On Thursday 28 October 2004 14:08, Alan Stern wrote: > On Thu, 28 Oct 2004, David Brownell wrote: > > I intend to split it up into multiple patches, but the split won't be on > functional grounds. It will just be based on the source arrangement > (like: everything in usb/core, everything in usb/host, and everything in > usb/gadget).
That can work. What I was getting at was that even the usbcore bits seemed to involve several functional changes ... and for something like this, it's a lot easier for someone (in particular me!) to see what's going on if the changes get factored a bit better. (Sometimes good delta comments solve that issue...) > > The "init1" then "init2" style bothers me. But then, so > > does the current "reset" then "start" style. The reset() > > stuff is the same logic that's now cloned (ugh!) over > > with the pci quirks. (Or actually, the stuff we _needed_ reset for is cloned. There's some other stuff in there now, which isn't cloned.) > The names will be altered. init1 will become something more like > usb_init_hcd. uninit1 will become something more like usb_put_hcd. > init2 will become something more like usb_register_hcd. Those names > aren't perfect because there's no get to match the put (the hcd doesn't > need to be reference counted because its embedded usb_bus already is) > and init includes memory allocation (I don't see any point in a separate > usb_alloc_hcd). That's an old-school API anyway -- soon it can vanish. I'll review this stuff when you send an updated patch (preferably against a Linus kernel!). > For now I'll leave the struct usb_hcd as the first member of the larger > driver-specific structures. Changing over to make the the driver-specific > part an extra add-on to the usb_hcd part (à la struct Scsi_Host) will > involve a fair number of changes to the HC drivers themselves. Having just seen your patch to change the API by requiring the use of type punning, I wonder whether we'd even really need to adopt the alloc_etherdev() model here. It's not like there'll ever be as many HCDs as there are Ethernet drivers, and we don't really have a "legacy driver" issue any more. - Dave ------------------------------------------------------- This SF.Net email is sponsored by: Sybase ASE Linux Express Edition - download now for FREE LinuxWorld Reader's Choice Award Winner for best database on Linux. http://ads.osdn.com/?ad_idU88&alloc_id065&op=click _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
