On Sun, Apr 23, 2017 at 4:10 AM, Simon Glass sjg@chromium.org wrote:

We should not call out to board code from drivers. With driver model, mmc_power_init() already has code to use a named regulator, but the legacy code path remains. Update the code to make this clear.

I don't like this patch as it described. (I will answer to your mail in the other thread later) Let me check more, possible it's okay to have.

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

drivers/mmc/mmc.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)

diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c index 72fc17716e..3cdf6a4f3b 100644 --- a/drivers/mmc/mmc.c +++ b/drivers/mmc/mmc.c @@ -1608,17 +1608,17 @@ static int mmc_send_if_cond(struct mmc *mmc) return 0; }

+#ifndef CONFIG_DM_MMC /* board-specific MMC power initializations. */ __weak void board_mmc_power_init(void) { } +#endif

static int mmc_power_init(struct mmc *mmc) {

•   board_mmc_power_init();
  

-#if defined(CONFIG_DM_MMC) && defined(CONFIG_DM_REGULATOR) && \

•   !defined(CONFIG_SPL_BUILD)
  

+#if defined(CONFIG_DM_MMC) +#if defined(CONFIG_DM_REGULATOR) && !defined(CONFIG_SPL_BUILD) struct udevice *vmmc_supply; int ret;

@@ -1635,6 +1635,13 @@ static int mmc_power_init(struct mmc *mmc) return ret; } #endif +#else /* !CONFIG_DM_MMC */

•   /*
  

•    * Driver model should use a regulator, as above, rather than calling
  

•    * out to board code.
  

•    */
  

•   board_mmc_power_init();
  

+#endif return 0; }

-- 2.12.2.816.g2cccc81164-goog

---

U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot