Merge branch 'topic/err_reporting' into for-linus

Signed-off-by: Vinod Koul <vinod.koul@intel.com>

Conflicts:
	drivers/dma/cppi41.c

41 files changed, 596 insertions, 285 deletions

diff --git a/Documentation/dmaengine/provider.txt b/Documentation/dmaengine/provider.txt
index 91ce82d5f0c4..c4fd47540b31 100644
--- a/Documentation/dmaengine/provider.txt
+++ b/Documentation/dmaengine/provider.txt
@@ -282,6 +282,17 @@ supported.
 			that is supposed to push the current
 			transaction descriptor to a pending queue, waiting
 			for issue_pending to be called.
+     - In this structure the function pointer callback_result can be
+       initialized in order for the submitter to be notified that a
+       transaction has completed. In the earlier code the function pointer
+       callback has been used. However it does not provide any status to the
+       transaction and will be deprecated. The result structure defined as
+       dmaengine_result that is passed in to callback_result has two fields:
+       + result: This provides the transfer result defined by
+		 dmaengine_tx_result. Either success or some error
+		 condition.
+       + residue: Provides the residue bytes of the transfer for those that
+		  support residue.
 
    * device_issue_pending
      - Takes the first transaction descriptor in the pending queue,
diff --git a/drivers/dma/at_hdmac.c b/drivers/dma/at_hdmac.c
index 53d22eb73b56..a4c8f80db29d 100644
--- a/drivers/dma/at_hdmac.c
+++ b/drivers/dma/at_hdmac.c
@@ -473,15 +473,11 @@ atc_chain_complete(struct at_dma_chan *atchan, struct at_desc *desc)
 	/* for cyclic transfers,
 	 * no need to replay callback function while stopping */
 	if (!atc_chan_is_cyclic(atchan)) {
-		dma_async_tx_callback	callback = txd->callback;
-		void			*param = txd->callback_param;
-
 		/*
 		 * The API requires that no submissions are done from a
 		 * callback, so we don't need to drop the lock here
 		 */
-		if (callback)
-			callback(param);
+		dmaengine_desc_get_callback_invoke(txd, NULL);
 	}
 
 	dma_run_dependencies(txd);
@@ -598,15 +594,12 @@ static void atc_handle_cyclic(struct at_dma_chan *atchan)
 {
 	struct at_desc			*first = atc_first_active(atchan);
 	struct dma_async_tx_descriptor	*txd = &first->txd;
-	dma_async_tx_callback		callback = txd->callback;
-	void				*param = txd->callback_param;
 
 	dev_vdbg(chan2dev(&atchan->chan_common),
 			"new cyclic period llp 0x%08x\n",
 			channel_readl(atchan, DSCR));
 
-	if (callback)
-		callback(param);
+	dmaengine_desc_get_callback_invoke(txd, NULL);
 }
 
 /*--  IRQ & Tasklet  ---------------------------------------------------*/
diff --git a/drivers/dma/at_xdmac.c b/drivers/dma/at_xdmac.c
index e434ffe7bc5c..2badc57a7f31 100644
--- a/drivers/dma/at_xdmac.c
+++ b/drivers/dma/at_xdmac.c
@@ -1572,8 +1572,8 @@ static void at_xdmac_handle_cyclic(struct at_xdmac_chan *atchan)
 	desc = list_first_entry(&atchan->xfers_list, struct at_xdmac_desc, xfer_node);
 	txd = &desc->tx_dma_desc;
 
-	if (txd->callback && (txd->flags & DMA_PREP_INTERRUPT))
-		txd->callback(txd->callback_param);
+	if (txd->flags & DMA_PREP_INTERRUPT)
+		dmaengine_desc_get_callback_invoke(txd, NULL);
 }
 
 static void at_xdmac_tasklet(unsigned long data)
@@ -1616,8 +1616,8 @@ static void at_xdmac_tasklet(unsigned long data)
 
 		if (!at_xdmac_chan_is_cyclic(atchan)) {
 			dma_cookie_complete(txd);
-			if (txd->callback && (txd->flags & DMA_PREP_INTERRUPT))
-				txd->callback(txd->callback_param);
+			if (txd->flags & DMA_PREP_INTERRUPT)
+				dmaengine_desc_get_callback_invoke(txd, NULL);
 		}
 
 		dma_run_dependencies(txd);
diff --git a/drivers/dma/coh901318.c b/drivers/dma/coh901318.c
index 472be1d09586..74794c9859f6 100644
--- a/drivers/dma/coh901318.c
+++ b/drivers/dma/coh901318.c
@@ -1875,8 +1875,7 @@ static void dma_tasklet(unsigned long data)
 	struct coh901318_chan *cohc = (struct coh901318_chan *) data;
 	struct coh901318_desc *cohd_fin;
 	unsigned long flags;
-	dma_async_tx_callback callback;
-	void *callback_param;
+	struct dmaengine_desc_callback cb;
 
 	dev_vdbg(COHC_2_DEV(cohc), "[%s] chan_id %d"
 		 " nbr_active_done %ld\n", __func__,
@@ -1891,8 +1890,7 @@ static void dma_tasklet(unsigned long data)
 		goto err;
 
 	/* locate callback to client */
-	callback = cohd_fin->desc.callback;
-	callback_param = cohd_fin->desc.callback_param;
+	dmaengine_desc_get_callback(&cohd_fin->desc, &cb);
 
 	/* sign this job as completed on the channel */
 	dma_cookie_complete(&cohd_fin->desc);
@@ -1907,8 +1905,7 @@ static void dma_tasklet(unsigned long data)
 	spin_unlock_irqrestore(&cohc->lock, flags);
 
 	/* Call the callback when we're done */
-	if (callback)
-		callback(callback_param);
+	dmaengine_desc_callback_invoke(&cb, NULL);
 
 	spin_lock_irqsave(&cohc->lock, flags);
 
diff --git a/drivers/dma/cppi41.c b/drivers/dma/cppi41.c
index 97f4d6c1b6b9..bac5f023013b 100644
--- a/drivers/dma/cppi41.c
+++ b/drivers/dma/cppi41.c
@@ -336,7 +336,7 @@ static irqreturn_t cppi41_irq(int irq, void *data)
 
 			c->residue = pd_trans_len(c->desc->pd6) - len;
 			dma_cookie_complete(&c->txd);
-			c->txd.callback(c->txd.callback_param);
+			dmaengine_desc_get_callback_invoke(&c->txd, NULL);
 
 			/* Paired with cppi41_dma_issue_pending */
 			pm_runtime_mark_last_busy(cdd->ddev.dev);
diff --git a/drivers/dma/dmaengine.h b/drivers/dma/dmaengine.h
index 17f983a4e9ba..882ff9448c3b 100644
--- a/drivers/dma/dmaengine.h
+++ b/drivers/dma/dmaengine.h
@@ -86,4 +86,88 @@ static inline void dma_set_residue(struct dma_tx_state *state, u32 residue)
 		state->residue = residue;
 }
 
+struct dmaengine_desc_callback {
+	dma_async_tx_callback callback;
+	dma_async_tx_callback_result callback_result;
+	void *callback_param;
+};
+
+/**
+ * dmaengine_desc_get_callback - get the passed in callback function
+ * @tx: tx descriptor
+ * @cb: temp struct to hold the callback info
+ *
+ * Fill the passed in cb struct with what's available in the passed in
+ * tx descriptor struct
+ * No locking is required.
+ */
+static inline void
+dmaengine_desc_get_callback(struct dma_async_tx_descriptor *tx,
+			    struct dmaengine_desc_callback *cb)
+{
+	cb->callback = tx->callback;
+	cb->callback_result = tx->callback_result;
+	cb->callback_param = tx->callback_param;
+}
+
+/**
+ * dmaengine_desc_callback_invoke - call the callback function in cb struct
+ * @cb: temp struct that is holding the callback info
+ * @result: transaction result
+ *
+ * Call the callback function provided in the cb struct with the parameter
+ * in the cb struct.
+ * Locking is dependent on the driver.
+ */
+static inline void
+dmaengine_desc_callback_invoke(struct dmaengine_desc_callback *cb,
+			       const struct dmaengine_result *result)
+{
+	struct dmaengine_result dummy_result = {
+		.result = DMA_TRANS_NOERROR,
+		.residue = 0
+	};
+
+	if (cb->callback_result) {
+		if (!result)
+			result = &dummy_result;
+		cb->callback_result(cb->callback_param, result);
+	} else if (cb->callback) {
+		cb->callback(cb->callback_param);
+	}
+}
+
+/**
+ * dmaengine_desc_get_callback_invoke - get the callback in tx descriptor and
+ * 					then immediately call the callback.
+ * @tx: dma async tx descriptor
+ * @result: transaction result
+ *
+ * Call dmaengine_desc_get_callback() and dmaengine_desc_callback_invoke()
+ * in a single function since no work is necessary in between for the driver.
+ * Locking is dependent on the driver.
+ */
+static inline void
+dmaengine_desc_get_callback_invoke(struct dma_async_tx_descriptor *tx,
+				   const struct dmaengine_result *result)
+{
+	struct dmaengine_desc_callback cb;
+
+	dmaengine_desc_get_callback(tx, &cb);
+	dmaengine_desc_callback_invoke(&cb, result);
+}
+
+/**
+ * dmaengine_desc_callback_valid - verify the callback is valid in cb
+ * @cb: callback info struct
+ *
+ * Return a bool that verifies whether callback in cb is valid or not.
+ * No locking is required.
+ */
+static inline bool
+dmaengine_desc_callback_valid(struct dmaengine_desc_callback *cb)
+{
+	return (cb->callback) ? true : false;
+}
+
 #endif
diff --git a/drivers/dma/dw/core.c b/drivers/dma/dw/core.c
index edf053f73a49..12eedd457193 100644
--- a/drivers/dma/dw/core.c
+++ b/drivers/dma/dw/core.c
@@ -270,20 +270,19 @@ static void
 dwc_descriptor_complete(struct dw_dma_chan *dwc, struct dw_desc *desc,
 		bool callback_required)
 {
-	dma_async_tx_callback		callback = NULL;
-	void				*param = NULL;
 	struct dma_async_tx_descriptor	*txd = &desc->txd;
 	struct dw_desc			*child;
 	unsigned long			flags;
+	struct dmaengine_desc_callback	cb;
 
 	dev_vdbg(chan2dev(&dwc->chan), "descriptor %u complete\n", txd->cookie);
 
 	spin_lock_irqsave(&dwc->lock, flags);
 	dma_cookie_complete(txd);
-	if (callback_required) {
-		callback = txd->callback;
-		param = txd->callback_param;
-	}
+	if (callback_required)
+		dmaengine_desc_get_callback(txd, &cb);
+	else
+		memset(&cb, 0, sizeof(cb));
 
 	/* async_tx_ack */
 	list_for_each_entry(child, &desc->tx_list, desc_node)
@@ -292,8 +291,7 @@ dwc_descriptor_complete(struct dw_dma_chan *dwc, struct dw_desc *desc,
 	dwc_desc_put(dwc, desc);
 	spin_unlock_irqrestore(&dwc->lock, flags);
 
-	if (callback)
-		callback(param);
+	dmaengine_desc_callback_invoke(&cb, NULL);
 }
 
 static void dwc_complete_all(struct dw_dma *dw, struct dw_dma_chan *dwc)
diff --git a/drivers/dma/ep93xx_dma.c b/drivers/dma/ep93xx_dma.c
index ca17e8751af2..d37e8dda8079 100644
--- a/drivers/dma/ep93xx_dma.c
+++ b/drivers/dma/ep93xx_dma.c
@@ -737,10 +737,10 @@ static void ep93xx_dma_tasklet(unsigned long data)
 {
 	struct ep93xx_dma_chan *edmac = (struct ep93xx_dma_chan *)data;
 	struct ep93xx_dma_desc *desc, *d;
-	dma_async_tx_callback callback = NULL;
-	void *callback_param = NULL;
+	struct dmaengine_desc_callback cb;
 	LIST_HEAD(list);
 
+	memset(&cb, 0, sizeof(cb));
 	spin_lock_irq(&edmac->lock);
 	/*
 	 * If dma_terminate_all() was called before we get to run, the active
@@ -755,8 +755,7 @@ static void ep93xx_dma_tasklet(unsigned long data)
 				dma_cookie_complete(&desc->txd);
 			list_splice_init(&edmac->active, &list);
 		}
-		callback = desc->txd.callback;
-		callback_param = desc->txd.callback_param;
+		dmaengine_desc_get_callback(&desc->txd, &cb);
 	}
 	spin_unlock_irq(&edmac->lock);
 
@@ -769,8 +768,7 @@ static void ep93xx_dma_tasklet(unsigned long data)
 		ep93xx_dma_desc_put(edmac, desc);
 	}
 
-	if (callback)
-		callback(callback_param);
+	dmaengine_desc_callback_invoke(&cb, NULL);
 }
 
 static irqreturn_t ep93xx_dma_interrupt(int irq, void *dev_id)
diff --git a/drivers/dma/fsl_raid.c b/drivers/dma/fsl_raid.c
index 496ff8e7d7f9..40c58ae80660 100644
--- a/drivers/dma/fsl_raid.c
+++ b/drivers/dma/fsl_raid.c
@@ -134,16 +134,8 @@ static void fsl_re_issue_pending(struct dma_chan *chan)
 
 static void fsl_re_desc_done(struct fsl_re_desc *desc)
 {
-	dma_async_tx_callback callback;
-	void *callback_param;
-
 	dma_cookie_complete(&desc->async_tx);
-
-	callback = desc->async_tx.callback;
-	callback_param = desc->async_tx.callback_param;
-	if (callback)
-		callback(callback_param);
-
+	dmaengine_desc_get_callback_invoke(&desc->async_tx, NULL);
 	dma_descriptor_unmap(&desc->async_tx);
 }
 
diff --git a/drivers/dma/fsldma.c b/drivers/dma/fsldma.c
index 6ccb787ba56d..87f6ab222d8c 100644
--- a/drivers/dma/fsldma.c
+++ b/drivers/dma/fsldma.c
@@ -517,11 +517,7 @@ static dma_cookie_t fsldma_run_tx_complete_actions(struct fsldma_chan *chan,
 		ret = txd->cookie;
 
 		/* Run the link descriptor callback function */
-		if (txd->callback) {
-			chan_dbg(chan, "LD %p callback\n", desc);
-			txd->callback(txd->callback_param);
-		}
-
+		dmaengine_desc_get_callback_invoke(txd, NULL);
 		dma_descriptor_unmap(txd);
 	}
 
diff --git a/drivers/dma/imx-dma.c b/drivers/dma/imx-dma.c
index a960608c0a4d..ab0fb804fb1e 100644
--- a/drivers/dma/imx-dma.c
+++ b/drivers/dma/imx-dma.c
@@ -663,9 +663,7 @@ static void imxdma_tasklet(unsigned long data)
 out:
 	spin_unlock_irqrestore(&imxdma->lock, flags);
 
-	if (desc->desc.callback)
-		desc->desc.callback(desc->desc.callback_param);
-
+	dmaengine_desc_get_callback_invoke(&desc->desc, NULL);
 }
 
 static int imxdma_terminate_all(struct dma_chan *chan)
diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
index 03ec76fc22ff..624facb6c8f4 100644
--- a/drivers/dma/imx-sdma.c
+++ b/drivers/dma/imx-sdma.c
@@ -650,8 +650,7 @@ static void sdma_event_disable(struct sdma_channel *sdmac, unsigned int event)
 
 static void sdma_handle_channel_loop(struct sdma_channel *sdmac)
 {
-	if (sdmac->desc.callback)
-		sdmac->desc.callback(sdmac->desc.callback_param);
+	dmaengine_desc_get_callback_invoke(&sdmac->desc, NULL);
 }
 
 static void sdma_update_channel_loop(struct sdma_channel *sdmac)
@@ -701,8 +700,8 @@ static void mxc_sdma_handle_channel_normal(struct sdma_channel *sdmac)
 		sdmac->status = DMA_COMPLETE;
 
 	dma_cookie_complete(&sdmac->desc);
-	if (sdmac->desc.callback)
-		sdmac->desc.callback(sdmac->desc.callback_param);
+
+	dmaengine_desc_get_callback_invoke(&sdmac->desc, NULL);
 }
 
 static void sdma_tasklet(unsigned long data)
diff --git a/drivers/dma/ioat/dma.c b/drivers/dma/ioat/dma.c
index bd09961443b1..49386ce04bf5 100644
--- a/drivers/dma/ioat/dma.c
+++ b/drivers/dma/ioat/dma.c
@@ -38,8 +38,54 @@
 
 #include "../dmaengine.h"
 
+static char *chanerr_str[] = {
+	"DMA Transfer Destination Address Error",
+	"Next Descriptor Address Error",
+	"Descriptor Error",
+	"Chan Address Value Error",
+	"CHANCMD Error",
+	"Chipset Uncorrectable Data Integrity Error",
+	"DMA Uncorrectable Data Integrity Error",
+	"Read Data Error",
+	"Write Data Error",
+	"Descriptor Control Error",
+	"Descriptor Transfer Size Error",
+	"Completion Address Error",
+	"Interrupt Configuration Error",
+	"Super extended descriptor Address Error",
+	"Unaffiliated Error",
+	"CRC or XOR P Error",
+	"XOR Q Error",
+	"Descriptor Count Error",
+	"DIF All F detect Error",
+	"Guard Tag verification Error",
+	"Application Tag verification Error",
+	"Reference Tag verification Error",
+	"Bundle Bit Error",
+	"Result DIF All F detect Error",
+	"Result Guard Tag verification Error",
+	"Result Application Tag verification Error",
+	"Result Reference Tag verification Error",
+	NULL
+};
+
 static void ioat_eh(struct ioatdma_chan *ioat_chan);
 
+static void ioat_print_chanerrs(struct ioatdma_chan *ioat_chan, u32 chanerr)
+{
+	int i;
+
+	for (i = 0; i < 32; i++) {
+		if ((chanerr >> i) & 1) {
+			if (chanerr_str[i]) {
+				dev_err(to_dev(ioat_chan), "Err(%d): %s\n",
+					i, chanerr_str[i]);
+			} else
+				break;
+		}
+	}
+}
+
 /**
  * ioat_dma_do_interrupt - handler used for single vector interrupt mode
  * @irq: interrupt id
@@ -568,12 +614,14 @@ static void __cleanup(struct ioatdma_chan *ioat_chan, dma_addr_t phys_complete)
 
 		tx = &desc->txd;
 		if (tx->cookie) {
+			struct dmaengine_result res;
+
 			dma_cookie_complete(tx);
 			dma_descriptor_unmap(tx);
-			if (tx->callback) {
-				tx->callback(tx->callback_param);
-				tx->callback = NULL;
-			}
+			res.result = DMA_TRANS_NOERROR;
+			dmaengine_desc_get_callback_invoke(tx, NULL);
+			tx->callback = NULL;
+			tx->callback_result = NULL;
 		}
 
 		if (tx->phys == phys_complete)
@@ -622,7 +670,8 @@ static void ioat_cleanup(struct ioatdma_chan *ioat_chan)
 	if (is_ioat_halted(*ioat_chan->completion)) {
 		u32 chanerr = readl(ioat_chan->reg_base + IOAT_CHANERR_OFFSET);
 
-		if (chanerr & IOAT_CHANERR_HANDLE_MASK) {
+		if (chanerr &
+		    (IOAT_CHANERR_HANDLE_MASK | IOAT_CHANERR_RECOVER_MASK)) {
 			mod_timer(&ioat_chan->timer, jiffies + IDLE_TIMEOUT);
 			ioat_eh(ioat_chan);
 		}
@@ -652,6 +701,61 @@ static void ioat_restart_channel(struct ioatdma_chan *ioat_chan)
 	__ioat_restart_chan(ioat_chan);
 }
 
+
+static void ioat_abort_descs(struct ioatdma_chan *ioat_chan)
+{
+	struct ioatdma_device *ioat_dma = ioat_chan->ioat_dma;
+	struct ioat_ring_ent *desc;
+	u16 active;
+	int idx = ioat_chan->tail, i;
+
+	/*
+	 * We assume that the failed descriptor has been processed.
+	 * Now we are just returning all the remaining submitted
+	 * descriptors to abort.
+	 */
+	active = ioat_ring_active(ioat_chan);
+
+	/* we skip the failed descriptor that tail points to */
+	for (i = 1; i < active; i++) {
+		struct dma_async_tx_descriptor *tx;
+
+		smp_read_barrier_depends();
+		prefetch(ioat_get_ring_ent(ioat_chan, idx + i + 1));
+		desc = ioat_get_ring_ent(ioat_chan, idx + i);
+
+		tx = &desc->txd;
+		if (tx->cookie) {
+			struct dmaengine_result res;
+
+			dma_cookie_complete(tx);
+			dma_descriptor_unmap(tx);
+			res.result = DMA_TRANS_ABORTED;
+			dmaengine_desc_get_callback_invoke(tx, &res);
+			tx->callback = NULL;
+			tx->callback_result = NULL;
+		}
+
+		/* skip extended descriptors */
+		if (desc_has_ext(desc)) {
+			WARN_ON(i + 1 >= active);
+			i++;
+		}
+
+		/* cleanup super extended descriptors */
+		if (desc->sed) {
+			ioat_free_sed(ioat_dma, desc->sed);
+			desc->sed = NULL;
+		}
+	}
+
+	smp_mb(); /* finish all descriptor reads before incrementing tail */
+	ioat_chan->tail = idx + active;
+
+	desc = ioat_get_ring_ent(ioat_chan, ioat_chan->tail);
+	ioat_chan->last_completion = *ioat_chan->completion = desc->txd.phys;
+}
+
 static void ioat_eh(struct ioatdma_chan *ioat_chan)
 {
 	struct pci_dev *pdev = to_pdev(ioat_chan);
@@ -662,6 +766,8 @@ static void ioat_eh(struct ioatdma_chan *ioat_chan)
 	u32 err_handled = 0;
 	u32 chanerr_int;
 	u32 chanerr;
+	bool abort = false;
+	struct dmaengine_result res;
 
 	/* cleanup so tail points to descriptor that caused the error */
 	if (ioat_cleanup_preamble(ioat_chan, &phys_complete))
@@ -697,30 +803,55 @@ static void ioat_eh(struct ioatdma_chan *ioat_chan)
 		break;
 	}
 
+	if (chanerr & IOAT_CHANERR_RECOVER_MASK) {
+		if (chanerr & IOAT_CHANERR_READ_DATA_ERR) {
+			res.result = DMA_TRANS_READ_FAILED;
+			err_handled |= IOAT_CHANERR_READ_DATA_ERR;
+		} else if (chanerr & IOAT_CHANERR_WRITE_DATA_ERR) {
+			res.result = DMA_TRANS_WRITE_FAILED;
+			err_handled |= IOAT_CHANERR_WRITE_DATA_ERR;
+		}
+
+		abort = true;
+	} else
+		res.result = DMA_TRANS_NOERROR;
+
 	/* fault on unhandled error or spurious halt */
 	if (chanerr ^ err_handled || chanerr == 0) {
 		dev_err(to_dev(ioat_chan), "%s: fatal error (%x:%x)\n",
 			__func__, chanerr, err_handled);
+		dev_err(to_dev(ioat_chan), "Errors handled:\n");
+		ioat_print_chanerrs(ioat_chan, err_handled);
+		dev_err(to_dev(ioat_chan), "Errors not handled:\n");
+		ioat_print_chanerrs(ioat_chan, (chanerr & ~err_handled));
+
 		BUG();
-	} else { /* cleanup the faulty descriptor */
-		tx = &desc->txd;
-		if (tx->cookie) {
-			dma_cookie_complete(tx);
-			dma_descriptor_unmap(tx);
-			if (tx->callback) {
-				tx->callback(tx->callback_param);
-				tx->callback = NULL;
-			}
-		}
 	}
 
-	writel(chanerr, ioat_chan->reg_base + IOAT_CHANERR_OFFSET);
-	pci_write_config_dword(pdev, IOAT_PCI_CHANERR_INT_OFFSET, chanerr_int);
+	/* cleanup the faulty descriptor since we are continuing */
+	tx = &desc->txd;
+	if (tx->cookie) {
+		dma_cookie_complete(tx);
+		dma_descriptor_unmap(tx);
+		dmaengine_desc_get_callback_invoke(tx, &res);
+		tx->callback = NULL;
+		tx->callback_result = NULL;
+	}
 
 	/* mark faulting descriptor as complete */
 	*ioat_chan->completion = desc->txd.phys;
 
 	spin_lock_bh(&ioat_chan->prep_lock);
+	/* we need abort all descriptors */
+	if (abort) {
+		ioat_abort_descs(ioat_chan);
+		/* clean up the channel, we could be in weird state */
+		ioat_reset_hw(ioat_chan);
+	}
+
+	writel(chanerr, ioat_chan->reg_base + IOAT_CHANERR_OFFSET);
+	pci_write_config_dword(pdev, IOAT_PCI_CHANERR_INT_OFFSET, chanerr_int);
+
 	ioat_restart_channel(ioat_chan);
 	spin_unlock_bh(&ioat_chan->prep_lock);
 }
@@ -753,10 +884,28 @@ void ioat_timer_event(unsigned long data)
 		chanerr = readl(ioat_chan->reg_base + IOAT_CHANERR_OFFSET);
 		dev_err(to_dev(ioat_chan), "%s: Channel halted (%x)\n",
 			__func__, chanerr);
-		if (test_bit(IOAT_RUN, &ioat_chan->state))
-			BUG_ON(is_ioat_bug(chanerr));
-		else /* we never got off the ground */
-			return;
+		dev_err(to_dev(ioat_chan), "Errors:\n");
+		ioat_print_chanerrs(ioat_chan, chanerr);
+
+		if (test_bit(IOAT_RUN, &ioat_chan->state)) {
+			spin_lock_bh(&ioat_chan->cleanup_lock);
+			spin_lock_bh(&ioat_chan->prep_lock);
+			set_bit(IOAT_CHAN_DOWN, &ioat_chan->state);
+			spin_unlock_bh(&ioat_chan->prep_lock);
+
+			ioat_abort_descs(ioat_chan);
+			dev_warn(to_dev(ioat_chan), "Reset channel...\n");
+			ioat_reset_hw(ioat_chan);
+			dev_warn(to_dev(ioat_chan), "Restart channel...\n");
+			ioat_restart_channel(ioat_chan);
+
+			spin_lock_bh(&ioat_chan->prep_lock);
+			clear_bit(IOAT_CHAN_DOWN, &ioat_chan->state);
+			spin_unlock_bh(&ioat_chan->prep_lock);
+			spin_unlock_bh(&ioat_chan->cleanup_lock);
+		}
+
+		return;
 	}
 
 	spin_lock_bh(&ioat_chan->cleanup_lock);
@@ -780,14 +929,26 @@ void ioat_timer_event(unsigned long data)
 		u32 chanerr;
 
 		chanerr = readl(ioat_chan->reg_base + IOAT_CHANERR_OFFSET);
-		dev_warn(to_dev(ioat_chan), "Restarting channel...\n");
-		dev_warn(to_dev(ioat_chan), "CHANSTS: %#Lx CHANERR: %#x\n",
-			 status, chanerr);
-		dev_warn(to_dev(ioat_chan), "Active descriptors: %d\n",
-			 ioat_ring_active(ioat_chan));
+		dev_err(to_dev(ioat_chan), "CHANSTS: %#Lx CHANERR: %#x\n",
+			status, chanerr);
+		dev_err(to_dev(ioat_chan), "Errors:\n");
+		ioat_print_chanerrs(ioat_chan, chanerr);
+
+		dev_dbg(to_dev(ioat_chan), "Active descriptors: %d\n",
+			ioat_ring_active(ioat_chan));
 
 		spin_lock_bh(&ioat_chan->prep_lock);
+		set_bit(IOAT_CHAN_DOWN, &ioat_chan->state);
+		spin_unlock_bh(&ioat_chan->prep_lock);
+
+		ioat_abort_descs(ioat_chan);
+		dev_warn(to_dev(ioat_chan), "Resetting channel...\n");
+		ioat_reset_hw(ioat_chan);
+		dev_warn(to_dev(ioat_chan), "Restarting channel...\n");
 		ioat_restart_channel(ioat_chan);
+
+		spin_lock_bh(&ioat_chan->prep_lock);
+		clear_bit(IOAT_CHAN_DOWN, &ioat_chan->state);
 		spin_unlock_bh(&ioat_chan->prep_lock);
 		spin_unlock_bh(&ioat_chan->cleanup_lock);
 		return;
diff --git a/drivers/dma/ioat/registers.h b/drivers/dma/ioat/registers.h
index 70534981a49b..48fa4cf9f64a 100644
--- a/drivers/dma/ioat/registers.h
+++ b/drivers/dma/ioat/registers.h
@@ -240,6 +240,8 @@
 #define IOAT_CHANERR_DESCRIPTOR_COUNT_ERR	0x40000
 
 #define IOAT_CHANERR_HANDLE_MASK (IOAT_CHANERR_XOR_P_OR_CRC_ERR | IOAT_CHANERR_XOR_Q_ERR)
+#define IOAT_CHANERR_RECOVER_MASK (IOAT_CHANERR_READ_DATA_ERR | \
+				   IOAT_CHANERR_WRITE_DATA_ERR)
 
 #define IOAT_CHANERR_MASK_OFFSET		0x2C	/* 32-bit Channel Error Register */
 
diff --git a/drivers/dma/iop-adma.c b/drivers/dma/iop-adma.c
index f039cfadf17b..a410657f7bcd 100644
--- a/drivers/dma/iop-adma.c
+++ b/drivers/dma/iop-adma.c
@@ -71,8 +71,7 @@ iop_adma_run_tx_complete_actions(struct iop_adma_desc_slot *desc,
 		/* call the callback (must not sleep or submit new
 		 * operations to this channel)
 		 */
-		if (tx->callback)
-			tx->callback(tx->callback_param);
+		dmaengine_desc_get_callback_invoke(tx, NULL);
 
 		dma_descriptor_unmap(tx);
 		if (desc->group_head)
diff --git a/drivers/dma/ipu/ipu_idmac.c b/drivers/dma/ipu/ipu_idmac.c
index b54f62de9232..ed76044ce4b9 100644
--- a/drivers/dma/ipu/ipu_idmac.c
+++ b/drivers/dma/ipu/ipu_idmac.c
@@ -1160,11 +1160,10 @@ static irqreturn_t idmac_interrupt(int irq, void *dev_id)
 	struct scatterlist **sg, *sgnext, *sgnew = NULL;
 	/* Next transfer descriptor */
 	struct idmac_tx_desc *desc, *descnew;
-	dma_async_tx_callback callback;
-	void *callback_param;
 	bool done = false;
 	u32 ready0, ready1, curbuf, err;
 	unsigned long flags;
+	struct dmaengine_desc_callback cb;
 
 	/* IDMAC has cleared the respective BUFx_RDY bit, we manage the buffer */
 
@@ -1278,12 +1277,12 @@ static irqreturn_t idmac_interrupt(int irq, void *dev_id)
 
 	if (likely(sgnew) &&
 	    ipu_submit_buffer(ichan, descnew, sgnew, ichan->active_buffer) < 0) {
-		callback = descnew->txd.callback;
-		callback_param = descnew->txd.callback_param;
+		dmaengine_desc_get_callback(&descnew->txd, &cb);
+
 		list_del_init(&descnew->list);
 		spin_unlock(&ichan->lock);
-		if (callback)
-			callback(callback_param);
+
+		dmaengine_desc_callback_invoke(&cb, NULL);
 		spin_lock(&ichan->lock);
 	}
 
@@ -1292,13 +1291,12 @@ static irqreturn_t idmac_interrupt(int irq, void *dev_id)
 	if (done)
 		dma_cookie_complete(&desc->txd);
 
-	callback = desc->txd.callback;
-	callback_param = desc->txd.callback_param;
+	dmaengine_desc_get_callback(&desc->txd, &cb);
 
 	spin_unlock(&ichan->lock);
 
-	if (done && (desc->txd.flags & DMA_PREP_INTERRUPT) && callback)
-		callback(callback_param);
+	if (done && (desc->txd.flags & DMA_PREP_INTERRUPT))
+		dmaengine_desc_callback_invoke(&cb, NULL);
 
 	return IRQ_HANDLED;
 }
diff --git a/drivers/dma/mic_x100_dma.c b/drivers/dma/mic_x100_dma.c
index 1502b24b7c7d..818255844a3c 100644
--- a/drivers/dma/mic_x100_dma.c
+++ b/drivers/dma/mic_x100_dma.c
@@ -104,10 +104,8 @@ static void mic_dma_cleanup(struct mic_dma_chan *ch)
 		tx = &ch->tx_array[last_tail];
 		if (tx->cookie) {
 			dma_cookie_complete(tx);
-			if (tx->callback) {
-				tx->callback(tx->callback_param);
-				tx->callback = NULL;
-			}
+			dmaengine_desc_get_callback_invoke(tx, NULL);
+			tx->callback = NULL;
 		}
 		last_tail = mic_dma_hw_ring_inc(last_tail);
 	}
diff --git a/drivers/dma/mmp_pdma.c b/drivers/dma/mmp_pdma.c
index f4b25fb0d040..eb3a1f42ab06 100644
--- a/drivers/dma/mmp_pdma.c
+++ b/drivers/dma/mmp_pdma.c
@@ -864,19 +864,15 @@ static void dma_do_tasklet(unsigned long data)
 	struct mmp_pdma_desc_sw *desc, *_desc;
 	LIST_HEAD(chain_cleanup);
 	unsigned long flags;
+	struct dmaengine_desc_callback cb;
 
 	if (chan->cyclic_first) {
-		dma_async_tx_callback cb = NULL;
-		void *cb_data = NULL;
-
 		spin_lock_irqsave(&chan->desc_lock, flags);
 		desc = chan->cyclic_first;
-		cb = desc->async_tx.callback;
-		cb_data = desc->async_tx.callback_param;
+		dmaengine_desc_get_callback(&desc->async_tx, &cb);
 		spin_unlock_irqrestore(&chan->desc_lock, flags);
 
-		if (cb)
-			cb(cb_data);
+		dmaengine_desc_callback_invoke(&cb, NULL);
 
 		return;
 	}
@@ -921,8 +917,8 @@ static void dma_do_tasklet(unsigned long data)
 		/* Remove from the list of transactions */
 		list_del(&desc->node);
 		/* Run the link descriptor callback function */
-		if (txd->callback)
-			txd->callback(txd->callback_param);
+		dmaengine_desc_get_callback(txd, &cb);
+		dmaengine_desc_callback_invoke(&cb, NULL);
 
 		dma_pool_free(chan->desc_pool, desc, txd->phys);
 	}
diff --git a/drivers/dma/mmp_tdma.c b/drivers/dma/mmp_tdma.c
index d7422b1bf406..13c68b6434ce 100644
--- a/drivers/dma/mmp_tdma.c
+++ b/drivers/dma/mmp_tdma.c
@@ -349,9 +349,7 @@ static void dma_do_tasklet(unsigned long data)
 {
 	struct mmp_tdma_chan *tdmac = (struct mmp_tdma_chan *)data;
 
-	if (tdmac->desc.callback)
-		tdmac->desc.callback(tdmac->desc.callback_param);
-
+	dmaengine_desc_get_callback_invoke(&tdmac->desc, NULL);
 }
 
 static void mmp_tdma_free_descriptor(struct mmp_tdma_chan *tdmac)
diff --git a/drivers/dma/mpc512x_dma.c b/drivers/dma/mpc512x_dma.c
index 9dd99ba18fce..dde713461a95 100644
--- a/drivers/dma/mpc512x_dma.c
+++ b/drivers/dma/mpc512x_dma.c
@@ -411,8 +411,7 @@ static void mpc_dma_process_completed(struct mpc_dma *mdma)
 		list_for_each_entry(mdesc, &list, node) {
 			desc = &mdesc->desc;
 
-			if (desc->callback)
-				desc->callback(desc->callback_param);
+			dmaengine_desc_get_callback_invoke(desc, NULL);
 
 			last_cookie = desc->cookie;
 			dma_run_dependencies(desc);
diff --git a/drivers/dma/mv_xor.c b/drivers/dma/mv_xor.c
index f4c9f98ec35e..f8b5e7424b3a 100644
--- a/drivers/dma/mv_xor.c
+++ b/drivers/dma/mv_xor.c
@@ -209,10 +209,7 @@ mv_desc_run_tx_complete_actions(struct mv_xor_desc_slot *desc,
 		/* call the callback (must not sleep or submit new
 		 * operations to this channel)
 		 */
-		if (desc->async_tx.callback)
-			desc->async_tx.callback(
-				desc->async_tx.callback_param);
-
+		dmaengine_desc_get_callback_invoke(&desc->async_tx, NULL);
 		dma_descriptor_unmap(&desc->async_tx);
 	}
 
diff --git a/drivers/dma/mxs-dma.c b/drivers/dma/mxs-dma.c
index 60de35251da5..50e64e113ffb 100644
--- a/drivers/dma/mxs-dma.c
+++ b/drivers/dma/mxs-dma.c
@@ -326,8 +326,7 @@ static void mxs_dma_tasklet(unsigned long data)
 {
 	struct mxs_dma_chan *mxs_chan = (struct mxs_dma_chan *) data;
 
-	if (mxs_chan->desc.callback)
-		mxs_chan->desc.callback(mxs_chan->desc.callback_param);
+	dmaengine_desc_get_callback_invoke(&mxs_chan->desc, NULL);
 }
 
 static int mxs_dma_irq_to_chan(struct mxs_dma_engine *mxs_dma, int irq)
diff --git a/drivers/dma/nbpfaxi.c b/drivers/dma/nbpfaxi.c
index 08c45c185549..09de71519d37 100644
--- a/drivers/dma/nbpfaxi.c
+++ b/drivers/dma/nbpfaxi.c
@@ -1102,8 +1102,7 @@ static void nbpf_chan_tasklet(unsigned long data)
 {
 	struct nbpf_channel *chan = (struct nbpf_channel *)data;
 	struct nbpf_desc *desc, *tmp;
-	dma_async_tx_callback callback;
-	void *param;
+	struct dmaengine_desc_callback cb;
 
 	while (!list_empty(&chan->done)) {
 		bool found = false, must_put, recycling = false;
@@ -1151,14 +1150,12 @@ static void nbpf_chan_tasklet(unsigned long data)
 			must_put = false;
 		}
 
-		callback = desc->async_tx.callback;
-		param = desc->async_tx.callback_param;
+		dmaengine_desc_get_callback(&desc->async_tx, &cb);
 
 		/* ack and callback completed descriptor */
 		spin_unlock_irq(&chan->lock);
 
-		if (callback)
-			callback(param);
+		dmaengine_desc_callback_invoke(&cb, NULL);
 
 		if (must_put)
 			nbpf_desc_put(desc);
diff --git a/drivers/dma/pch_dma.c b/drivers/dma/pch_dma.c
index 113605f6fe20..df95727dc2fb 100644
--- a/drivers/dma/pch_dma.c
+++ b/drivers/dma/pch_dma.c
@@ -357,14 +357,13 @@ static void pdc_chain_complete(struct pch_dma_chan *pd_chan,
 			       struct pch_dma_desc *desc)
 {
 	struct dma_async_tx_descriptor *txd = &desc->txd;
-	dma_async_tx_callback callback = txd->callback;
-	void *param = txd->callback_param;
+	struct dmaengine_desc_callback cb;
 
+	dmaengine_desc_get_callback(txd, &cb);
 	list_splice_init(&desc->tx_list, &pd_chan->free_list);
 	list_move(&desc->desc_node, &pd_chan->free_list);
 
-	if (callback)
-		callback(param);
+	dmaengine_desc_callback_invoke(&cb, NULL);
 }
 
 static void pdc_complete_all(struct pch_dma_chan *pd_chan)
diff --git a/drivers/dma/pl330.c b/drivers/dma/pl330.c
index 4fc3ffbd5ca0..1ecd4674aa23 100644
--- a/drivers/dma/pl330.c
+++ b/drivers/dma/pl330.c
@@ -2039,14 +2039,12 @@ static void pl330_tasklet(unsigned long data)
 	}
 
 	while (!list_empty(&pch->completed_list)) {
-		dma_async_tx_callback callback;
-		void *callback_param;
+		struct dmaengine_desc_callback cb;
 
 		desc = list_first_entry(&pch->completed_list,
 					struct dma_pl330_desc, node);
 
-		callback = desc->txd.callback;
-		callback_param = desc->txd.callback_param;
+		dmaengine_desc_get_callback(&desc->txd, &cb);
 
 		if (pch->cyclic) {
 			desc->status = PREP;
@@ -2064,9 +2062,9 @@ static void pl330_tasklet(unsigned long data)
 
 		dma_descriptor_unmap(&desc->txd);
 
-		if (callback) {
+		if (dmaengine_desc_callback_valid(&cb)) {
 			spin_unlock_irqrestore(&pch->lock, flags);
-			callback(callback_param);
+			dmaengine_desc_callback_invoke(&cb, NULL);
 			spin_lock_irqsave(&pch->lock, flags);
 		}
 	}
diff --git a/drivers/dma/ppc4xx/adma.c b/drivers/dma/ppc4xx/adma.c
index 66bd96724b2f..d45da34a0568 100644
--- a/drivers/dma/ppc4xx/adma.c
+++ b/drivers/dma/ppc4xx/adma.c
@@ -1485,10 +1485,7 @@ static dma_cookie_t ppc440spe_adma_run_tx_complete_actions(
 		/* call the callback (must not sleep or submit new
 		 * operations to this channel)
 		 */
-		if (desc->async_tx.callback)
-			desc->async_tx.callback(
-				desc->async_tx.callback_param);
-
+		dmaengine_desc_get_callback_invoke(&desc->async_tx, NULL);
 		dma_descriptor_unmap(&desc->async_tx);
 	}
 
diff --git a/drivers/dma/qcom/hidma.c b/drivers/dma/qcom/hidma.c
index b2374cd91e45..e244e10a94b5 100644
--- a/drivers/dma/qcom/hidma.c
+++ b/drivers/dma/qcom/hidma.c
@@ -111,6 +111,7 @@ static void hidma_process_completed(struct hidma_chan *mchan)
 	struct dma_async_tx_descriptor *desc;
 	dma_cookie_t last_cookie;
 	struct hidma_desc *mdesc;
+	struct hidma_desc *next;
 	unsigned long irqflags;
 	struct list_head list;
 
@@ -122,28 +123,36 @@ static void hidma_process_completed(struct hidma_chan *mchan)
 	spin_unlock_irqrestore(&mchan->lock, irqflags);
 
 	/* Execute callbacks and run dependencies */
-	list_for_each_entry(mdesc, &list, node) {
+	list_for_each_entry_safe(mdesc, next, &list, node) {
 		enum dma_status llstat;
+		struct dmaengine_desc_callback cb;
+		struct dmaengine_result result;
 
 		desc = &mdesc->desc;
+		last_cookie = desc->cookie;
 
 		spin_lock_irqsave(&mchan->lock, irqflags);
 		dma_cookie_complete(desc);
 		spin_unlock_irqrestore(&mchan->lock, irqflags);
 
 		llstat = hidma_ll_status(mdma->lldev, mdesc->tre_ch);
-		if (desc->callback && (llstat == DMA_COMPLETE))
-			desc->callback(desc->callback_param);
+		dmaengine_desc_get_callback(desc, &cb);
 
-		last_cookie = desc->cookie;
 		dma_run_dependencies(desc);
-	}
 
-	/* Free descriptors */
-	spin_lock_irqsave(&mchan->lock, irqflags);
-	list_splice_tail_init(&list, &mchan->free);
-	spin_unlock_irqrestore(&mchan->lock, irqflags);
+		spin_lock_irqsave(&mchan->lock, irqflags);
+		list_move(&mdesc->node, &mchan->free);
+
+		if (llstat == DMA_COMPLETE) {
+			mchan->last_success = last_cookie;
+			result.result = DMA_TRANS_NOERROR;
+		} else
+			result.result = DMA_TRANS_ABORTED;
+
+		spin_unlock_irqrestore(&mchan->lock, irqflags);
 
+		dmaengine_desc_callback_invoke(&cb, &result);
+	}
 }
 
 /*
@@ -238,6 +247,19 @@ static void hidma_issue_pending(struct dma_chan *dmach)
 		hidma_ll_start(dmadev->lldev);
 }
 
+static inline bool hidma_txn_is_success(dma_cookie_t cookie,
+		dma_cookie_t last_success, dma_cookie_t last_used)
+{
+	if (last_success <= last_used) {
+		if ((cookie <= last_success) || (cookie > last_used))
+			return true;
+	} else {
+		if ((cookie <= last_success) && (cookie > last_used))
+			return true;
+	}
+	return false;
+}
+
 static enum dma_status hidma_tx_status(struct dma_chan *dmach,
 				       dma_cookie_t cookie,
 				       struct dma_tx_state *txstate)
@@ -246,8 +268,13 @@ static enum dma_status hidma_tx_status(struct dma_chan *dmach,
 	enum dma_status ret;
 
 	ret = dma_cookie_status(dmach, cookie, txstate);
-	if (ret == DMA_COMPLETE)
-		return ret;
+	if (ret == DMA_COMPLETE) {
+		bool is_success;
+
+		is_success = hidma_txn_is_success(cookie, mchan->last_success,
+						  dmach->cookie);
+		return is_success ? ret : DMA_ERROR;
+	}
 
 	if (mchan->paused && (ret == DMA_IN_PROGRESS)) {
 		unsigned long flags;
@@ -398,6 +425,7 @@ static int hidma_terminate_channel(struct dma_chan *chan)
 	hidma_process_completed(mchan);
 
 	spin_lock_irqsave(&mchan->lock, irqflags);
+	mchan->last_success = 0;
 	list_splice_init(&mchan->active, &list);
 	list_splice_init(&mchan->prepared, &list);
 	list_splice_init(&mchan->completed, &list);
@@ -413,14 +441,9 @@ static int hidma_terminate_channel(struct dma_chan *chan)
 	/* return all user requests */
 	list_for_each_entry_safe(mdesc, tmp, &list, node) {
 		struct dma_async_tx_descriptor *txd = &mdesc->desc;
-		dma_async_tx_callback callback = mdesc->desc.callback;
-		void *param = mdesc->desc.callback_param;
 
 		dma_descriptor_unmap(txd);
-
-		if (callback)
-			callback(param);
-
+		dmaengine_desc_get_callback_invoke(txd, NULL);
 		dma_run_dependencies(txd);
 
 		/* move myself to free_list */
diff --git a/drivers/dma/qcom/hidma.h b/drivers/dma/qcom/hidma.h
index db413a5efc4e..e52e20716303 100644
--- a/drivers/dma/qcom/hidma.h
+++ b/drivers/dma/qcom/hidma.h
@@ -72,7 +72,6 @@ struct hidma_lldev {
 
 	u32 tre_write_offset;           /* TRE write location              */
 	struct tasklet_struct task;	/* task delivering notifications   */
-	struct tasklet_struct rst_task;	/* task to reset HW                */
 	DECLARE_KFIFO_PTR(handoff_fifo,
 		struct hidma_tre *);    /* pending TREs FIFO               */
 };
@@ -89,6 +88,7 @@ struct hidma_chan {
 	bool				allocated;
 	char				dbg_name[16];
 	u32				dma_sig;
+	dma_cookie_t			last_success;
 
 	/*
 	 * active descriptor on this channel
diff --git a/drivers/dma/qcom/hidma_ll.c b/drivers/dma/qcom/hidma_ll.c
index ad20dfb64c71..3224f24c577b 100644
--- a/drivers/dma/qcom/hidma_ll.c
+++ b/drivers/dma/qcom/hidma_ll.c
@@ -381,27 +381,6 @@ static int hidma_ll_reset(struct hidma_lldev *lldev)
 }
 
 /*
- * Abort all transactions and perform a reset.
- */
-static void hidma_ll_abort(unsigned long arg)
-{
-	struct hidma_lldev *lldev = (struct hidma_lldev *)arg;
-	u8 err_code = HIDMA_EVRE_STATUS_ERROR;
-	u8 err_info = 0xFF;
-	int rc;
-
-	hidma_cleanup_pending_tre(lldev, err_info, err_code);
-
-	/* reset the channel for recovery */
-	rc = hidma_ll_setup(lldev);
-	if (rc) {
-		dev_err(lldev->dev, "channel reinitialize failed after error\n");
-		return;
-	}
-	writel(ENABLE_IRQS, lldev->evca + HIDMA_EVCA_IRQ_EN_REG);
-}
-
-/*
  * The interrupt handler for HIDMA will try to consume as many pending
  * EVRE from the event queue as possible. Each EVRE has an associated
  * TRE that holds the user interface parameters. EVRE reports the
@@ -454,13 +433,18 @@ irqreturn_t hidma_ll_inthandler(int chirq, void *arg)
 
 	while (cause) {
 		if (cause & HIDMA_ERR_INT_MASK) {
-			dev_err(lldev->dev, "error 0x%x, resetting...\n",
+			dev_err(lldev->dev, "error 0x%x, disabling...\n",
 					cause);
 
 			/* Clear out pending interrupts */
 			writel(cause, lldev->evca + HIDMA_EVCA_IRQ_CLR_REG);
 
-			tasklet_schedule(&lldev->rst_task);
+			/* No further submissions. */
+			hidma_ll_disable(lldev);
+
+			/* Driver completes the txn and intimates the client.*/
+			hidma_cleanup_pending_tre(lldev, 0xFF,
+						  HIDMA_EVRE_STATUS_ERROR);
 			goto out;
 		}
 
@@ -808,7 +792,6 @@ struct hidma_lldev *hidma_ll_init(struct device *dev, u32 nr_tres,
 		return NULL;
 
 	spin_lock_init(&lldev->lock);
-	tasklet_init(&lldev->rst_task, hidma_ll_abort, (unsigned long)lldev);
 	tasklet_init(&lldev->task, hidma_ll_tre_complete, (unsigned long)lldev);
 	lldev->initialized = 1;
 	writel(ENABLE_IRQS, lldev->evca + HIDMA_EVCA_IRQ_EN_REG);
@@ -831,7 +814,6 @@ int hidma_ll_uninit(struct hidma_lldev *lldev)
 
 	required_bytes = sizeof(struct hidma_tre) * lldev->nr_tres;
 	tasklet_kill(&lldev->task);
-	tasklet_kill(&lldev->rst_task);
 	memset(lldev->trepool, 0, required_bytes);
 	lldev->trepool = NULL;
 	lldev->pending_tre_count = 0;
diff --git a/drivers/dma/sh/rcar-dmac.c b/drivers/dma/sh/rcar-dmac.c
index 0dd953884d1d..d1defa4646ba 100644
--- a/drivers/dma/sh/rcar-dmac.c
+++ b/drivers/dma/sh/rcar-dmac.c
@@ -1389,21 +1389,18 @@ static irqreturn_t rcar_dmac_isr_channel_thread(int irq, void *dev)
 {
 	struct rcar_dmac_chan *chan = dev;
 	struct rcar_dmac_desc *desc;
+	struct dmaengine_desc_callback cb;
 
 	spin_lock_irq(&chan->lock);
 
 	/* For cyclic transfers notify the user after every chunk. */
 	if (chan->desc.running && chan->desc.running->cyclic) {
-		dma_async_tx_callback callback;
-		void *callback_param;
-
 		desc = chan->desc.running;
-		callback = desc->async_tx.callback;
-		callback_param = desc->async_tx.callback_param;
+		dmaengine_desc_get_callback(&desc->async_tx, &cb);
 
-		if (callback) {
+		if (dmaengine_desc_callback_valid(&cb)) {
 			spin_unlock_irq(&chan->lock);
-			callback(callback_param);
+			dmaengine_desc_callback_invoke(&cb, NULL);
 			spin_lock_irq(&chan->lock);
 		}
 	}
@@ -1418,14 +1415,15 @@ static irqreturn_t rcar_dmac_isr_channel_thread(int irq, void *dev)
 		dma_cookie_complete(&desc->async_tx);
 		list_del(&desc->node);
 
-		if (desc->async_tx.callback) {
+		dmaengine_desc_get_callback(&desc->async_tx, &cb);
+		if (dmaengine_desc_callback_valid(&cb)) {
 			spin_unlock_irq(&chan->lock);
 			/*
 			 * We own the only reference to this descriptor, we can
 			 * safely dereference it without holding the channel
 			 * lock.
 			 */
-			desc->async_tx.callback(desc->async_tx.callback_param);
+			dmaengine_desc_callback_invoke(&cb, NULL);
 			spin_lock_irq(&chan->lock);
 		}
 
diff --git a/drivers/dma/sh/shdma-base.c b/drivers/dma/sh/shdma-base.c
index 10fcabad80f3..12fa48e380cf 100644
--- a/drivers/dma/sh/shdma-base.c
+++ b/drivers/dma/sh/shdma-base.c
@@ -330,10 +330,11 @@ static dma_async_tx_callback __ld_cleanup(struct shdma_chan *schan, bool all)
 	bool head_acked = false;
 	dma_cookie_t cookie = 0;
 	dma_async_tx_callback callback = NULL;
-	void *param = NULL;
+	struct dmaengine_desc_callback cb;
 	unsigned long flags;
 	LIST_HEAD(cyclic_list);
 
+	memset(&cb, 0, sizeof(cb));
 	spin_lock_irqsave(&schan->chan_lock, flags);
 	list_for_each_entry_safe(desc, _desc, &schan->ld_queue, node) {
 		struct dma_async_tx_descriptor *tx = &desc->async_tx;
@@ -367,8 +368,8 @@ static dma_async_tx_callback __ld_cleanup(struct shdma_chan *schan, bool all)
 		/* Call callback on the last chunk */
 		if (desc->mark == DESC_COMPLETED && tx->callback) {
 			desc->mark = DESC_WAITING;
+			dmaengine_desc_get_callback(tx, &cb);
 			callback = tx->callback;
-			param = tx->callback_param;
 			dev_dbg(schan->dev, "descriptor #%d@%p on %d callback\n",
 				tx->cookie, tx, schan->id);
 			BUG_ON(desc->chunks != 1);
@@ -430,8 +431,7 @@ static dma_async_tx_callback __ld_cleanup(struct shdma_chan *schan, bool all)
 
 	spin_unlock_irqrestore(&schan->chan_lock, flags);
 
-	if (callback)
-		callback(param);
+	dmaengine_desc_callback_invoke(&cb, NULL);
 
 	return callback;
 }
@@ -885,9 +885,9 @@ bool shdma_reset(struct shdma_dev *sdev)
 		/* Complete all  */
 		list_for_each_entry(sdesc, &dl, node) {
 			struct dma_async_tx_descriptor *tx = &sdesc->async_tx;
+
 			sdesc->mark = DESC_IDLE;
-			if (tx->callback)
-				tx->callback(tx->callback_param);
+			dmaengine_desc_get_callback_invoke(tx, NULL);
 		}
 
 		spin_lock(&schan->chan_lock);
diff --git a/drivers/dma/sirf-dma.c b/drivers/dma/sirf-dma.c
index d8bc3f2a71db..a96e4a480de5 100644
--- a/drivers/dma/sirf-dma.c
+++ b/drivers/dma/sirf-dma.c
@@ -360,9 +360,7 @@ static void sirfsoc_dma_process_completed(struct sirfsoc_dma *sdma)
 			list_for_each_entry(sdesc, &list, node) {
 				desc = &sdesc->desc;
 
-				if (desc->callback)
-					desc->callback(desc->callback_param);
-
+				dmaengine_desc_get_callback_invoke(desc, NULL);
 				last_cookie = desc->cookie;
 				dma_run_dependencies(desc);
 			}
@@ -388,8 +386,7 @@ static void sirfsoc_dma_process_completed(struct sirfsoc_dma *sdma)
 
 			desc = &sdesc->desc;
 			while (happened_cyclic != schan->completed_cyclic) {
-				if (desc->callback)
-					desc->callback(desc->callback_param);
+				dmaengine_desc_get_callback_invoke(desc, NULL);
 				schan->completed_cyclic++;
 			}
 		}
diff --git a/drivers/dma/ste_dma40.c b/drivers/dma/ste_dma40.c
index e43d2bbfd122..08f3d7be2df0 100644
--- a/drivers/dma/ste_dma40.c
+++ b/drivers/dma/ste_dma40.c
@@ -1570,8 +1570,7 @@ static void dma_tasklet(unsigned long data)
 	struct d40_desc *d40d;
 	unsigned long flags;
 	bool callback_active;
-	dma_async_tx_callback callback;
-	void *callback_param;
+	struct dmaengine_desc_callback cb;
 
 	spin_lock_irqsave(&d40c->lock, flags);
 
@@ -1598,8 +1597,7 @@ static void dma_tasklet(unsigned long data)
 
 	/* Callback to client */
 	callback_active = !!(d40d->txd.flags & DMA_PREP_INTERRUPT);
-	callback = d40d->txd.callback;
-	callback_param = d40d->txd.callback_param;
+	dmaengine_desc_get_callback(&d40d->txd, &cb);
 
 	if (!d40d->cyclic) {
 		if (async_tx_test_ack(&d40d->txd)) {
@@ -1620,8 +1618,8 @@ static void dma_tasklet(unsigned long data)
 
 	spin_unlock_irqrestore(&d40c->lock, flags);
 
-	if (callback_active && callback)
-		callback(callback_param);
+	if (callback_active)
+		dmaengine_desc_callback_invoke(&cb, NULL);
 
 	return;
 
diff --git a/drivers/dma/tegra20-apb-dma.c b/drivers/dma/tegra20-apb-dma.c
index 6ab9eb98588a..3722b9d8d9fe 100644
--- a/drivers/dma/tegra20-apb-dma.c
+++ b/drivers/dma/tegra20-apb-dma.c
@@ -655,8 +655,7 @@ static void handle_cont_sngl_cycle_dma_done(struct tegra_dma_channel *tdc,
 static void tegra_dma_tasklet(unsigned long data)
 {
 	struct tegra_dma_channel *tdc = (struct tegra_dma_channel *)data;
-	dma_async_tx_callback callback = NULL;
-	void *callback_param = NULL;
+	struct dmaengine_desc_callback cb;
 	struct tegra_dma_desc *dma_desc;
 	unsigned long flags;
 	int cb_count;
@@ -666,13 +665,12 @@ static void tegra_dma_tasklet(unsigned long data)
 		dma_desc  = list_first_entry(&tdc->cb_desc,
 					typeof(*dma_desc), cb_node);
 		list_del(&dma_desc->cb_node);
-		callback = dma_desc->txd.callback;
-		callback_param = dma_desc->txd.callback_param;
+		dmaengine_desc_get_callback(&dma_desc->txd, &cb);
 		cb_count = dma_desc->cb_count;
 		dma_desc->cb_count = 0;
 		spin_unlock_irqrestore(&tdc->lock, flags);
-		while (cb_count-- && callback)
-			callback(callback_param);
+		while (cb_count--)
+			dmaengine_desc_callback_invoke(&cb, NULL);
 		spin_lock_irqsave(&tdc->lock, flags);
 	}
 	spin_unlock_irqrestore(&tdc->lock, flags);
diff --git a/drivers/dma/timb_dma.c b/drivers/dma/timb_dma.c
index e82745aa42a8..896bafb7a532 100644
--- a/drivers/dma/timb_dma.c
+++ b/drivers/dma/timb_dma.c
@@ -226,8 +226,7 @@ static void __td_start_dma(struct timb_dma_chan *td_chan)
 
 static void __td_finish(struct timb_dma_chan *td_chan)
 {
-	dma_async_tx_callback		callback;
-	void				*param;
+	struct dmaengine_desc_callback	cb;
 	struct dma_async_tx_descriptor	*txd;
 	struct timb_dma_desc		*td_desc;
 
@@ -252,8 +251,7 @@ static void __td_finish(struct timb_dma_chan *td_chan)
 	dma_cookie_complete(txd);
 	td_chan->ongoing = false;
 
-	callback = txd->callback;
-	param = txd->callback_param;
+	dmaengine_desc_get_callback(txd, &cb);
 
 	list_move(&td_desc->desc_node, &td_chan->free_list);
 
@@ -262,8 +260,7 @@ static void __td_finish(struct timb_dma_chan *td_chan)
 	 * The API requires that no submissions are done from a
 	 * callback, so we don't need to drop the lock here
 	 */
-	if (callback)
-		callback(param);
+	dmaengine_desc_callback_invoke(&cb, NULL);
 }
 
 static u32 __td_ier_mask(struct timb_dma *td)
diff --git a/drivers/dma/txx9dmac.c b/drivers/dma/txx9dmac.c
index 7632290e7c14..4d8c7b9078fd 100644
--- a/drivers/dma/txx9dmac.c
+++ b/drivers/dma/txx9dmac.c
@@ -403,16 +403,14 @@ static void
 txx9dmac_descriptor_complete(struct txx9dmac_chan *dc,
 			     struct txx9dmac_desc *desc)
 {
-	dma_async_tx_callback callback;
-	void *param;
+	struct dmaengine_desc_callback cb;
 	struct dma_async_tx_descriptor *txd = &desc->txd;
 
 	dev_vdbg(chan2dev(&dc->chan), "descriptor %u %p complete\n",
 		 txd->cookie, desc);
 
 	dma_cookie_complete(txd);
-	callback = txd->callback;
-	param = txd->callback_param;
+	dmaengine_desc_get_callback(txd, &cb);
 
 	txx9dmac_sync_desc_for_cpu(dc, desc);
 	list_splice_init(&desc->tx_list, &dc->free_list);
@@ -423,8 +421,7 @@ txx9dmac_descriptor_complete(struct txx9dmac_chan *dc,
 	 * The API requires that no submissions are done from a
 	 * callback, so we don't need to drop the lock here
 	 */
-	if (callback)
-		callback(param);
+	dmaengine_desc_callback_invoke(&cb, NULL);
 	dma_run_dependencies(txd);
 }
 
diff --git a/drivers/dma/virt-dma.c b/drivers/dma/virt-dma.c
index a35c211857dd..e47fc9b0944f 100644
--- a/drivers/dma/virt-dma.c
+++ b/drivers/dma/virt-dma.c
@@ -87,8 +87,7 @@ static void vchan_complete(unsigned long arg)
 {
 	struct virt_dma_chan *vc = (struct virt_dma_chan *)arg;
 	struct virt_dma_desc *vd;
-	dma_async_tx_callback cb = NULL;
-	void *cb_data = NULL;
+	struct dmaengine_desc_callback cb;
 	LIST_HEAD(head);
 
 	spin_lock_irq(&vc->lock);
@@ -96,18 +95,17 @@ static void vchan_complete(unsigned long arg)
 	vd = vc->cyclic;
 	if (vd) {
 		vc->cyclic = NULL;
-		cb = vd->tx.callback;
-		cb_data = vd->tx.callback_param;
+		dmaengine_desc_get_callback(&vd->tx, &cb);
+	} else {
+		memset(&cb, 0, sizeof(cb));
 	}
 	spin_unlock_irq(&vc->lock);
 
-	if (cb)
-		cb(cb_data);
+	dmaengine_desc_callback_invoke(&cb, NULL);
 
 	while (!list_empty(&head)) {
 		vd = list_first_entry(&head, struct virt_dma_desc, node);
-		cb = vd->tx.callback;
-		cb_data = vd->tx.callback_param;
+		dmaengine_desc_get_callback(&vd->tx, &cb);
 
 		list_del(&vd->node);
 		if (dmaengine_desc_test_reuse(&vd->tx))
@@ -115,8 +113,7 @@ static void vchan_complete(unsigned long arg)
 		else
 			vc->desc_free(vd);
 
-		if (cb)
-			cb(cb_data);
+		dmaengine_desc_callback_invoke(&cb, NULL);
 	}
 }
 
diff --git a/drivers/dma/xgene-dma.c b/drivers/dma/xgene-dma.c
index 9cb93c5b655d..d66ed11baaec 100644
--- a/drivers/dma/xgene-dma.c
+++ b/drivers/dma/xgene-dma.c
@@ -608,8 +608,7 @@ static void xgene_dma_run_tx_complete_actions(struct xgene_dma_chan *chan,
 	dma_cookie_complete(tx);
 
 	/* Run the link descriptor callback function */
-	if (tx->callback)
-		tx->callback(tx->callback_param);
+	dmaengine_desc_get_callback_invoke(tx, NULL);
 
 	dma_descriptor_unmap(tx);
 
diff --git a/drivers/dma/xilinx/xilinx_dma.c b/drivers/dma/xilinx/xilinx_dma.c
index 4e223d094433..8288fe4d17c3 100644
--- a/drivers/dma/xilinx/xilinx_dma.c
+++ b/drivers/dma/xilinx/xilinx_dma.c
@@ -755,8 +755,7 @@ static void xilinx_dma_chan_desc_cleanup(struct xilinx_dma_chan *chan)
 	spin_lock_irqsave(&chan->lock, flags);
 
 	list_for_each_entry_safe(desc, next, &chan->done_list, node) {
-		dma_async_tx_callback callback;
-		void *callback_param;
+		struct dmaengine_desc_callback cb;
 
 		if (desc->cyclic) {
 			xilinx_dma_chan_handle_cyclic(chan, desc, &flags);
@@ -767,11 +766,10 @@ static void xilinx_dma_chan_desc_cleanup(struct xilinx_dma_chan *chan)
 		list_del(&desc->node);
 
 		/* Run the link descriptor callback function */
-		callback = desc->async_tx.callback;
-		callback_param = desc->async_tx.callback_param;
-		if (callback) {
+		dmaengine_desc_get_callback(&desc->async_tx, &cb);
+		if (dmaengine_desc_callback_valid(&cb)) {
 			spin_unlock_irqrestore(&chan->lock, flags);
-			callback(callback_param);
+			dmaengine_desc_callback_invoke(&cb, NULL);
 			spin_lock_irqsave(&chan->lock, flags);
 		}
 
diff --git a/drivers/ntb/ntb_transport.c b/drivers/ntb/ntb_transport.c
index d5c5894f252e..8601c10acf74 100644
--- a/drivers/ntb/ntb_transport.c
+++ b/drivers/ntb/ntb_transport.c
@@ -102,13 +102,16 @@ struct ntb_queue_entry {
 	void *buf;
 	unsigned int len;
 	unsigned int flags;
+	int retries;
+	int errors;
+	unsigned int tx_index;
+	unsigned int rx_index;
 
 	struct ntb_transport_qp *qp;
 	union {
 		struct ntb_payload_header __iomem *tx_hdr;
 		struct ntb_payload_header *rx_hdr;
 	};
-	unsigned int index;
 };
 
 struct ntb_rx_info {
@@ -259,6 +262,12 @@ enum {
 static void ntb_transport_rxc_db(unsigned long data);
 static const struct ntb_ctx_ops ntb_transport_ops;
 static struct ntb_client ntb_transport_client;
+static int ntb_async_tx_submit(struct ntb_transport_qp *qp,
+			       struct ntb_queue_entry *entry);
+static void ntb_memcpy_tx(struct ntb_queue_entry *entry, void __iomem *offset);
+static int ntb_async_rx_submit(struct ntb_queue_entry *entry, void *offset);
+static void ntb_memcpy_rx(struct ntb_queue_entry *entry, void *offset);
+
 
 static int ntb_transport_bus_match(struct device *dev,
 				   struct device_driver *drv)
@@ -1229,7 +1238,7 @@ static void ntb_complete_rxc(struct ntb_transport_qp *qp)
 			break;
 
 		entry->rx_hdr->flags = 0;
-		iowrite32(entry->index, &qp->rx_info->entry);
+		iowrite32(entry->rx_index, &qp->rx_info->entry);
 
 		cb_data = entry->cb_data;
 		len = entry->len;
@@ -1247,10 +1256,36 @@ static void ntb_complete_rxc(struct ntb_transport_qp *qp)
 	spin_unlock_irqrestore(&qp->ntb_rx_q_lock, irqflags);
 }
 
-static void ntb_rx_copy_callback(void *data)
+static void ntb_rx_copy_callback(void *data,
+				 const struct dmaengine_result *res)
 {
 	struct ntb_queue_entry *entry = data;
 
+	/* we need to check DMA results if we are using DMA */
+	if (res) {
+		enum dmaengine_tx_result dma_err = res->result;
+
+		switch (dma_err) {
+		case DMA_TRANS_READ_FAILED:
+		case DMA_TRANS_WRITE_FAILED:
+			entry->errors++;
+		case DMA_TRANS_ABORTED:
+		{
+			struct ntb_transport_qp *qp = entry->qp;
+			void *offset = qp->rx_buff + qp->rx_max_frame *
+					qp->rx_index;
+
+			ntb_memcpy_rx(entry, offset);
+			qp->rx_memcpy++;
+			return;
+		}
+
+		case DMA_TRANS_NOERROR:
+		default:
+			break;
+		}
+	}
+
 	entry->flags |= DESC_DONE_FLAG;
 
 	ntb_complete_rxc(entry->qp);
@@ -1266,10 +1301,10 @@ static void ntb_memcpy_rx(struct ntb_queue_entry *entry, void *offset)
 	/* Ensure that the data is fully copied out before clearing the flag */
 	wmb();
 
-	ntb_rx_copy_callback(entry);
+	ntb_rx_copy_callback(entry, NULL);
 }
 
-static void ntb_async_rx(struct ntb_queue_entry *entry, void *offset)
+static int ntb_async_rx_submit(struct ntb_queue_entry *entry, void *offset)
 {
 	struct dma_async_tx_descriptor *txd;
 	struct ntb_transport_qp *qp = entry->qp;
@@ -1282,13 +1317,6 @@ static void ntb_async_rx(struct ntb_queue_entry *entry, void *offset)
 	int retries = 0;
 
 	len = entry->len;
-
-	if (!chan)
-		goto err;
-
-	if (len < copy_bytes)
-		goto err;
-
 	device = chan->device;
 	pay_off = (size_t)offset & ~PAGE_MASK;
 	buff_off = (size_t)buf & ~PAGE_MASK;
@@ -1316,7 +1344,8 @@ static void ntb_async_rx(struct ntb_queue_entry *entry, void *offset)
 	unmap->from_cnt = 1;
 
 	for (retries = 0; retries < DMA_RETRIES; retries++) {
-		txd = device->device_prep_dma_memcpy(chan, unmap->addr[1],
+		txd = device->device_prep_dma_memcpy(chan,
+						     unmap->addr[1],
 						     unmap->addr[0], len,
 						     DMA_PREP_INTERRUPT);
 		if (txd)
@@ -1331,7 +1360,7 @@ static void ntb_async_rx(struct ntb_queue_entry *entry, void *offset)
 		goto err_get_unmap;
 	}
 
-	txd->callback = ntb_rx_copy_callback;
+	txd->callback_result = ntb_rx_copy_callback;
 	txd->callback_param = entry;
 	dma_set_unmap(txd, unmap);
 
@@ -1345,13 +1374,38 @@ static void ntb_async_rx(struct ntb_queue_entry *entry, void *offset)
 
 	qp->rx_async++;
 
-	return;
+	return 0;
 
 err_set_unmap:
 	dmaengine_unmap_put(unmap);
 err_get_unmap:
 	dmaengine_unmap_put(unmap);
 err:
+	return -ENXIO;
+}
+
+static void ntb_async_rx(struct ntb_queue_entry *entry, void *offset)
+{
+	struct ntb_transport_qp *qp = entry->qp;
+	struct dma_chan *chan = qp->rx_dma_chan;
+	int res;
+
+	if (!chan)
+		goto err;
+
+	if (entry->len < copy_bytes)
+		goto err;
+
+	res = ntb_async_rx_submit(entry, offset);
+	if (res < 0)
+		goto err;
+
+	if (!entry->retries)
+		qp->rx_async++;
+
+	return;
+
+err:
 	ntb_memcpy_rx(entry, offset);
 	qp->rx_memcpy++;
 }
@@ -1397,7 +1451,7 @@ static int ntb_process_rxc(struct ntb_transport_qp *qp)
 	}
 
 	entry->rx_hdr = hdr;
-	entry->index = qp->rx_index;
+	entry->rx_index = qp->rx_index;
 
 	if (hdr->len > entry->len) {
 		dev_dbg(&qp->ndev->pdev->dev,
@@ -1467,12 +1521,39 @@ static void ntb_transport_rxc_db(unsigned long data)
 	}
 }
 
-static void ntb_tx_copy_callback(void *data)
+static void ntb_tx_copy_callback(void *data,
+				 const struct dmaengine_result *res)
 {
 	struct ntb_queue_entry *entry = data;
 	struct ntb_transport_qp *qp = entry->qp;
 	struct ntb_payload_header __iomem *hdr = entry->tx_hdr;
 
+	/* we need to check DMA results if we are using DMA */
+	if (res) {
+		enum dmaengine_tx_result dma_err = res->result;
+
+		switch (dma_err) {
+		case DMA_TRANS_READ_FAILED:
+		case DMA_TRANS_WRITE_FAILED:
+			entry->errors++;
+		case DMA_TRANS_ABORTED:
+		{
+			void __iomem *offset =
+				qp->tx_mw + qp->tx_max_frame *
+				entry->tx_index;
+
+			/* resubmit via CPU */
+			ntb_memcpy_tx(entry, offset);
+			qp->tx_memcpy++;
+			return;
+		}
+
+		case DMA_TRANS_NOERROR:
+		default:
+			break;
+		}
+	}
+
 	iowrite32(entry->flags | DESC_DONE_FLAG, &hdr->flags);
 
 	ntb_peer_db_set(qp->ndev, BIT_ULL(qp->qp_num));
@@ -1507,40 +1588,25 @@ static void ntb_memcpy_tx(struct ntb_queue_entry *entry, void __iomem *offset)
 	/* Ensure that the data is fully copied out before setting the flags */
 	wmb();
 
-	ntb_tx_copy_callback(entry);
+	ntb_tx_copy_callback(entry, NULL);
 }
 
-static void ntb_async_tx(struct ntb_transport_qp *qp,
-			 struct ntb_queue_entry *entry)
+static int ntb_async_tx_submit(struct ntb_transport_qp *qp,
+			       struct ntb_queue_entry *entry)
 {
-	struct ntb_payload_header __iomem *hdr;
 	struct dma_async_tx_descriptor *txd;
 	struct dma_chan *chan = qp->tx_dma_chan;
 	struct dma_device *device;
+	size_t len = entry->len;
+	void *buf = entry->buf;
 	size_t dest_off, buff_off;
 	struct dmaengine_unmap_data *unmap;
 	dma_addr_t dest;
 	dma_cookie_t cookie;
-	void __iomem *offset;
-	size_t len = entry->len;
-	void *buf = entry->buf;
 	int retries = 0;
 
-	offset = qp->tx_mw + qp->tx_max_frame * qp->tx_index;
-	hdr = offset + qp->tx_max_frame - sizeof(struct ntb_payload_header);
-	entry->tx_hdr = hdr;
-
-	iowrite32(entry->len, &hdr->len);
-	iowrite32((u32)qp->tx_pkts, &hdr->ver);
-
-	if (!chan)
-		goto err;
-
-	if (len < copy_bytes)
-		goto err;
-
 	device = chan->device;
-	dest = qp->tx_mw_phys + qp->tx_max_frame * qp->tx_index;
+	dest = qp->tx_mw_phys + qp->tx_max_frame * entry->tx_index;
 	buff_off = (size_t)buf & ~PAGE_MASK;
 	dest_off = (size_t)dest & ~PAGE_MASK;
 
@@ -1560,8 +1626,9 @@ static void ntb_async_tx(struct ntb_transport_qp *qp,
 	unmap->to_cnt = 1;
 
 	for (retries = 0; retries < DMA_RETRIES; retries++) {
-		txd = device->device_prep_dma_memcpy(chan, dest, unmap->addr[0],
-						     len, DMA_PREP_INTERRUPT);
+		txd = device->device_prep_dma_memcpy(chan, dest,
+						     unmap->addr[0], len,
+						     DMA_PREP_INTERRUPT);
 		if (txd)
 			break;
 
@@ -1574,7 +1641,7 @@ static void ntb_async_tx(struct ntb_transport_qp *qp,
 		goto err_get_unmap;
 	}
 
-	txd->callback = ntb_tx_copy_callback;
+	txd->callback_result = ntb_tx_copy_callback;
 	txd->callback_param = entry;
 	dma_set_unmap(txd, unmap);
 
@@ -1585,14 +1652,48 @@ static void ntb_async_tx(struct ntb_transport_qp *qp,
 	dmaengine_unmap_put(unmap);
 
 	dma_async_issue_pending(chan);
-	qp->tx_async++;
 
-	return;
+	return 0;
 err_set_unmap:
 	dmaengine_unmap_put(unmap);
 err_get_unmap:
 	dmaengine_unmap_put(unmap);
 err:
+	return -ENXIO;
+}
+
+static void ntb_async_tx(struct ntb_transport_qp *qp,
+			 struct ntb_queue_entry *entry)
+{
+	struct ntb_payload_header __iomem *hdr;
+	struct dma_chan *chan = qp->tx_dma_chan;
+	void __iomem *offset;
+	int res;
+
+	entry->tx_index = qp->tx_index;
+	offset = qp->tx_mw + qp->tx_max_frame * entry->tx_index;
+	hdr = offset + qp->tx_max_frame - sizeof(struct ntb_payload_header);
+	entry->tx_hdr = hdr;
+
+	iowrite32(entry->len, &hdr->len);
+	iowrite32((u32)qp->tx_pkts, &hdr->ver);
+
+	if (!chan)
+		goto err;
+
+	if (entry->len < copy_bytes)
+		goto err;
+
+	res = ntb_async_tx_submit(qp, entry);
+	if (res < 0)
+		goto err;
+
+	if (!entry->retries)
+		qp->tx_async++;
+
+	return;
+
+err:
 	ntb_memcpy_tx(entry, offset);
 	qp->tx_memcpy++;
 }
@@ -1928,6 +2029,9 @@ int ntb_transport_rx_enqueue(struct ntb_transport_qp *qp, void *cb, void *data,
 	entry->buf = data;
 	entry->len = len;
 	entry->flags = 0;
+	entry->retries = 0;
+	entry->errors = 0;
+	entry->rx_index = 0;
 
 	ntb_list_add(&qp->ntb_rx_q_lock, &entry->entry, &qp->rx_pend_q);
 
@@ -1970,6 +2074,9 @@ int ntb_transport_tx_enqueue(struct ntb_transport_qp *qp, void *cb, void *data,
 	entry->buf = data;
 	entry->len = len;
 	entry->flags = 0;
+	entry->errors = 0;
+	entry->retries = 0;
+	entry->tx_index = 0;
 
 	rc = ntb_process_tx(qp, entry);
 	if (rc)
diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
index 30de0197263a..cc535a478bae 100644
--- a/include/linux/dmaengine.h
+++ b/include/linux/dmaengine.h
@@ -441,6 +441,21 @@ typedef bool (*dma_filter_fn)(struct dma_chan *chan, void *filter_param);
 
 typedef void (*dma_async_tx_callback)(void *dma_async_param);
 
+enum dmaengine_tx_result {
+	DMA_TRANS_NOERROR = 0,		/* SUCCESS */
+	DMA_TRANS_READ_FAILED,		/* Source DMA read failed */
+	DMA_TRANS_WRITE_FAILED,		/* Destination DMA write failed */
+	DMA_TRANS_ABORTED,		/* Op never submitted / aborted */
+};
+
+struct dmaengine_result {
+	enum dmaengine_tx_result result;
+	u32 residue;
+};
+
+typedef void (*dma_async_tx_callback_result)(void *dma_async_param,
+				const struct dmaengine_result *result);
+
 struct dmaengine_unmap_data {
 	u8 map_cnt;
 	u8 to_cnt;
@@ -478,6 +493,7 @@ struct dma_async_tx_descriptor {
 	dma_cookie_t (*tx_submit)(struct dma_async_tx_descriptor *tx);
 	int (*desc_free)(struct dma_async_tx_descriptor *tx);
 	dma_async_tx_callback callback;
+	dma_async_tx_callback_result callback_result;
 	void *callback_param;
 	struct dmaengine_unmap_data *unmap;
 #ifdef CONFIG_ASYNC_TX_ENABLE_CHANNEL_SWITCH