On 2023/6/27 上午 12:09, Stephen Boyd wrote:

Quoting Jacky Huang (2023-06-22 07:13:43)

quoted

From: Jacky Huang <ychuang3@nuvoton.com>

For the declaration of parent clocks, use struct clk_parent_data instead
of a string. Due to the change in the passed arguments, replace the usage
of devm_clk_hw_register_mux() with clk_hw_register_mux_parent_data() for
all cases.

Signed-off-by: Jacky Huang <ychuang3@nuvoton.com>
---
  drivers/clk/nuvoton/clk-ma35d1.c | 306 ++++++++++++++++++++++---------
  1 file changed, 219 insertions(+), 87 deletions(-)

diff --git a/drivers/clk/nuvoton/clk-ma35d1.c b/drivers/clk/nuvoton/clk-ma35d1.c
index 733750dda0f4..f1fe7edd21b5 100644
--- a/drivers/clk/nuvoton/clk-ma35d1.c
+++ b/drivers/clk/nuvoton/clk-ma35d1.c

@@ -63,167 +63,298 @@ static DEFINE_SPINLOCK(ma35d1_lock);
  #define PLL_MODE_SS             2
  
  static const struct clk_parent_data ca35clk_sel_clks[] = {
-       { .index = 0 },  /* HXT */
-       { .index = 1 },  /* CAPLL */
-       { .index = 2 }   /* DDRPLL */
+       { .fw_name = "hxt", },
+       { .fw_name = "capll", },
+       { .fw_name = "ddrpll", },

Is there any reason why this changed from numbers to strings? It is
faster to directly index vs. access the clock-names property. The
preference is to use the index field and only use fw_name if we're
migrating legacy clk drivers to use DT based lookups.

This is simply because I think strings are more readable than numbers.
Since using an index is the preferred method, I will change all occurrences
of ".fw_name" to ".index" and add comments to describe it.
As this patch was applied, I will create another patch relative to this one.


Best Regards,
Jacky Huang



_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel