Re: [PATCH v3 16/18] rockchip: Avoid #ifdefs in RK3399 SPL

Hi Jonas,

On Fri, 21 Jun 2024 at 00:45, Jonas Karlman jonas@kwiboo.se wrote:

Hi Simon,

On 2024-06-21 01:06, Simon Glass wrote:

...

The code here is confusing due to large blocks which are #ifdefed out. Add a function phase_sdram_init() which returns whether SDRAM init should happen in the current phase, using that as needed to control the code flow.

This increases code size by about 500 bytes in SPL when the cache is on, since it must call the rather large rockchip_sdram_size() function.

Signed-off-by: Simon Glass sjg@chromium.org

Changes in v3:

• Split out the refactoring into a separate patch
• Drop the non-dcache optimisation, since the cache should normally be on

drivers/ram/rockchip/sdram_rk3399.c | 47 ++++++++++++++++------------- 1 file changed, 26 insertions(+), 21 deletions(-)

diff --git a/drivers/ram/rockchip/sdram_rk3399.c b/drivers/ram/rockchip/sdram_rk3399.c index 3c4e20f4e80..2f37dd712e7 100644 --- a/drivers/ram/rockchip/sdram_rk3399.c +++ b/drivers/ram/rockchip/sdram_rk3399.c @@ -13,6 +13,7 @@ #include <log.h> #include <ram.h> #include <regmap.h> +#include <spl.h> #include <syscon.h> #include <asm/arch-rockchip/clock.h> #include <asm/arch-rockchip/cru.h> @@ -63,8 +64,6 @@ struct chan_info { };

struct dram_info { -#if defined(CONFIG_TPL_BUILD) || \

• (!defined(CONFIG_TPL) && defined(CONFIG_SPL_BUILD))
  u32 pwrup_srefresh_exit[2];
  struct chan_info chan[2];
  struct clk ddr_clk;
  

@@ -75,7 +74,6 @@ struct dram_info { struct rk3399_pmusgrf_regs *pmusgrf; struct rk3399_ddr_cic_regs *cic; const struct sdram_rk3399_ops *ops; -#endif struct ram_info info; struct rk3399_pmugrf_regs *pmugrf; }; @@ -92,9 +90,6 @@ struct sdram_rk3399_ops { struct rk3399_sdram_params *params); };

-#if defined(CONFIG_TPL_BUILD) || \

• (!defined(CONFIG_TPL) && defined(CONFIG_SPL_BUILD))
  

struct rockchip_dmc_plat { #if CONFIG_IS_ENABLED(OF_PLATDATA) struct dtd_rockchip_rk3399_dmc dtplat; @@ -191,6 +186,17 @@ struct io_setting { }, };

+/**

• • phase_sdram_init() - Check if this is the phase where SDRAM init happens
• • Returns: true to do SDRAM init in this phase, false to not
• */

+static bool phase_sdram_init(void) +{

• return spl_phase() == PHASE_TPL ||
  

•     (!IS_ENABLED(CONFIG_TPL) && !spl_in_proper());
  

+}

static struct io_setting * lpddr4_get_io_settings(const struct rk3399_sdram_params *params, u32 mr5) { @@ -3024,7 +3030,7 @@ static int rk3399_dmc_of_to_plat(struct udevice *dev) struct rockchip_dmc_plat *plat = dev_get_plat(dev); int ret;

• if (!CONFIG_IS_ENABLED(OF_REAL))
  

• if (!CONFIG_IS_ENABLED(OF_REAL) || !phase_sdram_init())
          return 0;
  
  ret = dev_read_u32_array(dev, "rockchip,sdram-params",
  

@@ -3138,22 +3144,24 @@ static int rk3399_dmc_init(struct udevice *dev)

  return 0;

} -#endif

static int rk3399_dmc_probe(struct udevice *dev) { struct dram_info *priv = dev_get_priv(dev);

-#if defined(CONFIG_TPL_BUILD) || \

• (!defined(CONFIG_TPL) && defined(CONFIG_SPL_BUILD))
  

• if (rk3399_dmc_init(dev))
  

•         return 0;
  

-#endif

• priv->pmugrf = syscon_get_first_range(ROCKCHIP_SYSCON_PMUGRF);
  

• debug("%s: pmugrf = %p\n", __func__, priv->pmugrf);
  

• priv->info.base = CFG_SYS_SDRAM_BASE;
  

• priv->info.size =
  

•         rockchip_sdram_size((phys_addr_t)&priv->pmugrf->os_reg2);
  

• if (phase_sdram_init()) {
  

•         if (rk3399_dmc_init(dev))
  

•                 return 0;
  

• } else {
  

•         priv->pmugrf = syscon_get_first_range(ROCKCHIP_SYSCON_PMUGRF);
  

•         debug("%s: pmugrf = %p\n", __func__, priv->pmugrf);
  

• }
  

• if (!CONFIG_IS_ENABLED(SYS_DCACHE_OFF)) {
  

This check should be dropped, this driver intent is to initialize dram in first phase (normally TPL), and report the size to any other phase.

This whole patch can be dropped :-) Here I am just trying to avoid code-size growth when the cache is off. But yes, it is not needed.

The main need for phase_sdram_init() and disable of DCACHE can probably be avoided if bob/kevin can move to use separate TPL and SPL instead of doing both dram init and everything else in SPL.

My best guess is that enabling of caches in SPL cause issue for bob and kevin because they only use SPL not TPL+SPL like (if I am not mistaken) all other Rockchip arm64 targets.

Using SPL-only was not something I tested when caches was enabled in SPL.

Maybe bob/kevin can be changed to also use TPL+SPL similar to all other RK3399 boards?

How U-Boot works on these chromebooks is still a mystery to me.

When coreboot is involved it would only load U-Boot proper and SPL or TPL+SPL would never be involved at all?

If I understand correctly SPL is only involved for bare metal boot, if this is the case then using TPL + back-to-brom to load SPL should probably work fine?, and would align all RK3399 boards to work in a similar way and reduce the need for special care/handling in the code.

These boards were the first rk3399 support we had in mainline, I believe. Everything else came later but conversions were not done for these existing boards.

I actually intensely dislike the back-to-brom stuff. If the ROM had an actual API (mmc_read(), etc.) then that would be fine. But just jumping back makes it hard to control what is going on.

It is interesting to hear that SPL-only is causing a problem. Do you have any inkling of what that might be? I could take a look at converting the boards to use TPL, but it does seem a bit sideways to the issue here. Is the bootrom doing some magic, perhaps? Is there decent documentation for the boot ROM these days?

Regards, Simon