commit 787f04bb6a ("imx: add USB2_BOOT type") broke get_boot_device()
for IMX8 which affects booting from SDP due to boot_instance being
non-zero.
Fix this by only using boot_instance for imx8ulp and imx9.
Fixes: 787f04bb6a ("imx: add USB2_BOOT type")
Signed-off-by: Tim Harvey <tharvey@gateworks.com>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Reviewed-by: Fabio Estevam <festevam@denx.de>
Decode ECSPI boot device in env_get_location() from i.MX8M ROMAPI tables.
This is necessary to correctly identify env is in SPI NOR when the system
boots from SPI NOR attached to ECSPI.
This reinstates change from commit:
e26d0152d6 ("ARM: imx: Decode ECSPI env location from i.MX8M ROMAPI tables")
which has been dropped in commit:
b0a284a7c9 ("imx: move get_boot_device to common file")
Fixes: b0a284a7c9 ("imx: move get_boot_device to common file")
Signed-off-by: Marek Vasut <marex@denx.de>
Reviewed-by: Fabio Estevam <festevam@denx.de>
i.MX8MN/P/ULP supports ROM API, they have almost same get_boot_device
implementation, so move to a common file. And when support i.MX9,
no need to include the other function copy.
Since sys_proto.h is included in imx_romapi.c, there will be build
warning for i.MX8M because wdog_regs not defined, so include imx-regs.h
in i.MX8M sys_proro.h
Signed-off-by: Peng Fan <peng.fan@nxp.com>
The ROM API is thoroughly undocumented, but apparently passing the xor
of the real arguments as an extra argument is required [1]. Also, we
need to do the "save gd/restore gd" dance. These are both error-prone,
and lead to a lot of code duplication.
Since both imx8m[np] and imx8ulp SOCs have this, add a separate
translation unit which is included precisely when the new
CONFIG_IMX8_ROMAPI symbol is set, which provide convenience wrappers
that take care of computing the xor value as well as doing the gd
dance, and that thus have a more intuitive API. Subsequent patches
will make use of these to reduce boilerplate.
[1] One wonders, for example, if the check is only applied to the
lower 32 bits, or if we're implicitly relying on all 64-bit pointer
values we're passing effectively have 0 in the upper 32 bits.
Reviewed-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
