<smaeul>
oh, I didn't know there was a v1.2... what DRAM chip do you have?
<smaeul>
wigyori_: besides the fact that the base address is wrong, the block 0x0 - 0x40000000 is ignored due to a weird quirk of the RISC-V MMU setup
<wigyori_>
mmind00: the one i have also has 1gb - thanks for the correct setting, the patch i've found was a different one. The kernel gets to boot, it freezes later on, but at least it's a start :)
<mmind00>
smaeul: I have that D1_DEV_DDR3_16X2_V1_2 mentioned in the wiki ... which I assumed means v1.2 ;-)
<smaeul>
the lowest virtual address in the linear map is assigned to the kernel's physical address, so all DRAM below the kernel load address is inaccessible
<wigyori_>
v1.2 is the one i have too
<wigyori_>
smaeul: i see
<smaeul>
I tried to fix that, but I don't understand the MMU init code well enough to do it properly, so my attempt just broke the page allocator
<mmind00>
smaeul: SKHynix H5TQ4G63EFR
<smaeul>
mmind00: that is 512MB/chip, so if you have two chips, you have 1GB
<mmind00>
smaeul: hmm, there is another one on the bottom ... though the u-boot only reports 512MB right now
<mmind00>
smaeul: ah, the boot0 thingy reports the 1GB, though u-boot seems to think it's 512MB
<smaeul>
yes, u-boot has it hardcoded (in arch/riscv/dts/sun20i-d1-nezha-u-boot.dtsi) because boot0 doesn't have a way to update the DRAM size in the DTB
<mmind00>
smaeul: ah that explains it ... so I can update my memory node :-D
<wigyori_>
smaeul: at this time, the safest bet would be to go for using 512mb memory node in the kernel dts?
<mmind00>
wigyori_: probably, or you can just inspect the memory chips on your board (likely one on top and one on the bottom) to get the capacity
<smaeul>
wigyori_: no, the kernel should be getting its DTB from U-Boot, not loaded from a file, so you son't need to put anything in the kernel DTS
<wigyori_>
smaeul: hmm, okay. I was using a fit image as a starter, but i'll then just use a simple uimage
<smaeul>
is it possible to create a FIT image that doesn't contain a DTB?
kailo has quit [Ping timeout: 260 seconds]
<wigyori_>
not sure if it would make sense, in my understanding an fit is to have the kernel and the dtb in one image file
<mmind00>
+maybe a initramfs
<wigyori_>
yep
<smaeul>
that doesn't really work though if you expect the firmware to patch the DTB at runtime
<mmind00>
smaeul: u-boot normally can patch the dtb it loads from a fit image
<smaeul>
yes, you could add a call to fdt_fixup_memory from somewhere in arch/riscv/lib/bootm.c, but currently U-Boot doesn't do that :)
<smaeul>
and I think I was confused earlier, when talking about v1.1 I really meant v1.2
<mmind00>
smaeul: but it is definitly great that with the wip-branches most things just work that I like in my setup (mostly nfsroot)
<smaeul>
great! that is good to hear :D
<wigyori_>
getting there, getting there - freezing now at sunxi-mmc init after 'allocated mmc-pwrseq'
rsalveti has quit [Quit: Connection closed for inactivity]
jacklsw has joined #riscv
fabs has quit [Ping timeout: 264 seconds]
nvmd has quit [Quit: Later, nerds.]
aburgess has quit [Remote host closed the connection]
aburgess has joined #riscv
wgrant has quit [Ping timeout: 268 seconds]
vagrantc has quit [Quit: leaving]
frost has joined #riscv
adjtm has joined #riscv
wgrant has joined #riscv
EchelonX has quit [Quit: Leaving]
PyroPeter has quit [Ping timeout: 250 seconds]
PyroPeter has joined #riscv
riff-IRC has quit [Quit: PROTO-IRC v0.73a (C) 1988 NetSoft - Built on 11-13-1988 on AT&T System V]