Hi Johan - On Mon, 14 Jan 2013, Johan Hedberg wrote:
Hi, I was initially only going to fix the thing in the last patch of this set but after starting to look at the code I found several other issues and hence the rest of the patches. The first three patches fix work queue handling. We should, whenever possible, avoid the system-global workqueue (the schedule_*work functions) and use hdev-specific ones instead.
There are also multiple system workqueues beyond the default system_wq, like system_long_wq and system_unbound_wq. Those may be a better fit and you wouldn't have to manage additional per-device workqueues.
However, we can't use hci_request() with the usual workqueue since that would block e.g. hci_send_cmd from completing so a second work queue is needed. The last patch fixes a race with setting the scan mode that I was seeing every once in a while through a connectable or discoverable test case of mgmt-tester failing. There are several other HCI commands with the same potential issue (like the class of device and the local name) but I haven't gotten around to fixing those yet.
Regards, -- Mat Martineau Employee of Qualcomm Innovation Center, Inc. Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation -- To unsubscribe from this list: send the line "unsubscribe linux-bluetooth" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
