Reviewing myself:
On Sun, 4 Nov 2012 12:32:04 +0100, Albert ARIBAUD albert.u.boot@aribaud.net wrote:
diff --git a/arch/arm/lib/crt0.S b/arch/arm/lib/crt0.S new file mode 100644 index 0000000..fd6bd92 --- /dev/null +++ b/arch/arm/lib/crt0.S
+clbss_l:cmp r0, r1 /* while not at end of BSS */
- strlo r2, [r0] /* clear 32-bit BSS word */
- addlo r0, r0, #4 /* move to next
Unclosed comment here. This causes the following code to be ignored:
- bhs clbss_l
+#ifdef CONFIG_SPL_BUILD
- bl coloured_LED_init
- bl red_led_on
+#endif
- /* call board_init_r(gd_t *id, ulong dest_addr) */
Which might explain why SPL builds did not complain about LED calls and why some boards failed.
Also:
+#if defined(CONFIG_NAND_SPL)
- ldr pc, _nand_boot
+#elif ! defined(CONFIG_SPL_BUILD)
- /* setup parameters for board_init_r */
- mov r0, r8 /* gd_t */
- ldr r1, [r8, #GD_RELOCADDR] /* dest_addr */
- /* call board_init_r */
- ldr pc, _board_init_r /* this is auto-relocated! */
+#endif
The #elif is wrong, as it causes CONFIG_SPL_BUILD builds to not call board_init_f() at all; actually, they overrun into whatever follows. This might have been what Sughosh (Cc:) hit with the hawkboard.
Amicalement,