Re: RT-friendly IRQ management in USB
Alan Stern <[email protected]>
| Newsgroups | gmane.linux.usb.devel |
|---|---|
| Message-ID | <[email protected]> |
On Fri, 6 Jul 2007, David Brownell wrote: > On Monday 02 July 2007, Alan Cox wrote: > > > Why not just bite the bullet and change the callback convention. The lock > > verification code should catch the cases that matter and which are > > overlooked on a code scan. You could also change the name of the callback > > to be sure it breaks anything out of tree that isn't fixed. > > The cases Alan identified are inside usbcore, mostly for root hubs. > Those would be easy to convert. Right; there are only a few of them. The hard part is converting the completion routines themselves, of which there are many. > By far the majority of URB completions are issued from HCDs, from > IRQ handlers ... which in RT kernels will be thread contexts, but > in normal kernels will be hardirq contexts. > > I'm not sure the lock validator would necessarily catch all the > relevant cases. Fault paths are troublesome... Agreed. For example, the fact that a spinlock is sometimes taken with IRQs disabled and sometimes with IRQs enabled won't necessarily show up in a casual test. > On the other hand, that *could* dovetail nicely with the proposed > change of the callback signature to take urb and status code, > rather than urb with urb->status ... :) Augh!! Too many changes, all at once! I can't face it! :-) Alan Stern ------------------------------------------------------------------------- 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