From: Martin Hundebøll <[email protected]>
The Max10 BMC on the Silicom n5010 PAC is slightly different than the
existing BMCs, so use a dedicated feature revision detect it.
Signed-off-by: Martin Hundebøll <[email protected]>
Reviewed-by: Moritz Fischer <[email protected]>
---
Changes since v3:
* Changed "BMC's" to "BMCs"
* Added Moritz' Reviewed-by
Changes since v2:
* None
Changes since v1:
* use feature revision from struct dfl_device instead of reading it
from io-mem
drivers/spi/spi-altera-dfl.c | 15 +++++++++++++--
1 file changed, 13 insertions(+), 2 deletions(-)
diff --git a/drivers/spi/spi-altera-dfl.c b/drivers/spi/spi-altera-dfl.c
index 3e32e4fe5895..f6cf7c8d9dac 100644
--- a/drivers/spi/spi-altera-dfl.c
+++ b/drivers/spi/spi-altera-dfl.c
@@ -111,6 +111,13 @@ static struct spi_board_info m10_bmc_info = {
.chip_select = 0,
};
+static struct spi_board_info m10_n5010_bmc_info = {
+ .modalias = "m10-n5010",
+ .max_speed_hz = 12500000,
+ .bus_num = 0,
+ .chip_select = 0,
+};
+
static void config_spi_master(void __iomem *base, struct spi_master *master)
{
u64 v;
@@ -130,6 +137,7 @@ static void config_spi_master(void __iomem *base, struct spi_master *master)
static int dfl_spi_altera_probe(struct dfl_device *dfl_dev)
{
+ struct spi_board_info *board_info = &m10_bmc_info;
struct device *dev = &dfl_dev->dev;
struct spi_master *master;
struct altera_spi *hw;
@@ -172,9 +180,12 @@ static int dfl_spi_altera_probe(struct dfl_device *dfl_dev)
goto exit;
}
- if (!spi_new_device(master, &m10_bmc_info)) {
+ if (dfl_dev->revision == FME_FEATURE_REV_MAX10_SPI_N5010)
+ board_info = &m10_n5010_bmc_info;
+
+ if (!spi_new_device(master, board_info)) {
dev_err(dev, "%s failed to create SPI device: %s\n",
- __func__, m10_bmc_info.modalias);
+ __func__, board_info->modalias);
}
return 0;
--
2.31.0
On 7/5/21 3:16 AM, Martin Hundebøll wrote:
> From: Martin Hundebøll <[email protected]>
>
> The Max10 BMC on the Silicom n5010 PAC is slightly different than the
> existing BMCs, so use a dedicated feature revision detect it.
>
> Signed-off-by: Martin Hundebøll <[email protected]>
> Reviewed-by: Moritz Fischer <[email protected]>
> ---
>
> Changes since v3:
> * Changed "BMC's" to "BMCs"
> * Added Moritz' Reviewed-by
>
> Changes since v2:
> * None
>
> Changes since v1:
> * use feature revision from struct dfl_device instead of reading it
> from io-mem
>
> drivers/spi/spi-altera-dfl.c | 15 +++++++++++++--
> 1 file changed, 13 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/spi/spi-altera-dfl.c b/drivers/spi/spi-altera-dfl.c
> index 3e32e4fe5895..f6cf7c8d9dac 100644
> --- a/drivers/spi/spi-altera-dfl.c
> +++ b/drivers/spi/spi-altera-dfl.c
> @@ -111,6 +111,13 @@ static struct spi_board_info m10_bmc_info = {
> .chip_select = 0,
> };
>
> +static struct spi_board_info m10_n5010_bmc_info = {
> + .modalias = "m10-n5010",
> + .max_speed_hz = 12500000,
> + .bus_num = 0,
> + .chip_select = 0,
> +};
Other then the modalias, this is exactly the same as m10_bmc_info.
Why not set platform_data?
> +
> static void config_spi_master(void __iomem *base, struct spi_master *master)
> {
> u64 v;
> @@ -130,6 +137,7 @@ static void config_spi_master(void __iomem *base, struct spi_master *master)
>
> static int dfl_spi_altera_probe(struct dfl_device *dfl_dev)
> {
> + struct spi_board_info *board_info = &m10_bmc_info;
> struct device *dev = &dfl_dev->dev;
> struct spi_master *master;
> struct altera_spi *hw;
> @@ -172,9 +180,12 @@ static int dfl_spi_altera_probe(struct dfl_device *dfl_dev)
> goto exit;
> }
>
> - if (!spi_new_device(master, &m10_bmc_info)) {
> + if (dfl_dev->revision == FME_FEATURE_REV_MAX10_SPI_N5010)
> + board_info = &m10_n5010_bmc_info;
The revision is board parameter, I think this check could be improved.
There should be a
#define FME_FATURE_REV_MAX10_SPI_D5005 0
And it checked here instead of setting above.
And -EINVAL returned if the revision is not known.
> +
> + if (!spi_new_device(master, board_info)) {
> dev_err(dev, "%s failed to create SPI device: %s\n",
> - __func__, m10_bmc_info.modalias);
> + __func__, board_info->modalias);
Why isn't this error handled ?
Tom
> }
>
> return 0;
On 06/07/2021 16.56, Tom Rix wrote:
>> diff --git a/drivers/spi/spi-altera-dfl.c b/drivers/spi/spi-altera-dfl.c
>> index 3e32e4fe5895..f6cf7c8d9dac 100644
>> --- a/drivers/spi/spi-altera-dfl.c
>> +++ b/drivers/spi/spi-altera-dfl.c
>> @@ -111,6 +111,13 @@ static struct spi_board_info m10_bmc_info = {
>> .chip_select = 0,
>> };
>> +static struct spi_board_info m10_n5010_bmc_info = {
>> + .modalias = "m10-n5010",
>> + .max_speed_hz = 12500000,
>> + .bus_num = 0,
>> + .chip_select = 0,
>> +};
>
> Other then the modalias, this is exactly the same as m10_bmc_info.
>
> Why not set platform_data?
So like this?
+static struct spi_board_info m10_n5010_bmc_info = {
+ .platform_data = "m10-n5010",
+ .max_speed_hz = 12500000,
+ .bus_num = 0,
+ .chip_select = 0,
+};
I don't see how that should improve the situation, but we might allocate
the board info on the stack and set modalias dynamically instead?
// Martin
On 7/14/21 4:33 AM, Martin Hundebøll wrote:
>
>
> On 06/07/2021 16.56, Tom Rix wrote:
>>> diff --git a/drivers/spi/spi-altera-dfl.c
>>> b/drivers/spi/spi-altera-dfl.c
>>> index 3e32e4fe5895..f6cf7c8d9dac 100644
>>> --- a/drivers/spi/spi-altera-dfl.c
>>> +++ b/drivers/spi/spi-altera-dfl.c
>>> @@ -111,6 +111,13 @@ static struct spi_board_info m10_bmc_info = {
>>> .chip_select = 0,
>>> };
>>> +static struct spi_board_info m10_n5010_bmc_info = {
>>> + .modalias = "m10-n5010",
>>> + .max_speed_hz = 12500000,
>>> + .bus_num = 0,
>>> + .chip_select = 0,
>>> +};
>>
>> Other then the modalias, this is exactly the same as m10_bmc_info.
>>
>> Why not set platform_data?
>
> So like this?
>
> +static struct spi_board_info m10_n5010_bmc_info = {
> + .platform_data = "m10-n5010",
> + .max_speed_hz = 12500000,
> + .bus_num = 0,
> + .chip_select = 0,
> +};
>
> I don't see how that should improve the situation, but we might allocate
> the board info on the stack and set modalias dynamically instead?
No, I mean that instead of have two *bmc_info's generalize the existing one.
This could be done by using the as yet unused platform_data field to
hold the identity as a bit/enum in an int.
Tom
>
> // Martin
>
On 14/07/2021 16.22, Tom Rix wrote:
>
> On 7/14/21 4:33 AM, Martin Hundebøll wrote:
>>
>>
>> On 06/07/2021 16.56, Tom Rix wrote:
>>>> diff --git a/drivers/spi/spi-altera-dfl.c b/drivers/spi/spi-altera-dfl.c
>>>> index 3e32e4fe5895..f6cf7c8d9dac 100644
>>>> --- a/drivers/spi/spi-altera-dfl.c
>>>> +++ b/drivers/spi/spi-altera-dfl.c
>>>> @@ -111,6 +111,13 @@ static struct spi_board_info m10_bmc_info = {
>>>> .chip_select = 0,
>>>> };
>>>> +static struct spi_board_info m10_n5010_bmc_info = {
>>>> + .modalias = "m10-n5010",
>>>> + .max_speed_hz = 12500000,
>>>> + .bus_num = 0,
>>>> + .chip_select = 0,
>>>> +};
>>>
>>> Other then the modalias, this is exactly the same as m10_bmc_info.
>>>
>>> Why not set platform_data?
>>
>> So like this?
>>
>> +static struct spi_board_info m10_n5010_bmc_info = {
>> + .platform_data = "m10-n5010",
>> + .max_speed_hz = 12500000,
>> + .bus_num = 0,
>> + .chip_select = 0,
>> +};
>>
>> I don't see how that should improve the situation, but we might allocate
>> the board info on the stack and set modalias dynamically instead?
>
> No, I mean that instead of have two *bmc_info's generalize the existing one.
>
> This could be done by using the as yet unused platform_data field to hold the identity as a bit/enum in an int.
But the existing one is global/static, so it would need
to be moved to the function/stack, yes?
In which case we could as easily keep the existing approach
of modalias, and avoid changing both drivers/fpga/dfl-n3000-nios.c
and drivers/mfd/intel-m10-bmc.c ...
// Martin