brcmfmac: fix regression in brcmf_sdio_txpkt_hdalign()

Recent change in brcmf_sdio_txpkt_hdalign() changed the
behavior and now always returns 0. This resulted in a
regression which basically renders the device useless.

Fixes: 270a6c1f ("brcmfmac: rework headroom check in .start_xmit()")
Reported-by: S. Gilles <sgilles@math.umd.edu>
Tested-by: S. Gilles <sgilles@math.umd.edu>
Signed-off-by: Arend van Spriel <arend.vanspriel@broadcom.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

drivers/net/wireless/broadcom/brcm80211/brcmfmac/sdio.c

@@ -2053,12 +2053,13 @@ static int brcmf_sdio_txpkt_hdalign(struct brcmf_sdio *bus, struct sk_buff *pkt)
 				atomic_inc(&stats->pktcow_failed);
 				return -ENOMEM;
 			}
+			head_pad = 0;
 		}
 		skb_push(pkt, head_pad);
 		dat_buf = (u8 *)(pkt->data);
 	}
 	memset(dat_buf, 0, head_pad + bus->tx_hdrlen);
-	return 0;
+	return head_pad;
 }
 
 /**