Re: [PATCH]usblp read race

Am Dienstag, 15. Mai 2007 00:24 schrieb Pete Zaitcev:
> Oliver, I've removed wait_event. Please have a look if you have a moment.
> This also should fix the endless loop upon a write error, which Jochen
> reported.
> 
> Unfortunately, the diffstat is unfavourable: +247 -155. This may be the
> fault of added comments, because the module itself lost about 700 bytes
> on x86_64.  I think the code is more regular and clean now, but
> please let me know if I muddied waters.
> 
> I made a special effort to preserve all oddities of the current driver,
> because it had a history of "cleanups" causing regressions. Again, this
> is something to pay attention to.

Hi,

firstly, you fixed the ordering race.
However:

1. usblp_read():

	rv = usblp_rwait_and_lock(usblp, !!(file->f_flags & O_NONBLOCK));
	if (rv < 0) {
		count = rv;
		goto done;
	}

If you get a signal, the mutex will _not_ be locked, yet it'll be unlocked.

2. usblp_write():	
	if ((rv = usb_submit_urb(usblp->writeurb, GFP_KERNEL)) < 0) {
			usblp->wcomplete = 1;
			usblp->wstatus = 0;
			mutex_unlock (&usblp->mut);
			return writecount ? writecount : -EIO;

At least ENOMEM we'd probably like to report as such.

	Regards
		Oliver

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
[email protected]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel