On Thu, Jun 30, 2016 at 01:08:10PM +0200, Hans de Goede wrote: > 2) spice_usbredir_channel_disconnect_device now running in a thread, rather > then being a synchronous part of the tear-down sequence, this can either be > due to someone calling spice_usbredir_channel_disconnect_device_async() or > someone calling spice_usbredir_channel_reset(). Note that if either happens > just before session_disconnect() gets called we've a problem as then > the channel will be torn down while disconnect is running async, looking > at your backtraces that is exactly what is happening. This all seems to be > fallout from making usb device connect / disconnect async because windows > is oh so terribly slow with this. At e3932bfebbfec7637 time (the commit which introduced the NULL session check), this issue was already existing, and spice_usbredir_channel_disconnect_device_async() hadn't been introduced yet, so this was not introduced by the usbdk/async device connection/disconnection work which has been done later. Does not change the fact that the bug exists :) Christophe
Attachment:
signature.asc
Description: PGP signature
_______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/spice-devel