On Sunday, December 01, 2013 12:39:19 PM Michael Chan wrote: > On Sun, 2013-12-01 at 02:34 +0100, Rafael J. Wysocki wrote: > > From: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx> > > > > Modify tg3_chip_reset() and tg3_close() to check if the PCI network > > adapter device is accessible at all in order to skip poking it or > > trying to handle a carrier loss in vain when that's not the case. > > Introduce a special PCI helper function pci_device_is_present() > > for this purpose. > > > > Of course, this uncovers the lack of the appropriate RTNL locking > > in tg3_suspend() and tg3_resume(), so add that locking in there > > too. > > > > These changes prevent tg3 from burning a CPU at 100% load level for > > solid several seconds after the Thunderbolt link is disconnected from > > a Matrox DS1 docking station. > > I'm not familiar with the DS1. Does the tg3 device get removed through > tg3_remove_one() in this case? Yes, this is a normal removal except that the device is gone physically when it happens. Thunderbolt cable disconnect is like a power failure for all devices on the link (that is, all devices in the DS1 in this particular case). > What happens when you reconnect the DS1? Well, it works. ACPIPHP just waits for the removal to complete and then it handles the hot-add event and the device works normally after that. The only problem is that without the patch the hot-removal takes a few seconds and quite a lot of CPU power due to the tight loop spinning in tg3_carrier_off() called by tg3_close(), while with the patch it just happens almost instantaneously. And in the Thunderbolt disconnect case doing the entire tg3_chip_reset() is pointless anyway, because the config space of the device is not available then. Thanks! -- I speak only for myself. Rafael J. Wysocki, Intel Open Source Technology Center. -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
