> -----Original Message----- > From: Thilo-Alexander Ginkel [mailto:thilo@xxxxxxxxxx] > Sent: Monday, March 12, 2012 3:11 PM > To: Xu, Andiry > Cc: linux-usb@xxxxxxxxxxxxxxx; Sarah Sharp > Subject: Re: USB timeout during S3 suspend w/ NEC Corporation uPD720200 > > On Mon, Mar 12, 2012 at 04:20, Xu, Andiry <Andiry.Xu@xxxxxxx> wrote: > >> xhci_hcd 0000:0d:00.0: Poll event ring: 4308439040 > >> xhci_hcd 0000:0d:00.0: op reg status = 0xffffffff > >> xhci_hcd 0000:0d:00.0: HW died, polling stopped. > > > > This means your host controller is dead after suspend: op reg status > > reads 0xffffffff indicates the xHC is not working at all, it's in > > a status like "plug out". > > Is my assumption correct that this happened during the suspend as it > seems to me that 30 s earlier the controller was still happily > supplying its state during the debug dump, right? > Well, seems everything is fine, but since there is no device connected to the host, I can not tell if it's running normally. I can see your host controller's firmware version is 40.15. I'm not sure if it's the latest version. Maybe you contact NEC for some firmware updates. > OTOH, if the controller is dead there is no point in waiting for it to > come back, right? The polling seems to stop according to the debug > message, but the the suspend still hangs. It would be interesting to > know if the controller would wake up again after a resume (which is > currently prevented by the blocking suspend, of course). > Once the HC is dead, I'm afraid there is no way to bring it back other than rmmod/modprobe again. > > What's the kernel version are you using? > > 3.2.9 x86_64 x86_64 - mainline with only your patch (and the timeout > tweak) applied > > I am not absolutely certain, but I think this did not happen with an > earlier Kernel version [most probably 3.0] (there were other S3 kernel > panics related to my 3G modem, though, so I can't tell exactly). > Bisecting is probably not an option due to the sporadic nature of the > issue. > That's interesting. From I can tell there is only one bug-fix related with suspend-resume since 3.0. It's for resume after suspend, so it should not affect your platform. Thanks, Andiry ��.n��������+%������w��{.n�����{���)�� jg��� �����ݢj����G���� ���j:+v���w�m������w�������h�����٥
