* vagrantc
adds odroid-c2 to the list (amlogic ???)
<vagrantc>
eesh. i may have to revert to u-boot 2022.04 for debian ... pinebook-pro-rk3399, odroid-c2 failed, pretty sure firefly-rk3288 will fail (failed with ~rc4) ... it did boot on the pinebook, but noticed some instability in usb (maybe unrelated)
<vagrantc>
hrm. wonder if i can figure out how to write from u-boot to downgrade u-boot :)
<vagrantc>
hrm. can't find the files i wrote ... foolishly upgrading without having a good backup plan.
Net147 has quit [Remote host closed the connection]
<marex>
sjg1: create dummy lpddr4_pmu_train_*mem*202006.bin files, then
<marex>
make -j imx8mp_dhcom_pdk2_defconfig ; make -j flash.bin
<marex>
that would generate you a flash.bin per imx8mp-u-boot.dtsi
Net147 has joined #u-boot
Net147 has joined #u-boot
Net147 has quit [Changing host]
<marex>
but the lpddr4_pmu_train_1d_imem_202006.bin would be placed at unaligned address (i.e. the u-boot-spl { align-end = <4>; }; has no impact
<marex>
also, if the no-expanded; is removed, the SPL is not bundled into the flash.bin at all
akaWolf has joined #u-boot
Net147 has quit [Remote host closed the connection]
Net147 has joined #u-boot
Net147 has joined #u-boot
Net147 has quit [Changing host]
<sjg1>
marex: Where is your fdtmap? Also you are creating separate pre-images and incorporating them as blocks, which defeats binman's ability to see the whole image
<sjg1>
marex: I added in an fdtmap and I don't see an align-end problem:
<sjg1>
marex: Yes imx-cfg is not setting the image-pos stuff properly
<marex>
image-pos stuff ?
<marex>
why is the u-boot-spl not even expanded at all ?
<marex>
note that the spl binary is 137kiB
<sjg1>
marex: OK, so image-pos is the absolute position of each entry in the image
<sjg1>
marex: Using mkimage defeats that, i.e. binman has no way of knowing where it ends up
<sjg1>
marex: I'm looking at why align-end doesn't work
<marex>
sjg1: I think the real problem is the "no-expanded"
<marex>
that forces the SPL to be included, but I think that's wrong
<marex>
without the no-expanded, I think the alignment does work, but SPL binary is not included
<sjg1>
marex: OK I'm just trying to figure things out
<sjg1>
hd /tmp/b/imx8mp_dhcom_pdk2/input.flash-bin.mkimage.imx-cfg |less
<sjg1>
shows no effect from align-end
<marex>
yep ;-)
<marex>
it's like the whole section handling/expansion is somehow not applied
<sjg1>
marex: Yes, it looks like it isn't, because the data is never packed. Binman doesn't support aligning with collect_contents_to_file(), which is what is used by imx_cfg (and other things)
<sjg1>
marex: There are no tests for that case as I had not thought of it. Will need to look at what is missing
<marex>
sjg1: ack
mmu_man has quit [Ping timeout: 248 seconds]
adip has quit [Quit: Lost terminal]
<sjg1>
marex: I think the right thing to do might be to convert Entry_mkimage to a parent class of Entry_section so that all the section things 'just happen'
<sjg1>
marex: I'll need to take a closer look though and spend some time on it, hopefully before Monday
aggi has joined #u-boot
akaWolf has quit [Ping timeout: 252 seconds]
Net147 has quit [Quit: Quit]
akaWolf has joined #u-boot
vagrantc has joined #u-boot
Net147 has joined #u-boot
Net147 has joined #u-boot
Net147 has quit [Changing host]
<marex>
sjg1: thanks
persmule has quit [Remote host closed the connection]
persmule has joined #u-boot
torez has quit [Quit: torez]
GNUtoo has quit [Ping timeout: 268 seconds]
GNUtoo has joined #u-boot
mmu_man has joined #u-boot
persmule has quit [Remote host closed the connection]
GNUtoo has quit [Remote host closed the connection]
<dsmith-work>
Greetings. Can someone point me to some documentation on bootm subcommands?
<dsmith-work>
I have a kernel and device tree in an image.ub, and I want to modify the device tree a bit before booting.
<hanetzer>
what is an image.ub, a fit?
<dsmith-work>
Yes
<dsmith-work>
SO I think I need to "fatload <params>" and then "bootm start <address>", "bootm loados", "boot fdt" ? How do I know where the fdt ends up so I can modify it with "fdt mknode" "fdt set" etc.?
<dsmith-work>
IS there an env variable that gets set?
<dsmith-work>
The help just says:
<dsmith-work>
"fdt - relocate flat device tree"
<dsmith-work>
<Forty-Bot>
maybe try imextract?
<dsmith-work>
Hmm. Looks interesting
<dsmith-work>
looks like "bootm fdt" sets $ftdaddr
dormito has quit [Ping timeout: 268 seconds]
<dsmith-work>
Yey! Worked!
Gravis has joined #u-boot
<dsmith-work>
The "bootm" sets $fdtaddr, $fileaddr $filesize, and the "bootm ftd" changes $fdtaddr
<marex>
that whole block is about applying DTOs from fitImage and then possibly adding last-minute modifications to the DT by running 'loaddtoscustom' script
<marex>
which should contain like ... 'fdt set ... ; fdt set ... ' kind of stuff
<marex>
looks new enough, you might be able to use 'load' command which is filesystem agnostic
<hanetzer>
dsmith-work: vendorboot? at least its relatively recent
* hanetzer
shudders at the horror of u-boot-2010.06 being *everywhere*
<dsmith-work>
SO the point of this was to make the qspi just one partition. Need to beat on it over temperature to qualify some alternate parts. Most stuff is very hard to get nowadays.
<dsmith-work>
Something else that was very difficult to find was the device tree partition size reg entry of "0" means the rest of the device. Finally found that in a kenrel header.
<marex>
hanetzer: atheros routers came with 1.1.3 :-)
<hanetzer>
marex: true.
<marex>
dsmith-work: you can also enable command line partition parsing in Linux and then just pass mtdparts= via kernel command line
<hanetzer>
marex: didya check out my paste? I think there's some real use there :)
<marex>
hanetzer: I did, is that openocd+gdb with this hisilicon hw ?
<hanetzer>
yee, plus pycortexmdebug
GNUtoo has quit [Ping timeout: 268 seconds]
GNUtoo has joined #u-boot
<marex>
hanetzer: is that openocd replacement ?
<hanetzer>
no, its a gdb script/extension in python, more commonly used for cortex-m debugging.
<hanetzer>
it loads svd files into gdb so you can inspect registers by name, and if you have it setup you can see individual fields by value *and* if you have enumeratedValues setup translate those into human names :)
<dsmith-work>
hanetzer: Yes, I looked
<marex>
hanetzer: nice
<hanetzer>
yeh. problem is, I don't think *any* vendor ships cortex-a svd files so you gotta make your own
<dsmith-work>
marex: Yes, I tried command line first. Didn't feel like figureing out how to patch the petalinux build stuff to eable that opion. Hacking the device tree seemed easier.. ;^}
persmule has quit [Remote host closed the connection]
persmule has joined #u-boot
persmule has quit [Remote host closed the connection]