On 25-10-22, 12:02, Tudor Ambarus wrote: > v2: > - reorder patches so that fixes come first -> easier to backport to > stable kernels. > - drop the devm_request_irq() patch as we had to disable the irq anyway > in remove() in order to avoid spurios IRQs. Using devm variant brings no > palpable benefit. > - reword pm_ptr commit message > > > at_hdmac driver had poor list handling and concurrency bugs. > We experienced calling of the completion call twice for the > same descriptor. Peter Rosin encountered the same while > reporting a different bug: > https://lore.kernel.org/lkml/13c6c9a2-6db5-c3bf-349b-4c127ad3496a@xxxxxxxxxx/ > > Two sets of tests were performed: > 1/ tested just the fixes, to make sure everything is fine and the > concurrency bugs are squashed even without the conversion to virt-dma. > All went fine. > 2/ tested the entire series including the conversion the virt-dma > All went fine. > > I tested NAND (prep_dma_memcpy), MMC (prep_dma_slave_sg), > usart (cyclic mode), dmatest (memcpy, memset). > With the conversion to virt-dma I replaced the election of a new transfer > in the tasklet with the election of the new transfer in the interrupt > handler. We should have a shorter idle window as we remove the scheduling > latency of the tasklet. Using mtd_speedtest showed similar performances > when using NAND with DMA. That could be because of using a low timming > mode on NAND. This does not apply on dmaengine-fixes, can you please rebase and resend -- ~Vinod
