Hello Naveen,
On 06/11/2014 01:38 PM, Naveen Krishna Ch wrote:
> Hello Javier,
>
> On 11 June 2014 16:43, Javier Martinez Canillas
> <javier.martinez@xxxxxxxxxxxxxxx> wrote:
>> Hello Naveen,
>>
>> Thanks a lot for your patches and sorry that I didn't review your prior two
>> versions but I didn't have the time yesterday.
>>
>> On 06/11/2014 08:31 AM, Naveen Krishna Chatradhi wrote:
>>> Currently, spi-s3c64xx.c needs "cs-gpio" chip select GPIO to be
>>> defined under "controller-data" node under each slave node.
>>>
>>
>> I think that the commit message is not clear enough about the intentions behind
>> your patch.
>>
>> It's not only that spi-s3c64xx needs a cs-gpio chip to be defined under
>> controller-data dev node while all other SPI drivers expects cs-gpios at the top
>> level, but more important that currently s3c64xx driver expects to have both
>> cs-gpio (singular) at the top level *and* cs-gpio in controller data which
>> doesn't make too much sense.
>>
>>> &spi_x {
>>> cs-gpios <>;
>>> ...
>>
>> As I said, currently it expects cs-gpio (singular) not cs-gpios (plural) as your
>> example. It's important to have a correct commit message so future code
>> archaeologists can have a proper picture of the situation if needed.
> Will add some explanation in the example, for it to be clear.
>>
>>> slave_node {
>>>
>>> controller-data {
>>> cs-gpio = <>;
>>> ...
>>> };
>>> ...
>>> };
>>> ...
>>> };
>>>
>>> Where as, SPI core and many other drivers uses "cs-gpios" for
>>> from device tree node.
>>>
>>> Hence, make changes in spi-s3c64xx.c driver to make use of
>>> "cs-gpios" from SPI node(parent) instead of "cs-gpio" defined in
>>> slaves "controller-data"(child) node.
>>>
>>
>> So, the problem is not that the binding is not consistent with other SPI drivers
>> (that would have been bad but acceptable IMHO) but that it is completely broken.
>> And since we have to fix it which means breaking the ABI anyways, it is better
>> to make it consistent with other drivers and SPI core.
> Will take this point for the while rewriting the commit messages
>>
>>> Also updates the Device tree Documentation.
>>>
>>
>> While I agree with others that a binding that has been broken for a year is a
>> binding that appears to not be used a lot, we should be more vocal about this
>> and why breaking backward compatibility is the right approach in this case.
>>
>> So, in the (theoretical?) case that users find that their platform stopped to
>> work with their current FDT, it will be easy for them to figure out what commit
>> break it, what was the motivation to break the ABI and what changes are needed
>> on their DTS to make it work again.
>>
>> Adding the sha1 of the culprit commit (3146bee spi: s3c64xx: Added provision for
>> dedicated cs pin) that puts us in this situation will also be useful. Since the
>> date of that commit is part of the rationale behind this change. (e.g: nobody
>> cared about this binding and so can be changed).
> Will take this point for the while rewriting the commit messages
Awesome, thanks.
>>
>>> Signed-off-by: Naveen Krishna Chatradhi <ch.naveen@xxxxxxxxxxx>
>>> Acked-by: Rob Herring <robh@xxxxxxxxxx>
>>> Cc: Javier Martinez Canillas <javier.martinez@xxxxxxxxxxxxxxx>
>>> Cc: Doug Anderson <dianders@xxxxxxxxxxxx>
>>> Cc: Tomasz Figa <t.figa@xxxxxxxxxxx>
>>> ---
>>> Changes since v2:
>>> 1. updated the gpios usage in Documentation
>>> 2. use the spi->cs_gpio in the driver, instead of parsing the node again.
>>> 3. Corrected error check of the of.node and during gpio_free
>>>
>>> .../devicetree/bindings/spi/spi-samsung.txt | 8 +++-----
>>> drivers/spi/spi-s3c64xx.c | 18 ++++++------------
>>> 2 files changed, 9 insertions(+), 17 deletions(-)
>>>
>>> diff --git a/Documentation/devicetree/bindings/spi/spi-samsung.txt b/Documentation/devicetree/bindings/spi/spi-samsung.txt
>>> index 86aa061..2d29dac 100644
>>> --- a/Documentation/devicetree/bindings/spi/spi-samsung.txt
>>> +++ b/Documentation/devicetree/bindings/spi/spi-samsung.txt
>>> @@ -42,15 +42,13 @@ Optional Board Specific Properties:
>>> - num-cs: Specifies the number of chip select lines supported. If
>>> not specified, the default number of chip select lines is set to 1.
>>>
>>> +- cs-gpios: should specify GPIOs used for chipselects (see spi-bus.txt)
>>> +
>>> SPI Controller specific data in SPI slave nodes:
>>>
>>> - The spi slave nodes should provide the following information which is required
>>> by the spi controller.
>>>
>>> - - cs-gpio: A gpio specifier that specifies the gpio line used as
>>> - the slave select line by the spi controller. The format of the gpio
>>> - specifier depends on the gpio controller.
>>> -
>>> - samsung,spi-feedback-delay: The sampling phase shift to be applied on the
>>> miso line (to account for any lag in the miso line). The following are the
>>> valid values.
>>> @@ -85,6 +83,7 @@ Example:
>>> #size-cells = <0>;
>>> pinctrl-names = "default";
>>> pinctrl-0 = <&spi0_bus>;
>>> + cs-gpios = <&gpa2 5 0>;
>>>
>>> w25q80bw@0 {
>>> #address-cells = <1>;
>>> @@ -94,7 +93,6 @@ Example:
>>> spi-max-frequency = <10000>;
>>>
>>> controller-data {
>>> - cs-gpio = <&gpa2 5 1 0 3>;
>>> samsung,spi-feedback-delay = <0>;
>>> };
>>>
>>> diff --git a/drivers/spi/spi-s3c64xx.c b/drivers/spi/spi-s3c64xx.c
>>> index 75a5696..842b148 100644
>>> --- a/drivers/spi/spi-s3c64xx.c
>>> +++ b/drivers/spi/spi-s3c64xx.c
>>> @@ -772,24 +772,19 @@ static struct s3c64xx_spi_csinfo *s3c64xx_get_slave_ctrldata(
>>>
>>> cs = kzalloc(sizeof(*cs), GFP_KERNEL);
>>> if (!cs) {
>>> - of_node_put(data_np);
>>
>> Why are you removing this?
>>
>> You are not removing the call to of_get_child_by_name(slave_np,
>> "controller-data") so on failure data_np reference counter still needs to be
>> decremented.
>>
>>> return ERR_PTR(-ENOMEM);
>>> }
>>>
>>> - /* The CS line is asserted/deasserted by the gpio pin */
>>> - if (sdd->cs_gpio)
>>> - cs->line = of_get_named_gpio(data_np, "cs-gpio", 0);
>>> -
>>> - if (!gpio_is_valid(cs->line)) {
>>> + if (!gpio_is_valid(spi->cs_gpio)) {
>>> dev_err(&spi->dev, "chip select gpio is not specified or invalid\n");
>>> - kfree(cs);
>>> - of_node_put(data_np);
>>
>> Again, it's wrong to remove this of_node_put(data_np) call.
> Actually, i wanted to move down the of_get_child_by_name() to before
> of_property_read_u32() call.
Yeah, the wrong removals seemed to me like a left over from a previous attempt
to reorganize the code.
>>
>>> return ERR_PTR(-EINVAL);
>>> }
>>> + cs->line = spi->cs_gpio;
>>>
>>
>> I wonder why are you keeping cs->line? AFAICT it's only used in
>> s3c64xx_spi_setup() to request the GPIO and since you get the struct spi_device
>> pointer as a parameter then you can just use spi->s_gpio instead.
> I'm trying not to touch the non-DT part of the code.
>
> struct s3c64xx_spi_csinfo *cs = spi->controller_data;
>
> This will update the cs->line and cs->fb_delay in case of non-DT.
I see, then I prefer the opposite and do something like this on s3c64xx_spi_probe():
if (!pdev->dev.of_node)
spi->cs_gpio = cs->line;
So, cs->line is only used by non-DT platforms to pass the GPIO to be used as
chip-select line. But after that use spi->cs_gpio should be used in every place.
If someday all the non-DT platforms are migrated to DT then only that assignment
has to be removed instead of several lines of code using cs->line currently.
Also is more consistent to use spi->cs_gpio like most SPI drivers do instead of
a custom structure member.
>>
>>> of_property_read_u32(data_np, "samsung,spi-feedback-delay", &fb_delay);
>>> cs->fb_delay = fb_delay;
>>> of_node_put(data_np);
>>> +
>>
>> Usually is better to not mix style and functionality changes in the same patch
>> since it makes harder to review.
>>
>>> return cs;
>>> }
>>>
>>> @@ -828,8 +823,6 @@ static int s3c64xx_spi_setup(struct spi_device *spi)
>>> cs->line, err);
>>> goto err_gpio_req;
>>> }
>>> -
>>> - spi->cs_gpio = cs->line;
>>> }
>>>
>>> spi_set_ctldata(spi, cs);
>>> @@ -884,7 +877,8 @@ setup_exit:
>>> /* setup() returns with device de-selected */
>>> writel(S3C64XX_SPI_SLAVE_SIG_INACT, sdd->regs + S3C64XX_SPI_SLAVE_SEL);
>>>
>>> - gpio_free(cs->line);
>>> + if (cs->line)
>>> + gpio_free(cs->line);
>>
>> Same here, we can just get rid of cs->line and use gpio_free(spi->cs_gpio) instead.
> This is to keep the non-DT part untouched.
I don't get why you don't want to change legacy platform code if that means
improving the driver. If you don't feel confident with your changes because you
don't have access to the non-DT platform to test then you can post your patches
as RFC or ask people to with access to those platforms to test and collect some
Tested-by tags before the patch is merged.
>>
>>> spi_set_ctldata(spi, NULL);
>>>
>>> err_gpio_req:
>>> @@ -1077,7 +1071,7 @@ static int s3c64xx_spi_probe(struct platform_device *pdev)
>>> sdd->sfr_start = mem_res->start;
>>> sdd->cs_gpio = true;
>>> if (pdev->dev.of_node) {
>>> - if (!of_find_property(pdev->dev.of_node, "cs-gpio", NULL))
>>> + if (!of_find_property(pdev->dev.of_node, "cs-gpios", NULL))
>>> sdd->cs_gpio = false;
>>
>> I think we can get rid of sdd->cs_gpio as well. You can always use
>> gpio_is_valid(spi->cs_gpio) or if (master->cs_gpios) to figure out if the SPI
>> device has a cs-gpios array associated.
> This is to keep the non-DT part untouched
sdd->cs_gpio has nothing to do with non-DT platforms AFAIU. It was introduced in
the commit that initially broke this binding and it's redundant information.
It's assigned to true on s3c64xx_spi_probe() and then set to false only if
of_find_property(pdev->dev.of_node, "cs-gpios", NULL) fails.
But you don't need an additional variable to get that info since
gpio_is_valid(spi->cs_gpio) or if (master->cs_gpios) can tell you the same AFAIK.
>>
>>>
>>> ret = of_alias_get_id(pdev->dev.of_node, "spi");
>>>
>>
>> Best regards,
>> Javier
> Will re-spin after waiting for a couple of more comments.
Perfect, thanks a lot!
> Thanks for your review.
>
>
Best regards,
Javier
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
