On Tue, Jan 14, 2014 at 1:47 PM, Dan Williams <dan.j.williams@xxxxxxxxx> wrote: > On Tue, Jan 14, 2014 at 1:42 PM, Sarah Sharp > <sarah.a.sharp@xxxxxxxxxxxxxxx> wrote: >> On Mon, Jan 13, 2014 at 08:57:38PM -0500, Alan Stern wrote: >>> On Mon, 13 Jan 2014, Sarah Sharp wrote: >>> >>> > On Mon, Jan 13, 2014 at 02:56:57PM -0800, Sarah Sharp wrote: >>> >>> > > I haven't looked at this too closely, but what happens if: >>> > > - the USB 2.0 roothub is registered >>> > > - userspace immediately sets autosuspend_delay to zero and >>> > > pm_qos_no_port_poweroff to zero >>> > > - usb_acpi_find_device changes the connect_type to something other than >>> > > unknown >>> > > - before usb_acpi_check_port_peer() is called, the port is suspended >>> > >>> > Or that call finishes, but no peer port is set yet, because the USB 3.0 >>> > roothub isn't registered yet. The USB 2.0 bus can be suspended before >>> > the USB 3.0 bus has finished registering, as I've seen from this thread: >>> > >>> > http://marc.info/?l=linux-usb&m=138914518219334&w=2 >>> >>> It's always possible for xhci-hcd to prevent the USB-2 root hub from >>> being suspended by calling pm_runtime_get_noresume. The corresponding >>> _put can be done after the USB-3 root hub has been registered. >> >> Sure, it's on my todo list to fix that. I just wondered if there were >> other race conditions present, given that I could find one off the top >> of my head. > > At least as far as this patchset is concerned there is no > race/requirement to hold off usb-2 root hub power management. The > patchset expects usb2 ports to suspend independent of their peer usb3 > port state, and forces usb3 ports to always resume before usb2 peers. > Let me qualify that, there is no requirement to hold off usb-2 root hub power management *unless* the root hub fails to maintain VBUS like an external hub. The patchset assumes that ACPI port power off needs to be called for both ports in a peer before VBUS goes away. My test platform seems to enforce that, but we may need to quirk platforms that violate this assumption. For that quirk we would need to disable port poweroff until ACPI enumeration completed. -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
