On Thu, Jul 20, 2023 at 02:29:07PM +0300, Nikita Shubin via B4 Relay wrote:
> From: Nikita Shubin <[email protected]>
>
> This adds an SoC driver for the ep93xx. Currently there
> is only one thing not fitting into any other framework,
> and that is the swlock setting.
>
> It's used for clock settings and restart.
> +#include <linux/clk-provider.h>
> +#include <linux/init.h>
> +#include <linux/kernel.h>
> +#include <linux/mfd/syscon.h>
> +#include <linux/of.h>
> +#include <linux/regmap.h>
> +#include <linux/slab.h>
> +#include <linux/sys_soc.h>
> +#include <linux/soc/cirrus/ep93xx.h>
...
> +#define EP93XX_SYSCON_SYSCFG_REV_MASK 0xf0000000
GENMASK() (you will need bits.h, and looking below you seem missed it anyway)
...
> + spin_lock_irqsave(&ep93xx_swlock, flags);
> +
> + regmap_read(map, EP93XX_SYSCON_DEVCFG, &val);
> + val &= ~clear_bits;
> + val |= set_bits;
> + regmap_write(map, EP93XX_SYSCON_SWLOCK, EP93XX_SWLOCK_MAGICK);
> + regmap_write(map, EP93XX_SYSCON_DEVCFG, val);
Is this sequence a must?
I.o.w. can you first supply magic and then update devcfg?
> + spin_unlock_irqrestore(&ep93xx_swlock, flags);
...
> +void ep93xx_swlocked_update_bits(struct regmap *map, unsigned int reg,
> + unsigned int mask, unsigned int val)
> +{
Same Q as above.
> +}
...
> + int rev = ep93xx_chip_revision(map);
> +
> + switch (rev) {
switch(ep93xx_chip_revision(map))
?
> + case EP93XX_CHIP_REV_D0:
> + return "D0";
> + case EP93XX_CHIP_REV_D1:
> + return "D1";
> + case EP93XX_CHIP_REV_E0:
> + return "E0";
> + case EP93XX_CHIP_REV_E1:
> + return "E1";
> + case EP93XX_CHIP_REV_E2:
> + return "E2";
> + default:
> + return "unknown";
> + }
...
> +static unsigned long __init calc_pll_rate(u64 rate, u32 config_word)
> +{
> + rate *= ((config_word >> 11) & GENMASK(4, 0)) + 1; /* X1FBD */
> + rate *= ((config_word >> 5) & GENMASK(5, 0)) + 1; /* X2FBD */
> + do_div(rate, (config_word & GENMASK(4, 0)) + 1); /* X2IPD */
> + rate >>= ((config_word >> 16) & 3); /* PS */
GENMASK()
> + return rate;
> +}
...
> + /* Multiplatform guard, only proceed on ep93xx */
> + if (!of_machine_is_compatible("cirrus,ep9301"))
> + return 0;
Why?
> + map = syscon_regmap_lookup_by_compatible("cirrus,ep9301-syscon");
> + if (IS_ERR(map))
> + return PTR_ERR(map);
Is this not enough?
...
> + if (!(value & EP93XX_SYSCON_CLKSET1_NBYP1))
> + clk_pll1_rate = EP93XX_EXT_CLK_RATE;
> + else
> + clk_pll1_rate = calc_pll_rate(EP93XX_EXT_CLK_RATE, value);
Positive conditionals in if-else are easier to be read.
...
> + clk_f_div = fclk_divisors[(value >> 25) & 0x7];
> + clk_h_div = hclk_divisors[(value >> 20) & 0x7];
> + clk_p_div = pclk_divisors[(value >> 18) & 0x3];
GENMASK() in all three?
...
> + np = of_find_node_by_path("/");
> + of_property_read_string(np, "model", &machine);
> + if (machine)
> + attrs->machine = kstrdup(machine, GFP_KERNEL);
No error check?
> + of_node_put(np);
Looks like NIH of_flat_dt_get_machine_name(). Can you rather make use of it as
it's done, e.g., here arch/riscv/kernel/setup.c:251?
--
With Best Regards,
Andy Shevchenko
Hello Andy!
On Fri, 2023-07-21 at 17:13 +0300, Andy Shevchenko wrote:
> On Thu, Jul 20, 2023 at 02:29:07PM +0300, Nikita Shubin via B4 Relay
> wrote:
> > From: Nikita Shubin <[email protected]>
> > + spin_lock_irqsave(&ep93xx_swlock, flags);
> > +
> > + regmap_read(map, EP93XX_SYSCON_DEVCFG, &val);
> > + val &= ~clear_bits;
> > + val |= set_bits;
> > + regmap_write(map, EP93XX_SYSCON_SWLOCK,
> > EP93XX_SWLOCK_MAGICK);
> > + regmap_write(map, EP93XX_SYSCON_DEVCFG, val);
>
> Is this sequence a must?
> I.o.w. can you first supply magic and then update devcfg?
>
Unfortunately it is a must to write EP93XX_SYSCON_SWLOCK and only then
the next write to swlocked registers will succeed.
On Fri, Jul 28, 2023 at 12:28 PM Nikita Shubin
<[email protected]> wrote:
>
> Hello Andy!
>
> On Fri, 2023-07-21 at 17:13 +0300, Andy Shevchenko wrote:
> > On Thu, Jul 20, 2023 at 02:29:07PM +0300, Nikita Shubin via B4 Relay
> > wrote:
> > > From: Nikita Shubin <[email protected]>
> > > + spin_lock_irqsave(&ep93xx_swlock, flags);
> > > +
> > > + regmap_read(map, EP93XX_SYSCON_DEVCFG, &val);
> > > + val &= ~clear_bits;
> > > + val |= set_bits;
> > > + regmap_write(map, EP93XX_SYSCON_SWLOCK,
> > > EP93XX_SWLOCK_MAGICK);
> > > + regmap_write(map, EP93XX_SYSCON_DEVCFG, val);
> >
> > Is this sequence a must?
> > I.o.w. can you first supply magic and then update devcfg?
> >
>
> Unfortunately it is a must to write EP93XX_SYSCON_SWLOCK and only then
> the next write to swlocked registers will succeed.
This doesn't answer my question. Can you first write a magic and then
_update_ the other register (update means RMW op)?
--
With Best Regards,
Andy Shevchenko