Date: Tue, 21 Oct 2003 20:10:25 +1000 From: Johny Mattsson <lonewolf-freebsd@earthmagic.org> To: freebsd-current@freebsd.org Subject: USB device detach events [PATCH - please test] Message-ID: <3F950611.8080302@earthmagic.org>
next in thread | raw e-mail | index | archive | help
This is a multi-part message in MIME format. --------------040900000008030903040403 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Hello, As an answer to my previous email, I now have patches fixing the missing USB device detach events. This works extremely well on my system, but it'd be great if I could get some wider exposure before I send-pr them. From my perusal/grepping of all the usb drivers, these patches shouldn't break anything, only allow usbd to receive DEVICE_DETACH events for all devices. When I continued chasing my original problem, I discovered that the usbd_add_dev_event(USB_EVENT_DEVICE_DETACH, dev) call in usb_subr.c was commented out. Bringing it back into play brought me a nice solid system freeze, followed by a reboot shortly thereafter. I figure that must be why it was commented out in the first place. As it turns out, the placement of that call is really bad. The 'dev' it hands out is in an inconsistent state if any subdevs existed (if there were no subdevs, everything works fine). I say existed, because at the original location of the "add event" call, the subdevs have already been detached. The usbd_add_dev_event function however calls the usbd_fill_deviceinfo, which wants to fiddle with the subdevs as well, resulting in Bad Things happening. My patch to usb_subr.c simply moves the usbd_add_dev_event call to just before the subdevs are disconnected. Also, the patch to usb.c guarantees (I believe) that if a DRIVER_DETACH event is added to the queue before the DEVICE_DETACH event has been processed, the DEVICE_DETACH is kept on the queue, resulting in expected behaviour by usbd. Comments are welcome. If I don't hear anything from anyone, I'll send-pr in a couple of days. Cheers, /Johny -- Johny Mattsson - System Designer ,-. ,-. ,-. There is no truth. http://www.earthmagic.org _.' `-' `-' There is only perception. --------------040900000008030903040403 Content-Type: text/plain; name="usb_subr.c.patch" Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="usb_subr.c.patch" --- usb_subr.c.org Tue Oct 21 17:44:48 2003 +++ usb_subr.c Tue Oct 21 19:40:44 2003 @@ -1353,6 +1353,8 @@ } #endif + usbd_add_dev_event(USB_EVENT_DEVICE_DETACH, dev); + if (dev->subdevs != NULL) { DPRINTFN(3,("usb_disconnect_port: disconnect subdevs\n")); for (i = 0; dev->subdevs[i]; i++) { @@ -1365,7 +1367,6 @@ } } - /*usbd_add_dev_event(USB_EVENT_DEVICE_DETACH, dev);*/ dev->bus->devices[dev->address] = NULL; up->device = NULL; usb_free_device(dev); --------------040900000008030903040403 Content-Type: text/plain; name="usb.c.patch" Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="usb.c.patch" --- usb.c.org Tue Oct 21 16:51:25 2003 +++ usb.c Tue Oct 21 19:42:00 2003 @@ -781,7 +781,8 @@ for (ueqi = TAILQ_FIRST(&usb_events); ueqi; ueqi = ueqi_next) { ueqi_next = TAILQ_NEXT(ueqi, next); if (ueqi->ue.u.ue_driver.ue_cookie.cookie == - uep->u.ue_device.udi_cookie.cookie) { + uep->u.ue_device.udi_cookie.cookie && + !USB_EVENT_IS_DETACH(ueqi->ue.ue_type)) { TAILQ_REMOVE(&usb_events, ueqi, next); free(ueqi, M_USBDEV); usb_nevents--; --------------040900000008030903040403--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3F950611.8080302>