Hi Bin,
On Thu, Sep 26, 2019 at 7:26 AM Bin Meng bmeng.cn@gmail.com wrote:
Hi Sagar,
On Thu, Sep 26, 2019 at 1:54 AM Sagar Kadam sagar.kadam@sifive.com wrote:
Hi Bin,
On Wed, Sep 18, 2019 at 1:23 PM Bin Meng bmeng.cn@gmail.com wrote:
Hi Sagar,
On Tue, Sep 10, 2019 at 11:44 PM Sagar Shrikant Kadam sagar.kadam@sifive.com wrote:
U-Boot currently is missing GPIO support for FU540-C000 SoC which is mounted on HiFive Unleashed A00 board. This patch is intended to add DM based GPIO controller driver in order to access GPIO pins within the SoC using GPIO command in U-Boot. More details on the GPIO controller within the SoC can be found at[1]
The driver is based above master branch of u-boot-riscv.git and provides a method to configure Input/Output mode of the GPIO pin along with an option to set or clear state of the GPIO pin. The patch is available in dev/sagark/gpio_v3 branch here[2].
GPIO device node added to the mainline bound device tree for HiFive Unleashed is available in dev/sagark/mlv5.3-rc5 branch of repo here[3].
This implementation is ported from linux driver submitted for review at [4].
More details of GPIO pin routing on J1 header is available in schematic document[5]
[1] https://static.dev.sifive.com/FU540-C000-v1.0.pdf [2] https://github.com/sagsifive/u-boot [3] https://github.com/sagsifive/riscv-linux-hifive/ [4] https://lkml.org/lkml/2018/10/9/1103 [5] https://static.dev.sifive.com/dev-kits/hifive-unleashed/hifive-unleashed-a00...
Driver Testing: #Set GPIO1 high. =>gpio set 1 Can be confirmed by probing pin No #24 on J1 Header or memory dump of gpio register space viz: #md 0x10060000
#Set GPIO1 low =>gpio clear 0
#Toggle GPIO1 =>gpio toggle 1 #Toggle value of GPIO1 =>gpio toggle 1 #Toggle value of GPIO1
#Configure pin as input =>gpio input 3 #Configure gpio line 3 as input.
#Error check =>gpio set 16 #Not a valid GPIO number for FU540-C000 GPIO: '16' not found Command 'gpio' failed: Error -22
I tested this:
=> gpio status -a Bank gpio@10060000: gpio@100600000: unknown gpio@100600001: unknown gpio@100600002: unknown gpio@100600003: unknown gpio@100600004: unknown gpio@100600005: unknown gpio@100600006: unknown gpio@100600007: unknown gpio@100600008: unknown gpio@100600009: unknown gpio@1006000010: unknown gpio@1006000011: unknown gpio@1006000012: unknown gpio@1006000013: unknown gpio@1006000014: unknown gpio@1006000015: unknown
The status is "unknown" for all gpio pins, which is wrong. It should be either input or output.
Thank you for your suggestions. The get_function operation is missing for this driver and so the status is unknown. I will implement it and send a revised version. Thanks for catching this. Please correct me if I am wrong, what I do see is that the gpio command uses the bank name appended before the GPIO number. So the bank_name as assigned in the driver probe function gets prefixed to the pin number and so it shows: gpio@100600000 gpio@100600001 and so on. I see that few driver's updates the uc_priv->bank_name in probe function with '_' as the separator between bank_name and pin number and so #gpio status -a will show it as :
Bank <bank_name>: <bank_name>_0: input : 1 [] <bank_name>_1: input : 1 [] ........ and so on
eg: In the current case here it will show as Bank gpio@10060000_: gpio@10060000_0 gpio@10060000_1 and so on.
Please let me know if this implementation is ok.
Also the gpio pin name is weird. I think we should use "0, 1, 2 ..."
The current implementation of the gpio_get_status function includes the base_name to the pin description. Truncating it here can help to get pin numbers as just numbers "0,1,2....". I will also include this if needed?
I think <bank_name>_0 is fine. Thanks!
Thanks, I will roll out the next patch with necessary changes.
BR, Sagar
Regards, Bin