On Fri, 3 Oct 2003, Matthias Urlichs wrote:
> OK, here's the debug output when I remove the keychain:
>
> Oct 3 17:41:32 linux kernel: hub 1-1:1.0: port 1, status 100, change 1, 12 Mb/s
> Oct 3 17:41:32 linux kernel: usb 1-1.1: USB disconnect, address 4
> Oct 3 17:41:32 linux kernel: usb 1-1.1: usb_disable_device nuking all URBs
> Oct 3 17:41:32 linux kernel: usb 1-1.1: unregistering interface 1-1.1:1.0
> Oct 3 17:41:32 linux kernel: bus usb: remove device 1-1.1:1.0
> Oct 3 17:41:32 linux kernel: usb-storage: storage_disconnect() called
> Oct 3 17:41:32 linux kernel: usb-storage: usb_stor_stop_transport called
> Oct 3 17:41:32 linux kernel: usb-storage: -- dissociate_dev
> Oct 3 17:41:32 linux kernel: usb-storage: -- sending exit command to thread
> Oct 3 17:41:32 linux kernel: usb-storage: *** thread awakened.
> Oct 3 17:41:32 linux kernel: usb-storage: -- exit command received
> Oct 3 17:41:32 linux kernel: usb-storage: -- usb_stor_release_resources finished
> Oct 3 17:41:32 linux kernel: drivers/usb/core/usb.c: usb_hotplug
> Oct 3 17:41:32 linux kernel: usb 1-1.1: unregistering device
> Oct 3 17:41:32 linux kernel: bus usb: remove device 1-1.1
> Oct 3 17:41:32 linux kernel: drivers/usb/core/usb.c: usb_hotplug
Hmm... I get a lot more debugging information. It turns out that's
because I'm using a development tree and it has some extra debugging
options turned on. You can try doing the same by putting
#define DEBUG 1
in place of
#undef DEBUG
in the following files:
lib/kobject.c, drivers/base/bus.c, drivers/base/class.c,
drivers/base/core.c, drivers/base/driver.c.
The important part of the control flow is where storage_disconnect() in
drivers/usb/storage/usb.c calls scsi_remove_host() in
drivers/scsi/hosts.c. That call occurs between the "dissociate_dev" and
the "sending exit command" log entries above, so we can be pretty sure
that it is getting made.
In turn, scsi_remove_host() calls class_device_unregister() and
device_del(), and those calls should appear in the debugging log once you
make the changes listed above. Somewhere in there is where the problem
lies, and it may take a little work to track it down exactly.
Alan Stern
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
