Merge branch 'dma40' into dmaengine

3 files changed, 758 insertions, 928 deletions

diff --git a/drivers/dma/ste_dma40.c b/drivers/dma/ste_dma40.c
index 6e1d46a65d0e..af955de035f4 100644
--- a/drivers/dma/ste_dma40.c
+++ b/drivers/dma/ste_dma40.c
@@ -68,6 +68,7 @@ enum d40_command {
  * @base: Pointer to memory area when the pre_alloc_lli's are not large
  * enough, IE bigger than the most common case, 1 dst and 1 src. NULL if
  * pre_alloc_lli is used.
+ * @dma_addr: DMA address, if mapped
  * @size: The size in bytes of the memory at base or the size of pre_alloc_lli.
  * @pre_alloc_lli: Pre allocated area for the most common case of transfers,
  * one buffer to one buffer.
@@ -75,6 +76,7 @@ enum d40_command {
 struct d40_lli_pool {
 	void	*base;
 	int	 size;
+	dma_addr_t	dma_addr;
 	/* Space for dst and src, plus an extra for padding */
 	u8	 pre_alloc_lli[3 * sizeof(struct d40_phy_lli)];
 };
@@ -94,7 +96,6 @@ struct d40_lli_pool {
  * during a transfer.
  * @node: List entry.
  * @is_in_client_list: true if the client owns this descriptor.
- * @is_hw_linked: true if this job will automatically be continued for
  * the previous one.
  *
  * This descriptor is used for both logical and physical transfers.
@@ -114,7 +115,7 @@ struct d40_desc {
 	struct list_head		 node;
 
 	bool				 is_in_client_list;
-	bool				 is_hw_linked;
+	bool				 cyclic;
 };
 
 /**
@@ -130,6 +131,7 @@ struct d40_desc {
  */
 struct d40_lcla_pool {
 	void		*base;
+	dma_addr_t	dma_addr;
 	void		*base_unaligned;
 	int		 pages;
 	spinlock_t	 lock;
@@ -303,9 +305,37 @@ struct d40_reg_val {
 	unsigned int val;
 };
 
-static int d40_pool_lli_alloc(struct d40_desc *d40d,
-			      int lli_len, bool is_log)
+static struct device *chan2dev(struct d40_chan *d40c)
 {
+	return &d40c->chan.dev->device;
+}
+
+static bool chan_is_physical(struct d40_chan *chan)
+{
+	return chan->log_num == D40_PHY_CHAN;
+}
+
+static bool chan_is_logical(struct d40_chan *chan)
+{
+	return !chan_is_physical(chan);
+}
+
+static void __iomem *chan_base(struct d40_chan *chan)
+{
+	return chan->base->virtbase + D40_DREG_PCBASE +
+	       chan->phy_chan->num * D40_DREG_PCDELTA;
+}
+
+#define d40_err(dev, format, arg...)		\
+	dev_err(dev, "[%s] " format, __func__, ## arg)
+
+#define chan_err(d40c, format, arg...)		\
+	d40_err(chan2dev(d40c), format, ## arg)
+
+static int d40_pool_lli_alloc(struct d40_chan *d40c, struct d40_desc *d40d,
+			      int lli_len)
+{
+	bool is_log = chan_is_logical(d40c);
 	u32 align;
 	void *base;
 
@@ -319,7 +349,7 @@ static int d40_pool_lli_alloc(struct d40_desc *d40d,
 		d40d->lli_pool.size = sizeof(d40d->lli_pool.pre_alloc_lli);
 		d40d->lli_pool.base = NULL;
 	} else {
-		d40d->lli_pool.size = ALIGN(lli_len * 2 * align, align);
+		d40d->lli_pool.size = lli_len * 2 * align;
 
 		base = kmalloc(d40d->lli_pool.size + align, GFP_NOWAIT);
 		d40d->lli_pool.base = base;
@@ -329,22 +359,37 @@ static int d40_pool_lli_alloc(struct d40_desc *d40d,
 	}
 
 	if (is_log) {
-		d40d->lli_log.src = PTR_ALIGN((struct d40_log_lli *) base,
-					      align);
-		d40d->lli_log.dst = PTR_ALIGN(d40d->lli_log.src + lli_len,
-					      align);
+		d40d->lli_log.src = PTR_ALIGN(base, align);
+		d40d->lli_log.dst = d40d->lli_log.src + lli_len;
+
+		d40d->lli_pool.dma_addr = 0;
 	} else {
-		d40d->lli_phy.src = PTR_ALIGN((struct d40_phy_lli *)base,
-					      align);
-		d40d->lli_phy.dst = PTR_ALIGN(d40d->lli_phy.src + lli_len,
-					      align);
+		d40d->lli_phy.src = PTR_ALIGN(base, align);
+		d40d->lli_phy.dst = d40d->lli_phy.src + lli_len;
+
+		d40d->lli_pool.dma_addr = dma_map_single(d40c->base->dev,
+							 d40d->lli_phy.src,
+							 d40d->lli_pool.size,
+							 DMA_TO_DEVICE);
+
+		if (dma_mapping_error(d40c->base->dev,
+				      d40d->lli_pool.dma_addr)) {
+			kfree(d40d->lli_pool.base);
+			d40d->lli_pool.base = NULL;
+			d40d->lli_pool.dma_addr = 0;
+			return -ENOMEM;
+		}
 	}
 
 	return 0;
 }
 
-static void d40_pool_lli_free(struct d40_desc *d40d)
+static void d40_pool_lli_free(struct d40_chan *d40c, struct d40_desc *d40d)
 {
+	if (d40d->lli_pool.dma_addr)
+		dma_unmap_single(d40c->base->dev, d40d->lli_pool.dma_addr,
+				 d40d->lli_pool.size, DMA_TO_DEVICE);
+
 	kfree(d40d->lli_pool.base);
 	d40d->lli_pool.base = NULL;
 	d40d->lli_pool.size = 0;
@@ -391,7 +436,7 @@ static int d40_lcla_free_all(struct d40_chan *d40c,
 	int i;
 	int ret = -EINVAL;
 
-	if (d40c->log_num == D40_PHY_CHAN)
+	if (chan_is_physical(d40c))
 		return 0;
 
 	spin_lock_irqsave(&d40c->base->lcla_pool.lock, flags);
@@ -430,7 +475,7 @@ static struct d40_desc *d40_desc_get(struct d40_chan *d40c)
 
 		list_for_each_entry_safe(d, _d, &d40c->client, node)
 			if (async_tx_test_ack(&d->txd)) {
-				d40_pool_lli_free(d);
+				d40_pool_lli_free(d40c, d);
 				d40_desc_remove(d);
 				desc = d;
 				memset(desc, 0, sizeof(*desc));
@@ -450,6 +495,7 @@ static struct d40_desc *d40_desc_get(struct d40_chan *d40c)
 static void d40_desc_free(struct d40_chan *d40c, struct d40_desc *d40d)
 {
 
+	d40_pool_lli_free(d40c, d40d);
 	d40_lcla_free_all(d40c, d40d);
 	kmem_cache_free(d40c->base->desc_slab, d40d);
 }
@@ -459,57 +505,128 @@ static void d40_desc_submit(struct d40_chan *d40c, struct d40_desc *desc)
 	list_add_tail(&desc->node, &d40c->active);
 }
 
-static void d40_desc_load(struct d40_chan *d40c, struct d40_desc *d40d)
+static void d40_phy_lli_load(struct d40_chan *chan, struct d40_desc *desc)
 {
-	int curr_lcla = -EINVAL, next_lcla;
+	struct d40_phy_lli *lli_dst = desc->lli_phy.dst;
+	struct d40_phy_lli *lli_src = desc->lli_phy.src;
+	void __iomem *base = chan_base(chan);
+
+	writel(lli_src->reg_cfg, base + D40_CHAN_REG_SSCFG);
+	writel(lli_src->reg_elt, base + D40_CHAN_REG_SSELT);
+	writel(lli_src->reg_ptr, base + D40_CHAN_REG_SSPTR);
+	writel(lli_src->reg_lnk, base + D40_CHAN_REG_SSLNK);
+
+	writel(lli_dst->reg_cfg, base + D40_CHAN_REG_SDCFG);
+	writel(lli_dst->reg_elt, base + D40_CHAN_REG_SDELT);
+	writel(lli_dst->reg_ptr, base + D40_CHAN_REG_SDPTR);
+	writel(lli_dst->reg_lnk, base + D40_CHAN_REG_SDLNK);
+}
 
-	if (d40c->log_num == D40_PHY_CHAN) {
-		d40_phy_lli_write(d40c->base->virtbase,
-				  d40c->phy_chan->num,
-				  d40d->lli_phy.dst,
-				  d40d->lli_phy.src);
-		d40d->lli_current = d40d->lli_len;
-	} else {
+static void d40_log_lli_to_lcxa(struct d40_chan *chan, struct d40_desc *desc)
+{
+	struct d40_lcla_pool *pool = &chan->base->lcla_pool;
+	struct d40_log_lli_bidir *lli = &desc->lli_log;
+	int lli_current = desc->lli_current;
+	int lli_len = desc->lli_len;
+	bool cyclic = desc->cyclic;
+	int curr_lcla = -EINVAL;
+	int first_lcla = 0;
+	bool linkback;
 
-		if ((d40d->lli_len - d40d->lli_current) > 1)
-			curr_lcla = d40_lcla_alloc_one(d40c, d40d);
+	/*
+	 * We may have partially running cyclic transfers, in case we did't get
+	 * enough LCLA entries.
+	 */
+	linkback = cyclic && lli_current == 0;
 
-		d40_log_lli_lcpa_write(d40c->lcpa,
-				       &d40d->lli_log.dst[d40d->lli_current],
-				       &d40d->lli_log.src[d40d->lli_current],
-				       curr_lcla);
+	/*
+	 * For linkback, we need one LCLA even with only one link, because we
+	 * can't link back to the one in LCPA space
+	 */
+	if (linkback || (lli_len - lli_current > 1)) {
+		curr_lcla = d40_lcla_alloc_one(chan, desc);
+		first_lcla = curr_lcla;
+	}
 
-		d40d->lli_current++;
-		for (; d40d->lli_current < d40d->lli_len; d40d->lli_current++) {
-			struct d40_log_lli *lcla;
+	/*
+	 * For linkback, we normally load the LCPA in the loop since we need to
+	 * link it to the second LCLA and not the first.  However, if we
+	 * couldn't even get a first LCLA, then we have to run in LCPA and
+	 * reload manually.
+	 */
+	if (!linkback || curr_lcla == -EINVAL) {
+		unsigned int flags = 0;
 
-			if (d40d->lli_current + 1 < d40d->lli_len)
-				next_lcla = d40_lcla_alloc_one(d40c, d40d);
-			else
-				next_lcla = -EINVAL;
+		if (curr_lcla == -EINVAL)
+			flags |= LLI_TERM_INT;
 
-			lcla = d40c->base->lcla_pool.base +
-				d40c->phy_chan->num * 1024 +
-				8 * curr_lcla * 2;
+		d40_log_lli_lcpa_write(chan->lcpa,
+				       &lli->dst[lli_current],
+				       &lli->src[lli_current],
+				       curr_lcla,
+				       flags);
+		lli_current++;
+	}
 
-			d40_log_lli_lcla_write(lcla,
-					       &d40d->lli_log.dst[d40d->lli_current],
-					       &d40d->lli_log.src[d40d->lli_current],
-					       next_lcla);
+	if (curr_lcla < 0)
+		goto out;
 
-			(void) dma_map_single(d40c->base->dev, lcla,
-					      2 * sizeof(struct d40_log_lli),
-					      DMA_TO_DEVICE);
+	for (; lli_current < lli_len; lli_current++) {
+		unsigned int lcla_offset = chan->phy_chan->num * 1024 +
+					   8 * curr_lcla * 2;
+		struct d40_log_lli *lcla = pool->base + lcla_offset;
+		unsigned int flags = 0;
+		int next_lcla;
 
-			curr_lcla = next_lcla;
+		if (lli_current + 1 < lli_len)
+			next_lcla = d40_lcla_alloc_one(chan, desc);
+		else
+			next_lcla = linkback ? first_lcla : -EINVAL;
 
-			if (curr_lcla == -EINVAL) {
-				d40d->lli_current++;
-				break;
-			}
+		if (cyclic || next_lcla == -EINVAL)
+			flags |= LLI_TERM_INT;
+
+		if (linkback && curr_lcla == first_lcla) {
+			/* First link goes in both LCPA and LCLA */
+			d40_log_lli_lcpa_write(chan->lcpa,
+					       &lli->dst[lli_current],
+					       &lli->src[lli_current],
+					       next_lcla, flags);
+		}
+
+		/*
+		 * One unused LCLA in the cyclic case if the very first
+		 * next_lcla fails...
+		 */
+		d40_log_lli_lcla_write(lcla,
+				       &lli->dst[lli_current],
+				       &lli->src[lli_current],
+				       next_lcla, flags);
+
+		dma_sync_single_range_for_device(chan->base->dev,
+					pool->dma_addr, lcla_offset,
+					2 * sizeof(struct d40_log_lli),
+					DMA_TO_DEVICE);
 
+		curr_lcla = next_lcla;
+
+		if (curr_lcla == -EINVAL || curr_lcla == first_lcla) {
+			lli_current++;
+			break;
 		}
 	}
+
+out:
+	desc->lli_current = lli_current;
+}
+
+static void d40_desc_load(struct d40_chan *d40c, struct d40_desc *d40d)
+{
+	if (chan_is_physical(d40c)) {
+		d40_phy_lli_load(d40c, d40d);
+		d40d->lli_current = d40d->lli_len;
+	} else
+		d40_log_lli_to_lcxa(d40c, d40d);
 }
 
 static struct d40_desc *d40_first_active_get(struct d40_chan *d40c)
@@ -543,18 +660,6 @@ static struct d40_desc *d40_first_queued(struct d40_chan *d40c)
 	return d;
 }
 
-static struct d40_desc *d40_last_queued(struct d40_chan *d40c)
-{
-	struct d40_desc *d;
-
-	if (list_empty(&d40c->queue))
-		return NULL;
-	list_for_each_entry(d, &d40c->queue, node)
-		if (list_is_last(&d->node, &d40c->queue))
-			break;
-	return d;
-}
-
 static int d40_psize_2_burst_size(bool is_log, int psize)
 {
 	if (is_log) {
@@ -666,9 +771,9 @@ static int d40_channel_execute_command(struct d40_chan *d40c,
 		}
 
 		if (i == D40_SUSPEND_MAX_IT) {
-			dev_err(&d40c->chan.dev->device,
-				"[%s]: unable to suspend the chl %d (log: %d) status %x\n",
-				__func__, d40c->phy_chan->num, d40c->log_num,
+			chan_err(d40c,
+				"unable to suspend the chl %d (log: %d) status %x\n",
+				d40c->phy_chan->num, d40c->log_num,
 				status);
 			dump_stack();
 			ret = -EBUSY;
@@ -701,17 +806,45 @@ static void d40_term_all(struct d40_chan *d40c)
 	d40c->busy = false;
 }
 
+static void __d40_config_set_event(struct d40_chan *d40c, bool enable,
+				   u32 event, int reg)
+{
+	void __iomem *addr = chan_base(d40c) + reg;
+	int tries;
+
+	if (!enable) {
+		writel((D40_DEACTIVATE_EVENTLINE << D40_EVENTLINE_POS(event))
+		       | ~D40_EVENTLINE_MASK(event), addr);
+		return;
+	}
+
+	/*
+	 * The hardware sometimes doesn't register the enable when src and dst
+	 * event lines are active on the same logical channel.  Retry to ensure
+	 * it does.  Usually only one retry is sufficient.
+	 */
+	tries = 100;
+	while (--tries) {
+		writel((D40_ACTIVATE_EVENTLINE << D40_EVENTLINE_POS(event))
+		       | ~D40_EVENTLINE_MASK(event), addr);
+
+		if (readl(addr) & D40_EVENTLINE_MASK(event))
+			break;
+	}
+
+	if (tries != 99)
+		dev_dbg(chan2dev(d40c),
+			"[%s] workaround enable S%cLNK (%d tries)\n",
+			__func__, reg == D40_CHAN_REG_SSLNK ? 'S' : 'D',
+			100 - tries);
+
+	WARN_ON(!tries);
+}
+
 static void d40_config_set_event(struct d40_chan *d40c, bool do_enable)
 {
-	u32 val;
 	unsigned long flags;
 
-	/* Notice, that disable requires the physical channel to be stopped */
-	if (do_enable)
-		val = D40_ACTIVATE_EVENTLINE;
-	else
-		val = D40_DEACTIVATE_EVENTLINE;
-
 	spin_lock_irqsave(&d40c->phy_chan->lock, flags);
 
 	/* Enable event line connected to device (or memcpy) */
@@ -719,20 +852,15 @@ static void d40_config_set_event(struct d40_chan *d40c, bool do_enable)
 	    (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_PERIPH)) {
 		u32 event = D40_TYPE_TO_EVENT(d40c->dma_cfg.src_dev_type);
 
-		writel((val << D40_EVENTLINE_POS(event)) |
-		       ~D40_EVENTLINE_MASK(event),
-		       d40c->base->virtbase + D40_DREG_PCBASE +
-		       d40c->phy_chan->num * D40_DREG_PCDELTA +
-		       D40_CHAN_REG_SSLNK);
+		__d40_config_set_event(d40c, do_enable, event,
+				       D40_CHAN_REG_SSLNK);
 	}
+
 	if (d40c->dma_cfg.dir !=  STEDMA40_PERIPH_TO_MEM) {
 		u32 event = D40_TYPE_TO_EVENT(d40c->dma_cfg.dst_dev_type);
 
-		writel((val << D40_EVENTLINE_POS(event)) |
-		       ~D40_EVENTLINE_MASK(event),
-		       d40c->base->virtbase + D40_DREG_PCBASE +
-		       d40c->phy_chan->num * D40_DREG_PCDELTA +
-		       D40_CHAN_REG_SDLNK);
+		__d40_config_set_event(d40c, do_enable, event,
+				       D40_CHAN_REG_SDLNK);
 	}
 
 	spin_unlock_irqrestore(&d40c->phy_chan->lock, flags);
@@ -740,15 +868,12 @@ static void d40_config_set_event(struct d40_chan *d40c, bool do_enable)
 
 static u32 d40_chan_has_events(struct d40_chan *d40c)
 {
+	void __iomem *chanbase = chan_base(d40c);
 	u32 val;
 
-	val = readl(d40c->base->virtbase + D40_DREG_PCBASE +
-		    d40c->phy_chan->num * D40_DREG_PCDELTA +
-		    D40_CHAN_REG_SSLNK);
+	val = readl(chanbase + D40_CHAN_REG_SSLNK);
+	val |= readl(chanbase + D40_CHAN_REG_SDLNK);
 
-	val |= readl(d40c->base->virtbase + D40_DREG_PCBASE +
-		     d40c->phy_chan->num * D40_DREG_PCDELTA +
-		     D40_CHAN_REG_SDLNK);
 	return val;
 }
 
@@ -771,7 +896,7 @@ static u32 d40_get_prmo(struct d40_chan *d40c)
 			= D40_DREG_PRMO_LCHAN_SRC_LOG_DST_LOG,
 	};
 
-	if (d40c->log_num == D40_PHY_CHAN)
+	if (chan_is_physical(d40c))
 		return phy_map[d40c->dma_cfg.mode_opt];
 	else
 		return log_map[d40c->dma_cfg.mode_opt];
@@ -785,7 +910,7 @@ static void d40_config_write(struct d40_chan *d40c)
 	/* Odd addresses are even addresses + 4 */
 	addr_base = (d40c->phy_chan->num % 2) * 4;
 	/* Setup channel mode to logical or physical */
-	var = ((u32)(d40c->log_num != D40_PHY_CHAN) + 1) <<
+	var = ((u32)(chan_is_logical(d40c)) + 1) <<
 		D40_CHAN_POS(d40c->phy_chan->num);
 	writel(var, d40c->base->virtbase + D40_DREG_PRMSE + addr_base);
 
@@ -794,30 +919,18 @@ static void d40_config_write(struct d40_chan *d40c)
 
 	writel(var, d40c->base->virtbase + D40_DREG_PRMOE + addr_base);
 
-	if (d40c->log_num != D40_PHY_CHAN) {
+	if (chan_is_logical(d40c)) {
+		int lidx = (d40c->phy_chan->num << D40_SREG_ELEM_LOG_LIDX_POS)
+			   & D40_SREG_ELEM_LOG_LIDX_MASK;
+		void __iomem *chanbase = chan_base(d40c);
+
 		/* Set default config for CFG reg */
-		writel(d40c->src_def_cfg,
-		       d40c->base->virtbase + D40_DREG_PCBASE +
-		       d40c->phy_chan->num * D40_DREG_PCDELTA +
-		       D40_CHAN_REG_SSCFG);
-		writel(d40c->dst_def_cfg,
-		       d40c->base->virtbase + D40_DREG_PCBASE +
-		       d40c->phy_chan->num * D40_DREG_PCDELTA +
-		       D40_CHAN_REG_SDCFG);
+		writel(d40c->src_def_cfg, chanbase + D40_CHAN_REG_SSCFG);
+		writel(d40c->dst_def_cfg, chanbase + D40_CHAN_REG_SDCFG);
 
 		/* Set LIDX for lcla */
-		writel((d40c->phy_chan->num << D40_SREG_ELEM_LOG_LIDX_POS) &
-		       D40_SREG_ELEM_LOG_LIDX_MASK,
-		       d40c->base->virtbase + D40_DREG_PCBASE +
-		       d40c->phy_chan->num * D40_DREG_PCDELTA +
-		       D40_CHAN_REG_SDELT);
-
-		writel((d40c->phy_chan->num << D40_SREG_ELEM_LOG_LIDX_POS) &
-		       D40_SREG_ELEM_LOG_LIDX_MASK,
-		       d40c->base->virtbase + D40_DREG_PCBASE +
-		       d40c->phy_chan->num * D40_DREG_PCDELTA +
-		       D40_CHAN_REG_SSELT);
-
+		writel(lidx, chanbase + D40_CHAN_REG_SSELT);
+		writel(lidx, chanbase + D40_CHAN_REG_SDELT);
 	}
 }
 
@@ -825,15 +938,15 @@ static u32 d40_residue(struct d40_chan *d40c)
 {
 	u32 num_elt;
 
-	if (d40c->log_num != D40_PHY_CHAN)
+	if (chan_is_logical(d40c))
 		num_elt = (readl(&d40c->lcpa->lcsp2) & D40_MEM_LCSP2_ECNT_MASK)
 			>> D40_MEM_LCSP2_ECNT_POS;
-	else
-		num_elt = (readl(d40c->base->virtbase + D40_DREG_PCBASE +
-				 d40c->phy_chan->num * D40_DREG_PCDELTA +
-				 D40_CHAN_REG_SDELT) &
-			   D40_SREG_ELEM_PHY_ECNT_MASK) >>
-			D40_SREG_ELEM_PHY_ECNT_POS;
+	else {
+		u32 val = readl(chan_base(d40c) + D40_CHAN_REG_SDELT);
+		num_elt = (val & D40_SREG_ELEM_PHY_ECNT_MASK)
+			  >> D40_SREG_ELEM_PHY_ECNT_POS;
+	}
+
 	return num_elt * (1 << d40c->dma_cfg.dst_info.data_width);
 }
 
@@ -841,20 +954,17 @@ static bool d40_tx_is_linked(struct d40_chan *d40c)
 {
 	bool is_link;
 
-	if (d40c->log_num != D40_PHY_CHAN)
+	if (chan_is_logical(d40c))
 		is_link = readl(&d40c->lcpa->lcsp3) &  D40_MEM_LCSP3_DLOS_MASK;
 	else
-		is_link = readl(d40c->base->virtbase + D40_DREG_PCBASE +
-				d40c->phy_chan->num * D40_DREG_PCDELTA +
-				D40_CHAN_REG_SDLNK) &
-			D40_SREG_LNK_PHYS_LNK_MASK;
+		is_link = readl(chan_base(d40c) + D40_CHAN_REG_SDLNK)
+			  & D40_SREG_LNK_PHYS_LNK_MASK;
+
 	return is_link;
 }
 
-static int d40_pause(struct dma_chan *chan)
+static int d40_pause(struct d40_chan *d40c)
 {
-	struct d40_chan *d40c =
-		container_of(chan, struct d40_chan, chan);
 	int res = 0;
 	unsigned long flags;
 
@@ -865,7 +975,7 @@ static int d40_pause(struct dma_chan *chan)
 
 	res = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
 	if (res == 0) {
-		if (d40c->log_num != D40_PHY_CHAN) {
+		if (chan_is_logical(d40c)) {
 			d40_config_set_event(d40c, false);
 			/* Resume the other logical channels if any */
 			if (d40_chan_has_events(d40c))
@@ -878,10 +988,8 @@ static int d40_pause(struct dma_chan *chan)
 	return res;
 }
 
-static int d40_resume(struct dma_chan *chan)
+static int d40_resume(struct d40_chan *d40c)
 {
-	struct d40_chan *d40c =
-		container_of(chan, struct d40_chan, chan);
 	int res = 0;
 	unsigned long flags;
 
@@ -891,7 +999,7 @@ static int d40_resume(struct dma_chan *chan)
 	spin_lock_irqsave(&d40c->lock, flags);
 
 	if (d40c->base->rev == 0)
-		if (d40c->log_num != D40_PHY_CHAN) {
+		if (chan_is_logical(d40c)) {
 			res = d40_channel_execute_command(d40c,
 							  D40_DMA_SUSPEND_REQ);
 			goto no_suspend;
@@ -900,7 +1008,7 @@ static int d40_resume(struct dma_chan *chan)
 	/* If bytes left to transfer or linked tx resume job */
 	if (d40_residue(d40c) || d40_tx_is_linked(d40c)) {
 
-		if (d40c->log_num != D40_PHY_CHAN)
+		if (chan_is_logical(d40c))
 			d40_config_set_event(d40c, true);
 
 		res = d40_channel_execute_command(d40c, D40_DMA_RUN);
@@ -911,75 +1019,20 @@ no_suspend:
 	return res;
 }
 
-static void d40_tx_submit_log(struct d40_chan *d40c, struct d40_desc *d40d)
+static int d40_terminate_all(struct d40_chan *chan)
 {
-	/* TODO: Write */
-}
-
-static void d40_tx_submit_phy(struct d40_chan *d40c, struct d40_desc *d40d)
-{
-	struct d40_desc *d40d_prev = NULL;
-	int i;
-	u32 val;
-
-	if (!list_empty(&d40c->queue))
-		d40d_prev = d40_last_queued(d40c);
-	else if (!list_empty(&d40c->active))
-		d40d_prev = d40_first_active_get(d40c);
-
-	if (!d40d_prev)
-		return;
-
-	/* Here we try to join this job with previous jobs */
-	val = readl(d40c->base->virtbase + D40_DREG_PCBASE +
-		    d40c->phy_chan->num * D40_DREG_PCDELTA +
-		    D40_CHAN_REG_SSLNK);
-
-	/* Figure out which link we're currently transmitting */
-	for (i = 0; i < d40d_prev->lli_len; i++)
-		if (val == d40d_prev->lli_phy.src[i].reg_lnk)
-			break;
-
-	val = readl(d40c->base->virtbase + D40_DREG_PCBASE +
-		    d40c->phy_chan->num * D40_DREG_PCDELTA +
-		    D40_CHAN_REG_SSELT) >> D40_SREG_ELEM_LOG_ECNT_POS;
-
-	if (i == (d40d_prev->lli_len - 1) && val > 0) {
-		/* Change the current one */
-		writel(virt_to_phys(d40d->lli_phy.src),
-		       d40c->base->virtbase + D40_DREG_PCBASE +
-		       d40c->phy_chan->num * D40_DREG_PCDELTA +
-		       D40_CHAN_REG_SSLNK);
-		writel(virt_to_phys(d40d->lli_phy.dst),
-		       d40c->base->virtbase + D40_DREG_PCBASE +
-		       d40c->phy_chan->num * D40_DREG_PCDELTA +
-		       D40_CHAN_REG_SDLNK);
-
-		d40d->is_hw_linked = true;
-
-	} else if (i < d40d_prev->lli_len) {
-		(void) dma_unmap_single(d40c->base->dev,
-					virt_to_phys(d40d_prev->lli_phy.src),
-					d40d_prev->lli_pool.size,
-					DMA_TO_DEVICE);
+	unsigned long flags;
+	int ret = 0;
 
-		/* Keep the settings */
-		val = d40d_prev->lli_phy.src[d40d_prev->lli_len - 1].reg_lnk &
-			~D40_SREG_LNK_PHYS_LNK_MASK;
-		d40d_prev->lli_phy.src[d40d_prev->lli_len - 1].reg_lnk =
-			val | virt_to_phys(d40d->lli_phy.src);
+	ret = d40_pause(chan);
+	if (!ret && chan_is_physical(chan))
+		ret = d40_channel_execute_command(chan, D40_DMA_STOP);
 
-		val = d40d_prev->lli_phy.dst[d40d_prev->lli_len - 1].reg_lnk &
-			~D40_SREG_LNK_PHYS_LNK_MASK;
-		d40d_prev->lli_phy.dst[d40d_prev->lli_len - 1].reg_lnk =
-			val | virt_to_phys(d40d->lli_phy.dst);
+	spin_lock_irqsave(&chan->lock, flags);
+	d40_term_all(chan);
+	spin_unlock_irqrestore(&chan->lock, flags);
 
-		(void) dma_map_single(d40c->base->dev,
-				      d40d_prev->lli_phy.src,
-				      d40d_prev->lli_pool.size,
-				      DMA_TO_DEVICE);
-		d40d->is_hw_linked = true;
-	}
+	return ret;
 }
 
 static dma_cookie_t d40_tx_submit(struct dma_async_tx_descriptor *tx)
@@ -990,8 +1043,6 @@ static dma_cookie_t d40_tx_submit(struct dma_async_tx_descriptor *tx)
 	struct d40_desc *d40d = container_of(tx, struct d40_desc, txd);
 	unsigned long flags;
 
-	(void) d40_pause(&d40c->chan);
-
 	spin_lock_irqsave(&d40c->lock, flags);
 
 	d40c->chan.cookie++;
@@ -1001,17 +1052,10 @@ static dma_cookie_t d40_tx_submit(struct dma_async_tx_descriptor *tx)
 
 	d40d->txd.cookie = d40c->chan.cookie;
 
-	if (d40c->log_num == D40_PHY_CHAN)
-		d40_tx_submit_phy(d40c, d40d);
-	else
-		d40_tx_submit_log(d40c, d40d);
-
 	d40_desc_queue(d40c, d40d);
 
 	spin_unlock_irqrestore(&d40c->lock, flags);
 
-	(void) d40_resume(&d40c->chan);
-
 	return tx->cookie;
 }
 
@@ -1020,7 +1064,7 @@ static int d40_start(struct d40_chan *d40c)
 	if (d40c->base->rev == 0) {
 		int err;
 
-		if (d40c->log_num != D40_PHY_CHAN) {
+		if (chan_is_logical(d40c)) {
 			err = d40_channel_execute_command(d40c,
 							  D40_DMA_SUSPEND_REQ);
 			if (err)
@@ -1028,7 +1072,7 @@ static int d40_start(struct d40_chan *d40c)
 		}
 	}
 
-	if (d40c->log_num != D40_PHY_CHAN)
+	if (chan_is_logical(d40c))
 		d40_config_set_event(d40c, true);
 
 	return d40_channel_execute_command(d40c, D40_DMA_RUN);
@@ -1051,21 +1095,14 @@ static struct d40_desc *d40_queue_start(struct d40_chan *d40c)
 		/* Add to active queue */
 		d40_desc_submit(d40c, d40d);
 
-		/*
-		 * If this job is already linked in hw,
-		 * do not submit it.
-		 */
-
-		if (!d40d->is_hw_linked) {
-			/* Initiate DMA job */
-			d40_desc_load(d40c, d40d);
+		/* Initiate DMA job */
+		d40_desc_load(d40c, d40d);
 
-			/* Start dma job */
-			err = d40_start(d40c);
+		/* Start dma job */
+		err = d40_start(d40c);
 
-			if (err)
-				return NULL;
-		}
+		if (err)
+			return NULL;
 	}
 
 	return d40d;
@@ -1082,17 +1119,36 @@ static void dma_tc_handle(struct d40_chan *d40c)
 	if (d40d == NULL)
 		return;
 
-	d40_lcla_free_all(d40c, d40d);
+	if (d40d->cyclic) {
+		/*
+		 * If this was a paritially loaded list, we need to reloaded
+		 * it, and only when the list is completed.  We need to check
+		 * for done because the interrupt will hit for every link, and
+		 * not just the last one.
+		 */
+		if (d40d->lli_current < d40d->lli_len
+		    && !d40_tx_is_linked(d40c)
+		    && !d40_residue(d40c)) {
+			d40_lcla_free_all(d40c, d40d);
+			d40_desc_load(d40c, d40d);
+			(void) d40_start(d40c);
 
-	if (d40d->lli_current < d40d->lli_len) {
-		d40_desc_load(d40c, d40d);
-		/* Start dma job */
-		(void) d40_start(d40c);
-		return;
-	}
+			if (d40d->lli_current == d40d->lli_len)
+				d40d->lli_current = 0;
+		}
+	} else {
+		d40_lcla_free_all(d40c, d40d);
 
-	if (d40_queue_start(d40c) == NULL)
-		d40c->busy = false;
+		if (d40d->lli_current < d40d->lli_len) {
+			d40_desc_load(d40c, d40d);
+			/* Start dma job */
+			(void) d40_start(d40c);
+			return;
+		}
+
+		if (d40_queue_start(d40c) == NULL)
+			d40c->busy = false;
+	}
 
 	d40c->pending_tx++;
 	tasklet_schedule(&d40c->tasklet);
@@ -1111,11 +1167,11 @@ static void dma_tasklet(unsigned long data)
 
 	/* Get first active entry from list */
 	d40d = d40_first_active_get(d40c);
-
 	if (d40d == NULL)
 		goto err;
 
-	d40c->completed = d40d->txd.cookie;
+	if (!d40d->cyclic)
+		d40c->completed = d40d->txd.cookie;
 
 	/*
 	 * If terminating a channel pending_tx is set to zero.
@@ -1130,16 +1186,18 @@ static void dma_tasklet(unsigned long data)
 	callback = d40d->txd.callback;
 	callback_param = d40d->txd.callback_param;
 
-	if (async_tx_test_ack(&d40d->txd)) {
-		d40_pool_lli_free(d40d);
-		d40_desc_remove(d40d);
-		d40_desc_free(d40c, d40d);
-	} else {
-		if (!d40d->is_in_client_list) {
+	if (!d40d->cyclic) {
+		if (async_tx_test_ack(&d40d->txd)) {
+			d40_pool_lli_free(d40c, d40d);
 			d40_desc_remove(d40d);
-			d40_lcla_free_all(d40c, d40d);
-			list_add_tail(&d40d->node, &d40c->client);
-			d40d->is_in_client_list = true;
+			d40_desc_free(d40c, d40d);
+		} else {
+			if (!d40d->is_in_client_list) {
+				d40_desc_remove(d40d);
+				d40_lcla_free_all(d40c, d40d);
+				list_add_tail(&d40d->node, &d40c->client);
+				d40d->is_in_client_list = true;
+			}
 		}
 	}
 
@@ -1216,9 +1274,8 @@ static irqreturn_t d40_handle_interrupt(int irq, void *data)
 		if (!il[row].is_error)
 			dma_tc_handle(d40c);
 		else
-			dev_err(base->dev,
-				"[%s] IRQ chan: %ld offset %d idx %d\n",
-				__func__, chan, il[row].offset, idx);
+			d40_err(base->dev, "IRQ chan: %ld offset %d idx %d\n",
+				chan, il[row].offset, idx);
 
 		spin_unlock(&d40c->lock);
 	}
@@ -1237,8 +1294,7 @@ static int d40_validate_conf(struct d40_chan *d40c,
 	bool is_log = conf->mode == STEDMA40_MODE_LOGICAL;
 
 	if (!conf->dir) {
-		dev_err(&d40c->chan.dev->device, "[%s] Invalid direction.\n",
-			__func__);
+		chan_err(d40c, "Invalid direction.\n");
 		res = -EINVAL;
 	}
 
@@ -1246,46 +1302,40 @@ static int d40_validate_conf(struct d40_chan *d40c,
 	    d40c->base->plat_data->dev_tx[conf->dst_dev_type] == 0 &&
 	    d40c->runtime_addr == 0) {
 
-		dev_err(&d40c->chan.dev->device,
-			"[%s] Invalid TX channel address (%d)\n",
-			__func__, conf->dst_dev_type);
+		chan_err(d40c, "Invalid TX channel address (%d)\n",
+			 conf->dst_dev_type);
 		res = -EINVAL;
 	}
 
 	if (conf->src_dev_type != STEDMA40_DEV_SRC_MEMORY &&
 	    d40c->base->plat_data->dev_rx[conf->src_dev_type] == 0 &&
 	    d40c->runtime_addr == 0) {
-		dev_err(&d40c->chan.dev->device,
-			"[%s] Invalid RX channel address (%d)\n",
-			__func__, conf->src_dev_type);
+		chan_err(d40c, "Invalid RX channel address (%d)\n",
+			conf->src_dev_type);
 		res = -EINVAL;
 	}
 
 	if (conf->dir == STEDMA40_MEM_TO_PERIPH &&
 	    dst_event_group == STEDMA40_DEV_DST_MEMORY) {
-		dev_err(&d40c->chan.dev->device, "[%s] Invalid dst\n",
-			__func__);
+		chan_err(d40c, "Invalid dst\n");
 		res = -EINVAL;
 	}
 
 	if (conf->dir == STEDMA40_PERIPH_TO_MEM &&
 	    src_event_group == STEDMA40_DEV_SRC_MEMORY) {
-		dev_err(&d40c->chan.dev->device, "[%s] Invalid src\n",
-			__func__);
+		chan_err(d40c, "Invalid src\n");
 		res = -EINVAL;
 	}
 
 	if (src_event_group == STEDMA40_DEV_SRC_MEMORY &&
 	    dst_event_group == STEDMA40_DEV_DST_MEMORY && is_log) {
-		dev_err(&d40c->chan.dev->device,
-			"[%s] No event line\n", __func__);
+		chan_err(d40c, "No event line\n");
 		res = -EINVAL;
 	}
 
 	if (conf->dir == STEDMA40_PERIPH_TO_PERIPH &&
 	    (src_event_group != dst_event_group)) {
-		dev_err(&d40c->chan.dev->device,
-			"[%s] Invalid event group\n", __func__);
+		chan_err(d40c, "Invalid event group\n");
 		res = -EINVAL;
 	}
 
@@ -1294,9 +1344,7 @@ static int d40_validate_conf(struct d40_chan *d40c,
 		 * DMAC HW supports it. Will be added to this driver,
 		 * in case any dma client requires it.
 		 */
-		dev_err(&d40c->chan.dev->device,
-			"[%s] periph to periph not supported\n",
-			__func__);
+		chan_err(d40c, "periph to periph not supported\n");
 		res = -EINVAL;
 	}
 
@@ -1309,9 +1357,7 @@ static int d40_validate_conf(struct d40_chan *d40c,
 		 * src (burst x width) == dst (burst x width)
 		 */
 
-		dev_err(&d40c->chan.dev->device,
-			"[%s] src (burst x width) != dst (burst x width)\n",
-			__func__);
+		chan_err(d40c, "src (burst x width) != dst (burst x width)\n");
 		res = -EINVAL;
 	}
 
@@ -1514,8 +1560,7 @@ static int d40_config_memcpy(struct d40_chan *d40c)
 		   dma_has_cap(DMA_SLAVE, cap)) {
 		d40c->dma_cfg = *d40c->base->plat_data->memcpy_conf_phy;
 	} else {
-		dev_err(&d40c->chan.dev->device, "[%s] No memcpy\n",
-			__func__);
+		chan_err(d40c, "No memcpy\n");
 		return -EINVAL;
 	}
 
@@ -1540,21 +1585,19 @@ static int d40_free_dma(struct d40_chan *d40c)
 	/* Release client owned descriptors */
 	if (!list_empty(&d40c->client))
 		list_for_each_entry_safe(d, _d, &d40c->client, node) {
-			d40_pool_lli_free(d);
+			d40_pool_lli_free(d40c, d);
 			d40_desc_remove(d);
 			d40_desc_free(d40c, d);
 		}
 
 	if (phy == NULL) {
-		dev_err(&d40c->chan.dev->device, "[%s] phy == null\n",
-			__func__);
+		chan_err(d40c, "phy == null\n");
 		return -EINVAL;
 	}
 
 	if (phy->allocated_src == D40_ALLOC_FREE &&
 	    phy->allocated_dst == D40_ALLOC_FREE) {
-		dev_err(&d40c->chan.dev->device, "[%s] channel already free\n",
-			__func__);
+		chan_err(d40c, "channel already free\n");
 		return -EINVAL;
 	}
 
@@ -1566,19 +1609,17 @@ static int d40_free_dma(struct d40_chan *d40c)
 		event = D40_TYPE_TO_EVENT(d40c->dma_cfg.src_dev_type);
 		is_src = true;
 	} else {
-		dev_err(&d40c->chan.dev->device,
-			"[%s] Unknown direction\n", __func__);
+		chan_err(d40c, "Unknown direction\n");
 		return -EINVAL;
 	}
 
 	res = d40_channel_execute_command(d40c, D40_DMA_SUSPEND_REQ);
 	if (res) {
-		dev_err(&d40c->chan.dev->device, "[%s] suspend failed\n",
-			__func__);
+		chan_err(d40c, "suspend failed\n");
 		return res;
 	}
 
-	if (d40c->log_num != D40_PHY_CHAN) {
+	if (chan_is_logical(d40c)) {
 		/* Release logical channel, deactivate the event line */
 
 		d40_config_set_event(d40c, false);
@@ -1594,9 +1635,8 @@ static int d40_free_dma(struct d40_chan *d40c)
 				res = d40_channel_execute_command(d40c,
 								  D40_DMA_RUN);
 				if (res) {
-					dev_err(&d40c->chan.dev->device,
-						"[%s] Executing RUN command\n",
-						__func__);
+					chan_err(d40c,
+						"Executing RUN command\n");
 					return res;
 				}
 			}
@@ -1609,8 +1649,7 @@ static int d40_free_dma(struct d40_chan *d40c)
 	/* Release physical channel */
 	res = d40_channel_execute_command(d40c, D40_DMA_STOP);
 	if (res) {
-		dev_err(&d40c->chan.dev->device,
-			"[%s] Failed to stop channel\n", __func__);
+		chan_err(d40c, "Failed to stop channel\n");
 		return res;
 	}
 	d40c->phy_chan = NULL;
@@ -1622,6 +1661,7 @@ static int d40_free_dma(struct d40_chan *d40c)
 
 static bool d40_is_paused(struct d40_chan *d40c)
 {
+	void __iomem *chanbase = chan_base(d40c);
 	bool is_paused = false;
 	unsigned long flags;
 	void __iomem *active_reg;
@@ -1630,7 +1670,7 @@ static bool d40_is_paused(struct d40_chan *d40c)
 
 	spin_lock_irqsave(&d40c->lock, flags);
 
-	if (d40c->log_num == D40_PHY_CHAN) {
+	if (chan_is_physical(d40c)) {
 		if (d40c->phy_chan->num % 2 == 0)
 			active_reg = d40c->base->virtbase + D40_DREG_ACTIVE;
 		else
@@ -1648,17 +1688,12 @@ static bool d40_is_paused(struct d40_chan *d40c)
 	if (d40c->dma_cfg.dir == STEDMA40_MEM_TO_PERIPH ||
 	    d40c->dma_cfg.dir == STEDMA40_MEM_TO_MEM) {
 		event = D40_TYPE_TO_EVENT(d40c->dma_cfg.dst_dev_type);
-		status = readl(d40c->base->virtbase + D40_DREG_PCBASE +
-			       d40c->phy_chan->num * D40_DREG_PCDELTA +
-			       D40_CHAN_REG_SDLNK);
+		status = readl(chanbase + D40_CHAN_REG_SDLNK);
 	} else if (d40c->dma_cfg.dir == STEDMA40_PERIPH_TO_MEM) {
 		event = D40_TYPE_TO_EVENT(d40c->dma_cfg.src_dev_type);
-		status = readl(d40c->base->virtbase + D40_DREG_PCBASE +
-			       d40c->phy_chan->num * D40_DREG_PCDELTA +
-			       D40_CHAN_REG_SSLNK);
+		status = readl(chanbase + D40_CHAN_REG_SSLNK);
 	} else {
-		dev_err(&d40c->chan.dev->device,
-			"[%s] Unknown direction\n", __func__);
+		chan_err(d40c, "Unknown direction\n");
 		goto _exit;
 	}
 
@@ -1688,114 +1723,184 @@ static u32 stedma40_residue(struct dma_chan *chan