On 15-01-21, 14:52, Dave Jiang wrote: > Sanjay reported the following kernel splat after running dmatest for stress > testing. The current code is giving up the spinlock in the middle of > a completion list walk, and that opens up opportunity for list corruption > if another thread touches the list at the same time. In order to make sure > the list is always protected, the hardware completed descriptors will be > put on a local list to be completed with callbacks from the outside of > the list lock. Applied, thanks -- ~Vinod
