Re: [linux-usb-devel] USB Storage oops in 2.5.69-bk8

[David Brownell <david-b@pacbell.net>]

[-- Attachment #1: Type: text/plain, Size: 766 bytes --]

Alan Stern wrote:
> On Sat, 17 May 2003, David Brownell wrote:
> 
> 
>>That reset couldn't re-assign the address, so of course it failed.
>>
>>In hub.c, usb_physical_reset_device() is leaving dev->state as
>>USB_STATE_CONFIGURED after resetting the port, and of course it's
>>not legal to set the address of a device in that state.
> 
> 
> Do you mean to say that this is a bug in usb_reset_device()?  Isn't it 
> always going to be true that after a port reset a device will need to be 
> assigned an address?

No; yes.  What I mean is that something like this patch
would seem to be needed, so USB_STATE_DEFAULT ("needs an
address") is set when the reset completes.  Untested, but
it compiles.

The enumeration logic still relies on hidden side effects.

- Dave



[-- Attachment #2: Diff --]
[-- Type: text/plain, Size: 691 bytes --]

--- 1.99/drivers/usb/core/hub.c	Mon Apr 21 10:47:47 2003
+++ edited/drivers/usb/core/hub.c	Sun May 18 16:31:02 2003
@@ -737,6 +737,9 @@
 		if (status != -1) {
 			usb_clear_port_feature(hub,
 				port + 1, USB_PORT_FEAT_C_RESET);
+			dev->state = status
+					? USB_STATE_NOTATTACHED
+					: USB_STATE_DEFAULT;
 			return status;
 		}
 
--- 1.204/drivers/usb/core/usb.c	Sun May  4 23:49:53 2003
+++ edited/drivers/usb/core/usb.c	Sun May 18 16:32:19 2003
@@ -1018,7 +1018,6 @@
 	dev->dev.dma_mask = parent->dma_mask;
 
 	/* it's not usable yet */
-	dev->state = USB_STATE_DEFAULT;
 
 	/* USB 2.0 section 5.5.3 talks about ep0 maxpacket ...
 	 * it's fixed size except for full speed devices.