diff --git a/[refs] b/[refs]
index 1b203b4c8d6c..b5c698a41713 100644
--- a/[refs]
+++ b/[refs]
@@ -1,2 +1,2 @@
 ---
-refs/heads/master: 054ee8fd39f1b5d50e803f126b63f400d631eea4
+refs/heads/master: 08db2a701bd63d0e36f06a29137bf016a907cf71
diff --git a/trunk/drivers/pci/quirks.c b/trunk/drivers/pci/quirks.c
index bbd9c2323d8c..5627ce1d2b32 100644
--- a/trunk/drivers/pci/quirks.c
+++ b/trunk/drivers/pci/quirks.c
@@ -356,7 +356,7 @@ static void piix4_mem_quirk(struct pci_dev *dev, const char *name, unsigned int
 /*
  * PIIX4 ACPI: Two IO regions pointed to by longwords at
  *	0x40 (64 bytes of ACPI registers)
- *	0x90 (32 bytes of SMB registers)
+ *	0x90 (16 bytes of SMB registers)
  * and a few strange programmable PIIX4 device resources.
  */
 static void __devinit quirk_piix4_acpi(struct pci_dev *dev)
@@ -366,7 +366,7 @@ static void __devinit quirk_piix4_acpi(struct pci_dev *dev)
 	pci_read_config_dword(dev, 0x40, &region);
 	quirk_io_region(dev, region, 64, PCI_BRIDGE_RESOURCES, "PIIX4 ACPI");
 	pci_read_config_dword(dev, 0x90, &region);
-	quirk_io_region(dev, region, 32, PCI_BRIDGE_RESOURCES+1, "PIIX4 SMB");
+	quirk_io_region(dev, region, 16, PCI_BRIDGE_RESOURCES+1, "PIIX4 SMB");
 
 	/* Device resource A has enables for some of the other ones */
 	pci_read_config_dword(dev, 0x5c, &res_a);
diff --git a/trunk/drivers/scsi/ahci.c b/trunk/drivers/scsi/ahci.c
index 25a47d7afa28..fe8187d6f58b 100644
--- a/trunk/drivers/scsi/ahci.c
+++ b/trunk/drivers/scsi/ahci.c
@@ -192,6 +192,7 @@ static void ahci_port_stop(struct ata_port *ap);
 static void ahci_tf_read(struct ata_port *ap, struct ata_taskfile *tf);
 static void ahci_qc_prep(struct ata_queued_cmd *qc);
 static u8 ahci_check_status(struct ata_port *ap);
+static u8 ahci_check_err(struct ata_port *ap);
 static inline int ahci_host_intr(struct ata_port *ap, struct ata_queued_cmd *qc);
 static void ahci_remove_one (struct pci_dev *pdev);
 
@@ -220,6 +221,7 @@ static const struct ata_port_operations ahci_ops = {
 
 	.check_status		= ahci_check_status,
 	.check_altstatus	= ahci_check_status,
+	.check_err		= ahci_check_err,
 	.dev_select		= ata_noop_dev_select,
 
 	.tf_read		= ahci_tf_read,
@@ -312,15 +314,8 @@ static int ahci_port_start(struct ata_port *ap)
 		return -ENOMEM;
 	memset(pp, 0, sizeof(*pp));
 
-	ap->pad = dma_alloc_coherent(dev, ATA_DMA_PAD_BUF_SZ, &ap->pad_dma, GFP_KERNEL);
-	if (!ap->pad) {
-		kfree(pp);
-		return -ENOMEM;
-	}
-
 	mem = dma_alloc_coherent(dev, AHCI_PORT_PRIV_DMA_SZ, &mem_dma, GFP_KERNEL);
 	if (!mem) {
-		dma_free_coherent(dev, ATA_DMA_PAD_BUF_SZ, ap->pad, ap->pad_dma);
 		kfree(pp);
 		return -ENOMEM;
 	}
@@ -396,7 +391,6 @@ static void ahci_port_stop(struct ata_port *ap)
 	ap->private_data = NULL;
 	dma_free_coherent(dev, AHCI_PORT_PRIV_DMA_SZ,
 			  pp->cmd_slot, pp->cmd_slot_dma);
-	dma_free_coherent(dev, ATA_DMA_PAD_BUF_SZ, ap->pad, ap->pad_dma);
 	kfree(pp);
 }
 
@@ -464,6 +458,13 @@ static u8 ahci_check_status(struct ata_port *ap)
 	return readl(mmio + PORT_TFDATA) & 0xFF;
 }
 
+static u8 ahci_check_err(struct ata_port *ap)
+{
+	void __iomem *mmio = (void __iomem *) ap->ioaddr.cmd_addr;
+
+	return (readl(mmio + PORT_TFDATA) >> 8) & 0xFF;
+}
+
 static void ahci_tf_read(struct ata_port *ap, struct ata_taskfile *tf)
 {
 	struct ahci_port_priv *pp = ap->private_data;
@@ -475,23 +476,23 @@ static void ahci_tf_read(struct ata_port *ap, struct ata_taskfile *tf)
 static void ahci_fill_sg(struct ata_queued_cmd *qc)
 {
 	struct ahci_port_priv *pp = qc->ap->private_data;
-	struct scatterlist *sg;
-	struct ahci_sg *ahci_sg;
+	unsigned int i;
 
 	VPRINTK("ENTER\n");
 
 	/*
 	 * Next, the S/G list.
 	 */
-	ahci_sg = pp->cmd_tbl_sg;
-	ata_for_each_sg(sg, qc) {
-		dma_addr_t addr = sg_dma_address(sg);
-		u32 sg_len = sg_dma_len(sg);
-
-		ahci_sg->addr = cpu_to_le32(addr & 0xffffffff);
-		ahci_sg->addr_hi = cpu_to_le32((addr >> 16) >> 16);
-		ahci_sg->flags_size = cpu_to_le32(sg_len - 1);
-		ahci_sg++;
+	for (i = 0; i < qc->n_elem; i++) {
+		u32 sg_len;
+		dma_addr_t addr;
+
+		addr = sg_dma_address(&qc->sg[i]);
+		sg_len = sg_dma_len(&qc->sg[i]);
+
+		pp->cmd_tbl_sg[i].addr = cpu_to_le32(addr & 0xffffffff);
+		pp->cmd_tbl_sg[i].addr_hi = cpu_to_le32((addr >> 16) >> 16);
+		pp->cmd_tbl_sg[i].flags_size = cpu_to_le32(sg_len - 1);
 	}
 }
 
@@ -608,7 +609,7 @@ static void ahci_eng_timeout(struct ata_port *ap)
 	 	 * not being called from the SCSI EH.
 	 	 */
 		qc->scsidone = scsi_finish_command;
-		ata_qc_complete(qc, AC_ERR_OTHER);
+		ata_qc_complete(qc, ATA_ERR);
 	}
 
 	spin_unlock_irqrestore(&host_set->lock, flags);
@@ -637,7 +638,7 @@ static inline int ahci_host_intr(struct ata_port *ap, struct ata_queued_cmd *qc)
 	if (status & PORT_IRQ_FATAL) {
 		ahci_intr_error(ap, status);
 		if (qc)
-			ata_qc_complete(qc, AC_ERR_OTHER);
+			ata_qc_complete(qc, ATA_ERR);
 	}
 
 	return 1;
diff --git a/trunk/drivers/scsi/libata-core.c b/trunk/drivers/scsi/libata-core.c
index 6cab14965cc8..5ca97605ff35 100644
--- a/trunk/drivers/scsi/libata-core.c
+++ b/trunk/drivers/scsi/libata-core.c
@@ -372,7 +372,7 @@ static void ata_tf_read_pio(struct ata_port *ap, struct ata_taskfile *tf)
 	struct ata_ioports *ioaddr = &ap->ioaddr;
 
 	tf->command = ata_check_status(ap);
-	tf->feature = inb(ioaddr->error_addr);
+	tf->feature = ata_chk_err(ap);
 	tf->nsect = inb(ioaddr->nsect_addr);
 	tf->lbal = inb(ioaddr->lbal_addr);
 	tf->lbam = inb(ioaddr->lbam_addr);
@@ -406,7 +406,7 @@ static void ata_tf_read_mmio(struct ata_port *ap, struct ata_taskfile *tf)
 	struct ata_ioports *ioaddr = &ap->ioaddr;
 
 	tf->command = ata_check_status(ap);
-	tf->feature = readb((void __iomem *)ioaddr->error_addr);
+	tf->feature = ata_chk_err(ap);
 	tf->nsect = readb((void __iomem *)ioaddr->nsect_addr);
 	tf->lbal = readb((void __iomem *)ioaddr->lbal_addr);
 	tf->lbam = readb((void __iomem *)ioaddr->lbam_addr);
@@ -526,6 +526,30 @@ u8 ata_altstatus(struct ata_port *ap)
 }
 
 
+/**
+ *	ata_chk_err - Read device error reg
+ *	@ap: port where the device is
+ *
+ *	Reads ATA taskfile error register for
+ *	currently-selected device and return its value.
+ *
+ *	Note: may NOT be used as the check_err() entry in
+ *	ata_port_operations.
+ *
+ *	LOCKING:
+ *	Inherited from caller.
+ */
+u8 ata_chk_err(struct ata_port *ap)
+{
+	if (ap->ops->check_err)
+		return ap->ops->check_err(ap);
+
+	if (ap->flags & ATA_FLAG_MMIO) {
+		return readb((void __iomem *) ap->ioaddr.error_addr);
+	}
+	return inb(ap->ioaddr.error_addr);
+}
+
 /**
  *	ata_tf_to_fis - Convert ATA taskfile to SATA FIS structure
  *	@tf: Taskfile to convert
@@ -878,8 +902,8 @@ static u8 ata_dev_try_classify(struct ata_port *ap, unsigned int device)
 
 	memset(&tf, 0, sizeof(tf));
 
+	err = ata_chk_err(ap);
 	ap->ops->tf_read(ap, &tf);
-	err = tf.feature;
 
 	dev->class = ATA_DEV_NONE;
 
@@ -1116,6 +1140,7 @@ static void ata_dev_identify(struct ata_port *ap, unsigned int device)
 	unsigned int major_version;
 	u16 tmp;
 	unsigned long xfer_modes;
+	u8 status;
 	unsigned int using_edd;
 	DECLARE_COMPLETION(wait);
 	struct ata_queued_cmd *qc;
@@ -1169,11 +1194,8 @@ static void ata_dev_identify(struct ata_port *ap, unsigned int device)
 	else
 		wait_for_completion(&wait);
 
-	spin_lock_irqsave(&ap->host_set->lock, flags);
-	ap->ops->tf_read(ap, &qc->tf);
-	spin_unlock_irqrestore(&ap->host_set->lock, flags);
-
-	if (qc->tf.command & ATA_ERR) {
+	status = ata_chk_status(ap);
+	if (status & ATA_ERR) {
 		/*
 		 * arg!  EDD works for all test cases, but seems to return
 		 * the ATA signature for some ATAPI devices.  Until the
@@ -1186,7 +1208,7 @@ static void ata_dev_identify(struct ata_port *ap, unsigned int device)
 		 * to have this problem.
 		 */
 		if ((using_edd) && (qc->tf.command == ATA_CMD_ID_ATA)) {
-			u8 err = qc->tf.feature;
+			u8 err = ata_chk_err(ap);
 			if (err & ATA_ABORTED) {
 				dev->class = ATA_DEV_ATAPI;
 				qc->cursg = 0;
@@ -2427,9 +2449,8 @@ static void ata_dev_init_params(struct ata_port *ap, struct ata_device *dev)
 static void ata_sg_clean(struct ata_queued_cmd *qc)
 {
 	struct ata_port *ap = qc->ap;
-	struct scatterlist *sg = qc->__sg;
+	struct scatterlist *sg = qc->sg;
 	int dir = qc->dma_dir;
-	void *pad_buf = NULL;
 
 	assert(qc->flags & ATA_QCFLAG_DMAMAP);
 	assert(sg != NULL);
@@ -2439,35 +2460,14 @@ static void ata_sg_clean(struct ata_queued_cmd *qc)
 
 	DPRINTK("unmapping %u sg elements\n", qc->n_elem);
 
-	/* if we padded the buffer out to 32-bit bound, and data
-	 * xfer direction is from-device, we must copy from the
-	 * pad buffer back into the supplied buffer
-	 */
-	if (qc->pad_len && !(qc->tf.flags & ATA_TFLAG_WRITE))
-		pad_buf = ap->pad + (qc->tag * ATA_DMA_PAD_SZ);
-
-	if (qc->flags & ATA_QCFLAG_SG) {
+	if (qc->flags & ATA_QCFLAG_SG)
 		dma_unmap_sg(ap->host_set->dev, sg, qc->n_elem, dir);
-		/* restore last sg */
-		sg[qc->orig_n_elem - 1].length += qc->pad_len;
-		if (pad_buf) {
-			struct scatterlist *psg = &qc->pad_sgent;
-			void *addr = kmap_atomic(psg->page, KM_IRQ0);
-			memcpy(addr + psg->offset, pad_buf, qc->pad_len);
-			kunmap_atomic(psg->page, KM_IRQ0);
-		}
-	} else {
+	else
 		dma_unmap_single(ap->host_set->dev, sg_dma_address(&sg[0]),
 				 sg_dma_len(&sg[0]), dir);
-		/* restore sg */
-		sg->length += qc->pad_len;
-		if (pad_buf)
-			memcpy(qc->buf_virt + sg->length - qc->pad_len,
-			       pad_buf, qc->pad_len);
-	}
 
 	qc->flags &= ~ATA_QCFLAG_DMAMAP;
-	qc->__sg = NULL;
+	qc->sg = NULL;
 }
 
 /**
@@ -2483,15 +2483,15 @@ static void ata_sg_clean(struct ata_queued_cmd *qc)
  */
 static void ata_fill_sg(struct ata_queued_cmd *qc)
 {
+	struct scatterlist *sg = qc->sg;
 	struct ata_port *ap = qc->ap;
-	struct scatterlist *sg;
-	unsigned int idx;
+	unsigned int idx, nelem;
 
-	assert(qc->__sg != NULL);
+	assert(sg != NULL);
 	assert(qc->n_elem > 0);
 
 	idx = 0;
-	ata_for_each_sg(sg, qc) {
+	for (nelem = qc->n_elem; nelem; nelem--,sg++) {
 		u32 addr, offset;
 		u32 sg_len, len;
 
@@ -2577,18 +2577,12 @@ void ata_qc_prep(struct ata_queued_cmd *qc)
 
 void ata_sg_init_one(struct ata_queued_cmd *qc, void *buf, unsigned int buflen)
 {
-	struct scatterlist *sg;
-
 	qc->flags |= ATA_QCFLAG_SINGLE;
 
-	memset(&qc->sgent, 0, sizeof(qc->sgent));
-	qc->__sg = &qc->sgent;
+	qc->sg = &qc->sgent;
 	qc->n_elem = 1;
-	qc->orig_n_elem = 1;
 	qc->buf_virt = buf;
-
-	sg = qc->__sg;
-	sg_init_one(sg, buf, buflen);
+	sg_init_one(qc->sg, buf, buflen);
 }
 
 /**
@@ -2609,9 +2603,8 @@ void ata_sg_init(struct ata_queued_cmd *qc, struct scatterlist *sg,
 		 unsigned int n_elem)
 {
 	qc->flags |= ATA_QCFLAG_SG;
-	qc->__sg = sg;
+	qc->sg = sg;
 	qc->n_elem = n_elem;
-	qc->orig_n_elem = n_elem;
 }
 
 /**
@@ -2631,32 +2624,9 @@ static int ata_sg_setup_one(struct ata_queued_cmd *qc)
 {
 	struct ata_port *ap = qc->ap;
 	int dir = qc->dma_dir;
-	struct scatterlist *sg = qc->__sg;
+	struct scatterlist *sg = qc->sg;
 	dma_addr_t dma_address;
 
-	/* we must lengthen transfers to end on a 32-bit boundary */
-	qc->pad_len = sg->length & 3;
-	if (qc->pad_len) {
-		void *pad_buf = ap->pad + (qc->tag * ATA_DMA_PAD_SZ);
-		struct scatterlist *psg = &qc->pad_sgent;
-
-		assert(qc->dev->class == ATA_DEV_ATAPI);
-
-		memset(pad_buf, 0, ATA_DMA_PAD_SZ);
-
-		if (qc->tf.flags & ATA_TFLAG_WRITE)
-			memcpy(pad_buf, qc->buf_virt + sg->length - qc->pad_len,
-			       qc->pad_len);
-
-		sg_dma_address(psg) = ap->pad_dma + (qc->tag * ATA_DMA_PAD_SZ);
-		sg_dma_len(psg) = ATA_DMA_PAD_SZ;
-		/* trim sg */
-		sg->length -= qc->pad_len;
-
-		DPRINTK("padding done, sg->length=%u pad_len=%u\n",
-			sg->length, qc->pad_len);
-	}
-
 	dma_address = dma_map_single(ap->host_set->dev, qc->buf_virt,
 				     sg->length, dir);
 	if (dma_mapping_error(dma_address))
@@ -2688,47 +2658,12 @@ static int ata_sg_setup_one(struct ata_queued_cmd *qc)
 static int ata_sg_setup(struct ata_queued_cmd *qc)
 {
 	struct ata_port *ap = qc->ap;
-	struct scatterlist *sg = qc->__sg;
-	struct scatterlist *lsg = &sg[qc->n_elem - 1];
+	struct scatterlist *sg = qc->sg;
 	int n_elem, dir;
 
 	VPRINTK("ENTER, ata%u\n", ap->id);
 	assert(qc->flags & ATA_QCFLAG_SG);
 
-	/* we must lengthen transfers to end on a 32-bit boundary */
-	qc->pad_len = lsg->length & 3;
-	if (qc->pad_len) {
-		void *pad_buf = ap->pad + (qc->tag * ATA_DMA_PAD_SZ);
-		struct scatterlist *psg = &qc->pad_sgent;
-		unsigned int offset;
-
-		assert(qc->dev->class == ATA_DEV_ATAPI);
-
-		memset(pad_buf, 0, ATA_DMA_PAD_SZ);
-
-		/*
-		 * psg->page/offset are used to copy to-be-written
-		 * data in this function or read data in ata_sg_clean.
-		 */
-		offset = lsg->offset + lsg->length - qc->pad_len;
-		psg->page = nth_page(lsg->page, offset >> PAGE_SHIFT);
-		psg->offset = offset_in_page(offset);
-
-		if (qc->tf.flags & ATA_TFLAG_WRITE) {
-			void *addr = kmap_atomic(psg->page, KM_IRQ0);
-			memcpy(pad_buf, addr + psg->offset, qc->pad_len);
-			kunmap_atomic(psg->page, KM_IRQ0);
-		}
-
-		sg_dma_address(psg) = ap->pad_dma + (qc->tag * ATA_DMA_PAD_SZ);
-		sg_dma_len(psg) = ATA_DMA_PAD_SZ;
-		/* trim last sg */
-		lsg->length -= qc->pad_len;
-
-		DPRINTK("padding done, sg[%d].length=%u pad_len=%u\n",
-			qc->n_elem - 1, lsg->length, qc->pad_len);
-	}
-
 	dir = qc->dma_dir;
 	n_elem = dma_map_sg(ap->host_set->dev, sg, qc->n_elem, dir);
 	if (n_elem < 1)
@@ -2750,7 +2685,7 @@ static int ata_sg_setup(struct ata_queued_cmd *qc)
  *	None.  (grabs host lock)
  */
 
-void ata_poll_qc_complete(struct ata_queued_cmd *qc, unsigned int err_mask)
+void ata_poll_qc_complete(struct ata_queued_cmd *qc, u8 drv_stat)
 {
 	struct ata_port *ap = qc->ap;
 	unsigned long flags;
@@ -2758,7 +2693,7 @@ void ata_poll_qc_complete(struct ata_queued_cmd *qc, unsigned int err_mask)
 	spin_lock_irqsave(&ap->host_set->lock, flags);
 	ap->flags &= ~ATA_FLAG_NOINTR;
 	ata_irq_on(ap);
-	ata_qc_complete(qc, err_mask);
+	ata_qc_complete(qc, drv_stat);
 	spin_unlock_irqrestore(&ap->host_set->lock, flags);
 }
 
@@ -2855,7 +2790,7 @@ static int ata_pio_complete (struct ata_port *ap)
 
 	ap->hsm_task_state = HSM_ST_IDLE;
 
-	ata_poll_qc_complete(qc, 0);
+	ata_poll_qc_complete(qc, drv_stat);
 
 	/* another command may start at this point */
 
@@ -3004,7 +2939,7 @@ static void ata_data_xfer(struct ata_port *ap, unsigned char *buf,
 static void ata_pio_sector(struct ata_queued_cmd *qc)
 {
 	int do_write = (qc->tf.flags & ATA_TFLAG_WRITE);
-	struct scatterlist *sg = qc->__sg;
+	struct scatterlist *sg = qc->sg;
 	struct ata_port *ap = qc->ap;
 	struct page *page;
 	unsigned int offset;
@@ -3054,7 +2989,7 @@ static void ata_pio_sector(struct ata_queued_cmd *qc)
 static void __atapi_pio_bytes(struct ata_queued_cmd *qc, unsigned int bytes)
 {
 	int do_write = (qc->tf.flags & ATA_TFLAG_WRITE);
-	struct scatterlist *sg = qc->__sg;
+	struct scatterlist *sg = qc->sg;
 	struct ata_port *ap = qc->ap;
 	struct page *page;
 	unsigned char *buf;
@@ -3087,7 +3022,7 @@ static void __atapi_pio_bytes(struct ata_queued_cmd *qc, unsigned int bytes)
 		return;
 	}
 
-	sg = &qc->__sg[qc->cursg];
+	sg = &qc->sg[qc->cursg];
 
 	page = sg->page;
 	offset = sg->offset + qc->cursg_ofs;
@@ -3223,15 +3158,18 @@ static void ata_pio_block(struct ata_port *ap)
 static void ata_pio_error(struct ata_port *ap)
 {
 	struct ata_queued_cmd *qc;
-
-	printk(KERN_WARNING "ata%u: PIO error\n", ap->id);
+	u8 drv_stat;
 
 	qc = ata_qc_from_tag(ap, ap->active_tag);
 	assert(qc != NULL);
 
+	drv_stat = ata_chk_status(ap);
+	printk(KERN_WARNING "ata%u: PIO error, drv_stat 0x%x\n",
+	       ap->id, drv_stat);
+
 	ap->hsm_task_state = HSM_ST_IDLE;
 
-	ata_poll_qc_complete(qc, AC_ERR_ATA_BUS);
+	ata_poll_qc_complete(qc, drv_stat | ATA_ERR);
 }
 
 static void ata_pio_task(void *_data)
@@ -3354,7 +3292,7 @@ static void ata_qc_timeout(struct ata_queued_cmd *qc)
 		       ap->id, qc->tf.command, drv_stat, host_stat);
 
 		/* complete taskfile transaction */
-		ata_qc_complete(qc, ac_err_mask(drv_stat));
+		ata_qc_complete(qc, drv_stat);
 		break;
 	}
 
@@ -3444,7 +3382,7 @@ struct ata_queued_cmd *ata_qc_new_init(struct ata_port *ap,
 
 	qc = ata_qc_new(ap);
 	if (qc) {
-		qc->__sg = NULL;
+		qc->sg = NULL;
 		qc->flags = 0;
 		qc->scsicmd = NULL;
 		qc->ap = ap;
@@ -3459,7 +3397,7 @@ struct ata_queued_cmd *ata_qc_new_init(struct ata_port *ap,
 	return qc;
 }
 
-int ata_qc_complete_noop(struct ata_queued_cmd *qc, unsigned int err_mask)
+int ata_qc_complete_noop(struct ata_queued_cmd *qc, u8 drv_stat)
 {
 	return 0;
 }
@@ -3518,7 +3456,7 @@ void ata_qc_free(struct ata_queued_cmd *qc)
  *	spin_lock_irqsave(host_set lock)
  */
 
-void ata_qc_complete(struct ata_queued_cmd *qc, unsigned int err_mask)
+void ata_qc_complete(struct ata_queued_cmd *qc, u8 drv_stat)
 {
 	int rc;
 
@@ -3535,7 +3473,7 @@ void ata_qc_complete(struct ata_queued_cmd *qc, unsigned int err_mask)
 	qc->flags &= ~ATA_QCFLAG_ACTIVE;
 
 	/* call completion callback */
-	rc = qc->complete_fn(qc, err_mask);
+	rc = qc->complete_fn(qc, drv_stat);
 
 	/* if callback indicates not to complete command (non-zero),
 	 * return immediately
@@ -3973,7 +3911,7 @@ inline unsigned int ata_host_intr (struct ata_port *ap,
 		ap->ops->irq_clear(ap);
 
 		/* complete taskfile transaction */
-		ata_qc_complete(qc, ac_err_mask(status));
+		ata_qc_complete(qc, status);
 		break;
 
 	default:
@@ -4068,7 +4006,7 @@ static void atapi_packet_task(void *_data)
 	/* sleep-wait for BSY to clear */
 	DPRINTK("busy wait\n");
 	if (ata_busy_sleep(ap, ATA_TMOUT_CDB_QUICK, ATA_TMOUT_CDB))
-		goto err_out_status;
+		goto err_out;
 
 	/* make sure DRQ is set */
 	status = ata_chk_status(ap);
@@ -4105,10 +4043,8 @@ static void atapi_packet_task(void *_data)
 
 	return;
 
-err_out_status:
-	status = ata_chk_status(ap);
 err_out:
-	ata_poll_qc_complete(qc, __ac_err_mask(status));
+	ata_poll_qc_complete(qc, ATA_ERR);
 }
 
 
@@ -4133,12 +4069,6 @@ int ata_port_start (struct ata_port *ap)
 	if (!ap->prd)
 		return -ENOMEM;
 
-	ap->pad = dma_alloc_coherent(dev, ATA_DMA_PAD_BUF_SZ, &ap->pad_dma, GFP_KERNEL);
-	if (!ap->pad) {
-		dma_free_coherent(dev, ATA_PRD_TBL_SZ, ap->prd, ap->prd_dma);
-		return -ENOMEM;
-	}
-
 	DPRINTK("prd alloc, virt %p, dma %llx\n", ap->prd, (unsigned long long) ap->prd_dma);
 
 	return 0;
@@ -4162,7 +4092,6 @@ void ata_port_stop (struct ata_port *ap)
 	struct device *dev = ap->host_set->dev;
 
 	dma_free_coherent(dev, ATA_PRD_TBL_SZ, ap->prd, ap->prd_dma);
-	dma_free_coherent(dev, ATA_DMA_PAD_BUF_SZ, ap->pad, ap->pad_dma);
 }
 
 void ata_host_stop (struct ata_host_set *host_set)
@@ -4938,6 +4867,7 @@ EXPORT_SYMBOL_GPL(ata_tf_to_fis);
 EXPORT_SYMBOL_GPL(ata_tf_from_fis);
 EXPORT_SYMBOL_GPL(ata_check_status);
 EXPORT_SYMBOL_GPL(ata_altstatus);
+EXPORT_SYMBOL_GPL(ata_chk_err);
 EXPORT_SYMBOL_GPL(ata_exec_command);
 EXPORT_SYMBOL_GPL(ata_port_start);
 EXPORT_SYMBOL_GPL(ata_port_stop);
diff --git a/trunk/drivers/scsi/libata-scsi.c b/trunk/drivers/scsi/libata-scsi.c
index 5574520bf1c2..89a04b1a5a0e 100644
--- a/trunk/drivers/scsi/libata-scsi.c
+++ b/trunk/drivers/scsi/libata-scsi.c
@@ -354,10 +354,10 @@ struct ata_queued_cmd *ata_scsi_qc_new(struct ata_port *ap,
 		qc->scsidone = done;
 
 		if (cmd->use_sg) {
-			qc->__sg = (struct scatterlist *) cmd->request_buffer;
+			qc->sg = (struct scatterlist *) cmd->request_buffer;
 			qc->n_elem = cmd->use_sg;
 		} else {
-			qc->__sg = &qc->sgent;
+			qc->sg = &qc->sgent;
 			qc->n_elem = 1;
 		}
 	} else {
@@ -560,7 +560,7 @@ void ata_gen_ata_desc_sense(struct ata_queued_cmd *qc)
 	 * Use ata_to_sense_error() to map status register bits
 	 * onto sense key, asc & ascq.
 	 */
-	if (tf->command & (ATA_BUSY | ATA_DF | ATA_ERR | ATA_DRQ)) {
+	if (unlikely(tf->command & (ATA_BUSY | ATA_DF | ATA_ERR | ATA_DRQ))) {
 		ata_to_sense_error(qc->ap->id, tf->command, tf->feature,
 				   &sb[1], &sb[2], &sb[3]);
 		sb[1] &= 0x0f;
@@ -635,7 +635,7 @@ void ata_gen_fixed_sense(struct ata_queued_cmd *qc)
 	 * Use ata_to_sense_error() to map status register bits
 	 * onto sense key, asc & ascq.
 	 */
-	if (tf->command & (ATA_BUSY | ATA_DF | ATA_ERR | ATA_DRQ)) {
+	if (unlikely(tf->command & (ATA_BUSY | ATA_DF | ATA_ERR | ATA_DRQ))) {
 		ata_to_sense_error(qc->ap->id, tf->command, tf->feature,
 				   &sb[2], &sb[12], &sb[13]);
 		sb[2] &= 0x0f;
@@ -644,11 +644,7 @@ void ata_gen_fixed_sense(struct ata_queued_cmd *qc)
 	sb[0] = 0x70;
 	sb[7] = 0x0a;
 
-	if (tf->flags & ATA_TFLAG_LBA48) {
-		/* TODO: find solution for LBA48 descriptors */
-	}
-
-	else if (tf->flags & ATA_TFLAG_LBA) {
+	if (tf->flags & ATA_TFLAG_LBA && !(tf->flags & ATA_TFLAG_LBA48)) {
 		/* A small (28b) LBA will fit in the 32b info field */
 		sb[0] |= 0x80;		/* set valid bit */
 		sb[3] = tf->device & 0x0f;
@@ -656,10 +652,6 @@ void ata_gen_fixed_sense(struct ata_queued_cmd *qc)
 		sb[5] = tf->lbam;
 		sb[6] = tf->lbal;
 	}
-
-	else {
-		/* TODO: C/H/S */
-	}
 }
 
 /**
@@ -701,16 +693,6 @@ int ata_scsi_slave_config(struct scsi_device *sdev)
 			 */
 			blk_queue_max_sectors(sdev->request_queue, 2048);
 		}
-
-		/*
-		 * SATA DMA transfers must be multiples of 4 byte, so
-		 * we need to pad ATAPI transfers using an extra sg.
-		 * Decrement max hw segments accordingly.
-		 */
-		if (dev->class == ATA_DEV_ATAPI) {
-			request_queue_t *q = sdev->request_queue;
-			blk_queue_max_hw_segments(q, q->max_hw_segments - 1);
-		}
 	}
 
 	return 0;	/* scsi layer doesn't check return value, sigh */
@@ -1217,12 +1199,10 @@ static unsigned int ata_scsi_rw_xlat(struct ata_queued_cmd *qc, const u8 *scsicm
 	return 1;
 }
 
-static int ata_scsi_qc_complete(struct ata_queued_cmd *qc,
-				unsigned int err_mask)
+static int ata_scsi_qc_complete(struct ata_queued_cmd *qc, u8 drv_stat)
 {
 	struct scsi_cmnd *cmd = qc->scsicmd;
-	u8 *cdb = cmd->cmnd;
- 	int need_sense = (err_mask != 0);
+ 	int need_sense = drv_stat & (ATA_ERR | ATA_BUSY | ATA_DRQ);
 
 	/* For ATA pass thru (SAT) commands, generate a sense block if
 	 * user mandated it or if there's an error.  Note that if we
@@ -1231,8 +1211,8 @@ static int ata_scsi_qc_complete(struct ata_queued_cmd *qc,
 	 * whether the command completed successfully or not. If there
 	 * was no error, SK, ASC and ASCQ will all be zero.
 	 */
-	if (((cdb[0] == ATA_16) || (cdb[0] == ATA_12)) &&
- 	    ((cdb[2] & 0x20) || need_sense)) {
+	if (((cmd->cmnd[0] == ATA_16) || (cmd->cmnd[0] == ATA_12)) &&
+ 	    ((cmd->cmnd[2] & 0x20) || need_sense)) {
  		ata_gen_ata_desc_sense(qc);
 	} else {
 		if (!need_sense) {
@@ -2015,13 +1995,21 @@ void atapi_request_sense(struct ata_port *ap, struct ata_device *dev,
 	DPRINTK("EXIT\n");
 }
 
-static int atapi_qc_complete(struct ata_queued_cmd *qc, unsigned int err_mask)
+static int atapi_qc_complete(struct ata_queued_cmd *qc, u8 drv_stat)
 {
 	struct scsi_cmnd *cmd = qc->scsicmd;
 
-	VPRINTK("ENTER, err_mask 0x%X\n", err_mask);
+	VPRINTK("ENTER, drv_stat == 0x%x\n", drv_stat);
+
+	if (unlikely(drv_stat & (ATA_BUSY | ATA_DRQ)))
+		/* FIXME: not quite right; we don't want the
+		 * translation of taskfile registers into
+		 * a sense descriptors, since that's only
+		 * correct for ATA, not ATAPI
+		 */
+		ata_gen_ata_desc_sense(qc);
 
-	if (unlikely(err_mask & AC_ERR_DEV)) {
+	else if (unlikely(drv_stat & ATA_ERR)) {
 		DPRINTK("request check condition\n");
 
 		/* FIXME: command completion with check condition
@@ -2038,14 +2026,6 @@ static int atapi_qc_complete(struct ata_queued_cmd *qc, unsigned int err_mask)
 		return 1;
 	}
 
-	else if (unlikely(err_mask))
-		/* FIXME: not quite right; we don't want the
-		 * translation of taskfile registers into
-		 * a sense descriptors, since that's only
-		 * correct for ATA, not ATAPI
-		 */
-		ata_gen_ata_desc_sense(qc);
-
 	else {
 		u8 *scsicmd = cmd->cmnd;
 
diff --git a/trunk/drivers/scsi/libata.h b/trunk/drivers/scsi/libata.h
index 10ecd9e15e4f..65c264b91136 100644
--- a/trunk/drivers/scsi/libata.h
+++ b/trunk/drivers/scsi/libata.h
@@ -39,7 +39,7 @@ struct ata_scsi_args {
 
 /* libata-core.c */
 extern int atapi_enabled;
-extern int ata_qc_complete_noop(struct ata_queued_cmd *qc, unsigned int err_mask);
+extern int ata_qc_complete_noop(struct ata_queued_cmd *qc, u8 drv_stat);
 extern struct ata_queued_cmd *ata_qc_new_init(struct ata_port *ap,
 				      struct ata_device *dev);
 extern void ata_rwcmd_protocol(struct ata_queued_cmd *qc);
diff --git a/trunk/drivers/scsi/pdc_adma.c b/trunk/drivers/scsi/pdc_adma.c
index 988d0d70a277..af99feb9d237 100644
--- a/trunk/drivers/scsi/pdc_adma.c
+++ b/trunk/drivers/scsi/pdc_adma.c
@@ -292,14 +292,14 @@ static void adma_eng_timeout(struct ata_port *ap)
 
 static int adma_fill_sg(struct ata_queued_cmd *qc)
 {
-	struct scatterlist *sg;
+	struct scatterlist *sg = qc->sg;
 	struct ata_port *ap = qc->ap;
 	struct adma_port_priv *pp = ap->private_data;
 	u8  *buf = pp->pkt;
-	int i = (2 + buf[3]) * 8;
+	int nelem, i = (2 + buf[3]) * 8;
 	u8 pFLAGS = pORD | ((qc->tf.flags & ATA_TFLAG_WRITE) ? pDIRO : 0);
 
-	ata_for_each_sg(sg, qc) {
+	for (nelem = 0; nelem < qc->n_elem; nelem++,sg++) {
 		u32 addr;
 		u32 len;
 
@@ -311,7 +311,7 @@ static int adma_fill_sg(struct ata_queued_cmd *qc)
 		*(__le32 *)(buf + i) = cpu_to_le32(len);
 		i += 4;
 
-		if (ata_sg_is_last(sg, qc))
+		if ((nelem + 1) == qc->n_elem)
 			pFLAGS |= pEND;
 		buf[i++] = pFLAGS;
 		buf[i++] = qc->dev->dma_mode & 0xf;
@@ -451,7 +451,7 @@ static inline unsigned int adma_intr_pkt(struct ata_host_set *host_set)
 		struct adma_port_priv *pp;
 		struct ata_queued_cmd *qc;
 		void __iomem *chan = ADMA_REGS(mmio_base, port_no);
-		u8 status = readb(chan + ADMA_STATUS);
+		u8 drv_stat = 0, status = readb(chan + ADMA_STATUS);
 
 		if (status == 0)
 			continue;
@@ -464,14 +464,11 @@ static inline unsigned int adma_intr_pkt(struct ata_host_set *host_set)
 			continue;
 		qc = ata_qc_from_tag(ap, ap->active_tag);
 		if (qc && (!(qc->tf.ctl & ATA_NIEN))) {
-			unsigned int err_mask = 0;
-
 			if ((status & (aPERR | aPSD | aUIRQ)))
-				err_mask = AC_ERR_OTHER;
+				drv_stat = ATA_ERR;
 			else if (pp->pkt[0] != cDONE)
-				err_mask = AC_ERR_OTHER;
-
-			ata_qc_complete(qc, err_mask);
+				drv_stat = ATA_ERR;
+			ata_qc_complete(qc, drv_stat);
 		}
 	}
 	return handled;
@@ -501,7 +498,7 @@ static inline unsigned int adma_intr_mmio(struct ata_host_set *host_set)
 		
 				/* complete taskfile transaction */
 				pp->state = adma_state_idle;
-				ata_qc_complete(qc, ac_err_mask(status));
+				ata_qc_complete(qc, status);
 				handled = 1;
 			}
 		}
diff --git a/trunk/drivers/scsi/sata_mv.c b/trunk/drivers/scsi/sata_mv.c
index a3ab14c79cdd..422e0b6f603a 100644
--- a/trunk/drivers/scsi/sata_mv.c
+++ b/trunk/drivers/scsi/sata_mv.c
@@ -258,6 +258,7 @@ struct mv_host_priv {
 static void mv_irq_clear(struct ata_port *ap);
 static u32 mv_scr_read(struct ata_port *ap, unsigned int sc_reg_in);
 static void mv_scr_write(struct ata_port *ap, unsigned int sc_reg_in, u32 val);
+static u8 mv_check_err(struct ata_port *ap);
 static void mv_phy_reset(struct ata_port *ap);
 static void mv_host_stop(struct ata_host_set *host_set);
 static int mv_port_start(struct ata_port *ap);
@@ -295,6 +296,7 @@ static const struct ata_port_operations mv_ops = {
 	.tf_load		= ata_tf_load,
 	.tf_read		= ata_tf_read,
 	.check_status		= ata_check_status,
+	.check_err		= mv_check_err,
 	.exec_command		= ata_exec_command,
 	.dev_select		= ata_std_dev_select,
 
@@ -783,24 +785,23 @@ static void mv_port_stop(struct ata_port *ap)
 static void mv_fill_sg(struct ata_queued_cmd *qc)
 {
 	struct mv_port_priv *pp = qc->ap->private_data;
-	unsigned int i = 0;
-	struct scatterlist *sg;
+	unsigned int i;
 
-	ata_for_each_sg(sg, qc) {
+	for (i = 0; i < qc->n_elem; i++) {
 		u32 sg_len;
 		dma_addr_t addr;
 
-		addr = sg_dma_address(sg);
-		sg_len = sg_dma_len(sg);
+		addr = sg_dma_address(&qc->sg[i]);
+		sg_len = sg_dma_len(&qc->sg[i]);
 
 		pp->sg_tbl[i].addr = cpu_to_le32(addr & 0xffffffff);
 		pp->sg_tbl[i].addr_hi = cpu_to_le32((addr >> 16) >> 16);
 		assert(0 == (sg_len & ~MV_DMA_BOUNDARY));
 		pp->sg_tbl[i].flags_size = cpu_to_le32(sg_len);
-		if (ata_sg_is_last(sg, qc))
-			pp->sg_tbl[i].flags_size |= cpu_to_le32(EPRD_FLAG_END_OF_TBL);
-
-		i++;
+	}
+	if (0 < qc->n_elem) {
+		pp->sg_tbl[qc->n_elem - 1].flags_size |= 
+			cpu_to_le32(EPRD_FLAG_END_OF_TBL);
 	}
 }
 
@@ -1066,7 +1067,6 @@ static void mv_host_intr(struct ata_host_set *host_set, u32 relevant,
 	struct ata_queued_cmd *qc;
 	u32 hc_irq_cause;
 	int shift, port, port0, hard_port, handled;
-	unsigned int err_mask;
 	u8 ata_status = 0;
 
 	if (hc == 0) {
@@ -1102,15 +1102,15 @@ static void mv_host_intr(struct ata_host_set *host_set, u32 relevant,
 			handled++;
 		}
 
-		err_mask = ac_err_mask(ata_status);
-
 		shift = port << 1;		/* (port * 2) */
 		if (port >= MV_PORTS_PER_HC) {
 			shift++;	/* skip bit 8 in the HC Main IRQ reg */
 		}
 		if ((PORT0_ERR << shift) & relevant) {
 			mv_err_intr(ap);
-			err_mask |= AC_ERR_OTHER;
+			/* OR in ATA_ERR to ensure libata knows we took one */
+			ata_status = readb((void __iomem *)
+					   ap->ioaddr.status_addr) | ATA_ERR;
 			handled++;
 		}
 		
@@ -1120,7 +1120,7 @@ static void mv_host_intr(struct ata_host_set *host_set, u32 relevant,
 				VPRINTK("port %u IRQ found for qc, "
 					"ata_status 0x%x\n", port,ata_status);
 				/* mark qc status appropriately */
-				ata_qc_complete(qc, err_mask);
+				ata_qc_complete(qc, ata_status);
 			}
 		}
 	}
@@ -1184,6 +1184,22 @@ static irqreturn_t mv_interrupt(int irq, void *dev_instance,
 	return IRQ_RETVAL(handled);
 }
 
+/**
+ *      mv_check_err - Return the error shadow register to caller.
+ *      @ap: ATA channel to manipulate
+ *
+ *      Marvell requires DMA to be stopped before accessing shadow
+ *      registers.  So we do that, then return the needed register.
+ *
+ *      LOCKING:
+ *      Inherited from caller.  FIXME: protect mv_stop_dma with lock?
+ */
+static u8 mv_check_err(struct ata_port *ap)
+{
+	mv_stop_dma(ap);		/* can't read shadow regs if DMA on */
+	return readb((void __iomem *) ap->ioaddr.error_addr);
+}
+
 /**
  *      mv_phy_reset - Perform eDMA reset followed by COMRESET
  *      @ap: ATA channel to manipulate
@@ -1296,7 +1312,7 @@ static void mv_eng_timeout(struct ata_port *ap)
 	 	 */
 		spin_lock_irqsave(&ap->host_set->lock, flags);
 		qc->scsidone = scsi_finish_command;
-		ata_qc_complete(qc, AC_ERR_OTHER);
+		ata_qc_complete(qc, ATA_ERR);
 		spin_unlock_irqrestore(&ap->host_set->lock, flags);
 	}
 }
diff --git a/trunk/drivers/scsi/sata_promise.c b/trunk/drivers/scsi/sata_promise.c
index 8f41702275db..63911f16b6ec 100644
--- a/trunk/drivers/scsi/sata_promise.c
+++ b/trunk/drivers/scsi/sata_promise.c
@@ -399,8 +399,7 @@ static void pdc_eng_timeout(struct ata_port *ap)
 	case ATA_PROT_DMA:
 	case ATA_PROT_NODATA:
 		printk(KERN_ERR "ata%u: command timeout\n", ap->id);
-		drv_stat = ata_wait_idle(ap);
-		ata_qc_complete(qc, __ac_err_mask(drv_stat));
+		ata_qc_complete(qc, ata_wait_idle(ap) | ATA_ERR);
 		break;
 
 	default:
@@ -409,7 +408,7 @@ static void pdc_eng_timeout(struct ata_port *ap)
 		printk(KERN_ERR "ata%u: unknown timeout, cmd 0x%x stat 0x%x\n",
 		       ap->id, qc->tf.command, drv_stat);
 
-		ata_qc_complete(qc, ac_err_mask(drv_stat));
+		ata_qc_complete(qc, drv_stat);
 		break;
 	}
 
@@ -421,21 +420,24 @@ static void pdc_eng_timeout(struct ata_port *ap)
 static inline unsigned int pdc_host_intr( struct ata_port *ap,
                                           struct ata_queued_cmd *qc)
 {
-	unsigned int handled = 0, err_mask = 0;
+	u8 status;
+	unsigned int handled = 0, have_err = 0;
 	u32 tmp;
 	void __iomem *mmio = (void __iomem *) ap->ioaddr.cmd_addr + PDC_GLOBAL_CTL;
 
 	tmp = readl(mmio);
 	if (tmp & PDC_ERR_MASK) {
-		err_mask = AC_ERR_DEV;
+		have_err = 1;
 		pdc_reset_port(ap);
 	}
 
 	switch (qc->tf.protocol) {
 	case ATA_PROT_DMA:
 	case ATA_PROT_NODATA:
-		err_mask |= ac_err_mask(ata_wait_idle(ap));
-		ata_qc_complete(qc, err_mask);
+		status = ata_wait_idle(ap);
+		if (have_err)
+			status |= ATA_ERR;
+		ata_qc_complete(qc, status);
 		handled = 1;
 		break;
 
diff --git a/trunk/drivers/scsi/sata_qstor.c b/trunk/drivers/scsi/sata_qstor.c
index 8c20229b01b4..1aaf3304d397 100644
--- a/trunk/drivers/scsi/sata_qstor.c
+++ b/trunk/drivers/scsi/sata_qstor.c
@@ -270,17 +270,16 @@ static void qs_scr_write (struct ata_port *ap, unsigned int sc_reg, u32 val)
 
 static void qs_fill_sg(struct ata_queued_cmd *qc)
 {
-	struct scatterlist *sg;
+	struct scatterlist *sg = qc->sg;
 	struct ata_port *ap = qc->ap;
 	struct qs_port_priv *pp = ap->private_data;
 	unsigned int nelem;
 	u8 *prd = pp->pkt + QS_CPB_BYTES;
 
-	assert(qc->__sg != NULL);
+	assert(sg != NULL);
 	assert(qc->n_elem > 0);
 
-	nelem = 0;
-	ata_for_each_sg(sg, qc) {
+	for (nelem = 0; nelem < qc->n_elem; nelem++,sg++) {
 		u64 addr;
 		u32 len;
 
@@ -294,7 +293,6 @@ static void qs_fill_sg(struct ata_queued_cmd *qc)
 
 		VPRINTK("PRD[%u] = (0x%llX, 0x%X)\n", nelem,
 					(unsigned long long)addr, len);
-		nelem++;
 	}
 }
 
@@ -402,12 +400,11 @@ static inline unsigned int qs_intr_pkt(struct ata_host_set *host_set)
 				qc = ata_qc_from_tag(ap, ap->active_tag);
 				if (qc && (!(qc->tf.ctl & ATA_NIEN))) {
 					switch (sHST) {
-					case 0: /* successful CPB */
+					case 0: /* sucessful CPB */
 					case 3: /* device error */
 						pp->state = qs_state_idle;
 						qs_enter_reg_mode(qc->ap);
-						ata_qc_complete(qc,
-							ac_err_mask(sDST));
+						ata_qc_complete(qc, sDST);
 						break;
 					default:
 						break;
@@ -444,7 +441,7 @@ static inline unsigned int qs_intr_mmio(struct ata_host_set *host_set)
 
 				/* complete taskfile transaction */
 				pp->state = qs_state_idle;
-				ata_qc_complete(qc, ac_err_mask(status));
+				ata_qc_complete(qc, status);
 				handled = 1;
 			}
 		}
diff --git a/trunk/drivers/scsi/sata_sil24.c b/trunk/drivers/scsi/sata_sil24.c
index 05ce84286aea..51855d3bac64 100644
--- a/trunk/drivers/scsi/sata_sil24.c
+++ b/trunk/drivers/scsi/sata_sil24.c
@@ -225,6 +225,7 @@ struct sil24_host_priv {
 };
 
 static u8 sil24_check_status(struct ata_port *ap);
+static u8 sil24_check_err(struct ata_port *ap);
 static u32 sil24_scr_read(struct ata_port *ap, unsigned sc_reg);
 static void sil24_scr_write(struct ata_port *ap, unsigned sc_reg, u32 val);
 static void sil24_tf_read(struct ata_port *ap, struct ata_taskfile *tf);
@@ -279,6 +280,7 @@ static const struct ata_port_operations sil24_ops = {
 
 	.check_status		= sil24_check_status,
 	.check_altstatus	= sil24_check_status,
+	.check_err		= sil24_check_err,
 	.dev_select		= ata_noop_dev_select,
 
 	.tf_read		= sil24_tf_read,
@@ -361,6 +363,12 @@ static u8 sil24_check_status(struct ata_port *ap)
 	return pp->tf.command;
 }
 
+static u8 sil24_check_err(struct ata_port *ap)
+{
+	struct sil24_port_priv *pp = ap->private_data;
+	return pp->tf.feature;
+}
+
 static int sil24_scr_map[] = {
 	[SCR_CONTROL]	= 0,
 	[SCR_STATUS]	= 1,
@@ -410,20 +418,15 @@ static void sil24_phy_reset(struct ata_port *ap)
 static inline void sil24_fill_sg(struct ata_queued_cmd *qc,
 				 struct sil24_cmd_block *cb)
 {
+	struct scatterlist *sg = qc->sg;
 	struct sil24_sge *sge = cb->sge;
-	struct scatterlist *sg;
-	unsigned int idx = 0;
+	unsigned i;
 
-	ata_for_each_sg(sg, qc) {
+	for (i = 0; i < qc->n_elem; i++, sg++, sge++) {
 		sge->addr = cpu_to_le64(sg_dma_address(sg));
 		sge->cnt = cpu_to_le32(sg_dma_len(sg));
-		if (ata_sg_is_last(sg, qc))
-			sge->flags = cpu_to_le32(SGE_TRM);
-		else
-			sge->flags = 0;
-
-		sge++;
-		idx++;
+		sge->flags = 0;
+		sge->flags = i < qc->n_elem - 1 ? 0 : cpu_to_le32(SGE_TRM);
 	}
 }
 
@@ -503,7 +506,7 @@ static void sil24_eng_timeout(struct ata_port *ap)
 
 	qc = ata_qc_from_tag(ap, ap->active_tag);
 	if (!qc) {
-		printk(KERN_ERR "ata%u: BUG: timeout without command\n",
+		printk(KERN_ERR "ata%u: BUG: tiemout without command\n",
 		       ap->id);
 		return;
 	}
@@ -517,7 +520,7 @@ static void sil24_eng_timeout(struct ata_port *ap)
 	 */
 	printk(KERN_ERR "ata%u: command timeout\n", ap->id);
 	qc->scsidone = scsi_finish_command;
-	ata_qc_complete(qc, AC_ERR_OTHER);
+	ata_qc_complete(qc, ATA_ERR);
 
 	sil24_reset_controller(ap);
 }
@@ -528,7 +531,6 @@ static void sil24_error_intr(struct ata_port *ap, u32 slot_stat)
 	struct sil24_port_priv *pp = ap->private_data;
 	void __iomem *port = (void __iomem *)ap->ioaddr.cmd_addr;
 	u32 irq_stat, cmd_err, sstatus, serror;
-	unsigned int err_mask;
 
 	irq_stat = readl(port + PORT_IRQ_STAT);
 	writel(irq_stat, port + PORT_IRQ_STAT);		/* clear irq */
@@ -556,18 +558,17 @@ static void sil24_error_intr(struct ata_port *ap, u32 slot_stat)
 		 * Device is reporting error, tf registers are valid.
 		 */
 		sil24_update_tf(ap);
-		err_mask = ac_err_mask(pp->tf.command);
 	} else {
 		/*
 		 * Other errors.  libata currently doesn't have any
 		 * mechanism to report these errors.  Just turn on
 		 * ATA_ERR.
 		 */
-		err_mask = AC_ERR_OTHER;
+		pp->tf.command = ATA_ERR;
 	}
 
 	if (qc)
-		ata_qc_complete(qc, err_mask);
+		ata_qc_complete(qc, pp->tf.command);
 
 	sil24_reset_controller(ap);
 }
@@ -592,7 +593,7 @@ static inline void sil24_host_intr(struct ata_port *ap)
 		sil24_update_tf(ap);
 
 		if (qc)
-			ata_qc_complete(qc, ac_err_mask(pp->tf.command));
+			ata_qc_complete(qc, pp->tf.command);
 	} else
 		sil24_error_intr(ap, slot_stat);
 }
diff --git a/trunk/drivers/scsi/sata_sx4.c b/trunk/drivers/scsi/sata_sx4.c
index 8e2e7b8308c0..af08f4f650c1 100644
--- a/trunk/drivers/scsi/sata_sx4.c
+++ b/trunk/drivers/scsi/sata_sx4.c
@@ -449,14 +449,14 @@ static inline void pdc20621_host_pkt(struct ata_taskfile *tf, u8 *buf,
 
 static void pdc20621_dma_prep(struct ata_queued_cmd *qc)
 {
-	struct scatterlist *sg;
+	struct scatterlist *sg = qc->sg;
 	struct ata_port *ap = qc->ap;
 	struct pdc_port_priv *pp = ap->private_data;
 	void __iomem *mmio = ap->host_set->mmio_base;
 	struct pdc_host_priv *hpriv = ap->host_set->private_data;
 	void __iomem *dimm_mmio = hpriv->dimm_mmio;
 	unsigned int portno = ap->port_no;
-	unsigned int i, idx, total_len = 0, sgt_len;
+	unsigned int i, last, idx, total_len = 0, sgt_len;
 	u32 *buf = (u32 *) &pp->dimm_buf[PDC_DIMM_HEADER_SZ];
 
 	assert(qc->flags & ATA_QCFLAG_DMAMAP);
@@ -469,11 +469,12 @@ static void pdc20621_dma_prep(struct ata_queued_cmd *qc)
 	/*
 	 * Build S/G table
 	 */
+	last = qc->n_elem;
 	idx = 0;
-	ata_for_each_sg(sg, qc) {
-		buf[idx++] = cpu_to_le32(sg_dma_address(sg));
-		buf[idx++] = cpu_to_le32(sg_dma_len(sg));
-		total_len += sg_dma_len(sg);
+	for (i = 0; i < last; i++) {
+		buf[idx++] = cpu_to_le32(sg_dma_address(&sg[i]));
+		buf[idx++] = cpu_to_le32(sg_dma_len(&sg[i]));
+		total_len += sg_dma_len(&sg[i]);
 	}
 	buf[idx - 1] |= cpu_to_le32(ATA_PRD_EOT);
 	sgt_len = idx * 4;
@@ -717,7 +718,7 @@ static inline unsigned int pdc20621_host_intr( struct ata_port *ap,
 			VPRINTK("ata%u: read hdma, 0x%x 0x%x\n", ap->id,
 				readl(mmio + 0x104), readl(mmio + PDC_HDMA_CTLSTAT));
 			/* get drive status; clear intr; complete txn */
-			ata_qc_complete(qc, ac_err_mask(ata_wait_idle(ap)));
+			ata_qc_complete(qc, ata_wait_idle(ap));
 			pdc20621_pop_hdma(qc);
 		}
 
@@ -755,7 +756,7 @@ static inline unsigned int pdc20621_host_intr( struct ata_port *ap,
 			VPRINTK("ata%u: write ata, 0x%x 0x%x\n", ap->id,
 				readl(mmio + 0x104), readl(mmio + PDC_HDMA_CTLSTAT));
 			/* get drive status; clear intr; complete txn */
-			ata_qc_complete(qc, ac_err_mask(ata_wait_idle(ap)));
+			ata_qc_complete(qc, ata_wait_idle(ap));
 			pdc20621_pop_hdma(qc);
 		}
 		handled = 1;
@@ -765,7 +766,7 @@ static inline unsigned int pdc20621_host_intr( struct ata_port *ap,
 
 		status = ata_busy_wait(ap, ATA_BUSY | ATA_DRQ, 1000);
 		DPRINTK("BUS_NODATA (drv_stat 0x%X)\n", status);
-		ata_qc_complete(qc, ac_err_mask(status));
+		ata_qc_complete(qc, status);
 		handled = 1;
 
 	} else {
@@ -880,7 +881,7 @@ static void pdc_eng_timeout(struct ata_port *ap)
 	case ATA_PROT_DMA:
 	case ATA_PROT_NODATA:
 		printk(KERN_ERR "ata%u: command timeout\n", ap->id);
-		ata_qc_complete(qc, __ac_err_mask(ata_wait_idle(ap)));
+		ata_qc_complete(qc, ata_wait_idle(ap) | ATA_ERR);
 		break;
 
 	default:
@@ -889,7 +890,7 @@ static void pdc_eng_timeout(struct ata_port *ap)
 		printk(KERN_ERR "ata%u: unknown timeout, cmd 0x%x stat 0x%x\n",
 		       ap->id, qc->tf.command, drv_stat);
 
-		ata_qc_complete(qc, ac_err_mask(drv_stat));
+		ata_qc_complete(qc, drv_stat);
 		break;
 	}
 
diff --git a/trunk/include/linux/libata.h b/trunk/include/linux/libata.h
index 6225b78fa65b..00a8a5738858 100644
--- a/trunk/include/linux/libata.h
+++ b/trunk/include/linux/libata.h
@@ -155,10 +155,6 @@ enum {
 	ATA_SHIFT_UDMA		= 0,
 	ATA_SHIFT_MWDMA		= 8,
 	ATA_SHIFT_PIO		= 11,
-
-	/* size of buffer to pad xfers ending on unaligned boundaries */
-	ATA_DMA_PAD_SZ		= 4,
-	ATA_DMA_PAD_BUF_SZ	= ATA_DMA_PAD_SZ * ATA_MAX_QUEUE,
 	
 	/* Masks for port functions */
 	ATA_PORT_PRIMARY	= (1 << 0),
@@ -176,13 +172,6 @@ enum hsm_task_states {
 	HSM_ST_ERR,
 };
 
-enum ata_completion_errors {
-	AC_ERR_OTHER		= (1 << 0),
-	AC_ERR_DEV		= (1 << 1),
-	AC_ERR_ATA_BUS		= (1 << 2),
-	AC_ERR_HOST_BUS		= (1 << 3),
-};
-
 /* forward declarations */
 struct scsi_device;
 struct ata_port_operations;
@@ -190,7 +179,7 @@ struct ata_port;
 struct ata_queued_cmd;
 
 /* typedefs */
-typedef int (*ata_qc_cb_t) (struct ata_queued_cmd *qc, unsigned int err_mask);
+typedef int (*ata_qc_cb_t) (struct ata_queued_cmd *qc, u8 drv_stat);
 
 struct ata_ioports {
 	unsigned long		cmd_addr;
@@ -253,12 +242,9 @@ struct ata_queued_cmd {
 	unsigned long		flags;		/* ATA_QCFLAG_xxx */
 	unsigned int		tag;
 	unsigned int		n_elem;
-	unsigned int		orig_n_elem;
 
 	int			dma_dir;
 
-	unsigned int		pad_len;
-
 	unsigned int		nsect;
 	unsigned int		cursect;
 
@@ -269,11 +255,9 @@ struct ata_queued_cmd {
 	unsigned int		cursg_ofs;
 
 	struct scatterlist	sgent;
-	struct scatterlist	pad_sgent;
 	void			*buf_virt;
 
-	/* DO NOT iterate over __sg manually, use ata_for_each_sg() */
-	struct scatterlist	*__sg;
+	struct scatterlist	*sg;
 
 	ata_qc_cb_t		complete_fn;
 
@@ -319,9 +303,6 @@ struct ata_port {
 	struct ata_prd		*prd;	 /* our SG list */
 	dma_addr_t		prd_dma; /* and its DMA mapping */
 
-	void			*pad;	/* array of DMA pad buffers */
-	dma_addr_t		pad_dma;
-
 	struct ata_ioports	ioaddr;	/* ATA cmd/ctl/dma register blocks */
 
 	u8			ctl;	/* cache of ATA control register */
@@ -366,6 +347,7 @@ struct ata_port_operations {
 	void (*exec_command)(struct ata_port *ap, const struct ata_taskfile *tf);
 	u8   (*check_status)(struct ata_port *ap);
 	u8   (*check_altstatus)(struct ata_port *ap);
+	u8   (*check_err)(struct ata_port *ap);
 	void (*dev_select)(struct ata_port *ap, unsigned int device);
 
 	void (*phy_reset) (struct ata_port *ap);
@@ -452,6 +434,7 @@ extern void ata_noop_dev_select (struct ata_port *ap, unsigned int device);
 extern void ata_std_dev_select (struct ata_port *ap, unsigned int device);
 extern u8 ata_check_status(struct ata_port *ap);
 extern u8 ata_altstatus(struct ata_port *ap);
+extern u8 ata_chk_err(struct ata_port *ap);
 extern void ata_exec_command(struct ata_port *ap, const struct ata_taskfile *tf);
 extern int ata_port_start (struct ata_port *ap);
 extern void ata_port_stop (struct ata_port *ap);
@@ -472,7 +455,7 @@ extern void ata_bmdma_start (struct ata_queued_cmd *qc);
 extern void ata_bmdma_stop(struct ata_queued_cmd *qc);
 extern u8   ata_bmdma_status(struct ata_port *ap);
 extern void ata_bmdma_irq_clear(struct ata_port *ap);
-extern void ata_qc_complete(struct ata_queued_cmd *qc, unsigned int err_mask);
+extern void ata_qc_complete(struct ata_queued_cmd *qc, u8 drv_stat);
 extern void ata_eng_timeout(struct ata_port *ap);
 extern void ata_scsi_simulate(u16 *id, struct scsi_cmnd *cmd,
 			      void (*done)(struct scsi_cmnd *));
@@ -524,31 +507,6 @@ extern int pci_test_config_bits(struct pci_dev *pdev, const struct pci_bits *bit
 #endif /* CONFIG_PCI */
 
 
-static inline int
-ata_sg_is_last(struct scatterlist *sg, struct ata_queued_cmd *qc)
-{
-	if (sg == &qc->pad_sgent)
-		return 1;
-	if (qc->pad_len)
-		return 0;
-	if (((sg - qc->__sg) + 1) == qc->n_elem)
-		return 1;
-	return 0;
-}
-
-static inline struct scatterlist *
-ata_qc_next_sg(struct scatterlist *sg, struct ata_queued_cmd *qc)
-{
-	if (sg == &qc->pad_sgent)
-		return NULL;
-	if (++sg - qc->__sg < qc->n_elem)
-		return sg;
-	return qc->pad_len ? &qc->pad_sgent : NULL;
-}
-
-#define ata_for_each_sg(sg, qc) \
-	for (sg = qc->__sg; sg; sg = ata_qc_next_sg(sg, qc))
-
 static inline unsigned int ata_tag_valid(unsigned int tag)
 {
 	return (tag < ATA_MAX_QUEUE) ? 1 : 0;
@@ -760,21 +718,4 @@ static inline int ata_try_flush_cache(const struct ata_device *dev)
 	       ata_id_has_flush_ext(dev->id);
 }
 
-static inline unsigned int ac_err_mask(u8 status)
-{
-	if (status & ATA_BUSY)
-		return AC_ERR_ATA_BUS;
-	if (status & (ATA_ERR | ATA_DF))
-		return AC_ERR_DEV;
-	return 0;
-}
-
-static inline unsigned int __ac_err_mask(u8 status)
-{
-	unsigned int mask = ac_err_mask(status);
-	if (mask == 0)
-		return AC_ERR_OTHER;
-	return mask;
-}
-
 #endif /* __LINUX_LIBATA_H__ */