1 files changed, 771 insertions, 353 deletions

diff --git a/drivers/dma/ioat/dma_v3.c b/drivers/dma/ioat/dma_v3.c
index d0f49909847..b9b38a1cf92 100644
--- a/drivers/dma/ioat/dma_v3.c
+++ b/drivers/dma/ioat/dma_v3.c
@@ -55,34 +55,41 @@
 /*
  * Support routines for v3+ hardware
  */
-
+#include <linux/module.h>
 #include <linux/pci.h>
 #include <linux/gfp.h>
 #include <linux/dmaengine.h>
 #include <linux/dma-mapping.h>
+#include <linux/prefetch.h>
+#include "../dmaengine.h"
 #include "registers.h"
 #include "hw.h"
 #include "dma.h"
 #include "dma_v2.h"
 
+extern struct kmem_cache *ioat3_sed_cache;
+
 /* ioat hardware assumes at least two sources for raid operations */
 #define src_cnt_to_sw(x) ((x) + 2)
 #define src_cnt_to_hw(x) ((x) - 2)
+#define ndest_to_sw(x) ((x) + 1)
+#define ndest_to_hw(x) ((x) - 1)
+#define src16_cnt_to_sw(x) ((x) + 9)
+#define src16_cnt_to_hw(x) ((x) - 9)
 
 /* provide a lookup table for setting the source address in the base or
  * extended descriptor of an xor or pq descriptor
  */
-static const u8 xor_idx_to_desc __read_mostly = 0xd0;
-static const u8 xor_idx_to_field[] __read_mostly = { 1, 4, 5, 6, 7, 0, 1, 2 };
-static const u8 pq_idx_to_desc __read_mostly = 0xf8;
-static const u8 pq_idx_to_field[] __read_mostly = { 1, 4, 5, 0, 1, 2, 4, 5 };
+static const u8 xor_idx_to_desc = 0xe0;
+static const u8 xor_idx_to_field[] = { 1, 4, 5, 6, 7, 0, 1, 2 };
+static const u8 pq_idx_to_desc = 0xf8;
+static const u8 pq16_idx_to_desc[] = { 0, 0, 1, 1, 1, 1, 1, 1, 1,
+				       2, 2, 2, 2, 2, 2, 2 };
+static const u8 pq_idx_to_field[] = { 1, 4, 5, 0, 1, 2, 4, 5 };
+static const u8 pq16_idx_to_field[] = { 1, 4, 1, 2, 3, 4, 5, 6, 7,
+					0, 1, 2, 3, 4, 5, 6 };
 
-static dma_addr_t xor_get_src(struct ioat_raw_descriptor *descs[2], int idx)
-{
-	struct ioat_raw_descriptor *raw = descs[xor_idx_to_desc >> idx & 1];
-
-	return raw->field[xor_idx_to_field[idx]];
-}
+static void ioat3_eh(struct ioat2_dma_chan *ioat);
 
 static void xor_set_src(struct ioat_raw_descriptor *descs[2],
 			dma_addr_t addr, u32 offset, int idx)
@@ -99,6 +106,13 @@ static dma_addr_t pq_get_src(struct ioat_raw_descriptor *descs[2], int idx)
 	return raw->field[pq_idx_to_field[idx]];
 }
 
+static dma_addr_t pq16_get_src(struct ioat_raw_descriptor *desc[3], int idx)
+{
+	struct ioat_raw_descriptor *raw = desc[pq16_idx_to_desc[idx]];
+
+	return raw->field[pq16_idx_to_field[idx]];
+}
+
 static void pq_set_src(struct ioat_raw_descriptor *descs[2],
 		       dma_addr_t addr, u32 offset, u8 coef, int idx)
 {
@@ -109,122 +123,159 @@ static void pq_set_src(struct ioat_raw_descriptor *descs[2],
 	pq->coef[idx] = coef;
 }
 
-static void ioat3_dma_unmap(struct ioat2_dma_chan *ioat,
-			    struct ioat_ring_ent *desc, int idx)
+static bool is_jf_ioat(struct pci_dev *pdev)
 {
-	struct ioat_chan_common *chan = &ioat->base;
-	struct pci_dev *pdev = chan->device->pdev;
-	size_t len = desc->len;
-	size_t offset = len - desc->hw->size;
-	struct dma_async_tx_descriptor *tx = &desc->txd;
-	enum dma_ctrl_flags flags = tx->flags;
-
-	switch (desc->hw->ctl_f.op) {
-	case IOAT_OP_COPY:
-		if (!desc->hw->ctl_f.null) /* skip 'interrupt' ops */
-			ioat_dma_unmap(chan, flags, len, desc->hw);
-		break;
-	case IOAT_OP_FILL: {
-		struct ioat_fill_descriptor *hw = desc->fill;
+	switch (pdev->device) {
+	case PCI_DEVICE_ID_INTEL_IOAT_JSF0:
+	case PCI_DEVICE_ID_INTEL_IOAT_JSF1:
+	case PCI_DEVICE_ID_INTEL_IOAT_JSF2:
+	case PCI_DEVICE_ID_INTEL_IOAT_JSF3:
+	case PCI_DEVICE_ID_INTEL_IOAT_JSF4:
+	case PCI_DEVICE_ID_INTEL_IOAT_JSF5:
+	case PCI_DEVICE_ID_INTEL_IOAT_JSF6:
+	case PCI_DEVICE_ID_INTEL_IOAT_JSF7:
+	case PCI_DEVICE_ID_INTEL_IOAT_JSF8:
+	case PCI_DEVICE_ID_INTEL_IOAT_JSF9:
+		return true;
+	default:
+		return false;
+	}
+}
 
-		if (!(flags & DMA_COMPL_SKIP_DEST_UNMAP))
-			ioat_unmap(pdev, hw->dst_addr - offset, len,
-				   PCI_DMA_FROMDEVICE, flags, 1);
-		break;
+static bool is_snb_ioat(struct pci_dev *pdev)
+{
+	switch (pdev->device) {
+	case PCI_DEVICE_ID_INTEL_IOAT_SNB0:
+	case PCI_DEVICE_ID_INTEL_IOAT_SNB1:
+	case PCI_DEVICE_ID_INTEL_IOAT_SNB2:
+	case PCI_DEVICE_ID_INTEL_IOAT_SNB3:
+	case PCI_DEVICE_ID_INTEL_IOAT_SNB4:
+	case PCI_DEVICE_ID_INTEL_IOAT_SNB5:
+	case PCI_DEVICE_ID_INTEL_IOAT_SNB6:
+	case PCI_DEVICE_ID_INTEL_IOAT_SNB7:
+	case PCI_DEVICE_ID_INTEL_IOAT_SNB8:
+	case PCI_DEVICE_ID_INTEL_IOAT_SNB9:
+		return true;
+	default:
+		return false;
 	}
-	case IOAT_OP_XOR_VAL:
-	case IOAT_OP_XOR: {
-		struct ioat_xor_descriptor *xor = desc->xor;
-		struct ioat_ring_ent *ext;
-		struct ioat_xor_ext_descriptor *xor_ex = NULL;
-		int src_cnt = src_cnt_to_sw(xor->ctl_f.src_cnt);
-		struct ioat_raw_descriptor *descs[2];
-		int i;
+}
 
-		if (src_cnt > 5) {
-			ext = ioat2_get_ring_ent(ioat, idx + 1);
-			xor_ex = ext->xor_ex;
-		}
+static bool is_ivb_ioat(struct pci_dev *pdev)
+{
+	switch (pdev->device) {
+	case PCI_DEVICE_ID_INTEL_IOAT_IVB0:
+	case PCI_DEVICE_ID_INTEL_IOAT_IVB1:
+	case PCI_DEVICE_ID_INTEL_IOAT_IVB2:
+	case PCI_DEVICE_ID_INTEL_IOAT_IVB3:
+	case PCI_DEVICE_ID_INTEL_IOAT_IVB4:
+	case PCI_DEVICE_ID_INTEL_IOAT_IVB5:
+	case PCI_DEVICE_ID_INTEL_IOAT_IVB6:
+	case PCI_DEVICE_ID_INTEL_IOAT_IVB7:
+	case PCI_DEVICE_ID_INTEL_IOAT_IVB8:
+	case PCI_DEVICE_ID_INTEL_IOAT_IVB9:
+		return true;
+	default:
+		return false;
+	}
 
-		if (!(flags & DMA_COMPL_SKIP_SRC_UNMAP)) {
-			descs[0] = (struct ioat_raw_descriptor *) xor;
-			descs[1] = (struct ioat_raw_descriptor *) xor_ex;
-			for (i = 0; i < src_cnt; i++) {
-				dma_addr_t src = xor_get_src(descs, i);
+}
 
-				ioat_unmap(pdev, src - offset, len,
-					   PCI_DMA_TODEVICE, flags, 0);
-			}
+static bool is_hsw_ioat(struct pci_dev *pdev)
+{
+	switch (pdev->device) {
+	case PCI_DEVICE_ID_INTEL_IOAT_HSW0:
+	case PCI_DEVICE_ID_INTEL_IOAT_HSW1:
+	case PCI_DEVICE_ID_INTEL_IOAT_HSW2:
+	case PCI_DEVICE_ID_INTEL_IOAT_HSW3:
+	case PCI_DEVICE_ID_INTEL_IOAT_HSW4:
+	case PCI_DEVICE_ID_INTEL_IOAT_HSW5:
+	case PCI_DEVICE_ID_INTEL_IOAT_HSW6:
+	case PCI_DEVICE_ID_INTEL_IOAT_HSW7:
+	case PCI_DEVICE_ID_INTEL_IOAT_HSW8:
+	case PCI_DEVICE_ID_INTEL_IOAT_HSW9:
+		return true;
+	default:
+		return false;
+	}
 
-			/* dest is a source in xor validate operations */
-			if (xor->ctl_f.op == IOAT_OP_XOR_VAL) {
-				ioat_unmap(pdev, xor->dst_addr - offset, len,
-					   PCI_DMA_TODEVICE, flags, 1);
-				break;
-			}
-		}
+}
 
-		if (!(flags & DMA_COMPL_SKIP_DEST_UNMAP))
-			ioat_unmap(pdev, xor->dst_addr - offset, len,
-				   PCI_DMA_FROMDEVICE, flags, 1);
-		break;
+static bool is_xeon_cb32(struct pci_dev *pdev)
+{
+	return is_jf_ioat(pdev) || is_snb_ioat(pdev) || is_ivb_ioat(pdev) ||
+		is_hsw_ioat(pdev);
+}
+
+static bool is_bwd_ioat(struct pci_dev *pdev)
+{
+	switch (pdev->device) {
+	case PCI_DEVICE_ID_INTEL_IOAT_BWD0:
+	case PCI_DEVICE_ID_INTEL_IOAT_BWD1:
+	case PCI_DEVICE_ID_INTEL_IOAT_BWD2:
+	case PCI_DEVICE_ID_INTEL_IOAT_BWD3:
+		return true;
+	default:
+		return false;
 	}
-	case IOAT_OP_PQ_VAL:
-	case IOAT_OP_PQ: {
-		struct ioat_pq_descriptor *pq = desc->pq;
-		struct ioat_ring_ent *ext;
-		struct ioat_pq_ext_descriptor *pq_ex = NULL;
-		int src_cnt = src_cnt_to_sw(pq->ctl_f.src_cnt);
-		struct ioat_raw_descriptor *descs[2];
-		int i;
+}
 
-		if (src_cnt > 3) {
-			ext = ioat2_get_ring_ent(ioat, idx + 1);
-			pq_ex = ext->pq_ex;
-		}
+static bool is_bwd_noraid(struct pci_dev *pdev)
+{
+	switch (pdev->device) {
+	case PCI_DEVICE_ID_INTEL_IOAT_BWD2:
+	case PCI_DEVICE_ID_INTEL_IOAT_BWD3:
+		return true;
+	default:
+		return false;
+	}
 
-		/* in the 'continue' case don't unmap the dests as sources */
-		if (dmaf_p_disabled_continue(flags))
-			src_cnt--;
-		else if (dmaf_continue(flags))
-			src_cnt -= 3;
-
-		if (!(flags & DMA_COMPL_SKIP_SRC_UNMAP)) {
-			descs[0] = (struct ioat_raw_descriptor *) pq;
-			descs[1] = (struct ioat_raw_descriptor *) pq_ex;
-			for (i = 0; i < src_cnt; i++) {
-				dma_addr_t src = pq_get_src(descs, i);
-
-				ioat_unmap(pdev, src - offset, len,
-					   PCI_DMA_TODEVICE, flags, 0);
-			}
+}
 
-			/* the dests are sources in pq validate operations */
-			if (pq->ctl_f.op == IOAT_OP_XOR_VAL) {
-				if (!(flags & DMA_PREP_PQ_DISABLE_P))
-					ioat_unmap(pdev, pq->p_addr - offset,
-						   len, PCI_DMA_TODEVICE, flags, 0);
-				if (!(flags & DMA_PREP_PQ_DISABLE_Q))
-					ioat_unmap(pdev, pq->q_addr - offset,
-						   len, PCI_DMA_TODEVICE, flags, 0);
-				break;
-			}
-		}
+static void pq16_set_src(struct ioat_raw_descriptor *desc[3],
+			dma_addr_t addr, u32 offset, u8 coef, unsigned idx)
+{
+	struct ioat_pq_descriptor *pq = (struct ioat_pq_descriptor *)desc[0];
+	struct ioat_pq16a_descriptor *pq16 =
+		(struct ioat_pq16a_descriptor *)desc[1];
+	struct ioat_raw_descriptor *raw = desc[pq16_idx_to_desc[idx]];
 
-		if (!(flags & DMA_COMPL_SKIP_DEST_UNMAP)) {
-			if (!(flags & DMA_PREP_PQ_DISABLE_P))
-				ioat_unmap(pdev, pq->p_addr - offset, len,
-					   PCI_DMA_BIDIRECTIONAL, flags, 1);
-			if (!(flags & DMA_PREP_PQ_DISABLE_Q))
-				ioat_unmap(pdev, pq->q_addr - offset, len,
-					   PCI_DMA_BIDIRECTIONAL, flags, 1);
-		}
-		break;
-	}
-	default:
-		dev_err(&pdev->dev, "%s: unknown op type: %#x\n",
-			__func__, desc->hw->ctl_f.op);
+	raw->field[pq16_idx_to_field[idx]] = addr + offset;
+
+	if (idx < 8)
+		pq->coef[idx] = coef;
+	else
+		pq16->coef[idx - 8] = coef;
+}
+
+static struct ioat_sed_ent *
+ioat3_alloc_sed(struct ioatdma_device *device, unsigned int hw_pool)
+{
+	struct ioat_sed_ent *sed;
+	gfp_t flags = __GFP_ZERO | GFP_ATOMIC;
+
+	sed = kmem_cache_alloc(ioat3_sed_cache, flags);
+	if (!sed)
+		return NULL;
+
+	sed->hw_pool = hw_pool;
+	sed->hw = dma_pool_alloc(device->sed_hw_pool[hw_pool],
+				 flags, &sed->dma);
+	if (!sed->hw) {
+		kmem_cache_free(ioat3_sed_cache, sed);
+		return NULL;
 	}
+
+	return sed;
+}
+
+static void ioat3_free_sed(struct ioatdma_device *device, struct ioat_sed_ent *sed)
+{
+	if (!sed)
+		return;
+
+	dma_pool_free(device->sed_hw_pool[sed->hw_pool], sed->hw, sed->dma);
+	kmem_cache_free(ioat3_sed_cache, sed);
 }
 
 static bool desc_has_ext(struct ioat_ring_ent *desc)
@@ -248,6 +299,63 @@ static bool desc_has_ext(struct ioat_ring_ent *desc)
 	return false;
 }
 
+static u64 ioat3_get_current_completion(struct ioat_chan_common *chan)
+{
+	u64 phys_complete;
+	u64 completion;
+
+	completion = *chan->completion;
+	phys_complete = ioat_chansts_to_addr(completion);
+
+	dev_dbg(to_dev(chan), "%s: phys_complete: %#llx\n", __func__,
+		(unsigned long long) phys_complete);
+
+	return phys_complete;
+}
+
+static bool ioat3_cleanup_preamble(struct ioat_chan_common *chan,
+				   u64 *phys_complete)
+{
+	*phys_complete = ioat3_get_current_completion(chan);
+	if (*phys_complete == chan->last_completion)
+		return false;
+
+	clear_bit(IOAT_COMPLETION_ACK, &chan->state);
+	mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
+
+	return true;
+}
+
+static void
+desc_get_errstat(struct ioat2_dma_chan *ioat, struct ioat_ring_ent *desc)
+{
+	struct ioat_dma_descriptor *hw = desc->hw;
+
+	switch (hw->ctl_f.op) {
+	case IOAT_OP_PQ_VAL:
+	case IOAT_OP_PQ_VAL_16S:
+	{
+		struct ioat_pq_descriptor *pq = desc->pq;
+
+		/* check if there's error written */
+		if (!pq->dwbes_f.wbes)
+			return;
+
+		/* need to set a chanerr var for checking to clear later */
+
+		if (pq->dwbes_f.p_val_err)
+			*desc->result |= SUM_CHECK_P_RESULT;
+
+		if (pq->dwbes_f.q_val_err)
+			*desc->result |= SUM_CHECK_Q_RESULT;
+
+		return;
+	}
+	default:
+		return;
+	}
+}
+
 /**
  * __cleanup - reclaim used descriptors
  * @ioat: channel (ring) to clean
@@ -255,9 +363,10 @@ static bool desc_has_ext(struct ioat_ring_ent *desc)
  * The difference from the dma_v2.c __cleanup() is that this routine
  * handles extended descriptors and dma-unmapping raid operations.
  */
-static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
+static void __cleanup(struct ioat2_dma_chan *ioat, dma_addr_t phys_complete)
 {
 	struct ioat_chan_common *chan = &ioat->base;
+	struct ioatdma_device *device = chan->device;
 	struct ioat_ring_ent *desc;
 	bool seen_current = false;
 	int idx = ioat->tail, i;
@@ -266,6 +375,16 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 	dev_dbg(to_dev(chan), "%s: head: %#x tail: %#x issued: %#x\n",
 		__func__, ioat->head, ioat->tail, ioat->issued);
 
+	/*
+	 * At restart of the channel, the completion address and the
+	 * channel status will be 0 due to starting a new chain. Since
+	 * it's new chain and the first descriptor "fails", there is
+	 * nothing to clean up. We do not want to reap the entire submitted
+	 * chain due to this 0 address value and then BUG.
+	 */
+	if (!phys_complete)
+		return;
+
 	active = ioat2_ring_active(ioat);
 	for (i = 0; i < active && !seen_current; i++) {
 		struct dma_async_tx_descriptor *tx;
@@ -274,11 +393,15 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 		prefetch(ioat2_get_ring_ent(ioat, idx + i + 1));
 		desc = ioat2_get_ring_ent(ioat, idx + i);
 		dump_desc_dbg(ioat, desc);
+
+		/* set err stat if we are using dwbes */
+		if (device->cap & IOAT_CAP_DWBES)
+			desc_get_errstat(ioat, desc);
+
 		tx = &desc->txd;
 		if (tx->cookie) {
-			chan->completed_cookie = tx->cookie;
-			ioat3_dma_unmap(ioat, desc, idx + i);
-			tx->cookie = 0;
+			dma_cookie_complete(tx);
+			dma_descriptor_unmap(tx);
 			if (tx->callback) {
 				tx->callback(tx->callback_param);
 				tx->callback = NULL;
@@ -293,6 +416,12 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 			BUG_ON(i + 1 >= active);
 			i++;
 		}
+
+		/* cleanup super extended descriptors */
+		if (desc->sed) {
+			ioat3_free_sed(device, desc->sed);
+			desc->sed = NULL;
+		}
 	}
 	smp_mb(); /* finish all descriptor reads before incrementing tail */
 	ioat->tail = idx + i;
@@ -313,89 +442,127 @@ static void __cleanup(struct ioat2_dma_chan *ioat, unsigned long phys_complete)
 static void ioat3_cleanup(struct ioat2_dma_chan *ioat)
 {
 	struct ioat_chan_common *chan = &ioat->base;
-	unsigned long phys_complete;
+	u64 phys_complete;
 
 	spin_lock_bh(&chan->cleanup_lock);
-	if (ioat_cleanup_preamble(chan, &phys_complete))
+
+	if (ioat3_cleanup_preamble(chan, &phys_complete))
 		__cleanup(ioat, phys_complete);
+
+	if (is_ioat_halted(*chan->completion)) {
+		u32 chanerr = readl(chan->reg_base + IOAT_CHANERR_OFFSET);
+
+		if (chanerr & IOAT_CHANERR_HANDLE_MASK) {
+			mod_timer(&chan->timer, jiffies + IDLE_TIMEOUT);
+			ioat3_eh(ioat);
+		}
+	}
+
 	spin_unlock_bh(&chan->cleanup_lock);
 }
 
 static void ioat3_cleanup_event(unsigned long data)
 {
 	struct ioat2_dma_chan *ioat = to_ioat2_chan((void *) data);
+	struct ioat_chan_common *chan = &ioat->base;
 
 	ioat3_cleanup(ioat);
+	if (!test_bit(IOAT_RUN, &chan->state))
+		return;
 	writew(IOAT_CHANCTRL_RUN, ioat->base.reg_base + IOAT_CHANCTRL_OFFSET);
 }
 
 static void ioat3_restart_channel(struct ioat2_dma_chan *ioat)
 {
 	struct ioat_chan_common *chan = &ioat->base;
-	unsigned long phys_complete;
+	u64 phys_complete;
 
 	ioat2_quiesce(chan, 0);
-	if (ioat_cleanup_preamble(chan, &phys_complete))
+	if (ioat3_cleanup_preamble(chan, &phys_complete))
 		__cleanup(ioat, phys_complete);
 
 	__ioat2_restart_chan(ioat);
 }
 
-static void ioat3_timer_event(unsigned long data)
+static void ioat3_eh(struct ioat2_dma_chan *ioat)
 {
-	struct ioat2_dma_chan *ioat = to_ioat2_chan((void *) data);
 	struct ioat_chan_common *chan = &ioat->base;
+	struct pci_dev *pdev = to_pdev(chan);
+	struct ioat_dma_descriptor *hw;
+	u64 phys_complete;
+	struct ioat_ring_ent *desc;
+	u32 err_handled = 0;
+	u32 chanerr_int;
+	u32 chanerr;
 
-	if (test_bit(IOAT_COMPLETION_PENDING, &chan->state)) {
-		unsigned long phys_complete;
-		u64 status;
+	/* cleanup so tail points to descriptor that caused the error */
+	if (ioat3_cleanup_preamble(chan, &phys_complete))
+		__cleanup(ioat, phys_complete);
 
-		status = ioat_chansts(chan);
+	chanerr = readl(chan->reg_base + IOAT_CHANERR_OFFSET);
+	pci_read_config_dword(pdev, IOAT_PCI_CHANERR_INT_OFFSET, &chanerr_int);
 
-		/* when halted due to errors check for channel
-		 * programming errors before advancing the completion state
-		 */
-		if (is_ioat_halted(status)) {
-			u32 chanerr;
-
-			chanerr = readl(chan->reg_base + IOAT_CHANERR_OFFSET);
-			dev_err(to_dev(chan), "%s: Channel halted (%x)\n",
-				__func__, chanerr);
-			if (test_bit(IOAT_RUN, &chan->state))
-				BUG_ON(is_ioat_bug(chanerr));
-			else /* we never got off the ground */
-				return;
-		}
+	dev_dbg(to_dev(chan), "%s: error = %x:%x\n",
+		__func__, chanerr, chanerr_int);
 
-		/* if we haven't made progress and we have already
-		 * acknowledged a pending completion once, then be more
-		 * forceful with a restart
-		 */
-		spin_lock_bh(&chan->cleanup_lock);
-		if (ioat_cleanup_preamble(chan, &phys_complete))
-			__cleanup(ioat, phys_complete);
-		else if (test_bit(IOAT_COMPLETION_ACK, &chan->state)) {
-			spin_lock_bh(&ioat->prep_lock);
-			ioat3_restart_channel(ioat);
-			spin_unlock_bh(&ioat->prep_lock);
-		} else {
-			set_bit(IOAT_COMPLETION_ACK, &chan->state);
-			mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
+	desc = ioat2_get_ring_ent(ioat, ioat->tail);
+	hw = desc->hw;
+	dump_desc_dbg(ioat, desc);
+
+	switch (hw->ctl_f.op) {
+	case IOAT_OP_XOR_VAL:
+		if (chanerr & IOAT_CHANERR_XOR_P_OR_CRC_ERR) {
+			*desc->result |= SUM_CHECK_P_RESULT;
+			err_handled |= IOAT_CHANERR_XOR_P_OR_CRC_ERR;
 		}
-		spin_unlock_bh(&chan->cleanup_lock);
-	} else {
-		u16 active;
+		break;
+	case IOAT_OP_PQ_VAL:
+	case IOAT_OP_PQ_VAL_16S:
+		if (chanerr & IOAT_CHANERR_XOR_P_OR_CRC_ERR) {
+			*desc->result |= SUM_CHECK_P_RESULT;
+			err_handled |= IOAT_CHANERR_XOR_P_OR_CRC_ERR;
+		}
+		if (chanerr & IOAT_CHANERR_XOR_Q_ERR) {
+			*desc->result |= SUM_CHECK_Q_RESULT;
+			err_handled |= IOAT_CHANERR_XOR_Q_ERR;
+		}
+		break;
+	}
+
+	/* fault on unhandled error or spurious halt */
+	if (chanerr ^ err_handled || chanerr == 0) {
+		dev_err(to_dev(chan), "%s: fatal error (%x:%x)\n",
+			__func__, chanerr, err_handled);
+		BUG();
+	}
+
+	writel(chanerr, chan->reg_base + IOAT_CHANERR_OFFSET);
+	pci_write_config_dword(pdev, IOAT_PCI_CHANERR_INT_OFFSET, chanerr_int);
+
+	/* mark faulting descriptor as complete */
+	*chan->completion = desc->txd.phys;
+
+	spin_lock_bh(&ioat->prep_lock);
+	ioat3_restart_channel(ioat);
+	spin_unlock_bh(&ioat->prep_lock);
+}
 
+static void check_active(struct ioat2_dma_chan *ioat)
+{
+	struct ioat_chan_common *chan = &ioat->base;
+
+	if (ioat2_ring_active(ioat)) {
+		mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
+		return;
+	}
+
+	if (test_and_clear_bit(IOAT_CHAN_ACTIVE, &chan->state))
+		mod_timer(&chan->timer, jiffies + IDLE_TIMEOUT);
+	else if (ioat->alloc_order > ioat_get_alloc_order()) {
 		/* if the ring is idle, empty, and oversized try to step
 		 * down the size
 		 */
-		spin_lock_bh(&chan->cleanup_lock);
-		spin_lock_bh(&ioat->prep_lock);
-		active = ioat2_ring_active(ioat);
-		if (active == 0 && ioat->alloc_order > ioat_get_alloc_order())
-			reshape_ring(ioat, ioat->alloc_order-1);
-		spin_unlock_bh(&ioat->prep_lock);
-		spin_unlock_bh(&chan->cleanup_lock);
+		reshape_ring(ioat, ioat->alloc_order - 1);
 
 		/* keep shrinking until we get back to our minimum
 		 * default size
@@ -403,65 +570,76 @@ static void ioat3_timer_event(unsigned long data)
 		if (ioat->alloc_order > ioat_get_alloc_order())
 			mod_timer(&chan->timer, jiffies + IDLE_TIMEOUT);
 	}
+
 }
 
-static enum dma_status
-ioat3_tx_status(struct dma_chan *c, dma_cookie_t cookie,
-		struct dma_tx_state *txstate)
+static void ioat3_timer_event(unsigned long data)
 {
-	struct ioat2_dma_chan *ioat = to_ioat2_chan(c);
+	struct ioat2_dma_chan *ioat = to_ioat2_chan((void *) data);
+	struct ioat_chan_common *chan = &ioat->base;
+	dma_addr_t phys_complete;
+	u64 status;
 
-	if (ioat_tx_status(c, cookie, txstate) == DMA_SUCCESS)
-		return DMA_SUCCESS;
+	status = ioat_chansts(chan);
 
-	ioat3_cleanup(ioat);
+	/* when halted due to errors check for channel
+	 * programming errors before advancing the completion state
+	 */
+	if (is_ioat_halted(status)) {
+		u32 chanerr;
+
+		chanerr = readl(chan->reg_base + IOAT_CHANERR_OFFSET);
+		dev_err(to_dev(chan), "%s: Channel halted (%x)\n",
+			__func__, chanerr);
+		if (test_bit(IOAT_RUN, &chan->state))
+			BUG_ON(is_ioat_bug(chanerr));
+		else /* we never got off the ground */
+			return;
+	}
 
-	return ioat_tx_status(c, cookie, txstate);
+	/* if we haven't made progress and we have already
+	 * acknowledged a pending completion once, then be more
+	 * forceful with a restart
+	 */
+	spin_lock_bh(&chan->cleanup_lock);
+	if (ioat_cleanup_preamble(chan, &phys_complete))
+		__cleanup(ioat, phys_complete);
+	else if (test_bit(IOAT_COMPLETION_ACK, &chan->state)) {
+		spin_lock_bh(&ioat->prep_lock);
+		ioat3_restart_channel(ioat);
+		spin_unlock_bh(&ioat->prep_lock);
+		spin_unlock_bh(&chan->cleanup_lock);
+		return;
+	} else {
+		set_bit(IOAT_COMPLETION_ACK, &chan->state);
+		mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
+	}
+
+
+	if (ioat2_ring_active(ioat))
+		mod_timer(&chan->timer, jiffies + COMPLETION_TIMEOUT);
+	else {
+		spin_lock_bh(&ioat->prep_lock);
+		check_active(ioat);
+		spin_unlock_bh(&ioat->prep_lock);
+	}
+	spin_unlock_bh(&chan->cleanup_lock);
 }
 
-static struct dma_async_tx_descriptor *
-ioat3_prep_memset_lock(struct dma_chan *c, dma_addr_t dest, int value,
-		       size_t len, unsigned long flags)
+static enum dma_status
+ioat3_tx_status(struct dma_chan *c, dma_cookie_t cookie,
+		struct dma_tx_state *txstate)
 {
 	struct ioat2_dma_chan *ioat = to_ioat2_chan(c);
-	struct ioat_ring_ent *desc;
-	size_t total_len = len;
-	struct ioat_fill_descriptor *fill;
-	u64 src_data = (0x0101010101010101ULL) * (value & 0xff);
-	int num_descs, idx, i;
-
-	num_descs = ioat2_xferlen_to_descs(ioat, len);
-	if (likely(num_descs) && ioat2_check_space_lock(ioat, num_descs) == 0)
-		idx = ioat->head;
-	else
-		return NULL;
-	i = 0;
-	do {
-		size_t xfer_size = min_t(size_t, len, 1 << ioat->xfercap_log);
-
-		desc = ioat2_get_ring_ent(ioat, idx + i);
-		fill = desc->fill;
-
-		fill->size = xfer_size;
-		fill->src_data = src_data;
-		fill->dst_addr = dest;
-		fill->ctl = 0;
-		fill->ctl_f.op = IOAT_OP_FILL;
+	enum dma_status ret;
 
-		len -= xfer_size;
-		dest += xfer_size;
-		dump_desc_dbg(ioat, desc);
-	} while (++i < num_descs);
+	ret = dma_cookie_status(c, cookie, txstate);
+	if (ret == DMA_COMPLETE)
+		return ret;
 
-	desc->txd.flags = flags;
-	desc->len = total_len;
-	fill->ctl_f.int_en = !!(flags & DMA_PREP_INTERRUPT);
-	fill->ctl_f.fence = !!(flags & DMA_PREP_FENCE);
-	fill->ctl_f.compl_write = 1;
-	dump_desc_dbg(ioat, desc);
+	ioat3_cleanup(ioat);
 
-	/* we leave the channel locked to ensure in order submission */
-	return &desc->txd;
+	return dma_cookie_status(c, cookie, txstate);
 }
 
 static struct dma_async_tx_descriptor *
@@ -587,7 +765,8 @@ dump_pq_desc_dbg(struct ioat2_dma_chan *ioat, struct ioat_ring_ent *desc, struct
 	int i;
 
 	dev_dbg(dev, "desc[%d]: (%#llx->%#llx) flags: %#x"
-		" sz: %#x ctl: %#x (op: %d int: %d compl: %d pq: '%s%s' src_cnt: %d)\n",
+		" sz: %#10.8x ctl: %#x (op: %#x int: %d compl: %d pq: '%s%s'"
+		" src_cnt: %d)\n",
 		desc_id(desc), (unsigned long long) desc->txd.phys,
 		(unsigned long long) (pq_ex ? pq_ex->next : pq->next),
 		desc->txd.flags, pq->size, pq->ctl, pq->ctl_f.op, pq->ctl_f.int_en,
@@ -599,6 +778,42 @@ dump_pq_desc_dbg(struct ioat2_dma_chan *ioat, struct ioat_ring_ent *desc, struct
 			(unsigned long long) pq_get_src(descs, i), pq->coef[i]);
 	dev_dbg(dev, "\tP: %#llx\n", pq->p_addr);
 	dev_dbg(dev, "\tQ: %#llx\n", pq->q_addr);
+	dev_dbg(dev, "\tNEXT: %#llx\n", pq->next);
+}
+
+static void dump_pq16_desc_dbg(struct ioat2_dma_chan *ioat,
+			       struct ioat_ring_ent *desc)
+{
+	struct device *dev = to_dev(&ioat->base);
+	struct ioat_pq_descriptor *pq = desc->pq;
+	struct ioat_raw_descriptor *descs[] = { (void *)pq,
+						(void *)pq,
+						(void *)pq };
+	int src_cnt = src16_cnt_to_sw(pq->ctl_f.src_cnt);
+	int i;
+
+	if (desc->sed) {
+		descs[1] = (void *)desc->sed->hw;
+		descs[2] = (void *)desc->sed->hw + 64;
+	}
+
+	dev_dbg(dev, "desc[%d]: (%#llx->%#llx) flags: %#x"
+		" sz: %#x ctl: %#x (op: %#x int: %d compl: %d pq: '%s%s'"
+		" src_cnt: %d)\n",
+		desc_id(desc), (unsigned long long) desc->txd.phys,
+		(unsigned long long) pq->next,
+		desc->txd.flags, pq->size, pq->ctl,
+		pq->ctl_f.op, pq->ctl_f.int_en,
+		pq->ctl_f.compl_write,
+		pq->ctl_f.p_disable ? "" : "p", pq->ctl_f.q_disable ? "" : "q",
+		pq->ctl_f.src_cnt);
+	for (i = 0; i < src_cnt; i++) {
+		dev_dbg(dev, "\tsrc[%d]: %#llx coef: %#x\n", i,
+			(unsigned long long) pq16_get_src(descs, i),
+			pq->coef[i]);
+	}
+	dev_dbg(dev, "\tP: %#llx\n", pq->p_addr);
+	dev_dbg(dev, "\tQ: %#llx\n", pq->q_addr);
 }
 
 static struct dma_async_tx_descriptor *
@@ -609,6 +824,7 @@ __ioat3_prep_pq_lock(struct dma_chan *c, enum sum_check_flags *result,
 {
 	struct ioat2_dma_chan *ioat = to_ioat2_chan(c);
 	struct ioat_chan_common *chan = &ioat->base;
+	struct ioatdma_device *device = chan->device;
 	struct ioat_ring_ent *compl_desc;
 	struct ioat_ring_ent *desc;
 	struct ioat_ring_ent *ext;
@@ -619,6 +835,7 @@ __ioat3_prep_pq_lock(struct dma_chan *c, enum sum_check_flags *result,
 	u32 offset = 0;
 	u8 op = result ? IOAT_OP_PQ_VAL : IOAT_OP_PQ;
 	int i, s, idx, with_ext, num_descs;
+	int cb32 = (device->version < IOAT_VER_3_3) ? 1 : 0;
 
 	dev_dbg(to_dev(chan), "%s\n", __func__);
 	/* the engine requires at least two sources (we provide
@@ -644,7 +861,7 @@ __ioat3_prep_pq_lock(struct dma_chan *c, enum sum_check_flags *result,
 	 * order.
 	 */
 	if (likely(num_descs) &&
-	    ioat2_check_space_lock(ioat, num_descs+1) == 0)
+	    ioat2_check_space_lock(ioat, num_descs + cb32) == 0)
 		idx = ioat->head;
 	else
 		return NULL;
@@ -682,6 +899,9 @@ __ioat3_prep_pq_lock(struct dma_chan *c, enum sum_check_flags *result,
 		pq->q_addr = dst[1] + offset;
 		pq->ctl = 0;
 		pq->ctl_f.op = op;
+		/* we turn on descriptor write back error status */
+		if (device->cap & IOAT_CAP_DWBES)
+			pq->ctl_f.wb_en = result ? 1 : 0;
 		pq->ctl_f.src_cnt = src_cnt_to_hw(s);
 		pq->ctl_f.p_disable = !!(flags & DMA_PREP_PQ_DISABLE_P);
 		pq->ctl_f.q_disable = !!(flags & DMA_PREP_PQ_DISABLE_Q);
@@ -698,22 +918,140 @@ __ioat3_prep_pq_lock(struct dma_chan *c, enum sum_check_flags *result,
 	pq->ctl_f.fence = !!(flags & DMA_PREP_FENCE);
 	dump_pq_desc_dbg(ioat, desc, ext);
 
-	/* completion descriptor carries interrupt bit */
-	compl_desc = ioat2_get_ring_ent(ioat, idx + i);
-	compl_desc->txd.flags = flags & DMA_PREP_INTERRUPT;
-	hw = compl_desc->hw;
-	hw->ctl = 0;
-	hw->ctl_f.null = 1;
-	hw->ctl_f.int_en = !!(flags & DMA_PREP_INTERRUPT);
-	hw->ctl_f.compl_write = 1;
-	hw->size = NULL_DESC_BUFFER_SIZE;
-	dump_desc_dbg(ioat, compl_desc);
+	if (!cb32) {
+		pq->ctl_f.int_en = !!(flags & DMA_PREP_INTERRUPT);
+		pq->ctl_f.compl_write = 1;
+		compl_desc = desc;
+	} else {
+		/* completion descriptor carries interrupt bit */
+		compl_desc = ioat2_get_ring_ent(ioat, idx + i);
+		compl_desc->txd.flags = flags & DMA_PREP_INTERRUPT;
+		hw = compl_desc->hw;
+		hw->ctl = 0;
+		hw->ctl_f.null = 1;
+		hw->ctl_f.int_en = !!(flags & DMA_PREP_INTERRUPT);
+		hw->ctl_f.compl_write = 1;
+		hw->size = NULL_DESC_BUFFER_SIZE;
+		dump_desc_dbg(ioat, compl_desc);
+	}
+
 
 	/* we leave the channel locked to ensure in order submission */
 	return &compl_desc->txd;
 }
 
 static struct dma_async_tx_descriptor *
+__ioat3_prep_pq16_lock(struct dma_chan *c, enum sum_check_flags *result,
+		       const dma_addr_t *dst, const dma_addr_t *src,
+		       unsigned int src_cnt, const unsigned char *scf,
+		       size_t len, unsigned long flags)
+{
+	struct ioat2_dma_chan *ioat = to_ioat2_chan(c);
+	struct ioat_chan_common *chan = &ioat->base;
+	struct ioatdma_device *device = chan->device;
+	struct ioat_ring_ent *desc;
+	size_t total_len = len;
+	struct ioat_pq_descriptor *pq;
+	u32 offset = 0;
+	u8 op;
+	int i, s, idx, num_descs;
+
+	/* this function is only called with 9-16 sources */
+	op = result ? IOAT_OP_PQ_VAL_16S : IOAT_OP_PQ_16S;
+
+	dev_dbg(to_dev(chan), "%s\n", __func__);
+
+	num_descs = ioat2_xferlen_to_descs(ioat, len);
+
+	/*
+	 * 16 source pq is only available on cb3.3 and has no completion
+	 * write hw bug.
+	 */
+	if (num_descs && ioat2_check_space_lock(ioat, num_descs) == 0)
+		idx = ioat->head;
+	else
+		return NULL;
+
+	i = 0;
+
+	do {
+		struct ioat_raw_descriptor *descs[4];
+		size_t xfer_size = min_t(size_t, len, 1 << ioat->xfercap_log);
+
+		desc = ioat2_get_ring_ent(ioat, idx + i);
+		pq = desc->pq;
+
+		descs[0] = (struct ioat_raw_descriptor *) pq;
+
+		desc->sed = ioat3_alloc_sed(device, (src_cnt-2) >> 3);
+		if (!desc->sed) {
+			dev_err(to_dev(chan),
+				"%s: no free sed entries\n", __func__);
+			return NULL;
+		}
+
+		pq->sed_addr = desc->sed->dma;
+		desc->sed->parent = desc;
+
+		descs[1] = (struct ioat_raw_descriptor *)desc->sed->hw;
+		descs[2] = (void *)descs[1] + 64;
+
+		for (s = 0; s < src_cnt; s++)
+			pq16_set_src(descs, src[s], offset, scf[s], s);
+
+		/* see the comment for dma_maxpq in include/linux/dmaengine.h */
+		if (dmaf_p_disabled_continue(flags))
+			pq16_set_src(descs, dst[1], offset, 1, s++);
+		else if (dmaf_continue(flags)) {
+			pq16_set_src(descs, dst[0], offset, 0, s++);
+			pq16_set_src(descs, dst[1], offset, 1, s++);
+			pq16_set_src(descs, dst[1], offset, 0, s++);
+		}
+
+		pq->size = xfer_size;
+		pq->p_addr = dst[0] + offset;
+		pq->q_addr = dst[1] + offset;
+		pq->ctl = 0;
+		pq->ctl_f.op = op;
+		pq->ctl_f.src_cnt = src16_cnt_to_hw(s);
+		/* we turn on descriptor write back error status */
+		if (device->cap & IOAT_CAP_DWBES)
+			pq->ctl_f.wb_en = result ? 1 : 0;
+		pq->ctl_f.p_disable = !!(flags & DMA_PREP_PQ_DISABLE_P);
+		pq->ctl_f.q_disable = !!(flags & DMA_PREP_PQ_DISABLE_Q);
+
+		len -= xfer_size;
+		offset += xfer_size;
+	} while (++i < num_descs);
+
+	/* last pq descriptor carries the unmap parameters and fence bit */
+	desc->txd.flags = flags;
+	desc->len = total_len;
+	if (result)
+		desc->result = result;
+	pq->ctl_f.fence = !!(flags & DMA_PREP_FENCE);
+
+	/* with cb3.3 we should be able to do completion w/o a null desc */
+	pq->ctl_f.int_en = !!(flags & DMA_PREP_INTERRUPT);
+	pq->ctl_f.compl_write = 1;
+
+	dump_pq16_desc_dbg(ioat, desc);
+
+	/* we leave the channel locked to ensure in order submission */
+	return &desc->txd;
+}
+
+static int src_cnt_flags(unsigned int src_cnt, unsigned long flags)
+{
+	if (dmaf_p_disabled_continue(flags))
+		return src_cnt + 1;
+	else if (dmaf_continue(flags))
+		return src_cnt + 3;
+	else
+		return src_cnt;
+}
+
+static struct dma_async_tx_descriptor *
 ioat3_prep_pq(struct dma_chan *chan, dma_addr_t *dst, dma_addr_t *src,
 	      unsigned int src_cnt, const unsigned char *scf, size_t len,
 	      unsigned long flags)
@@ -737,11 +1075,20 @@ ioat3_prep_pq(struct dma_chan *chan, dma_addr_t *dst, dma_addr_t *src,
 		single_source_coef[0] = scf[0];
 		single_source_coef[1] = 0;
 
-		return __ioat3_prep_pq_lock(chan, NULL, dst, single_source, 2,
-					    single_source_coef, len, flags);
-	} else
-		return __ioat3_prep_pq_lock(chan, NULL, dst, src, src_cnt, scf,
-					    len, flags);
+		return src_cnt_flags(src_cnt, flags) > 8 ?
+			__ioat3_prep_pq16_lock(chan, NULL, dst, single_source,
+					       2, single_source_coef, len,
+					       flags) :
+			__ioat3_prep_pq_lock(chan, NULL, dst, single_source, 2,
+					     single_source_coef, len, flags);
+
+	} else {
+		return src_cnt_flags(src_cnt, flags) > 8 ?
+			__ioat3_prep_pq16_lock(chan, NULL, dst, src, src_cnt,
+					       scf, len, flags) :
+			__ioat3_prep_pq_lock(chan, NULL, dst, src, src_cnt,
+					     scf, len, flags);
+	}
 }
 
 struct dma_async_tx_descriptor *
@@ -760,8 +1107,11 @@ ioat3_prep_pq_val(struct dma_chan *chan, dma_addr_t *pq, dma_addr_t *src,
 	 */
 	*pqres = 0;
 
-	return __ioat3_prep_pq_lock(chan, pqres, pq, src, src_cnt, scf, len,
-				    flags);
+	return src_cnt_flags(src_cnt, flags) > 8 ?
+		__ioat3_prep_pq16_lock(chan, pqres, pq, src, src_cnt, scf, len,
+				       flags) :
+		__ioat3_prep_pq_lock(chan, pqres, pq, src, src_cnt, scf, len,
+				     flags);
 }
 
 static struct dma_async_tx_descriptor *
@@ -776,8 +1126,11 @@ ioat3_prep_pqxor(struct dma_chan *chan, dma_addr_t dst, dma_addr_t *src,
 	flags |= DMA_PREP_PQ_DISABLE_Q;
 	pq[1] = dst; /* specify valid address for disabled result */
 
-	return __ioat3_prep_pq_lock(chan, NULL, pq, src, src_cnt, scf, len,
-				    flags);
+	return src_cnt_flags(src_cnt, flags) > 8 ?
+		__ioat3_prep_pq16_lock(chan, NULL, pq, src, src_cnt, scf, len,
+				       flags) :
+		__ioat3_prep_pq_lock(chan, NULL, pq, src, src_cnt, scf, len,
+				     flags);
 }
 
 struct dma_async_tx_descriptor *
@@ -798,8 +1151,11 @@ ioat3_prep_pqxor_val(struct dma_chan *chan, dma_addr_t *src,
 	flags |= DMA_PREP_PQ_DISABLE_Q;
 	pq[1] = pq[0]; /* specify valid address for disabled result */
 
-	return __ioat3_prep_pq_lock(chan, result, pq, &src[1], src_cnt - 1, scf,
-				    len, flags);
+	return src_cnt_flags(src_cnt, flags) > 8 ?
+		__ioat3_prep_pq16_lock(chan, result, pq, &src[1], src_cnt - 1,
+				       scf, len, flags) :
+		__ioat3_prep_pq_lock(chan, result, pq, &src[1], src_cnt - 1,
+				     scf, len, flags);
 }
 
 static struct dma_async_tx_descriptor *
@@ -833,7 +1189,7 @@ ioat3_prep_interrupt_lock(struct dma_chan *c, unsigned long flags)
 	return &desc->txd;
 }
 
-static void __devinit ioat3_dma_test_callback(void *dma_async_param)
+static void ioat3_dma_test_callback(void *dma_async_param)
 {
 	struct completion *cmp = dma_async_param;
 
@@ -841,14 +1197,14 @@ static void __devinit ioat3_dma_test_callback(void *dma_async_param)
 }
 
 #define IOAT_NUM_SRC_TEST 6 /* must be <= 8 */
-static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
+static int ioat_xor_val_self_test(struct ioatdma_device *device)
 {
 	int i, src_idx;
 	struct page *dest;
 	struct page *xor_srcs[IOAT_NUM_SRC_TEST];
 	struct page *xor_val_srcs[IOAT_NUM_SRC_TEST + 1];
 	dma_addr_t dma_srcs[IOAT_NUM_SRC_TEST + 1];
-	dma_addr_t dma_addr, dest_dma;
+	dma_addr_t dest_dma;
 	struct dma_async_tx_descriptor *tx;
 	struct dma_chan *dma_chan;
 	dma_cookie_t cookie;
@@ -860,6 +1216,7 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 	unsigned long tmo;
 	struct device *dev = &device->pdev->dev;
 	struct dma_device *dma = &device->common;
+	u8 op = 0;
 
 	dev_dbg(dev, "%s\n", __func__);
 
@@ -905,6 +1262,8 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 	}
 
 	/* test xor */
+	op = IOAT_OP_XOR;
+
 	dest_dma = dma_map_page(dev, dest, 0, PAGE_SIZE, DMA_FROM_DEVICE);
 	for (i = 0; i < IOAT_NUM_SRC_TEST; i++)
 		dma_srcs[i] = dma_map_page(dev, xor_srcs[i], 0, PAGE_SIZE,
@@ -916,7 +1275,7 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 	if (!tx) {
 		dev_err(dev, "Self-test xor prep failed\n");
 		err = -ENODEV;
-		goto free_resources;
+		goto dma_unmap;
 	}
 
 	async_tx_ack(tx);
@@ -927,18 +1286,22 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 	if (cookie < 0) {
 		dev_err(dev, "Self-test xor setup failed\n");
 		err = -ENODEV;
-		goto free_resources;
+		goto dma_unmap;
 	}
 	dma->device_issue_pending(dma_chan);
 
 	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
 
-	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
+	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_COMPLETE) {
 		dev_err(dev, "Self-test xor timed out\n");
 		err = -ENODEV;
-		goto free_resources;
+		goto dma_unmap;
 	}
 
+	dma_unmap_page(dev, dest_dma, PAGE_SIZE, DMA_FROM_DEVICE);
+	for (i = 0; i < IOAT_NUM_SRC_TEST; i++)
+		dma_unmap_page(dev, dma_srcs[i], PAGE_SIZE, DMA_TO_DEVICE);
+
 	dma_sync_single_for_cpu(dev, dest_dma, PAGE_SIZE, DMA_FROM_DEVICE);
 	for (i = 0; i < (PAGE_SIZE / sizeof(u32)); i++) {
 		u32 *ptr = page_address(dest);
@@ -948,12 +1311,14 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 			goto free_resources;
 		}
 	}
-	dma_sync_single_for_device(dev, dest_dma, PAGE_SIZE, DMA_TO_DEVICE);
+	dma_sync_single_for_device(dev, dest_dma, PAGE_SIZE, DMA_FROM_DEVICE);
 
 	/* skip validate if the capability is not present */
 	if (!dma_has_cap(DMA_XOR_VAL, dma_chan->device->cap_mask))
 		goto free_resources;
 
+	op = IOAT_OP_XOR_VAL;
+
 	/* validate the sources with the destintation page */
 	for (i = 0; i < IOAT_NUM_SRC_TEST; i++)
 		xor_val_srcs[i] = xor_srcs[i];
@@ -970,7 +1335,7 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 	if (!tx) {
 		dev_err(dev, "Self-test zero prep failed\n");
 		err = -ENODEV;
-		goto free_resources;
+		goto dma_unmap;
 	}
 
 	async_tx_ack(tx);
@@ -981,69 +1346,32 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 	if (cookie < 0) {
 		dev_err(dev, "Self-test zero setup failed\n");
 		err = -ENODEV;
-		goto free_resources;
+		goto dma_unmap;
 	}
 	dma->device_issue_pending(dma_chan);
 
 	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
 
-	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
+	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_COMPLETE) {
 		dev_err(dev, "Self-test validate timed out\n");
 		err = -ENODEV;
-		goto free_resources;
+		goto dma_unmap;
 	}
 
+	for (i = 0; i < IOAT_NUM_SRC_TEST + 1; i++)
+		dma_unmap_page(dev, dma_srcs[i], PAGE_SIZE, DMA_TO_DEVICE);
+
 	if (xor_val_result != 0) {
 		dev_err(dev, "Self-test validate failed compare\n");
 		err = -ENODEV;
 		goto free_resources;
 	}
 
-	/* skip memset if the capability is not present */
-	if (!dma_has_cap(DMA_MEMSET, dma_chan->device->cap_mask))
-		goto free_resources;
-
-	/* test memset */
-	dma_addr = dma_map_page(dev, dest, 0,
-			PAGE_SIZE, DMA_FROM_DEVICE);
-	tx = dma->device_prep_dma_memset(dma_chan, dma_addr, 0, PAGE_SIZE,
-					 DMA_PREP_INTERRUPT);
-	if (!tx) {
-		dev_err(dev, "Self-test memset prep failed\n");
-		err = -ENODEV;
-		goto free_resources;
-	}
-
-	async_tx_ack(tx);
-	init_completion(&cmp);
-	tx->callback = ioat3_dma_test_callback;
-	tx->callback_param = &cmp;
-	cookie = tx->tx_submit(tx);
-	if (cookie < 0) {
-		dev_err(dev, "Self-test memset setup failed\n");
-		err = -ENODEV;
-		goto free_resources;
-	}
-	dma->device_issue_pending(dma_chan);
-
-	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
-
-	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
-		dev_err(dev, "Self-test memset timed out\n");
-		err = -ENODEV;
-		goto free_resources;
-	}
-
-	for (i = 0; i < PAGE_SIZE/sizeof(u32); i++) {
-		u32 *ptr = page_address(dest);
-		if (ptr[i]) {
-			dev_err(dev, "Self-test memset failed compare\n");
-			err = -ENODEV;
-			goto free_resources;
-		}
-	}
+	memset(page_address(dest), 0, PAGE_SIZE);
 
 	/* test for non-zero parity sum */
+	op = IOAT_OP_XOR_VAL;
+
 	xor_val_result = 0;
 	for (i = 0; i < IOAT_NUM_SRC_TEST + 1; i++)
 		dma_srcs[i] = dma_map_page(dev, xor_val_srcs[i], 0, PAGE_SIZE,
@@ -1054,7 +1382,7 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 	if (!tx) {
 		dev_err(dev, "Self-test 2nd zero prep failed\n");
 		err = -ENODEV;
-		goto free_resources;
+		goto dma_unmap;
 	}
 
 	async_tx_ack(tx);
@@ -1065,24 +1393,39 @@ static int __devinit ioat_xor_val_self_test(struct ioatdma_device *device)
 	if (cookie < 0) {
 		dev_err(dev, "Self-test  2nd zero setup failed\n");
 		err = -ENODEV;
-		goto free_resources;
+		goto dma_unmap;
 	}
 	dma->device_issue_pending(dma_chan);
 
 	tmo = wait_for_completion_timeout(&cmp, msecs_to_jiffies(3000));
 
-	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_SUCCESS) {
+	if (dma->device_tx_status(dma_chan, cookie, NULL) != DMA_COMPLETE) {
 		dev_err(dev, "Self-test 2nd validate timed out\n");
 		err = -ENODEV;
-		goto free_resources;
+		goto dma_unmap;
 	}
 
 	if (xor_val_result != SUM_CHECK_P_RESULT) {
 		dev_err(dev, "Self-test validate failed compare\n");
 		err = -ENODEV;
-		goto free_resources;
+		goto dma_unmap;
 	}
 
+	for (i = 0; i < IOAT_NUM_SRC_TEST + 1; i++)
+		dma_unmap_page(dev, dma_srcs[i], PAGE_SIZE, DMA_TO_DEVICE);
+
+	goto free_resources;
+dma_unmap:
+	if (op == IOAT_OP_XOR) {
+		dma_unmap_page(dev, dest_dma, PAGE_SIZE, DMA_FROM_DEVICE);
+		for (i = 0; i < IOAT_NUM_SRC_TEST; i++)
+			dma_unmap_page(dev, dma_srcs[i], PAGE_SIZE,
+				       DMA_TO_DEVICE);
+	} else if (op == IOAT_OP_XOR_VAL) {
+		for (i = 0; i < IOAT_NUM_SRC_TEST + 1; i++)
+			dma_unmap_page(dev, dma_srcs[i], PAGE_SIZE,
+				       DMA_TO_DEVICE);
+	}
 free_resources:
 	dma->device_free_chan_resources(dma_chan);
 out:
@@ -1093,7 +1436,7 @@ out:
 	return err;
 }
 
-static int __devinit ioat3_dma_self_test(struct ioatdma_device *device)
+static int ioat3_dma_self_test(struct ioatdma_device *device)
 {
 	int rc = ioat_dma_self_test(device);
 
@@ -1107,6 +1450,40 @@ static int __devinit ioat3_dma_self_test(struct ioatdma_device *device)
 	return 0;
 }
 
+static int ioat3_irq_reinit(struct ioatdma_device *device)
+{
+	struct pci_dev *pdev = device->pdev;
+	int irq = pdev->irq, i;
+
+	if (!is_bwd_ioat(pdev))
+		return 0;
+
+	switch (device->irq_mode) {
+	case IOAT_MSIX:
+		for (i = 0; i < device->common.chancnt; i++) {
+			struct msix_entry *msix = &device->msix_entries[i];
+			struct ioat_chan_common *chan;
+
+			chan = ioat_chan_by_index(device, i);
+			devm_free_irq(&pdev->dev, msix->vector, chan);
+		}
+
+		pci_disable_msix(pdev);
+		break;
+	case IOAT_MSI:
+		pci_disable_msi(pdev);
+		/* fall through */
+	case IOAT_INTX:
+		devm_free_irq(&pdev->dev, irq, device);
+		break;
+	default:
+		return 0;
+	}
+	device->irq_mode = IOAT_NOIRQ;
+
+	return ioat_dma_setup_interrupts(device);
+}
+
 static int ioat3_reset_hw(struct ioat_chan_common *chan)
 {
 	/* throw away whatever the channel was doing and get it
@@ -1123,30 +1500,66 @@ static int ioat3_reset_hw(struct ioat_chan_common *chan)
 	chanerr = readl(chan->reg_base + IOAT_CHANERR_OFFSET);
 	writel(chanerr, chan->reg_base + IOAT_CHANERR_OFFSET);
 
-	/* -= IOAT ver.3 workarounds =- */
-	/* Write CHANERRMSK_INT with 3E07h to mask out the errors
-	 * that can cause stability issues for IOAT ver.3, and clear any
-	 * pending errors
-	 */
-	pci_write_config_dword(pdev, IOAT_PCI_CHANERRMASK_INT_OFFSET, 0x3e07);
-	err = pci_read_config_dword(pdev, IOAT_PCI_CHANERR_INT_OFFSET, &chanerr);
-	if (err) {
-		dev_err(&pdev->dev, "channel error register unreachable\n");
-		return err;
+	if (device->version < IOAT_VER_3_3) {
+		/* clear any pending errors */
+		err = pci_read_config_dword(pdev,
+				IOAT_PCI_CHANERR_INT_OFFSET, &chanerr);
+		if (err) {
+			dev_err(&pdev->dev,
+				"channel error register unreachable\n");
+			return err;
+		}
+		pci_write_config_dword(pdev,
+				IOAT_PCI_CHANERR_INT_OFFSET, chanerr);
+
+		/* Clear DMAUNCERRSTS Cfg-Reg Parity Error status bit
+		 * (workaround for spurious config parity error after restart)
+		 */
+		pci_read_config_word(pdev, IOAT_PCI_DEVICE_ID_OFFSET, &dev_id);
+		if (dev_id == PCI_DEVICE_ID_INTEL_IOAT_TBG0) {
+			pci_write_config_dword(pdev,
+					       IOAT_PCI_DMAUNCERRSTS_OFFSET,
+					       0x10);
+		}
 	}
-	pci_write_config_dword(pdev, IOAT_PCI_CHANERR_INT_OFFSET, chanerr);
 
-	/* Clear DMAUNCERRSTS Cfg-Reg Parity Error status bit
-	 * (workaround for spurious config parity error after restart)
-	 */
-	pci_read_config_word(pdev, IOAT_PCI_DEVICE_ID_OFFSET, &dev_id);
-	if (dev_id == PCI_DEVICE_ID_INTEL_IOAT_TBG0)
-		pci_write_config_dword(pdev, IOAT_PCI_DMAUNCERRSTS_OFFSET, 0x10);
+	err = ioat2_reset_sync(chan, msecs_to_jiffies(200));
+	if (!err)
+		err = ioat3_irq_reinit(device);
+
+	if (err)
+		dev_err(&pdev->dev, "Failed to reset: %d\n", err);
 
-	return ioat2_reset_sync(chan, msecs_to_jiffies(200));
+	return err;
 }
 
-int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
+static void ioat3_intr_quirk(struct ioatdma_device *device)
+{
+	struct dma_device *dma;
+	struct dma_chan *c;
+	struct ioat_chan_common *chan;
+	u32 errmask;
+
+	dma = &device->common;
+
+	/*
+	 * if we have descriptor write back error status, we mask the
+	 * error interrupts
+	 */
+	if (device->cap & IOAT_CAP_DWBES) {
+		list_for_each_entry(c, &dma->channels, device_node) {
+			chan = to_chan_common(c);
+			errmask = readl(chan->reg_base +
+					IOAT_CHANERR_MASK_OFFSET);
+			errmask |= IOAT_CHANERR_XOR_P_OR_CRC_ERR |
+				   IOAT_CHANERR_XOR_Q_ERR;
+			writel(errmask, chan->reg_base +
+					IOAT_CHANERR_MASK_OFFSET);
+		}
+	}
+}
+
+int ioat3_dma_probe(struct ioatdma_device *device, int dca)
 {
 	struct pci_dev *pdev = device->pdev;
 	int dca_en = system_has_dca_enabled(pdev);
@@ -1155,11 +1568,11 @@ int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 	struct ioat_chan_common *chan;
 	bool is_raid_device = false;
 	int err;
-	u32 cap;
 
 	device->enumerate_channels = ioat2_enumerate_channels;
 	device->reset_hw = ioat3_reset_hw;
 	device->self_test = ioat3_dma_self_test;
+	device->intr_quirk = ioat3_intr_quirk;
 	dma = &device->common;
 	dma->device_prep_dma_memcpy = ioat2_dma_prep_memcpy_lock;
 	dma->device_issue_pending = ioat2_issue_pending;
@@ -1169,16 +1582,18 @@ int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 	dma_cap_set(DMA_INTERRUPT, dma->cap_mask);
 	dma->device_prep_dma_interrupt = ioat3_prep_interrupt_lock;
 
-	cap = readl(device->reg_base + IOAT_DMA_CAP_OFFSET);
+	device->cap = readl(device->reg_base + IOAT_DMA_CAP_OFFSET);
+
+	if (is_xeon_cb32(pdev) || is_bwd_noraid(pdev))
+		device->cap &= ~(IOAT_CAP_XOR | IOAT_CAP_PQ | IOAT_CAP_RAID16SS);
 
 	/* dca is incompatible with raid operations */
-	if (dca_en && (cap & (IOAT_CAP_XOR|IOAT_CAP_PQ)))
-		cap &= ~(IOAT_CAP_XOR|IOAT_CAP_PQ);
+	if (dca_en && (device->cap & (IOAT_CAP_XOR|IOAT_CAP_PQ)))
+		device->cap &= ~(IOAT_CAP_XOR|IOAT_CAP_PQ);
 
-	if (cap & IOAT_CAP_XOR) {
+	if (device->cap & IOAT_CAP_XOR) {
 		is_raid_device = true;
 		dma->max_xor = 8;
-		dma->xor_align = 6;
 
 		dma_cap_set(DMA_XOR, dma->cap_mask);
 		dma->device_prep_dma_xor = ioat3_prep_xor;
@@ -1186,53 +1601,56 @@ int __devinit ioat3_dma_probe(struct ioatdma_device *device, int dca)
 		dma_cap_set(DMA_XOR_VAL, dma->cap_mask);
 		dma->device_prep_dma_xor_val = ioat3_prep_xor_val;
 	}
-	if (cap & IOAT_CAP_PQ) {
+
+	if (device->cap & IOAT_CAP_PQ) {
 		is_raid_device = true;
-		dma_set_maxpq(dma, 8, 0);
-		dma->pq_align = 6;
 
-		dma_cap_set(DMA_PQ, dma->cap_mask);
 		dma->device_prep_dma_pq = ioat3_prep_pq;
-
-		dma_cap_set(DMA_PQ_VAL, dma->cap_mask);
 		dma->device_prep_dma_pq_val = ioat3_prep_pq_val;
+		dma_cap_set(DMA_PQ, dma->cap_mask);
+		dma_cap_set(DMA_PQ_VAL, dma->cap_mask);
 
-		if (!(cap & IOAT_CAP_XOR)) {
-			dma->max_xor = 8;
-			dma->xor_align = 6;
+		if (device->cap & IOAT_CAP_RAID16SS) {
+			dma_set_maxpq(dma, 16, 0);
+		} else {
+			dma_set_maxpq(dma, 8, 0);
+		}
 
-			dma_cap_set(DMA_XOR, dma->cap_mask);
+		if (!(device->cap & IOAT_CAP_XOR)) {
 			dma->device_prep_dma_xor = ioat3_prep_pqxor;
-
-			dma_cap_set(DMA_XOR_VAL, dma->cap_mask);
 			dma->device_prep_dma_xor_val = ioat3_prep_pqxor_val;
+			dma_cap_set(DMA_XOR, dma->cap_mask);
+			dma_cap_set(DMA_XOR_VAL, dma->cap_mask);
+
+			if (device->cap & IOAT_CAP_RAID16SS) {
+				dma->max_xor = 16;
+			} else {
+				dma->max_xor = 8;
+			}
 		}
 	}
-	if (is_raid_device && (cap & IOAT_CAP_FILL_BLOCK)) {
-		dma_cap_set(DMA_MEMSET, dma->cap_mask);
-		dma->device_prep_dma_memset = ioat3_prep_memset_lock;
-	}
 
+	dma->device_tx_status = ioat3_tx_status;
+	device->cleanup_fn = ioat3_cleanup_event;
+	device->timer_fn = ioat3_timer_event;
 
-	if (is_raid_device) {
-		dma->device_tx_status = ioat3_tx_status;
-		device->cleanup_fn = ioat3_cleanup_event;
-		device->timer_fn = ioat3_timer_event;
-	} else {
-		dma->device_tx_status = ioat_dma_tx_status;
-		device->cleanup_fn = ioat2_cleanup_event;
-		device->timer_fn = ioat2_timer_event;
-	}
+	/* starting with CB3.3 super extended descriptors are supported */
+	if (device->cap & IOAT_CAP_RAID16SS) {
+		char pool_name[14];
+		int i;
 
-	#ifdef CONFIG_ASYNC_TX_DISABLE_PQ_VAL_DMA
-	dma_cap_clear(DMA_PQ_VAL, dma->cap_mask);
-	dma->device_prep_dma_pq_val = NULL;
-	#endif
+		for (i = 0; i < MAX_SED_POOLS; i++) {
+			snprintf(pool_name, 14, "ioat_hw%d_sed", i);
 
-	#ifdef CONFIG_ASYNC_TX_DISABLE_XOR_VAL_DMA
-	dma_cap_clear(DMA_XOR_VAL, dma->cap_mask);
-	dma->device_prep_dma_xor_val = NULL;
-	#endif
+			/* allocate SED DMA pool */
+			device->sed_hw_pool[i] = dmam_pool_create(pool_name,
+					&pdev->dev,
+					SED_SIZE * (i + 1), 64, 0);
+			if (!device->sed_hw_pool[i])
+				return -ENOMEM;
+
+		}
+	}
 
 	err = ioat_probe(device);
 	if (err)