Hi Simon,
On Wed, 2023-02-08 at 18:08 -0700, Simon Glass wrote:
On Wed, 8 Feb 2023 at 12:04, Henrik Grimler henrik@grimler.se wrote:
Otherwise non-ChromeOS samsung devices, like the odroid boards, are stuck in a bootloop if CONFIG_CROS_EC is not enabled:
<...> MMC: SAMSUNG SDHCI: 2, EXYNOS DWMMC: 0 Loading Environment from MMC... *** Warning - bad CRC, using default environment
cros-ec communications failure -96
Please reset with Power+Refresh
Cannot init cros-ec device resetting ...
Issue started after commit e44d7e73fe0d ("dm: core: Switch uclass_*_device_err to use uclass_*_device_check").
Signed-off-by: Henrik Grimler henrik@grimler.se
board/samsung/common/board.c | 27 ++++++++++++++------------- 1 file changed, 14 insertions(+), 13 deletions(-)
diff --git a/board/samsung/common/board.c b/board/samsung/common/board.c index 16ce5cb89253..78161b807bac 100644 --- a/board/samsung/common/board.c +++ b/board/samsung/common/board.c @@ -217,26 +217,27 @@ int checkboard(void) #ifdef CONFIG_BOARD_LATE_INIT int board_late_init(void) { - struct udevice *dev; - int ret; int mmcbootdev = get_boot_mmc_dev(); char mmcbootdev_str[16];
- ret = uclass_first_device_err(UCLASS_CROS_EC, &dev); - if (ret && ret != -ENODEV) { - /* Force console on */ - gd->flags &= ~GD_FLG_SILENT;
- printf("cros-ec communications failure %d\n", ret); - puts("\nPlease reset with Power+Refresh\n\n"); - panic("Cannot init cros-ec device"); - return -1; - }
printf("Boot device: MMC(%u)\n", mmcbootdev); sprintf(mmcbootdev_str, "%u", mmcbootdev); env_set("mmcbootdev", mmcbootdev_str);
+ if (IS_ENABLED(CONFIG_CROS_EC)) { + struct udevice *dev; + int ret = uclass_first_device_err(UCLASS_CROS_EC, &dev); + if (ret && ret != -ENODEV) { + /* Force console on */ + gd->flags &= ~GD_FLG_SILENT;
+ printf("cros-ec communications failure %d\n", ret); + puts("\nPlease reset with Power+Refresh\n\n"); + panic("Cannot init cros-ec device"); + return -1; + } + }
return 0; } #endif -- 2.30.2
Reviewed-by: Simon Glass sjg@chromium.org
Another option would be to change the check for -ENODEV to also check for -EEPFNOSUPPORT (missing uclass).
Thanks for reviewing! Checking for that return code works as well, and would require a smaller patch, I will send a v2 and use that suggestion instead.
Best regards, Henrik Grimler