| Message ID | 1394125539-28062-1-git-send-email-ckeepax@opensource.wolfsonmicro.com |
|---|---|
| State | Superseded, archived |
| Headers | show |
On Thu, Mar 06, 2014 at 05:05:39PM +0000, Charles Keepax wrote: > The following patch added support for spi controllers with a dedicated > chip select pin: > > commit 3146beec21b64f4551fcf0ac148381d54dc41b1b > spi: s3c64xx: Added provision for dedicated cs pin > > It updated the device tree binding to require a "cs-gpio" property to be > specified on the spi controller node if chip selects will be given as > GPIOs per slave, rather than the controller having a dedicated internal > chip select pin. No, it doesn't - it's saying that if the device has a "cs-gpio" property then to use that as the chip select. It's not a boolean, it's a GPIO specifier. Looking at the code it looks like the intention is to search all children for a cs-gpio during the controller probe, it's possible that this isn't working correctly. > Apologies if I missed something but there are a couple of > things I am not sure about in the original patch, there > seems to be no special handling for the dedicated case, like > set a bit that enables this or some such, which implies that > parts with dedicated cs pins will always update them even > if cs-gpio is specified. In which case wasn't what existed > before reasonable? The chip select signal within the IP always needs to be manipulated for the hardware to work regardless of a GPIO being used, if a GPIO is being used as the actual chip select then the expecation is that the pinmux will be set up so that the signal isn't actually brought out of the device.
On Fri, Mar 07, 2014 at 10:48:41AM +0800, Mark Brown wrote: > On Thu, Mar 06, 2014 at 05:05:39PM +0000, Charles Keepax wrote: > > > The following patch added support for spi controllers with a dedicated > > chip select pin: > > > > commit 3146beec21b64f4551fcf0ac148381d54dc41b1b > > spi: s3c64xx: Added provision for dedicated cs pin > > > > It updated the device tree binding to require a "cs-gpio" property to be > > specified on the spi controller node if chip selects will be given as > > GPIOs per slave, rather than the controller having a dedicated internal > > chip select pin. > > No, it doesn't - it's saying that if the device has a "cs-gpio" property > then to use that as the chip select. It's not a boolean, it's a GPIO > specifier. Looking at the code it looks like the intention is to search > all children for a cs-gpio during the controller probe, it's possible > that this isn't working correctly. That is basically part of my question is the current setup doing what it is intended to? The Samsung binding has controller-data blocks on each of the slaves that specify the gpio for that slave. @@ -1326,7 +1340,11 @@ static int s3c64xx_spi_probe(struct platform_device *pdev) sdd->cntrlr_info = sci; sdd->pdev = 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)) + sdd->cs_gpio = false; This part of the original patch adds the check, I guess the mistake could be an assumption that of_find_property will recursively check all the children, which it won't. This will just check the controller node itself for a cs-gpio property but as these are specified within sub-nodes they won't be found. Hence currently you need to add one at the controller level. Thanks, Charles -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, Mar 07, 2014 at 09:19:08AM +0000, Charles Keepax wrote: > On Fri, Mar 07, 2014 at 10:48:41AM +0800, Mark Brown wrote: > > On Thu, Mar 06, 2014 at 05:05:39PM +0000, Charles Keepax wrote: > > > It updated the device tree binding to require a "cs-gpio" property to be > > > specified on the spi controller node if chip selects will be given as > > > GPIOs per slave, rather than the controller having a dedicated internal > > > chip select pin. > > No, it doesn't - it's saying that if the device has a "cs-gpio" property > > then to use that as the chip select. It's not a boolean, it's a GPIO > > specifier. Looking at the code it looks like the intention is to search > > all children for a cs-gpio during the controller probe, it's possible > > that this isn't working correctly. > That is basically part of my question is the current setup doing > what it is intended to? The Samsung binding has controller-data > blocks on each of the slaves that specify the gpio for that > slave. Right, which is also clearly the intention of the code.
diff --git a/Documentation/devicetree/bindings/spi/spi-samsung.txt b/Documentation/devicetree/bindings/spi/spi-samsung.txt index 86aa061..2f0167d 100644 --- a/Documentation/devicetree/bindings/spi/spi-samsung.txt +++ b/Documentation/devicetree/bindings/spi/spi-samsung.txt @@ -42,6 +42,11 @@ 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-gpio: Specifies that the spi controller will give a chip select gpio + for each slave node. Absence of this indicates that the controller has a + dedicated chip internal select pin and any GPIOs specified on the slaves + will be ignored. + SPI Controller specific data in SPI slave nodes: - The spi slave nodes should provide the following information which is required @@ -85,6 +90,7 @@ Example: #size-cells = <0>; pinctrl-names = "default"; pinctrl-0 = <&spi0_bus>; + cs-gpio; w25q80bw@0 { #address-cells = <1>;
The following patch added support for spi controllers with a dedicated chip select pin: commit 3146beec21b64f4551fcf0ac148381d54dc41b1b spi: s3c64xx: Added provision for dedicated cs pin It updated the device tree binding to require a "cs-gpio" property to be specified on the spi controller node if chip selects will be given as GPIOs per slave, rather than the controller having a dedicated internal chip select pin. This patch updates the device tree binding documentation to match this. Signed-off-by: Charles Keepax <ckeepax@opensource.wolfsonmicro.com> --- Hi, Apologies if I missed something but there are a couple of things I am not sure about in the original patch, there seems to be no special handling for the dedicated case, like set a bit that enables this or some such, which implies that parts with dedicated cs pins will always update them even if cs-gpio is specified. In which case wasn't what existed before reasonable? If the concern is that someone would try to use both the internal chip select and a GPIO for two slaves on the same controller, causing the first slave to always get chip selected wouldn't a better solution be to require all slaves to give a chip select GPIO if one does? Finally, if we really want to specify in the DT for the controller that it has an internal chip select wouldn't it be better to invert the logic and call it something like "dedicated-cs" as then older bindings that currently use GPIOs wouldn't break. I thought it better to do a patch to update the documentation rather than change the binding again, but personally I would lean on the side of updating the binding if others are in favour of it? Thanks, Charles .../devicetree/bindings/spi/spi-samsung.txt | 6 ++++++ 1 files changed, 6 insertions(+), 0 deletions(-)