[linux-pm] Question on USB re-enumeration

[Alan Stern]

On Sun, 15 Nov 2009, Sriram V wrote:

> Hi,
> 
> I have a couple of questions of re-enumeration of usb devices on
> suspend-resume
> 
> I have a USB Device permanently connected to the host controller.
> 
> Upon resume - Should i usb device get re-enumerated?

If the host controller lost power was reinitialized during the suspend 
then yes.  Otherwise no.  In particular, devices should always get 
re-enumerated when resuming from hibernation.  But when resuming from 
suspend-to-RAM, it depends on the motherboard.

> If so, does Linux choose a different address for accessing it?

If the device's firmware hasn't changed then the address remains the 
same.  Otherwise a different address is used.

Read Documentation/usb/persist.txt.

> If i have a USB harddisk, I am accessing it using a device node say
> /dev/sdb1
> The harddisk is mounted as /dev/sdb1.
> 
> Now, i go into suspend and then resume - If i need to re-enumerate the
> devices
> It could be detected as /dev/sdc1... or something.
> 
> If that is the case i cannot access my harddisk - since im assuming
> /dev/sdb1.
> 
> What im seeing here is - My usb device is permanently connected to the usb
> host via an external phy.
> 
>  When i suspend - i turn off the clocks and upon resume - turn on the clocks
> and reset the phy.
> 
>  For some time - i am able to access the device, What happens is - The
> device gets re-enumerated to some other device node.
> 
> I am not sure if this is a device issue or a resume function issue.
> 
> 
> Please advice.

Enable CONFIG_USB_DEBUG in your kernel and post the dmesg log showing
what happens during suspend and resume.

Alan Stern