Merge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx

* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx: (63 commits)
  ARM: PL08x: cleanup comments
  Update CONFIG_MD_RAID6_PQ to CONFIG_RAID6_PQ in drivers/dma/iop-adma.c
  ARM: PL08x: fix a warning
  Fix dmaengine_submit() return type
  dmaengine: at_hdmac: fix race while monitoring channel status
  dmaengine: at_hdmac: flags located in first descriptor
  dmaengine: at_hdmac: use subsys_initcall instead of module_init
  dmaengine: at_hdmac: no need set ACK in new descriptor
  dmaengine: at_hdmac: trivial add precision to unmapping comment
  dmaengine: at_hdmac: use dma_address to program DMA hardware
  pch_dma: support new device ML7213 IOH
  ARM: PL08x: prevent dma_set_runtime_config() reconfiguring memcpy channels
  ARM: PL08x: allow dma_set_runtime_config() to return errors
  ARM: PL08x: fix locking between prepare function and submit function
  ARM: PL08x: introduce 'phychan_hold' to hold on to physical channels
  ARM: PL08x: put txd's on the pending list in pl08x_tx_submit()
  ARM: PL08x: rename 'desc_list' as 'pend_list'
  ARM: PL08x: implement unmapping of memcpy buffers
  ARM: PL08x: store prep_* flags in async_tx structure
  ARM: PL08x: shrink srcbus/dstbus in txd structure
  ...

arch/arm/plat-nomadik/include/plat/ste_dma40.h

@@ -13,6 +13,14 @@
 #include <linux/workqueue.h>
 #include <linux/interrupt.h>
 
+/*
+ * Maxium size for a single dma descriptor
+ * Size is limited to 16 bits.
+ * Size is in the units of addr-widths (1,2,4,8 bytes)
+ * Larger transfers will be split up to multiple linked desc
+ */
+#define STEDMA40_MAX_SEG_SIZE 0xFFFF
+
 /* dev types for memcpy */
 #define STEDMA40_DEV_DST_MEMORY (-1)
 #define	STEDMA40_DEV_SRC_MEMORY (-1)

drivers/dma/Kconfig

@@ -200,11 +200,16 @@ config PL330_DMA
 	  platform_data for a dma-pl330 device.
 
 config PCH_DMA
-	tristate "Topcliff (Intel EG20T) PCH DMA support"
+	tristate "Intel EG20T PCH / OKI SEMICONDUCTOR ML7213 IOH DMA support"
 	depends on PCI && X86
 	select DMA_ENGINE
 	help
-	  Enable support for the Topcliff (Intel EG20T) PCH DMA engine.
+	  Enable support for Intel EG20T PCH DMA engine.
+
+	  This driver also can be used for OKI SEMICONDUCTOR ML7213 IOH(Input/
+	  Output Hub) which is for IVI(In-Vehicle Infotainment) use.
+	  ML7213 is companion chip for Intel Atom E6xx series.
+	  ML7213 is completely compatible for Intel EG20T PCH.
 
 config IMX_SDMA
 	tristate "i.MX SDMA support"

drivers/dma/at_hdmac.c

@@ -253,7 +253,7 @@ atc_chain_complete(struct at_dma_chan *atchan, struct at_desc *desc)
 	/* move myself to free_list */
 	list_move(&desc->desc_node, &atchan->free_list);
 
-	/* unmap dma addresses */
+	/* unmap dma addresses (not on slave channels) */
 	if (!atchan->chan_common.private) {
 		struct device *parent = chan2parent(&atchan->chan_common);
 		if (!(txd->flags & DMA_COMPL_SKIP_DEST_UNMAP)) {
@@ -583,7 +583,6 @@ atc_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src,
 		desc->lli.ctrlb = ctrlb;
 
 		desc->txd.cookie = 0;
-		async_tx_ack(&desc->txd);
 
 		if (!first) {
 			first = desc;
@@ -604,7 +603,7 @@ atc_prep_dma_memcpy(struct dma_chan *chan, dma_addr_t dest, dma_addr_t src,
 	/* set end-of-link to the last link descriptor of list*/
 	set_desc_eol(desc);
 
-	desc->txd.flags = flags; /* client is in control of this ack */
+	first->txd.flags = flags; /* client is in control of this ack */
 
 	return &first->txd;
 
@@ -670,7 +669,7 @@ atc_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 			if (!desc)
 				goto err_desc_get;
 
-			mem = sg_phys(sg);
+			mem = sg_dma_address(sg);
 			len = sg_dma_len(sg);
 			mem_width = 2;
 			if (unlikely(mem & 3 || len & 3))
@@ -712,7 +711,7 @@ atc_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 			if (!desc)
 				goto err_desc_get;
 
-			mem = sg_phys(sg);
+			mem = sg_dma_address(sg);
 			len = sg_dma_len(sg);
 			mem_width = 2;
 			if (unlikely(mem & 3 || len & 3))
@@ -749,8 +748,8 @@ atc_prep_slave_sg(struct dma_chan *chan, struct scatterlist *sgl,
 	first->txd.cookie = -EBUSY;
 	first->len = total_len;
 
-	/* last link descriptor of list is responsible of flags */
-	prev->txd.flags = flags; /* client is in control of this ack */
+	/* first link descriptor of list is responsible of flags */
+	first->txd.flags = flags; /* client is in control of this ack */
 
 	return &first->txd;
 
@@ -854,11 +853,11 @@ static void atc_issue_pending(struct dma_chan *chan)
 
 	dev_vdbg(chan2dev(chan), "issue_pending\n");
 
-	if (!atc_chan_is_enabled(atchan)) {
 	spin_lock_bh(&atchan->lock);
+	if (!atc_chan_is_enabled(atchan)) {
 		atc_advance_work(atchan);
-		spin_unlock_bh(&atchan->lock);
 	}
+	spin_unlock_bh(&atchan->lock);
 }
 
 /**
@@ -1210,7 +1209,7 @@ static int __init at_dma_init(void)
 {
 	return platform_driver_probe(&at_dma_driver, at_dma_probe);
 }
-module_init(at_dma_init);
+subsys_initcall(at_dma_init);
 
 static void __exit at_dma_exit(void)
 {

drivers/dma/fsldma.c

 /*
  * Freescale MPC85xx, MPC83xx DMA Engine support
  *
- * Copyright (C) 2007 Freescale Semiconductor, Inc. All rights reserved.
+ * Copyright (C) 2007-2010 Freescale Semiconductor, Inc. All rights reserved.
  *
  * Author:
  *   Zhang Wei <wei.zhang@freescale.com>, Jul 2007
@@ -1324,6 +1324,8 @@ static int __devinit fsldma_of_probe(struct platform_device *op,
 	fdev->common.device_control = fsl_dma_device_control;
 	fdev->common.dev = &op->dev;
 
+	dma_set_mask(&(op->dev), DMA_BIT_MASK(36));
+
 	dev_set_drvdata(&op->dev, fdev);
 
 	/*

drivers/dma/intel_mid_dma.c

@@ -664,11 +664,20 @@ static struct dma_async_tx_descriptor *intel_mid_dma_prep_memcpy(
 	/*calculate CTL_LO*/
 	ctl_lo.ctl_lo = 0;
 	ctl_lo.ctlx.int_en = 1;
-	ctl_lo.ctlx.dst_tr_width = mids->dma_slave.dst_addr_width;
-	ctl_lo.ctlx.src_tr_width = mids->dma_slave.src_addr_width;
 	ctl_lo.ctlx.dst_msize = mids->dma_slave.src_maxburst;
 	ctl_lo.ctlx.src_msize = mids->dma_slave.dst_maxburst;
 
+	/*
+	 * Here we need some translation from "enum dma_slave_buswidth"
+	 * to the format for our dma controller
+	 *		standard	intel_mid_dmac's format
+	 *		 1 Byte			0b000
+	 *		 2 Bytes		0b001
+	 *		 4 Bytes		0b010
+	 */
+	ctl_lo.ctlx.dst_tr_width = mids->dma_slave.dst_addr_width / 2;
+	ctl_lo.ctlx.src_tr_width = mids->dma_slave.src_addr_width / 2;
+
 	if (mids->cfg_mode == LNW_DMA_MEM_TO_MEM) {
 		ctl_lo.ctlx.tt_fc = 0;
 		ctl_lo.ctlx.sinc = 0;
@@ -746,9 +755,19 @@ static struct dma_async_tx_descriptor *intel_mid_dma_prep_slave_sg(
 	BUG_ON(!mids);
 
 	if (!midc->dma->pimr_mask) {
-		pr_debug("MDMA: SG list is not supported by this controller\n");
+		/* We can still handle sg list with only one item */
+		if (sg_len == 1) {
+			txd = intel_mid_dma_prep_memcpy(chan,
+						mids->dma_slave.dst_addr,
+						mids->dma_slave.src_addr,
+						sgl->length,
+						flags);
+			return txd;
+		} else {
+			pr_warn("MDMA: SG list is not supported by this controller\n");
 			return  NULL;
 		}
+	}
 
 	pr_debug("MDMA: SG Length = %d, direction = %d, Flags = %#lx\n",
 			sg_len, direction, flags);
@@ -758,6 +777,7 @@ static struct dma_async_tx_descriptor *intel_mid_dma_prep_slave_sg(
 		pr_err("MDMA: Prep memcpy failed\n");
 		return NULL;
 	}
+
 	desc = to_intel_mid_dma_desc(txd);
 	desc->dirn = direction;
 	ctl_lo.ctl_lo = desc->ctl_lo;
@@ -1021,11 +1041,6 @@ static irqreturn_t intel_mid_dma_interrupt(int irq, void *data)
 
 	/*DMA Interrupt*/
 	pr_debug("MDMA:Got an interrupt on irq %d\n", irq);
-	if (!mid) {
-		pr_err("ERR_MDMA:null pointer mid\n");
-		return -EINVAL;
-	}
-
 	pr_debug("MDMA: Status %x, Mask %x\n", tfr_status, mid->intr_mask);
 	tfr_status &= mid->intr_mask;
 	if (tfr_status) {

drivers/dma/iop-adma.c

@@ -1261,7 +1261,7 @@ iop_adma_xor_val_self_test(struct iop_adma_device *device)
 	return err;
 }
 
-#ifdef CONFIG_MD_RAID6_PQ
+#ifdef CONFIG_RAID6_PQ
 static int __devinit
 iop_adma_pq_zero_sum_self_test(struct iop_adma_device *device)
 {
@@ -1584,7 +1584,7 @@ static int __devinit iop_adma_probe(struct platform_device *pdev)
 
 	if (dma_has_cap(DMA_PQ, dma_dev->cap_mask) &&
 	    dma_has_cap(DMA_PQ_VAL, dma_dev->cap_mask)) {
-		#ifdef CONFIG_MD_RAID6_PQ
+		#ifdef CONFIG_RAID6_PQ
 		ret = iop_adma_pq_zero_sum_self_test(adev);
 		dev_dbg(&pdev->dev, "pq self test returned %d\n", ret);
 		#else

drivers/dma/pch_dma.c

 /*
  * Topcliff PCH DMA controller driver
  * Copyright (c) 2010 Intel Corporation
+ * Copyright (C) 2011 OKI SEMICONDUCTOR CO., LTD.
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License version 2 as
@@ -921,12 +922,19 @@ static void __devexit pch_dma_remove(struct pci_dev *pdev)
 }
 
 /* PCI Device ID of DMA device */
-#define PCI_DEVICE_ID_PCH_DMA_8CH        0x8810
-#define PCI_DEVICE_ID_PCH_DMA_4CH        0x8815
+#define PCI_VENDOR_ID_ROHM             0x10DB
+#define PCI_DEVICE_ID_EG20T_PCH_DMA_8CH        0x8810
+#define PCI_DEVICE_ID_EG20T_PCH_DMA_4CH        0x8815
+#define PCI_DEVICE_ID_ML7213_DMA1_8CH	0x8026
+#define PCI_DEVICE_ID_ML7213_DMA2_8CH	0x802B
+#define PCI_DEVICE_ID_ML7213_DMA3_4CH	0x8034
 
 static const struct pci_device_id pch_dma_id_table[] = {
-	{ PCI_VDEVICE(INTEL, PCI_DEVICE_ID_PCH_DMA_8CH), 8 },
-	{ PCI_VDEVICE(INTEL, PCI_DEVICE_ID_PCH_DMA_4CH), 4 },
+	{ PCI_VDEVICE(INTEL, PCI_DEVICE_ID_EG20T_PCH_DMA_8CH), 8 },
+	{ PCI_VDEVICE(INTEL, PCI_DEVICE_ID_EG20T_PCH_DMA_4CH), 4 },
+	{ PCI_VDEVICE(ROHM, PCI_DEVICE_ID_ML7213_DMA1_8CH), 8}, /* UART Video */
+	{ PCI_VDEVICE(ROHM, PCI_DEVICE_ID_ML7213_DMA2_8CH), 8}, /* PCMIF SPI */
+	{ PCI_VDEVICE(ROHM, PCI_DEVICE_ID_ML7213_DMA3_4CH), 4}, /* FPGA */
 	{ 0, },
 };
 
@@ -954,6 +962,7 @@ static void __exit pch_dma_exit(void)
 module_init(pch_dma_init);
 module_exit(pch_dma_exit);
 
-MODULE_DESCRIPTION("Topcliff PCH DMA controller driver");
+MODULE_DESCRIPTION("Intel EG20T PCH / OKI SEMICONDUCTOR ML7213 IOH "
+		   "DMA controller driver");
 MODULE_AUTHOR("Yong Wang <yong.y.wang@intel.com>");
 MODULE_LICENSE("GPL v2");

drivers/dma/ste_dma40.c

 /*
- * Copyright (C) ST-Ericsson SA 2007-2010
+ * Copyright (C) Ericsson AB 2007-2008
+ * Copyright (C) ST-Ericsson SA 2008-2010
  * Author: Per Forlin <per.forlin@stericsson.com> for ST-Ericsson
  * Author: Jonas Aaberg <jonas.aberg@stericsson.com> for ST-Ericsson
  * License terms: GNU General Public License (GPL) version 2
@@ -554,8 +555,66 @@ static struct d40_desc *d40_last_queued(struct d40_chan *d40c)
 	return d;
 }
 
-/* Support functions for logical channels */
+static int d40_psize_2_burst_size(bool is_log, int psize)
+{
+	if (is_log) {
+		if (psize == STEDMA40_PSIZE_LOG_1)
+			return 1;
+	} else {
+		if (psize == STEDMA40_PSIZE_PHY_1)
+			return 1;
+	}
+
+	return 2 << psize;
+}
+
+/*
+ * The dma only supports transmitting packages up to
+ * STEDMA40_MAX_SEG_SIZE << data_width. Calculate the total number of
+ * dma elements required to send the entire sg list
+ */
+static int d40_size_2_dmalen(int size, u32 data_width1, u32 data_width2)
+{
+	int dmalen;
+	u32 max_w = max(data_width1, data_width2);
+	u32 min_w = min(data_width1, data_width2);
+	u32 seg_max = ALIGN(STEDMA40_MAX_SEG_SIZE << min_w, 1 << max_w);
+
+	if (seg_max > STEDMA40_MAX_SEG_SIZE)
+		seg_max -= (1 << max_w);
+
+	if (!IS_ALIGNED(size, 1 << max_w))
+		return -EINVAL;
+
+	if (size <= seg_max)
+		dmalen = 1;
+	else {
+		dmalen = size / seg_max;
+		if (dmalen * seg_max < size)
+			dmalen++;
+	}
+	return dmalen;
+}
+
+static int d40_sg_2_dmalen(struct scatterlist *sgl, int sg_len,
+			   u32 data_width1, u32 data_width2)
+{
+	struct scatterlist *sg;
+	int i;
+	int len = 0;
+	int ret;
+
+	for_each_sg(sgl, sg, sg_len, i) {
+		ret = d40_size_2_dmalen(sg_dma_len(sg),
+					data_width1, data_width2);
+		if (ret < 0)
+			return ret;
+		len += ret;
+	}
+	return len;
+}
 
+/* Support functions for logical channels */
 
 static int d40_channel_execute_command(struct d40_chan *d40c,
 				       enum d40_command command)
@@ -1241,6 +1300,21 @@ static int d40_validate_conf(struct d40_chan *d40c,
 		res = -EINVAL;
 	}
 
+	if (d40_psize_2_burst_size(is_log, conf->src_info.psize) *
+	    (1 << conf->src_info.data_width) !=
+	    d40_psize_2_burst_size(is_log, conf->dst_info.psize) *
+	    (1 << conf->dst_info.data_width)) {
+		/*
+		 * The DMAC hardware only supports
+		 * 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__);
+		res = -EINVAL;
+	}
+
 	return res;
 }
 
@@ -1638,13 +1712,21 @@ struct dma_async_tx_descriptor *stedma40_memcpy_sg(struct dma_chan *chan,
 	if (d40d == NULL)
 		goto err;
 
-	d40d->lli_len = sgl_len;
+	d40d->lli_len = d40_sg_2_dmalen(sgl_dst, sgl_len,
+					d40c->dma_cfg.src_info.data_width,
+					d40c->dma_cfg.dst_info.data_width);
+	if (d40d->lli_len < 0) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Unaligned size\n", __func__);
+		goto err;
+	}
+
 	d40d->lli_current = 0;
 	d40d->txd.flags = dma_flags;
 
 	if (d40c->log_num != D40_PHY_CHAN) {
 
-		if (d40_pool_lli_alloc(d40d, sgl_len, true) < 0) {
+		if (d40_pool_lli_alloc(d40d, d40d->lli_len, true) < 0) {
 			dev_err(&d40c->chan.dev->device,
 				"[%s] Out of memory\n", __func__);
 			goto err;
@@ -1654,15 +1736,17 @@ struct dma_async_tx_descriptor *stedma40_memcpy_sg(struct dma_chan *chan,
 					 sgl_len,
 					 d40d->lli_log.src,
 					 d40c->log_def.lcsp1,
-					 d40c->dma_cfg.src_info.data_width);
+					 d40c->dma_cfg.src_info.data_width,
+					 d40c->dma_cfg.dst_info.data_width);
 
 		(void) d40_log_sg_to_lli(sgl_dst,
 					 sgl_len,
 					 d40d->lli_log.dst,
 					 d40c->log_def.lcsp3,
-					 d40c->dma_cfg.dst_info.data_width);
+					 d40c->dma_cfg.dst_info.data_width,
+					 d40c->dma_cfg.src_info.data_width);
 	} else {
-		if (d40_pool_lli_alloc(d40d, sgl_len, false) < 0) {
+		if (d40_pool_lli_alloc(d40d, d40d->lli_len, false) < 0) {
 			dev_err(&d40c->chan.dev->device,
 				"[%s] Out of memory\n", __func__);
 			goto err;
@@ -1675,6 +1759,7 @@ struct dma_async_tx_descriptor *stedma40_memcpy_sg(struct dma_chan *chan,
 					virt_to_phys(d40d->lli_phy.src),
 					d40c->src_def_cfg,
 					d40c->dma_cfg.src_info.data_width,
+					d40c->dma_cfg.dst_info.data_width,
 					d40c->dma_cfg.src_info.psize);
 
 		if (res < 0)
@@ -1687,6 +1772,7 @@ struct dma_async_tx_descriptor *stedma40_memcpy_sg(struct dma_chan *chan,
 					virt_to_phys(d40d->lli_phy.dst),
 					d40c->dst_def_cfg,
 					d40c->dma_cfg.dst_info.data_width,
+					d40c->dma_cfg.src_info.data_width,
 					d40c->dma_cfg.dst_info.psize);
 
 		if (res < 0)
@@ -1826,7 +1912,6 @@ static struct dma_async_tx_descriptor *d40_prep_memcpy(struct dma_chan *chan,
 	struct d40_chan *d40c = container_of(chan, struct d40_chan,
 					     chan);
 	unsigned long flags;
-	int err = 0;
 
 	if (d40c->phy_chan == NULL) {
 		dev_err(&d40c->chan.dev->device,
@@ -1844,6 +1929,15 @@ static struct dma_async_tx_descriptor *d40_prep_memcpy(struct dma_chan *chan,
 	}
 
 	d40d->txd.flags = dma_flags;
+	d40d->lli_len = d40_size_2_dmalen(size,
+					  d40c->dma_cfg.src_info.data_width,
+					  d40c->dma_cfg.dst_info.data_width);
+	if (d40d->lli_len < 0) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Unaligned size\n", __func__);
+		goto err;
+	}
+
 
 	dma_async_tx_descriptor_init(&d40d->txd, chan);
 
@@ -1851,37 +1945,40 @@ static struct dma_async_tx_descriptor *d40_prep_memcpy(struct dma_chan *chan,
 
 	if (d40c->log_num != D40_PHY_CHAN) {
 
-		if (d40_pool_lli_alloc(d40d, 1, true) < 0) {
+		if (d40_pool_lli_alloc(d40d, d40d->lli_len, true) < 0) {
 			dev_err(&d40c->chan.dev->device,
 				"[%s] Out of memory\n", __func__);
 			goto err;
 		}
-		d40d->lli_len = 1;
 		d40d->lli_current = 0;
 
-		d40_log_fill_lli(d40d->lli_log.src,
+		if (d40_log_buf_to_lli(d40d->lli_log.src,
 				       src,
 				       size,
 				       d40c->log_def.lcsp1,
 				       d40c->dma_cfg.src_info.data_width,
-				 true);
+				       d40c->dma_cfg.dst_info.data_width,
+				       true) == NULL)
+			goto err;
 
-		d40_log_fill_lli(d40d->lli_log.dst,
+		if (d40_log_buf_to_lli(d40d->lli_log.dst,
 				       dst,
 				       size,
 				       d40c->log_def.lcsp3,
 				       d40c->dma_cfg.dst_info.data_width,
-				 true);
+				       d40c->dma_cfg.src_info.data_width,
+				       true) == NULL)
+			goto err;
 
 	} else {
 
-		if (d40_pool_lli_alloc(d40d, 1, false) < 0) {
+		if (d40_pool_lli_alloc(d40d, d40d->lli_len, false) < 0) {
 			dev_err(&d40c->chan.dev->device,
 				"[%s] Out of memory\n", __func__);
 			goto err;
 		}
 
-		err = d40_phy_fill_lli(d40d->lli_phy.src,
+		if (d40_phy_buf_to_lli(d40d->lli_phy.src,
 				       src,
 				       size,
 				       d40c->dma_cfg.src_info.psize,
@@ -1889,11 +1986,11 @@ static struct dma_async_tx_descriptor *d40_prep_memcpy(struct dma_chan *chan,
 				       d40c->src_def_cfg,
 				       true,
 				       d40c->dma_cfg.src_info.data_width,
-				       false);
-		if (err)
-			goto err_fill_lli;
+				       d40c->dma_cfg.dst_info.data_width,
+				       false) == NULL)
+			goto err;
 
-		err = d40_phy_fill_lli(d40d->lli_phy.dst,
+		if (d40_phy_buf_to_lli(d40d->lli_phy.dst,
 				       dst,
 				       size,
 				       d40c->dma_cfg.dst_info.psize,
@@ -1901,10 +1998,9 @@ static struct dma_async_tx_descriptor *d40_prep_memcpy(struct dma_chan *chan,
 				       d40c->dst_def_cfg,
 				       true,
 				       d40c->dma_cfg.dst_info.data_width,
-				       false);
-
-		if (err)
-			goto err_fill_lli;
+				       d40c->dma_cfg.src_info.data_width,
+				       false) == NULL)
+			goto err;
 
 		(void) dma_map_single(d40c->base->dev, d40d->lli_phy.src,
 				      d40d->lli_pool.size, DMA_TO_DEVICE);
@@ -1913,9 +2009,6 @@ static struct dma_async_tx_descriptor *d40_prep_memcpy(struct dma_chan *chan,
 	spin_unlock_irqrestore(&d40c->lock, flags);
 	return &d40d->txd;
 
-err_fill_lli:
-	dev_err(&d40c->chan.dev->device,
-		"[%s] Failed filling in PHY LLI\n", __func__);
 err:
 	if (d40d)
 		d40_desc_free(d40c, d40d);
@@ -1945,13 +2038,21 @@ static int d40_prep_slave_sg_log(struct d40_desc *d40d,
 	dma_addr_t dev_addr = 0;
 	int total_size;
 
-	if (d40_pool_lli_alloc(d40d, sg_len, true) < 0) {
+	d40d->lli_len = d40_sg_2_dmalen(sgl, sg_len,
+					d40c->dma_cfg.src_info.data_width,
+					d40c->dma_cfg.dst_info.data_width);
+	if (d40d->lli_len < 0) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Unaligned size\n", __func__);
+		return -EINVAL;
+	}
+
+	if (d40_pool_lli_alloc(d40d, d40d->lli_len, true) < 0) {
 		dev_err(&d40c->chan.dev->device,
 			"[%s] Out of memory\n", __func__);
 		return -ENOMEM;
 	}
 
-	d40d->lli_len = sg_len;
 	d40d->lli_current = 0;
 
 	if (direction == DMA_FROM_DEVICE)
@@ -1993,13 +2094,21 @@ static int d40_prep_slave_sg_phy(struct d40_desc *d40d,
 	dma_addr_t dst_dev_addr;
 	int res;
 
-	if (d40_pool_lli_alloc(d40d, sgl_len, false) < 0) {
+	d40d->lli_len = d40_sg_2_dmalen(sgl, sgl_len,
+					d40c->dma_cfg.src_info.data_width,
+					d40c->dma_cfg.dst_info.data_width);
+	if (d40d->lli_len < 0) {
+		dev_err(&d40c->chan.dev->device,
+			"[%s] Unaligned size\n", __func__);
+		return -EINVAL;
+	}
+
+	if (d40_pool_lli_alloc(d40d, d40d->lli_len, false) < 0) {
 		dev_err(&d40c->chan.dev->device,
 			"[%s] Out of memory\n", __func__);
 		return -ENOMEM;
 	}
 
-	d40d->lli_len = sgl_len;
 	d40d->lli_current = 0;
 
 	if (direction == DMA_FROM_DEVICE) {
@@ -2024,6 +2133,7 @@ static int d40_prep_slave_sg_phy(struct d40_desc *d40d,
 				virt_to_phys(d40d->lli_phy.src),
 				d40c->src_def_cfg,
 				d40c->dma_cfg.src_info.data_width,
+				d40c->dma_cfg.dst_info.data_width,
 				d40c->dma_cfg.src_info.psize);
 	if (res < 0)
 		return res;
@@ -2035,6 +2145,7 @@ static int d40_prep_slave_sg_phy(struct d40_desc *d40d,
 				virt_to_phys(d40d->lli_phy.dst),
 				d40c->dst_def_cfg,
 				d40c->dma_cfg.dst_info.data_width,
+				d40c->dma_cfg.src_info.data_width,
 				d40c->dma_cfg.dst_info.psize);
 	if (res < 0)
 		return res;
@@ -2244,6 +2355,8 @@ static void d40_set_runtime_config(struct dma_chan *chan,
 			psize = STEDMA40_PSIZE_PHY_8;
 		else if (config_maxburst >= 4)
 			psize = STEDMA40_PSIZE_PHY_4;
+		else if (config_maxburst >= 2)
+			psize = STEDMA40_PSIZE_PHY_2;
 		else
 			psize = STEDMA40_PSIZE_PHY_1;
 	}

drivers/dma/ste_dma40_ll.c

 /*
  * Copyright (C) ST-Ericsson SA 2007-2010
- * Author: Per Friden <per.friden@stericsson.com> for ST-Ericsson
+ * Author: Per Forlin <per.forlin@stericsson.com> for ST-Ericsson
  * Author: Jonas Aaberg <jonas.aberg@stericsson.com> for ST-Ericsson
  * License terms: GNU General Public License (GPL) version 2
  */
@@ -122,7 +122,7 @@ void d40_phy_cfg(struct stedma40_chan_cfg *cfg,
 	*dst_cfg = dst;
 }
 
-int d40_phy_fill_lli(struct d40_phy_lli *lli,
+static int d40_phy_fill_lli(struct d40_phy_lli *lli,
 			    dma_addr_t data,
 			    u32 data_size,
 			    int psize,
@@ -139,13 +139,6 @@ int d40_phy_fill_lli(struct d40_phy_lli *lli,
 	else
 		num_elems = 2 << psize;
 
-	/*
-	 * Size is 16bit. data_width is 8, 16, 32 or 64 bit
-	 * Block large than 64 KiB must be split.
-	 */
-	if (data_size > (0xffff << data_width))
-		return -EINVAL;
-
 	/* Must be aligned */
 	if (!IS_ALIGNED(data, 0x1 << data_width))
 		return -EINVAL;
@@ -187,55 +180,118 @@ int d40_phy_fill_lli(struct d40_phy_lli *lli,
 	return 0;
 }
 
+static int d40_seg_size(int size, int data_width1, int data_width2)
+{
+	u32 max_w = max(data_width1, data_width2);
+	u32 min_w = min(data_width1, data_width2);
+	u32 seg_max = ALIGN(STEDMA40_MAX_SEG_SIZE << min_w, 1 << max_w);
+
+	if (seg_max > STEDMA40_MAX_SEG_SIZE)
+		seg_max -= (1 << max_w);
+
+	if (size <= seg_max)
+		return size;
+
+	if (size <= 2 * seg_max)
+		return ALIGN(size / 2, 1 << max_w);
+
+	return seg_max;
+}
+
+struct d40_phy_lli *d40_phy_buf_to_lli(struct d40_phy_lli *lli,
+				       dma_addr_t addr,
+				       u32 size,
+				       int psize,
+				       dma_addr_t lli_phys,
+				       u32 reg_cfg,
+				       bool term_int,
+				       u32 data_width1,
+				       u32 data_width2,
+				       bool is_device)
+{
+	int err;
+	dma_addr_t next = lli_phys;
+	int size_rest = size;
+	int size_seg = 0;
+
+	do {
+		size_seg = d40_seg_size(size_rest, data_width1, data_width2);
+		size_rest -= size_seg;
+
+		if (term_int && size_rest == 0)
+			next = 0;
+		else
+			next = ALIGN(next + sizeof(struct d40_phy_lli),
+				     D40_LLI_ALIGN);
+
+		err = d40_phy_fill_lli(lli,
+				       addr,
+				       size_seg,
+				       psize,
+				       next,
+				       reg_cfg,
+				       !next,
+				       data_width1,
+				       is_device);
+
+		if (err)
+			goto err;
+
+		lli++;
+		if (!is_device)
+			addr += size_seg;
+	} while (size_rest);
+
+	return lli;
+
+ err:
+	return NULL;
+}
+
 int d40_phy_sg_to_lli(struct scatterlist *sg,
 		      int sg_len,
 		      dma_addr_t target,
-		      struct d40_phy_lli *lli,
+		      struct d40_phy_lli *lli_sg,
 		      dma_addr_t lli_phys,
 		      u32 reg_cfg,
-		      u32 data_width,
+		      u32 data_width1,
+		      u32 data_width2,
 		      int psize)
 {
 	int total_size = 0;
 	int i;
 	struct scatterlist *current_sg = sg;
-	dma_addr_t next_lli_phys;
 	dma_addr_t dst;
-	int err = 0;
+	struct d40_phy_lli *lli = lli_sg;
+	dma_addr_t l_phys = lli_phys;
 
 	for_each_sg(sg, current_sg, sg_len, i) {
 
 		total_size += sg_dma_len(current_sg);
 
-		/* If this scatter list entry is the last one, no next link */
-		if (sg_len - 1 == i)
-			next_lli_phys = 0;
-		else
-			next_lli_phys = ALIGN(lli_phys + (i + 1) *
-					      sizeof(struct d40_phy_lli),
-					      D40_LLI_ALIGN);
-
 		if (target)
 			dst = target;
 		else
 			dst = sg_phys(current_sg);
 
-		err = d40_phy_fill_lli(&lli[i],
+		l_phys = ALIGN(lli_phys + (lli - lli_sg) *
+			       sizeof(struct d40_phy_lli), D40_LLI_ALIGN);
+
+		lli = d40_phy_buf_to_lli(lli,
 					 dst,
 					 sg_dma_len(current_sg),
 					 psize,
-				       next_lli_phys,
+					 l_phys,
 					 reg_cfg,
-				       !next_lli_phys,
-				       data_width,
+					 sg_len - 1 == i,
+					 data_width1,
+					 data_width2,
 					 target == dst);
-		if (err)
-			goto err;
+		if (lli == NULL)
+			return -EINVAL;
 	}
 
 	return total_size;
-err:
-	return err;
 }
 
 
@@ -315,7 +371,7 @@ void d40_log_lli_lcla_write(struct d40_log_lli *lcla,
 	writel(lli_dst->lcsp13, &lcla[1].lcsp13);
 }
 
-void d40_log_fill_lli(struct d40_log_lli *lli,
+static void d40_log_fill_lli(struct d40_log_lli *lli,
 			     dma_addr_t data, u32 data_size,
 			     u32 reg_cfg,
 			     u32 data_width,
@@ -326,6 +382,9 @@ void d40_log_fill_lli(struct d40_log_lli *lli,
 	/* The number of elements to transfer */
 	lli->lcsp02 = ((data_size >> data_width) <<
 		       D40_MEM_LCSP0_ECNT_POS) & D40_MEM_LCSP0_ECNT_MASK;
+
+	BUG_ON((data_size >> data_width) > STEDMA40_MAX_SEG_SIZE);
+
 	/* 16 LSBs address of the current element */
 	lli->lcsp02 |= data & D40_MEM_LCSP0_SPTR_MASK;
 	/* 16 MSBs address of the current element */
@@ -348,55 +407,94 @@ int d40_log_sg_to_dev(struct scatterlist *sg,
 	int total_size = 0;
 	struct scatterlist *current_sg = sg;
 	int i;
+	struct d40_log_lli *lli_src = lli->src;
+	struct d40_log_lli *lli_dst = lli->dst;
 
 	for_each_sg(sg, current_sg, sg_len, i) {
 		total_size += sg_dma_len(current_sg);
 
 		if (direction == DMA_TO_DEVICE) {
-			d40_log_fill_lli(&lli->src[i],
+			lli_src =
+				d40_log_buf_to_lli(lli_src,
 						   sg_phys(current_sg),
 						   sg_dma_len(current_sg),
 						   lcsp->lcsp1, src_data_width,
+						   dst_data_width,
 						   true);
-			d40_log_fill_lli(&lli->dst[i],
+			lli_dst =
+				d40_log_buf_to_lli(lli_dst,
 						   dev_addr,
 						   sg_dma_len(current_sg),
 						   lcsp->lcsp3, dst_data_width,
+						   src_data_width,
 						   false);
 		} else {
-			d40_log_fill_lli(&lli->dst[i],
+			lli_dst =
+				d40_log_buf_to_lli(lli_dst,
 						   sg_phys(current_sg),
 						   sg_dma_len(current_sg),
 						   lcsp->lcsp3, dst_data_width,
+						   src_data_width,
 						   true);
-			d40_log_fill_lli(&lli->src[i],
+			lli_src =
+				d40_log_buf_to_lli(lli_src,
 						   dev_addr,
 						   sg_dma_len(current_sg),
 						   lcsp->lcsp1, src_data_width,
+						   dst_data_width,
 						   false);
 		}
 	}
 	return total_size;
 }
 
+struct d40_log_lli *d40_log_buf_to_lli(struct d40_log_lli *lli_sg,
+				       dma_addr_t addr,
+				       int size,
+				       u32 lcsp13, /* src or dst*/
+				       u32 data_width1,
+				       u32 data_width2,
+				       bool addr_inc)
+{
+	struct d40_log_lli *lli = lli_sg;
+	int size_rest = size;
+	int size_seg = 0;
+
+	do {
+		size_seg = d40_seg_size(size_rest, data_width1, data_width2);
+		size_rest -= size_seg;
+
+		d40_log_fill_lli(lli,
+				 addr,
+				 size_seg,
+				 lcsp13, data_width1,
+				 addr_inc);
+		if (addr_inc)
+			addr += size_seg;
+		lli++;
+	} while (size_rest);
+
+	return lli;
+}
+
 int d40_log_sg_to_lli(struct scatterlist *sg,
 		      int sg_len,
 		      struct d40_log_lli *lli_sg,
 		      u32 lcsp13, /* src or dst*/
-		      u32 data_width)
+		      u32 data_width1, u32 data_width2)
 {
 	int total_size = 0;
 	struct scatterlist *current_sg = sg;
 	int i;
+	struct d40_log_lli *lli = lli_sg;
 
 	for_each_sg(sg, current_sg, sg_len, i) {
 		total_size += sg_dma_len(current_sg);
-
-		d40_log_fill_lli(&lli_sg[i],
+		lli = d40_log_buf_to_lli(lli,
 					 sg_phys(current_sg),
 					 sg_dma_len(current_sg),
-				 lcsp13, data_width,
-				 true);
+					 lcsp13,
+					 data_width1, data_width2, true);
 	}
 	return total_size;
 }

drivers/dma/ste_dma40_ll.h

@@ -292,17 +292,19 @@ int d40_phy_sg_to_lli(struct scatterlist *sg,
 		      struct d40_phy_lli *lli,
 		      dma_addr_t lli_phys,
 		      u32 reg_cfg,
-		      u32 data_width,
+		      u32 data_width1,
+		      u32 data_width2,
 		      int psize);
 
-int d40_phy_fill_lli(struct d40_phy_lli *lli,
+struct d40_phy_lli *d40_phy_buf_to_lli(struct d40_phy_lli *lli,
 				       dma_addr_t data,
 				       u32 data_size,
 				       int psize,
 				       dma_addr_t next_lli,
 				       u32 reg_cfg,
 				       bool term_int,
-		     u32 data_width,
+				       u32 data_width1,
+				       u32 data_width2,
 				       bool is_device);
 
 void d40_phy_lli_write(void __iomem *virtbase,
@@ -312,11 +314,11 @@ void d40_phy_lli_write(void __iomem *virtbase,
 
 /* Logical channels */
 
-void d40_log_fill_lli(struct d40_log_lli *lli,
-		      dma_addr_t data,
-		      u32 data_size,
-		      u32 reg_cfg,
-		      u32 data_width,
+struct d40_log_lli *d40_log_buf_to_lli(struct d40_log_lli *lli_sg,
+				       dma_addr_t addr,
+				       int size,
+				       u32 lcsp13, /* src or dst*/
+				       u32 data_width1, u32 data_width2,
 				       bool addr_inc);
 
 int d40_log_sg_to_dev(struct scatterlist *sg,
@@ -332,7 +334,7 @@ int d40_log_sg_to_lli(struct scatterlist *sg,
 		      int sg_len,
 		      struct d40_log_lli *lli_sg,
 		      u32 lcsp13, /* src or dst*/
-		      u32 data_width);
+		      u32 data_width1, u32 data_width2);
 
 void d40_log_lli_lcpa_write(struct d40_log_lli_full *lcpa,
 			    struct d40_log_lli *lli_dst,

include/linux/amba/pl08x.h

@@ -12,7 +12,6 @@
  *
  * Please credit ARM.com
  * Documentation: ARM DDI 0196D
- *
  */
 
 #ifndef AMBA_PL08X_H
@@ -22,6 +21,15 @@
 #include <linux/dmaengine.h>
 #include <linux/interrupt.h>
 
+struct pl08x_lli;
+struct pl08x_driver_data;
+
+/* Bitmasks for selecting AHB ports for DMA transfers */
+enum {
+	PL08X_AHB1 = (1 << 0),
+	PL08X_AHB2 = (1 << 1)
+};
+
 /**
  * struct pl08x_channel_data - data structure to pass info between
  * platform and PL08x driver regarding channel configuration
@@ -46,8 +54,10 @@
  * @circular_buffer: whether the buffer passed in is circular and
  * shall simply be looped round round (like a record baby round
  * round round round)
- * @single: the device connected to this channel will request single
- * DMA transfers, not bursts. (Bursts are default.)
+ * @single: the device connected to this channel will request single DMA
+ * transfers, not bursts. (Bursts are default.)
+ * @periph_buses: the device connected to this channel is accessible via
+ * these buses (use PL08X_AHB1 | PL08X_AHB2).
  */
 struct pl08x_channel_data {
 	char *bus_id;
@@ -55,10 +65,10 @@ struct pl08x_channel_data {
 	int max_signal;
 	u32 muxval;
 	u32 cctl;
-	u32 ccfg;
 	dma_addr_t addr;
 	bool circular_buffer;
 	bool single;
+	u8 periph_buses;
 };
 
 /**
@@ -67,24 +77,23 @@ struct pl08x_channel_data {
  * @addr: current address
  * @maxwidth: the maximum width of a transfer on this bus
  * @buswidth: the width of this bus in bytes: 1, 2 or 4
- * @fill_bytes: bytes required to fill to the next bus memory
- * boundary
+ * @fill_bytes: bytes required to fill to the next bus memory boundary
  */
 struct pl08x_bus_data {
 	dma_addr_t addr;
 	u8 maxwidth;
 	u8 buswidth;
-	u32 fill_bytes;
+	size_t fill_bytes;
 };
 
 /**
  * struct pl08x_phy_chan - holder for the physical channels
  * @id: physical index to this channel
  * @lock: a lock to use when altering an instance of this struct
- * @signal: the physical signal (aka channel) serving this
- * physical channel right now
- * @serving: the virtual channel currently being served by this
- * physical channel
+ * @signal: the physical signal (aka channel) serving this physical channel
+ * right now
+ * @serving: the virtual channel currently being served by this physical
+ * channel
  */
 struct pl08x_phy_chan {
 	unsigned int id;
@@ -92,11 +101,6 @@ struct pl08x_phy_chan {
 	spinlock_t lock;
 	int signal;
 	struct pl08x_dma_chan *serving;
-	u32 csrc;
-	u32 cdst;
-	u32 clli;
-	u32 cctl;
-	u32 ccfg;
 };
 
 /**
@@ -108,26 +112,23 @@ struct pl08x_txd {
 	struct dma_async_tx_descriptor tx;
 	struct list_head node;
 	enum dma_data_direction	direction;
-	struct pl08x_bus_data srcbus;
-	struct pl08x_bus_data dstbus;
-	int len;
+	dma_addr_t src_addr;
+	dma_addr_t dst_addr;
+	size_t len;
 	dma_addr_t llis_bus;
-	void *llis_va;
-	struct pl08x_channel_data *cd;
-	bool active;
+	struct pl08x_lli *llis_va;
+	/* Default cctl value for LLIs */
+	u32 cctl;
 	/*
 	 * Settings to be put into the physical channel when we
-	 * trigger this txd
+	 * trigger this txd.  Other registers are in llis_va[0].
 	 */
-	u32 csrc;
-	u32 cdst;
-	u32 clli;
-	u32 cctl;
+	u32 ccfg;
 };
 
 /**
- * struct pl08x_dma_chan_state - holds the PL08x specific virtual
- * channel states
+ * struct pl08x_dma_chan_state - holds the PL08x specific virtual channel
+ * states
  * @PL08X_CHAN_IDLE: the channel is idle
  * @PL08X_CHAN_RUNNING: the channel has allocated a physical transport
  * channel and is running a transfer on it
@@ -147,6 +148,8 @@ enum pl08x_dma_chan_state {
  * struct pl08x_dma_chan - this structure wraps a DMA ENGINE channel
  * @chan: wrappped abstract channel
  * @phychan: the physical channel utilized by this channel, if there is one
+ * @phychan_hold: if non-zero, hold on to the physical channel even if we
+ * have no pending entries
  * @tasklet: tasklet scheduled by the IRQ to handle actual work etc
  * @name: name of channel
  * @cd: channel platform data
@@ -154,53 +157,49 @@ enum pl08x_dma_chan_state {
  * @runtime_direction: current direction of this channel according to
  * runtime config
  * @lc: last completed transaction on this channel
- * @desc_list: queued transactions pending on this channel
+ * @pend_list: queued transactions pending on this channel
  * @at: active transaction on this channel
- * @lockflags: sometimes we let a lock last between two function calls,
- * especially prep/submit, and then we need to store the IRQ flags
- * in the channel state, here
  * @lock: a lock for this channel data
  * @host: a pointer to the host (internal use)
  * @state: whether the channel is idle, paused, running etc
  * @slave: whether this channel is a device (slave) or for memcpy
- * @waiting: a TX descriptor on this channel which is waiting for
- * a physical channel to become available
+ * @waiting: a TX descriptor on this channel which is waiting for a physical
+ * channel to become available
  */
 struct pl08x_dma_chan {
 	struct dma_chan chan;
 	struct pl08x_phy_chan *phychan;
+	int phychan_hold;
 	struct tasklet_struct tasklet;
 	char *name;
 	struct pl08x_channel_data *cd;
 	dma_addr_t runtime_addr;
 	enum dma_data_direction	runtime_direction;
-	atomic_t last_issued;
 	dma_cookie_t lc;
-	struct list_head desc_list;
+	struct list_head pend_list;
 	struct pl08x_txd *at;
-	unsigned long lockflags;
 	spinlock_t lock;
-	void *host;
+	struct pl08x_driver_data *host;
 	enum pl08x_dma_chan_state state;
 	bool slave;
 	struct pl08x_txd *waiting;
 };
 
 /**
- * struct pl08x_platform_data - the platform configuration for the
- * PL08x PrimeCells.
+ * struct pl08x_platform_data - the platform configuration for the PL08x
+ * PrimeCells.
  * @slave_channels: the channels defined for the different devices on the
  * platform, all inclusive, including multiplexed channels. The available
- * physical channels will be multiplexed around these signals as they
- * are requested, just enumerate all possible channels.
- * @get_signal: request a physical signal to be used for a DMA
- * transfer immediately: if there is some multiplexing or similar blocking
- * the use of the channel the transfer can be denied by returning
- * less than zero, else it returns the allocated signal number
+ * physical channels will be multiplexed around these signals as they are
+ * requested, just enumerate all possible channels.
+ * @get_signal: request a physical signal to be used for a DMA transfer
+ * immediately: if there is some multiplexing or similar blocking the use
+ * of the channel the transfer can be denied by returning less than zero,
+ * else it returns the allocated signal number
  * @put_signal: indicate to the platform that this physical signal is not
  * running any DMA transfer and multiplexing can be recycled
- * @bus_bit_lli: Bit[0] of the address indicated which AHB bus master the
- * LLI addresses are on 0/1 Master 1/2.
+ * @lli_buses: buses which LLIs can be fetched from: PL08X_AHB1 | PL08X_AHB2
+ * @mem_buses: buses which memory can be accessed from: PL08X_AHB1 | PL08X_AHB2
  */
 struct pl08x_platform_data {
 	struct pl08x_channel_data *slave_channels;
@@ -208,6 +207,8 @@ struct pl08x_platform_data {
 	struct pl08x_channel_data memcpy_channel;
 	int (*get_signal)(struct pl08x_dma_chan *);
 	void (*put_signal)(struct pl08x_dma_chan *);
+	u8 lli_buses;
+	u8 mem_buses;
 };
 
 #ifdef CONFIG_AMBA_PL08X

include/linux/dmaengine.h

@@ -532,7 +532,7 @@ static inline int dmaengine_resume(struct dma_chan *chan)
 	return dmaengine_device_control(chan, DMA_RESUME, 0);
 }
 
-static inline int dmaengine_submit(struct dma_async_tx_descriptor *desc)
+static inline dma_cookie_t dmaengine_submit(struct dma_async_tx_descriptor *desc)
 {
 	return desc->tx_submit(desc);
 }