diff mbox series

[RFC,4/4] ufs: set power domain performance state when scaling gears

Message ID 20220401145820.1003826-5-krzysztof.kozlowski@linaro.org
State New
Headers show
Series ufs: set power domain performance state when scaling gears | expand

Commit Message

Krzysztof Kozlowski April 1, 2022, 2:58 p.m. UTC
Scaling gears requires not only scaling clocks, but also voltage levels,
e.g. via performance states.

USe the provided OPP table, to set proper OPP frequency which through
required-opps will trigger performance state change.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
---
 drivers/scsi/ufs/ufshcd-pltfrm.c |  6 +++++
 drivers/scsi/ufs/ufshcd.c        | 42 +++++++++++++++++++++++++-------
 drivers/scsi/ufs/ufshcd.h        |  3 +++
 3 files changed, 42 insertions(+), 9 deletions(-)

Comments

Krzysztof Kozlowski April 3, 2022, 5:46 p.m. UTC | #1
On 01/04/2022 21:47, Bjorn Andersson wrote:
> On Fri 01 Apr 07:58 PDT 2022, Krzysztof Kozlowski wrote:
> 
>> Scaling gears requires not only scaling clocks, but also voltage levels,
>> e.g. via performance states.
>>
>> USe the provided OPP table, to set proper OPP frequency which through
>> required-opps will trigger performance state change.
>>
> 
> This looks quite nice! Just two questions about the path looking forward.
> 
> If we where to extend the opp core to allow specifying the clock rate
> for some N first clocks (similar to how e.g. regulators are handled) it
> seems possible to extend this to replace the freq-table property as
> well. Would you agree?

Yes, although that might be trickier. The frequency is a key. I'll take
a look whether it could be changed to multiple values like the voltage.

> 
> 
> The other missing required feature (in this area) from the upstream UFS
> driver is the ability of voting for interconnect bandwidth. Based on
> your path it would be trivial to specify different values for the votes
> for each speed, but looking at downstream [1] (each row represents the
> vote for the two paths in KB/s) indicates a more complex relationship
> between gear and voted bandwidth.
> 
> This was the reason I suggested that perhaps we need to key the
> opp-table based on the gear? But I don't think there would be any issue
> detecting this in runtime...
> 
> [1] https://github.com/MiCode/kernel_devicetree/blob/zeus-s-oss/qcom/waipio.dtsi#L1982

It should be doable with current bindings, assuming that gear is some
imaginary frequency. We have two interconnects for UFS (the DDR and CPU)
and OPP bindings allow to specify opp-peak-kBps and opp-avg-kBps for all
of interconnects. IOW, the opp-peak-kBps will have two values and
opp-avg-kBps as well.

What would be still missing is scaling clocks.

interconnects = <&ddr>, <&cpu>
interconnect-names = "ufs-ddr", "cpu-ufs";
opp-table {
  // gear 1 or some core clock frequency?
  opp-1 {
    opp-hz = /bits/ 64 <75000000>, <0>, <0>, <75000000> ....;
    opp-avg-kBps = <922 1000>;
    opp-peak-kBps = <0 0>;
    required-opps = <&rpmpd_opp_low_svs>;
  }
}

arch/arm64/boot/dts/qcom/sdm630.dtsi already uses it.

I think still the problem is scaling of multiple clocks, depending on
the gear.

Best regards,
Krzysztof
diff mbox series

Patch

diff --git a/drivers/scsi/ufs/ufshcd-pltfrm.c b/drivers/scsi/ufs/ufshcd-pltfrm.c
index cca4b2181a81..c8f19b54be92 100644
--- a/drivers/scsi/ufs/ufshcd-pltfrm.c
+++ b/drivers/scsi/ufs/ufshcd-pltfrm.c
@@ -360,6 +360,12 @@  int ufshcd_pltfrm_init(struct platform_device *pdev,
 		goto dealloc_host;
 	}
 
+	if (devm_pm_opp_of_add_table(dev))
+		dev_dbg(dev, "no OPP table (%d), no performance state control\n",
+			err);
+	else
+		hba->use_pm_opp = true;
+
 	ufshcd_init_lanes_per_dir(hba);
 
 	err = ufshcd_init(hba, mmio_base, irq);
diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index 3f9caafa91bf..84912db86da8 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -1164,11 +1164,16 @@  static int ufshcd_wait_for_doorbell_clr(struct ufs_hba *hba,
 static int ufshcd_scale_gear(struct ufs_hba *hba, bool scale_up)
 {
 	int ret = 0;
+	struct ufs_clk_info *clki;
+	unsigned long pm_opp_target_rate;
 	struct ufs_pa_layer_attr new_pwr_info;
 
+	clki = list_first_entry(&hba->clk_list_head, struct ufs_clk_info, list);
+
 	if (scale_up) {
 		memcpy(&new_pwr_info, &hba->clk_scaling.saved_pwr_info.info,
 		       sizeof(struct ufs_pa_layer_attr));
+		pm_opp_target_rate = clki->max_freq;
 	} else {
 		memcpy(&new_pwr_info, &hba->pwr_info,
 		       sizeof(struct ufs_pa_layer_attr));
@@ -1184,6 +1189,13 @@  static int ufshcd_scale_gear(struct ufs_hba *hba, bool scale_up)
 			new_pwr_info.gear_tx = hba->clk_scaling.min_gear;
 			new_pwr_info.gear_rx = hba->clk_scaling.min_gear;
 		}
+		pm_opp_target_rate = clki->min_freq;
+	}
+
+	if (hba->use_pm_opp && scale_up) {
+		ret = dev_pm_opp_set_rate(hba->dev, pm_opp_target_rate);
+		if (ret)
+			return ret;
 	}
 
 	/* check if the power mode needs to be changed or not? */
@@ -1194,6 +1206,11 @@  static int ufshcd_scale_gear(struct ufs_hba *hba, bool scale_up)
 			hba->pwr_info.gear_tx, hba->pwr_info.gear_rx,
 			new_pwr_info.gear_tx, new_pwr_info.gear_rx);
 
+	if (ret && hba->use_pm_opp && scale_up)
+		dev_pm_opp_set_rate(hba->dev, hba->devfreq->previous_freq);
+	else if (hba->use_pm_opp && !scale_up)
+		ret = dev_pm_opp_set_rate(hba->dev, pm_opp_target_rate);
+
 	return ret;
 }
 
@@ -1435,9 +1452,11 @@  static int ufshcd_devfreq_init(struct ufs_hba *hba)
 	if (list_empty(clk_list))
 		return 0;
 
-	clki = list_first_entry(clk_list, struct ufs_clk_info, list);
-	dev_pm_opp_add(hba->dev, clki->min_freq, 0);
-	dev_pm_opp_add(hba->dev, clki->max_freq, 0);
+	if (!hba->use_pm_opp) {
+		clki = list_first_entry(clk_list, struct ufs_clk_info, list);
+		dev_pm_opp_add(hba->dev, clki->min_freq, 0);
+		dev_pm_opp_add(hba->dev, clki->max_freq, 0);
+	}
 
 	ufshcd_vops_config_scaling_param(hba, &hba->vps->devfreq_profile,
 					 &hba->vps->ondemand_data);
@@ -1449,8 +1468,10 @@  static int ufshcd_devfreq_init(struct ufs_hba *hba)
 		ret = PTR_ERR(devfreq);
 		dev_err(hba->dev, "Unable to register with devfreq %d\n", ret);
 
-		dev_pm_opp_remove(hba->dev, clki->min_freq);
-		dev_pm_opp_remove(hba->dev, clki->max_freq);
+		if (!hba->use_pm_opp) {
+			dev_pm_opp_remove(hba->dev, clki->min_freq);
+			dev_pm_opp_remove(hba->dev, clki->max_freq);
+		}
 		return ret;
 	}
 
@@ -1462,7 +1483,6 @@  static int ufshcd_devfreq_init(struct ufs_hba *hba)
 static void ufshcd_devfreq_remove(struct ufs_hba *hba)
 {
 	struct list_head *clk_list = &hba->clk_list_head;
-	struct ufs_clk_info *clki;
 
 	if (!hba->devfreq)
 		return;
@@ -1470,9 +1490,13 @@  static void ufshcd_devfreq_remove(struct ufs_hba *hba)
 	devfreq_remove_device(hba->devfreq);
 	hba->devfreq = NULL;
 
-	clki = list_first_entry(clk_list, struct ufs_clk_info, list);
-	dev_pm_opp_remove(hba->dev, clki->min_freq);
-	dev_pm_opp_remove(hba->dev, clki->max_freq);
+	if (!hba->use_pm_opp) {
+		struct ufs_clk_info *clki;
+
+		clki = list_first_entry(clk_list, struct ufs_clk_info, list);
+		dev_pm_opp_remove(hba->dev, clki->min_freq);
+		dev_pm_opp_remove(hba->dev, clki->max_freq);
+	}
 }
 
 static void __ufshcd_suspend_clkscaling(struct ufs_hba *hba)
diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
index 88c20f3608c2..3bd02095897f 100644
--- a/drivers/scsi/ufs/ufshcd.h
+++ b/drivers/scsi/ufs/ufshcd.h
@@ -776,6 +776,8 @@  struct ufs_hba_monitor {
  * @auto_bkops_enabled: to track whether bkops is enabled in device
  * @vreg_info: UFS device voltage regulator information
  * @clk_list_head: UFS host controller clocks list node head
+ * @use_pm_opp: whether OPP table is provided and scaling gears should trigger
+ *              setting OPP
  * @pwr_info: holds current power mode
  * @max_pwr_info: keeps the device max valid pwm
  * @clk_scaling_lock: used to serialize device commands and clock scaling
@@ -894,6 +896,7 @@  struct ufs_hba {
 	bool auto_bkops_enabled;
 	struct ufs_vreg_info vreg_info;
 	struct list_head clk_list_head;
+	bool use_pm_opp;
 
 	/* Number of requests aborts */
 	int req_abort_count;