2023-04-13 07:51:55

by Zheng Wang

Subject: [RESEND PATCH] usb: renesas_usbhs: Fix use after free bug in usbhs_remove due to race condition

In usbhs_probe, &priv->notify_hotplug_work is bound with
usbhsc_notify_hotplug. It will be started then.

If we remove the driver which will call usbhs_remove
to make cleanup, there may be a unfinished work.

The possible sequence is as follows:

Fix it by finishing the work before cleanup in the usbhs_remove

CPU0 CPU1

|usbhsc_notify_hotplug
usbhs_remove |
usbhs_mod_remove |
usbhs_mod_gadget_remove|
kfree(gpriv); |
|usbhsc_hotplug
|usbhs_mod_call start
|usbhsg_start
|usbhsg_try_start
|//use gpriv
Fixes: bc57381e6347 ("usb: renesas_usbhs: use delayed_work instead of work_struct")
Signed-off-by: Zheng Wang <[email protected]>
---
drivers/usb/renesas_usbhs/common.c | 1 +
1 file changed, 1 insertion(+)

diff --git a/drivers/usb/renesas_usbhs/common.c b/drivers/usb/renesas_usbhs/common.c
index 96f3939a65e2..17a0987ef4f5 100644
--- a/drivers/usb/renesas_usbhs/common.c
+++ b/drivers/usb/renesas_usbhs/common.c
@@ -768,6 +768,7 @@ static int usbhs_remove(struct platform_device *pdev)

dev_dbg(&pdev->dev, "usb remove\n");

+ cancel_delayed_work_sync(&priv->notify_hotplug_work);
/* power off */
if (!usbhs_get_dparam(priv, runtime_pwctrl))
usbhsc_power_ctrl(priv, 0);
--
2.25.1

2023-04-13 12:31:49

by Yoshihiro Shimoda

Subject: RE: [RESEND PATCH] usb: renesas_usbhs: Fix use after free bug in usbhs_remove due to race condition

Hi,

> From: Zheng Wang, Sent: Thursday, April 13, 2023 4:49 PM
>
> In usbhs_probe, &priv->notify_hotplug_work is bound with
> usbhsc_notify_hotplug. It will be started then.
>
> If we remove the driver which will call usbhs_remove
> to make cleanup, there may be a unfinished work.
>
> The possible sequence is as follows:
>
> Fix it by finishing the work before cleanup in the usbhs_remove
>
> CPU0 CPU1
>
> |usbhsc_notify_hotplug
> usbhs_remove |
> usbhs_mod_remove |
> usbhs_mod_gadget_remove|
> kfree(gpriv); |
> |usbhsc_hotplug
> |usbhs_mod_call start
> |usbhsg_start
> |usbhsg_try_start
> |//use gpriv

Thank you for the patch!

You should adjust the figure like followings because Greg mentioned it on other patch [1].
-----
CPU0 CPU1

| usbhsc_notify_hotplug
usbhs_remove |
usbhs_mod_remove |
usbhs_mod_gadget_remove |
kfree(gpriv); |
| usbhsc_hotplug
| usbhs_mod_call start
| usbhsg_start
| usbhsg_try_start
| //use gpriv
-----

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

Best regards,
Yoshihiro Shimoda

> Fixes: bc57381e6347 ("usb: renesas_usbhs: use delayed_work instead of work_struct")
> Signed-off-by: Zheng Wang <[email protected]>
> ---
> drivers/usb/renesas_usbhs/common.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/drivers/usb/renesas_usbhs/common.c b/drivers/usb/renesas_usbhs/common.c
> index 96f3939a65e2..17a0987ef4f5 100644
> --- a/drivers/usb/renesas_usbhs/common.c
> +++ b/drivers/usb/renesas_usbhs/common.c
> @@ -768,6 +768,7 @@ static int usbhs_remove(struct platform_device *pdev)
>
> dev_dbg(&pdev->dev, "usb remove\n");
>
> + cancel_delayed_work_sync(&priv->notify_hotplug_work);
> /* power off */
> if (!usbhs_get_dparam(priv, runtime_pwctrl))
> usbhsc_power_ctrl(priv, 0);
> --
> 2.25.1

2023-04-13 15:50:14

by Zheng Hacker

Subject: Re: [RESEND PATCH] usb: renesas_usbhs: Fix use after free bug in usbhs_remove due to race condition

Yoshihiro Shimoda <[email protected]> 于2023年4月13日周四 20:19写道：
>
> Hi,
>
> > From: Zheng Wang, Sent: Thursday, April 13, 2023 4:49 PM
> >
> > In usbhs_probe, &priv->notify_hotplug_work is bound with
> > usbhsc_notify_hotplug. It will be started then.
> >
> > If we remove the driver which will call usbhs_remove
> > to make cleanup, there may be a unfinished work.
> >
> > The possible sequence is as follows:
> >
> > Fix it by finishing the work before cleanup in the usbhs_remove
> >
> > CPU0 CPU1
> >
> > |usbhsc_notify_hotplug
> > usbhs_remove |
> > usbhs_mod_remove |
> > usbhs_mod_gadget_remove|
> > kfree(gpriv); |
> > |usbhsc_hotplug
> > |usbhs_mod_call start
> > |usbhsg_start
> > |usbhsg_try_start
> > |//use gpriv
>
> Thank you for the patch!
>
> You should adjust the figure like followings because Greg mentioned it on other patch [1].

Thanks for your advice. Will apply in the next.

Best regards,
Zheng

> -----
> CPU0 CPU1
>
> | usbhsc_notify_hotplug
> usbhs_remove |
> usbhs_mod_remove |
> usbhs_mod_gadget_remove |
> kfree(gpriv); |
> | usbhsc_hotplug
> | usbhs_mod_call start
> | usbhsg_start
> | usbhsg_try_start
> | //use gpriv
> -----
>
> [1]
> https://lore.kernel.org/lkml/[email protected]/
>
> Best regards,
> Yoshihiro Shimoda
>
> > Fixes: bc57381e6347 ("usb: renesas_usbhs: use delayed_work instead of work_struct")
> > Signed-off-by: Zheng Wang <[email protected]>
> > ---
> > drivers/usb/renesas_usbhs/common.c | 1 +
> > 1 file changed, 1 insertion(+)
> >
> > diff --git a/drivers/usb/renesas_usbhs/common.c b/drivers/usb/renesas_usbhs/common.c
> > index 96f3939a65e2..17a0987ef4f5 100644
> > --- a/drivers/usb/renesas_usbhs/common.c
> > +++ b/drivers/usb/renesas_usbhs/common.c
> > @@ -768,6 +768,7 @@ static int usbhs_remove(struct platform_device *pdev)
> >
> > dev_dbg(&pdev->dev, "usb remove\n");
> >
> > + cancel_delayed_work_sync(&priv->notify_hotplug_work);
> > /* power off */
> > if (!usbhs_get_dparam(priv, runtime_pwctrl))
> > usbhsc_power_ctrl(priv, 0);
> > --
> > 2.25.1
>