On Fri, 2008-11-14 at 18:05 -0500, Bob Copeland wrote: > On Fri, Nov 14, 2008 at 5:44 PM, Luis R. Rodriguez > <lrodriguez@xxxxxxxxxxx> wrote: > > mac80211's ieee80211_register_hw() is often called within the > > probe path so it cannot assume the device's driver structure > > has been attached yet so to create a workqueue instead of > > using driver->name use the wiphy's phy%d name. The name doesn't > > really matter anyway. > > > > This should fix sporadic oopses found when we race to beat the > > driver pointer setting. Not even sure how this was working properly. > > I'd still like to identify the race first. This patch won't hurt > anything, but I'm not confident it will solve any problem either (the > pci_dev->driver setting I was seeing after probe was for a different > struct driver). Also it could make the bug harder to track down. > > In the reported case, wiphy_name will be empty string which isn't too > helpful for the workqueue name either, though it probably won't crash. That can't be right! wiphy_name is assigned right in wiphy_new()! johannes
Attachment:
signature.asc
Description: This is a digitally signed message part
