Index: sys/dev/ata/chipsets/ata-ahci.c
===================================================================
--- sys/dev/ata/chipsets/ata-ahci.c	(revision 242159)
+++ sys/dev/ata/chipsets/ata-ahci.c	(working copy)
@@ -555,8 +555,10 @@
     if (request->status & ATA_S_ERROR)  
 	request->error = tf_data >> 8;
 
-    /* on control commands read back registers to the request struct */
-    if (request->flags & ATA_R_CONTROL) {
+    /* Read back registers to the request struct. */
+    if ((request->flags & ATA_R_ATAPI) == 0 &&
+	((request->status & ATA_S_ERROR) ||
+	 (request->flags & (ATA_R_CONTROL | ATA_R_NEEDRESULT)))) {
 	u_int8_t *fis = ch->dma.work + ATA_AHCI_FB_OFFSET + 0x40;
 
 	request->u.ata.count = fis[12] | ((u_int16_t)fis[13] << 8);
Index: sys/dev/ata/chipsets/ata-siliconimage.c
===================================================================
--- sys/dev/ata/chipsets/ata-siliconimage.c	(revision 242159)
+++ sys/dev/ata/chipsets/ata-siliconimage.c	(working copy)
@@ -658,8 +658,10 @@
 	}
     }
 
-    /* on control commands read back registers to the request struct */
-    if (request->flags & ATA_R_CONTROL) {
+    /* Read back registers to the request struct. */
+    if ((request->flags & ATA_R_ATAPI) == 0 &&
+	((request->status & ATA_S_ERROR) ||
+	 (request->flags & (ATA_R_CONTROL | ATA_R_NEEDRESULT)))) {
 	request->u.ata.count = prb->fis[12] | ((u_int16_t)prb->fis[13] << 8);
 	request->u.ata.lba = prb->fis[4] | ((u_int64_t)prb->fis[5] << 8) |
 			     ((u_int64_t)prb->fis[6] << 16);
Index: sys/dev/ata/ata-all.c
===================================================================
--- sys/dev/ata/ata-all.c	(revision 242159)
+++ sys/dev/ata/ata-all.c	(working copy)
@@ -1483,6 +1483,8 @@
 		request->u.ata.lba |= ((uint64_t)ccb->ataio.cmd.lba_high << 16) |
 				      ((uint64_t)ccb->ataio.cmd.lba_mid << 8) |
 				       (uint64_t)ccb->ataio.cmd.lba_low;
+		if (ccb->ataio.cmd.flags & CAM_ATAIO_NEEDRESULT)
+			request->flags |= ATA_R_NEEDRESULT;
 		if ((ccb->ccb_h.flags & CAM_DIR_MASK) != CAM_DIR_NONE &&
 		    ccb->ataio.cmd.flags & CAM_ATAIO_DMA)
 			request->flags |= ATA_R_DMA;
Index: sys/dev/ata/ata-all.h
===================================================================
--- sys/dev/ata/ata-all.h	(revision 242159)
+++ sys/dev/ata/ata-all.h	(working copy)
@@ -396,6 +396,7 @@
 #define         ATA_R_REQUEUE           0x00000400
 #define         ATA_R_THREAD            0x00000800
 #define         ATA_R_DIRECT            0x00001000
+#define         ATA_R_NEEDRESULT        0x00002000
 
 #define         ATA_R_ATAPI16           0x00010000
 #define         ATA_R_ATAPI_INTR        0x00020000
Index: sys/dev/ata/ata-lowlevel.c
===================================================================
--- sys/dev/ata/ata-lowlevel.c	(revision 242159)
+++ sys/dev/ata/ata-lowlevel.c	(working copy)
@@ -116,6 +116,7 @@
 		} while (request->status & ATA_S_BUSY && timeout--);
 		if (request->status & ATA_S_ERROR)
 		    request->error = ATA_IDX_INB(ch, ATA_ERROR);
+		ch->hw.tf_read(request);
 		goto begin_finished;
 	    }
 
@@ -253,8 +254,9 @@
 	if (request->flags & ATA_R_TIMEOUT)
 	    goto end_finished;
 
-	/* on control commands read back registers to the request struct */
-	if (request->flags & ATA_R_CONTROL) {
+	/* Read back registers to the request struct. */
+	if ((request->status & ATA_S_ERROR) ||
+	    (request->flags & (ATA_R_CONTROL | ATA_R_NEEDRESULT))) {
 	    ch->hw.tf_read(request);
 	}
 
@@ -332,6 +334,12 @@
 	else if (!(request->flags & ATA_R_TIMEOUT))
 	    request->donecount = request->bytecount;
 
+	/* Read back registers to the request struct. */
+	if ((request->status & ATA_S_ERROR) ||
+	    (request->flags & (ATA_R_CONTROL | ATA_R_NEEDRESULT))) {
+	    ch->hw.tf_read(request);
+	}
+
 	/* release SG list etc */
 	ch->dma.unload(request);