On Thu, 1 Sep 2016, Vaibhav Hiremath wrote:
> >> I can put prints to trace the execution flow, lets see what comes out
> >> from it.
> > How easily can you reproduce the problem?
>
> Very easily. I would say 7-8 times out of 10.
Then you can add a printk in hub_activate() just before the call to
On Thursday 01 September 2016 07:59 PM, Alan Stern wrote:
On Thu, 1 Sep 2016, Viresh Kumar wrote:
On 31-08-16, 12:46, Alan Stern wrote:
On Wed, 31 Aug 2016, Viresh Kumar wrote:
On 05-08-16, 11:51, Alan Stern wrote:
+++ usb-4.x/drivers/usb/core/hub.c
@@ -1052,7 +1052,7 @@ static void hub_a
On Thu, 1 Sep 2016, Viresh Kumar wrote:
> On 31-08-16, 12:46, Alan Stern wrote:
> > On Wed, 31 Aug 2016, Viresh Kumar wrote:
> >
> > > On 05-08-16, 11:51, Alan Stern wrote:
> > > > +++ usb-4.x/drivers/usb/core/hub.c
> > > > @@ -1052,7 +1052,7 @@ static void hub_activate(struct usb_hub
> > > >
>
On Thursday 01 September 2016 10:32 AM, Viresh Kumar wrote:
On 31-08-16, 12:46, Alan Stern wrote:
On Wed, 31 Aug 2016, Viresh Kumar wrote:
On 05-08-16, 11:51, Alan Stern wrote:
+++ usb-4.x/drivers/usb/core/hub.c
@@ -1052,7 +1052,7 @@ static void hub_activate(struct usb_hub
/* Continue
On 31-08-16, 12:46, Alan Stern wrote:
> On Wed, 31 Aug 2016, Viresh Kumar wrote:
>
> > On 05-08-16, 11:51, Alan Stern wrote:
> > > +++ usb-4.x/drivers/usb/core/hub.c
> > > @@ -1052,7 +1052,7 @@ static void hub_activate(struct usb_hub
> > >
> > > /* Continue a partial initialization */
> > >
On Wed, 31 Aug 2016, Viresh Kumar wrote:
> On 05-08-16, 11:51, Alan Stern wrote:
> > +++ usb-4.x/drivers/usb/core/hub.c
> > @@ -1052,7 +1052,7 @@ static void hub_activate(struct usb_hub
> >
> > /* Continue a partial initialization */
> > if (type == HUB_INIT2 || type == HUB_INIT3) {
> >
On 05-08-16, 11:51, Alan Stern wrote:
> The locking in hub_activate() is not adequate to provide full mutual
> exclusion with hub_quiesce(). The subroutine locks the hub's
> usb_interface, but the callers of hub_quiesce() (such as
> hub_pre_reset() and hub_event()) hold the lock to the hub's
> usb
The locking in hub_activate() is not adequate to provide full mutual
exclusion with hub_quiesce(). The subroutine locks the hub's
usb_interface, but the callers of hub_quiesce() (such as
hub_pre_reset() and hub_event()) hold the lock to the hub's
usb_device.
This patch changes hub_activate() to m
