Re: [linux-usb-devel] [PATCH] 2.6.18-rc6-mm2 - usb_resume_both() - fix suspend/resume

[Alan Stern <stern@rowland.harvard.edu>]

On Tue, 19 Sep 2006, Jiri Kosina wrote:

> With 2.6.18-rc6-mm2 it is possible on my IBM T42p to do suspend/resume 
> cycle only once. The second time, suspend fails with
> 
> 	usb_hcd_pci_suspend(): ehci_pci_suspend+0x0/0xd0 [ehci_hcd]() returns -22
> 	pci_device_suspend(): usb_hcd_pci_suspend+0x0/0x280() returns -22
> 	suspend_device(): pci_device_suspend+0x0/0x70() returns -22
> 
> or, when ehci is unloaded, uhci fails in a similar way:
> 
> 	usb_hcd_pci_suspend(): uhci_suspend+0x0/0xf0 [uhci_hcd]() returns -16
> 	pci_device_suspend(): usb_hcd_pci_suspend+0x0/0x1c0() returns -16
> 	suspend_device(): pci_device_suspend+0x0/0x70() returns -16
> 
> (by the way this looks to me like a mess, that the user could be presented 
> with either -EINVAL or -EBUSY, in case of the same error, depending on the 
> device driver which reports it. I will probably submit a consolidating 
> patch later this evening).
> 
> After some time I figured out that usb_resume_both() is not resuming 
> devices properly - it calls resume_device() only for those devices which 
> are USB_STATE_SUSPENDED (in which case we want to propagate the resume 
> recursively up the device tree), but other cases (most notably 
> USB_STATE_CONFIGURED) stay unhadled.
> 
> This patch fixes my issue, now the suspend/resume can be done arbitrary 
> number of times without any error.
> 
> Signed-off-by: Jiri Kosina <jikos@jikos.cz>
> 
> --- linux-2.6.18-rc6-mm2.orig/drivers/usb/core/driver.c	2006-09-14 16:20:48.000000000 +0200
> +++ linux-2.6.18-rc6-mm2/drivers/usb/core/driver.c	2006-09-19 03:03:12.000000000 +0200
> @@ -1083,6 +1083,8 @@ int usb_resume_both(struct usb_device *u
>  			status = resume_device(udev);
>  		if (parent)
>  			mutex_unlock(&parent->pm_mutex);
> +	} else {
> +		status = resume_device(udev);
>  	}
>  
>  	/* Now the parent won't suspend until we are finished */

Oddly enough, I identified exactly the same problem last week and 
submitted a nearly-identical patch just a few minutes ago.  The problem 
occurs only when CONFIG_USB_SUSPEND isn't set, and it stems from the fact 
that without this addition, the device's dev.power.power_state.event field 
doesn't get set properly during the resume.

Alan Stern

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/