The first slot in the ParamRAM of EDMA holds the current active
subtransfer. Depending on the direction we read either the source or
the destination address from there. In the internal psets we have the
address of the buffer(s).

In the cyclic case we only use the internal pset[0] which holds the
start address of the circular buffer and calculate the remaining room
to the end of the buffer.

In the SG case we read the current address and compare it to the
internal psets address and length.

- If the current address is outside of this range, the pset has been
  processed already and we mark it done, update the residue_stat value
  and process the next set. That avoids that we need to walk all
  processed psets for every invocation of tx_status.

- If its inside the range we know that we look at the current active
  set and stop the walk.

- In case of intermediate transfers we update the stats in the
  interrupt callback function before starting the next batch of
  transfers. The tx_status callback and the interrupt callback are
  serialized via vchan.lock.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
---
 drivers/dma/edma.c |   67 +++++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 65 insertions(+), 2 deletions(-)

Index: linux-2.6/drivers/dma/edma.c
===================================================================
--- linux-2.6.orig/drivers/dma/edma.c
+++ linux-2.6/drivers/dma/edma.c
@@ -71,7 +71,11 @@ struct edma_desc {
 	int				absync;
 	int				pset_nr;
 	int				processed;
+	int				processed_stat;
 	u32				residue;
+	u32				sg_len;
+	u32				residue_stat;
+	struct edma_chan		*echan;
 	struct edma_pset		pset[0];
 };
 
@@ -144,11 +148,13 @@ static void edma_execute(struct edma_cha
 	/* Find out how many left */
 	left = edesc->pset_nr - edesc->processed;
 	nslots = min(MAX_NR_SG, left);
+	edesc->sg_len = 0;
 
 	/* Write descriptor PaRAM set(s) */
 	for (i = 0; i < nslots; i++) {
 		j = i + edesc->processed;
 		edma_write_slot(echan->slot[i], &edesc->pset[j].param);
+		edesc->sg_len += edesc->pset[j].len;
 		dev_dbg(echan->vchan.chan.device->dev,
 			"\n pset[%d]:\n"
 			"  chnum\t%d\n"
@@ -433,6 +439,7 @@ static struct dma_async_tx_descriptor *e
 	edesc->pset_nr = sg_len;
 	edesc->residue = 0;
 	edesc->direction = direction;
+	edesc->echan = echan;
 
 	/* Allocate a PaRAM slot, if needed */
 	nslots = min_t(unsigned, MAX_NR_SG, sg_len);
@@ -478,6 +485,7 @@ static struct dma_async_tx_descriptor *e
 		if (i == sg_len - 1)
 			edesc->pset[i].param.opt |= TCINTEN;
 	}
+	edesc->residue_stat = edesc->residue;
 
 	return vchan_tx_prep(&echan->vchan, &edesc->vdesc, tx_flags);
 }
@@ -545,8 +553,9 @@ static struct dma_async_tx_descriptor *e
 
 	edesc->cyclic = 1;
 	edesc->pset_nr = nslots;
-	edesc->residue = buf_len;
+	edesc->residue = edesc->residue_stat = buf_len;
 	edesc->direction = direction;
+	edesc->echan = echan;
 
 	dev_dbg(dev, "%s: nslots=%d\n", __func__, nslots);
 	dev_dbg(dev, "%s: period_len=%d\n", __func__, period_len);
@@ -648,6 +657,12 @@ static void edma_callback(unsigned ch_nu
 				edma_execute(echan);
 			} else {
 				dev_dbg(dev, "Intermediate transfer complete on channel %d\n", ch_num);
+
+				/* Update statistics for tx_status */
+				edesc->residue -= edesc->sg_len;
+				edesc->residue_stat = edesc->residue;
+				edesc->processed_stat = edesc->processed;
+
 				edma_execute(echan);
 			}
 		}
@@ -775,6 +790,54 @@ static void edma_issue_pending(struct dm
 	spin_unlock_irqrestore(&echan->vchan.lock, flags);
 }
 
+static u32 edma_residue(struct edma_desc *edesc)
+{
+	bool dst = edesc->direction == DMA_DEV_TO_MEM;
+	struct edma_pset *pset = edesc->pset;
+	dma_addr_t done, pos;
+	int i;
+
+	/*
+	 * We always read the dst/src position from the first RamPar
+	 * pset. That's the one which is active now.
+	 */
+	pos = edma_get_position(edesc->echan->slot[0], dst);
+
+	/*
+	 * Cyclic is simple. Just subtract pset[0].addr from pos.
+	 *
+	 * We never update edesc->residue in the cyclic case, so we
+	 * can tell the remaining room to the end of the circular
+	 * buffer.
+	 */
+	if (edesc->cyclic) {
+		done = pos - pset->addr;
+		edesc->residue_stat = edesc->residue - done;
+		return edesc->residue_stat;
+	}
+
+	/*
+	 * For SG operation we catch up with the last processed
+	 * status.
+	 */
+	pset += edesc->processed_stat;
+
+	for (i = edesc->processed_stat; i < edesc->processed; i++, pset++) {
+		/*
+		 * If we are inside this pset address range, we know
+		 * this is the active one. Get the current delta and
+		 * stop walking the psets.
+		 */
+		if (pos >= pset->addr && pos < pset->addr + pset->len)
+			return edesc->residue_stat - (pos - pset->addr);
+
+		/* Otherwise mark it done and update residue_stat. */
+		edesc->processed_stat++;
+		edesc->residue_stat -= pset->len;
+	}
+	return edesc->residue_stat;
+}
+
 /* Check request completion status */
 static enum dma_status edma_tx_status(struct dma_chan *chan,
 				      dma_cookie_t cookie,
@@ -791,7 +854,7 @@ static enum dma_status edma_tx_status(st
 
 	spin_lock_irqsave(&echan->vchan.lock, flags);
 	if (echan->edesc && echan->edesc->vdesc.tx.cookie == cookie)
-		txstate->residue = echan->edesc->residue;
+		txstate->residue = edma_residue(echan->edesc);
 	else if ((vdesc = vchan_find_desc(&echan->vchan, cookie)))
 		txstate->residue = to_edma_desc(&vdesc->tx)->residue;
 	spin_unlock_irqrestore(&echan->vchan.lock, flags);