On Thu, Sep 18, 2003 at 07:35:30PM +0900, Shin-ichiro KAWASAKI wrote: > > Greg KH wrote: > > >On Fri, Sep 12, 2003 at 01:31:28PM +0900, Shin-ichiro KAWASAKI wrote: > > > >>I wish pl2303.c would accept write request while the device is busy. > > > > > >Patches are always welcome :) > > I tried to make my wish come true on 2.4. Here's the patch. > Five functions including pl2303_write() and pl2303_write_room() were copied > from drivers/usb/serial/visor.c shamelessly :-) > > This patch avoids the trouble I complained, and works well with my > pl2303 device. Could you examine if this patch is appropirate or not?
No, it really isn't as you are just reducing the chance that the problem will occur. You still can run out of urbs, and fail a write, correct? So this isn't a correct solution, sorry. But it's a nice fix for some people who want to get this to work most of the time :) > (What I said before was already discussed in later August in this mailing > list, but I have not read those mails when I posted my complaint, > sorry.) > > >To do this on 2.6 would be quite easy, 2.4 will be a bit tougher. > > I don't understand why it's easier on 2.6. > I need some explanation (or some documents to refer). Look at how the visor driver handles urbs in 2.6. There is no more urb pool that can run out. We create the urbs on the fly, and let the garbage collection clean them up when they are finished by both the driver and the usb core. This makes for a much cleaner and smaller driver. Try doing this to the pl2303 driver on 2.6 and see if it solves your problem. Hope this helps, greg k-h ------------------------------------------------------- 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
