[PATCH RESEND] usb: mtu3: fix system suspend in gadget mode after soft disconnect

[PATCH RESEND] usb: mtu3: fix system suspend in gadget mode after soft disconnect

[Ryan Brue]

ssusb_gadget_suspend() removes the D+ pullup for system suspend by calling
mtu3_dev_on_off(mtu, 0) directly. That writes the hardware SOFT_CONN bit
but leaves mtu->softconnect alone, deliberately: ssusb_gadget_resume()
needs it to decide whether to pull up again.

wait_for_ip_sleep() then asks ssusb_gadget_ip_sleep_check() whether the IP
can sleep, and that function tests mtu->softconnect -- the gadget's intent
rather than the state of the pin. So it answers "pulled up, the IP can
sleep" for a device that was just disconnected, and the poll for
SSUSB_IP_SLEEP_STS runs to its 100 ms timeout:

  mtu3 11271000.usb: gadget (high-speed) pullup D-
  mtu3 11271000.usb: ip sleep failed!!!
  mtu3 11271000.usb: gadget (high-speed) pullup D+
  mtu3 11271000.usb: PM: dpm_run_callback(): mtu3_suspend returns -16
  PM: Some devices failed to suspend, or early wake event detected

and the suspend fails. The rule the hardware follows is stated in that
same function: the IP cannot sleep while D+ is not pulled up.

On an MT8173 tablet (Amazon Fire HD 10 2017, dr_mode = "peripheral") this
makes system suspend impossible whenever VBUS is present, which on a board
whose only USB port is also its charging input means whenever it is
charging. Unbinding the UDC, which clears mtu->softconnect through
mtu3_gadget_pullup() and so makes the check agree with the pin again, is
enough to let the same suspend succeed with the driver still bound and the
charger still attached:

  gadget bound   (softconnect=1): suspend returns -16, "ip sleep failed!!!"
  gadget unbound (softconnect=0): suspend completes, 20 s of 20 s

So record what mtu3_dev_on_off() actually wrote and test that instead.
Runtime suspend is unaffected: it does not soft disconnect, so the pullup
is still set and the IP-sleep poll is still performed.

Fixes: e69027c25361 ("usb: mtu3: allow system suspend during active gadget connection")
Assisted-by: LLM
Signed-off-by: Ryan Brue <ryanbrue.dev@gmail.com>
---
 drivers/usb/mtu3/mtu3.h      | 1 +
 drivers/usb/mtu3/mtu3_core.c | 3 ++-
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/usb/mtu3/mtu3.h b/drivers/usb/mtu3/mtu3.h
index ba5a63669e5f..7f4fb87d8438 100644
--- a/drivers/usb/mtu3/mtu3.h
+++ b/drivers/usb/mtu3/mtu3.h
@@ -361,6 +361,7 @@ struct mtu3 {
 	unsigned is_self_powered:1;
 	unsigned test_mode:1;
 	unsigned softconnect:1;
+	unsigned pullup:1;
 	unsigned u1_enable:1;
 	unsigned u2_enable:1;
 	unsigned u3_capable:1;
diff --git a/drivers/usb/mtu3/mtu3_core.c b/drivers/usb/mtu3/mtu3_core.c
index a40bf5bad2d5..5ce61240194f 100644
--- a/drivers/usb/mtu3/mtu3_core.c
+++ b/drivers/usb/mtu3/mtu3_core.c
@@ -346,6 +346,7 @@ void mtu3_dev_on_off(struct mtu3 *mtu, int is_on)
 	else
 		mtu3_hs_softconn_set(mtu, is_on);
 
+	mtu->pullup = !!is_on;
 	dev_info(mtu->dev, "gadget (%s) pullup D%s\n",
 		usb_speed_string(mtu->speed), is_on ? "+" : "-");
 }
@@ -1023,7 +1024,7 @@ bool ssusb_gadget_ip_sleep_check(struct ssusb_mtk *ssusb)
 		return true;
 
 	/* device is started and pullup D+, ip can sleep */
-	if (mtu->is_active && mtu->softconnect)
+	if (mtu->is_active && mtu->pullup)
 		return true;
 
 	/* ip can't sleep if not pullup D+ when support device mode */

---
base-commit: df2908090cda368b01ff43709f51890076c56157
change-id: 20260908-rbrue-suez-upstreaming-mtu3-fix-suspend-usb-gadget-mode-c69c5104b25b

Best regards,
--
Ryan Brue <ryanbrue.dev@gmail.com>
-- 
Ryan Brue <ryanbrue.dev@gmail.com>

Re: [PATCH RESEND] usb: mtu3: fix system suspend in gadget mode after soft disconnect

[Ryan Brue]

Hi Greg,

Gentle ping on this one, there hasn't been any replies since the resend 
on Sept 29. It fixes system suspend on MT8173 in gadget mode after a 
soft disconnect. I've been running it on an Amazon Fire HD 10 (2017) 
since September.

Thanks,
Ryan

Re: [PATCH RESEND] usb: mtu3: fix system suspend in gadget mode after soft disconnect

[Greg Kroah-Hartman]

On Sun, Nov 01, 2026 at 09:43:20PM -0600, Ryan Brue wrote:
> Hi Greg,
> 
> Gentle ping on this one, there hasn't been any replies since the resend on
> Sept 29. It fixes system suspend on MT8173 in gadget mode after a soft
> disconnect. I've been running it on an Amazon Fire HD 10 (2017) since
> September.

Sorry, I see no context here :(

Re: [PATCH RESEND] usb: mtu3: fix system suspend in gadget mode after soft disconnect

[Greg Kroah-Hartman]

On Sun, Nov 01, 2026 at 09:43:20PM -0600, Ryan Brue wrote:
> Hi Greg,
> 
> Gentle ping on this one, there hasn't been any replies since the resend on
> Sept 29. It fixes system suspend on MT8173 in gadget mode after a soft
> disconnect. I've been running it on an Amazon Fire HD 10 (2017) since
> September.

Sorry, I see no context here :(

[PATCH RESEND] usb: mtu3: fix system suspend in gadget mode after soft disconnect

[Ryan Brue]

ssusb_gadget_suspend() removes the D+ pullup for system suspend by calling
mtu3_dev_on_off(mtu, 0) directly. That writes the hardware SOFT_CONN bit
but leaves mtu->softconnect alone, deliberately: ssusb_gadget_resume()
needs it to decide whether to pull up again.

wait_for_ip_sleep() then asks ssusb_gadget_ip_sleep_check() whether the IP
can sleep, and that function tests mtu->softconnect -- the gadget's intent
rather than the state of the pin. So it answers "pulled up, the IP can
sleep" for a device that was just disconnected, and the poll for
SSUSB_IP_SLEEP_STS runs to its 100 ms timeout:

  mtu3 11271000.usb: gadget (high-speed) pullup D-
  mtu3 11271000.usb: ip sleep failed!!!
  mtu3 11271000.usb: gadget (high-speed) pullup D+
  mtu3 11271000.usb: PM: dpm_run_callback(): mtu3_suspend returns -16
  PM: Some devices failed to suspend, or early wake event detected

and the suspend fails. The rule the hardware follows is stated in that
same function: the IP cannot sleep while D+ is not pulled up.

On an MT8173 tablet (Amazon Fire HD 10 2017, dr_mode = "peripheral") this
makes system suspend impossible whenever VBUS is present, which on a board
whose only USB port is also its charging input means whenever it is
charging. Unbinding the UDC, which clears mtu->softconnect through
mtu3_gadget_pullup() and so makes the check agree with the pin again, is
enough to let the same suspend succeed with the driver still bound and the
charger still attached:

  gadget bound   (softconnect=1): suspend returns -16, "ip sleep failed!!!"
  gadget unbound (softconnect=0): suspend completes, 20 s of 20 s

So record what mtu3_dev_on_off() actually wrote and test that instead.
Runtime suspend is unaffected: it does not soft disconnect, so the pullup
is still set and the IP-sleep poll is still performed.

Fixes: e69027c25361 ("usb: mtu3: allow system suspend during active gadget connection")
Assisted-by: LLM
Signed-off-by: Ryan Brue <ryanbrue.dev@gmail.com>
---
 drivers/usb/mtu3/mtu3.h      | 1 +
 drivers/usb/mtu3/mtu3_core.c | 3 ++-
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/usb/mtu3/mtu3.h b/drivers/usb/mtu3/mtu3.h
index ba5a63669e5f..7f4fb87d8438 100644
--- a/drivers/usb/mtu3/mtu3.h
+++ b/drivers/usb/mtu3/mtu3.h
@@ -361,6 +361,7 @@ struct mtu3 {
 	unsigned is_self_powered:1;
 	unsigned test_mode:1;
 	unsigned softconnect:1;
+	unsigned pullup:1;
 	unsigned u1_enable:1;
 	unsigned u2_enable:1;
 	unsigned u3_capable:1;
diff --git a/drivers/usb/mtu3/mtu3_core.c b/drivers/usb/mtu3/mtu3_core.c
index a40bf5bad2d5..5ce61240194f 100644
--- a/drivers/usb/mtu3/mtu3_core.c
+++ b/drivers/usb/mtu3/mtu3_core.c
@@ -346,6 +346,7 @@ void mtu3_dev_on_off(struct mtu3 *mtu, int is_on)
 	else
 		mtu3_hs_softconn_set(mtu, is_on);
 
+	mtu->pullup = !!is_on;
 	dev_info(mtu->dev, "gadget (%s) pullup D%s\n",
 		usb_speed_string(mtu->speed), is_on ? "+" : "-");
 }
@@ -1023,7 +1024,7 @@ bool ssusb_gadget_ip_sleep_check(struct ssusb_mtk *ssusb)
 		return true;
 
 	/* device is started and pullup D+, ip can sleep */
-	if (mtu->is_active && mtu->softconnect)
+	if (mtu->is_active && mtu->pullup)
 		return true;
 
 	/* ip can't sleep if not pullup D+ when support device mode */

---
base-commit: df2908090cda368b01ff43709f51890076c56157
change-id: 20260908-rbrue-suez-upstreaming-mtu3-fix-suspend-usb-gadget-mode-c69c5104b25b

Best regards,
--
Ryan Brue <ryanbrue.dev@gmail.com>
-- 
Ryan Brue <ryanbrue.dev@gmail.com>

Re: [PATCH RESEND] usb: mtu3: fix system suspend in gadget mode after soft disconnect

[sashiko-bot]

> ssusb_gadget_suspend() removes the D+ pullup for system suspend by calling
> mtu3_dev_on_off(mtu, 0) directly. That writes the hardware SOFT_CONN bit
> but leaves mtu->softconnect alone, deliberately: ssusb_gadget_resume()
> needs it to decide whether to pull up again.
> 
> wait_for_ip_sleep() then asks ssusb_gadget_ip_sleep_check() whether the IP
> can sleep, and that function tests mtu->softconnect -- the gadget's intent
> rather than the state of the pin. So it answers "pulled up, the IP can
> sleep" for a device that was just disconnected, and the poll for
> SSUSB_IP_SLEEP_STS runs to its 100 ms timeout:
> 
>   mtu3 11271000.usb: gadget (high-speed) pullup D-
>   mtu3 11271000.usb: ip sleep failed!!!
>   mtu3 11271000.usb: gadget (high-speed) pullup D+
>   mtu3 11271000.usb: PM: dpm_run_callback(): mtu3_suspend returns -16
> [ ... ]
> is still set and the IP-sleep poll is still performed.
> 
> Fixes: e69027c25361 ("usb: mtu3: allow system suspend during active gadget connection")
> Assisted-by: LLM
> Signed-off-by: Ryan Brue <ryanbrue.dev@gmail.com>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261102-rbrue-suez-upstreaming-mtu3-fix-suspend-usb-gadget-mode-v1-1-70a74a392015@gmail.com?part=1