[PATCH 1/2] Fix a bug about BUG_ON() on DMA engine capability DMA_INTERRUPT.

[PATCH 1/2] Fix a bug about BUG_ON() on DMA engine capability DMA_INTERRUPT.

[Zhang Wei]

The device->device_prep_dma_interrupt function is used by
DMA_INTERRUPT capability, not DMA_ZERO_SUM.

Signed-off-by: Zhang Wei <wei.zhang@freescale.com>
---
 drivers/dma/dmaengine.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
index 2996523..8db0e7f 100644
--- a/drivers/dma/dmaengine.c
+++ b/drivers/dma/dmaengine.c
@@ -357,7 +357,7 @@ int dma_async_device_register(struct dma_device *device)
 		!device->device_prep_dma_zero_sum);
 	BUG_ON(dma_has_cap(DMA_MEMSET, device->cap_mask) &&
 		!device->device_prep_dma_memset);
-	BUG_ON(dma_has_cap(DMA_ZERO_SUM, device->cap_mask) &&
+	BUG_ON(dma_has_cap(DMA_INTERRUPT, device->cap_mask) &&
 		!device->device_prep_dma_interrupt);
 
 	BUG_ON(!device->device_alloc_chan_resources);
-- 
1.5.4

[PATCH 2/2] Add device_prep_dma_interrupt support to fsldma.c

[Zhang Wei]

This is a bug that I assigned DMA_INTERRUPT capability to fsldma
but missing device_prep_dma_interrupt function. For a bug in
dmaengine.c the driver passed BUG_ON() checking. The patch fixes it.

Signed-off-by: Zhang Wei <wei.zhang@freescale.com>
---
 drivers/dma/fsldma.c |   27 +++++++++++++++++++++++++++
 1 files changed, 27 insertions(+), 0 deletions(-)

diff --git a/drivers/dma/fsldma.c b/drivers/dma/fsldma.c
index 5dfedf3..5428f81 100644
--- a/drivers/dma/fsldma.c
+++ b/drivers/dma/fsldma.c
@@ -406,6 +406,32 @@ static void fsl_dma_free_chan_resources(struct dma_chan *chan)
 	dma_pool_destroy(fsl_chan->desc_pool);
 }
 
+static struct dma_async_tx_descriptor *fsl_dma_prep_interrupt(
+						struct dma_chan *chan)
+{
+	struct fsl_dma_chan *fsl_chan;
+	struct fsl_desc_sw *new;
+
+	if (!chan)
+		return NULL;
+
+	fsl_chan = to_fsl_chan(chan);
+
+	new = fsl_dma_alloc_descriptor(fsl_chan);
+	if (!new) {
+		dev_err(fsl_chan->dev, "No free memory for link descriptor\n");
+		return NULL;
+	}
+
+	new->async_tx.cookie = -EBUSY;
+	new->async_tx.ack = 0;
+
+	/* Set End-of-link to the last link descriptor of new list*/
+	set_ld_eol(fsl_chan, new);
+
+	return &new->async_tx;
+}
+
 static struct dma_async_tx_descriptor *fsl_dma_prep_memcpy(
 	struct dma_chan *chan, dma_addr_t dma_dest, dma_addr_t dma_src,
 	size_t len, unsigned long flags)
@@ -1020,6 +1046,7 @@ static int __devinit of_fsl_dma_probe(struct of_device *dev,
 	dma_cap_set(DMA_INTERRUPT, fdev->common.cap_mask);
 	fdev->common.device_alloc_chan_resources = fsl_dma_alloc_chan_resources;
 	fdev->common.device_free_chan_resources = fsl_dma_free_chan_resources;
+	fdev->common.device_prep_dma_interrupt = fsl_dma_prep_interrupt;
 	fdev->common.device_prep_dma_memcpy = fsl_dma_prep_memcpy;
 	fdev->common.device_is_tx_complete = fsl_dma_is_complete;
 	fdev->common.device_issue_pending = fsl_dma_memcpy_issue_pending;
-- 
1.5.4

Re: [PATCH 2/2] Add device_prep_dma_interrupt support to fsldma.c

[Dan Williams]

On Mon, Mar 10, 2008 at 8:25 PM, Zhang Wei <wei.zhang@freescale.com> wrote:
> This is a bug that I assigned DMA_INTERRUPT capability to fsldma
>  but missing device_prep_dma_interrupt function. For a bug in
>  dmaengine.c the driver passed BUG_ON() checking. The patch fixes it.
>
>  Signed-off-by: Zhang Wei <wei.zhang@freescale.com>
>  ---
>   drivers/dma/fsldma.c |   27 +++++++++++++++++++++++++++
>   1 files changed, 27 insertions(+), 0 deletions(-)
>
>  diff --git a/drivers/dma/fsldma.c b/drivers/dma/fsldma.c
>  index 5dfedf3..5428f81 100644
>  --- a/drivers/dma/fsldma.c
>  +++ b/drivers/dma/fsldma.c
>  @@ -406,6 +406,32 @@ static void fsl_dma_free_chan_resources(struct dma_chan *chan)
>         dma_pool_destroy(fsl_chan->desc_pool);
>   }
>
>  +static struct dma_async_tx_descriptor *fsl_dma_prep_interrupt(
>  +                                               struct dma_chan *chan)
>  +{
>  +       struct fsl_dma_chan *fsl_chan;
>  +       struct fsl_desc_sw *new;
>  +
>  +       if (!chan)
>  +               return NULL;
>  +
>  +       fsl_chan = to_fsl_chan(chan);
>  +
>  +       new = fsl_dma_alloc_descriptor(fsl_chan);
>  +       if (!new) {
>  +               dev_err(fsl_chan->dev, "No free memory for link descriptor\n");
>  +               return NULL;
>  +       }
>  +
>  +       new->async_tx.cookie = -EBUSY;
>  +       new->async_tx.ack = 0;
>  +
>  +       /* Set End-of-link to the last link descriptor of new list*/
>  +       set_ld_eol(fsl_chan, new);

Question, is 'set_ld_eol' safe to call on descriptors that may not be
the last in the list?  For example what about the following sequence:

/* prepare two descriptors */
tx1 = fsl_dma_prep_interrupt(chan);
tx2 = fsl_dma_prep_memcpy(chan);

/* submit out of order */
tx2->tx_submit(tx2);
tx1->tx_submit(tx1);

This is only a concern if you plan to support channel switching at
some point.  For example, switching from a memcpy channel to an xor
channel.

Thanks,
Dan

RE: [PATCH 2/2] Add device_prep_dma_interrupt support to fsldma.c

[Zhang Wei]

Hi, Dan, 

> -----Original Message-----
> From: dan.j.williams@gmail.com 
> >  +
> >  +       /* Set End-of-link to the last link descriptor of 
> new list*/
> >  +       set_ld_eol(fsl_chan, new);
> 
> Question, is 'set_ld_eol' safe to call on descriptors that may not be
> the last in the list?  For example what about the following sequence:

set_ld_eol() is a safe function, which is only for preparing the tx descriptors
lists. When adding prepared tx descriptior list to DMA channel tx list,
the function append_ld_queue() should be called.

> 
> /* prepare two descriptors */
> tx1 = fsl_dma_prep_interrupt(chan);
> tx2 = fsl_dma_prep_memcpy(chan);
> 
> /* submit out of order */
> tx2->tx_submit(tx2);
> tx1->tx_submit(tx1);
> 
> This is only a concern if you plan to support channel switching at
> some point.  For example, switching from a memcpy channel to an xor
> channel.
> 
It's no problem. :) In fact, I've added out of order testing codes in fsl_dma_self_test()
 function.

But I have a question about device_prep_dma_interrupt(), which is no way to assign
 dest and src address. Is it a null tx action dma_async_tx_descriptor except to trigger
 an interrupt?

Thanks!
Wei.

Re: [PATCH 2/2] Add device_prep_dma_interrupt support to fsldma.c

[Dan Williams]

On Tue, Mar 11, 2008 at 7:28 PM, Zhang Wei <Wei.Zhang@freescale.com> wrote:
[..]
>  But I have a question about device_prep_dma_interrupt(), which is no way to assign
>   dest and src address. Is it a null tx action dma_async_tx_descriptor except to trigger
>   an interrupt?
>

Yes, it is for sequences where we are scheduling a chain of operations
of indeterminate length and want a callback after they all complete.
See ops_run_biofill() in drivers/md/raid5.c for an example.

>  Thanks!
>  Wei.

--
Dan