using binman fails boot
Simon Glass
sjg at chromium.org
Sun Jul 18 04:22:12 CEST 2021
Hi Tim,
On Fri, 16 Jul 2021 at 17:15, Tim Harvey <tharvey at gateworks.com> wrote:
>
> On Fri, Jul 16, 2021 at 3:11 PM Simon Glass <sjg at chromium.org> wrote:
> >
> > () which has
> > Hi Tim,
> >
> > On Fri, 16 Jul 2021 at 15:43, Tim Harvey <tharvey at gateworks.com> wrote:
> > >
> > > On Thu, Jul 15, 2021 at 9:30 PM Simon Glass <sjg at chromium.org> wrote:
> > > >
> > > > Hi Tim,
> > > >
> > > > On Thu, 15 Jul 2021 at 16:58, Tim Harvey <tharvey at gateworks.com> wrote:
> > > > >
> > > > > Greetings,
> > > > >
> > > > > I'm taking a look at moving imx8mm-venice to use binman for packaging.
> > > > > After doing so U-Boot proper fails to boot:
> > > > >
> > > > > U-Boot SPL 2021.07-00475-g1126252f40 (Jul 15 2021 - 11:09:02 -0700)
> > > > > GSC : v58 0xf098 RST:VIN Thermal Protection Disabled
> > > > > Model : GW7300-00-B1B
> > > > > Serial : 852420
> > > > > MFGDate : 10-26-2020
> > > > > RTC : 122
> > > > > PMIC : MP5416
> > > > > DRAM : LPDDR4 1 GiB
> > > > > WDT: Not starting
> > > > > Trying to boot from MMC1
> > > > > DTB : imx8mm-venice-gw73xx-0x
> > > > >
> > > > >
> > > > > U-Boot 2021.07-00475-g1126252f40 (Jul 15 2021 - 11:09:02 -0700)
> > > > >
> > > > > CPU: Freescale i.MX8MMQ rev1.0 1600 MHz (running at 1200 MHz)
> > > > > CPU: Industrial temperature grade (-40C to 105C) at 43C
> > > > > Reset cause: POR
> > > > > Model: Gateworks Venice GW73xx-0x i.MX8MM Development Kit
> > > > > DRAM: 1 GiB
> > > > > temp : 38.3C
> > > > > vdd_bat : 0.000V
> > > > > vdd_vin : 15.731V
> > > > > vdd_adc1: 0.000V
> > > > > vdd_adc2: 0.000V
> > > > > vdd_dram: 1.093V
> > > > > vdd_1p2 : 1.193V
> > > > > vdd_1p0 : 0.985V
> > > > > vdd_2p5 : 2.470V
> > > > > vdd_3p3 : 3.250V
> > > > > vdd_0p95: 0.948V
> > > > > vdd_1p8 : 1.799V
> > > > > vdd_gsc : 3.262V
> > > > > initcall sequence 000000007ffc4f58 failed at call 0000000040255910 (err=-2)
> > > > > ### ERROR ### Please RESET the board ###
> > > > >
> > > > > Any ideas what this could be?
> > > >
> > > > I don't have much idea. What is the initcall that is failing? Can you
> > > > check u-boot.map ? That might give a clue as to what is failing. I
> > > > assume the DT is passed to U-Boot somehow from SPL?
> > > >
> > >
> > > Simon,
> > >
> > > Thanks for the help!
> > >
> > > The initcall addr doesn't match anything in u-boot.map (maybe
> > > u-boot.map doesn't show what's in lib/binman.o?) but I was able to
> >
> > It certainly should show up, but if you have CONFIG_LTO enabled lots
> > of functions disappear. Still if you get an initcall address I would
> > expect a function to be present. Make sure you use the unallocated
> > address.
>
> I'm not sure what you mean by 'Make sure you use the unallocated address'
Sorry I mean unrelocated address. After U-Boot relocates the addresses
change but it still prints out the original addresses.
>
> >
> > > track it down to initr_binman() failing due to
> > > binman_init()->find_image_node(&binman->image)' returning -EINVAL.
> > > This is because my imx8mm-venice-gw73xx-0x-uboot.dtsi doesn't have a
> > > binman node (my CONFIG_DEFAULT_DEVICE_TREE did but not my actual
> > > dtbs). So I have it working now!
> >
> > OK good progress! Perhaps we should put an error message in initr_binman() ?
> >
>
> sure - I just sent a patch
>
> > >
> > > > >
> > > <snip>
> > > > >
> > > > > A follow-on question is that I would like to investigate using binman
> > > > > in the SPL to dynamically access the IMX8M ddr training blobs so that
> > > > > we don't have to waste padding space taking them onto the end of the
> > > > > SPL which is currently done. The lpddr4 training blobs I'm using
> > > > > currently take up 57k without padding compared to 81k with padding.
> > > > > The location of them is handled in ddr_load_train_firmware.
> > > > >
> > > > > If I add the following to my SPL:
> > > > > diff --git a/board/gateworks/venice/spl.c b/board/gateworks/venice/spl.c
> > > > > index d0a490b0e6..62eb67fa5e 100644
> > > > > --- a/board/gateworks/venice/spl.c
> > > > > +++ b/board/gateworks/venice/spl.c
> > > > > @@ -3,6 +3,7 @@
> > > > > * Copyright 2021 Gateworks Corporation
> > > > > */
> > > > >
> > > > > +#include <binman_sym.h>
> > > > > #include <common.h>
> > > > > #include <cpu_func.h>
> > > > > #include <hang.h>
> > > > > @@ -252,6 +253,8 @@ static int power_init_board(void)
> > > > > return 0;
> > > > > }
> > > > >
> > > > > +binman_sym_declare(ulong, blob_1, image_pos);
> > > > > +
> > > > > void board_init_f(ulong dummy)
> > > > > {
> > > > > struct udevice *dev;
> > > > > @@ -291,6 +294,8 @@ void board_init_f(ulong dummy)
> > > > > gpio_request(PCIE_RSTN, "perst#");
> > > > > gpio_direction_output(PCIE_RSTN, 0);
> > > > >
> > > > > + printf("%s: blob_1:0x%0lx\n", __func__, binman_sym(ulong,
> > > > > blob_1, image_pos));
> > > > > +
> > > > > /* GSC */
> > > > > dram_sz = gsc_init(0);
> > > > >
> > > > > I get 'blob_1:0x0' which is not what I expected.
> > > > >
> > > > > If I understand correctly binman is using linker symbols to determine
> > > > > where things are in the image? What I don't quite understand is what
> > > > > symbols are valid to use assuming my dtsi above. The binman.rst docs
> > > > > talk use 'u_boot_any' as an example which apparently can match
> > > > > 'u-boot.bin', 'u-boot.img', and 'u-boot-nodtb.bin' but I can't find
> > > > > the code that somehow translates this meaning.
> > > >
> > > > Actually any symbol can be used. It basically depends on the name of
> > > > the entry in your image description. So here it would be
> > > > blob-ext at 1...I think that translates to blob_ext_1 but I'm not sure
> > > > about the @. You could try blob-ext-1 instead. It does not know about
> > > > phandles or labels.
> > > >
> > > > If you pass BINMAN_VERBOSE=4 to the build you should see it talking
> > > > about writing symbols into the SPL image.
> > > >
> > >
> > > For the following:
> > > u-boot-spl-ddr {
> > > filename = "u-boot-spl-ddr.bin";
> > > pad-byte = <0xff>;
> > > align-size = <4>;
> > > align = <4>;
> > >
> > > u-boot-spl {
> > > align-end = <4>;
> > > };
> > >
> > > blob-ext at 1 {
> > > filename = "lpddr4_pmu_train_1d_imem.bin";
> > > size = <0x8000>;
> > > };
> > >
> > > blob-ext at 2 {
> > > filename = "lpddr4_pmu_train_1d_dmem.bin";
> > > size = <0x4000>;
> > > };
> > >
> > > blob-ext at 3 {
> > > filename = "lpddr4_pmu_train_2d_imem.bin";
> > > size = <0x8000>;
> > > };
> > >
> > > blob-ext at 4 {
> > > filename = "lpddr4_pmu_train_2d_dmem.bin";
> > > size = <0x4000>;
> > > };
> > > };
> > >
> > > I tried 'blob_ext_1' and 'blob_ext1' and both formats resolve to 0x0.
> > > The 'ext-blob' is an entry type supported by binman so if I had
> > > multiple they must be called blob-ext at 1, blob-ext at 2, ... right?
> > >
> > > The entry_name used in binman_sym_declare/binman_sym certainly can't
> > > support non C varname characters so '-' and '@' characters must get
> > > translated somewhere. Where would that be done in order to figure out
> > > what to use?
> >
> > If you want to look at the internals, see section.py LookupSymbol().
> >
> > It takes the ELF symbol and replaces _ by - but does not (cannot)
> > replace _ with @. So I think you'll have to use - instead of @
> >
> > I suppose we could do the search in the other direction (take the
> > entry and try to find the symbol that matches it), but I'd need to
> > think about it. A simple translation is easier.
> >
> > In this case binman should really give an error for your chosen entry
> > name (blob-ext at 4) but it doesn't know you are using it as a symbol. I
> > think it should complain about this (see the Warning in section.py
> > LookupSymbol()) but apparently it does not in your case.
>
> But isn't blob-ext at 4 a correct name? I can't use 'blob-ext-4' as
> that's an unknown entry type.
Well you can use any name and specify the type:
my-name {
type = "blob-ext";
};
>
> >
> > If you can push your tree somewhere (with this problem) I'll see if I
> > can figure out why.
> >
>
> Sure, I pushed it to
> https://github.com/Gateworks/uboot-venice/tree/WIP-venice-binman
> make imx8mm_venice_defconfig
> make
OK
>
> > >
> > > BINMAN_VERBOSE=4 indeed prints out a tone of stuff but I'm not seeing
> > > anything for 'blob' below that would seem to indicate one node name vs
> > > another:
> >
> > Oops you need BINMAN_VERBOSE=5 - see elf.py LookupAndWriteSymbols()
> > which has tout.Debug() which is level 5.
> >
>
> LookupAndWriteSymbols ends up doing nothing because
> syms.get('__image_copy_start') returns None.
Well that is likely the problem.
Regards,
Simon
More information about the U-Boot
mailing list