On Thu, Mar 17, 2022 at 04:31:15PM +0100, Marek Vasut wrote:
On 3/17/22 16:13, Tommaso Merciai wrote:
On Thu, Mar 17, 2022 at 01:38:18PM +0100, Tommaso Merciai wrote:
On Thu, Mar 17, 2022 at 12:58:31PM +0100, Marek Vasut wrote:
On 3/17/22 08:39, Tommaso Merciai wrote:
On Wed, Mar 16, 2022 at 09:54:34PM +0100, Marek Vasut wrote:
On 3/16/22 16:27, Tommaso Merciai wrote: > Add function enable_pwm_clk into in clock_imx8mm.c. This > function first configure, then enable pwm clock from clock control > register. The following configuration is used: > > source(0) -> 24 MHz ref clock > div(0) -> no division for this clock > > References: > - iMX8MMRM.pdf p 303 > > Signed-off-by: Tommaso Merciai tommaso.merciai@amarulasolutions.com > --- > Changes since v1: > - Fix enable_pwm_clk function implementation. Now is generic for all pwm clks > > arch/arm/mach-imx/imx8m/clock_imx8mm.c | 53 ++++++++++++++++++++++++++ > 1 file changed, 53 insertions(+)
Why is this not in drivers/clk/imx/ DM driver ?
Hi Marek, All function that enable/configure clk from CCGR are in arch/arm/mach-imx/imx8m/clock_imx8mm.c.
These seems to be CCGR:
$ grep -C 2 '0x4[0-9a-f]{3}' drivers/clk/imx/clk-imx8mm.c | sed "s@^.@@"
clk_dm(IMX8MM_CLK_ECSPI1_ROOT, imx_clk_gate4("ecspi1_root_clk", "ecspi1", base + 0x4070, 0)); clk_dm(IMX8MM_CLK_ECSPI2_ROOT, imx_clk_gate4("ecspi2_root_clk", "ecspi2", base + 0x4080, 0)); clk_dm(IMX8MM_CLK_ECSPI3_ROOT, imx_clk_gate4("ecspi3_root_clk", "ecspi3", base + 0x4090, 0)); clk_dm(IMX8MM_CLK_I2C1_ROOT, imx_clk_gate4("i2c1_root_clk", "i2c1", base + 0x4170, 0)); clk_dm(IMX8MM_CLK_I2C2_ROOT, imx_clk_gate4("i2c2_root_clk", "i2c2", base + 0x4180, 0)); clk_dm(IMX8MM_CLK_I2C3_ROOT, imx_clk_gate4("i2c3_root_clk", "i2c3", base + 0x4190, 0)); clk_dm(IMX8MM_CLK_I2C4_ROOT, imx_clk_gate4("i2c4_root_clk", "i2c4", base + 0x41a0, 0)); clk_dm(IMX8MM_CLK_OCOTP_ROOT, imx_clk_gate4("ocotp_root_clk", "ipg_root", base + 0x4220, 0)); clk_dm(IMX8MM_CLK_USDHC1_ROOT, imx_clk_gate4("usdhc1_root_clk", "usdhc1", base + 0x4510, 0)); clk_dm(IMX8MM_CLK_USDHC2_ROOT, imx_clk_gate4("usdhc2_root_clk", "usdhc2", base + 0x4520, 0)); clk_dm(IMX8MM_CLK_WDOG1_ROOT, imx_clk_gate4("wdog1_root_clk", "wdog", base + 0x4530, 0)); clk_dm(IMX8MM_CLK_WDOG2_ROOT, imx_clk_gate4("wdog2_root_clk", "wdog", base + 0x4540, 0)); clk_dm(IMX8MM_CLK_WDOG3_ROOT, imx_clk_gate4("wdog3_root_clk", "wdog", base + 0x4550, 0)); clk_dm(IMX8MM_CLK_USDHC3_ROOT, imx_clk_gate4("usdhc3_root_clk", "usdhc3", base + 0x45e0, 0)); clk_dm(IMX8MM_CLK_QSPI_ROOT, imx_clk_gate4("qspi_root_clk", "qspi", base + 0x42f0, 0)); clk_dm(IMX8MM_CLK_USB1_CTRL_ROOT, imx_clk_gate4("usb1_ctrl_root_clk", "usb_bus", base + 0x44d0, 0));
/* clks not needed in SPL stage */
clk_dm(IMX8MM_CLK_ENET1_ROOT, imx_clk_gate4("enet1_root_clk", "enet_axi", base + 0x40a0, 0)); endif
For that I continue to put here the implementation. After we can port the clk dm part to manipulate clock in drivers/clk/imx/ DM driver. What do you think about? Let me know.
Seems like the clk_dm part is already in place and all you have to do is extend it.
Hi Marek, I test the solution using DM model, it work:
u-boot=> clk dump
24000000 2 | |-- pwm1 24000000 3 | | `-- pwm1_root_clk 24000000 0 | |-- pwm2 24000000 0 | | `-- pwm2_root_clk 24000000 0 | |-- pwm3 24000000 0 | | `-- pwm3_root_clk 24000000 0 | |-- pwm4 24000000 0 | | `-- pwm4_root_clk
I test it using the following call on a dummy driver:
ret = uclass_get_device_by_name(UCLASS_PWM, "pwm@30660000", &pwm); if (ret) printk("Failed to get pwm dev\n");
ret = clk_get_by_name(pwm, "per", &per_clk); if (ret) { printf("Failed to get per_clk\n"); return ret; } ret = clk_enable(&per_clk); if (ret) { printf("Failed to enable per_clk\n"); return ret; }
ret = clk_get_by_name(pwm, "ipg", &ipg_clk); if (ret) { printf("Failed to get ipg_clk\n"); return ret; } ret = clk_enable(&ipg_clk); if (ret) { printf("Failed to enable ipg_clk\n"); return ret; }
It's better to keep both the solutions or only based on DM model? I think we can put this initialization into drivers/pwm/pwm-imx.c imx_pwm_of_to_plat function. What do you think about? Let me know.
DM is the preferred way, non-DM is fading away.
Hi All, I'm working on pwm-imx drv, what do you think about move pwm-imx-utils and put all the 2 function inside pwm-imx.c under an ifndef CONFIG_DM_PWM. Using DM on pwm-imx make pwm-imx-utils.c no sense to exist.
Example:
#ifndef CONFIG_DM_PWM int pwm_imx_get_parms(int period_ns, int duty_ns, unsigned long *period_c, unsigned long *duty_c, unsigned long *prescale){ ....
struct pwm_regs *pwm_id_to_reg(int pwm_id) { ...
+ all non dm implementation
#else
struct imx_pwm_priv { struct pwm_regs *regs; bool invert; struct clk *per_clk; struct clk *ipg_clk; };
int pwm_dm_imx_get_parms(struct imx_pwm_priv *priv, int period_ns, int duty_ns, unsigned long *period_c, unsigned long *duty_c, unsigned long *prescale) { ....
+ all dm implementation
#endif
What do you think about? Let me know.
Thanks Tommaso