On Tue, 3 Jun 2014, Pantelis Koukousoulas wrote: > On Mon, Jun 2, 2014 at 6:04 PM, Alan Stern <[email protected]> wrote: > > No, this is completely wrong. When the driver uses SuperSpeed support, > > there are two hcds: the high speed one and the SuperSpeed one. They > > are different structures and they both need to be initialized. > > > > It sounds like you have confused struct dummy (there's only one of > > these) with struct dummy_hcd (there are two of these). > > I think that indeed I was completely wrong but I wasn't confused about the > structures, but regarding some details of what is executed when, in which > order and in which context. > > The reason was that when I simply removed the timer / list initialization > from dummy_start_ss(), (because they belong to the shared struct dummy_hcd > so they better only be initialized in one place)
That's the same mistake as before -- struct dummy_hcd _isn't_ shared. Only struct dummy is shared. > I got a hard lockup > during testing, > when I started dummy_hcd in superspeed mode and added a device. > > I think that now I understand why this happened, but I will read the > relevant bits > from dummy_hcd and xhci once more (especially what happens if there is > a suspend / resume between the initialization of the highspeed hcd and the > superspeed hcd) and come back with a different approach that will hopefully > be both correct and DRY. The current version of xhci-pci prevents any suspends from occurring between the initialization of the two hcds; see commit bcffae7708eb. You might need to make dummy-hcd do the same thing. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
