Hi,
Changes since v2:
- Fix dma-common.yaml documentation patch and extend the description of the
dma-channel-mask array
- The edma documentation now includes information on the dma-channel-mask array
size for EDMAs with 32 or 64 channels
Changes since v1:
- Extend the common dma-channel-mask to uint32-array to be usable for
controllers with more than 32 channels
- Use the dma-channel-mask instead custom property for available channels for
EDMA.
The original patch was part of the EDMA multicore usage series.
Rob: I'm not sure if I got the dma-common.yaml update correctly...
EDMAs can have 32 or 64 channels depending on the SoC, the dma-channel-mask
needs to be an array to be usable for the driver.
Regards,
Peter
---
Peter Ujfalusi (3):
dt-bindings: dmaengine: dma-common: Change dma-channel-mask to
uint32-array
dt-bindings: dma: ti-edma: Document dma-channel-mask for EDMA
dmaengine: ti: edma: Add support for handling reserved channels
.../devicetree/bindings/dma/dma-common.yaml | 9 ++-
.../devicetree/bindings/dma/ti-edma.txt | 8 +++
drivers/dma/ti/edma.c | 59 +++++++++++++++++--
3 files changed, 69 insertions(+), 7 deletions(-)
--
Peter
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
Like paRAM slots, channels could be used by other cores and in this case
we need to make sure that the driver do not alter these channels.
Handle the generic dma-channel-mask property to mark channels in a bitmap
which can not be used by Linux and convert the legacy rsv_chans if it is
provided by platform_data.
Signed-off-by: Peter Ujfalusi <[email protected]>
---
drivers/dma/ti/edma.c | 59 ++++++++++++++++++++++++++++++++++++++-----
1 file changed, 53 insertions(+), 6 deletions(-)
diff --git a/drivers/dma/ti/edma.c b/drivers/dma/ti/edma.c
index ba7c4f07fcd6..03c9c6296006 100644
--- a/drivers/dma/ti/edma.c
+++ b/drivers/dma/ti/edma.c
@@ -260,6 +260,13 @@ struct edma_cc {
*/
unsigned long *slot_inuse;
+ /*
+ * For tracking reserved channels used by DSP.
+ * If the bit is cleared, the channel is allocated to be used by DSP
+ * and Linux must not touch it.
+ */
+ unsigned long *channels_mask;
+
struct dma_device dma_slave;
struct dma_device *dma_memcpy;
struct edma_chan *slave_chans;
@@ -716,6 +723,12 @@ static int edma_alloc_channel(struct edma_chan *echan,
struct edma_cc *ecc = echan->ecc;
int channel = EDMA_CHAN_SLOT(echan->ch_num);
+ if (!test_bit(echan->ch_num, ecc->channels_mask)) {
+ dev_err(ecc->dev, "Channel%d is reserved, can not be used!\n",
+ echan->ch_num);
+ return -EINVAL;
+ }
+
/* ensure access through shadow region 0 */
edma_or_array2(ecc, EDMA_DRAE, 0, EDMA_REG_ARRAY_INDEX(channel),
EDMA_CHANNEL_BIT(channel));
@@ -2250,7 +2263,7 @@ static int edma_probe(struct platform_device *pdev)
struct edma_soc_info *info = pdev->dev.platform_data;
s8 (*queue_priority_mapping)[2];
int i, off;
- const s16 (*rsv_slots)[2];
+ const s16 (*reserved)[2];
const s16 (*xbar_chans)[2];
int irq;
char *irq_name;
@@ -2331,15 +2344,32 @@ static int edma_probe(struct platform_device *pdev)
if (!ecc->slot_inuse)
return -ENOMEM;
+ ecc->channels_mask = devm_kcalloc(dev,
+ BITS_TO_LONGS(ecc->num_channels),
+ sizeof(unsigned long), GFP_KERNEL);
+ if (!ecc->channels_mask)
+ return -ENOMEM;
+
+ /* Mark all channels available initially */
+ bitmap_fill(ecc->channels_mask, ecc->num_channels);
+
ecc->default_queue = info->default_queue;
if (info->rsv) {
/* Set the reserved slots in inuse list */
- rsv_slots = info->rsv->rsv_slots;
- if (rsv_slots) {
- for (i = 0; rsv_slots[i][0] != -1; i++)
- bitmap_set(ecc->slot_inuse, rsv_slots[i][0],
- rsv_slots[i][1]);
+ reserved = info->rsv->rsv_slots;
+ if (reserved) {
+ for (i = 0; reserved[i][0] != -1; i++)
+ bitmap_set(ecc->slot_inuse, reserved[i][0],
+ reserved[i][1]);
+ }
+
+ /* Clear channels not usable for Linux */
+ reserved = info->rsv->rsv_chans;
+ if (reserved) {
+ for (i = 0; reserved[i][0] != -1; i++)
+ bitmap_clear(ecc->channels_mask, reserved[i][0],
+ reserved[i][1]);
}
}
@@ -2399,6 +2429,7 @@ static int edma_probe(struct platform_device *pdev)
if (!ecc->legacy_mode) {
int lowest_priority = 0;
+ unsigned int array_max;
struct of_phandle_args tc_args;
ecc->tc_list = devm_kcalloc(dev, ecc->num_tc,
@@ -2420,6 +2451,18 @@ static int edma_probe(struct platform_device *pdev)
info->default_queue = i;
}
}
+
+ /* See if we have optional dma-channel-mask array */
+ array_max = DIV_ROUND_UP(ecc->num_channels, BITS_PER_TYPE(u32));
+ ret = of_property_read_variable_u32_array(node,
+ "dma-channel-mask",
+ (u32 *)ecc->channels_mask,
+ 1, array_max);
+ if (ret > 0 && ret != array_max)
+ dev_warn(dev, "dma-channel-mask is not complete.\n");
+ else if (ret == -EOVERFLOW || ret == -ENODATA)
+ dev_warn(dev,
+ "dma-channel-mask is out of range or empty\n");
}
/* Event queue priority mapping */
@@ -2437,6 +2480,10 @@ static int edma_probe(struct platform_device *pdev)
edma_dma_init(ecc, legacy_mode);
for (i = 0; i < ecc->num_channels; i++) {
+ /* Do not touch reserved channels */
+ if (!test_bit(i, ecc->channels_mask))
+ continue;
+
/* Assign all channels to the default queue */
edma_assign_channel_eventq(&ecc->slave_chans[i],
info->default_queue);
--
Peter
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
Make the dma-channel-mask to be usable for controllers with more than 32
channels.
Signed-off-by: Peter Ujfalusi <[email protected]>
---
Documentation/devicetree/bindings/dma/dma-common.yaml | 9 ++++++++-
1 file changed, 8 insertions(+), 1 deletion(-)
diff --git a/Documentation/devicetree/bindings/dma/dma-common.yaml b/Documentation/devicetree/bindings/dma/dma-common.yaml
index ed0a49a6f020..4527f20301ff 100644
--- a/Documentation/devicetree/bindings/dma/dma-common.yaml
+++ b/Documentation/devicetree/bindings/dma/dma-common.yaml
@@ -25,11 +25,18 @@ properties:
Used to provide DMA controller specific information.
dma-channel-mask:
- $ref: /schemas/types.yaml#definitions/uint32
description:
Bitmask of available DMA channels in ascending order that are
not reserved by firmware and are available to the
kernel. i.e. first channel corresponds to LSB.
+ The first item in the array is for channels 0-31, the second is for
+ channels 32-63, etc.
+ allOf:
+ - $ref: /schemas/types.yaml#/definitions/uint32-array
+ items:
+ minItems: 1
+ # Should be enough
+ maxItems: 255
dma-channels:
$ref: /schemas/types.yaml#definitions/uint32
--
Peter
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
Similarly to paRAM slots, channels can be used by other cores.
The common dma-channel-mask property can be used for specifying the
available channels.
Signed-off-by: Peter Ujfalusi <[email protected]>
---
Documentation/devicetree/bindings/dma/ti-edma.txt | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/Documentation/devicetree/bindings/dma/ti-edma.txt b/Documentation/devicetree/bindings/dma/ti-edma.txt
index 4bbc94d829c8..014187088020 100644
--- a/Documentation/devicetree/bindings/dma/ti-edma.txt
+++ b/Documentation/devicetree/bindings/dma/ti-edma.txt
@@ -42,6 +42,11 @@ Optional properties:
- ti,edma-reserved-slot-ranges: PaRAM slot ranges which should not be used by
the driver, they are allocated to be used by for example the
DSP. See example.
+- dma-channel-mask: Mask of usable channels.
+ Single uint32 for EDMA with 32 channels, array of two uint32 for
+ EDMA with 64 channels. See example and
+ Documentation/devicetree/bindings/dma/dma-common.yaml
+
------------------------------------------------------------------------------
eDMA3 Transfer Controller
@@ -91,6 +96,9 @@ edma: edma@49000000 {
ti,edma-memcpy-channels = <20 21>;
/* The following PaRAM slots are reserved: 35-44 and 100-109 */
ti,edma-reserved-slot-ranges = <35 10>, <100 10>;
+ /* The following channels are reserved: 35-44 */
+ dma-channel-mask = <0xffffffff>, /* Channel 0-31 */
+ <0xffffe007>; /* Channel 32-63 */
};
edma_tptc0: tptc@49800000 {
--
Peter
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
On Thu, Sep 26, 2019 at 02:19:52PM +0300, Peter Ujfalusi wrote:
> Make the dma-channel-mask to be usable for controllers with more than 32
> channels.
>
> Signed-off-by: Peter Ujfalusi <[email protected]>
> ---
> Documentation/devicetree/bindings/dma/dma-common.yaml | 9 ++++++++-
> 1 file changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/Documentation/devicetree/bindings/dma/dma-common.yaml b/Documentation/devicetree/bindings/dma/dma-common.yaml
> index ed0a49a6f020..4527f20301ff 100644
> --- a/Documentation/devicetree/bindings/dma/dma-common.yaml
> +++ b/Documentation/devicetree/bindings/dma/dma-common.yaml
> @@ -25,11 +25,18 @@ properties:
> Used to provide DMA controller specific information.
>
> dma-channel-mask:
> - $ref: /schemas/types.yaml#definitions/uint32
> description:
> Bitmask of available DMA channels in ascending order that are
> not reserved by firmware and are available to the
> kernel. i.e. first channel corresponds to LSB.
> + The first item in the array is for channels 0-31, the second is for
> + channels 32-63, etc.
> + allOf:
> + - $ref: /schemas/types.yaml#/definitions/uint32-array
> + items:
> + minItems: 1
> + # Should be enough
> + maxItems: 255
'items' has to be a separate sub-schema from $ref to have any effect:
allOf:
- $ref: /schemas/types.yaml#/definitions/uint32-array
items:
minItems: 1
# Should be enough
maxItems: 255
Or (note the added '-'):
allOf:
- $ref: /schemas/types.yaml#/definitions/uint32-array
- items:
minItems: 1
# Should be enough
maxItems: 255
The first way is my preference.
Rob
On Thu, Sep 26, 2019 at 02:19:53PM +0300, Peter Ujfalusi wrote:
> Similarly to paRAM slots, channels can be used by other cores.
>
> The common dma-channel-mask property can be used for specifying the
> available channels.
>
> Signed-off-by: Peter Ujfalusi <[email protected]>
> ---
> Documentation/devicetree/bindings/dma/ti-edma.txt | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> diff --git a/Documentation/devicetree/bindings/dma/ti-edma.txt b/Documentation/devicetree/bindings/dma/ti-edma.txt
> index 4bbc94d829c8..014187088020 100644
> --- a/Documentation/devicetree/bindings/dma/ti-edma.txt
> +++ b/Documentation/devicetree/bindings/dma/ti-edma.txt
> @@ -42,6 +42,11 @@ Optional properties:
> - ti,edma-reserved-slot-ranges: PaRAM slot ranges which should not be used by
> the driver, they are allocated to be used by for example the
> DSP. See example.
> +- dma-channel-mask: Mask of usable channels.
> + Single uint32 for EDMA with 32 channels, array of two uint32 for
> + EDMA with 64 channels. See example and
> + Documentation/devicetree/bindings/dma/dma-common.yaml
> +
>
> ------------------------------------------------------------------------------
> eDMA3 Transfer Controller
> @@ -91,6 +96,9 @@ edma: edma@49000000 {
> ti,edma-memcpy-channels = <20 21>;
> /* The following PaRAM slots are reserved: 35-44 and 100-109 */
> ti,edma-reserved-slot-ranges = <35 10>, <100 10>;
> + /* The following channels are reserved: 35-44 */
> + dma-channel-mask = <0xffffffff>, /* Channel 0-31 */
> + <0xffffe007>; /* Channel 32-63 */
Doesn't matter yet, but you have a mismatch here with the schema. While
the <> around each int or not doesn't matter for the dtb, it does for
the schema.
dma-channel-mask = <0xffffffff>, <0xffffe007>;
minItems: 1
maxItems: 255
dma-channel-mask = <0xffffffff 0xffffe007>;
items:
minItems: 1
maxItems: 255
I think the latter case is slightly more logical here as you have 1
thing (a mask). If had N of something (like interrupts), then the former
makes sense.
Rob
On 27/09/2019 23.48, Rob Herring wrote:
> On Thu, Sep 26, 2019 at 02:19:53PM +0300, Peter Ujfalusi wrote:
>> Similarly to paRAM slots, channels can be used by other cores.
>>
>> The common dma-channel-mask property can be used for specifying the
>> available channels.
>>
>> Signed-off-by: Peter Ujfalusi <[email protected]>
>> ---
>> Documentation/devicetree/bindings/dma/ti-edma.txt | 8 ++++++++
>> 1 file changed, 8 insertions(+)
>>
>> diff --git a/Documentation/devicetree/bindings/dma/ti-edma.txt b/Documentation/devicetree/bindings/dma/ti-edma.txt
>> index 4bbc94d829c8..014187088020 100644
>> --- a/Documentation/devicetree/bindings/dma/ti-edma.txt
>> +++ b/Documentation/devicetree/bindings/dma/ti-edma.txt
>> @@ -42,6 +42,11 @@ Optional properties:
>> - ti,edma-reserved-slot-ranges: PaRAM slot ranges which should not be used by
>> the driver, they are allocated to be used by for example the
>> DSP. See example.
>> +- dma-channel-mask: Mask of usable channels.
>> + Single uint32 for EDMA with 32 channels, array of two uint32 for
>> + EDMA with 64 channels. See example and
>> + Documentation/devicetree/bindings/dma/dma-common.yaml
>> +
>>
>> ------------------------------------------------------------------------------
>> eDMA3 Transfer Controller
>> @@ -91,6 +96,9 @@ edma: edma@49000000 {
>> ti,edma-memcpy-channels = <20 21>;
>> /* The following PaRAM slots are reserved: 35-44 and 100-109 */
>> ti,edma-reserved-slot-ranges = <35 10>, <100 10>;
>> + /* The following channels are reserved: 35-44 */
>> + dma-channel-mask = <0xffffffff>, /* Channel 0-31 */
>> + <0xffffe007>; /* Channel 32-63 */
>
> Doesn't matter yet, but you have a mismatch here with the schema. While
> the <> around each int or not doesn't matter for the dtb, it does for
> the schema.
>
> dma-channel-mask = <0xffffffff>, <0xffffe007>;
> minItems: 1
> maxItems: 255
>
> dma-channel-mask = <0xffffffff 0xffffe007>;
> items:
> minItems: 1
> maxItems: 255
>
> I think the latter case is slightly more logical here as you have 1
> thing (a mask). If had N of something (like interrupts), then the former
> makes sense.
So, in dma-common.yaml:
allOf:
- $ref: /schemas/types.yaml#/definitions/uint32-array
items:
minItems: 1
# Should be enough
maxItems: 255
and here in the example:
dma-channel-mask = <0xffffffff /* Channel 0-31 */
0xffffe007>; /* Channel 32-63 */
I'll send an updated series with these changes.
Thanks,
- Péter
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki