On Thu, 15 Sep 2016, Ulf Hansson wrote: > > The situation isn't any better. At the start of the trace, > > the device is in runtime suspend but there are many attempts to > > communicate with it, all of which fail. > > It's really weird. Have this driver ever worked!? :-) Probably not. Or at least, not with runtime PM. > > Then a little less than an hour after the trace started, the device was > > resumed. At that point it started working okay. Until there was a > > spontaneous disconnect. > > > > The device reconnected, but after 3 seconds it was runtime suspended > > again -- and the I/O attempts continued. Some time later there was > > another runtime resume, and the device began working again. Until > > another spontaneous disconnect occurred. And so on... > > > > Ulf, we really need to figure out why the autosuspends are occurring > > and why the I/O doesn't stop while the device is suspended. > > Okay, let's see. > > I had another look in the rtsx_usb_sdmmc driver. Apparently it > registers a led classdev. Updating the led is done from a work, by > calling rtsx_usb_turn_on|off_led(), which do access the usb device. > These calls are not properly managed by runtime PM, so I have fixed > those according to below change: > > --- > drivers/mmc/host/rtsx_usb_sdmmc.c | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/drivers/mmc/host/rtsx_usb_sdmmc.c > b/drivers/mmc/host/rtsx_usb_sdmmc.c > index 6c71fc9..a59c7fa 100644 > --- a/drivers/mmc/host/rtsx_usb_sdmmc.c > +++ b/drivers/mmc/host/rtsx_usb_sdmmc.c > @@ -1314,6 +1314,7 @@ static void rtsx_usb_update_led(struct work_struct *work) > container_of(work, struct rtsx_usb_sdmmc, led_work); > struct rtsx_ucr *ucr = host->ucr; > > + pm_runtime_get_sync(sdmmc_dev(host)); > mutex_lock(&ucr->dev_mutex); > > if (host->led.brightness == LED_OFF) > @@ -1322,6 +1323,7 @@ static void rtsx_usb_update_led(struct work_struct *work) > rtsx_usb_turn_on_led(ucr); > > mutex_unlock(&ucr->dev_mutex); > + pm_runtime_put(sdmmc_dev(host)); > } > #endif > > -- > > Although, I doubt the above is the main reason to the issues we see. I don't know -- it could well be the reason. The symptoms are definitely what you would expect to see if some thread was doing I/O without calling the pm_runtime_* routines. > Instead I think somehow the parent device (usb device) isn't being > properly managed through runtime PM, but not due to wrong deployment > in the mmc core nor in the rtsx_usb_driver, but at some place else. > :-) > > I started looking for calls to pm_suspend_ignore_children(dev, true), > which would decouple the usb device from the mmc platform device from > a runtime PM point of view. I found one suspicious case! > > drivers/usb/storage/realtek_cr.c: > pm_suspend_ignore_children(&us->pusb_intf->dev, true); > > As I am not so familiar with USB, I can't really tell why the above > exists, although perhaps just removing that line would be worth a > try!? No, the realtek_cr driver has no connection with this. It's a sub-module of the usb_storage driver; it uses the SCSI interface, not the MMC interface. > If neither of the above works, the next step could be to start > checking error codes in the mmc core and in the rtsx_usb_sdmmc driver, > from the calls to pm_runtime_get|put() and pm_runtime_enable(). Let's see what this patch does. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
