On Mon, Aug 13, 2018 at 12:10 AM Sascha Hauer <s.hauer@xxxxxxxxxxxxxx> wrote: > > On Fri, Aug 10, 2018 at 12:04:26PM -0700, Andrey Smirnov wrote: > > For both SoCs data found in SBMR registers reflects only the boot > > source that was selected via pins of fuses and not the final boot > > source that ended up being used by MaskROM code. Original i.MX7 boot > > source detection implementation worked around that fact by having a > > special code to correctly handle "Manufacturing Mode". > > > > MaskROM in i.MX8MQ changed what SoC uses as recovery device and > > switched it to be USDHC2. It also made recovery device switch always > > enabled. Since correct actual boot source detection is important to > > being able to properly boot i.MX8MQ (due to not using DCD to > > initialize RAM), change the code to handle described exception. > > > > Instead of trying to adapt original i.MX7 code with yet another > > special case if(), change the whole thing to do what U-Boot does on > > i.MX7 and i.MX8MQ and use "Boot information for software" provided by > > recent (found in i.MX7 and i.MX8MQ) versions of MaskROM. > > > > Signed-off-by: Andrey Smirnov <andrew.smirnov@xxxxxxxxx> > > --- > > arch/arm/mach-imx/boot.c | 97 +++++++++++++--------------------------- > > 1 file changed, 30 insertions(+), 67 deletions(-) > > > > diff --git a/arch/arm/mach-imx/boot.c b/arch/arm/mach-imx/boot.c > > index 45170ab10..f1fc40479 100644 > > --- a/arch/arm/mach-imx/boot.c > > +++ b/arch/arm/mach-imx/boot.c > > @@ -28,6 +28,7 @@ > > #include <mach/imx6-regs.h> > > #include <mach/imx7-regs.h> > > #include <mach/vf610-regs.h> > > +#include <mach/imx8mq.h> > > > > > > static void > > @@ -424,32 +425,12 @@ void imx6_boot_save_loc(void) > > imx_boot_save_loc(imx6_get_boot_source); > > } > > > > -#define IMX7_SRC_SBMR1 0x58 > > -#define IMX7_SRC_SBMR2 0x70 > > +#define IMX7_BOOT_SW_INFO_POINTER_ADDR 0x000001E8 > > +#define IMX8M_BOOT_SW_INFO_POINTER_ADDR_A0 0x000009e8 > > +#define IMX8M_BOOT_SW_INFO_POINTER_ADDR_B0 0x00000968 > > Uargh. Whatever they smoke @NXP, I hope it at least makes them feel > good. We are provided a "Boot information for software" pointer whose > address depends on the mask ROM version? AFAIK, yes. What I've been told by NXP folks is: "This could potentially change with each ROM release update" (this meaning the location address), so I am assuming this all depends on where linker would place that global variable next time they recompile the MaskROM code. > So to use the information we must first get the mask ROM version which itself is found at varying > offsets in the mask ROM? True, although this seems like a genuine screw-up, whereas "NXP's Moving Boot Info" sounds more like a "works as intended" kind of scenario. > Needless to say that my i.MX8MQ reference manual specifies the pointer > address to 0x1e8 which is obviously copied from the i.MX7 Manual without > adjusting. Yeah, I discovered that the hard way. I brought it up to NXP folks and there's supposed to be a ticket, to fix that discrepancy in the next revision of the RM, somewhere within their corporate ticket tracking system. Thanks, Andrey Smirnov _______________________________________________ barebox mailing list barebox@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/barebox
