Summary: | USB stack stall cause complete system input loss | ||
---|---|---|---|
Product: | Base System | Reporter: | Arnaud <lacombar> |
Component: | usb | Assignee: | freebsd-usb (Nobody) <usb> |
Status: | Closed Feedback Timeout | ||
Severity: | Affects Only Me | CC: | miwi |
Priority: | Normal | ||
Version: | 8.2-STABLE | ||
Hardware: | Any | ||
OS: | Any |
Description
Arnaud
2011-04-29 21:20:08 UTC
this bug is tightly related to amd64/156726. Considering the following scenario: 1) plug a snd_uaudio(4)-based device 2) open aumix(1) on, pcmN, the mixer device attached to the previously attached device 3) unplug the snd_uaudio(4)-based device [the USB stack busy wait pcmN detach] 4) close aumix(1) [pcmN finally detach] If there is no USB activity (ie, no attach/detach, but "normal" hid activity), the USB stack recovers. If another USB device is attached/detached while the stack busy wait on pcmN detachment (between step 3 and step 4), events ends up being queued. Once step 4 is done, events gets processed, however, once completed the USB stall, ie. no more mouse or keyboard. At this point, I do not see any USB related message while the system halts. Responsible Changed From-To: freebsd-amd64->freebsd-usb reclassify. On Monday 02 May 2011 02:26:29 linimon@freebsd.org wrote: > Synopsis: USB stack stall cause complete system input loss > > Responsible-Changed-From-To: freebsd-amd64->freebsd-usb > Responsible-Changed-By: linimon > Responsible-Changed-When: Mon May 2 00:25:49 UTC 2011 > Responsible-Changed-Why: > reclassify. > > http://www.freebsd.org/cgi/query-pr.cgi?pr=156725 Hi, All attach and detach methods are run from a single thread, the root HUB thread. If any device driver is blocking the detach routine for some reason, other devices on the same USB controller will be prevented from enumerating aswell. --HPS |