2019-09-05 20:05:33

by Radhey Shyam Pandey

[permalink] [raw]
Subject: [PATCH -next 3/8] dmaengine: xilinx_dma: Introduce xilinx_dma_get_residue

From: Nicholas Graumann <[email protected]>

Introduce a function that can calculate residues for IPs that support it:
AXI DMA and CDMA.

Signed-off-by: Nicholas Graumann <[email protected]>
Signed-off-by: Radhey Shyam Pandey <[email protected]>
---
drivers/dma/xilinx/xilinx_dma.c | 75 ++++++++++++++++++++++++++++++-----------
1 file changed, 56 insertions(+), 19 deletions(-)

diff --git a/drivers/dma/xilinx/xilinx_dma.c b/drivers/dma/xilinx/xilinx_dma.c
index 9909bfb..4094adb 100644
--- a/drivers/dma/xilinx/xilinx_dma.c
+++ b/drivers/dma/xilinx/xilinx_dma.c
@@ -787,6 +787,51 @@ static void xilinx_dma_free_chan_resources(struct dma_chan *dchan)
}

/**
+ * xilinx_dma_get_residue - Compute residue for a given descriptor
+ * @chan: Driver specific dma channel
+ * @desc: dma transaction descriptor
+ *
+ * Return: The number of residue bytes for the descriptor.
+ */
+static u32 xilinx_dma_get_residue(struct xilinx_dma_chan *chan,
+ struct xilinx_dma_tx_descriptor *desc)
+{
+ struct xilinx_cdma_tx_segment *cdma_seg;
+ struct xilinx_axidma_tx_segment *axidma_seg;
+ struct xilinx_cdma_desc_hw *cdma_hw;
+ struct xilinx_axidma_desc_hw *axidma_hw;
+ struct list_head *entry;
+ u32 residue = 0;
+
+ /**
+ * VDMA and simple mode do not support residue reporting, so the
+ * residue field will always be 0.
+ */
+ if (chan->xdev->dma_config->dmatype == XDMA_TYPE_VDMA || !chan->has_sg)
+ return residue;
+
+ list_for_each(entry, &desc->segments) {
+ if (chan->xdev->dma_config->dmatype == XDMA_TYPE_CDMA) {
+ cdma_seg = list_entry(entry,
+ struct xilinx_cdma_tx_segment,
+ node);
+ cdma_hw = &cdma_seg->hw;
+ residue += (cdma_hw->control - cdma_hw->status) &
+ chan->xdev->max_buffer_len;
+ } else {
+ axidma_seg = list_entry(entry,
+ struct xilinx_axidma_tx_segment,
+ node);
+ axidma_hw = &axidma_seg->hw;
+ residue += (axidma_hw->control - axidma_hw->status) &
+ chan->xdev->max_buffer_len;
+ }
+ }
+
+ return residue;
+}
+
+/**
* xilinx_dma_chan_handle_cyclic - Cyclic dma callback
* @chan: Driver specific dma channel
* @desc: dma transaction descriptor
@@ -995,33 +1040,22 @@ static enum dma_status xilinx_dma_tx_status(struct dma_chan *dchan,
{
struct xilinx_dma_chan *chan = to_xilinx_chan(dchan);
struct xilinx_dma_tx_descriptor *desc;
- struct xilinx_axidma_tx_segment *segment;
- struct xilinx_axidma_desc_hw *hw;
enum dma_status ret;
unsigned long flags;
- u32 residue = 0;

ret = dma_cookie_status(dchan, cookie, txstate);
if (ret == DMA_COMPLETE || !txstate)
return ret;

- if (chan->xdev->dma_config->dmatype == XDMA_TYPE_AXIDMA) {
- spin_lock_irqsave(&chan->lock, flags);
+ spin_lock_irqsave(&chan->lock, flags);

- desc = list_last_entry(&chan->active_list,
- struct xilinx_dma_tx_descriptor, node);
- if (chan->has_sg) {
- list_for_each_entry(segment, &desc->segments, node) {
- hw = &segment->hw;
- residue += (hw->control - hw->status) &
- chan->xdev->max_buffer_len;
- }
- }
- spin_unlock_irqrestore(&chan->lock, flags);
+ desc = list_last_entry(&chan->active_list,
+ struct xilinx_dma_tx_descriptor, node);
+ chan->residue = xilinx_dma_get_residue(chan, desc);

- chan->residue = residue;
- dma_set_residue(txstate, chan->residue);
- }
+ spin_unlock_irqrestore(&chan->lock, flags);
+
+ dma_set_residue(txstate, chan->residue);

return ret;
}
@@ -2701,12 +2735,15 @@ static int xilinx_dma_probe(struct platform_device *pdev)
xilinx_dma_prep_dma_cyclic;
xdev->common.device_prep_interleaved_dma =
xilinx_dma_prep_interleaved;
- /* Residue calculation is supported by only AXI DMA */
+ /* Residue calculation is supported by only AXI DMA and CDMA */
xdev->common.residue_granularity =
DMA_RESIDUE_GRANULARITY_SEGMENT;
} else if (xdev->dma_config->dmatype == XDMA_TYPE_CDMA) {
dma_cap_set(DMA_MEMCPY, xdev->common.cap_mask);
xdev->common.device_prep_dma_memcpy = xilinx_cdma_prep_memcpy;
+ /* Residue calculation is supported by only AXI DMA and CDMA */
+ xdev->common.residue_granularity =
+ DMA_RESIDUE_GRANULARITY_SEGMENT;
} else {
xdev->common.device_prep_interleaved_dma =
xilinx_vdma_dma_prep_interleaved;
--
2.7.4


2019-09-26 09:59:54

by Vinod Koul

[permalink] [raw]
Subject: Re: [PATCH -next 3/8] dmaengine: xilinx_dma: Introduce xilinx_dma_get_residue

On 05-09-19, 22:06, Radhey Shyam Pandey wrote:
> From: Nicholas Graumann <[email protected]>
>
> Introduce a function that can calculate residues for IPs that support it:
> AXI DMA and CDMA.
>
> Signed-off-by: Nicholas Graumann <[email protected]>
> Signed-off-by: Radhey Shyam Pandey <[email protected]>
> ---
> drivers/dma/xilinx/xilinx_dma.c | 75 ++++++++++++++++++++++++++++++-----------
> 1 file changed, 56 insertions(+), 19 deletions(-)
>
> diff --git a/drivers/dma/xilinx/xilinx_dma.c b/drivers/dma/xilinx/xilinx_dma.c
> index 9909bfb..4094adb 100644
> --- a/drivers/dma/xilinx/xilinx_dma.c
> +++ b/drivers/dma/xilinx/xilinx_dma.c
> @@ -787,6 +787,51 @@ static void xilinx_dma_free_chan_resources(struct dma_chan *dchan)
> }
>
> /**
> + * xilinx_dma_get_residue - Compute residue for a given descriptor
> + * @chan: Driver specific dma channel
> + * @desc: dma transaction descriptor
> + *
> + * Return: The number of residue bytes for the descriptor.
> + */
> +static u32 xilinx_dma_get_residue(struct xilinx_dma_chan *chan,
> + struct xilinx_dma_tx_descriptor *desc)
> +{
> + struct xilinx_cdma_tx_segment *cdma_seg;
> + struct xilinx_axidma_tx_segment *axidma_seg;
> + struct xilinx_cdma_desc_hw *cdma_hw;
> + struct xilinx_axidma_desc_hw *axidma_hw;
> + struct list_head *entry;
> + u32 residue = 0;
> +
> + /**

it should be:
/*
* comment...

> + * VDMA and simple mode do not support residue reporting, so the
> + * residue field will always be 0.
> + */
> + if (chan->xdev->dma_config->dmatype == XDMA_TYPE_VDMA || !chan->has_sg)
> + return residue;

why not check this in status callback?

> +
> + list_for_each(entry, &desc->segments) {
> + if (chan->xdev->dma_config->dmatype == XDMA_TYPE_CDMA) {
> + cdma_seg = list_entry(entry,
> + struct xilinx_cdma_tx_segment,
> + node);
> + cdma_hw = &cdma_seg->hw;
> + residue += (cdma_hw->control - cdma_hw->status) &
> + chan->xdev->max_buffer_len;
> + } else {
> + axidma_seg = list_entry(entry,
> + struct xilinx_axidma_tx_segment,
> + node);
> + axidma_hw = &axidma_seg->hw;
> + residue += (axidma_hw->control - axidma_hw->status) &
> + chan->xdev->max_buffer_len;
> + }
> + }
> +
> + return residue;
> +}
> +
> +/**
> * xilinx_dma_chan_handle_cyclic - Cyclic dma callback
> * @chan: Driver specific dma channel
> * @desc: dma transaction descriptor
> @@ -995,33 +1040,22 @@ static enum dma_status xilinx_dma_tx_status(struct dma_chan *dchan,
> {
> struct xilinx_dma_chan *chan = to_xilinx_chan(dchan);
> struct xilinx_dma_tx_descriptor *desc;
> - struct xilinx_axidma_tx_segment *segment;
> - struct xilinx_axidma_desc_hw *hw;
> enum dma_status ret;
> unsigned long flags;
> - u32 residue = 0;
>
> ret = dma_cookie_status(dchan, cookie, txstate);
> if (ret == DMA_COMPLETE || !txstate)
> return ret;
>
> - if (chan->xdev->dma_config->dmatype == XDMA_TYPE_AXIDMA) {
> - spin_lock_irqsave(&chan->lock, flags);
> + spin_lock_irqsave(&chan->lock, flags);
>
> - desc = list_last_entry(&chan->active_list,
> - struct xilinx_dma_tx_descriptor, node);
> - if (chan->has_sg) {
> - list_for_each_entry(segment, &desc->segments, node) {
> - hw = &segment->hw;
> - residue += (hw->control - hw->status) &
> - chan->xdev->max_buffer_len;
> - }
> - }
> - spin_unlock_irqrestore(&chan->lock, flags);
> + desc = list_last_entry(&chan->active_list,
> + struct xilinx_dma_tx_descriptor, node);
> + chan->residue = xilinx_dma_get_residue(chan, desc);
>
> - chan->residue = residue;
> - dma_set_residue(txstate, chan->residue);
> - }
> + spin_unlock_irqrestore(&chan->lock, flags);
> +
> + dma_set_residue(txstate, chan->residue);
>
> return ret;
> }
> @@ -2701,12 +2735,15 @@ static int xilinx_dma_probe(struct platform_device *pdev)
> xilinx_dma_prep_dma_cyclic;
> xdev->common.device_prep_interleaved_dma =
> xilinx_dma_prep_interleaved;
> - /* Residue calculation is supported by only AXI DMA */
> + /* Residue calculation is supported by only AXI DMA and CDMA */
> xdev->common.residue_granularity =
> DMA_RESIDUE_GRANULARITY_SEGMENT;
> } else if (xdev->dma_config->dmatype == XDMA_TYPE_CDMA) {
> dma_cap_set(DMA_MEMCPY, xdev->common.cap_mask);
> xdev->common.device_prep_dma_memcpy = xilinx_cdma_prep_memcpy;
> + /* Residue calculation is supported by only AXI DMA and CDMA */
> + xdev->common.residue_granularity =
> + DMA_RESIDUE_GRANULARITY_SEGMENT;
> } else {
> xdev->common.device_prep_interleaved_dma =
> xilinx_vdma_dma_prep_interleaved;
> --
> 2.7.4

--
~Vinod

2019-09-26 10:21:02

by Radhey Shyam Pandey

[permalink] [raw]
Subject: RE: [PATCH -next 3/8] dmaengine: xilinx_dma: Introduce xilinx_dma_get_residue

> -----Original Message-----
> From: Vinod Koul <[email protected]>
> Sent: Thursday, September 26, 2019 2:31 AM
> To: Radhey Shyam Pandey <[email protected]>
> Cc: [email protected]; Michal Simek <[email protected]>;
> [email protected]; [email protected]; Appana Durga
> Kedareswara Rao <[email protected]>; [email protected];
> [email protected]; [email protected]
> Subject: Re: [PATCH -next 3/8] dmaengine: xilinx_dma: Introduce
> xilinx_dma_get_residue
>
> On 05-09-19, 22:06, Radhey Shyam Pandey wrote:
> > From: Nicholas Graumann <[email protected]>
> >
> > Introduce a function that can calculate residues for IPs that support it:
> > AXI DMA and CDMA.
> >
> > Signed-off-by: Nicholas Graumann <[email protected]>
> > Signed-off-by: Radhey Shyam Pandey
> <[email protected]>
> > ---
> > drivers/dma/xilinx/xilinx_dma.c | 75
> > ++++++++++++++++++++++++++++++-----------
> > 1 file changed, 56 insertions(+), 19 deletions(-)
> >
> > diff --git a/drivers/dma/xilinx/xilinx_dma.c
> > b/drivers/dma/xilinx/xilinx_dma.c index 9909bfb..4094adb 100644
> > --- a/drivers/dma/xilinx/xilinx_dma.c
> > +++ b/drivers/dma/xilinx/xilinx_dma.c
> > @@ -787,6 +787,51 @@ static void xilinx_dma_free_chan_resources(struct
> > dma_chan *dchan) }
> >
> > /**
> > + * xilinx_dma_get_residue - Compute residue for a given descriptor
> > + * @chan: Driver specific dma channel
> > + * @desc: dma transaction descriptor
> > + *
> > + * Return: The number of residue bytes for the descriptor.
> > + */
> > +static u32 xilinx_dma_get_residue(struct xilinx_dma_chan *chan,
> > + struct xilinx_dma_tx_descriptor *desc) {
> > + struct xilinx_cdma_tx_segment *cdma_seg;
> > + struct xilinx_axidma_tx_segment *axidma_seg;
> > + struct xilinx_cdma_desc_hw *cdma_hw;
> > + struct xilinx_axidma_desc_hw *axidma_hw;
> > + struct list_head *entry;
> > + u32 residue = 0;
> > +
> > + /**
>
> it should be:
> /*
> * comment...
>
Thanks for pointing it out. I will fix it in v2.

> > + * VDMA and simple mode do not support residue reporting, so the
> > + * residue field will always be 0.
> > + */
> > + if (chan->xdev->dma_config->dmatype == XDMA_TYPE_VDMA ||
> !chan->has_sg)
> > + return residue;
>
> why not check this in status callback?
Assuming we mean to move vdma and non-sg check to xilinx_dma_tx_status.
Just a thought- Keeping this check in xilinx_dma_get_residue provides
an abstraction and caller can simply call this func with knowing about
IP config specific residue calculation. Considering this point does it
looks ok ?
>
> > +
> > + list_for_each(entry, &desc->segments) {
> > + if (chan->xdev->dma_config->dmatype ==
> XDMA_TYPE_CDMA) {
> > + cdma_seg = list_entry(entry,
> > + struct xilinx_cdma_tx_segment,
> > + node);
> > + cdma_hw = &cdma_seg->hw;
> > + residue += (cdma_hw->control - cdma_hw->status)
> &
> > + chan->xdev->max_buffer_len;
> > + } else {
> > + axidma_seg = list_entry(entry,
> > + struct
> xilinx_axidma_tx_segment,
> > + node);
> > + axidma_hw = &axidma_seg->hw;
> > + residue += (axidma_hw->control - axidma_hw-
> >status) &
> > + chan->xdev->max_buffer_len;
> > + }
> > + }
> > +
> > + return residue;
> > +}
> > +
> > +/**
> > * xilinx_dma_chan_handle_cyclic - Cyclic dma callback
> > * @chan: Driver specific dma channel
> > * @desc: dma transaction descriptor
> > @@ -995,33 +1040,22 @@ static enum dma_status
> > xilinx_dma_tx_status(struct dma_chan *dchan, {
> > struct xilinx_dma_chan *chan = to_xilinx_chan(dchan);
> > struct xilinx_dma_tx_descriptor *desc;
> > - struct xilinx_axidma_tx_segment *segment;
> > - struct xilinx_axidma_desc_hw *hw;
> > enum dma_status ret;
> > unsigned long flags;
> > - u32 residue = 0;
> >
> > ret = dma_cookie_status(dchan, cookie, txstate);
> > if (ret == DMA_COMPLETE || !txstate)
> > return ret;
> >
> > - if (chan->xdev->dma_config->dmatype == XDMA_TYPE_AXIDMA) {
> > - spin_lock_irqsave(&chan->lock, flags);
> > + spin_lock_irqsave(&chan->lock, flags);
> >
> > - desc = list_last_entry(&chan->active_list,
> > - struct xilinx_dma_tx_descriptor, node);
> > - if (chan->has_sg) {
> > - list_for_each_entry(segment, &desc->segments,
> node) {
> > - hw = &segment->hw;
> > - residue += (hw->control - hw->status) &
> > - chan->xdev->max_buffer_len;
> > - }
> > - }
> > - spin_unlock_irqrestore(&chan->lock, flags);
> > + desc = list_last_entry(&chan->active_list,
> > + struct xilinx_dma_tx_descriptor, node);
> > + chan->residue = xilinx_dma_get_residue(chan, desc);
> >
> > - chan->residue = residue;
> > - dma_set_residue(txstate, chan->residue);
> > - }
> > + spin_unlock_irqrestore(&chan->lock, flags);
> > +
> > + dma_set_residue(txstate, chan->residue);
> >
> > return ret;
> > }
> > @@ -2701,12 +2735,15 @@ static int xilinx_dma_probe(struct
> platform_device *pdev)
> > xilinx_dma_prep_dma_cyclic;
> > xdev->common.device_prep_interleaved_dma =
> > xilinx_dma_prep_interleaved;
> > - /* Residue calculation is supported by only AXI DMA */
> > + /* Residue calculation is supported by only AXI DMA and
> CDMA */
> > xdev->common.residue_granularity =
> >
> DMA_RESIDUE_GRANULARITY_SEGMENT;
> > } else if (xdev->dma_config->dmatype == XDMA_TYPE_CDMA) {
> > dma_cap_set(DMA_MEMCPY, xdev->common.cap_mask);
> > xdev->common.device_prep_dma_memcpy =
> xilinx_cdma_prep_memcpy;
> > + /* Residue calculation is supported by only AXI DMA and
> CDMA */
> > + xdev->common.residue_granularity =
> > +
> DMA_RESIDUE_GRANULARITY_SEGMENT;
> > } else {
> > xdev->common.device_prep_interleaved_dma =
> > xilinx_vdma_dma_prep_interleaved;
> > --
> > 2.7.4
>
> --
> ~Vinod

2019-09-26 17:22:44

by Vinod Koul

[permalink] [raw]
Subject: Re: [PATCH -next 3/8] dmaengine: xilinx_dma: Introduce xilinx_dma_get_residue

On 26-09-19, 05:52, Radhey Shyam Pandey wrote:

> > > + * VDMA and simple mode do not support residue reporting, so the
> > > + * residue field will always be 0.
> > > + */
> > > + if (chan->xdev->dma_config->dmatype == XDMA_TYPE_VDMA ||
> > !chan->has_sg)
> > > + return residue;
> >
> > why not check this in status callback?
> Assuming we mean to move vdma and non-sg check to xilinx_dma_tx_status.
> Just a thought- Keeping this check in xilinx_dma_get_residue provides
> an abstraction and caller can simply call this func with knowing about
> IP config specific residue calculation. Considering this point does it
> looks ok ?

well you are checking either way, so calling the lower level function
only when you need it makes more sense!

--
~Vinod

2019-09-27 05:21:58

by Radhey Shyam Pandey

[permalink] [raw]
Subject: RE: [PATCH -next 3/8] dmaengine: xilinx_dma: Introduce xilinx_dma_get_residue

> -----Original Message-----
> From: Vinod Koul <[email protected]>
> Sent: Thursday, September 26, 2019 10:48 PM
> To: Radhey Shyam Pandey <[email protected]>
> Cc: [email protected]; Michal Simek <[email protected]>;
> [email protected]; [email protected]; Appana Durga
> Kedareswara Rao <[email protected]>; [email protected];
> [email protected]; [email protected]
> Subject: Re: [PATCH -next 3/8] dmaengine: xilinx_dma: Introduce
> xilinx_dma_get_residue
>
> On 26-09-19, 05:52, Radhey Shyam Pandey wrote:
>
> > > > + * VDMA and simple mode do not support residue reporting, so the
> > > > + * residue field will always be 0.
> > > > + */
> > > > + if (chan->xdev->dma_config->dmatype == XDMA_TYPE_VDMA ||
> > > !chan->has_sg)
> > > > + return residue;
> > >
> > > why not check this in status callback?
> > Assuming we mean to move vdma and non-sg check to
> xilinx_dma_tx_status.
> > Just a thought- Keeping this check in xilinx_dma_get_residue provides
> > an abstraction and caller can simply call this func with knowing about
> > IP config specific residue calculation. Considering this point does it
> > looks ok ?
>
> well you are checking either way, so calling the lower level function
> only when you need it makes more sense!

Sure, will do it in v2.
>
> --
> ~Vinod

2019-10-01 12:04:53

by Radhey Shyam Pandey

[permalink] [raw]
Subject: RE: [PATCH -next 3/8] dmaengine: xilinx_dma: Introduce xilinx_dma_get_residue

> -----Original Message-----
> From: Radhey Shyam Pandey
> Sent: Friday, September 27, 2019 10:46 AM
> To: Vinod Koul <[email protected]>
> Cc: [email protected]; Michal Simek <[email protected]>;
> [email protected]; [email protected]; Appana Durga
> Kedareswara Rao <[email protected]>; [email protected];
> [email protected]; [email protected]
> Subject: RE: [PATCH -next 3/8] dmaengine: xilinx_dma: Introduce
> xilinx_dma_get_residue
>
> > -----Original Message-----
> > From: Vinod Koul <[email protected]>
> > Sent: Thursday, September 26, 2019 10:48 PM
> > To: Radhey Shyam Pandey <[email protected]>
> > Cc: [email protected]; Michal Simek <[email protected]>;
> > [email protected]; [email protected]; Appana Durga
> > Kedareswara Rao <[email protected]>; [email protected];
> > [email protected]; [email protected]
> > Subject: Re: [PATCH -next 3/8] dmaengine: xilinx_dma: Introduce
> > xilinx_dma_get_residue
> >
> > On 26-09-19, 05:52, Radhey Shyam Pandey wrote:
> >
> > > > > + * VDMA and simple mode do not support residue reporting,
> so the
> > > > > + * residue field will always be 0.
> > > > > + */
> > > > > + if (chan->xdev->dma_config->dmatype ==
> XDMA_TYPE_VDMA ||
> > > > !chan->has_sg)
> > > > > + return residue;
> > > >
> > > > why not check this in status callback?
> > > Assuming we mean to move vdma and non-sg check to
> > xilinx_dma_tx_status.
> > > Just a thought- Keeping this check in xilinx_dma_get_residue
> > > provides an abstraction and caller can simply call this func with
> > > knowing about IP config specific residue calculation. Considering
> > > this point does it looks ok ?
> >
> > well you are checking either way, so calling the lower level function
> > only when you need it makes more sense!
>
> Sure, will do it in v2.

Just noticed that xilinx_dma_get_residue() is called at the multiple
places i.e one in device_tx_status and other in _complete_descriptor.
To avoid code duplication, I will create a helper function to check
if residue calculation is supported.

> >
> > --
> > ~Vinod