Skip to contentSkip to conversation

clk: gcc-sc8280xp: fix broken suspend

PATCH In review Johan Hovold 4 participants 7 messages Aug 5, 2022 – Aug 23, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

cover clk: gcc-sc8280xp: fix broken suspend

The Qualcomm PCIe driver currently does not implement suspend at all so we need to mark the GDSCs as always-on to prevent genpd from disabling them.

Similarly, the Qualcomm dwc3 USB suspend implementation is also incomplete and the controller doesn't currently survive a suspend cycle unless the GDSC is kept on. Note that this has nothing to with whether wakeup is enabled or not (cf. [1]).

With these two workarounds, we have somewhat functional suspend on the SC8280XP reference design and Lenovo Thinkpad X13s until the missing driver support is in place (even USB remote wakeup works with [2] applied).

Johan

[1] https://lore.kernel.org/all/[email protected]/
[2] https://lore.kernel.org/all/[email protected]/

Johan Hovold (2):
clk: gcc-sc8280xp: keep PCIe power-domains always-on
clk: gcc-sc8280xp: keep USB power-domains always-on

 drivers/clk/qcom/gcc-sc8280xp.c | 21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)
Signature · 1 line
2.35.1
#1 lore raw

PATCH 1/2 clk: gcc-sc8280xp: keep PCIe power-domains always-on

The Qualcomm PCIe driver does not yet implement suspend so to keep the PCIe power domains always-on for now to avoid crashing during resume.

---
 drivers/clk/qcom/gcc-sc8280xp.c | 14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)
drivers/clk/qcom/gcc-sc8280xp.c +9 -5
index a2f3ffcc5849..eaeada42e13a 100644
@@ -6768,6 +6768,10 @@ static struct gdsc pcie_1_tunnel_gdsc = {
 	.flags = VOTABLE,
 };
 
+/*
+ * The Qualcomm PCIe driver does not yet implement suspend so to keep the
+ * PCIe power domains always-on for now.
+ */
 static struct gdsc pcie_2a_gdsc = {
 	.gdscr = 0x9d004,
 	.collapse_ctrl = 0x52128,
@@ -6776,7 +6780,7 @@ static struct gdsc pcie_2a_gdsc = {
 		.name = "pcie_2a_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
-	.flags = VOTABLE,
+	.flags = VOTABLE | ALWAYS_ON,
 };
 
 static struct gdsc pcie_2b_gdsc = {
@@ -6787,7 +6791,7 @@ static struct gdsc pcie_2b_gdsc = {
 		.name = "pcie_2b_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
-	.flags = VOTABLE,
+	.flags = VOTABLE | ALWAYS_ON,
 };
 
 static struct gdsc pcie_3a_gdsc = {
@@ -6798,7 +6802,7 @@ static struct gdsc pcie_3a_gdsc = {
 		.name = "pcie_3a_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
-	.flags = VOTABLE,
+	.flags = VOTABLE | ALWAYS_ON,
 };
 
 static struct gdsc pcie_3b_gdsc = {
@@ -6809,7 +6813,7 @@ static struct gdsc pcie_3b_gdsc = {
 		.name = "pcie_3b_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
-	.flags = VOTABLE,
+	.flags = VOTABLE | ALWAYS_ON,
 };
 
 static struct gdsc pcie_4_gdsc = {
@@ -6820,7 +6824,7 @@ static struct gdsc pcie_4_gdsc = {
 		.name = "pcie_4_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
-	.flags = VOTABLE,
+	.flags = VOTABLE | ALWAYS_ON,
 };
 
 static struct gdsc ufs_card_gdsc = {
Signature · 1 line
2.35.1
Reviewed-by on v1 1/2
#2 lore raw

Re: [PATCH 1/2] clk: gcc-sc8280xp: keep PCIe power-domains always-on

Quoted 4 lines · Johan Hovold

The Qualcomm PCIe driver does not yet implement suspend so to keep the PCIe power domains always-on for now to avoid crashing during resume.

Thanks,
Mani

Quoted 67 lines
---
 drivers/clk/qcom/gcc-sc8280xp.c | 14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)
drivers/clk/qcom/gcc-sc8280xp.c +9 -5
index a2f3ffcc5849..eaeada42e13a 100644
@@ -6768,6 +6768,10 @@ static struct gdsc pcie_1_tunnel_gdsc = {
 	.flags = VOTABLE,
 };
 
+/*
+ * The Qualcomm PCIe driver does not yet implement suspend so to keep the
+ * PCIe power domains always-on for now.
+ */
 static struct gdsc pcie_2a_gdsc = {
 	.gdscr = 0x9d004,
 	.collapse_ctrl = 0x52128,
@@ -6776,7 +6780,7 @@ static struct gdsc pcie_2a_gdsc = {
 		.name = "pcie_2a_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
-	.flags = VOTABLE,
+	.flags = VOTABLE | ALWAYS_ON,
 };
 
 static struct gdsc pcie_2b_gdsc = {
@@ -6787,7 +6791,7 @@ static struct gdsc pcie_2b_gdsc = {
 		.name = "pcie_2b_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
-	.flags = VOTABLE,
+	.flags = VOTABLE | ALWAYS_ON,
 };
 
 static struct gdsc pcie_3a_gdsc = {
@@ -6798,7 +6802,7 @@ static struct gdsc pcie_3a_gdsc = {
 		.name = "pcie_3a_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
-	.flags = VOTABLE,
+	.flags = VOTABLE | ALWAYS_ON,
 };
 
 static struct gdsc pcie_3b_gdsc = {
@@ -6809,7 +6813,7 @@ static struct gdsc pcie_3b_gdsc = {
 		.name = "pcie_3b_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
-	.flags = VOTABLE,
+	.flags = VOTABLE | ALWAYS_ON,
 };
 
 static struct gdsc pcie_4_gdsc = {
@@ -6820,7 +6824,7 @@ static struct gdsc pcie_4_gdsc = {
 		.name = "pcie_4_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
-	.flags = VOTABLE,
+	.flags = VOTABLE | ALWAYS_ON,
 };
 
 static struct gdsc ufs_card_gdsc = {
Signature · 1 line
2.35.1
Signature · 1 line
மணிவண்ணன் சதாசிவம்
#3 lore raw

PATCH 2/2 clk: gcc-sc8280xp: keep USB power-domains always-on

In reply to Johan Hovold

The Qualcomm DWC3 driver suspend implementation appears to be incomplete for SC8280XP so keep the USB power domains always-on for now so that the controller survives a suspend cycle.

---
 drivers/clk/qcom/gcc-sc8280xp.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/clk/qcom/gcc-sc8280xp.c +7 -0
index eaeada42e13a..7768e6901dcc 100644
@@ -6843,12 +6843,17 @@ static struct gdsc ufs_phy_gdsc = {
 	.pwrsts = PWRSTS_OFF_ON,
 };
 
+/*
+ * The Qualcomm DWC3 driver suspend implementation appears to be incomplete
+ * for sc8280xp so keep the USB power domains always-on for now.
+ */
 static struct gdsc usb30_mp_gdsc = {
 	.gdscr = 0xab004,
 	.pd = {
 		.name = "usb30_mp_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
+	.flags = ALWAYS_ON,
 };
 
 static struct gdsc usb30_prim_gdsc = {
@@ -6857,6 +6862,7 @@ static struct gdsc usb30_prim_gdsc = {
 		.name = "usb30_prim_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
+	.flags = ALWAYS_ON,
 };
 
 static struct gdsc usb30_sec_gdsc = {
@@ -6865,6 +6871,7 @@ static struct gdsc usb30_sec_gdsc = {
 		.name = "usb30_sec_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
+	.flags = ALWAYS_ON,
 };
 
 static struct clk_regmap *gcc_sc8280xp_clocks[] = {
Signature · 1 line
2.35.1
Reviewed-by on v1 2/2
#4 lore raw

Re: [PATCH 2/2] clk: gcc-sc8280xp: keep USB power-domains always-on

Quoted 5 lines · Johan Hovold

The Qualcomm DWC3 driver suspend implementation appears to be incomplete for SC8280XP so keep the USB power domains always-on for now so that the controller survives a suspend cycle.

Thanks,
Mani

Quoted 45 lines
---
 drivers/clk/qcom/gcc-sc8280xp.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/clk/qcom/gcc-sc8280xp.c +7 -0
index eaeada42e13a..7768e6901dcc 100644
@@ -6843,12 +6843,17 @@ static struct gdsc ufs_phy_gdsc = {
 	.pwrsts = PWRSTS_OFF_ON,
 };
 
+/*
+ * The Qualcomm DWC3 driver suspend implementation appears to be incomplete
+ * for sc8280xp so keep the USB power domains always-on for now.
+ */
 static struct gdsc usb30_mp_gdsc = {
 	.gdscr = 0xab004,
 	.pd = {
 		.name = "usb30_mp_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
+	.flags = ALWAYS_ON,
 };
 
 static struct gdsc usb30_prim_gdsc = {
@@ -6857,6 +6862,7 @@ static struct gdsc usb30_prim_gdsc = {
 		.name = "usb30_prim_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
+	.flags = ALWAYS_ON,
 };
 
 static struct gdsc usb30_sec_gdsc = {
@@ -6865,6 +6871,7 @@ static struct gdsc usb30_sec_gdsc = {
 		.name = "usb30_sec_gdsc",
 	},
 	.pwrsts = PWRSTS_OFF_ON,
+	.flags = ALWAYS_ON,
 };
 
 static struct clk_regmap *gcc_sc8280xp_clocks[] = {
Signature · 1 line
2.35.1
Signature · 1 line
மணிவண்ணன் சதாசிவம்
maintainer
#5 lore raw
In reply to Johan Hovold

Quoting Johan Hovold (2022-08-05 05:12:48)

Quoted 9 lines

The Qualcomm PCIe driver currently does not implement suspend at all so we need to mark the GDSCs as always-on to prevent genpd from disabling them.

Similarly, the Qualcomm dwc3 USB suspend implementation is also incomplete and the controller doesn't currently survive a suspend cycle unless the GDSC is kept on. Note that this has nothing to with whether wakeup is enabled or not (cf. [1]).

Quoted 4 lines

With these two workarounds, we have somewhat functional suspend on the SC8280XP reference design and Lenovo Thinkpad X13s until the missing driver support is in place (even USB remote wakeup works with [2] applied).

Are these urgently needed for this release or does suspend not really work yet on sc8280xp? I'm trying to understand if we need to apply these to the fixes tree (in which case why isn't there a Fixes tag targetting whatever commit broke suspend) or if they can wait for the next release and come through clk-next.

#6 lore raw
Quoted 15 lines · Stephen Boyd

Quoting Johan Hovold (2022-08-05 05:12:48)

Quoted 9 lines

The Qualcomm PCIe driver currently does not implement suspend at all so we need to mark the GDSCs as always-on to prevent genpd from disabling them.

Similarly, the Qualcomm dwc3 USB suspend implementation is also incomplete and the controller doesn't currently survive a suspend cycle unless the GDSC is kept on. Note that this has nothing to with whether wakeup is enabled or not (cf. [1]).

Quoted 4 lines

With these two workarounds, we have somewhat functional suspend on the SC8280XP reference design and Lenovo Thinkpad X13s until the missing driver support is in place (even USB remote wakeup works with [2] applied).

Quoted 5 lines

Are these urgently needed for this release or does suspend not really work yet on sc8280xp? I'm trying to understand if we need to apply these to the fixes tree (in which case why isn't there a Fixes tag targetting whatever commit broke suspend) or if they can wait for the next release and come through clk-next.

They are urgently needed as sc8280xp does not survive a suspend cycle without them (e.g. crashes on resume) and this is not something that users expect when closing the lid of their laptops.

This is not a regression, but I guess I could have pointed to the commit adding the sc8280xp clock driver and GDSC power-domain definitions:

Fixes d65d005f9a6c ("clk: qcom: add sc8280xp GCC driver")

As several Qualcomm drivers simply ignore system PM and, for example, leave clocks enabled during suspend, we also need to keep the power-domains enabled.

I noticed that Bjorn applied these for 6.1, but I believe this is 6.0 material (or at the least the USB one is as PCI support for sc8280xp is being added in 6.1).

Johan