Add boot command "bootcmd_ram" to support autoboot from RAM.
This feature could be useful at the very initial state of chip design
when there is only a minimal set of peripheral. (e.g. without mmc and mac ..etc)
The kernel image is default to be loaded at 0x2000000 via debug port,
and the following script serves as an example:
spl()
{
cmd="riscv64-linux-gdb -q \
-ex \"target remote $host:$port\" \
-ex \"load\" \
-ex \"thread apply all set \\\$pc=&_start\" \
-ex \"thread apply all set \\\$a0=\\\$mhartid\" \
-ex \"thread apply all set \\\$a1=<dtb address>\" \
-ex \"restore u-boot.itb binary 0x200000\" \
-ex \"restore Image binary 0x2000000\" \
-ex \"c\" \
spl/u-boot-spl
"
echo $cmd
eval $cmd
}
The address where the kernel is loaded can be altered by
changing the value of KERNEL_IMAGE_ADDR.
Signed-off-by: Leo Yu-Chi Liang <ycliang@andestech.com>
Reviewed-by: Rick Chen <rick@andestech.com>
/* Increase max gunzip size */
#define CONFIG_SYS_BOOTM_LEN (64 << 20)
+/* Support autoboot from RAM (kernel image is loaded via debug port) */
+#define KERNEL_IMAGE_ADDR "0x2000000 "
+#define BOOTENV_DEV_NAME_RAM(devtypeu, devtypel, instance) \
+ "ram "
+#define BOOTENV_DEV_RAM(devtypeu, devtypel, instance) \
+ "bootcmd_ram=" \
+ "booti " \
+ KERNEL_IMAGE_ADDR \
+ "- $fdtcontroladdr\0"
+
/* When we use RAM as ENV */
/* Enable distro boot */
#define BOOT_TARGET_DEVICES(func) \
func(MMC, mmc, 0) \
- func(DHCP, dhcp, na)
+ func(DHCP, dhcp, na) \
+ func(RAM, ram, na)
#include <config_distro_bootcmd.h>
#define CONFIG_EXTRA_ENV_SETTINGS \