LinuxLists.cc - [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is created

[permalink] [raw]

Subject: Re: [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is created

On Sat, Jul 8, 2017 at 11:41 AM, Jan Kiszka <[email protected]> wrote:
> From: Jan Kiszka <[email protected]>
>
> Avoid hogging chip select GPIOs just because they are listed for the
> master. They might be mulitplexed and, if no slave device is attached,
> used for different purposes. Moreover, this strategy avoids having to
> allocate a cs_gpiods structure.
>
> Tested on the IOT2000 where the second SPI bus is connected to an
> Arduino-compatible connector and multiplexed between SPI, GPIO and PWM
> usage.

Can we first switch the driver to use GPIO descriptors instead of
plain integers?

>
> Signed-off-by: Jan Kiszka <[email protected]>
> ---
> drivers/spi/spi-pxa2xx.c | 59 +++++++++++++++++-------------------------------
> 1 file changed, 21 insertions(+), 38 deletions(-)
>
> diff --git a/drivers/spi/spi-pxa2xx.c b/drivers/spi/spi-pxa2xx.c
> index 38d053682892..be991266a6ce 100644
> --- a/drivers/spi/spi-pxa2xx.c
> +++ b/drivers/spi/spi-pxa2xx.c
> @@ -1213,21 +1213,33 @@ static int setup_cs(struct spi_device *spi, struct chip_data *chip,
> struct pxa2xx_spi_chip *chip_info)
> {
> struct driver_data *drv_data = spi_master_get_devdata(spi->master);
> + struct device *pdev = &drv_data->pdev->dev;
> + struct gpio_desc *gpiod;
> int err = 0;
> + int count;
>
> if (chip == NULL)
> return 0;
>
> - if (drv_data->cs_gpiods) {
> - struct gpio_desc *gpiod;
> + count = gpiod_count(pdev, "cs");
> + if (count > 0) {
> + if (spi->chip_select >= count)
> + return -EINVAL;
> +
> + gpiod = gpiod_get_index(pdev, "cs", spi->chip_select,
> + GPIOD_OUT_HIGH);
> + if (IS_ERR(gpiod)) {
> + /* Means use native chip select */
> + if (PTR_ERR(gpiod) == -ENOENT)
> + return 0;
>
> - gpiod = drv_data->cs_gpiods[spi->chip_select];
> - if (gpiod) {
> - chip->gpio_cs = desc_to_gpio(gpiod);
> - chip->gpio_cs_inverted = spi->mode & SPI_CS_HIGH;
> - gpiod_set_value(gpiod, chip->gpio_cs_inverted);
> + return PTR_ERR(gpiod);
> }
>
> + chip->gpio_cs = desc_to_gpio(gpiod);
> + chip->gpio_cs_inverted = spi->mode & SPI_CS_HIGH;
> + gpiod_set_value(gpiod, chip->gpio_cs_inverted);
> +
> return 0;
> }
>
> @@ -1415,8 +1427,7 @@ static void cleanup(struct spi_device *spi)
> if (!chip)
> return;
>
> - if (drv_data->ssp_type != CE4100_SSP && !drv_data->cs_gpiods &&
> - gpio_is_valid(chip->gpio_cs))
> + if (drv_data->ssp_type != CE4100_SSP && gpio_is_valid(chip->gpio_cs))
> gpio_free(chip->gpio_cs);
>
> kfree(chip);
> @@ -1752,38 +1763,10 @@ static int pxa2xx_spi_probe(struct platform_device *pdev)
> master->num_chipselect = platform_info->num_chipselect;
>
> count = gpiod_count(&pdev->dev, "cs");
> - if (count > 0) {
> - int i;
> -
> + if (count > 0)
> master->num_chipselect = max_t(int, count,
> master->num_chipselect);
>
> - drv_data->cs_gpiods = devm_kcalloc(&pdev->dev,
> - master->num_chipselect, sizeof(struct gpio_desc *),
> - GFP_KERNEL);
> - if (!drv_data->cs_gpiods) {
> - status = -ENOMEM;
> - goto out_error_clock_enabled;
> - }
> -
> - for (i = 0; i < master->num_chipselect; i++) {
> - struct gpio_desc *gpiod;
> -
> - gpiod = devm_gpiod_get_index(dev, "cs", i,
> - GPIOD_OUT_HIGH);
> - if (IS_ERR(gpiod)) {
> - /* Means use native chip select */
> - if (PTR_ERR(gpiod) == -ENOENT)
> - continue;
> -
> - status = (int)PTR_ERR(gpiod);
> - goto out_error_clock_enabled;
> - } else {
> - drv_data->cs_gpiods[i] = gpiod;
> - }
> - }
> - }
> -
> tasklet_init(&drv_data->pump_transfers, pump_transfers,
> (unsigned long)drv_data);
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-spi" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html

--
With Best Regards,
Andy Shevchenko

2017-07-09 09:30:52

[permalink] [raw]

Subject: Re: [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is created

On 2017-07-08 23:48, Andy Shevchenko wrote:
> On Sat, Jul 8, 2017 at 11:41 AM, Jan Kiszka <[email protected]> wrote:
>> From: Jan Kiszka <[email protected]>
>>
>> Avoid hogging chip select GPIOs just because they are listed for the
>> master. They might be mulitplexed and, if no slave device is attached,
>> used for different purposes. Moreover, this strategy avoids having to
>> allocate a cs_gpiods structure.
>>
>> Tested on the IOT2000 where the second SPI bus is connected to an
>> Arduino-compatible connector and multiplexed between SPI, GPIO and PWM
>> usage.
>
> Can we first switch the driver to use GPIO descriptors instead of
> plain integers?

-ENOPARSE

>
>>
>> Signed-off-by: Jan Kiszka <[email protected]>
>> ---
>> drivers/spi/spi-pxa2xx.c | 59 +++++++++++++++++-------------------------------
>> 1 file changed, 21 insertions(+), 38 deletions(-)
>>
>> diff --git a/drivers/spi/spi-pxa2xx.c b/drivers/spi/spi-pxa2xx.c
>> index 38d053682892..be991266a6ce 100644
>> --- a/drivers/spi/spi-pxa2xx.c
>> +++ b/drivers/spi/spi-pxa2xx.c
>> @@ -1213,21 +1213,33 @@ static int setup_cs(struct spi_device *spi, struct chip_data *chip,
>> struct pxa2xx_spi_chip *chip_info)
>> {
>> struct driver_data *drv_data = spi_master_get_devdata(spi->master);
>> + struct device *pdev = &drv_data->pdev->dev;
>> + struct gpio_desc *gpiod;
>> int err = 0;
>> + int count;
>>
>> if (chip == NULL)
>> return 0;
>>
>> - if (drv_data->cs_gpiods) {
>> - struct gpio_desc *gpiod;
>> + count = gpiod_count(pdev, "cs");
>> + if (count > 0) {
>> + if (spi->chip_select >= count)
>> + return -EINVAL;
>> +
>> + gpiod = gpiod_get_index(pdev, "cs", spi->chip_select,
>> + GPIOD_OUT_HIGH);
>> + if (IS_ERR(gpiod)) {
>> + /* Means use native chip select */
>> + if (PTR_ERR(gpiod) == -ENOENT)
>> + return 0;
>>
>> - gpiod = drv_data->cs_gpiods[spi->chip_select];
>> - if (gpiod) {
>> - chip->gpio_cs = desc_to_gpio(gpiod);
>> - chip->gpio_cs_inverted = spi->mode & SPI_CS_HIGH;
>> - gpiod_set_value(gpiod, chip->gpio_cs_inverted);
>> + return PTR_ERR(gpiod);
>> }
>>
>> + chip->gpio_cs = desc_to_gpio(gpiod);
>> + chip->gpio_cs_inverted = spi->mode & SPI_CS_HIGH;
>> + gpiod_set_value(gpiod, chip->gpio_cs_inverted);
>> +
>> return 0;
>> }
>>
>> @@ -1415,8 +1427,7 @@ static void cleanup(struct spi_device *spi)
>> if (!chip)
>> return;
>>
>> - if (drv_data->ssp_type != CE4100_SSP && !drv_data->cs_gpiods &&
>> - gpio_is_valid(chip->gpio_cs))
>> + if (drv_data->ssp_type != CE4100_SSP && gpio_is_valid(chip->gpio_cs))
>> gpio_free(chip->gpio_cs);
>>
>> kfree(chip);
>> @@ -1752,38 +1763,10 @@ static int pxa2xx_spi_probe(struct platform_device *pdev)
>> master->num_chipselect = platform_info->num_chipselect;
>>
>> count = gpiod_count(&pdev->dev, "cs");
>> - if (count > 0) {
>> - int i;
>> -
>> + if (count > 0)
>> master->num_chipselect = max_t(int, count,
>> master->num_chipselect);
>>
>> - drv_data->cs_gpiods = devm_kcalloc(&pdev->dev,
>> - master->num_chipselect, sizeof(struct gpio_desc *),
>> - GFP_KERNEL);
>> - if (!drv_data->cs_gpiods) {
>> - status = -ENOMEM;
>> - goto out_error_clock_enabled;
>> - }
>> -
>> - for (i = 0; i < master->num_chipselect; i++) {
>> - struct gpio_desc *gpiod;
>> -
>> - gpiod = devm_gpiod_get_index(dev, "cs", i,
>> - GPIOD_OUT_HIGH);
>> - if (IS_ERR(gpiod)) {
>> - /* Means use native chip select */
>> - if (PTR_ERR(gpiod) == -ENOENT)
>> - continue;
>> -
>> - status = (int)PTR_ERR(gpiod);
>> - goto out_error_clock_enabled;
>> - } else {
>> - drv_data->cs_gpiods[i] = gpiod;
>> - }
>> - }
>> - }
>> -
>> tasklet_init(&drv_data->pump_transfers, pump_transfers,
>> (unsigned long)drv_data);
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-spi" in
>> the body of a message to [email protected]
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
>
>

Attachments:

2017-07-09 09:55:34

[permalink] [raw]

Subject: Re: [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is created

On Sun, Jul 9, 2017 at 12:30 PM, Jan Kiszka <[email protected]> wrote:
> On 2017-07-08 23:48, Andy Shevchenko wrote:
>> On Sat, Jul 8, 2017 at 11:41 AM, Jan Kiszka <[email protected]> wrote:
>>> From: Jan Kiszka <[email protected]>
>>>
>>> Avoid hogging chip select GPIOs just because they are listed for the
>>> master. They might be mulitplexed and, if no slave device is attached,
>>> used for different purposes. Moreover, this strategy avoids having to
>>> allocate a cs_gpiods structure.
>>>
>>> Tested on the IOT2000 where the second SPI bus is connected to an
>>> Arduino-compatible connector and multiplexed between SPI, GPIO and PWM
>>> usage.
>>
>> Can we first switch the driver to use GPIO descriptors instead of
>> plain integers?
>
> -ENOPARSE

In code you are trying to modify there is a mix of plain integers and
GPIO descriptors (and two APIs).
Can we just convert it to use GPIO descriptors API?

>
>>
>>>
>>> Signed-off-by: Jan Kiszka <[email protected]>
>>> ---
>>> drivers/spi/spi-pxa2xx.c | 59 +++++++++++++++++-------------------------------
>>> 1 file changed, 21 insertions(+), 38 deletions(-)
>>>
>>> diff --git a/drivers/spi/spi-pxa2xx.c b/drivers/spi/spi-pxa2xx.c
>>> index 38d053682892..be991266a6ce 100644
>>> --- a/drivers/spi/spi-pxa2xx.c
>>> +++ b/drivers/spi/spi-pxa2xx.c
>>> @@ -1213,21 +1213,33 @@ static int setup_cs(struct spi_device *spi, struct chip_data *chip,
>>> struct pxa2xx_spi_chip *chip_info)
>>> {
>>> struct driver_data *drv_data = spi_master_get_devdata(spi->master);
>>> + struct device *pdev = &drv_data->pdev->dev;
>>> + struct gpio_desc *gpiod;
>>> int err = 0;
>>> + int count;
>>>
>>> if (chip == NULL)
>>> return 0;
>>>
>>> - if (drv_data->cs_gpiods) {
>>> - struct gpio_desc *gpiod;
>>> + count = gpiod_count(pdev, "cs");
>>> + if (count > 0) {
>>> + if (spi->chip_select >= count)
>>> + return -EINVAL;
>>> +
>>> + gpiod = gpiod_get_index(pdev, "cs", spi->chip_select,
>>> + GPIOD_OUT_HIGH);
>>> + if (IS_ERR(gpiod)) {
>>> + /* Means use native chip select */
>>> + if (PTR_ERR(gpiod) == -ENOENT)
>>> + return 0;
>>>
>>> - gpiod = drv_data->cs_gpiods[spi->chip_select];
>>> - if (gpiod) {
>>> - chip->gpio_cs = desc_to_gpio(gpiod);
>>> - chip->gpio_cs_inverted = spi->mode & SPI_CS_HIGH;
>>> - gpiod_set_value(gpiod, chip->gpio_cs_inverted);
>>> + return PTR_ERR(gpiod);
>>> }
>>>
>>> + chip->gpio_cs = desc_to_gpio(gpiod);
>>> + chip->gpio_cs_inverted = spi->mode & SPI_CS_HIGH;
>>> + gpiod_set_value(gpiod, chip->gpio_cs_inverted);
>>> +
>>> return 0;
>>> }
>>>
>>> @@ -1415,8 +1427,7 @@ static void cleanup(struct spi_device *spi)
>>> if (!chip)
>>> return;
>>>
>>> - if (drv_data->ssp_type != CE4100_SSP && !drv_data->cs_gpiods &&
>>> - gpio_is_valid(chip->gpio_cs))
>>> + if (drv_data->ssp_type != CE4100_SSP && gpio_is_valid(chip->gpio_cs))
>>> gpio_free(chip->gpio_cs);
>>>
>>> kfree(chip);
>>> @@ -1752,38 +1763,10 @@ static int pxa2xx_spi_probe(struct platform_device *pdev)
>>> master->num_chipselect = platform_info->num_chipselect;
>>>
>>> count = gpiod_count(&pdev->dev, "cs");
>>> - if (count > 0) {
>>> - int i;
>>> -
>>> + if (count > 0)
>>> master->num_chipselect = max_t(int, count,
>>> master->num_chipselect);
>>>
>>> - drv_data->cs_gpiods = devm_kcalloc(&pdev->dev,
>>> - master->num_chipselect, sizeof(struct gpio_desc *),
>>> - GFP_KERNEL);
>>> - if (!drv_data->cs_gpiods) {
>>> - status = -ENOMEM;
>>> - goto out_error_clock_enabled;
>>> - }
>>> -
>>> - for (i = 0; i < master->num_chipselect; i++) {
>>> - struct gpio_desc *gpiod;
>>> -
>>> - gpiod = devm_gpiod_get_index(dev, "cs", i,
>>> - GPIOD_OUT_HIGH);
>>> - if (IS_ERR(gpiod)) {
>>> - /* Means use native chip select */
>>> - if (PTR_ERR(gpiod) == -ENOENT)
>>> - continue;
>>> -
>>> - status = (int)PTR_ERR(gpiod);
>>> - goto out_error_clock_enabled;
>>> - } else {
>>> - drv_data->cs_gpiods[i] = gpiod;
>>> - }
>>> - }
>>> - }
>>> -
>>> tasklet_init(&drv_data->pump_transfers, pump_transfers,
>>> (unsigned long)drv_data);
>>>
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-spi" in
>>> the body of a message to [email protected]
>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>
>>
>>
>
>

--
With Best Regards,
Andy Shevchenko

2017-07-09 10:31:19

[permalink] [raw]

Subject: Re: [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is created

On 2017-07-09 11:55, Andy Shevchenko wrote:
> On Sun, Jul 9, 2017 at 12:30 PM, Jan Kiszka <[email protected]> wrote:
>> On 2017-07-08 23:48, Andy Shevchenko wrote:
>>> On Sat, Jul 8, 2017 at 11:41 AM, Jan Kiszka <[email protected]> wrote:
>>>> From: Jan Kiszka <[email protected]>
>>>>
>>>> Avoid hogging chip select GPIOs just because they are listed for the
>>>> master. They might be mulitplexed and, if no slave device is attached,
>>>> used for different purposes. Moreover, this strategy avoids having to
>>>> allocate a cs_gpiods structure.
>>>>
>>>> Tested on the IOT2000 where the second SPI bus is connected to an
>>>> Arduino-compatible connector and multiplexed between SPI, GPIO and PWM
>>>> usage.
>>>
>>> Can we first switch the driver to use GPIO descriptors instead of
>>> plain integers?
>>
>> -ENOPARSE
>
> In code you are trying to modify there is a mix of plain integers and
> GPIO descriptors (and two APIs).
> Can we just convert it to use GPIO descriptors API?

That wasn't helpful either for someone not doing gpio development the
whole day. I suppose you wanted to suggest the conversion of
chip_data::chip_cs to struct gpio_desc * and the consistent usage of
gpiod_*, instead of gpio_*. That's not necessarily a precondition, but
it's a reasonable cleanup.

Jan

Attachments:

2017-07-10 12:09:32

by Mark Brown

[permalink] [raw]

Subject: Re: [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is created

On Sun, Jul 09, 2017 at 12:48:10AM +0300, Andy Shevchenko wrote:
> On Sat, Jul 8, 2017 at 11:41 AM, Jan Kiszka <[email protected]> wrote:

> > Tested on the IOT2000 where the second SPI bus is connected to an
> > Arduino-compatible connector and multiplexed between SPI, GPIO and PWM
> > usage.

> Can we first switch the driver to use GPIO descriptors instead of
> plain integers?

Only if you also convert the SPI core to use descriptors, Chris Packham
was looking at that but he needed update the ep93xx drivers among others
and getting reviewers was hard.

Attachments:

(No filename) (560.00 B)
signature.asc (488.00 B)
Download all attachments

2017-07-10 17:32:32

[permalink] [raw]

Subject: Re: [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is created

On 2017-07-10 14:09, Mark Brown wrote:
> On Sun, Jul 09, 2017 at 12:48:10AM +0300, Andy Shevchenko wrote:
>> On Sat, Jul 8, 2017 at 11:41 AM, Jan Kiszka <[email protected]> wrote:
>
>>> Tested on the IOT2000 where the second SPI bus is connected to an
>>> Arduino-compatible connector and multiplexed between SPI, GPIO and PWM
>>> usage.
>
>> Can we first switch the driver to use GPIO descriptors instead of
>> plain integers?
>
> Only if you also convert the SPI core to use descriptors, Chris Packham
> was looking at that but he needed update the ep93xx drivers among others
> and getting reviewers was hard.
>

IIUC, we can't convert completely due to some legacy boards providing
their CS lines as integers. But even then, a few more API usages can be
converted.

While looking into this, I noticed that this patch violated the formal
rule to never release a GPIO with the old API when it was requested with
the new one. That's at least stated in the docs, even though gpio_free
is equivalent to gpiod_put. Fixed that already, but I need to find some
time to retest.

Jan

Attachments:

2017-07-17 16:07:20

by Mark Brown

[permalink] [raw]

Subject: Applied "spi: pxa2xx: Only claim CS GPIOs when the slave device is created" to the spi tree

The patch

spi: pxa2xx: Only claim CS GPIOs when the slave device is created

has been applied to the spi tree at

git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git

All being well this means that it will be integrated into the linux-next
tree (usually sometime in the next 24 hours) and sent to Linus during
the next merge window (or sooner if it is a bug fix), however if
problems are discovered then the patch may be dropped or reverted.

You may get further e-mails resulting from automated or manual testing
and review of the tree, please engage with people reporting problems and
send followup patches addressing any issues that are reported if needed.

If any updates are required or you are submitting further changes they
should be sent as incremental updates against current git, existing
patches will not be replaced.

Please add any relevant lists and maintainers to the CCs when replying
to this mail.

Thanks,
Mark

>From 676a4e3bab445d53fca4756865e2c0e2a87c38d6 Mon Sep 17 00:00:00 2001
From: Jan Kiszka <[email protected]>
Date: Sat, 8 Jul 2017 10:41:18 +0200
Subject: [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is
created

Avoid hogging chip select GPIOs just because they are listed for the
master. They might be mulitplexed and, if no slave device is attached,
used for different purposes. Moreover, this strategy avoids having to
allocate a cs_gpiods structure.

Tested on the IOT2000 where the second SPI bus is connected to an
Arduino-compatible connector and multiplexed between SPI, GPIO and PWM
usage.

Signed-off-by: Jan Kiszka <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
---
drivers/spi/spi-pxa2xx.c | 59 +++++++++++++++++-------------------------------
1 file changed, 21 insertions(+), 38 deletions(-)

diff --git a/drivers/spi/spi-pxa2xx.c b/drivers/spi/spi-pxa2xx.c
index 38d053682892..be991266a6ce 100644
--- a/drivers/spi/spi-pxa2xx.c
+++ b/drivers/spi/spi-pxa2xx.c
@@ -1213,21 +1213,33 @@ static int setup_cs(struct spi_device *spi, struct chip_data *chip,
struct pxa2xx_spi_chip *chip_info)
{
struct driver_data *drv_data = spi_master_get_devdata(spi->master);
+ struct device *pdev = &drv_data->pdev->dev;
+ struct gpio_desc *gpiod;
int err = 0;
+ int count;

if (chip == NULL)
return 0;

- if (drv_data->cs_gpiods) {
- struct gpio_desc *gpiod;
+ count = gpiod_count(pdev, "cs");
+ if (count > 0) {
+ if (spi->chip_select >= count)
+ return -EINVAL;
+
+ gpiod = gpiod_get_index(pdev, "cs", spi->chip_select,
+ GPIOD_OUT_HIGH);
+ if (IS_ERR(gpiod)) {
+ /* Means use native chip select */
+ if (PTR_ERR(gpiod) == -ENOENT)
+ return 0;

- gpiod = drv_data->cs_gpiods[spi->chip_select];
- if (gpiod) {
- chip->gpio_cs = desc_to_gpio(gpiod);
- chip->gpio_cs_inverted = spi->mode & SPI_CS_HIGH;
- gpiod_set_value(gpiod, chip->gpio_cs_inverted);
+ return PTR_ERR(gpiod);
}

+ chip->gpio_cs = desc_to_gpio(gpiod);
+ chip->gpio_cs_inverted = spi->mode & SPI_CS_HIGH;
+ gpiod_set_value(gpiod, chip->gpio_cs_inverted);
+
return 0;
}

@@ -1415,8 +1427,7 @@ static void cleanup(struct spi_device *spi)
if (!chip)
return;

- if (drv_data->ssp_type != CE4100_SSP && !drv_data->cs_gpiods &&
- gpio_is_valid(chip->gpio_cs))
+ if (drv_data->ssp_type != CE4100_SSP && gpio_is_valid(chip->gpio_cs))
gpio_free(chip->gpio_cs);

kfree(chip);
@@ -1752,38 +1763,10 @@ static int pxa2xx_spi_probe(struct platform_device *pdev)
master->num_chipselect = platform_info->num_chipselect;

count = gpiod_count(&pdev->dev, "cs");
- if (count > 0) {
- int i;
-
+ if (count > 0)
master->num_chipselect = max_t(int, count,
master->num_chipselect);

- drv_data->cs_gpiods = devm_kcalloc(&pdev->dev,
- master->num_chipselect, sizeof(struct gpio_desc *),
- GFP_KERNEL);
- if (!drv_data->cs_gpiods) {
- status = -ENOMEM;
- goto out_error_clock_enabled;
- }
-
- for (i = 0; i < master->num_chipselect; i++) {
- struct gpio_desc *gpiod;
-
- gpiod = devm_gpiod_get_index(dev, "cs", i,
- GPIOD_OUT_HIGH);
- if (IS_ERR(gpiod)) {
- /* Means use native chip select */
- if (PTR_ERR(gpiod) == -ENOENT)
- continue;
-
- status = (int)PTR_ERR(gpiod);
- goto out_error_clock_enabled;
- } else {
- drv_data->cs_gpiods[i] = gpiod;
- }
- }
- }
-
tasklet_init(&drv_data->pump_transfers, pump_transfers,
(unsigned long)drv_data);

--
2.13.2

2017-07-22 22:07:13

[permalink] [raw]

Subject: Re: [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is created

2017-07-24 10:45:01

[permalink] [raw]

Subject: Re: [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is created

+Cc: Mika

On Sat, Jul 8, 2017 at 11:41 AM, Jan Kiszka <[email protected]> wrote:
> From: Jan Kiszka <[email protected]>
>
> Avoid hogging chip select GPIOs just because they are listed for the
> master. They might be mulitplexed and, if no slave device is attached,
> used for different purposes. Moreover, this strategy avoids having to
> allocate a cs_gpiods structure.
>
> Tested on the IOT2000 where the second SPI bus is connected to an
> Arduino-compatible connector and multiplexed between SPI, GPIO and PWM
> usage.
>

This breaks all systems which are using _DSD.

While I'm looking for fix, I get feeling that the approach itself is not right,

So, for now I would vote for immediate revert and then rethink what we
can do here.

> Signed-off-by: Jan Kiszka <[email protected]>
> ---
> drivers/spi/spi-pxa2xx.c | 59 +++++++++++++++++-------------------------------
> 1 file changed, 21 insertions(+), 38 deletions(-)
>
> diff --git a/drivers/spi/spi-pxa2xx.c b/drivers/spi/spi-pxa2xx.c
> index 38d053682892..be991266a6ce 100644
> --- a/drivers/spi/spi-pxa2xx.c
> +++ b/drivers/spi/spi-pxa2xx.c
> @@ -1213,21 +1213,33 @@ static int setup_cs(struct spi_device *spi, struct chip_data *chip,
> struct pxa2xx_spi_chip *chip_info)
> {
> struct driver_data *drv_data = spi_master_get_devdata(spi->master);
> + struct device *pdev = &drv_data->pdev->dev;
> + struct gpio_desc *gpiod;
> int err = 0;
> + int count;
>
> if (chip == NULL)
> return 0;
>
> - if (drv_data->cs_gpiods) {
> - struct gpio_desc *gpiod;
> + count = gpiod_count(pdev, "cs");
> + if (count > 0) {
> + if (spi->chip_select >= count)
> + return -EINVAL;
> +
> + gpiod = gpiod_get_index(pdev, "cs", spi->chip_select,
> + GPIOD_OUT_HIGH);
> + if (IS_ERR(gpiod)) {
> + /* Means use native chip select */
> + if (PTR_ERR(gpiod) == -ENOENT)
> + return 0;
>
> - gpiod = drv_data->cs_gpiods[spi->chip_select];
> - if (gpiod) {
> - chip->gpio_cs = desc_to_gpio(gpiod);
> - chip->gpio_cs_inverted = spi->mode & SPI_CS_HIGH;
> - gpiod_set_value(gpiod, chip->gpio_cs_inverted);
> + return PTR_ERR(gpiod);
> }
>
> + chip->gpio_cs = desc_to_gpio(gpiod);
> + chip->gpio_cs_inverted = spi->mode & SPI_CS_HIGH;
> + gpiod_set_value(gpiod, chip->gpio_cs_inverted);
> +
> return 0;
> }
>
> @@ -1415,8 +1427,7 @@ static void cleanup(struct spi_device *spi)
> if (!chip)
> return;
>
> - if (drv_data->ssp_type != CE4100_SSP && !drv_data->cs_gpiods &&
> - gpio_is_valid(chip->gpio_cs))
> + if (drv_data->ssp_type != CE4100_SSP && gpio_is_valid(chip->gpio_cs))
> gpio_free(chip->gpio_cs);
>
> kfree(chip);
> @@ -1752,38 +1763,10 @@ static int pxa2xx_spi_probe(struct platform_device *pdev)
> master->num_chipselect = platform_info->num_chipselect;
>
> count = gpiod_count(&pdev->dev, "cs");
> - if (count > 0) {
> - int i;
> -
> + if (count > 0)
> master->num_chipselect = max_t(int, count,
> master->num_chipselect);
>
> - drv_data->cs_gpiods = devm_kcalloc(&pdev->dev,
> - master->num_chipselect, sizeof(struct gpio_desc *),
> - GFP_KERNEL);
> - if (!drv_data->cs_gpiods) {
> - status = -ENOMEM;
> - goto out_error_clock_enabled;
> - }
> -
> - for (i = 0; i < master->num_chipselect; i++) {
> - struct gpio_desc *gpiod;
> -
> - gpiod = devm_gpiod_get_index(dev, "cs", i,
> - GPIOD_OUT_HIGH);
> - if (IS_ERR(gpiod)) {
> - /* Means use native chip select */
> - if (PTR_ERR(gpiod) == -ENOENT)
> - continue;
> -
> - status = (int)PTR_ERR(gpiod);
> - goto out_error_clock_enabled;
> - } else {
> - drv_data->cs_gpiods[i] = gpiod;
> - }
> - }
> - }
> -
> tasklet_init(&drv_data->pump_transfers, pump_transfers,
> (unsigned long)drv_data);
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-spi" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html

--
With Best Regards,
Andy Shevchenko

2017-07-24 10:53:45

[permalink] [raw]

Subject: Re: [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is created

On 2017-07-24 12:44, Andy Shevchenko wrote:
> +Cc: Mika
>
> On Sat, Jul 8, 2017 at 11:41 AM, Jan Kiszka <[email protected]> wrote:
>> From: Jan Kiszka <[email protected]>
>>
>> Avoid hogging chip select GPIOs just because they are listed for the
>> master. They might be mulitplexed and, if no slave device is attached,
>> used for different purposes. Moreover, this strategy avoids having to
>> allocate a cs_gpiods structure.
>>
>> Tested on the IOT2000 where the second SPI bus is connected to an
>> Arduino-compatible connector and multiplexed between SPI, GPIO and PWM
>> usage.
>>
>
> This breaks all systems which are using _DSD.

Err, can you elaborate? Worked fine here with _DSD on the IOT2000.

>
> While I'm looking for fix, I get feeling that the approach itself is not right,
>
> So, for now I would vote for immediate revert and then rethink what we
> can do here.

I'm fine with reverting because the patch wasn't clean anyway (mixed old
and new GPIO API) - aside from whatever you found in addition. I had an
update pending but, as you are looking into this anyway, I'm sure your
patches will be more holistic.

Thanks,
Jan

Attachments:

2017-07-24 11:03:07

[permalink] [raw]

Subject: Re: [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is created

On Mon, Jul 24, 2017 at 1:53 PM, Jan Kiszka <[email protected]> wrote:
> On 2017-07-24 12:44, Andy Shevchenko wrote:
>> +Cc: Mika
>>
>> On Sat, Jul 8, 2017 at 11:41 AM, Jan Kiszka <[email protected]> wrote:
>>> From: Jan Kiszka <[email protected]>
>>>
>>> Avoid hogging chip select GPIOs just because they are listed for the
>>> master. They might be mulitplexed and, if no slave device is attached,
>>> used for different purposes. Moreover, this strategy avoids having to
>>> allocate a cs_gpiods structure.
>>>
>>> Tested on the IOT2000 where the second SPI bus is connected to an
>>> Arduino-compatible connector and multiplexed between SPI, GPIO and PWM
>>> usage.

>> This breaks all systems which are using _DSD.
>
> Err, can you elaborate? Worked fine here with _DSD on the IOT2000.

Sure, the setup() function can be called several times for the same
chip (as written in the comment inside the function).
Definitely your code doesn't follow this, since gpiod_get_index() is
returning -EBUSY when called 2+ time, that's what I got on all my
tests.

>> While I'm looking for fix, I get feeling that the approach itself is not right,
>>
>> So, for now I would vote for immediate revert and then rethink what we
>> can do here.
>
> I'm fine with reverting because the patch wasn't clean anyway (mixed old
> and new GPIO API) - aside from whatever you found in addition.

> I had an
> update pending but, as you are looking into this anyway, I'm sure your
> patches will be more holistic.

Please, send it as RFC, because it might have something we can use/re-use.

--
With Best Regards,
Andy Shevchenko

2017-07-24 11:07:33

[permalink] [raw]

Subject: Re: [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is created

On 2017-07-24 13:02, Andy Shevchenko wrote:
> On Mon, Jul 24, 2017 at 1:53 PM, Jan Kiszka <[email protected]> wrote:
>> On 2017-07-24 12:44, Andy Shevchenko wrote:
>>> +Cc: Mika
>>>
>>> On Sat, Jul 8, 2017 at 11:41 AM, Jan Kiszka <[email protected]> wrote:
>>>> From: Jan Kiszka <[email protected]>
>>>>
>>>> Avoid hogging chip select GPIOs just because they are listed for the
>>>> master. They might be mulitplexed and, if no slave device is attached,
>>>> used for different purposes. Moreover, this strategy avoids having to
>>>> allocate a cs_gpiods structure.
>>>>
>>>> Tested on the IOT2000 where the second SPI bus is connected to an
>>>> Arduino-compatible connector and multiplexed between SPI, GPIO and PWM
>>>> usage.
>
>>> This breaks all systems which are using _DSD.
>>
>> Err, can you elaborate? Worked fine here with _DSD on the IOT2000.
>
> Sure, the setup() function can be called several times for the same
> chip (as written in the comment inside the function).
> Definitely your code doesn't follow this, since gpiod_get_index() is
> returning -EBUSY when called 2+ time, that's what I got on all my
> tests.

Ah, multiple devices on the same controller - I only had one.

>
>>> While I'm looking for fix, I get feeling that the approach itself is not right,
>>>
>>> So, for now I would vote for immediate revert and then rethink what we
>>> can do here.
>>
>> I'm fine with reverting because the patch wasn't clean anyway (mixed old
>> and new GPIO API) - aside from whatever you found in addition.
>
>> I had an
>> update pending but, as you are looking into this anyway, I'm sure your
>> patches will be more holistic.
>
> Please, send it as RFC, because it might have something we can use/re-use.
>

OK, will dig them out later.

Jan

Attachments:

2017-07-24 11:14:57

[permalink] [raw]

Subject: Re: [PATCH] spi: pxa2xx: Only claim CS GPIOs when the slave device is created

On Mon, Jul 24, 2017 at 2:06 PM, Jan Kiszka <[email protected]> wrote:
> On 2017-07-24 13:02, Andy Shevchenko wrote:
>> On Mon, Jul 24, 2017 at 1:53 PM, Jan Kiszka <[email protected]> wrote:
>>> On 2017-07-24 12:44, Andy Shevchenko wrote:
>>>> +Cc: Mika
>>>>
>>>> On Sat, Jul 8, 2017 at 11:41 AM, Jan Kiszka <[email protected]> wrote:
>>>>> From: Jan Kiszka <[email protected]>
>>>>>
>>>>> Avoid hogging chip select GPIOs just because they are listed for the
>>>>> master. They might be mulitplexed and, if no slave device is attached,
>>>>> used for different purposes. Moreover, this strategy avoids having to
>>>>> allocate a cs_gpiods structure.
>>>>>
>>>>> Tested on the IOT2000 where the second SPI bus is connected to an
>>>>> Arduino-compatible connector and multiplexed between SPI, GPIO and PWM
>>>>> usage.
>>
>>>> This breaks all systems which are using _DSD.
>>>
>>> Err, can you elaborate? Worked fine here with _DSD on the IOT2000.
>>
>> Sure, the setup() function can be called several times for the same
>> chip (as written in the comment inside the function).
>> Definitely your code doesn't follow this, since gpiod_get_index() is
>> returning -EBUSY when called 2+ time, that's what I got on all my
>> tests.
>
> Ah, multiple devices on the same controller - I only had one.

Nope, one. Since we are talking about recurrent call to
gpiod_get_index() with the same index.

--
With Best Regards,
Andy Shevchenko

2017-07-24 13:03:51