2020-07-03 22:52:03

by Martin Blumenstingl

[permalink] [raw]
Subject: [PATCH for-5.8 v2] usb: dwc2: Add missing cleanups when usb_add_gadget_udc() fails

Call dwc2_debugfs_exit() and dwc2_hcd_remove() (if the HCD was enabled
earlier) when usb_add_gadget_udc() has failed. This ensures that the
debugfs entries created by dwc2_debugfs_init() as well as the HCD are
cleaned up in the error path.

Fixes: 207324a321a866 ("usb: dwc2: Postponed gadget registration to the udc class driver")
Signed-off-by: Martin Blumenstingl <[email protected]>
---
Changes since v1 at [0]
- also cleanup the HCD as suggested by Minas (thank you!)
- updated the subject accordingly


[0] https://patchwork.kernel.org/patch/11631381/


drivers/usb/dwc2/platform.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/usb/dwc2/platform.c b/drivers/usb/dwc2/platform.c
index c347d93eae64..9febae441069 100644
--- a/drivers/usb/dwc2/platform.c
+++ b/drivers/usb/dwc2/platform.c
@@ -582,12 +582,16 @@ static int dwc2_driver_probe(struct platform_device *dev)
retval = usb_add_gadget_udc(hsotg->dev, &hsotg->gadget);
if (retval) {
dwc2_hsotg_remove(hsotg);
- goto error_init;
+ goto error_debugfs;
}
}
#endif /* CONFIG_USB_DWC2_PERIPHERAL || CONFIG_USB_DWC2_DUAL_ROLE */
return 0;

+error_debugfs:
+ dwc2_debugfs_exit(hsotg);
+ if (hsotg->hcd_enabled)
+ dwc2_hcd_remove(hsotg);
error_init:
if (hsotg->params.activate_stm_id_vb_detection)
regulator_disable(hsotg->usb33d);
--
2.27.0


2020-07-04 06:03:29

by Minas Harutyunyan

[permalink] [raw]
Subject: Re: [PATCH for-5.8 v2] usb: dwc2: Add missing cleanups when usb_add_gadget_udc() fails



On 7/4/2020 2:50 AM, Martin Blumenstingl wrote:
> Call dwc2_debugfs_exit() and dwc2_hcd_remove() (if the HCD was enabled
> earlier) when usb_add_gadget_udc() has failed. This ensures that the
> debugfs entries created by dwc2_debugfs_init() as well as the HCD are
> cleaned up in the error path.
>
> Fixes: 207324a321a866 ("usb: dwc2: Postponed gadget registration to the udc class driver")
> Signed-off-by: Martin Blumenstingl <[email protected]>

Acked-by: Minas Harutyunyan <[email protected]>

> ---
> Changes since v1 at [0]
> - also cleanup the HCD as suggested by Minas (thank you!)
> - updated the subject accordingly
>
>
> [0] https://urldefense.com/v3/__https://patchwork.kernel.org/patch/11631381/__;!!A4F2R9G_pg!K0ZJIzYVYtrwy2VgxTI28Z_qA995uqmg1dFrMNx3XMQ653ROMKPkt_zQdJe1OOk$
>
>
> drivers/usb/dwc2/platform.c | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/usb/dwc2/platform.c b/drivers/usb/dwc2/platform.c
> index c347d93eae64..9febae441069 100644
> --- a/drivers/usb/dwc2/platform.c
> +++ b/drivers/usb/dwc2/platform.c
> @@ -582,12 +582,16 @@ static int dwc2_driver_probe(struct platform_device *dev)
> retval = usb_add_gadget_udc(hsotg->dev, &hsotg->gadget);
> if (retval) {
> dwc2_hsotg_remove(hsotg);
> - goto error_init;
> + goto error_debugfs;
> }
> }
> #endif /* CONFIG_USB_DWC2_PERIPHERAL || CONFIG_USB_DWC2_DUAL_ROLE */
> return 0;
>
> +error_debugfs:
> + dwc2_debugfs_exit(hsotg);
> + if (hsotg->hcd_enabled)
> + dwc2_hcd_remove(hsotg);
> error_init:
> if (hsotg->params.activate_stm_id_vb_detection)
> regulator_disable(hsotg->usb33d);
>

2020-07-26 10:07:49

by Minas Harutyunyan

[permalink] [raw]
Subject: Re: [PATCH for-5.8 v2] usb: dwc2: Add missing cleanups when usb_add_gadget_udc() fails

Hi Martin,

On 7/4/2020 2:50 AM, Martin Blumenstingl wrote:
> Call dwc2_debugfs_exit() and dwc2_hcd_remove() (if the HCD was enabled
> earlier) when usb_add_gadget_udc() has failed. This ensures that the
> debugfs entries created by dwc2_debugfs_init() as well as the HCD are
> cleaned up in the error path.
>
> Fixes: 207324a321a866 ("usb: dwc2: Postponed gadget registration to the udc class driver")
> Signed-off-by: Martin Blumenstingl <[email protected]>
> ---
> Changes since v1 at [0]
> - also cleanup the HCD as suggested by Minas (thank you!)
> - updated the subject accordingly
>
>
> [0] https://urldefense.com/v3/__https://patchwork.kernel.org/patch/11631381/__;!!A4F2R9G_pg!K0ZJIzYVYtrwy2VgxTI28Z_qA995uqmg1dFrMNx3XMQ653ROMKPkt_zQdJe1OOk$
>
>
> drivers/usb/dwc2/platform.c | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/usb/dwc2/platform.c b/drivers/usb/dwc2/platform.c
> index c347d93eae64..9febae441069 100644
> --- a/drivers/usb/dwc2/platform.c
> +++ b/drivers/usb/dwc2/platform.c
> @@ -582,12 +582,16 @@ static int dwc2_driver_probe(struct platform_device *dev)
> retval = usb_add_gadget_udc(hsotg->dev, &hsotg->gadget);
> if (retval) {
> dwc2_hsotg_remove(hsotg);
> - goto error_init;
> + goto error_debugfs;
> }
> }
> #endif /* CONFIG_USB_DWC2_PERIPHERAL || CONFIG_USB_DWC2_DUAL_ROLE */
> return 0;
>

Kernel test robot found issue:
>> warning: unused label 'error_debugfs' [-Wunused-label]
error_debugfs:
^~~~~~~~~~~~~~
1 warning generated.

So, 'error_debugfs:' label should be under #if/#endif:

#if IS_ENABLED(CONFIG_USB_DWC2_PERIPHERAL) || \
IS_ENABLED(CONFIG_USB_DWC2_DUAL_ROLE)
error_debugfs:
#endif /* CONFIG_USB_DWC2_PERIPHERAL || CONFIG_USB_DWC2_DUAL_ROLE */

Or you have other suggestion?

Could you please fix this issue and submit new version of patch.

Thanks,
Minas


> +error_debugfs:
> + dwc2_debugfs_exit(hsotg);
> + if (hsotg->hcd_enabled)
> + dwc2_hcd_remove(hsotg);
> error_init:
> if (hsotg->params.activate_stm_id_vb_detection)
> regulator_disable(hsotg->usb33d);
>

2020-07-27 19:00:12

by Martin Blumenstingl

[permalink] [raw]
Subject: Re: [PATCH for-5.8 v2] usb: dwc2: Add missing cleanups when usb_add_gadget_udc() fails

Hello Minas,

On Sun, Jul 26, 2020 at 12:04 PM Minas Harutyunyan
<[email protected]> wrote:
[...]
> Kernel test robot found issue:
> >> warning: unused label 'error_debugfs' [-Wunused-label]
> error_debugfs:
> ^~~~~~~~~~~~~~
> 1 warning generated.
>
> So, 'error_debugfs:' label should be under #if/#endif:
>
> #if IS_ENABLED(CONFIG_USB_DWC2_PERIPHERAL) || \
> IS_ENABLED(CONFIG_USB_DWC2_DUAL_ROLE)
> error_debugfs:
> #endif /* CONFIG_USB_DWC2_PERIPHERAL || CONFIG_USB_DWC2_DUAL_ROLE */
>
> Or you have other suggestion?
unfortunately I have no better idea

> Could you please fix this issue and submit new version of patch.
I'm going to cover everything I add inside the same #if (not just the
error label).
if any additional label is added (above) in the future then the code
that I'm adding must not be executed when that #if evaluates to false

v3 is coming in the next few minutes.


Best regards,
Martin