Il 02/07/2014 13:12, ChenLiang ha scritto:
This patch avoid freeing dbs by dma_complete when dma_aio_cancel is running. Because dma_complete also will be called by dma_bdrv_cb. So double free will never happen.
Yes, you already said that. But I'm not sure _why_ the patch avoids freeing dbs, especially since the patch adds a free in dma_aio_cancel itself. My patch shouldn't have any effect. Where is my analysis of it wrong?
Paolo
