This is v3 of the shdma driver split. It is no longer trying to be
generic, instead it is kept sh-specific, until we have better conditions
for making it generic. Apart from moving to an own directory with the
rest of shdma driver files, it has been updated to the top of the current
slave-dma tree. The comment from Sascha Hauer to v2 of this patch-set
has also been addressed - thanks! Two more client drivers have also been
added - fsi and renesas_usbhs. Would be really great to get it into 3.5 -
it is blocking the addition of one more DMAC variant.
Guennadi Liakhovetski (9):
dmaengine: add an shdma-base library
dma: shdma: prepare for conversion to the shdma base library
mmc: sh_mmcif: remove unneeded struct sh_mmcif_dma, prepare to shdma
conversion
mmc: sh_mobile_sdhi: prepare for conversion to the shdma base library
serial: sh-sci: prepare for conversion to the shdma base library
ASoC: siu: prepare for conversion to the shdma base library
usb: renesas_usbhs: prepare for conversion to the shdma base library
ASoC: fsi: prepare for conversion to the shdma base library
dma: shdma: convert to the shdma base library
drivers/dma/sh/Makefile | 1 +
drivers/dma/sh/shdma-base.c | 868 ++++++++++++++++++++++++++++
drivers/dma/sh/shdma.c | 1122 +++++++++----------------------------
drivers/dma/sh/shdma.h | 44 +-
drivers/mmc/host/sh_mmcif.c | 24 +-
drivers/mmc/host/sh_mobile_sdhi.c | 8 +-
drivers/tty/serial/sh-sci.c | 8 +-
drivers/usb/renesas_usbhs/fifo.c | 10 +-
include/linux/mmc/sh_mmcif.h | 8 +-
include/linux/sh_dma.h | 33 +-
include/linux/shdma-base.h | 123 ++++
sound/soc/sh/fsi.c | 4 +-
sound/soc/sh/siu_pcm.c | 8 +-
13 files changed, 1326 insertions(+), 935 deletions(-)
create mode 100644 drivers/dma/sh/shdma-base.c
create mode 100644 include/linux/shdma-base.h
--
1.7.2.5
Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/
Signed-off-by: Guennadi Liakhovetski <[email protected]>
Cc: Greg Kroah-Hartman <[email protected]>
---
drivers/usb/renesas_usbhs/fifo.c | 10 +++++-----
1 files changed, 5 insertions(+), 5 deletions(-)
diff --git a/drivers/usb/renesas_usbhs/fifo.c b/drivers/usb/renesas_usbhs/fifo.c
index 6ec7f83..fc597a4 100644
--- a/drivers/usb/renesas_usbhs/fifo.c
+++ b/drivers/usb/renesas_usbhs/fifo.c
@@ -994,7 +994,7 @@ static bool usbhsf_dma_filter(struct dma_chan *chan, void *param)
*
* usbhs doesn't recognize id = 0 as valid DMA
*/
- if (0 == slave->slave_id)
+ if (0 == slave->shdma_slave.slave_id)
return false;
chan->private = slave;
@@ -1173,8 +1173,8 @@ int usbhs_fifo_probe(struct usbhs_priv *priv)
fifo->port = D0FIFO;
fifo->sel = D0FIFOSEL;
fifo->ctr = D0FIFOCTR;
- fifo->tx_slave.slave_id = usbhs_get_dparam(priv, d0_tx_id);
- fifo->rx_slave.slave_id = usbhs_get_dparam(priv, d0_rx_id);
+ fifo->tx_slave.shdma_slave.slave_id = usbhs_get_dparam(priv, d0_tx_id);
+ fifo->rx_slave.shdma_slave.slave_id = usbhs_get_dparam(priv, d0_rx_id);
/* D1FIFO */
fifo = usbhsf_get_d1fifo(priv);
@@ -1182,8 +1182,8 @@ int usbhs_fifo_probe(struct usbhs_priv *priv)
fifo->port = D1FIFO;
fifo->sel = D1FIFOSEL;
fifo->ctr = D1FIFOCTR;
- fifo->tx_slave.slave_id = usbhs_get_dparam(priv, d1_tx_id);
- fifo->rx_slave.slave_id = usbhs_get_dparam(priv, d1_rx_id);
+ fifo->tx_slave.shdma_slave.slave_id = usbhs_get_dparam(priv, d1_tx_id);
+ fifo->rx_slave.shdma_slave.slave_id = usbhs_get_dparam(priv, d1_rx_id);
return 0;
}
--
1.7.2.5
This patch extracts code from shdma.c, that does not directly deal with
hardware implementation details and can be re-used with diverse DMA
controller variants, found on SH-based SoCs.
Signed-off-by: Guennadi Liakhovetski <[email protected]>
Cc: Sascha Hauer <[email protected]>
---
drivers/dma/sh/Makefile | 1 +
drivers/dma/sh/shdma-base.c | 868 +++++++++++++++++++++++++++++++++++++++++++
include/linux/shdma-base.h | 123 ++++++
3 files changed, 992 insertions(+), 0 deletions(-)
create mode 100644 drivers/dma/sh/shdma-base.c
create mode 100644 include/linux/shdma-base.h
diff --git a/drivers/dma/sh/Makefile b/drivers/dma/sh/Makefile
index ad4981a..54ae957 100644
--- a/drivers/dma/sh/Makefile
+++ b/drivers/dma/sh/Makefile
@@ -1 +1,2 @@
+obj-$(CONFIG_SH_DMAE) += shdma-base.o
obj-$(CONFIG_SH_DMAE) += shdma.o
diff --git a/drivers/dma/sh/shdma-base.c b/drivers/dma/sh/shdma-base.c
new file mode 100644
index 0000000..ff060d0
--- /dev/null
+++ b/drivers/dma/sh/shdma-base.c
@@ -0,0 +1,868 @@
+/*
+ * Dmaengine driver base library for DMA controllers, found on SH-based SoCs
+ *
+ * extracted from shdma.c
+ *
+ * Copyright (C) 2011-2012 Guennadi Liakhovetski <[email protected]>
+ * Copyright (C) 2009 Nobuhiro Iwamatsu <[email protected]>
+ * Copyright (C) 2009 Renesas Solutions, Inc. All rights reserved.
+ * Copyright (C) 2007 Freescale Semiconductor, Inc. All rights reserved.
+ *
+ * This is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/delay.h>
+#include <linux/shdma-base.h>
+#include <linux/dmaengine.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/module.h>
+#include <linux/pm_runtime.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+
+#include "../dmaengine.h"
+
+/* DMA descriptor control */
+enum shdma_desc_status {
+ DESC_IDLE,
+ DESC_PREPARED,
+ DESC_SUBMITTED,
+ DESC_COMPLETED, /* completed, have to call callback */
+ DESC_WAITING, /* callback called, waiting for ack / re-submit */
+};
+
+#define NR_DESCS_PER_CHANNEL 32
+
+#define to_shdma_chan(c) container_of(c, struct shdma_chan, dma_chan)
+#define to_shdma_dev(d) container_of(d, struct shdma_dev, dma_dev)
+
+/*
+ * For slave DMA we assume, that there is a finite number of DMA slaves in the
+ * system, and that each such slave can only use a finite number of channels.
+ * We use slave channel IDs to make sure, that no such slave channel ID is
+ * allocated more than once.
+ */
+static unsigned int slave_num = 256;
+module_param(slave_num, uint, 0444);
+
+/* A bitmask with slave_num bits */
+static unsigned long *shdma_slave_used;
+
+/* Called under spin_lock_irq(&schan->chan_lock") */
+static void shdma_chan_xfer_ld_queue(struct shdma_chan *schan)
+{
+ struct shdma_dev *sdev = to_shdma_dev(schan->dma_chan.device);
+ const struct shdma_ops *ops = sdev->ops;
+ struct shdma_desc *sdesc;
+
+ /* DMA work check */
+ if (ops->channel_busy(schan))
+ return;
+
+ /* Find the first not transferred descriptor */
+ list_for_each_entry(sdesc, &schan->ld_queue, node)
+ if (sdesc->mark == DESC_SUBMITTED) {
+ ops->start_xfer(schan, sdesc);
+ break;
+ }
+}
+
+static dma_cookie_t shdma_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+ struct shdma_desc *chunk, *c, *desc =
+ container_of(tx, struct shdma_desc, async_tx),
+ *last = desc;
+ struct shdma_chan *schan = to_shdma_chan(tx->chan);
+ struct shdma_slave *slave = tx->chan->private;
+ dma_async_tx_callback callback = tx->callback;
+ dma_cookie_t cookie;
+ bool power_up;
+
+ spin_lock_irq(&schan->chan_lock);
+
+ power_up = list_empty(&schan->ld_queue);
+
+ cookie = dma_cookie_assign(tx);
+
+ /* Mark all chunks of this descriptor as submitted, move to the queue */
+ list_for_each_entry_safe(chunk, c, desc->node.prev, node) {
+ /*
+ * All chunks are on the global ld_free, so, we have to find
+ * the end of the chain ourselves
+ */
+ if (chunk != desc && (chunk->mark == DESC_IDLE ||
+ chunk->async_tx.cookie > 0 ||
+ chunk->async_tx.cookie == -EBUSY ||
+ &chunk->node == &schan->ld_free))
+ break;
+ chunk->mark = DESC_SUBMITTED;
+ /* Callback goes to the last chunk */
+ chunk->async_tx.callback = NULL;
+ chunk->cookie = cookie;
+ list_move_tail(&chunk->node, &schan->ld_queue);
+ last = chunk;
+
+ dev_dbg(schan->dev, "submit #%d@%p on %d\n",
+ tx->cookie, &last->async_tx, schan->id);
+ }
+
+ last->async_tx.callback = callback;
+ last->async_tx.callback_param = tx->callback_param;
+
+ if (power_up) {
+ int ret;
+ schan->pm_state = SHDMA_PM_BUSY;
+
+ ret = pm_runtime_get(schan->dev);
+
+ spin_unlock_irq(&schan->chan_lock);
+ if (ret < 0)
+ dev_err(schan->dev, "%s(): GET = %d\n", __func__, ret);
+
+ pm_runtime_barrier(schan->dev);
+
+ spin_lock_irq(&schan->chan_lock);
+
+ /* Have we been reset, while waiting? */
+ if (schan->pm_state != SHDMA_PM_ESTABLISHED) {
+ struct shdma_dev *sdev =
+ to_shdma_dev(schan->dma_chan.device);
+ const struct shdma_ops *ops = sdev->ops;
+ dev_dbg(schan->dev, "Bring up channel %d\n",
+ schan->id);
+ /*
+ * TODO: .xfer_setup() might fail on some platforms.
+ * Make it int then, on error remove chunks from the
+ * queue again
+ */
+ ops->setup_xfer(schan, slave);
+
+ if (schan->pm_state == SHDMA_PM_PENDING)
+ shdma_chan_xfer_ld_queue(schan);
+ schan->pm_state = SHDMA_PM_ESTABLISHED;
+ }
+ } else {
+ /*
+ * Tell .device_issue_pending() not to run the queue, interrupts
+ * will do it anyway
+ */
+ schan->pm_state = SHDMA_PM_PENDING;
+ }
+
+ spin_unlock_irq(&schan->chan_lock);
+
+ return cookie;
+}
+
+/* Called with desc_lock held */
+static struct shdma_desc *shdma_get_desc(struct shdma_chan *schan)
+{
+ struct shdma_desc *sdesc;
+
+ list_for_each_entry(sdesc, &schan->ld_free, node)
+ if (sdesc->mark != DESC_PREPARED) {
+ BUG_ON(sdesc->mark != DESC_IDLE);
+ list_del(&sdesc->node);
+ return sdesc;
+ }
+
+ return NULL;
+}
+
+static int shdma_alloc_chan_resources(struct dma_chan *chan)
+{
+ struct shdma_chan *schan = to_shdma_chan(chan);
+ struct shdma_dev *sdev = to_shdma_dev(schan->dma_chan.device);
+ const struct shdma_ops *ops = sdev->ops;
+ struct shdma_desc *desc;
+ struct shdma_slave *slave = chan->private;
+ int ret, i;
+
+ /*
+ * This relies on the guarantee from dmaengine that alloc_chan_resources
+ * never runs concurrently with itself or free_chan_resources.
+ */
+ if (slave) {
+ if (slave->slave_id >= slave_num) {
+ ret = -EINVAL;
+ goto evalid;
+ }
+
+ if (test_and_set_bit(slave->slave_id, shdma_slave_used)) {
+ ret = -EBUSY;
+ goto etestused;
+ }
+
+ ret = ops->set_slave(schan, slave);
+ if (ret < 0)
+ goto esetslave;
+ }
+
+ schan->desc = kcalloc(NR_DESCS_PER_CHANNEL,
+ sdev->desc_size, GFP_KERNEL);
+ if (!schan->desc) {
+ ret = -ENOMEM;
+ goto edescalloc;
+ }
+ schan->desc_num = NR_DESCS_PER_CHANNEL;
+
+ for (i = 0; i < NR_DESCS_PER_CHANNEL; i++) {
+ desc = ops->embedded_desc(schan->desc, i);
+ dma_async_tx_descriptor_init(&desc->async_tx,
+ &schan->dma_chan);
+ desc->async_tx.tx_submit = shdma_tx_submit;
+ desc->mark = DESC_IDLE;
+
+ list_add(&desc->node, &schan->ld_free);
+ }
+
+ return NR_DESCS_PER_CHANNEL;
+
+edescalloc:
+ if (slave)
+esetslave:
+ clear_bit(slave->slave_id, shdma_slave_used);
+etestused:
+evalid:
+ chan->private = NULL;
+ return ret;
+}
+
+static dma_async_tx_callback __ld_cleanup(struct shdma_chan *schan, bool all)
+{
+ struct shdma_desc *desc, *_desc;
+ /* Is the "exposed" head of a chain acked? */
+ bool head_acked = false;
+ dma_cookie_t cookie = 0;
+ dma_async_tx_callback callback = NULL;
+ void *param = NULL;
+ unsigned long flags;
+
+ 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;
+
+ BUG_ON(tx->cookie > 0 && tx->cookie != desc->cookie);
+ BUG_ON(desc->mark != DESC_SUBMITTED &&
+ desc->mark != DESC_COMPLETED &&
+ desc->mark != DESC_WAITING);
+
+ /*
+ * queue is ordered, and we use this loop to (1) clean up all
+ * completed descriptors, and to (2) update descriptor flags of
+ * any chunks in a (partially) completed chain
+ */
+ if (!all && desc->mark == DESC_SUBMITTED &&
+ desc->cookie != cookie)
+ break;
+
+ if (tx->cookie > 0)
+ cookie = tx->cookie;
+
+ if (desc->mark == DESC_COMPLETED && desc->chunks == 1) {
+ if (schan->dma_chan.completed_cookie != desc->cookie - 1)
+ dev_dbg(schan->dev,
+ "Completing cookie %d, expected %d\n",
+ desc->cookie,
+ schan->dma_chan.completed_cookie + 1);
+ schan->dma_chan.completed_cookie = desc->cookie;
+ }
+
+ /* Call callback on the last chunk */
+ if (desc->mark == DESC_COMPLETED && tx->callback) {
+ desc->mark = DESC_WAITING;
+ 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);
+ break;
+ }
+
+ if (tx->cookie > 0 || tx->cookie == -EBUSY) {
+ if (desc->mark == DESC_COMPLETED) {
+ BUG_ON(tx->cookie < 0);
+ desc->mark = DESC_WAITING;
+ }
+ head_acked = async_tx_test_ack(tx);
+ } else {
+ switch (desc->mark) {
+ case DESC_COMPLETED:
+ desc->mark = DESC_WAITING;
+ /* Fall through */
+ case DESC_WAITING:
+ if (head_acked)
+ async_tx_ack(&desc->async_tx);
+ }
+ }
+
+ dev_dbg(schan->dev, "descriptor %p #%d completed.\n",
+ tx, tx->cookie);
+
+ if (((desc->mark == DESC_COMPLETED ||
+ desc->mark == DESC_WAITING) &&
+ async_tx_test_ack(&desc->async_tx)) || all) {
+ /* Remove from ld_queue list */
+ desc->mark = DESC_IDLE;
+
+ list_move(&desc->node, &schan->ld_free);
+
+ if (list_empty(&schan->ld_queue)) {
+ dev_dbg(schan->dev, "Bring down channel %d\n", schan->id);
+ pm_runtime_put(schan->dev);
+ schan->pm_state = SHDMA_PM_ESTABLISHED;
+ }
+ }
+ }
+
+ if (all && !callback)
+ /*
+ * Terminating and the loop completed normally: forgive
+ * uncompleted cookies
+ */
+ schan->dma_chan.completed_cookie = schan->dma_chan.cookie;
+
+ spin_unlock_irqrestore(&schan->chan_lock, flags);
+
+ if (callback)
+ callback(param);
+
+ return callback;
+}
+
+/*
+ * shdma_chan_ld_cleanup - Clean up link descriptors
+ *
+ * Clean up the ld_queue of DMA channel.
+ */
+static void shdma_chan_ld_cleanup(struct shdma_chan *schan, bool all)
+{
+ while (__ld_cleanup(schan, all))
+ ;
+}
+
+/*
+ * shdma_free_chan_resources - Free all resources of the channel.
+ */
+static void shdma_free_chan_resources(struct dma_chan *chan)
+{
+ struct shdma_chan *schan = to_shdma_chan(chan);
+ struct shdma_dev *sdev = to_shdma_dev(chan->device);
+ const struct shdma_ops *ops = sdev->ops;
+ LIST_HEAD(list);
+
+ /* Protect against ISR */
+ spin_lock_irq(&schan->chan_lock);
+ ops->halt_channel(schan);
+ spin_unlock_irq(&schan->chan_lock);
+
+ /* Now no new interrupts will occur */
+
+ /* Prepared and not submitted descriptors can still be on the queue */
+ if (!list_empty(&schan->ld_queue))
+ shdma_chan_ld_cleanup(schan, true);
+
+ if (chan->private) {
+ /* The caller is holding dma_list_mutex */
+ struct shdma_slave *slave = chan->private;
+ clear_bit(slave->slave_id, shdma_slave_used);
+ chan->private = NULL;
+ }
+
+ spin_lock_irq(&schan->chan_lock);
+
+ list_splice_init(&schan->ld_free, &list);
+ schan->desc_num = 0;
+
+ spin_unlock_irq(&schan->chan_lock);
+
+ kfree(schan->desc);
+}
+
+/**
+ * shdma_add_desc - get, set up and return one transfer descriptor
+ * @schan: DMA channel
+ * @flags: DMA transfer flags
+ * @dst: destination DMA address, incremented when direction equals
+ * DMA_DEV_TO_MEM or DMA_MEM_TO_MEM
+ * @src: source DMA address, incremented when direction equals
+ * DMA_MEM_TO_DEV or DMA_MEM_TO_MEM
+ * @len: DMA transfer length
+ * @first: if NULL, set to the current descriptor and cookie set to -EBUSY
+ * @direction: needed for slave DMA to decide which address to keep constant,
+ * equals DMA_MEM_TO_MEM for MEMCPY
+ * Returns 0 or an error
+ * Locks: called with desc_lock held
+ */
+static struct shdma_desc *shdma_add_desc(struct shdma_chan *schan,
+ unsigned long flags, dma_addr_t *dst, dma_addr_t *src, size_t *len,
+ struct shdma_desc **first, enum dma_transfer_direction direction)
+{
+ struct shdma_dev *sdev = to_shdma_dev(schan->dma_chan.device);
+ const struct shdma_ops *ops = sdev->ops;
+ struct shdma_desc *new;
+ size_t copy_size = *len;
+
+ if (!copy_size)
+ return NULL;
+
+ /* Allocate the link descriptor from the free list */
+ new = shdma_get_desc(schan);
+ if (!new) {
+ dev_err(schan->dev, "No free link descriptor available\n");
+ return NULL;
+ }
+
+ ops->desc_setup(schan, new, *src, *dst, ©_size);
+
+ if (!*first) {
+ /* First desc */
+ new->async_tx.cookie = -EBUSY;
+ *first = new;
+ } else {
+ /* Other desc - invisible to the user */
+ new->async_tx.cookie = -EINVAL;
+ }
+
+ dev_dbg(schan->dev,
+ "chaining (%u/%u)@%x -> %x with %p, cookie %d\n",
+ copy_size, *len, *src, *dst, &new->async_tx,
+ new->async_tx.cookie);
+
+ new->mark = DESC_PREPARED;
+ new->async_tx.flags = flags;
+ new->direction = direction;
+
+ *len -= copy_size;
+ if (direction == DMA_MEM_TO_MEM || direction == DMA_MEM_TO_DEV)
+ *src += copy_size;
+ if (direction == DMA_MEM_TO_MEM || direction == DMA_DEV_TO_MEM)
+ *dst += copy_size;
+
+ return new;
+}
+
+/*
+ * shdma_prep_sg - prepare transfer descriptors from an SG list
+ *
+ * Common routine for public (MEMCPY) and slave DMA. The MEMCPY case is also
+ * converted to scatter-gather to guarantee consistent locking and a correct
+ * list manipulation. For slave DMA direction carries the usual meaning, and,
+ * logically, the SG list is RAM and the addr variable contains slave address,
+ * e.g., the FIFO I/O register. For MEMCPY direction equals DMA_MEM_TO_MEM
+ * and the SG list contains only one element and points at the source buffer.
+ */
+static struct dma_async_tx_descriptor *shdma_prep_sg(struct shdma_chan *schan,
+ struct scatterlist *sgl, unsigned int sg_len, dma_addr_t *addr,
+ enum dma_transfer_direction direction, unsigned long flags)
+{
+ struct scatterlist *sg;
+ struct shdma_desc *first = NULL, *new = NULL /* compiler... */;
+ LIST_HEAD(tx_list);
+ int chunks = 0;
+ unsigned long irq_flags;
+ int i;
+
+ for_each_sg(sgl, sg, sg_len, i)
+ chunks += DIV_ROUND_UP(sg_dma_len(sg), schan->max_xfer_len);
+
+ /* Have to lock the whole loop to protect against concurrent release */
+ spin_lock_irqsave(&schan->chan_lock, irq_flags);
+
+ /*
+ * Chaining:
+ * first descriptor is what user is dealing with in all API calls, its
+ * cookie is at first set to -EBUSY, at tx-submit to a positive
+ * number
+ * if more than one chunk is needed further chunks have cookie = -EINVAL
+ * the last chunk, if not equal to the first, has cookie = -ENOSPC
+ * all chunks are linked onto the tx_list head with their .node heads
+ * only during this function, then they are immediately spliced
+ * back onto the free list in form of a chain
+ */
+ for_each_sg(sgl, sg, sg_len, i) {
+ dma_addr_t sg_addr = sg_dma_address(sg);
+ size_t len = sg_dma_len(sg);
+
+ if (!len)
+ goto err_get_desc;
+
+ do {
+ dev_dbg(schan->dev, "Add SG #%d@%p[%d], dma %llx\n",
+ i, sg, len, (unsigned long long)sg_addr);
+
+ if (direction == DMA_DEV_TO_MEM)
+ new = shdma_add_desc(schan, flags,
+ &sg_addr, addr, &len, &first,
+ direction);
+ else
+ new = shdma_add_desc(schan, flags,
+ addr, &sg_addr, &len, &first,
+ direction);
+ if (!new)
+ goto err_get_desc;
+
+ new->chunks = chunks--;
+ list_add_tail(&new->node, &tx_list);
+ } while (len);
+ }
+
+ if (new != first)
+ new->async_tx.cookie = -ENOSPC;
+
+ /* Put them back on the free list, so, they don't get lost */
+ list_splice_tail(&tx_list, &schan->ld_free);
+
+ spin_unlock_irqrestore(&schan->chan_lock, irq_flags);
+
+ return &first->async_tx;
+
+err_get_desc:
+ list_for_each_entry(new, &tx_list, node)
+ new->mark = DESC_IDLE;
+ list_splice(&tx_list, &schan->ld_free);
+
+ spin_unlock_irqrestore(&schan->chan_lock, irq_flags);
+
+ return NULL;
+}
+
+static struct dma_async_tx_descriptor *shdma_prep_memcpy(
+ struct dma_chan *chan, dma_addr_t dma_dest, dma_addr_t dma_src,
+ size_t len, unsigned long flags)
+{
+ struct shdma_chan *schan = to_shdma_chan(chan);
+ struct scatterlist sg;
+
+ if (!chan || !len)
+ return NULL;
+
+ BUG_ON(!schan->desc_num);
+
+ sg_init_table(&sg, 1);
+ sg_set_page(&sg, pfn_to_page(PFN_DOWN(dma_src)), len,
+ offset_in_page(dma_src));
+ sg_dma_address(&sg) = dma_src;
+ sg_dma_len(&sg) = len;
+
+ return shdma_prep_sg(schan, &sg, 1, &dma_dest, DMA_MEM_TO_MEM, flags);
+}
+
+static struct dma_async_tx_descriptor *shdma_prep_slave_sg(
+ struct dma_chan *chan, struct scatterlist *sgl, unsigned int sg_len,
+ enum dma_transfer_direction direction, unsigned long flags, void *context)
+{
+ struct shdma_chan *schan = to_shdma_chan(chan);
+ struct shdma_dev *sdev = to_shdma_dev(schan->dma_chan.device);
+ const struct shdma_ops *ops = sdev->ops;
+ struct shdma_slave *slave = chan->private;
+ dma_addr_t slave_addr;
+
+ if (!chan)
+ return NULL;
+
+ BUG_ON(!schan->desc_num);
+
+ /* Someone calling slave DMA on a generic channel? */
+ if (!slave || !sg_len) {
+ dev_warn(schan->dev, "%s: bad parameter: %p, %d, %d\n",
+ __func__, slave, sg_len, slave ? slave->slave_id : -1);
+ return NULL;
+ }
+
+ slave_addr = ops->slave_addr(schan);
+
+ return shdma_prep_sg(schan, sgl, sg_len, &slave_addr,
+ direction, flags);
+}
+
+static int shdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+ unsigned long arg)
+{
+ struct shdma_chan *schan = to_shdma_chan(chan);
+ struct shdma_dev *sdev = to_shdma_dev(chan->device);
+ const struct shdma_ops *ops = sdev->ops;
+ unsigned long flags;
+
+ /* Only supports DMA_TERMINATE_ALL */
+ if (cmd != DMA_TERMINATE_ALL)
+ return -ENXIO;
+
+ if (!chan)
+ return -EINVAL;
+
+ spin_lock_irqsave(&schan->chan_lock, flags);
+
+ ops->halt_channel(schan);
+
+ spin_unlock_irqrestore(&schan->chan_lock, flags);
+
+ shdma_chan_ld_cleanup(schan, true);
+
+ return 0;
+}
+
+static void shdma_issue_pending(struct dma_chan *chan)
+{
+ struct shdma_chan *schan = to_shdma_chan(chan);
+
+ spin_lock_irq(&schan->chan_lock);
+ if (schan->pm_state == SHDMA_PM_ESTABLISHED)
+ shdma_chan_xfer_ld_queue(schan);
+ else
+ schan->pm_state = SHDMA_PM_PENDING;
+ spin_unlock_irq(&schan->chan_lock);
+}
+
+static enum dma_status shdma_tx_status(struct dma_chan *chan,
+ dma_cookie_t cookie,
+ struct dma_tx_state *txstate)
+{
+ struct shdma_chan *schan = to_shdma_chan(chan);
+ enum dma_status status;
+ unsigned long flags;
+
+ shdma_chan_ld_cleanup(schan, false);
+
+ spin_lock_irqsave(&schan->chan_lock, flags);
+
+ status = dma_cookie_status(chan, cookie, txstate);
+
+ /*
+ * If we don't find cookie on the queue, it has been aborted and we have
+ * to report error
+ */
+ if (status != DMA_SUCCESS) {
+ struct shdma_desc *sdesc;
+ status = DMA_ERROR;
+ list_for_each_entry(sdesc, &schan->ld_queue, node)
+ if (sdesc->cookie == cookie) {
+ status = DMA_IN_PROGRESS;
+ break;
+ }
+ }
+
+ spin_unlock_irqrestore(&schan->chan_lock, flags);
+
+ return status;
+}
+
+/* Called from error IRQ or NMI */
+bool shdma_reset(struct shdma_dev *sdev)
+{
+ const struct shdma_ops *ops = sdev->ops;
+ struct shdma_chan *schan;
+ unsigned int handled = 0;
+ int i;
+
+ /* Reset all channels */
+ shdma_for_each_chan(schan, sdev, i) {
+ struct shdma_desc *sdesc;
+ LIST_HEAD(dl);
+
+ if (!schan)
+ continue;
+
+ spin_lock(&schan->chan_lock);
+
+ /* Stop the channel */
+ ops->halt_channel(schan);
+
+ list_splice_init(&schan->ld_queue, &dl);
+
+ if (!list_empty(&dl)) {
+ dev_dbg(schan->dev, "Bring down channel %d\n", schan->id);
+ pm_runtime_put(schan->dev);
+ }
+ schan->pm_state = SHDMA_PM_ESTABLISHED;
+
+ spin_unlock(&schan->chan_lock);
+
+ /* 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);
+ }
+
+ spin_lock(&schan->chan_lock);
+ list_splice(&dl, &schan->ld_free);
+ spin_unlock(&schan->chan_lock);
+
+ handled++;
+ }
+
+ return !!handled;
+}
+EXPORT_SYMBOL(shdma_reset);
+
+static irqreturn_t chan_irq(int irq, void *dev)
+{
+ struct shdma_chan *schan = dev;
+ const struct shdma_ops *ops =
+ to_shdma_dev(schan->dma_chan.device)->ops;
+ irqreturn_t ret;
+
+ spin_lock(&schan->chan_lock);
+
+ ret = ops->chan_irq(schan, irq) ? IRQ_WAKE_THREAD : IRQ_NONE;
+
+ spin_unlock(&schan->chan_lock);
+
+ return ret;
+}
+
+static irqreturn_t chan_irqt(int irq, void *dev)
+{
+ struct shdma_chan *schan = dev;
+ const struct shdma_ops *ops =
+ to_shdma_dev(schan->dma_chan.device)->ops;
+ struct shdma_desc *sdesc;
+
+ spin_lock_irq(&schan->chan_lock);
+ list_for_each_entry(sdesc, &schan->ld_queue, node) {
+ if (sdesc->mark == DESC_SUBMITTED &&
+ ops->desc_completed(schan, sdesc)) {
+ dev_dbg(schan->dev, "done #%d@%p\n",
+ sdesc->async_tx.cookie, &sdesc->async_tx);
+ sdesc->mark = DESC_COMPLETED;
+ break;
+ }
+ }
+ /* Next desc */
+ shdma_chan_xfer_ld_queue(schan);
+ spin_unlock_irq(&schan->chan_lock);
+
+ shdma_chan_ld_cleanup(schan, false);
+
+ return IRQ_HANDLED;
+}
+
+int shdma_request_irq(struct shdma_chan *schan, int irq,
+ unsigned long flags, const char *name)
+{
+ int ret = request_threaded_irq(irq, chan_irq, chan_irqt,
+ flags, name, schan);
+
+ schan->irq = ret < 0 ? ret : irq;
+
+ return ret;
+}
+EXPORT_SYMBOL(shdma_request_irq);
+
+void shdma_free_irq(struct shdma_chan *schan)
+{
+ if (schan->irq >= 0)
+ free_irq(schan->irq, schan);
+}
+EXPORT_SYMBOL(shdma_free_irq);
+
+void shdma_chan_probe(struct shdma_dev *sdev,
+ struct shdma_chan *schan, int id)
+{
+ schan->pm_state = SHDMA_PM_ESTABLISHED;
+
+ /* reference struct dma_device */
+ schan->dma_chan.device = &sdev->dma_dev;
+ dma_cookie_init(&schan->dma_chan);
+
+ schan->dev = sdev->dma_dev.dev;
+ schan->id = id;
+
+ if (!schan->max_xfer_len)
+ schan->max_xfer_len = PAGE_SIZE;
+
+ spin_lock_init(&schan->chan_lock);
+
+ /* Init descripter manage list */
+ INIT_LIST_HEAD(&schan->ld_queue);
+ INIT_LIST_HEAD(&schan->ld_free);
+
+ /* Add the channel to DMA device channel list */
+ list_add_tail(&schan->dma_chan.device_node,
+ &sdev->dma_dev.channels);
+ sdev->schan[sdev->dma_dev.chancnt++] = schan;
+}
+EXPORT_SYMBOL(shdma_chan_probe);
+
+void shdma_chan_remove(struct shdma_chan *schan)
+{
+ list_del(&schan->dma_chan.device_node);
+}
+EXPORT_SYMBOL(shdma_chan_remove);
+
+int shdma_init(struct device *dev, struct shdma_dev *sdev,
+ int chan_num)
+{
+ struct dma_device *dma_dev = &sdev->dma_dev;
+
+ /*
+ * Require all call-backs for now, they can trivially be made optional
+ * later as required
+ */
+ if (!sdev->ops ||
+ !sdev->desc_size ||
+ !sdev->ops->embedded_desc ||
+ !sdev->ops->start_xfer ||
+ !sdev->ops->setup_xfer ||
+ !sdev->ops->set_slave ||
+ !sdev->ops->desc_setup ||
+ !sdev->ops->slave_addr ||
+ !sdev->ops->channel_busy ||
+ !sdev->ops->halt_channel ||
+ !sdev->ops->desc_completed)
+ return -EINVAL;
+
+ sdev->schan = kcalloc(chan_num, sizeof(*sdev->schan), GFP_KERNEL);
+ if (!sdev->schan)
+ return -ENOMEM;
+
+ INIT_LIST_HEAD(&dma_dev->channels);
+
+ /* Common and MEMCPY operations */
+ dma_dev->device_alloc_chan_resources
+ = shdma_alloc_chan_resources;
+ dma_dev->device_free_chan_resources = shdma_free_chan_resources;
+ dma_dev->device_prep_dma_memcpy = shdma_prep_memcpy;
+ dma_dev->device_tx_status = shdma_tx_status;
+ dma_dev->device_issue_pending = shdma_issue_pending;
+
+ /* Compulsory for DMA_SLAVE fields */
+ dma_dev->device_prep_slave_sg = shdma_prep_slave_sg;
+ dma_dev->device_control = shdma_control;
+
+ dma_dev->dev = dev;
+
+ return 0;
+}
+EXPORT_SYMBOL(shdma_init);
+
+void shdma_cleanup(struct shdma_dev *sdev)
+{
+ kfree(sdev->schan);
+}
+EXPORT_SYMBOL(shdma_cleanup);
+
+static int __init shdma_enter(void)
+{
+ shdma_slave_used = kzalloc(DIV_ROUND_UP(slave_num, BITS_PER_LONG) *
+ sizeof(long), GFP_KERNEL);
+ if (!shdma_slave_used)
+ return -ENOMEM;
+ return 0;
+}
+module_init(shdma_enter);
+
+static void __exit shdma_exit(void)
+{
+ kfree(shdma_slave_used);
+}
+module_exit(shdma_exit);
+
+MODULE_LICENSE("GPL v2");
+MODULE_DESCRIPTION("SH-DMA driver base library");
+MODULE_AUTHOR("Guennadi Liakhovetski <[email protected]>");
diff --git a/include/linux/shdma-base.h b/include/linux/shdma-base.h
new file mode 100644
index 0000000..83efd13
--- /dev/null
+++ b/include/linux/shdma-base.h
@@ -0,0 +1,123 @@
+/*
+ * Dmaengine driver base library for DMA controllers, found on SH-based SoCs
+ *
+ * extracted from shdma.c and headers
+ *
+ * Copyright (C) 2011-2012 Guennadi Liakhovetski <[email protected]>
+ * Copyright (C) 2009 Nobuhiro Iwamatsu <[email protected]>
+ * Copyright (C) 2009 Renesas Solutions, Inc. All rights reserved.
+ * Copyright (C) 2007 Freescale Semiconductor, Inc. All rights reserved.
+ *
+ * This is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef SHDMA_BASE_H
+#define SHDMA_BASE_H
+
+#include <linux/dmaengine.h>
+#include <linux/interrupt.h>
+#include <linux/list.h>
+#include <linux/types.h>
+
+/**
+ * shdma_pm_state - DMA channel PM state
+ * SHDMA_PM_ESTABLISHED: either idle or during data transfer
+ * SHDMA_PM_BUSY: during the transfer preparation, when we have to
+ * drop the lock temporarily
+ * SHDMA_PM_PENDING: transfers pending
+ */
+enum shdma_pm_state {
+ SHDMA_PM_ESTABLISHED,
+ SHDMA_PM_BUSY,
+ SHDMA_PM_PENDING,
+};
+
+struct device;
+
+/*
+ * Drivers, using this library are expected to embed struct shdma_dev,
+ * struct shdma_chan, struct shdma_desc, and struct shdma_slave
+ * in their respective device, channel, descriptor and slave objects.
+ */
+
+struct shdma_slave {
+ unsigned int slave_id;
+};
+
+struct shdma_desc {
+ struct list_head node;
+ struct dma_async_tx_descriptor async_tx;
+ enum dma_transfer_direction direction;
+ dma_cookie_t cookie;
+ int chunks;
+ int mark;
+};
+
+struct shdma_chan {
+ spinlock_t chan_lock; /* Channel operation lock */
+ struct list_head ld_queue; /* Link descriptors queue */
+ struct list_head ld_free; /* Free link descriptors */
+ struct dma_chan dma_chan; /* DMA channel */
+ struct device *dev; /* Channel device */
+ void *desc; /* buffer for descriptor array */
+ int desc_num; /* desc count */
+ size_t max_xfer_len; /* max transfer length */
+ int id; /* Raw id of this channel */
+ int irq; /* Channel IRQ */
+ enum shdma_pm_state pm_state;
+};
+
+/**
+ * struct shdma_ops - simple DMA driver operations
+ * desc_completed: return true, if this is the descriptor, that just has
+ * completed (atomic)
+ * halt_channel: stop DMA channel operation (atomic)
+ * channel_busy: return true, if the channel is busy (atomic)
+ * slave_addr: return slave DMA address
+ * desc_setup: set up the hardware specific descriptor portion (atomic)
+ * set_slave: bind channel to a slave
+ * setup_xfer: configure channel hardware for operation (atomic)
+ * start_xfer: start the DMA transfer (atomic)
+ * embedded_desc: return Nth struct shdma_desc pointer from the
+ * descriptor array
+ * chan_irq: process channel IRQ, return true if a transfer has
+ * completed (atomic)
+ */
+struct shdma_ops {
+ bool (*desc_completed)(struct shdma_chan *, struct shdma_desc *);
+ void (*halt_channel)(struct shdma_chan *);
+ bool (*channel_busy)(struct shdma_chan *);
+ dma_addr_t (*slave_addr)(struct shdma_chan *);
+ int (*desc_setup)(struct shdma_chan *, struct shdma_desc *,
+ dma_addr_t, dma_addr_t, size_t *);
+ int (*set_slave)(struct shdma_chan *, struct shdma_slave *);
+ void (*setup_xfer)(struct shdma_chan *, struct shdma_slave *);
+ void (*start_xfer)(struct shdma_chan *, struct shdma_desc *);
+ struct shdma_desc *(*embedded_desc)(void *, int);
+ bool (*chan_irq)(struct shdma_chan *, int);
+};
+
+struct shdma_dev {
+ struct dma_device dma_dev;
+ struct shdma_chan **schan;
+ const struct shdma_ops *ops;
+ size_t desc_size;
+};
+
+#define shdma_for_each_chan(c, d, i) for (i = 0, c = (d)->schan[0]; \
+ i < (d)->dma_dev.chancnt; c = (d)->schan[++i])
+
+int shdma_request_irq(struct shdma_chan *, int,
+ unsigned long, const char *);
+void shdma_free_irq(struct shdma_chan *);
+bool shdma_reset(struct shdma_dev *sdev);
+void shdma_chan_probe(struct shdma_dev *sdev,
+ struct shdma_chan *schan, int id);
+void shdma_chan_remove(struct shdma_chan *schan);
+int shdma_init(struct device *dev, struct shdma_dev *sdev,
+ int chan_num);
+void shdma_cleanup(struct shdma_dev *sdev);
+
+#endif
--
1.7.2.5
Now that all users have been updated to use the embedded in struct
sh_mmcif_plat_data DMA slave IDs, struct sh_mmcif_dma is no longer needed
and can be removed. This also makes preparation to the shdma base library
conversion easier.
Signed-off-by: Guennadi Liakhovetski <[email protected]>
Cc: Chris Ball <[email protected]>
---
drivers/mmc/host/sh_mmcif.c | 24 ++++++++++--------------
include/linux/mmc/sh_mmcif.h | 8 +-------
2 files changed, 11 insertions(+), 21 deletions(-)
diff --git a/drivers/mmc/host/sh_mmcif.c b/drivers/mmc/host/sh_mmcif.c
index aafaf0b..25cf203 100644
--- a/drivers/mmc/host/sh_mmcif.c
+++ b/drivers/mmc/host/sh_mmcif.c
@@ -385,31 +385,27 @@ static void sh_mmcif_request_dma(struct sh_mmcif_host *host,
host->dma_active = false;
/* We can only either use DMA for both Tx and Rx or not use it at all */
- if (pdata->dma) {
- dev_warn(&host->pd->dev,
- "Update your platform to use embedded DMA slave IDs\n");
- tx = &pdata->dma->chan_priv_tx;
- rx = &pdata->dma->chan_priv_rx;
- } else {
- tx = &host->dma_slave_tx;
- tx->slave_id = pdata->slave_id_tx;
- rx = &host->dma_slave_rx;
- rx->slave_id = pdata->slave_id_rx;
- }
- if (tx->slave_id > 0 && rx->slave_id > 0) {
+ tx = &host->dma_slave_tx;
+ tx->shdma_slave.slave_id = pdata->slave_id_tx;
+ rx = &host->dma_slave_rx;
+ rx->shdma_slave.slave_id = pdata->slave_id_rx;
+
+ if (tx->shdma_slave.slave_id > 0 && rx->shdma_slave.slave_id > 0) {
dma_cap_mask_t mask;
dma_cap_zero(mask);
dma_cap_set(DMA_SLAVE, mask);
- host->chan_tx = dma_request_channel(mask, sh_mmcif_filter, tx);
+ host->chan_tx = dma_request_channel(mask, sh_mmcif_filter,
+ &tx->shdma_slave);
dev_dbg(&host->pd->dev, "%s: TX: got channel %p\n", __func__,
host->chan_tx);
if (!host->chan_tx)
return;
- host->chan_rx = dma_request_channel(mask, sh_mmcif_filter, rx);
+ host->chan_rx = dma_request_channel(mask, sh_mmcif_filter,
+ &rx->shdma_slave);
dev_dbg(&host->pd->dev, "%s: RX: got channel %p\n", __func__,
host->chan_rx);
diff --git a/include/linux/mmc/sh_mmcif.h b/include/linux/mmc/sh_mmcif.h
index 05f0e3d..c37956c 100644
--- a/include/linux/mmc/sh_mmcif.h
+++ b/include/linux/mmc/sh_mmcif.h
@@ -32,17 +32,11 @@
* 1111 : Peripheral clock (sup_pclk set '1')
*/
-struct sh_mmcif_dma {
- struct sh_dmae_slave chan_priv_tx;
- struct sh_dmae_slave chan_priv_rx;
-};
-
struct sh_mmcif_plat_data {
void (*set_pwr)(struct platform_device *pdev, int state);
void (*down_pwr)(struct platform_device *pdev);
int (*get_cd)(struct platform_device *pdef);
- struct sh_mmcif_dma *dma; /* Deprecated. Instead */
- unsigned int slave_id_tx; /* use embedded slave_id_[tr]x */
+ unsigned int slave_id_tx; /* embedded slave_id_[tr]x */
unsigned int slave_id_rx;
u8 sup_pclk; /* 1 :SH7757, 0: SH7724/SH7372 */
unsigned long caps;
--
1.7.2.5
Signed-off-by: Guennadi Liakhovetski <[email protected]>
Cc: Alan Cox <[email protected]>
---
drivers/tty/serial/sh-sci.c | 8 ++++----
1 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/tty/serial/sh-sci.c b/drivers/tty/serial/sh-sci.c
index bf461cf..49e3b2a 100644
--- a/drivers/tty/serial/sh-sci.c
+++ b/drivers/tty/serial/sh-sci.c
@@ -1579,9 +1579,9 @@ static bool filter(struct dma_chan *chan, void *slave)
struct sh_dmae_slave *param = slave;
dev_dbg(chan->device->dev, "%s: slave ID %d\n", __func__,
- param->slave_id);
+ param->shdma_slave.slave_id);
- chan->private = param;
+ chan->private = ¶m->shdma_slave;
return true;
}
@@ -1620,7 +1620,7 @@ static void sci_request_dma(struct uart_port *port)
param = &s->param_tx;
/* Slave ID, e.g., SHDMA_SLAVE_SCIF0_TX */
- param->slave_id = s->cfg->dma_slave_tx;
+ param->shdma_slave.slave_id = s->cfg->dma_slave_tx;
s->cookie_tx = -EINVAL;
chan = dma_request_channel(mask, filter, param);
@@ -1648,7 +1648,7 @@ static void sci_request_dma(struct uart_port *port)
param = &s->param_rx;
/* Slave ID, e.g., SHDMA_SLAVE_SCIF0_RX */
- param->slave_id = s->cfg->dma_slave_rx;
+ param->shdma_slave.slave_id = s->cfg->dma_slave_rx;
chan = dma_request_channel(mask, filter, param);
dev_dbg(port->dev, "%s: RX: got channel %p\n", __func__, chan);
--
1.7.2.5
Signed-off-by: Guennadi Liakhovetski <[email protected]>
Cc: Liam Girdwood <[email protected]>
Cc: Mark Brown <[email protected]>
---
sound/soc/sh/siu_pcm.c | 8 ++++----
1 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/sound/soc/sh/siu_pcm.c b/sound/soc/sh/siu_pcm.c
index 5cfcc65..3fdc801 100644
--- a/sound/soc/sh/siu_pcm.c
+++ b/sound/soc/sh/siu_pcm.c
@@ -330,12 +330,12 @@ static bool filter(struct dma_chan *chan, void *slave)
{
struct sh_dmae_slave *param = slave;
- pr_debug("%s: slave ID %d\n", __func__, param->slave_id);
+ pr_debug("%s: slave ID %d\n", __func__, param->shdma_slave.slave_id);
if (unlikely(param->dma_dev != chan->device->dev))
return false;
- chan->private = param;
+ chan->private = ¶m->shdma_slave;
return true;
}
@@ -360,12 +360,12 @@ static int siu_pcm_open(struct snd_pcm_substream *ss)
if (ss->stream == SNDRV_PCM_STREAM_PLAYBACK) {
siu_stream = &port_info->playback;
param = &siu_stream->param;
- param->slave_id = port ? pdata->dma_slave_tx_b :
+ param->shdma_slave.slave_id = port ? pdata->dma_slave_tx_b :
pdata->dma_slave_tx_a;
} else {
siu_stream = &port_info->capture;
param = &siu_stream->param;
- param->slave_id = port ? pdata->dma_slave_rx_b :
+ param->shdma_slave.slave_id = port ? pdata->dma_slave_rx_b :
pdata->dma_slave_rx_a;
}
--
1.7.2.5
Signed-off-by: Guennadi Liakhovetski <[email protected]>
Cc: Liam Girdwood <[email protected]>
Cc: Mark Brown <[email protected]>
---
sound/soc/sh/fsi.c | 4 ++--
1 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/sound/soc/sh/fsi.c b/sound/soc/sh/fsi.c
index 74ed2df..c28314b 100644
--- a/sound/soc/sh/fsi.c
+++ b/sound/soc/sh/fsi.c
@@ -1532,8 +1532,8 @@ static void fsi_handler_init(struct fsi_priv *fsi)
fsi->capture.priv = fsi;
if (fsi->info->tx_id) {
- fsi->playback.slave.slave_id = fsi->info->tx_id;
- fsi->playback.handler = &fsi_dma_push_handler;
+ fsi->playback.slave.shdma_slave.slave_id = fsi->info->tx_id;
+ fsi->playback.handler = &fsi_dma_push_handler;
}
}
--
1.7.2.5
Signed-off-by: Guennadi Liakhovetski <[email protected]>
Cc: Chris Ball <[email protected]>
---
drivers/mmc/host/sh_mobile_sdhi.c | 8 ++++----
1 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/mmc/host/sh_mobile_sdhi.c b/drivers/mmc/host/sh_mobile_sdhi.c
index 934b68e..b8d7004 100644
--- a/drivers/mmc/host/sh_mobile_sdhi.c
+++ b/drivers/mmc/host/sh_mobile_sdhi.c
@@ -145,10 +145,10 @@ static int __devinit sh_mobile_sdhi_probe(struct platform_device *pdev)
mmc_data->cd_gpio = p->cd_gpio;
if (p->dma_slave_tx > 0 && p->dma_slave_rx > 0) {
- priv->param_tx.slave_id = p->dma_slave_tx;
- priv->param_rx.slave_id = p->dma_slave_rx;
- priv->dma_priv.chan_priv_tx = &priv->param_tx;
- priv->dma_priv.chan_priv_rx = &priv->param_rx;
+ priv->param_tx.shdma_slave.slave_id = p->dma_slave_tx;
+ priv->param_rx.shdma_slave.slave_id = p->dma_slave_rx;
+ priv->dma_priv.chan_priv_tx = &priv->param_tx.shdma_slave;
+ priv->dma_priv.chan_priv_rx = &priv->param_rx.shdma_slave;
priv->dma_priv.alignment_shift = 1; /* 2-byte alignment */
mmc_data->dma = &priv->dma_priv;
}
--
1.7.2.5
By placing an anonymous union at the top of struct sh_dmae_slave we can
transparently prepare all device and client drivers for the upcoming
shdma-base conversion.
Signed-off-by: Guennadi Liakhovetski <[email protected]>
---
include/linux/sh_dma.h | 8 ++++++--
1 files changed, 6 insertions(+), 2 deletions(-)
diff --git a/include/linux/sh_dma.h b/include/linux/sh_dma.h
index 425450b..e081e8e 100644
--- a/include/linux/sh_dma.h
+++ b/include/linux/sh_dma.h
@@ -10,12 +10,16 @@
#ifndef SH_DMA_H
#define SH_DMA_H
-#include <linux/list.h>
#include <linux/dmaengine.h>
+#include <linux/list.h>
+#include <linux/shdma-base.h>
/* Used by slave DMA clients to request DMA to/from a specific peripheral */
struct sh_dmae_slave {
- unsigned int slave_id; /* Set by the platform */
+ union {
+ unsigned int slave_id; /* Set by the platform */
+ struct shdma_slave shdma_slave;
+ };
struct device *dma_dev; /* Set by the platform */
const struct sh_dmae_slave_config *config; /* Set by the driver */
};
--
1.7.2.5
The shdma base library has originally been extracted from the shdma driver,
which now can be converted to actually use it.
Signed-off-by: Guennadi Liakhovetski <[email protected]>
---
drivers/dma/sh/shdma.c | 1122 ++++++++++++------------------------------------
drivers/dma/sh/shdma.h | 44 +-
include/linux/sh_dma.h | 33 +-
3 files changed, 302 insertions(+), 897 deletions(-)
diff --git a/drivers/dma/sh/shdma.c b/drivers/dma/sh/shdma.c
index 8ab4a1f..c393b35 100644
--- a/drivers/dma/sh/shdma.c
+++ b/drivers/dma/sh/shdma.c
@@ -3,6 +3,7 @@
*
* base is drivers/dma/flsdma.c
*
+ * Copyright (C) 2011-2012 Guennadi Liakhovetski <[email protected]>
* Copyright (C) 2009 Nobuhiro Iwamatsu <[email protected]>
* Copyright (C) 2009 Renesas Solutions, Inc. All rights reserved.
* Copyright (C) 2007 Freescale Semiconductor, Inc. All rights reserved.
@@ -34,18 +35,12 @@
#include "../dmaengine.h"
#include "shdma.h"
-/* DMA descriptor control */
-enum sh_dmae_desc_status {
- DESC_IDLE,
- DESC_PREPARED,
- DESC_SUBMITTED,
- DESC_COMPLETED, /* completed, have to call callback */
- DESC_WAITING, /* callback called, waiting for ack / re-submit */
-};
+#define SH_DMAE_DRV_NAME "sh-dma-engine"
-#define NR_DESCS_PER_CHANNEL 32
/* Default MEMCPY transfer size = 2^2 = 4 bytes */
#define LOG2_DEFAULT_XFER_SIZE 2
+#define SH_DMA_SLAVE_NUMBER 256
+#define SH_DMA_TCR_MAX (16 * 1024 * 1024 - 1)
/*
* Used for write-side mutual exclusion for the global device list,
@@ -54,18 +49,12 @@ enum sh_dmae_desc_status {
static DEFINE_SPINLOCK(sh_dmae_lock);
static LIST_HEAD(sh_dmae_devices);
-/* A bitmask with bits enough for enum sh_dmae_slave_chan_id */
-static unsigned long sh_dmae_slave_used[BITS_TO_LONGS(SH_DMA_SLAVE_NUMBER)];
-
-static void sh_dmae_chan_ld_cleanup(struct sh_dmae_chan *sh_chan, bool all);
-static void sh_chan_xfer_ld_queue(struct sh_dmae_chan *sh_chan);
-
static void chclr_write(struct sh_dmae_chan *sh_dc, u32 data)
{
struct sh_dmae_device *shdev = to_sh_dev(sh_dc);
__raw_writel(data, shdev->chan_reg +
- shdev->pdata->channel[sh_dc->id].chclr_offset);
+ shdev->pdata->channel[sh_dc->shdma_chan.id].chclr_offset);
}
static void sh_dmae_writel(struct sh_dmae_chan *sh_dc, u32 data, u32 reg)
@@ -155,11 +144,11 @@ static int sh_dmae_rst(struct sh_dmae_device *shdev)
spin_unlock_irqrestore(&sh_dmae_lock, flags);
if (dmaor & (DMAOR_AE | DMAOR_NMIF)) {
- dev_warn(shdev->common.dev, "Can't initialize DMAOR.\n");
+ dev_warn(shdev->shdma_dev.dma_dev.dev, "Can't initialize DMAOR.\n");
return -EIO;
}
if (shdev->pdata->dmaor_init & ~dmaor)
- dev_warn(shdev->common.dev,
+ dev_warn(shdev->shdma_dev.dma_dev.dev,
"DMAOR=0x%x hasn't latched the initial value 0x%x.\n",
dmaor, shdev->pdata->dmaor_init);
return 0;
@@ -224,15 +213,6 @@ static void dmae_start(struct sh_dmae_chan *sh_chan)
chcr_write(sh_chan, chcr & ~CHCR_TE);
}
-static void dmae_halt(struct sh_dmae_chan *sh_chan)
-{
- struct sh_dmae_device *shdev = to_sh_dev(sh_chan);
- u32 chcr = chcr_read(sh_chan);
-
- chcr &= ~(CHCR_DE | CHCR_TE | shdev->chcr_ie_bit);
- chcr_write(sh_chan, chcr);
-}
-
static void dmae_init(struct sh_dmae_chan *sh_chan)
{
/*
@@ -261,7 +241,7 @@ static int dmae_set_dmars(struct sh_dmae_chan *sh_chan, u16 val)
{
struct sh_dmae_device *shdev = to_sh_dev(sh_chan);
struct sh_dmae_pdata *pdata = shdev->pdata;
- const struct sh_dmae_channel *chan_pdata = &pdata->channel[sh_chan->id];
+ const struct sh_dmae_channel *chan_pdata = &pdata->channel[sh_chan->shdma_chan.id];
u16 __iomem *addr = shdev->dmars;
unsigned int shift = chan_pdata->dmars_bit;
@@ -282,706 +262,142 @@ static int dmae_set_dmars(struct sh_dmae_chan *sh_chan, u16 val)
return 0;
}
-static dma_cookie_t sh_dmae_tx_submit(struct dma_async_tx_descriptor *tx)
+static void sh_dmae_start_xfer(struct shdma_chan *schan,
+ struct shdma_desc *sdesc)
{
- struct sh_desc *desc = tx_to_sh_desc(tx), *chunk, *last = desc, *c;
- struct sh_dmae_chan *sh_chan = to_sh_chan(tx->chan);
- struct sh_dmae_slave *param = tx->chan->private;
- dma_async_tx_callback callback = tx->callback;
- dma_cookie_t cookie;
- bool power_up;
-
- spin_lock_irq(&sh_chan->desc_lock);
-
- if (list_empty(&sh_chan->ld_queue))
- power_up = true;
- else
- power_up = false;
-
- cookie = dma_cookie_assign(tx);
-
- /* Mark all chunks of this descriptor as submitted, move to the queue */
- list_for_each_entry_safe(chunk, c, desc->node.prev, node) {
- /*
- * All chunks are on the global ld_free, so, we have to find
- * the end of the chain ourselves
- */
- if (chunk != desc && (chunk->mark == DESC_IDLE ||
- chunk->async_tx.cookie > 0 ||
- chunk->async_tx.cookie == -EBUSY ||
- &chunk->node == &sh_chan->ld_free))
- break;
- chunk->mark = DESC_SUBMITTED;
- /* Callback goes to the last chunk */
- chunk->async_tx.callback = NULL;
- chunk->cookie = cookie;
- list_move_tail(&chunk->node, &sh_chan->ld_queue);
- last = chunk;
- }
-
- last->async_tx.callback = callback;
- last->async_tx.callback_param = tx->callback_param;
-
- dev_dbg(sh_chan->dev, "submit #%d@%p on %d: %x[%d] -> %x\n",
- tx->cookie, &last->async_tx, sh_chan->id,
- desc->hw.sar, desc->hw.tcr, desc->hw.dar);
-
- if (power_up) {
- sh_chan->pm_state = DMAE_PM_BUSY;
-
- pm_runtime_get(sh_chan->dev);
-
- spin_unlock_irq(&sh_chan->desc_lock);
-
- pm_runtime_barrier(sh_chan->dev);
-
- spin_lock_irq(&sh_chan->desc_lock);
-
- /* Have we been reset, while waiting? */
- if (sh_chan->pm_state != DMAE_PM_ESTABLISHED) {
- dev_dbg(sh_chan->dev, "Bring up channel %d\n",
- sh_chan->id);
- if (param) {
- const struct sh_dmae_slave_config *cfg =
- param->config;
-
- dmae_set_dmars(sh_chan, cfg->mid_rid);
- dmae_set_chcr(sh_chan, cfg->chcr);
- } else {
- dmae_init(sh_chan);
- }
-
- if (sh_chan->pm_state == DMAE_PM_PENDING)
- sh_chan_xfer_ld_queue(sh_chan);
- sh_chan->pm_state = DMAE_PM_ESTABLISHED;
- }
- } else {
- sh_chan->pm_state = DMAE_PM_PENDING;
- }
-
- spin_unlock_irq(&sh_chan->desc_lock);
-
- return cookie;
+ struct sh_dmae_chan *sh_chan = container_of(schan, struct sh_dmae_chan,
+ shdma_chan);
+ struct sh_dmae_desc *sh_desc = container_of(sdesc,
+ struct sh_dmae_desc, shdma_desc);
+ dev_dbg(sh_chan->shdma_chan.dev, "Queue #%d to %d: %u@%x -> %x\n",
+ sdesc->async_tx.cookie, sh_chan->shdma_chan.id,
+ sh_desc->hw.tcr, sh_desc->hw.sar, sh_desc->hw.dar);
+ /* Get the ld start address from ld_queue */
+ dmae_set_reg(sh_chan, &sh_desc->hw);
+ dmae_start(sh_chan);
}
-/* Called with desc_lock held */
-static struct sh_desc *sh_dmae_get_desc(struct sh_dmae_chan *sh_chan)
+static bool sh_dmae_channel_busy(struct shdma_chan *schan)
{
- struct sh_desc *desc;
-
- list_for_each_entry(desc, &sh_chan->ld_free, node)
- if (desc->mark != DESC_PREPARED) {
- BUG_ON(desc->mark != DESC_IDLE);
- list_del(&desc->node);
- return desc;
- }
-
- return NULL;
+ struct sh_dmae_chan *sh_chan = container_of(schan, struct sh_dmae_chan,
+ shdma_chan);
+ return dmae_is_busy(sh_chan);
}
-static const struct sh_dmae_slave_config *sh_dmae_find_slave(
- struct sh_dmae_chan *sh_chan, struct sh_dmae_slave *param)
+static void sh_dmae_setup_xfer(struct shdma_chan *schan,
+ struct shdma_slave *sslave)
{
- struct sh_dmae_device *shdev = to_sh_dev(sh_chan);
- struct sh_dmae_pdata *pdata = shdev->pdata;
- int i;
+ struct sh_dmae_chan *sh_chan = container_of(schan, struct sh_dmae_chan,
+ shdma_chan);
- if (param->slave_id >= SH_DMA_SLAVE_NUMBER)
- return NULL;
-
- for (i = 0; i < pdata->slave_num; i++)
- if (pdata->slave[i].slave_id == param->slave_id)
- return pdata->slave + i;
-
- return NULL;
-}
-
-static int sh_dmae_alloc_chan_resources(struct dma_chan *chan)
-{
- struct sh_dmae_chan *sh_chan = to_sh_chan(chan);
- struct sh_desc *desc;
- struct sh_dmae_slave *param = chan->private;
- int ret;
+ if (sslave) {
+ struct sh_dmae_slave *slave = container_of(sslave,
+ struct sh_dmae_slave, shdma_slave);
+ const struct sh_dmae_slave_config *cfg =
+ slave->config;
- /*
- * This relies on the guarantee from dmaengine that alloc_chan_resources
- * never runs concurrently with itself or free_chan_resources.
- */
- if (param) {
- const struct sh_dmae_slave_config *cfg;
-
- cfg = sh_dmae_find_slave(sh_chan, param);
- if (!cfg) {
- ret = -EINVAL;
- goto efindslave;
- }
-
- if (test_and_set_bit(param->slave_id, sh_dmae_slave_used)) {
- ret = -EBUSY;
- goto etestused;
- }
-
- param->config = cfg;
- }
-
- while (sh_chan->descs_allocated < NR_DESCS_PER_CHANNEL) {
- desc = kzalloc(sizeof(struct sh_desc), GFP_KERNEL);
- if (!desc)
- break;
- dma_async_tx_descriptor_init(&desc->async_tx,
- &sh_chan->common);
- desc->async_tx.tx_submit = sh_dmae_tx_submit;
- desc->mark = DESC_IDLE;
-
- list_add(&desc->node, &sh_chan->ld_free);
- sh_chan->descs_allocated++;
- }
-
- if (!sh_chan->descs_allocated) {
- ret = -ENOMEM;
- goto edescalloc;
- }
-
- return sh_chan->descs_allocated;
-
-edescalloc:
- if (param)
- clear_bit(param->slave_id, sh_dmae_slave_used);
-etestused:
-efindslave:
- chan->private = NULL;
- return ret;
-}
-
-/*
- * sh_dma_free_chan_resources - Free all resources of the channel.
- */
-static void sh_dmae_free_chan_resources(struct dma_chan *chan)
-{
- struct sh_dmae_chan *sh_chan = to_sh_chan(chan);
- struct sh_desc *desc, *_desc;
- LIST_HEAD(list);
-
- /* Protect against ISR */
- spin_lock_irq(&sh_chan->desc_lock);
- dmae_halt(sh_chan);
- spin_unlock_irq(&sh_chan->desc_lock);
-
- /* Now no new interrupts will occur */
-
- /* Prepared and not submitted descriptors can still be on the queue */
- if (!list_empty(&sh_chan->ld_queue))
- sh_dmae_chan_ld_cleanup(sh_chan, true);
-
- if (chan->private) {
- /* The caller is holding dma_list_mutex */
- struct sh_dmae_slave *param = chan->private;
- clear_bit(param->slave_id, sh_dmae_slave_used);
- chan->private = NULL;
- }
-
- spin_lock_irq(&sh_chan->desc_lock);
-
- list_splice_init(&sh_chan->ld_free, &list);
- sh_chan->descs_allocated = 0;
-
- spin_unlock_irq(&sh_chan->desc_lock);
-
- list_for_each_entry_safe(desc, _desc, &list, node)
- kfree(desc);
-}
-
-/**
- * sh_dmae_add_desc - get, set up and return one transfer descriptor
- * @sh_chan: DMA channel
- * @flags: DMA transfer flags
- * @dest: destination DMA address, incremented when direction equals
- * DMA_DEV_TO_MEM
- * @src: source DMA address, incremented when direction equals
- * DMA_MEM_TO_DEV
- * @len: DMA transfer length
- * @first: if NULL, set to the current descriptor and cookie set to -EBUSY
- * @direction: needed for slave DMA to decide which address to keep constant,
- * equals DMA_MEM_TO_MEM for MEMCPY
- * Returns 0 or an error
- * Locks: called with desc_lock held
- */
-static struct sh_desc *sh_dmae_add_desc(struct sh_dmae_chan *sh_chan,
- unsigned long flags, dma_addr_t *dest, dma_addr_t *src, size_t *len,
- struct sh_desc **first, enum dma_transfer_direction direction)
-{
- struct sh_desc *new;
- size_t copy_size;
-
- if (!*len)
- return NULL;
-
- /* Allocate the link descriptor from the free list */
- new = sh_dmae_get_desc(sh_chan);
- if (!new) {
- dev_err(sh_chan->dev, "No free link descriptor available\n");
- return NULL;
- }
-
- copy_size = min(*len, (size_t)SH_DMA_TCR_MAX + 1);
-
- new->hw.sar = *src;
- new->hw.dar = *dest;
- new->hw.tcr = copy_size;
-
- if (!*first) {
- /* First desc */
- new->async_tx.cookie = -EBUSY;
- *first = new;
+ dmae_set_dmars(sh_chan, cfg->mid_rid);
+ dmae_set_chcr(sh_chan, cfg->chcr);
} else {
- /* Other desc - invisible to the user */
- new->async_tx.cookie = -EINVAL;
+ dmae_init(sh_chan);
}
-
- dev_dbg(sh_chan->dev,
- "chaining (%u/%u)@%x -> %x with %p, cookie %d, shift %d\n",
- copy_size, *len, *src, *dest, &new->async_tx,
- new->async_tx.cookie, sh_chan->xmit_shift);
-
- new->mark = DESC_PREPARED;
- new->async_tx.flags = flags;
- new->direction = direction;
-
- *len -= copy_size;
- if (direction == DMA_MEM_TO_MEM || direction == DMA_MEM_TO_DEV)
- *src += copy_size;
- if (direction == DMA_MEM_TO_MEM || direction == DMA_DEV_TO_MEM)
- *dest += copy_size;
-
- return new;
}
-/*
- * sh_dmae_prep_sg - prepare transfer descriptors from an SG list
- *
- * Common routine for public (MEMCPY) and slave DMA. The MEMCPY case is also
- * converted to scatter-gather to guarantee consistent locking and a correct
- * list manipulation. For slave DMA direction carries the usual meaning, and,
- * logically, the SG list is RAM and the addr variable contains slave address,
- * e.g., the FIFO I/O register. For MEMCPY direction equals DMA_MEM_TO_MEM
- * and the SG list contains only one element and points at the source buffer.
- */
-static struct dma_async_tx_descriptor *sh_dmae_prep_sg(struct sh_dmae_chan *sh_chan,
- struct scatterlist *sgl, unsigned int sg_len, dma_addr_t *addr,
- enum dma_transfer_direction direction, unsigned long flags)
+static const struct sh_dmae_slave_config *dmae_find_slave(
+ struct sh_dmae_chan *sh_chan, struct sh_dmae_slave *slave)
{
- struct scatterlist *sg;
- struct sh_desc *first = NULL, *new = NULL /* compiler... */;
- LIST_HEAD(tx_list);
- int chunks = 0;
- unsigned long irq_flags;
+ struct sh_dmae_device *shdev = to_sh_dev(sh_chan);
+ struct sh_dmae_pdata *pdata = shdev->pdata;
+ const struct sh_dmae_slave_config *cfg;
int i;
- if (!sg_len)
+ if (slave->shdma_slave.slave_id >= SH_DMA_SLAVE_NUMBER)
return NULL;
- for_each_sg(sgl, sg, sg_len, i)
- chunks += (sg_dma_len(sg) + SH_DMA_TCR_MAX) /
- (SH_DMA_TCR_MAX + 1);
-
- /* Have to lock the whole loop to protect against concurrent release */
- spin_lock_irqsave(&sh_chan->desc_lock, irq_flags);
-
- /*
- * Chaining:
- * first descriptor is what user is dealing with in all API calls, its
- * cookie is at first set to -EBUSY, at tx-submit to a positive
- * number
- * if more than one chunk is needed further chunks have cookie = -EINVAL
- * the last chunk, if not equal to the first, has cookie = -ENOSPC
- * all chunks are linked onto the tx_list head with their .node heads
- * only during this function, then they are immediately spliced
- * back onto the free list in form of a chain
- */
- for_each_sg(sgl, sg, sg_len, i) {
- dma_addr_t sg_addr = sg_dma_address(sg);
- size_t len = sg_dma_len(sg);
-
- if (!len)
- goto err_get_desc;
-
- do {
- dev_dbg(sh_chan->dev, "Add SG #%d@%p[%d], dma %llx\n",
- i, sg, len, (unsigned long long)sg_addr);
-
- if (direction == DMA_DEV_TO_MEM)
- new = sh_dmae_add_desc(sh_chan, flags,
- &sg_addr, addr, &len, &first,
- direction);
- else
- new = sh_dmae_add_desc(sh_chan, flags,
- addr, &sg_addr, &len, &first,
- direction);
- if (!new)
- goto err_get_desc;
-
- new->chunks = chunks--;
- list_add_tail(&new->node, &tx_list);
- } while (len);
- }
-
- if (new != first)
- new->async_tx.cookie = -ENOSPC;
-
- /* Put them back on the free list, so, they don't get lost */
- list_splice_tail(&tx_list, &sh_chan->ld_free);
-
- spin_unlock_irqrestore(&sh_chan->desc_lock, irq_flags);
-
- return &first->async_tx;
-
-err_get_desc:
- list_for_each_entry(new, &tx_list, node)
- new->mark = DESC_IDLE;
- list_splice(&tx_list, &sh_chan->ld_free);
-
- spin_unlock_irqrestore(&sh_chan->desc_lock, irq_flags);
+ for (i = 0, cfg = pdata->slave; i < pdata->slave_num; i++, cfg++)
+ if (cfg->slave_id == slave->shdma_slave.slave_id)
+ return cfg;
return NULL;
}
-static struct dma_async_tx_descriptor *sh_dmae_prep_memcpy(
- struct dma_chan *chan, dma_addr_t dma_dest, dma_addr_t dma_src,
- size_t len, unsigned long flags)
+static int sh_dmae_set_slave(struct shdma_chan *schan,
+ struct shdma_slave *sslave)
{
- struct sh_dmae_chan *sh_chan;
- struct scatterlist sg;
-
- if (!chan || !len)
- return NULL;
-
- sh_chan = to_sh_chan(chan);
-
- sg_init_table(&sg, 1);
- sg_set_page(&sg, pfn_to_page(PFN_DOWN(dma_src)), len,
- offset_in_page(dma_src));
- sg_dma_address(&sg) = dma_src;
- sg_dma_len(&sg) = len;
-
- return sh_dmae_prep_sg(sh_chan, &sg, 1, &dma_dest, DMA_MEM_TO_MEM,
- flags);
-}
-
-static struct dma_async_tx_descriptor *sh_dmae_prep_slave_sg(
- struct dma_chan *chan, struct scatterlist *sgl, unsigned int sg_len,
- enum dma_transfer_direction direction, unsigned long flags,
- void *context)
-{
- struct sh_dmae_slave *param;
- struct sh_dmae_chan *sh_chan;
- dma_addr_t slave_addr;
-
- if (!chan)
- return NULL;
-
- sh_chan = to_sh_chan(chan);
- param = chan->private;
-
- /* Someone calling slave DMA on a public channel? */
- if (!param || !sg_len) {
- dev_warn(sh_chan->dev, "%s: bad parameter: %p, %d, %d\n",
- __func__, param, sg_len, param ? param->slave_id : -1);
- return NULL;
- }
-
- slave_addr = param->config->addr;
-
- /*
- * if (param != NULL), this is a successfully requested slave channel,
- * therefore param->config != NULL too.
- */
- return sh_dmae_prep_sg(sh_chan, sgl, sg_len, &slave_addr,
- direction, flags);
-}
-
-static int sh_dmae_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
- unsigned long arg)
-{
- struct sh_dmae_chan *sh_chan = to_sh_chan(chan);
- unsigned long flags;
-
- /* Only supports DMA_TERMINATE_ALL */
- if (cmd != DMA_TERMINATE_ALL)
- return -ENXIO;
-
- if (!chan)
- return -EINVAL;
-
- spin_lock_irqsave(&sh_chan->desc_lock, flags);
- dmae_halt(sh_chan);
-
- if (!list_empty(&sh_chan->ld_queue)) {
- /* Record partial transfer */
- struct sh_desc *desc = list_entry(sh_chan->ld_queue.next,
- struct sh_desc, node);
- desc->partial = (desc->hw.tcr - sh_dmae_readl(sh_chan, TCR)) <<
- sh_chan->xmit_shift;
- }
- spin_unlock_irqrestore(&sh_chan->desc_lock, flags);
+ struct sh_dmae_chan *sh_chan = container_of(schan, struct sh_dmae_chan,
+ shdma_chan);
+ struct sh_dmae_slave *slave = container_of(sslave, struct sh_dmae_slave,
+ shdma_slave);
+ const struct sh_dmae_slave_config *cfg = dmae_find_slave(sh_chan, slave);
+ if (!cfg)
+ return -ENODEV;
- sh_dmae_chan_ld_cleanup(sh_chan, true);
+ slave->config = cfg;
return 0;
}
-static dma_async_tx_callback __ld_cleanup(struct sh_dmae_chan *sh_chan, bool all)
+static void dmae_halt(struct sh_dmae_chan *sh_chan)
{
- struct sh_desc *desc, *_desc;
- /* Is the "exposed" head of a chain acked? */
- bool head_acked = false;
- dma_cookie_t cookie = 0;
- dma_async_tx_callback callback = NULL;
- void *param = NULL;
- unsigned long flags;
-
- spin_lock_irqsave(&sh_chan->desc_lock, flags);
- list_for_each_entry_safe(desc, _desc, &sh_chan->ld_queue, node) {
- struct dma_async_tx_descriptor *tx = &desc->async_tx;
-
- BUG_ON(tx->cookie > 0 && tx->cookie != desc->cookie);
- BUG_ON(desc->mark != DESC_SUBMITTED &&
- desc->mark != DESC_COMPLETED &&
- desc->mark != DESC_WAITING);
-
- /*
- * queue is ordered, and we use this loop to (1) clean up all
- * completed descriptors, and to (2) update descriptor flags of
- * any chunks in a (partially) completed chain
- */
- if (!all && desc->mark == DESC_SUBMITTED &&
- desc->cookie != cookie)
- break;
-
- if (tx->cookie > 0)
- cookie = tx->cookie;
-
- if (desc->mark == DESC_COMPLETED && desc->chunks == 1) {
- if (sh_chan->common.completed_cookie != desc->cookie - 1)
- dev_dbg(sh_chan->dev,
- "Completing cookie %d, expected %d\n",
- desc->cookie,
- sh_chan->common.completed_cookie + 1);
- sh_chan->common.completed_cookie = desc->cookie;
- }
-
- /* Call callback on the last chunk */
- if (desc->mark == DESC_COMPLETED && tx->callback) {
- desc->mark = DESC_WAITING;
- callback = tx->callback;
- param = tx->callback_param;
- dev_dbg(sh_chan->dev, "descriptor #%d@%p on %d callback\n",
- tx->cookie, tx, sh_chan->id);
- BUG_ON(desc->chunks != 1);
- break;
- }
-
- if (tx->cookie > 0 || tx->cookie == -EBUSY) {
- if (desc->mark == DESC_COMPLETED) {
- BUG_ON(tx->cookie < 0);
- desc->mark = DESC_WAITING;
- }
- head_acked = async_tx_test_ack(tx);
- } else {
- switch (desc->mark) {
- case DESC_COMPLETED:
- desc->mark = DESC_WAITING;
- /* Fall through */
- case DESC_WAITING:
- if (head_acked)
- async_tx_ack(&desc->async_tx);
- }
- }
-
- dev_dbg(sh_chan->dev, "descriptor %p #%d completed.\n",
- tx, tx->cookie);
-
- if (((desc->mark == DESC_COMPLETED ||
- desc->mark == DESC_WAITING) &&
- async_tx_test_ack(&desc->async_tx)) || all) {
- /* Remove from ld_queue list */
- desc->mark = DESC_IDLE;
-
- list_move(&desc->node, &sh_chan->ld_free);
-
- if (list_empty(&sh_chan->ld_queue)) {
- dev_dbg(sh_chan->dev, "Bring down channel %d\n", sh_chan->id);
- pm_runtime_put(sh_chan->dev);
- }
- }
- }
-
- if (all && !callback)
- /*
- * Terminating and the loop completed normally: forgive
- * uncompleted cookies
- */
- sh_chan->common.completed_cookie = sh_chan->common.cookie;
-
- spin_unlock_irqrestore(&sh_chan->desc_lock, flags);
-
- if (callback)
- callback(param);
+ struct sh_dmae_device *shdev = to_sh_dev(sh_chan);
+ u32 chcr = chcr_read(sh_chan);
- return callback;
+ chcr &= ~(CHCR_DE | CHCR_TE | shdev->chcr_ie_bit);
+ chcr_write(sh_chan, chcr);
}
-/*
- * sh_chan_ld_cleanup - Clean up link descriptors
- *
- * This function cleans up the ld_queue of DMA channel.
- */
-static void sh_dmae_chan_ld_cleanup(struct sh_dmae_chan *sh_chan, bool all)
+static int sh_dmae_desc_setup(struct shdma_chan *schan,
+ struct shdma_desc *sdesc,
+ dma_addr_t src, dma_addr_t dst, size_t *len)
{
- while (__ld_cleanup(sh_chan, all))
- ;
-}
+ struct sh_dmae_desc *sh_desc = container_of(sdesc,
+ struct sh_dmae_desc, shdma_desc);
-/* Called under spin_lock_irq(&sh_chan->desc_lock) */
-static void sh_chan_xfer_ld_queue(struct sh_dmae_chan *sh_chan)
-{
- struct sh_desc *desc;
+ if (*len > schan->max_xfer_len)
+ *len = schan->max_xfer_len;
- /* DMA work check */
- if (dmae_is_busy(sh_chan))
- return;
-
- /* Find the first not transferred descriptor */
- list_for_each_entry(desc, &sh_chan->ld_queue, node)
- if (desc->mark == DESC_SUBMITTED) {
- dev_dbg(sh_chan->dev, "Queue #%d to %d: %u@%x -> %x\n",
- desc->async_tx.cookie, sh_chan->id,
- desc->hw.tcr, desc->hw.sar, desc->hw.dar);
- /* Get the ld start address from ld_queue */
- dmae_set_reg(sh_chan, &desc->hw);
- dmae_start(sh_chan);
- break;
- }
-}
+ sh_desc->hw.sar = src;
+ sh_desc->hw.dar = dst;
+ sh_desc->hw.tcr = *len;
-static void sh_dmae_memcpy_issue_pending(struct dma_chan *chan)
-{
- struct sh_dmae_chan *sh_chan = to_sh_chan(chan);
-
- spin_lock_irq(&sh_chan->desc_lock);
- if (sh_chan->pm_state == DMAE_PM_ESTABLISHED)
- sh_chan_xfer_ld_queue(sh_chan);
- else
- sh_chan->pm_state = DMAE_PM_PENDING;
- spin_unlock_irq(&sh_chan->desc_lock);
+ return 0;
}
-static enum dma_status sh_dmae_tx_status(struct dma_chan *chan,
- dma_cookie_t cookie,
- struct dma_tx_state *txstate)
+static void sh_dmae_halt(struct shdma_chan *schan)
{
- struct sh_dmae_chan *sh_chan = to_sh_chan(chan);
- enum dma_status status;
- unsigned long flags;
-
- sh_dmae_chan_ld_cleanup(sh_chan, false);
-
- spin_lock_irqsave(&sh_chan->desc_lock, flags);
-
- status = dma_cookie_status(chan, cookie, txstate);
-
- /*
- * If we don't find cookie on the queue, it has been aborted and we have
- * to report error
- */
- if (status != DMA_SUCCESS) {
- struct sh_desc *desc;
- status = DMA_ERROR;
- list_for_each_entry(desc, &sh_chan->ld_queue, node)
- if (desc->cookie == cookie) {
- status = DMA_IN_PROGRESS;
- break;
- }
- }
-
- spin_unlock_irqrestore(&sh_chan->desc_lock, flags);
-
- return status;
+ struct sh_dmae_chan *sh_chan = container_of(schan, struct sh_dmae_chan,
+ shdma_chan);
+ dmae_halt(sh_chan);
}
-static irqreturn_t sh_dmae_interrupt(int irq, void *data)
+static bool sh_dmae_chan_irq(struct shdma_chan *schan, int irq)
{
- irqreturn_t ret = IRQ_NONE;
- struct sh_dmae_chan *sh_chan = data;
- u32 chcr;
-
- spin_lock(&sh_chan->desc_lock);
-
- chcr = chcr_read(sh_chan);
+ struct sh_dmae_chan *sh_chan = container_of(schan, struct sh_dmae_chan,
+ shdma_chan);
- if (chcr & CHCR_TE) {
- /* DMA stop */
- dmae_halt(sh_chan);
-
- ret = IRQ_HANDLED;
- tasklet_schedule(&sh_chan->tasklet);
- }
+ if (!(chcr_read(sh_chan) & CHCR_TE))
+ return false;
- spin_unlock(&sh_chan->desc_lock);
+ /* DMA stop */
+ dmae_halt(sh_chan);
- return ret;
+ return true;
}
/* Called from error IRQ or NMI */
static bool sh_dmae_reset(struct sh_dmae_device *shdev)
{
- unsigned int handled = 0;
- int i;
+ bool ret;
/* halt the dma controller */
sh_dmae_ctl_stop(shdev);
/* We cannot detect, which channel caused the error, have to reset all */
- for (i = 0; i < SH_DMAC_MAX_CHANNELS; i++) {
- struct sh_dmae_chan *sh_chan = shdev->chan[i];
- struct sh_desc *desc;
- LIST_HEAD(dl);
-
- if (!sh_chan)
- continue;
-
- spin_lock(&sh_chan->desc_lock);
-
- /* Stop the channel */
- dmae_halt(sh_chan);
-
- list_splice_init(&sh_chan->ld_queue, &dl);
-
- if (!list_empty(&dl)) {
- dev_dbg(sh_chan->dev, "Bring down channel %d\n", sh_chan->id);
- pm_runtime_put(sh_chan->dev);
- }
- sh_chan->pm_state = DMAE_PM_ESTABLISHED;
-
- spin_unlock(&sh_chan->desc_lock);
-
- /* Complete all */
- list_for_each_entry(desc, &dl, node) {
- struct dma_async_tx_descriptor *tx = &desc->async_tx;
- desc->mark = DESC_IDLE;
- if (tx->callback)
- tx->callback(tx->callback_param);
- }
-
- spin_lock(&sh_chan->desc_lock);
- list_splice(&dl, &sh_chan->ld_free);
- spin_unlock(&sh_chan->desc_lock);
-
- handled++;
- }
+ ret = shdma_reset(&shdev->shdma_dev);
sh_dmae_rst(shdev);
- return !!handled;
+ return ret;
}
static irqreturn_t sh_dmae_err(int irq, void *data)
@@ -991,35 +407,24 @@ static irqreturn_t sh_dmae_err(int irq, void *data)
if (!(dmaor_read(shdev) & DMAOR_AE))
return IRQ_NONE;
- sh_dmae_reset(data);
+ sh_dmae_reset(shdev);
return IRQ_HANDLED;
}
-static void dmae_do_tasklet(unsigned long data)
+static bool sh_dmae_desc_completed(struct shdma_chan *schan,
+ struct shdma_desc *sdesc)
{
- struct sh_dmae_chan *sh_chan = (struct sh_dmae_chan *)data;
- struct sh_desc *desc;
+ struct sh_dmae_chan *sh_chan = container_of(schan,
+ struct sh_dmae_chan, shdma_chan);
+ struct sh_dmae_desc *sh_desc = container_of(sdesc,
+ struct sh_dmae_desc, shdma_desc);
u32 sar_buf = sh_dmae_readl(sh_chan, SAR);
u32 dar_buf = sh_dmae_readl(sh_chan, DAR);
- spin_lock_irq(&sh_chan->desc_lock);
- list_for_each_entry(desc, &sh_chan->ld_queue, node) {
- if (desc->mark == DESC_SUBMITTED &&
- ((desc->direction == DMA_DEV_TO_MEM &&
- (desc->hw.dar + desc->hw.tcr) == dar_buf) ||
- (desc->hw.sar + desc->hw.tcr) == sar_buf)) {
- dev_dbg(sh_chan->dev, "done #%d@%p dst %u\n",
- desc->async_tx.cookie, &desc->async_tx,
- desc->hw.dar);
- desc->mark = DESC_COMPLETED;
- break;
- }
- }
- /* Next desc */
- sh_chan_xfer_ld_queue(sh_chan);
- spin_unlock_irq(&sh_chan->desc_lock);
-
- sh_dmae_chan_ld_cleanup(sh_chan, false);
+ return (sdesc->direction == DMA_DEV_TO_MEM &&
+ (sh_desc->hw.dar + sh_desc->hw.tcr) == dar_buf) ||
+ (sdesc->direction != DMA_DEV_TO_MEM &&
+ (sh_desc->hw.sar + sh_desc->hw.tcr) == sar_buf);
}
static bool sh_dmae_nmi_notify(struct sh_dmae_device *shdev)
@@ -1073,97 +478,174 @@ static struct notifier_block sh_dmae_nmi_notifier __read_mostly = {
static int __devinit sh_dmae_chan_probe(struct sh_dmae_device *shdev, int id,
int irq, unsigned long flags)
{
- int err;
const struct sh_dmae_channel *chan_pdata = &shdev->pdata->channel[id];
- struct platform_device *pdev = to_platform_device(shdev->common.dev);
- struct sh_dmae_chan *new_sh_chan;
+ struct shdma_dev *sdev = &shdev->shdma_dev;
+ struct platform_device *pdev = to_platform_device(sdev->dma_dev.dev);
+ struct sh_dmae_chan *sh_chan;
+ struct shdma_chan *schan;
+ int err;
- /* alloc channel */
- new_sh_chan = kzalloc(sizeof(struct sh_dmae_chan), GFP_KERNEL);
- if (!new_sh_chan) {
- dev_err(shdev->common.dev,
+ sh_chan = kzalloc(sizeof(struct sh_dmae_chan), GFP_KERNEL);
+ if (!sh_chan) {
+ dev_err(sdev->dma_dev.dev,
"No free memory for allocating dma channels!\n");
return -ENOMEM;
}
- new_sh_chan->pm_state = DMAE_PM_ESTABLISHED;
-
- /* reference struct dma_device */
- new_sh_chan->common.device = &shdev->common;
- dma_cookie_init(&new_sh_chan->common);
+ schan = &sh_chan->shdma_chan;
+ schan->max_xfer_len = SH_DMA_TCR_MAX + 1;
- new_sh_chan->dev = shdev->common.dev;
- new_sh_chan->id = id;
- new_sh_chan->irq = irq;
- new_sh_chan->base = shdev->chan_reg + chan_pdata->offset / sizeof(u32);
+ shdma_chan_probe(sdev, schan, id);
- /* Init DMA tasklet */
- tasklet_init(&new_sh_chan->tasklet, dmae_do_tasklet,
- (unsigned long)new_sh_chan);
-
- spin_lock_init(&new_sh_chan->desc_lock);
-
- /* Init descripter manage list */
- INIT_LIST_HEAD(&new_sh_chan->ld_queue);
- INIT_LIST_HEAD(&new_sh_chan->ld_free);
-
- /* Add the channel to DMA device channel list */
- list_add_tail(&new_sh_chan->common.device_node,
- &shdev->common.channels);
- shdev->common.chancnt++;
+ sh_chan->base = shdev->chan_reg + chan_pdata->offset / sizeof(u32);
+ /* set up channel irq */
if (pdev->id >= 0)
- snprintf(new_sh_chan->dev_id, sizeof(new_sh_chan->dev_id),
- "sh-dmae%d.%d", pdev->id, new_sh_chan->id);
+ snprintf(sh_chan->dev_id, sizeof(sh_chan->dev_id),
+ "sh-dmae%d.%d", pdev->id, id);
else
- snprintf(new_sh_chan->dev_id, sizeof(new_sh_chan->dev_id),
- "sh-dma%d", new_sh_chan->id);
+ snprintf(sh_chan->dev_id, sizeof(sh_chan->dev_id),
+ "sh-dma%d", id);
- /* set up channel irq */
- err = request_irq(irq, &sh_dmae_interrupt, flags,
- new_sh_chan->dev_id, new_sh_chan);
+ err = shdma_request_irq(schan, irq, flags, sh_chan->dev_id);
if (err) {
- dev_err(shdev->common.dev, "DMA channel %d request_irq error "
- "with return %d\n", id, err);
+ dev_err(sdev->dma_dev.dev,
+ "DMA channel %d request_irq error %d\n",
+ id, err);
goto err_no_irq;
}
- shdev->chan[id] = new_sh_chan;
+ shdev->chan[id] = sh_chan;
return 0;
err_no_irq:
/* remove from dmaengine device node */
- list_del(&new_sh_chan->common.device_node);
- kfree(new_sh_chan);
+ shdma_chan_remove(schan);
+ kfree(sh_chan);
return err;
}
static void sh_dmae_chan_remove(struct sh_dmae_device *shdev)
{
+ struct dma_device *dma_dev = &shdev->shdma_dev.dma_dev;
+ struct shdma_chan *schan;
int i;
- for (i = shdev->common.chancnt - 1 ; i >= 0 ; i--) {
- if (shdev->chan[i]) {
- struct sh_dmae_chan *sh_chan = shdev->chan[i];
+ shdma_for_each_chan(schan, &shdev->shdma_dev, i) {
+ struct sh_dmae_chan *sh_chan = container_of(schan,
+ struct sh_dmae_chan, shdma_chan);
+ BUG_ON(!schan);
- free_irq(sh_chan->irq, sh_chan);
+ shdma_free_irq(&sh_chan->shdma_chan);
- list_del(&sh_chan->common.device_node);
- kfree(sh_chan);
- shdev->chan[i] = NULL;
+ shdma_chan_remove(schan);
+ kfree(sh_chan);
+ }
+ dma_dev->chancnt = 0;
+}
+
+static void sh_dmae_shutdown(struct platform_device *pdev)
+{
+ struct sh_dmae_device *shdev = platform_get_drvdata(pdev);
+ sh_dmae_ctl_stop(shdev);
+}
+
+static int sh_dmae_runtime_suspend(struct device *dev)
+{
+ return 0;
+}
+
+static int sh_dmae_runtime_resume(struct device *dev)
+{
+ struct sh_dmae_device *shdev = dev_get_drvdata(dev);
+
+ return sh_dmae_rst(shdev);
+}
+
+#ifdef CONFIG_PM
+static int sh_dmae_suspend(struct device *dev)
+{
+ return 0;
+}
+
+static int sh_dmae_resume(struct device *dev)
+{
+ struct sh_dmae_device *shdev = dev_get_drvdata(dev);
+ int i, ret;
+
+ ret = sh_dmae_rst(shdev);
+ if (ret < 0)
+ dev_err(dev, "Failed to reset!\n");
+
+ for (i = 0; i < shdev->pdata->channel_num; i++) {
+ struct sh_dmae_chan *sh_chan = shdev->chan[i];
+ struct sh_dmae_slave *param = sh_chan->shdma_chan.dma_chan.private;
+
+ if (!sh_chan->shdma_chan.desc_num)
+ continue;
+
+ if (param) {
+ const struct sh_dmae_slave_config *cfg = param->config;
+ dmae_set_dmars(sh_chan, cfg->mid_rid);
+ dmae_set_chcr(sh_chan, cfg->chcr);
+ } else {
+ dmae_init(sh_chan);
}
}
- shdev->common.chancnt = 0;
+
+ return 0;
}
+#else
+#define sh_dmae_suspend NULL
+#define sh_dmae_resume NULL
+#endif
-static int __init sh_dmae_probe(struct platform_device *pdev)
+const struct dev_pm_ops sh_dmae_pm = {
+ .suspend = sh_dmae_suspend,
+ .resume = sh_dmae_resume,
+ .runtime_suspend = sh_dmae_runtime_suspend,
+ .runtime_resume = sh_dmae_runtime_resume,
+};
+
+static dma_addr_t sh_dmae_slave_addr(struct shdma_chan *schan)
+{
+ struct sh_dmae_slave *param = schan->dma_chan.private;
+
+ /*
+ * Implicit BUG_ON(!param)
+ * if (param != NULL), this is a successfully requested slave channel,
+ * therefore param->config != NULL too.
+ */
+ return param->config->addr;
+}
+
+static struct shdma_desc *sh_dmae_embedded_desc(void *buf, int i)
+{
+ return &((struct sh_dmae_desc *)buf)[i].shdma_desc;
+}
+
+static const struct shdma_ops sh_dmae_shdma_ops = {
+ .desc_completed = sh_dmae_desc_completed,
+ .halt_channel = sh_dmae_halt,
+ .channel_busy = sh_dmae_channel_busy,
+ .slave_addr = sh_dmae_slave_addr,
+ .desc_setup = sh_dmae_desc_setup,
+ .set_slave = sh_dmae_set_slave,
+ .setup_xfer = sh_dmae_setup_xfer,
+ .start_xfer = sh_dmae_start_xfer,
+ .embedded_desc = sh_dmae_embedded_desc,
+ .chan_irq = sh_dmae_chan_irq,
+};
+
+static int __devinit sh_dmae_probe(struct platform_device *pdev)
{
struct sh_dmae_pdata *pdata = pdev->dev.platform_data;
unsigned long irqflags = IRQF_DISABLED,
- chan_flag[SH_DMAC_MAX_CHANNELS] = {};
- int errirq, chan_irq[SH_DMAC_MAX_CHANNELS];
+ chan_flag[SH_DMAE_MAX_CHANNELS] = {};
+ int errirq, chan_irq[SH_DMAE_MAX_CHANNELS];
int err, i, irq_cnt = 0, irqres = 0, irq_cap = 0;
struct sh_dmae_device *shdev;
+ struct dma_device *dma_dev;
struct resource *chan, *dmars, *errirq_res, *chanirq_res;
/* get platform data */
@@ -1211,6 +693,8 @@ static int __init sh_dmae_probe(struct platform_device *pdev)
goto ealloc;
}
+ dma_dev = &shdev->shdma_dev.dma_dev;
+
shdev->chan_reg = ioremap(chan->start, resource_size(chan));
if (!shdev->chan_reg)
goto emapchan;
@@ -1220,8 +704,23 @@ static int __init sh_dmae_probe(struct platform_device *pdev)
goto emapdmars;
}
+ if (!pdata->slave_only)
+ dma_cap_set(DMA_MEMCPY, dma_dev->cap_mask);
+ if (pdata->slave && pdata->slave_num)
+ dma_cap_set(DMA_SLAVE, dma_dev->cap_mask);
+
+ /* Default transfer size of 32 bytes requires 32-byte alignment */
+ dma_dev->copy_align = LOG2_DEFAULT_XFER_SIZE;
+
+ shdev->shdma_dev.ops = &sh_dmae_shdma_ops;
+ shdev->shdma_dev.desc_size = sizeof(struct sh_dmae_desc);
+ err = shdma_init(&pdev->dev, &shdev->shdma_dev,
+ pdata->channel_num);
+ if (err < 0)
+ goto eshdma;
+
/* platform data */
- shdev->pdata = pdata;
+ shdev->pdata = pdev->dev.platform_data;
if (pdata->chcr_offset)
shdev->chcr_offset = pdata->chcr_offset;
@@ -1235,10 +734,10 @@ static int __init sh_dmae_probe(struct platform_device *pdev)
platform_set_drvdata(pdev, shdev);
- shdev->common.dev = &pdev->dev;
-
pm_runtime_enable(&pdev->dev);
- pm_runtime_get_sync(&pdev->dev);
+ err = pm_runtime_get_sync(&pdev->dev);
+ if (err < 0)
+ dev_err(&pdev->dev, "%s(): GET = %d\n", __func__, err);
spin_lock_irq(&sh_dmae_lock);
list_add_tail_rcu(&shdev->node, &sh_dmae_devices);
@@ -1249,27 +748,6 @@ static int __init sh_dmae_probe(struct platform_device *pdev)
if (err)
goto rst_err;
- INIT_LIST_HEAD(&shdev->common.channels);
-
- if (!pdata->slave_only)
- dma_cap_set(DMA_MEMCPY, shdev->common.cap_mask);
- if (pdata->slave && pdata->slave_num)
- dma_cap_set(DMA_SLAVE, shdev->common.cap_mask);
-
- shdev->common.device_alloc_chan_resources
- = sh_dmae_alloc_chan_resources;
- shdev->common.device_free_chan_resources = sh_dmae_free_chan_resources;
- shdev->common.device_prep_dma_memcpy = sh_dmae_prep_memcpy;
- shdev->common.device_tx_status = sh_dmae_tx_status;
- shdev->common.device_issue_pending = sh_dmae_memcpy_issue_pending;
-
- /* Compulsory for DMA_SLAVE fields */
- shdev->common.device_prep_slave_sg = sh_dmae_prep_slave_sg;
- shdev->common.device_control = sh_dmae_control;
-
- /* Default transfer size of 32 bytes requires 32-byte alignment */
- shdev->common.copy_align = LOG2_DEFAULT_XFER_SIZE;
-
#if defined(CONFIG_CPU_SH4) || defined(CONFIG_ARCH_SHMOBILE)
chanirq_res = platform_get_resource(pdev, IORESOURCE_IRQ, 1);
@@ -1301,7 +779,7 @@ static int __init sh_dmae_probe(struct platform_device *pdev)
!platform_get_resource(pdev, IORESOURCE_IRQ, 1)) {
/* Special case - all multiplexed */
for (; irq_cnt < pdata->channel_num; irq_cnt++) {
- if (irq_cnt < SH_DMAC_MAX_CHANNELS) {
+ if (irq_cnt < SH_DMAE_MAX_CHANNELS) {
chan_irq[irq_cnt] = chanirq_res->start;
chan_flag[irq_cnt] = IRQF_SHARED;
} else {
@@ -1312,7 +790,7 @@ static int __init sh_dmae_probe(struct platform_device *pdev)
} else {
do {
for (i = chanirq_res->start; i <= chanirq_res->end; i++) {
- if (irq_cnt >= SH_DMAC_MAX_CHANNELS) {
+ if (irq_cnt >= SH_DMAE_MAX_CHANNELS) {
irq_cap = 1;
break;
}
@@ -1328,7 +806,7 @@ static int __init sh_dmae_probe(struct platform_device *pdev)
chan_irq[irq_cnt++] = i;
}
- if (irq_cnt >= SH_DMAC_MAX_CHANNELS)
+ if (irq_cnt >= SH_DMAE_MAX_CHANNELS)
break;
chanirq_res = platform_get_resource(pdev,
@@ -1346,14 +824,19 @@ static int __init sh_dmae_probe(struct platform_device *pdev)
if (irq_cap)
dev_notice(&pdev->dev, "Attempting to register %d DMA "
"channels when a maximum of %d are supported.\n",
- pdata->channel_num, SH_DMAC_MAX_CHANNELS);
+ pdata->channel_num, SH_DMAE_MAX_CHANNELS);
pm_runtime_put(&pdev->dev);
- dma_async_device_register(&shdev->common);
+ err = dma_async_device_register(&shdev->shdma_dev.dma_dev);
+ if (err < 0)
+ goto edmadevreg;
return err;
+edmadevreg:
+ pm_runtime_get(&pdev->dev);
+
chan_probe_err:
sh_dmae_chan_remove(shdev);
@@ -1369,10 +852,11 @@ rst_err:
pm_runtime_put(&pdev->dev);
pm_runtime_disable(&pdev->dev);
+ platform_set_drvdata(pdev, NULL);
+ shdma_cleanup(&shdev->shdma_dev);
+eshdma:
if (dmars)
iounmap(shdev->dmars);
-
- platform_set_drvdata(pdev, NULL);
emapdmars:
iounmap(shdev->chan_reg);
synchronize_rcu();
@@ -1387,13 +871,14 @@ ermrdmars:
return err;
}
-static int __exit sh_dmae_remove(struct platform_device *pdev)
+static int __devexit sh_dmae_remove(struct platform_device *pdev)
{
struct sh_dmae_device *shdev = platform_get_drvdata(pdev);
+ struct dma_device *dma_dev = &shdev->shdma_dev.dma_dev;
struct resource *res;
int errirq = platform_get_irq(pdev, 0);
- dma_async_device_unregister(&shdev->common);
+ dma_async_device_unregister(dma_dev);
if (errirq > 0)
free_irq(errirq, shdev);
@@ -1402,11 +887,11 @@ static int __exit sh_dmae_remove(struct platform_device *pdev)
list_del_rcu(&shdev->node);
spin_unlock_irq(&sh_dmae_lock);
- /* channel data remove */
- sh_dmae_chan_remove(shdev);
-
pm_runtime_disable(&pdev->dev);
+ sh_dmae_chan_remove(shdev);
+ shdma_cleanup(&shdev->shdma_dev);
+
if (shdev->dmars)
iounmap(shdev->dmars);
iounmap(shdev->chan_reg);
@@ -1426,77 +911,14 @@ static int __exit sh_dmae_remove(struct platform_device *pdev)
return 0;
}
-static void sh_dmae_shutdown(struct platform_device *pdev)
-{
- struct sh_dmae_device *shdev = platform_get_drvdata(pdev);
- sh_dmae_ctl_stop(shdev);
-}
-
-static int sh_dmae_runtime_suspend(struct device *dev)
-{
- return 0;
-}
-
-static int sh_dmae_runtime_resume(struct device *dev)
-{
- struct sh_dmae_device *shdev = dev_get_drvdata(dev);
-
- return sh_dmae_rst(shdev);
-}
-
-#ifdef CONFIG_PM
-static int sh_dmae_suspend(struct device *dev)
-{
- return 0;
-}
-
-static int sh_dmae_resume(struct device *dev)
-{
- struct sh_dmae_device *shdev = dev_get_drvdata(dev);
- int i, ret;
-
- ret = sh_dmae_rst(shdev);
- if (ret < 0)
- dev_err(dev, "Failed to reset!\n");
-
- for (i = 0; i < shdev->pdata->channel_num; i++) {
- struct sh_dmae_chan *sh_chan = shdev->chan[i];
- struct sh_dmae_slave *param = sh_chan->common.private;
-
- if (!sh_chan->descs_allocated)
- continue;
-
- if (param) {
- const struct sh_dmae_slave_config *cfg = param->config;
- dmae_set_dmars(sh_chan, cfg->mid_rid);
- dmae_set_chcr(sh_chan, cfg->chcr);
- } else {
- dmae_init(sh_chan);
- }
- }
-
- return 0;
-}
-#else
-#define sh_dmae_suspend NULL
-#define sh_dmae_resume NULL
-#endif
-
-const struct dev_pm_ops sh_dmae_pm = {
- .suspend = sh_dmae_suspend,
- .resume = sh_dmae_resume,
- .runtime_suspend = sh_dmae_runtime_suspend,
- .runtime_resume = sh_dmae_runtime_resume,
-};
-
static struct platform_driver sh_dmae_driver = {
- .remove = __exit_p(sh_dmae_remove),
- .shutdown = sh_dmae_shutdown,
- .driver = {
+ .driver = {
.owner = THIS_MODULE,
- .name = "sh-dma-engine",
.pm = &sh_dmae_pm,
+ .name = SH_DMAE_DRV_NAME,
},
+ .remove = __devexit_p(sh_dmae_remove),
+ .shutdown = sh_dmae_shutdown,
};
static int __init sh_dmae_init(void)
@@ -1521,4 +943,4 @@ module_exit(sh_dmae_exit);
MODULE_AUTHOR("Nobuhiro Iwamatsu <[email protected]>");
MODULE_DESCRIPTION("Renesas SH DMA Engine driver");
MODULE_LICENSE("GPL");
-MODULE_ALIAS("platform:sh-dma-engine");
+MODULE_ALIAS("platform:" SH_DMAE_DRV_NAME);
diff --git a/drivers/dma/sh/shdma.h b/drivers/dma/sh/shdma.h
index 0b1d2c1..840e47d 100644
--- a/drivers/dma/sh/shdma.h
+++ b/drivers/dma/sh/shdma.h
@@ -13,42 +13,27 @@
#ifndef __DMA_SHDMA_H
#define __DMA_SHDMA_H
+#include <linux/shdma-base.h>
#include <linux/dmaengine.h>
#include <linux/interrupt.h>
#include <linux/list.h>
-#define SH_DMAC_MAX_CHANNELS 20
-#define SH_DMA_SLAVE_NUMBER 256
-#define SH_DMA_TCR_MAX 0x00FFFFFF /* 16MB */
+#define SH_DMAE_MAX_CHANNELS 20
+#define SH_DMAE_TCR_MAX 0x00FFFFFF /* 16MB */
struct device;
-enum dmae_pm_state {
- DMAE_PM_ESTABLISHED,
- DMAE_PM_BUSY,
- DMAE_PM_PENDING,
-};
-
struct sh_dmae_chan {
- spinlock_t desc_lock; /* Descriptor operation lock */
- struct list_head ld_queue; /* Link descriptors queue */
- struct list_head ld_free; /* Link descriptors free */
- struct dma_chan common; /* DMA common channel */
- struct device *dev; /* Channel device */
- struct tasklet_struct tasklet; /* Tasklet */
- int descs_allocated; /* desc count */
+ struct shdma_chan shdma_chan;
int xmit_shift; /* log_2(bytes_per_xfer) */
- int irq;
- int id; /* Raw id of this channel */
u32 __iomem *base;
char dev_id[16]; /* unique name per DMAC of channel */
int pm_error;
- enum dmae_pm_state pm_state;
};
struct sh_dmae_device {
- struct dma_device common;
- struct sh_dmae_chan *chan[SH_DMAC_MAX_CHANNELS];
+ struct shdma_dev shdma_dev;
+ struct sh_dmae_chan *chan[SH_DMAE_MAX_CHANNELS];
struct sh_dmae_pdata *pdata;
struct list_head node;
u32 __iomem *chan_reg;
@@ -57,10 +42,21 @@ struct sh_dmae_device {
u32 chcr_ie_bit;
};
-#define to_sh_chan(chan) container_of(chan, struct sh_dmae_chan, common)
+struct sh_dmae_regs {
+ u32 sar; /* SAR / source address */
+ u32 dar; /* DAR / destination address */
+ u32 tcr; /* TCR / transfer count */
+};
+
+struct sh_dmae_desc {
+ struct sh_dmae_regs hw;
+ struct shdma_desc shdma_desc;
+};
+
+#define to_sh_chan(chan) container_of(chan, struct sh_dmae_chan, shdma_chan)
#define to_sh_desc(lh) container_of(lh, struct sh_desc, node)
#define tx_to_sh_desc(tx) container_of(tx, struct sh_desc, async_tx)
-#define to_sh_dev(chan) container_of(chan->common.device,\
- struct sh_dmae_device, common)
+#define to_sh_dev(chan) container_of(chan->shdma_chan.dma_chan.device,\
+ struct sh_dmae_device, shdma_dev.dma_dev)
#endif /* __DMA_SHDMA_H */
diff --git a/include/linux/sh_dma.h b/include/linux/sh_dma.h
index e081e8e..7c8ca41 100644
--- a/include/linux/sh_dma.h
+++ b/include/linux/sh_dma.h
@@ -13,34 +13,21 @@
#include <linux/dmaengine.h>
#include <linux/list.h>
#include <linux/shdma-base.h>
+#include <linux/types.h>
+
+struct device;
/* Used by slave DMA clients to request DMA to/from a specific peripheral */
struct sh_dmae_slave {
- union {
- unsigned int slave_id; /* Set by the platform */
- struct shdma_slave shdma_slave;
- };
- struct device *dma_dev; /* Set by the platform */
- const struct sh_dmae_slave_config *config; /* Set by the driver */
-};
-
-struct sh_dmae_regs {
- u32 sar; /* SAR / source address */
- u32 dar; /* DAR / destination address */
- u32 tcr; /* TCR / transfer count */
-};
-
-struct sh_desc {
- struct sh_dmae_regs hw;
- struct list_head node;
- struct dma_async_tx_descriptor async_tx;
- enum dma_transfer_direction direction;
- dma_cookie_t cookie;
- size_t partial;
- int chunks;
- int mark;
+ struct shdma_slave shdma_slave; /* Set by the platform */
+ struct device *dma_dev; /* Set by the platform */
+ const struct sh_dmae_slave_config *config; /* Set by the driver */
};
+/*
+ * Supplied by platforms to specify, how a DMA channel has to be configured for
+ * a certain peripheral
+ */
struct sh_dmae_slave_config {
unsigned int slave_id;
dma_addr_t addr;
--
1.7.2.5
On Wed, May 09, 2012 at 05:09:18PM +0200, Guennadi Liakhovetski wrote:
> Signed-off-by: Guennadi Liakhovetski <[email protected]>
Acked-by: Mark Brown <[email protected]>
On Wed, May 09, 2012 at 05:09:20PM +0200, Guennadi Liakhovetski wrote:
> Signed-off-by: Guennadi Liakhovetski <[email protected]>
Acked-by: Mark Brown <[email protected]>
At Wed, 09 May 2012 17:09:19 +0200,
Guennadi wrote:
>
> Signed-off-by: Guennadi Liakhovetski <[email protected]>
> Cc: Greg Kroah-Hartman <[email protected]>
> ---
Acked-by: Kuninori Morimoto <[email protected]>
> drivers/usb/renesas_usbhs/fifo.c | 10 +++++-----
> 1 files changed, 5 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/usb/renesas_usbhs/fifo.c b/drivers/usb/renesas_usbhs/fifo.c
> index 6ec7f83..fc597a4 100644
> --- a/drivers/usb/renesas_usbhs/fifo.c
> +++ b/drivers/usb/renesas_usbhs/fifo.c
> @@ -994,7 +994,7 @@ static bool usbhsf_dma_filter(struct dma_chan *chan, void *param)
> *
> * usbhs doesn't recognize id = 0 as valid DMA
> */
> - if (0 == slave->slave_id)
> + if (0 == slave->shdma_slave.slave_id)
> return false;
>
> chan->private = slave;
> @@ -1173,8 +1173,8 @@ int usbhs_fifo_probe(struct usbhs_priv *priv)
> fifo->port = D0FIFO;
> fifo->sel = D0FIFOSEL;
> fifo->ctr = D0FIFOCTR;
> - fifo->tx_slave.slave_id = usbhs_get_dparam(priv, d0_tx_id);
> - fifo->rx_slave.slave_id = usbhs_get_dparam(priv, d0_rx_id);
> + fifo->tx_slave.shdma_slave.slave_id = usbhs_get_dparam(priv, d0_tx_id);
> + fifo->rx_slave.shdma_slave.slave_id = usbhs_get_dparam(priv, d0_rx_id);
>
> /* D1FIFO */
> fifo = usbhsf_get_d1fifo(priv);
> @@ -1182,8 +1182,8 @@ int usbhs_fifo_probe(struct usbhs_priv *priv)
> fifo->port = D1FIFO;
> fifo->sel = D1FIFOSEL;
> fifo->ctr = D1FIFOCTR;
> - fifo->tx_slave.slave_id = usbhs_get_dparam(priv, d1_tx_id);
> - fifo->rx_slave.slave_id = usbhs_get_dparam(priv, d1_rx_id);
> + fifo->tx_slave.shdma_slave.slave_id = usbhs_get_dparam(priv, d1_tx_id);
> + fifo->rx_slave.shdma_slave.slave_id = usbhs_get_dparam(priv, d1_rx_id);
>
> return 0;
> }
> --
> 1.7.2.5
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-sh" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
---
Kuninori Morimoto
$B!V$d$j$?$$$H;W$C$F$$$k$s$@$1$I!W$=$s$J$s$8$c;W$$B3$1$F0l@8=*$o$k$h!#(B
$B$d$k%d%D$O$d$k$&$($K!"$d$jB3$1$F$$$k$s$@!#@.8y$7$F$b!"$^$@$d$jB3$1$F$$$k!#(B
$B!]!!;3K\0lNO!J:n2H!K(B
At Wed, 09 May 2012 17:09:20 +0200,
Guennadi wrote:
>
> Signed-off-by: Guennadi Liakhovetski <[email protected]>
> Cc: Liam Girdwood <[email protected]>
> Cc: Mark Brown <[email protected]>
> ---
Acked-by: Kuninori Morimoto <[email protected]>
> sound/soc/sh/fsi.c | 4 ++--
> 1 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/sound/soc/sh/fsi.c b/sound/soc/sh/fsi.c
> index 74ed2df..c28314b 100644
> --- a/sound/soc/sh/fsi.c
> +++ b/sound/soc/sh/fsi.c
> @@ -1532,8 +1532,8 @@ static void fsi_handler_init(struct fsi_priv *fsi)
> fsi->capture.priv = fsi;
>
> if (fsi->info->tx_id) {
> - fsi->playback.slave.slave_id = fsi->info->tx_id;
> - fsi->playback.handler = &fsi_dma_push_handler;
> + fsi->playback.slave.shdma_slave.slave_id = fsi->info->tx_id;
> + fsi->playback.handler = &fsi_dma_push_handler;
> }
> }
>
> --
> 1.7.2.5
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-sh" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, May 09, 2012 at 04:59:14PM -0700, [email protected] wrote:
> At Wed, 09 May 2012 17:09:19 +0200,
> Guennadi wrote:
> >
> > Signed-off-by: Guennadi Liakhovetski <[email protected]>
> > Cc: Greg Kroah-Hartman <[email protected]>
> > ---
>
> Acked-by: Kuninori Morimoto <[email protected]>
Acked-by: Greg Kroah-Hartman <[email protected]>
On Wed, May 09, 2012 at 05:09:17PM +0200, Guennadi Liakhovetski wrote:
> Signed-off-by: Guennadi Liakhovetski <[email protected]>
> Cc: Alan Cox <[email protected]>
> ---
> drivers/tty/serial/sh-sci.c | 8 ++++----
> 1 files changed, 4 insertions(+), 4 deletions(-)
>
Acked-by: Paul Mundt <[email protected]>
On Wed, 2012-05-09 at 17:09 +0200, Guennadi Liakhovetski wrote:
> This patch extracts code from shdma.c, that does not directly deal with
> hardware implementation details and can be re-used with diverse DMA
> controller variants, found on SH-based SoCs.
Have you see recent Russell's virtual dma patches [1]? I like the idea
and simplifies things for implementation in each driver.
Said that, do we need this libarry. Can you see if you can use/enhance
that one.
[snip]
>
> Signed-off-by: Guennadi Liakhovetski <[email protected]>
> Cc: Sascha Hauer <[email protected]>
> ---
> drivers/dma/sh/Makefile | 1 +
> drivers/dma/sh/shdma-base.c | 868 +++++++++++++++++++++++++++++++++++++++++++
> include/linux/shdma-base.h | 123 ++++++
> 3 files changed, 992 insertions(+), 0 deletions(-)
> create mode 100644 drivers/dma/sh/shdma-base.c
> create mode 100644 include/linux/shdma-base.h
>
> diff --git a/drivers/dma/sh/Makefile b/drivers/dma/sh/Makefile
> index ad4981a..54ae957 100644
> --- a/drivers/dma/sh/Makefile
> +++ b/drivers/dma/sh/Makefile
> @@ -1 +1,2 @@
> +obj-$(CONFIG_SH_DMAE) += shdma-base.o
> obj-$(CONFIG_SH_DMAE) += shdma.o
> diff --git a/drivers/dma/sh/shdma-base.c b/drivers/dma/sh/shdma-base.c
> new file mode 100644
> index 0000000..ff060d0
> --- /dev/null
> +++ b/drivers/dma/sh/shdma-base.c
> @@ -0,0 +1,868 @@
> +/*
> + * Dmaengine driver base library for DMA controllers, found on SH-based SoCs
> + *
> + * extracted from shdma.c
> + *
> + * Copyright (C) 2011-2012 Guennadi Liakhovetski <[email protected]>
> + * Copyright (C) 2009 Nobuhiro Iwamatsu <[email protected]>
> + * Copyright (C) 2009 Renesas Solutions, Inc. All rights reserved.
> + * Copyright (C) 2007 Freescale Semiconductor, Inc. All rights reserved.
> + *
> + * This is free software; you can redistribute it and/or modify
> + * it under the terms of version 2 of the GNU General Public License as
> + * published by the Free Software Foundation.
> + */
> +
> +#include <linux/delay.h>
> +#include <linux/shdma-base.h>
> +#include <linux/dmaengine.h>
> +#include <linux/init.h>
> +#include <linux/interrupt.h>
> +#include <linux/module.h>
> +#include <linux/pm_runtime.h>
> +#include <linux/slab.h>
> +#include <linux/spinlock.h>
> +
> +#include "../dmaengine.h"
> +
> +/* DMA descriptor control */
> +enum shdma_desc_status {
> + DESC_IDLE,
> + DESC_PREPARED,
> + DESC_SUBMITTED,
> + DESC_COMPLETED, /* completed, have to call callback */
> + DESC_WAITING, /* callback called, waiting for ack / re-submit */
> +};
> +
> +#define NR_DESCS_PER_CHANNEL 32
> +
> +#define to_shdma_chan(c) container_of(c, struct shdma_chan, dma_chan)
> +#define to_shdma_dev(d) container_of(d, struct shdma_dev, dma_dev)
> +
> +/*
> + * For slave DMA we assume, that there is a finite number of DMA slaves in the
> + * system, and that each such slave can only use a finite number of channels.
> + * We use slave channel IDs to make sure, that no such slave channel ID is
> + * allocated more than once.
> + */
> +static unsigned int slave_num = 256;
> +module_param(slave_num, uint, 0444);
> +
> +/* A bitmask with slave_num bits */
> +static unsigned long *shdma_slave_used;
> +
> +/* Called under spin_lock_irq(&schan->chan_lock") */
> +static void shdma_chan_xfer_ld_queue(struct shdma_chan *schan)
> +{
> + struct shdma_dev *sdev = to_shdma_dev(schan->dma_chan.device);
> + const struct shdma_ops *ops = sdev->ops;
> + struct shdma_desc *sdesc;
> +
> + /* DMA work check */
> + if (ops->channel_busy(schan))
> + return;
> +
> + /* Find the first not transferred descriptor */
> + list_for_each_entry(sdesc, &schan->ld_queue, node)
> + if (sdesc->mark == DESC_SUBMITTED) {
> + ops->start_xfer(schan, sdesc);
> + break;
> + }
> +}
> +
> +static dma_cookie_t shdma_tx_submit(struct dma_async_tx_descriptor *tx)
> +{
> + struct shdma_desc *chunk, *c, *desc =
> + container_of(tx, struct shdma_desc, async_tx),
> + *last = desc;
> + struct shdma_chan *schan = to_shdma_chan(tx->chan);
> + struct shdma_slave *slave = tx->chan->private;
> + dma_async_tx_callback callback = tx->callback;
> + dma_cookie_t cookie;
> + bool power_up;
> +
> + spin_lock_irq(&schan->chan_lock);
> +
> + power_up = list_empty(&schan->ld_queue);
> +
> + cookie = dma_cookie_assign(tx);
> +
> + /* Mark all chunks of this descriptor as submitted, move to the queue */
> + list_for_each_entry_safe(chunk, c, desc->node.prev, node) {
> + /*
> + * All chunks are on the global ld_free, so, we have to find
> + * the end of the chain ourselves
> + */
> + if (chunk != desc && (chunk->mark == DESC_IDLE ||
> + chunk->async_tx.cookie > 0 ||
> + chunk->async_tx.cookie == -EBUSY ||
> + &chunk->node == &schan->ld_free))
> + break;
> + chunk->mark = DESC_SUBMITTED;
> + /* Callback goes to the last chunk */
> + chunk->async_tx.callback = NULL;
> + chunk->cookie = cookie;
> + list_move_tail(&chunk->node, &schan->ld_queue);
> + last = chunk;
> +
> + dev_dbg(schan->dev, "submit #%d@%p on %d\n",
> + tx->cookie, &last->async_tx, schan->id);
> + }
> +
> + last->async_tx.callback = callback;
> + last->async_tx.callback_param = tx->callback_param;
> +
> + if (power_up) {
> + int ret;
> + schan->pm_state = SHDMA_PM_BUSY;
> +
> + ret = pm_runtime_get(schan->dev);
any reason why this is here and not in issue pending?
> +
> + spin_unlock_irq(&schan->chan_lock);
> + if (ret < 0)
> + dev_err(schan->dev, "%s(): GET = %d\n", __func__, ret);
> +
> + pm_runtime_barrier(schan->dev);
> +
> + spin_lock_irq(&schan->chan_lock);
> +
> + /* Have we been reset, while waiting? */
> + if (schan->pm_state != SHDMA_PM_ESTABLISHED) {
> + struct shdma_dev *sdev =
> + to_shdma_dev(schan->dma_chan.device);
> + const struct shdma_ops *ops = sdev->ops;
> + dev_dbg(schan->dev, "Bring up channel %d\n",
> + schan->id);
> + /*
> + * TODO: .xfer_setup() might fail on some platforms.
> + * Make it int then, on error remove chunks from the
> + * queue again
> + */
> + ops->setup_xfer(schan, slave);
> +
> + if (schan->pm_state == SHDMA_PM_PENDING)
> + shdma_chan_xfer_ld_queue(schan);
> + schan->pm_state = SHDMA_PM_ESTABLISHED;
> + }
> + } else {
> + /*
> + * Tell .device_issue_pending() not to run the queue, interrupts
> + * will do it anyway
> + */
> + schan->pm_state = SHDMA_PM_PENDING;
> + }
> +
> + spin_unlock_irq(&schan->chan_lock);
> +
> + return cookie;
> +}
> +
> +/* Called with desc_lock held */
> +static struct shdma_desc *shdma_get_desc(struct shdma_chan *schan)
> +{
> + struct shdma_desc *sdesc;
> +
> + list_for_each_entry(sdesc, &schan->ld_free, node)
> + if (sdesc->mark != DESC_PREPARED) {
> + BUG_ON(sdesc->mark != DESC_IDLE);
> + list_del(&sdesc->node);
> + return sdesc;
> + }
> +
> + return NULL;
> +}
> +
> +static int shdma_alloc_chan_resources(struct dma_chan *chan)
> +{
> + struct shdma_chan *schan = to_shdma_chan(chan);
> + struct shdma_dev *sdev = to_shdma_dev(schan->dma_chan.device);
> + const struct shdma_ops *ops = sdev->ops;
> + struct shdma_desc *desc;
> + struct shdma_slave *slave = chan->private;
> + int ret, i;
> +
> + /*
> + * This relies on the guarantee from dmaengine that alloc_chan_resources
> + * never runs concurrently with itself or free_chan_resources.
> + */
> + if (slave) {
> + if (slave->slave_id >= slave_num) {
> + ret = -EINVAL;
> + goto evalid;
> + }
> +
> + if (test_and_set_bit(slave->slave_id, shdma_slave_used)) {
> + ret = -EBUSY;
> + goto etestused;
> + }
> +
> + ret = ops->set_slave(schan, slave);
> + if (ret < 0)
> + goto esetslave;
> + }
> +
> + schan->desc = kcalloc(NR_DESCS_PER_CHANNEL,
> + sdev->desc_size, GFP_KERNEL);
> + if (!schan->desc) {
> + ret = -ENOMEM;
> + goto edescalloc;
> + }
> + schan->desc_num = NR_DESCS_PER_CHANNEL;
> +
> + for (i = 0; i < NR_DESCS_PER_CHANNEL; i++) {
> + desc = ops->embedded_desc(schan->desc, i);
> + dma_async_tx_descriptor_init(&desc->async_tx,
> + &schan->dma_chan);
> + desc->async_tx.tx_submit = shdma_tx_submit;
> + desc->mark = DESC_IDLE;
> +
> + list_add(&desc->node, &schan->ld_free);
> + }
> +
> + return NR_DESCS_PER_CHANNEL;
> +
> +edescalloc:
> + if (slave)
> +esetslave:
> + clear_bit(slave->slave_id, shdma_slave_used);
> +etestused:
> +evalid:
> + chan->private = NULL;
> + return ret;
> +}
> +
> +static dma_async_tx_callback __ld_cleanup(struct shdma_chan *schan, bool all)
> +{
> + struct shdma_desc *desc, *_desc;
> + /* Is the "exposed" head of a chain acked? */
> + bool head_acked = false;
> + dma_cookie_t cookie = 0;
> + dma_async_tx_callback callback = NULL;
> + void *param = NULL;
> + unsigned long flags;
> +
> + 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;
> +
> + BUG_ON(tx->cookie > 0 && tx->cookie != desc->cookie);
> + BUG_ON(desc->mark != DESC_SUBMITTED &&
> + desc->mark != DESC_COMPLETED &&
> + desc->mark != DESC_WAITING);
> +
> + /*
> + * queue is ordered, and we use this loop to (1) clean up all
> + * completed descriptors, and to (2) update descriptor flags of
> + * any chunks in a (partially) completed chain
> + */
> + if (!all && desc->mark == DESC_SUBMITTED &&
> + desc->cookie != cookie)
> + break;
> +
> + if (tx->cookie > 0)
> + cookie = tx->cookie;
> +
> + if (desc->mark == DESC_COMPLETED && desc->chunks == 1) {
> + if (schan->dma_chan.completed_cookie != desc->cookie - 1)
> + dev_dbg(schan->dev,
> + "Completing cookie %d, expected %d\n",
> + desc->cookie,
> + schan->dma_chan.completed_cookie + 1);
> + schan->dma_chan.completed_cookie = desc->cookie;
> + }
> +
> + /* Call callback on the last chunk */
> + if (desc->mark == DESC_COMPLETED && tx->callback) {
> + desc->mark = DESC_WAITING;
> + 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);
> + break;
> + }
> +
> + if (tx->cookie > 0 || tx->cookie == -EBUSY) {
> + if (desc->mark == DESC_COMPLETED) {
> + BUG_ON(tx->cookie < 0);
> + desc->mark = DESC_WAITING;
> + }
> + head_acked = async_tx_test_ack(tx);
> + } else {
> + switch (desc->mark) {
> + case DESC_COMPLETED:
> + desc->mark = DESC_WAITING;
> + /* Fall through */
> + case DESC_WAITING:
> + if (head_acked)
> + async_tx_ack(&desc->async_tx);
> + }
> + }
> +
> + dev_dbg(schan->dev, "descriptor %p #%d completed.\n",
> + tx, tx->cookie);
> +
> + if (((desc->mark == DESC_COMPLETED ||
> + desc->mark == DESC_WAITING) &&
> + async_tx_test_ack(&desc->async_tx)) || all) {
> + /* Remove from ld_queue list */
> + desc->mark = DESC_IDLE;
> +
> + list_move(&desc->node, &schan->ld_free);
> +
> + if (list_empty(&schan->ld_queue)) {
> + dev_dbg(schan->dev, "Bring down channel %d\n", schan->id);
> + pm_runtime_put(schan->dev);
> + schan->pm_state = SHDMA_PM_ESTABLISHED;
> + }
> + }
> + }
> +
> + if (all && !callback)
> + /*
> + * Terminating and the loop completed normally: forgive
> + * uncompleted cookies
> + */
> + schan->dma_chan.completed_cookie = schan->dma_chan.cookie;
> +
> + spin_unlock_irqrestore(&schan->chan_lock, flags);
> +
> + if (callback)
> + callback(param);
> +
> + return callback;
> +}
> +
> +/*
> + * shdma_chan_ld_cleanup - Clean up link descriptors
> + *
> + * Clean up the ld_queue of DMA channel.
> + */
> +static void shdma_chan_ld_cleanup(struct shdma_chan *schan, bool all)
> +{
> + while (__ld_cleanup(schan, all))
> + ;
> +}
> +
> +/*
> + * shdma_free_chan_resources - Free all resources of the channel.
> + */
> +static void shdma_free_chan_resources(struct dma_chan *chan)
> +{
> + struct shdma_chan *schan = to_shdma_chan(chan);
> + struct shdma_dev *sdev = to_shdma_dev(chan->device);
> + const struct shdma_ops *ops = sdev->ops;
> + LIST_HEAD(list);
> +
> + /* Protect against ISR */
> + spin_lock_irq(&schan->chan_lock);
> + ops->halt_channel(schan);
> + spin_unlock_irq(&schan->chan_lock);
> +
> + /* Now no new interrupts will occur */
> +
> + /* Prepared and not submitted descriptors can still be on the queue */
> + if (!list_empty(&schan->ld_queue))
> + shdma_chan_ld_cleanup(schan, true);
> +
> + if (chan->private) {
> + /* The caller is holding dma_list_mutex */
> + struct shdma_slave *slave = chan->private;
> + clear_bit(slave->slave_id, shdma_slave_used);
> + chan->private = NULL;
> + }
> +
> + spin_lock_irq(&schan->chan_lock);
> +
> + list_splice_init(&schan->ld_free, &list);
> + schan->desc_num = 0;
> +
> + spin_unlock_irq(&schan->chan_lock);
> +
> + kfree(schan->desc);
> +}
> +
> +/**
> + * shdma_add_desc - get, set up and return one transfer descriptor
> + * @schan: DMA channel
> + * @flags: DMA transfer flags
> + * @dst: destination DMA address, incremented when direction equals
> + * DMA_DEV_TO_MEM or DMA_MEM_TO_MEM
> + * @src: source DMA address, incremented when direction equals
> + * DMA_MEM_TO_DEV or DMA_MEM_TO_MEM
> + * @len: DMA transfer length
> + * @first: if NULL, set to the current descriptor and cookie set to -EBUSY
> + * @direction: needed for slave DMA to decide which address to keep constant,
> + * equals DMA_MEM_TO_MEM for MEMCPY
> + * Returns 0 or an error
> + * Locks: called with desc_lock held
> + */
> +static struct shdma_desc *shdma_add_desc(struct shdma_chan *schan,
> + unsigned long flags, dma_addr_t *dst, dma_addr_t *src, size_t *len,
> + struct shdma_desc **first, enum dma_transfer_direction direction)
> +{
> + struct shdma_dev *sdev = to_shdma_dev(schan->dma_chan.device);
> + const struct shdma_ops *ops = sdev->ops;
> + struct shdma_desc *new;
> + size_t copy_size = *len;
> +
> + if (!copy_size)
> + return NULL;
> +
> + /* Allocate the link descriptor from the free list */
> + new = shdma_get_desc(schan);
> + if (!new) {
> + dev_err(schan->dev, "No free link descriptor available\n");
> + return NULL;
> + }
> +
> + ops->desc_setup(schan, new, *src, *dst, ©_size);
> +
> + if (!*first) {
> + /* First desc */
> + new->async_tx.cookie = -EBUSY;
> + *first = new;
> + } else {
> + /* Other desc - invisible to the user */
> + new->async_tx.cookie = -EINVAL;
> + }
> +
> + dev_dbg(schan->dev,
> + "chaining (%u/%u)@%x -> %x with %p, cookie %d\n",
> + copy_size, *len, *src, *dst, &new->async_tx,
> + new->async_tx.cookie);
> +
> + new->mark = DESC_PREPARED;
> + new->async_tx.flags = flags;
> + new->direction = direction;
> +
> + *len -= copy_size;
> + if (direction == DMA_MEM_TO_MEM || direction == DMA_MEM_TO_DEV)
> + *src += copy_size;
> + if (direction == DMA_MEM_TO_MEM || direction == DMA_DEV_TO_MEM)
> + *dst += copy_size;
> +
> + return new;
> +}
> +
> +/*
> + * shdma_prep_sg - prepare transfer descriptors from an SG list
> + *
> + * Common routine for public (MEMCPY) and slave DMA. The MEMCPY case is also
> + * converted to scatter-gather to guarantee consistent locking and a correct
> + * list manipulation. For slave DMA direction carries the usual meaning, and,
> + * logically, the SG list is RAM and the addr variable contains slave address,
> + * e.g., the FIFO I/O register. For MEMCPY direction equals DMA_MEM_TO_MEM
> + * and the SG list contains only one element and points at the source buffer.
> + */
> +static struct dma_async_tx_descriptor *shdma_prep_sg(struct shdma_chan *schan,
> + struct scatterlist *sgl, unsigned int sg_len, dma_addr_t *addr,
> + enum dma_transfer_direction direction, unsigned long flags)
> +{
> + struct scatterlist *sg;
> + struct shdma_desc *first = NULL, *new = NULL /* compiler... */;
> + LIST_HEAD(tx_list);
> + int chunks = 0;
> + unsigned long irq_flags;
> + int i;
> +
> + for_each_sg(sgl, sg, sg_len, i)
> + chunks += DIV_ROUND_UP(sg_dma_len(sg), schan->max_xfer_len);
> +
> + /* Have to lock the whole loop to protect against concurrent release */
> + spin_lock_irqsave(&schan->chan_lock, irq_flags);
> +
> + /*
> + * Chaining:
> + * first descriptor is what user is dealing with in all API calls, its
> + * cookie is at first set to -EBUSY, at tx-submit to a positive
> + * number
> + * if more than one chunk is needed further chunks have cookie = -EINVAL
> + * the last chunk, if not equal to the first, has cookie = -ENOSPC
> + * all chunks are linked onto the tx_list head with their .node heads
> + * only during this function, then they are immediately spliced
> + * back onto the free list in form of a chain
> + */
> + for_each_sg(sgl, sg, sg_len, i) {
> + dma_addr_t sg_addr = sg_dma_address(sg);
> + size_t len = sg_dma_len(sg);
> +
> + if (!len)
> + goto err_get_desc;
> +
> + do {
> + dev_dbg(schan->dev, "Add SG #%d@%p[%d], dma %llx\n",
> + i, sg, len, (unsigned long long)sg_addr);
> +
> + if (direction == DMA_DEV_TO_MEM)
> + new = shdma_add_desc(schan, flags,
> + &sg_addr, addr, &len, &first,
> + direction);
> + else
> + new = shdma_add_desc(schan, flags,
> + addr, &sg_addr, &len, &first,
> + direction);
> + if (!new)
> + goto err_get_desc;
> +
> + new->chunks = chunks--;
> + list_add_tail(&new->node, &tx_list);
> + } while (len);
> + }
> +
> + if (new != first)
> + new->async_tx.cookie = -ENOSPC;
> +
> + /* Put them back on the free list, so, they don't get lost */
> + list_splice_tail(&tx_list, &schan->ld_free);
> +
> + spin_unlock_irqrestore(&schan->chan_lock, irq_flags);
> +
> + return &first->async_tx;
> +
> +err_get_desc:
> + list_for_each_entry(new, &tx_list, node)
> + new->mark = DESC_IDLE;
> + list_splice(&tx_list, &schan->ld_free);
> +
> + spin_unlock_irqrestore(&schan->chan_lock, irq_flags);
> +
> + return NULL;
> +}
> +
> +static struct dma_async_tx_descriptor *shdma_prep_memcpy(
> + struct dma_chan *chan, dma_addr_t dma_dest, dma_addr_t dma_src,
> + size_t len, unsigned long flags)
> +{
> + struct shdma_chan *schan = to_shdma_chan(chan);
> + struct scatterlist sg;
> +
> + if (!chan || !len)
> + return NULL;
> +
> + BUG_ON(!schan->desc_num);
> +
> + sg_init_table(&sg, 1);
> + sg_set_page(&sg, pfn_to_page(PFN_DOWN(dma_src)), len,
> + offset_in_page(dma_src));
> + sg_dma_address(&sg) = dma_src;
> + sg_dma_len(&sg) = len;
> +
> + return shdma_prep_sg(schan, &sg, 1, &dma_dest, DMA_MEM_TO_MEM, flags);
> +}
> +
> +static struct dma_async_tx_descriptor *shdma_prep_slave_sg(
> + struct dma_chan *chan, struct scatterlist *sgl, unsigned int sg_len,
> + enum dma_transfer_direction direction, unsigned long flags, void *context)
> +{
> + struct shdma_chan *schan = to_shdma_chan(chan);
> + struct shdma_dev *sdev = to_shdma_dev(schan->dma_chan.device);
> + const struct shdma_ops *ops = sdev->ops;
> + struct shdma_slave *slave = chan->private;
> + dma_addr_t slave_addr;
> +
> + if (!chan)
> + return NULL;
> +
> + BUG_ON(!schan->desc_num);
> +
> + /* Someone calling slave DMA on a generic channel? */
> + if (!slave || !sg_len) {
> + dev_warn(schan->dev, "%s: bad parameter: %p, %d, %d\n",
> + __func__, slave, sg_len, slave ? slave->slave_id : -1);
> + return NULL;
> + }
> +
> + slave_addr = ops->slave_addr(schan);
> +
> + return shdma_prep_sg(schan, sgl, sg_len, &slave_addr,
> + direction, flags);
> +}
> +
> +static int shdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> + unsigned long arg)
> +{
> + struct shdma_chan *schan = to_shdma_chan(chan);
> + struct shdma_dev *sdev = to_shdma_dev(chan->device);
> + const struct shdma_ops *ops = sdev->ops;
> + unsigned long flags;
> +
> + /* Only supports DMA_TERMINATE_ALL */
> + if (cmd != DMA_TERMINATE_ALL)
> + return -ENXIO;
> +
> + if (!chan)
> + return -EINVAL;
> +
> + spin_lock_irqsave(&schan->chan_lock, flags);
> +
> + ops->halt_channel(schan);
> +
> + spin_unlock_irqrestore(&schan->chan_lock, flags);
> +
> + shdma_chan_ld_cleanup(schan, true);
> +
> + return 0;
> +}
> +
> +static void shdma_issue_pending(struct dma_chan *chan)
> +{
> + struct shdma_chan *schan = to_shdma_chan(chan);
> +
> + spin_lock_irq(&schan->chan_lock);
> + if (schan->pm_state == SHDMA_PM_ESTABLISHED)
> + shdma_chan_xfer_ld_queue(schan);
> + else
> + schan->pm_state = SHDMA_PM_PENDING;
> + spin_unlock_irq(&schan->chan_lock);
> +}
> +
> +static enum dma_status shdma_tx_status(struct dma_chan *chan,
> + dma_cookie_t cookie,
> + struct dma_tx_state *txstate)
> +{
> + struct shdma_chan *schan = to_shdma_chan(chan);
> + enum dma_status status;
> + unsigned long flags;
> +
> + shdma_chan_ld_cleanup(schan, false);
> +
> + spin_lock_irqsave(&schan->chan_lock, flags);
> +
> + status = dma_cookie_status(chan, cookie, txstate);
> +
> + /*
> + * If we don't find cookie on the queue, it has been aborted and we have
> + * to report error
> + */
> + if (status != DMA_SUCCESS) {
> + struct shdma_desc *sdesc;
> + status = DMA_ERROR;
> + list_for_each_entry(sdesc, &schan->ld_queue, node)
> + if (sdesc->cookie == cookie) {
> + status = DMA_IN_PROGRESS;
> + break;
> + }
> + }
> +
> + spin_unlock_irqrestore(&schan->chan_lock, flags);
> +
> + return status;
> +}
> +
> +/* Called from error IRQ or NMI */
> +bool shdma_reset(struct shdma_dev *sdev)
> +{
> + const struct shdma_ops *ops = sdev->ops;
> + struct shdma_chan *schan;
> + unsigned int handled = 0;
> + int i;
> +
> + /* Reset all channels */
> + shdma_for_each_chan(schan, sdev, i) {
> + struct shdma_desc *sdesc;
> + LIST_HEAD(dl);
> +
> + if (!schan)
> + continue;
> +
> + spin_lock(&schan->chan_lock);
> +
> + /* Stop the channel */
> + ops->halt_channel(schan);
> +
> + list_splice_init(&schan->ld_queue, &dl);
> +
> + if (!list_empty(&dl)) {
> + dev_dbg(schan->dev, "Bring down channel %d\n", schan->id);
> + pm_runtime_put(schan->dev);
> + }
> + schan->pm_state = SHDMA_PM_ESTABLISHED;
> +
> + spin_unlock(&schan->chan_lock);
> +
> + /* 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);
> + }
> +
> + spin_lock(&schan->chan_lock);
> + list_splice(&dl, &schan->ld_free);
> + spin_unlock(&schan->chan_lock);
> +
> + handled++;
> + }
> +
> + return !!handled;
> +}
> +EXPORT_SYMBOL(shdma_reset);
> +
> +static irqreturn_t chan_irq(int irq, void *dev)
> +{
> + struct shdma_chan *schan = dev;
> + const struct shdma_ops *ops =
> + to_shdma_dev(schan->dma_chan.device)->ops;
> + irqreturn_t ret;
> +
> + spin_lock(&schan->chan_lock);
> +
> + ret = ops->chan_irq(schan, irq) ? IRQ_WAKE_THREAD : IRQ_NONE;
> +
> + spin_unlock(&schan->chan_lock);
> +
> + return ret;
> +}
> +
> +static irqreturn_t chan_irqt(int irq, void *dev)
> +{
> + struct shdma_chan *schan = dev;
> + const struct shdma_ops *ops =
> + to_shdma_dev(schan->dma_chan.device)->ops;
> + struct shdma_desc *sdesc;
> +
> + spin_lock_irq(&schan->chan_lock);
> + list_for_each_entry(sdesc, &schan->ld_queue, node) {
> + if (sdesc->mark == DESC_SUBMITTED &&
> + ops->desc_completed(schan, sdesc)) {
> + dev_dbg(schan->dev, "done #%d@%p\n",
> + sdesc->async_tx.cookie, &sdesc->async_tx);
> + sdesc->mark = DESC_COMPLETED;
> + break;
> + }
> + }
> + /* Next desc */
> + shdma_chan_xfer_ld_queue(schan);
> + spin_unlock_irq(&schan->chan_lock);
> +
> + shdma_chan_ld_cleanup(schan, false);
> +
> + return IRQ_HANDLED;
> +}
> +
> +int shdma_request_irq(struct shdma_chan *schan, int irq,
> + unsigned long flags, const char *name)
> +{
> + int ret = request_threaded_irq(irq, chan_irq, chan_irqt,
> + flags, name, schan);
> +
> + schan->irq = ret < 0 ? ret : irq;
> +
> + return ret;
> +}
> +EXPORT_SYMBOL(shdma_request_irq);
> +
> +void shdma_free_irq(struct shdma_chan *schan)
> +{
> + if (schan->irq >= 0)
> + free_irq(schan->irq, schan);
> +}
> +EXPORT_SYMBOL(shdma_free_irq);
> +
> +void shdma_chan_probe(struct shdma_dev *sdev,
> + struct shdma_chan *schan, int id)
> +{
> + schan->pm_state = SHDMA_PM_ESTABLISHED;
> +
> + /* reference struct dma_device */
> + schan->dma_chan.device = &sdev->dma_dev;
> + dma_cookie_init(&schan->dma_chan);
> +
> + schan->dev = sdev->dma_dev.dev;
> + schan->id = id;
> +
> + if (!schan->max_xfer_len)
> + schan->max_xfer_len = PAGE_SIZE;
> +
> + spin_lock_init(&schan->chan_lock);
> +
> + /* Init descripter manage list */
> + INIT_LIST_HEAD(&schan->ld_queue);
> + INIT_LIST_HEAD(&schan->ld_free);
> +
> + /* Add the channel to DMA device channel list */
> + list_add_tail(&schan->dma_chan.device_node,
> + &sdev->dma_dev.channels);
> + sdev->schan[sdev->dma_dev.chancnt++] = schan;
> +}
> +EXPORT_SYMBOL(shdma_chan_probe);
> +
> +void shdma_chan_remove(struct shdma_chan *schan)
> +{
> + list_del(&schan->dma_chan.device_node);
> +}
> +EXPORT_SYMBOL(shdma_chan_remove);
> +
> +int shdma_init(struct device *dev, struct shdma_dev *sdev,
> + int chan_num)
> +{
> + struct dma_device *dma_dev = &sdev->dma_dev;
> +
> + /*
> + * Require all call-backs for now, they can trivially be made optional
> + * later as required
> + */
> + if (!sdev->ops ||
> + !sdev->desc_size ||
> + !sdev->ops->embedded_desc ||
> + !sdev->ops->start_xfer ||
> + !sdev->ops->setup_xfer ||
> + !sdev->ops->set_slave ||
> + !sdev->ops->desc_setup ||
> + !sdev->ops->slave_addr ||
> + !sdev->ops->channel_busy ||
> + !sdev->ops->halt_channel ||
> + !sdev->ops->desc_completed)
> + return -EINVAL;
> +
> + sdev->schan = kcalloc(chan_num, sizeof(*sdev->schan), GFP_KERNEL);
> + if (!sdev->schan)
> + return -ENOMEM;
> +
> + INIT_LIST_HEAD(&dma_dev->channels);
> +
> + /* Common and MEMCPY operations */
> + dma_dev->device_alloc_chan_resources
> + = shdma_alloc_chan_resources;
> + dma_dev->device_free_chan_resources = shdma_free_chan_resources;
> + dma_dev->device_prep_dma_memcpy = shdma_prep_memcpy;
> + dma_dev->device_tx_status = shdma_tx_status;
> + dma_dev->device_issue_pending = shdma_issue_pending;
> +
> + /* Compulsory for DMA_SLAVE fields */
> + dma_dev->device_prep_slave_sg = shdma_prep_slave_sg;
> + dma_dev->device_control = shdma_control;
> +
> + dma_dev->dev = dev;
> +
> + return 0;
> +}
> +EXPORT_SYMBOL(shdma_init);
> +
> +void shdma_cleanup(struct shdma_dev *sdev)
> +{
> + kfree(sdev->schan);
> +}
> +EXPORT_SYMBOL(shdma_cleanup);
> +
> +static int __init shdma_enter(void)
> +{
> + shdma_slave_used = kzalloc(DIV_ROUND_UP(slave_num, BITS_PER_LONG) *
> + sizeof(long), GFP_KERNEL);
> + if (!shdma_slave_used)
> + return -ENOMEM;
> + return 0;
> +}
> +module_init(shdma_enter);
> +
> +static void __exit shdma_exit(void)
> +{
> + kfree(shdma_slave_used);
> +}
> +module_exit(shdma_exit);
> +
> +MODULE_LICENSE("GPL v2");
> +MODULE_DESCRIPTION("SH-DMA driver base library");
> +MODULE_AUTHOR("Guennadi Liakhovetski <[email protected]>");
> diff --git a/include/linux/shdma-base.h b/include/linux/shdma-base.h
> new file mode 100644
> index 0000000..83efd13
> --- /dev/null
> +++ b/include/linux/shdma-base.h
> @@ -0,0 +1,123 @@
> +/*
> + * Dmaengine driver base library for DMA controllers, found on SH-based SoCs
> + *
> + * extracted from shdma.c and headers
> + *
> + * Copyright (C) 2011-2012 Guennadi Liakhovetski <[email protected]>
> + * Copyright (C) 2009 Nobuhiro Iwamatsu <[email protected]>
> + * Copyright (C) 2009 Renesas Solutions, Inc. All rights reserved.
> + * Copyright (C) 2007 Freescale Semiconductor, Inc. All rights reserved.
> + *
> + * This is free software; you can redistribute it and/or modify
> + * it under the terms of version 2 of the GNU General Public License as
> + * published by the Free Software Foundation.
> + */
> +
> +#ifndef SHDMA_BASE_H
> +#define SHDMA_BASE_H
> +
> +#include <linux/dmaengine.h>
> +#include <linux/interrupt.h>
> +#include <linux/list.h>
> +#include <linux/types.h>
> +
> +/**
> + * shdma_pm_state - DMA channel PM state
> + * SHDMA_PM_ESTABLISHED: either idle or during data transfer
> + * SHDMA_PM_BUSY: during the transfer preparation, when we have to
> + * drop the lock temporarily
> + * SHDMA_PM_PENDING: transfers pending
> + */
> +enum shdma_pm_state {
> + SHDMA_PM_ESTABLISHED,
> + SHDMA_PM_BUSY,
> + SHDMA_PM_PENDING,
> +};
> +
> +struct device;
> +
> +/*
> + * Drivers, using this library are expected to embed struct shdma_dev,
> + * struct shdma_chan, struct shdma_desc, and struct shdma_slave
> + * in their respective device, channel, descriptor and slave objects.
> + */
> +
> +struct shdma_slave {
> + unsigned int slave_id;
this should be moved to struct dma_slave_config
> +};
> +
> +struct shdma_desc {
> + struct list_head node;
> + struct dma_async_tx_descriptor async_tx;
> + enum dma_transfer_direction direction;
> + dma_cookie_t cookie;
> + int chunks;
> + int mark;
> +};
> +
> +struct shdma_chan {
> + spinlock_t chan_lock; /* Channel operation lock */
> + struct list_head ld_queue; /* Link descriptors queue */
> + struct list_head ld_free; /* Free link descriptors */
> + struct dma_chan dma_chan; /* DMA channel */
> + struct device *dev; /* Channel device */
> + void *desc; /* buffer for descriptor array */
> + int desc_num; /* desc count */
> + size_t max_xfer_len; /* max transfer length */
> + int id; /* Raw id of this channel */
> + int irq; /* Channel IRQ */
> + enum shdma_pm_state pm_state;
> +};
> +
> +/**
> + * struct shdma_ops - simple DMA driver operations
> + * desc_completed: return true, if this is the descriptor, that just has
> + * completed (atomic)
> + * halt_channel: stop DMA channel operation (atomic)
> + * channel_busy: return true, if the channel is busy (atomic)
> + * slave_addr: return slave DMA address
> + * desc_setup: set up the hardware specific descriptor portion (atomic)
> + * set_slave: bind channel to a slave
> + * setup_xfer: configure channel hardware for operation (atomic)
> + * start_xfer: start the DMA transfer (atomic)
> + * embedded_desc: return Nth struct shdma_desc pointer from the
> + * descriptor array
> + * chan_irq: process channel IRQ, return true if a transfer has
> + * completed (atomic)
> + */
> +struct shdma_ops {
> + bool (*desc_completed)(struct shdma_chan *, struct shdma_desc *);
> + void (*halt_channel)(struct shdma_chan *);
> + bool (*channel_busy)(struct shdma_chan *);
> + dma_addr_t (*slave_addr)(struct shdma_chan *);
> + int (*desc_setup)(struct shdma_chan *, struct shdma_desc *,
> + dma_addr_t, dma_addr_t, size_t *);
> + int (*set_slave)(struct shdma_chan *, struct shdma_slave *);
> + void (*setup_xfer)(struct shdma_chan *, struct shdma_slave *);
> + void (*start_xfer)(struct shdma_chan *, struct shdma_desc *);
> + struct shdma_desc *(*embedded_desc)(void *, int);
> + bool (*chan_irq)(struct shdma_chan *, int);
> +};
> +
> +struct shdma_dev {
> + struct dma_device dma_dev;
> + struct shdma_chan **schan;
> + const struct shdma_ops *ops;
> + size_t desc_size;
> +};
> +
> +#define shdma_for_each_chan(c, d, i) for (i = 0, c = (d)->schan[0]; \
> + i < (d)->dma_dev.chancnt; c = (d)->schan[++i])
> +
> +int shdma_request_irq(struct shdma_chan *, int,
> + unsigned long, const char *);
> +void shdma_free_irq(struct shdma_chan *);
> +bool shdma_reset(struct shdma_dev *sdev);
> +void shdma_chan_probe(struct shdma_dev *sdev,
> + struct shdma_chan *schan, int id);
> +void shdma_chan_remove(struct shdma_chan *schan);
> +int shdma_init(struct device *dev, struct shdma_dev *sdev,
> + int chan_num);
> +void shdma_cleanup(struct shdma_dev *sdev);
> +
> +#endif
--
~Vinod
On Wed, 2012-05-09 at 17:09 +0200, Guennadi Liakhovetski wrote:
> By placing an anonymous union at the top of struct sh_dmae_slave we can
> transparently prepare all device and client drivers for the upcoming
> shdma-base conversion.
>
> Signed-off-by: Guennadi Liakhovetski <[email protected]>
> ---
> include/linux/sh_dma.h | 8 ++++++--
> 1 files changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/include/linux/sh_dma.h b/include/linux/sh_dma.h
> index 425450b..e081e8e 100644
> --- a/include/linux/sh_dma.h
> +++ b/include/linux/sh_dma.h
> @@ -10,12 +10,16 @@
> #ifndef SH_DMA_H
> #define SH_DMA_H
>
> -#include <linux/list.h>
> #include <linux/dmaengine.h>
> +#include <linux/list.h>
> +#include <linux/shdma-base.h>
>
> /* Used by slave DMA clients to request DMA to/from a specific peripheral */
> struct sh_dmae_slave {
> - unsigned int slave_id; /* Set by the platform */
> + union {
> + unsigned int slave_id; /* Set by the platform */
> + struct shdma_slave shdma_slave;
again, this should be move to dma_slave_config
> + };
> struct device *dma_dev; /* Set by the platform */
> const struct sh_dmae_slave_config *config; /* Set by the driver */
> };
--
~Vinod
On Wed, 2012-05-09 at 17:09 +0200, Guennadi Liakhovetski wrote:
> Signed-off-by: Guennadi Liakhovetski <[email protected]>
> Cc: Alan Cox <[email protected]>
> ---
> drivers/tty/serial/sh-sci.c | 8 ++++----
> 1 files changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/tty/serial/sh-sci.c b/drivers/tty/serial/sh-sci.c
> index bf461cf..49e3b2a 100644
> --- a/drivers/tty/serial/sh-sci.c
> +++ b/drivers/tty/serial/sh-sci.c
> @@ -1579,9 +1579,9 @@ static bool filter(struct dma_chan *chan, void *slave)
> struct sh_dmae_slave *param = slave;
>
> dev_dbg(chan->device->dev, "%s: slave ID %d\n", __func__,
> - param->slave_id);
> + param->shdma_slave.slave_id);
>
> - chan->private = param;
> + chan->private = ¶m->shdma_slave;
passing stuff using chan->private is a dead road, so don't use it.
> return true;
> }
>
> @@ -1620,7 +1620,7 @@ static void sci_request_dma(struct uart_port *port)
> param = &s->param_tx;
>
> /* Slave ID, e.g., SHDMA_SLAVE_SCIF0_TX */
> - param->slave_id = s->cfg->dma_slave_tx;
> + param->shdma_slave.slave_id = s->cfg->dma_slave_tx;
>
> s->cookie_tx = -EINVAL;
> chan = dma_request_channel(mask, filter, param);
> @@ -1648,7 +1648,7 @@ static void sci_request_dma(struct uart_port *port)
> param = &s->param_rx;
>
> /* Slave ID, e.g., SHDMA_SLAVE_SCIF0_RX */
> - param->slave_id = s->cfg->dma_slave_rx;
> + param->shdma_slave.slave_id = s->cfg->dma_slave_rx;
>
> chan = dma_request_channel(mask, filter, param);
> dev_dbg(port->dev, "%s: RX: got channel %p\n", __func__, chan);
On Tue, 15 May 2012, Vinod Koul wrote:
> On Wed, 2012-05-09 at 17:09 +0200, Guennadi Liakhovetski wrote:
> > Signed-off-by: Guennadi Liakhovetski <[email protected]>
> > Cc: Alan Cox <[email protected]>
> > ---
> > drivers/tty/serial/sh-sci.c | 8 ++++----
> > 1 files changed, 4 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/tty/serial/sh-sci.c b/drivers/tty/serial/sh-sci.c
> > index bf461cf..49e3b2a 100644
> > --- a/drivers/tty/serial/sh-sci.c
> > +++ b/drivers/tty/serial/sh-sci.c
> > @@ -1579,9 +1579,9 @@ static bool filter(struct dma_chan *chan, void *slave)
> > struct sh_dmae_slave *param = slave;
> >
> > dev_dbg(chan->device->dev, "%s: slave ID %d\n", __func__,
> > - param->slave_id);
> > + param->shdma_slave.slave_id);
> >
> > - chan->private = param;
> > + chan->private = ¶m->shdma_slave;
> passing stuff using chan->private is a dead road, so don't use it.
As you see, it _is_ already done now and has been that way since some
time. We do want to change this, but not with this patch series.
> > return true;
> > }
> >
> > @@ -1620,7 +1620,7 @@ static void sci_request_dma(struct uart_port *port)
> > param = &s->param_tx;
> >
> > /* Slave ID, e.g., SHDMA_SLAVE_SCIF0_TX */
> > - param->slave_id = s->cfg->dma_slave_tx;
> > + param->shdma_slave.slave_id = s->cfg->dma_slave_tx;
> >
> > s->cookie_tx = -EINVAL;
> > chan = dma_request_channel(mask, filter, param);
> > @@ -1648,7 +1648,7 @@ static void sci_request_dma(struct uart_port *port)
> > param = &s->param_rx;
> >
> > /* Slave ID, e.g., SHDMA_SLAVE_SCIF0_RX */
> > - param->slave_id = s->cfg->dma_slave_rx;
> > + param->shdma_slave.slave_id = s->cfg->dma_slave_rx;
> >
> > chan = dma_request_channel(mask, filter, param);
> > dev_dbg(port->dev, "%s: RX: got channel %p\n", __func__, chan);
Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/
Hi Vinod
Thanks for your review.
On Tue, 15 May 2012, Vinod Koul wrote:
> On Wed, 2012-05-09 at 17:09 +0200, Guennadi Liakhovetski wrote:
> > This patch extracts code from shdma.c, that does not directly deal with
> > hardware implementation details and can be re-used with diverse DMA
> > controller variants, found on SH-based SoCs.
> Have you see recent Russell's virtual dma patches [1]? I like the idea
> and simplifies things for implementation in each driver.
> Said that, do we need this libarry. Can you see if you can use/enhance
> that one.
No, I haven't seen it. I had a look at it, and it doesn't seem to provide
an immediate solution to our problem. It can be used as a part of one, but
a significant amount of work would still be required to convert shdma to a
virtual-channel based design.
Let me try to describe, what such a virtual channel solution would look
like:
1. One (virtual) DMA channel is allocated per client
2. As clients issue dma_request_channel(), the correct channel is selected
3. The client configures the channel with a call to
dmaengine_slave_config()
4. As transfers are prepared and issued they are queued in per-virtual
channel lists
5. Each virtual channel links to a list of suitable DMA devices and
channels
6. As a new transfer request on a specific virtual channel has to be
processed, a suitable free physical channel is picked up, configured
and the request is sent out
If this my understanding is right, this should be doable, but not within
the current 3.5 time-frame. Further, if we first do this conversion, we'll
still have to split the driver similar to what is proposed by this patch
series, but then it would be largely a new effort, most of the work
invested into this so far will be lost. OTOH, if we now push the shdma
driver split, we enable the merge of the SUDMAC support and porting to the
virtual channel design should also become easier, because the hardware and
the software parts of the driver will already be separated and presumably
the hardware part will not need to be changed.
One more thing, before going into detail and replying to your comments:
this split doesn't change driver's functionality, it simply splits it into
two parts. So, most questions, that you ask in this your review would also
apply to the current driver. This patch just extracts hardware-neutral
code from the shdma driver into a separate file. I think, any functional
changes should be done separately from this driver-split series.
[snip]
> > +static dma_cookie_t shdma_tx_submit(struct dma_async_tx_descriptor *tx)
> > +{
> > + struct shdma_desc *chunk, *c, *desc =
> > + container_of(tx, struct shdma_desc, async_tx),
> > + *last = desc;
> > + struct shdma_chan *schan = to_shdma_chan(tx->chan);
> > + struct shdma_slave *slave = tx->chan->private;
> > + dma_async_tx_callback callback = tx->callback;
> > + dma_cookie_t cookie;
> > + bool power_up;
> > +
> > + spin_lock_irq(&schan->chan_lock);
> > +
> > + power_up = list_empty(&schan->ld_queue);
> > +
> > + cookie = dma_cookie_assign(tx);
> > +
> > + /* Mark all chunks of this descriptor as submitted, move to the queue */
> > + list_for_each_entry_safe(chunk, c, desc->node.prev, node) {
> > + /*
> > + * All chunks are on the global ld_free, so, we have to find
> > + * the end of the chain ourselves
> > + */
> > + if (chunk != desc && (chunk->mark == DESC_IDLE ||
> > + chunk->async_tx.cookie > 0 ||
> > + chunk->async_tx.cookie == -EBUSY ||
> > + &chunk->node == &schan->ld_free))
> > + break;
> > + chunk->mark = DESC_SUBMITTED;
> > + /* Callback goes to the last chunk */
> > + chunk->async_tx.callback = NULL;
> > + chunk->cookie = cookie;
> > + list_move_tail(&chunk->node, &schan->ld_queue);
> > + last = chunk;
> > +
> > + dev_dbg(schan->dev, "submit #%d@%p on %d\n",
> > + tx->cookie, &last->async_tx, schan->id);
> > + }
> > +
> > + last->async_tx.callback = callback;
> > + last->async_tx.callback_param = tx->callback_param;
> > +
> > + if (power_up) {
> > + int ret;
> > + schan->pm_state = SHDMA_PM_BUSY;
> > +
> > + ret = pm_runtime_get(schan->dev);
> any reason why this is here and not in issue pending?
Sorry, we have discussed this multiple times already. See, for example,
http://thread.gmane.org/gmane.linux.kernel/1181989/focus=12187
[snip]
> > +/*
> > + * Drivers, using this library are expected to embed struct shdma_dev,
> > + * struct shdma_chan, struct shdma_desc, and struct shdma_slave
> > + * in their respective device, channel, descriptor and slave objects.
> > + */
> > +
> > +struct shdma_slave {
> > + unsigned int slave_id;
> this should be moved to struct dma_slave_config
Again, this would be a functional change. And struct dma_slave_config
doesn't have a slave-ID field in it, which is the only thing we need for
now. Instead it has a bunch of other fields, of which none is so far used
by this driver.
Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/
On Tue, 15 May 2012, Vinod Koul wrote:
> On Wed, 2012-05-09 at 17:09 +0200, Guennadi Liakhovetski wrote:
> > By placing an anonymous union at the top of struct sh_dmae_slave we can
> > transparently prepare all device and client drivers for the upcoming
> > shdma-base conversion.
> >
> > Signed-off-by: Guennadi Liakhovetski <[email protected]>
> > ---
> > include/linux/sh_dma.h | 8 ++++++--
> > 1 files changed, 6 insertions(+), 2 deletions(-)
> >
> > diff --git a/include/linux/sh_dma.h b/include/linux/sh_dma.h
> > index 425450b..e081e8e 100644
> > --- a/include/linux/sh_dma.h
> > +++ b/include/linux/sh_dma.h
> > @@ -10,12 +10,16 @@
> > #ifndef SH_DMA_H
> > #define SH_DMA_H
> >
> > -#include <linux/list.h>
> > #include <linux/dmaengine.h>
> > +#include <linux/list.h>
> > +#include <linux/shdma-base.h>
> >
> > /* Used by slave DMA clients to request DMA to/from a specific peripheral */
> > struct sh_dmae_slave {
> > - unsigned int slave_id; /* Set by the platform */
> > + union {
> > + unsigned int slave_id; /* Set by the platform */
> > + struct shdma_slave shdma_slave;
> again, this should be move to dma_slave_config
Same reasoning as in patch 1: I think such changes should be done
separately.
> > + };
> > struct device *dma_dev; /* Set by the platform */
> > const struct sh_dmae_slave_config *config; /* Set by the driver */
> > };
Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/
On Fri, 2012-05-18 at 10:21 +0200, Guennadi Liakhovetski wrote:
> Hi Vinod
>
> Thanks for your review.
sorry for the late reply, was away for few days
>
> On Tue, 15 May 2012, Vinod Koul wrote:
>
> > On Wed, 2012-05-09 at 17:09 +0200, Guennadi Liakhovetski wrote:
> > > This patch extracts code from shdma.c, that does not directly deal with
> > > hardware implementation details and can be re-used with diverse DMA
> > > controller variants, found on SH-based SoCs.
> > Have you see recent Russell's virtual dma patches [1]? I like the idea
> > and simplifies things for implementation in each driver.
> > Said that, do we need this libarry. Can you see if you can use/enhance
> > that one.
>
> No, I haven't seen it. I had a look at it, and it doesn't seem to provide
> an immediate solution to our problem. It can be used as a part of one, but
> a significant amount of work would still be required to convert shdma to a
> virtual-channel based design.
>
> Let me try to describe, what such a virtual channel solution would look
> like:
>
> 1. One (virtual) DMA channel is allocated per client
> 2. As clients issue dma_request_channel(), the correct channel is selected
> 3. The client configures the channel with a call to
> dmaengine_slave_config()
> 4. As transfers are prepared and issued they are queued in per-virtual
> channel lists
> 5. Each virtual channel links to a list of suitable DMA devices and
> channels
> 6. As a new transfer request on a specific virtual channel has to be
> processed, a suitable free physical channel is picked up, configured
> and the request is sent out
>
> If this my understanding is right, this should be doable, but not within
> the current 3.5 time-frame. Further, if we first do this conversion, we'll
> still have to split the driver similar to what is proposed by this patch
> series, but then it would be largely a new effort, most of the work
> invested into this so far will be lost. OTOH, if we now push the shdma
> driver split, we enable the merge of the SUDMAC support and porting to the
> virtual channel design should also become easier, because the hardware and
> the software parts of the driver will already be separated and presumably
> the hardware part will not need to be changed.
>
> One more thing, before going into detail and replying to your comments:
> this split doesn't change driver's functionality, it simply splits it into
> two parts. So, most questions, that you ask in this your review would also
> apply to the current driver. This patch just extracts hardware-neutral
> code from the shdma driver into a separate file. I think, any functional
> changes should be done separately from this driver-split series.
Yes most of my comments do apply to current driver.
IMO update to a driver should fix existing problems, that why I pointed
them out.
Let me know if you plan to fixing them or I will queue these up for 3.6
>
> [snip]
>
> > > +static dma_cookie_t shdma_tx_submit(struct dma_async_tx_descriptor *tx)
> > > +{
> > > + struct shdma_desc *chunk, *c, *desc =
> > > + container_of(tx, struct shdma_desc, async_tx),
> > > + *last = desc;
> > > + struct shdma_chan *schan = to_shdma_chan(tx->chan);
> > > + struct shdma_slave *slave = tx->chan->private;
> > > + dma_async_tx_callback callback = tx->callback;
> > > + dma_cookie_t cookie;
> > > + bool power_up;
> > > +
> > > + spin_lock_irq(&schan->chan_lock);
> > > +
> > > + power_up = list_empty(&schan->ld_queue);
> > > +
> > > + cookie = dma_cookie_assign(tx);
> > > +
> > > + /* Mark all chunks of this descriptor as submitted, move to the queue */
> > > + list_for_each_entry_safe(chunk, c, desc->node.prev, node) {
> > > + /*
> > > + * All chunks are on the global ld_free, so, we have to find
> > > + * the end of the chain ourselves
> > > + */
> > > + if (chunk != desc && (chunk->mark == DESC_IDLE ||
> > > + chunk->async_tx.cookie > 0 ||
> > > + chunk->async_tx.cookie == -EBUSY ||
> > > + &chunk->node == &schan->ld_free))
> > > + break;
> > > + chunk->mark = DESC_SUBMITTED;
> > > + /* Callback goes to the last chunk */
> > > + chunk->async_tx.callback = NULL;
> > > + chunk->cookie = cookie;
> > > + list_move_tail(&chunk->node, &schan->ld_queue);
> > > + last = chunk;
> > > +
> > > + dev_dbg(schan->dev, "submit #%d@%p on %d\n",
> > > + tx->cookie, &last->async_tx, schan->id);
> > > + }
> > > +
> > > + last->async_tx.callback = callback;
> > > + last->async_tx.callback_param = tx->callback_param;
> > > +
> > > + if (power_up) {
> > > + int ret;
> > > + schan->pm_state = SHDMA_PM_BUSY;
> > > +
> > > + ret = pm_runtime_get(schan->dev);
> > any reason why this is here and not in issue pending?
>
> Sorry, we have discussed this multiple times already. See, for example,
>
> http://thread.gmane.org/gmane.linux.kernel/1181989/focus=12187
>
> [snip]
>
> > > +/*
> > > + * Drivers, using this library are expected to embed struct shdma_dev,
> > > + * struct shdma_chan, struct shdma_desc, and struct shdma_slave
> > > + * in their respective device, channel, descriptor and slave objects.
> > > + */
> > > +
> > > +struct shdma_slave {
> > > + unsigned int slave_id;
> > this should be moved to struct dma_slave_config
>
> Again, this would be a functional change. And struct dma_slave_config
> doesn't have a slave-ID field in it, which is the only thing we need for
> now. Instead it has a bunch of other fields, of which none is so far used
> by this driver.
>
> Thanks
> Guennadi
> ---
> Guennadi Liakhovetski, Ph.D.
> Freelance Open-Source Software Developer
> http://www.open-technology.de/
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
--
~Vinod
Hi Vinod
On Fri, 25 May 2012, Vinod Koul wrote:
> On Fri, 2012-05-18 at 10:21 +0200, Guennadi Liakhovetski wrote:
> > Hi Vinod
> >
> > Thanks for your review.
> sorry for the late reply, was away for few days
> >
> > On Tue, 15 May 2012, Vinod Koul wrote:
> >
> > > On Wed, 2012-05-09 at 17:09 +0200, Guennadi Liakhovetski wrote:
> > > > This patch extracts code from shdma.c, that does not directly deal with
> > > > hardware implementation details and can be re-used with diverse DMA
> > > > controller variants, found on SH-based SoCs.
> > > Have you see recent Russell's virtual dma patches [1]? I like the idea
> > > and simplifies things for implementation in each driver.
> > > Said that, do we need this libarry. Can you see if you can use/enhance
> > > that one.
> >
> > No, I haven't seen it. I had a look at it, and it doesn't seem to provide
> > an immediate solution to our problem. It can be used as a part of one, but
> > a significant amount of work would still be required to convert shdma to a
> > virtual-channel based design.
> >
> > Let me try to describe, what such a virtual channel solution would look
> > like:
> >
> > 1. One (virtual) DMA channel is allocated per client
> > 2. As clients issue dma_request_channel(), the correct channel is selected
> > 3. The client configures the channel with a call to
> > dmaengine_slave_config()
> > 4. As transfers are prepared and issued they are queued in per-virtual
> > channel lists
> > 5. Each virtual channel links to a list of suitable DMA devices and
> > channels
> > 6. As a new transfer request on a specific virtual channel has to be
> > processed, a suitable free physical channel is picked up, configured
> > and the request is sent out
> >
> > If this my understanding is right, this should be doable, but not within
> > the current 3.5 time-frame. Further, if we first do this conversion, we'll
> > still have to split the driver similar to what is proposed by this patch
> > series, but then it would be largely a new effort, most of the work
> > invested into this so far will be lost. OTOH, if we now push the shdma
> > driver split, we enable the merge of the SUDMAC support and porting to the
> > virtual channel design should also become easier, because the hardware and
> > the software parts of the driver will already be separated and presumably
> > the hardware part will not need to be changed.
> >
> > One more thing, before going into detail and replying to your comments:
> > this split doesn't change driver's functionality, it simply splits it into
> > two parts. So, most questions, that you ask in this your review would also
> > apply to the current driver. This patch just extracts hardware-neutral
> > code from the shdma driver into a separate file. I think, any functional
> > changes should be done separately from this driver-split series.
> Yes most of my comments do apply to current driver.
> IMO update to a driver should fix existing problems, that why I pointed
> them out.
>
> Let me know if you plan to fixing them or I will queue these up for 3.6
Please, queue these patches for the next merge window, we will work on
improving the driver after this split.
Thanks
Guennadi
> > [snip]
> >
> > > > +static dma_cookie_t shdma_tx_submit(struct dma_async_tx_descriptor *tx)
> > > > +{
> > > > + struct shdma_desc *chunk, *c, *desc =
> > > > + container_of(tx, struct shdma_desc, async_tx),
> > > > + *last = desc;
> > > > + struct shdma_chan *schan = to_shdma_chan(tx->chan);
> > > > + struct shdma_slave *slave = tx->chan->private;
> > > > + dma_async_tx_callback callback = tx->callback;
> > > > + dma_cookie_t cookie;
> > > > + bool power_up;
> > > > +
> > > > + spin_lock_irq(&schan->chan_lock);
> > > > +
> > > > + power_up = list_empty(&schan->ld_queue);
> > > > +
> > > > + cookie = dma_cookie_assign(tx);
> > > > +
> > > > + /* Mark all chunks of this descriptor as submitted, move to the queue */
> > > > + list_for_each_entry_safe(chunk, c, desc->node.prev, node) {
> > > > + /*
> > > > + * All chunks are on the global ld_free, so, we have to find
> > > > + * the end of the chain ourselves
> > > > + */
> > > > + if (chunk != desc && (chunk->mark == DESC_IDLE ||
> > > > + chunk->async_tx.cookie > 0 ||
> > > > + chunk->async_tx.cookie == -EBUSY ||
> > > > + &chunk->node == &schan->ld_free))
> > > > + break;
> > > > + chunk->mark = DESC_SUBMITTED;
> > > > + /* Callback goes to the last chunk */
> > > > + chunk->async_tx.callback = NULL;
> > > > + chunk->cookie = cookie;
> > > > + list_move_tail(&chunk->node, &schan->ld_queue);
> > > > + last = chunk;
> > > > +
> > > > + dev_dbg(schan->dev, "submit #%d@%p on %d\n",
> > > > + tx->cookie, &last->async_tx, schan->id);
> > > > + }
> > > > +
> > > > + last->async_tx.callback = callback;
> > > > + last->async_tx.callback_param = tx->callback_param;
> > > > +
> > > > + if (power_up) {
> > > > + int ret;
> > > > + schan->pm_state = SHDMA_PM_BUSY;
> > > > +
> > > > + ret = pm_runtime_get(schan->dev);
> > > any reason why this is here and not in issue pending?
> >
> > Sorry, we have discussed this multiple times already. See, for example,
> >
> > http://thread.gmane.org/gmane.linux.kernel/1181989/focus=12187
> >
> > [snip]
> >
> > > > +/*
> > > > + * Drivers, using this library are expected to embed struct shdma_dev,
> > > > + * struct shdma_chan, struct shdma_desc, and struct shdma_slave
> > > > + * in their respective device, channel, descriptor and slave objects.
> > > > + */
> > > > +
> > > > +struct shdma_slave {
> > > > + unsigned int slave_id;
> > > this should be moved to struct dma_slave_config
> >
> > Again, this would be a functional change. And struct dma_slave_config
> > doesn't have a slave-ID field in it, which is the only thing we need for
> > now. Instead it has a bunch of other fields, of which none is so far used
> > by this driver.
> >
> > Thanks
> > Guennadi
> > ---
> > Guennadi Liakhovetski, Ph.D.
> > Freelance Open-Source Software Developer
> > http://www.open-technology.de/
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> > the body of a message to [email protected]
> > More majordomo info at http://vger.kernel.org/majordomo-info.html
> > Please read the FAQ at http://www.tux.org/lkml/
>
>
> --
> ~Vinod
>
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/
On Friday 25 May 2012 01:45 PM, Vinod Koul wrote:
> On Fri, 2012-05-18 at 10:21 +0200, Guennadi Liakhovetski wrote:
> +/*
>>>> + * Drivers, using this library are expected to embed struct shdma_dev,
>>>> + * struct shdma_chan, struct shdma_desc, and struct shdma_slave
>>>> + * in their respective device, channel, descriptor and slave objects.
>>>> + */
>>>> +
>>>> +struct shdma_slave {
>>>> + unsigned int slave_id;
>>> this should be moved to struct dma_slave_config
>> Again, this would be a functional change. And struct dma_slave_config
>> doesn't have a slave-ID field in it, which is the only thing we need for
>> now. Instead it has a bunch of other fields, of which none is so far used
>> by this driver.
>>
In Nvidia's tegra SOC also require to pass the slave id from client to
dma driver. I have pushed the change to make it part of struct
dma_slave_config on the patch
[PATCH V4 1/2] dma: dmaengine: add slave req id in slave_config
So hope you can use the same if that patch get merged.
Hi Laxman
On Fri, 25 May 2012, Laxman Dewangan wrote:
> On Friday 25 May 2012 01:45 PM, Vinod Koul wrote:
> > On Fri, 2012-05-18 at 10:21 +0200, Guennadi Liakhovetski wrote:
>
> > +/*
> > > > > + * Drivers, using this library are expected to embed struct
> > > > > shdma_dev,
> > > > > + * struct shdma_chan, struct shdma_desc, and struct shdma_slave
> > > > > + * in their respective device, channel, descriptor and slave objects.
> > > > > + */
> > > > > +
> > > > > +struct shdma_slave {
> > > > > + unsigned int slave_id;
> > > > this should be moved to struct dma_slave_config
> > > Again, this would be a functional change. And struct dma_slave_config
> > > doesn't have a slave-ID field in it, which is the only thing we need for
> > > now. Instead it has a bunch of other fields, of which none is so far used
> > > by this driver.
> > >
>
> In Nvidia's tegra SOC also require to pass the slave id from client to dma
> driver. I have pushed the change to make it part of struct dma_slave_config
> on the patch
> [PATCH V4 1/2] dma: dmaengine: add slave req id in slave_config
>
> So hope you can use the same if that patch get merged.
Thanks for letting me know. After both our series get merged, when working
on removal of .private use, switching to your new .slave_id will really be
the way to go for us.
Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/
Hi Vinod
On Fri, 25 May 2012, Guennadi Liakhovetski wrote:
> Hi Vinod
>
> On Fri, 25 May 2012, Vinod Koul wrote:
[snip]
> > Let me know if you plan to fixing them or I will queue these up for 3.6
>
> Please, queue these patches for the next merge window, we will work on
> improving the driver after this split.
It's been a month since this mail - do I see it right, that these patches
are still not in your slave-dma tree? Having them there would (1) get them
some exposure and testing and (2) eventually speed up building on top of
them and fixing existing driver's shortcomings. Any approximate idea when
you'd be able to pull them in?
Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/