From f5006eba1c3736a687553bccba1f67cc6d21fb18 Mon Sep 17 00:00:00 2001
From: Miod Vallat <miod@cvs.openbsd.org>
Date: Wed, 25 Aug 2004 20:57:39 +0000
Subject: Set SCMD_PAD in non-integral sector DMA transfers, or the transfer
 never completes; from the old driver, forgotten during the replacement here.

---
 sys/arch/hp300/dev/spc.c | 17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)

(limited to 'sys/arch/hp300')

diff --git a/sys/arch/hp300/dev/spc.c b/sys/arch/hp300/dev/spc.c
index be2ad639bd8..dc04a869107 100644
--- a/sys/arch/hp300/dev/spc.c
+++ b/sys/arch/hp300/dev/spc.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: spc.c,v 1.5 2004/08/21 18:01:55 miod Exp $ */
+/* $OpenBSD: spc.c,v 1.6 2004/08/25 20:57:38 miod Exp $ */
 /* $NetBSD: spc.c,v 1.2 2003/11/17 14:37:59 tsutsui Exp $ */
 
 /*
@@ -212,6 +212,21 @@ spc_dio_dmago(void *arg)
 	cmd = SCMD_XFR;
 	len = sc->sc_dleft;
 
+	if ((len & (DEV_BSIZE -1)) != 0) {
+		cmd |= SCMD_PAD;
+#if 0
+		/*
+		 * XXX - If we don't do this, the last 2 or 4 bytes
+		 * (depending on word/lword DMA) of a read get trashed.
+		 * It looks like it is necessary for the DMA to complete
+		 * before the SPC goes into "pad mode"???  Note: if we
+		 * also do this on a write, the request never completes.
+		 */
+		if ((dsc->sc_dflags & SCSI_DATAIN) != 0)
+			len += 2;
+#endif
+	}
+
 	spc_write(TCH, len >> 16);
 	spc_write(TCM, len >>  8);
 	spc_write(TCL, len);
-- 
cgit v1.2.3