On Fri, 28 May 2004, Greg KH wrote:

> I'm holding off on the kill_urb() patch for 2.6.7.  I like the idea, and
> will reconsider it for 2.6.8.  Sound ok?

Okay.

> Everything else I have from you I've either applied, or lost in the mess
> somewhere.  If I have lost anything else, please feel free to resend it.

Hmm...  There were two more patches which I _thought_ had been sent out, 
but they don't seem to be in the archives.  I'll send them again.


> > I'm not entirely pleased with the decision to use an 8-bit counter to keep 
> > track of URB usage, since it requires an additional spinlock during the 
> > submit_urb and giveback_urb paths.  Those could be eliminated if the 
> > counter were converted to an atomic_t -- an easy change to make if you 
> > think it would be a significant improvement.
> 
> I thought we objected to an atomic_t for some reason, but can't remember
> it now...  Anyone else?

Oliver objected because he thought a single bit could do the job.  But it 
can't, since the counter needs to be able to go up to 2.  (The counter is 
still equal to 1 while the completion handler runs, and if the handler 
resubmits then the counter has to be set to 2.)

As far as I can recall, the only other objection was to adding extra
fields to struct urb.  Dedicating a 4-byte (or more on some architectures)
atomic_t to a counter that is not likely ever to go above 2 might seem
excessive.  But the kernel generally seems to favor these time-over-space
tradeoffs.

Alan Stern



-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to