk_names = (char*[]){ > @@ -2475,6 +2521,7 @@ static const struct of_device_id q6v5_of_match[] = { > { .compatible = "qcom,msm8998-mss-pil", .data = &msm8998_mss}, > { .compatible = "qcom,sc7180-mss-pil", .data = &sc7180_mss}, > { .compatible = "qcom,sc7280-mss-pil", .data = &sc7280_mss}, > + { .compatible = "qcom,sdm660-mss-pil", .data = &sdm660_mss}, > { .compatible = "qcom,sdm845-mss-pil", .data = &sdm845_mss}, > { }, > };
On Wed, 21 Jun 2023 20:50:44 +0300, Alexey Minnekhanov wrote: > Mention sdm660-mss-pil in compatibles list. > > Applied, thanks! [1/3] dt-bindings: remoteproc: qcom,msm8996-mss-pil: Add SDM660 compatible commit: ed50ac266f67829d4732c8ca61ef1953c2cc63d0 [2/3] remoteproc: qcom: q6v5-mss: Add support for SDM630/636/660 commit: 10b6fec2c8c99b5d1ccbcd070da1fc8e96da0046 Best regards,
diff --git a/Documentation/devicetree/bindings/remoteproc/qcom,msm8996-mss-pil.yaml b/Documentation/devicetree/bindings/remoteproc/qcom,msm8996-mss-pil.yaml index c1ac6ca1e759..09da5616e1e5 100644 --- a/Documentation/devicetree/bindings/remoteproc/qcom,msm8996-mss-pil.yaml +++ b/Documentation/devicetree/bindings/remoteproc/qcom,msm8996-mss-pil.yaml @@ -19,6 +19,7 @@ properties: enum: - qcom,msm8996-mss-pil - qcom,msm8998-mss-pil + - qcom,sdm660-mss-pil - qcom,sdm845-mss-pil reg: @@ -245,7 +246,9 @@ allOf: - if: properties: compatible: - const: qcom,msm8998-mss-pil + enum: + - qcom,msm8998-mss-pil + - qcom,sdm660-mss-pil then: properties: clocks: