On Mon, Dec 23, 2002 at 11:55:49AM -0800, Greg KH wrote:
> On Sun, Dec 22, 2002 at 07:49:48PM +0100, Henning Meier-Geinitz wrote:
> >
> > Decrease module usage count even if device was disconnected while
> > open. Check if disconnected before read, write, and ioctl.
>
> Hm, how about this simple patch to fix all of the module reference
> counting problems. Let me know if this works for you.
It fixes the module counting problem, i.e. the module count is 0 after
disconnect + close even if close_scanner complains "scanner.c:
close_scanner(0): invalid scn_minor".
However, it doesn't fix the crash with read/write after
disconnect ("Unable to handle kernel paging request at virtual address
5a5a5a5a"). The crash happens everytime when disconnecting the scanner
during actual scanning.
That's because of "scn = file->private_data;" in scanner_read and
scanner_write. The structure to which scn points now was kfreed in
disconnect_scanner...
I haven't found any way to "tell" read/write that the device was
disconnected without keeping the scn data until close_scanner() is
called. That's why I didn't include the "owner: THIS_MODULE" in this
patch.
Any ideas to do that without keeping the scn data around until the
final close?
Bye,
Henning
-------------------------------------------------------
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
