On Fri, Feb 09, 2018 at 03:55:58PM +0100, Julia Lawall wrote:

quoted

quoted

-	ring->txd = dma_alloc_coherent(hsdma2dev(hsdma),
-				       pc->sz_ring, &ring->tphys,
-				       GFP_ATOMIC | __GFP_ZERO);
+	ring->txd = dma_zalloc_coherent(hsdma2dev(hsdma), pc->sz_ring,
+					&ring->tphys, GFP_ATOMIC);
 	if (!ring->txd)
 		return -ENOMEM;

-	memset(ring->txd, 0, pc->sz_ring);

This should be folded back..

Sorry, I'm not sure what you mean by "folded back".

Since the original patch is also up for review in this series, it doesn't
help much to add a change and fix on top in a series. The fix should be
folded back into the original commit. IIRC Fengguang's bot also recommends
so..