00:00
bkeys has quit [*.net *.split]
00:00
warren has quit [*.net *.split]
00:00
sharkcz has quit [*.net *.split]
00:02
warren has joined #fedora-riscv
00:02
sharkcz has joined #fedora-riscv
00:04
bkeys has joined #fedora-riscv
06:36
tekkamanninja has joined #fedora-riscv
06:42
King_InuYasha has quit [Read error: Connection reset by peer]
06:42
King_InuYasha has joined #fedora-riscv
08:31
jcajka has joined #fedora-riscv
15:11
bkeys has quit [Remote host closed the connection]
15:40
bkeys has joined #fedora-riscv
16:27
bkeys has quit [Quit: bkeys]
16:28
bkeys has joined #fedora-riscv
16:35
Denis has joined #fedora-riscv
16:36
<
Denis >
I have tried the latest WIP build of Fedora for Nezha D1, and it seems to be hard-coded to assume 2GB of RAM
16:37
<
davidlt[m] >
I don't think Fu Wei is online right now. He produced those disk images IIRC.
16:38
<
Denis >
which does not block me entirely from running the image on the board (it is possible using a much older kernel), but if anybody is interested to take a bug report, I can provide diagnostics
16:38
<
davidlt[m] >
You should fine that on StarFive GitHub repos.
16:42
<
Denis >
davidlt[m]: it looks like a different manufacturer, Nezha D1 is made by Allwinner
16:42
<
davidlt[m] >
Ops, sorry :)
16:42
<
davidlt[m] >
My brain went on the autopilot :D
16:43
<
davidlt[m] >
I was reading about StarFive board and somehow everything became StarFive boards :)
16:43
<
davidlt[m] >
Those are also produced by Fu Wei.
16:44
<
davidlt[m] >
I am not sure where to file a bug report for it.
16:52
<
Denis >
davidlt[m]: it looks like the best next attempt would be during Beijing business hours then
17:10
bkeys has quit [Remote host closed the connection]
17:13
bkeys has joined #fedora-riscv
17:31
<
tekkamanninja >
yes, I am online
17:32
<
tekkamanninja >
this is fuwei speaking
17:32
<
tekkamanninja >
sorry for the my nickname
17:33
<
tekkamanninja >
Denis: sorry for the "hard-coded", this is in dtb
17:33
<
Denis >
oh, nice to meet you
17:33
<
tekkamanninja >
you can load different dtb for your board(1GB or 512MB)
17:35
<
Denis >
tekkamanninja: I tried that, it works as expected only if I boot the old kernel (5.4.61) with its own "aw" DT files via extlinux
17:36
<
tekkamanninja >
yes, the new kernel doesn't support wifi and hdmi
17:36
<
tekkamanninja >
even RTC
17:38
<
Denis >
tekkamanninja: that's fine, this is one of several Buildbot workers, so they run headless anyway. only the networking is required
17:39
<
tekkamanninja >
yes, this is the latest one with ethernet enabled
17:39
<
Denis >
so the problem is, if I try to boot a newer kernel, it either does not boot or thinks it has 2GB of RAM
17:40
<
tekkamanninja >
with this image, wifi works on 5.4.61
17:40
<
Denis >
I can explain what I tried if you have time
17:40
<
tekkamanninja >
you just need to modify grub.conf in boot partition
17:41
<
Denis >
in theory, yes
17:41
<
tekkamanninja >
devicetree /dtb-5.16.0-rc3/allwinner/sun20i-d1-nezha-2g.dtb to devicetree /dtb-5.16.0-rc3/allwinner/sun20i-d1-nezha-1g.dtb
17:42
<
tekkamanninja >
have you tried ?
17:42
<
Denis >
yes, it boots and after booting it "sees" 2GB
17:42
<
tekkamanninja >
sorry , the one you have is dtb-5.16.0-rc2
17:43
<
Denis >
sed --in-place 's#\(devicetree .*\)/sun20i-d1-nezha.dtb#\1/sun20i-d1-nezha-1g.dtb#' /boot/grub.cfg
17:43
<
tekkamanninja >
OMG, it seems got info from u-boot, I only tried 2GB
17:44
<
Denis >
either the "1GB" DT file does not specify 1GB, or the kernel ignores that
17:44
<
tekkamanninja >
I need to check this with 512B
17:44
<
Denis >
on the u-boot note, u-boot claims 2GB upfront as well
17:44
<
tekkamanninja >
I need to check this with 512MB, I have one ,
17:44
<
Denis >
this is u-boot output: "DRAM: 2 GiB"
17:46
<
tekkamanninja >
or I can build one for you, please let me know your email , I can send it to you
17:47
<
Denis >
whichever suits you, let me know
17:49
<
Denis >
one problem with the old 5.4.61 is the load average never goes below 1.0 and the system feels very slow
17:50
<
Denis >
the new 5.14.16 does not have this problem, but the RAM size issue blocks me from using it
17:52
<
Denis >
other than that, HDMI, WiFi, Bluetooth and other peripherals are irrelevant for the purpose, so long as the board boots from an SD card or even from USB, brings up Ethernet and runs as a server
17:58
<
tekkamanninja >
don't expect WiFi, Bluetooth can work, the driver is incompatible with upstream kernel , and no one have time to update it
17:59
<
tekkamanninja >
HDMI may work if I have some time
18:00
<
Denis >
yes, that's perfectly acceptable
18:03
<
davidlt[m] >
Yeah, Allwinner will be problematic for a long time
18:06
<
tekkamanninja >
for 1GB board
18:06
<
tekkamanninja >
davidlt[m]: yes, still
18:07
<
tekkamanninja >
just do a dd : dd if=u-boot-1GB.toc1 of=/dev/sdX bs=512 seek=32800
18:08
<
tekkamanninja >
into your SD, that should work, I have tried that
18:08
<
tekkamanninja >
Denis: :-) ^^
18:11
<
tekkamanninja >
I may need to add more config for different ddr size
18:11
bkeys has quit [Remote host closed the connection]
18:11
bkeys has joined #fedora-riscv
18:11
bkeys has quit [Client Quit]
18:12
bkeys has joined #fedora-riscv
18:13
<
Denis >
alright, the new bootloader should be in the right place in the middle of unallocated disk space, and there is a backup of the previous contents, so I am ready to reboot
18:14
<
tekkamanninja >
I have tried it on my board , that works
18:14
<
Denis >
this time u-boot says "DRAM: 1 GiB"
18:15
<
tekkamanninja >
and you just need to wait kernel boot up
18:16
<
tekkamanninja >
[ 0.000000] Memory: 948320K/1046528K available (11523K kernel code, 6566K rwdata, 6144K rodata, 2544K init, 604K bss, 98208K reserved, 0K cma-reserved)
18:16
<
tekkamanninja >
on my board
18:17
<
Denis >
it is paused now. what is the right thing to do to boot 5.14.16: give it sun20i-d1-nezha.dtb as it was originally or sun20i-d1-nezha-1g.dtb?
18:17
<
Denis >
grub.cfg still has sun20i-d1-nezha-1g.dtb, I guess
18:18
<
tekkamanninja >
you can ignore this, since ddr size comes from uboot
18:18
<
Denis >
okay, trying with whatever is there
18:19
<
tekkamanninja >
o, 5.14.16 can not boot by grub
18:19
<
tekkamanninja >
because it is lack of efi stub support
18:19
<
Denis >
5.16.0-rc2 it says
18:20
<
tekkamanninja >
yes, grub only boot mainline kernel(with patches) for now
18:20
<
Denis >
5.14.16 is the newest from the disk image I had before today's upgrade (September version)
18:20
<
davidlt[m] >
5.14.X has EFI STUB support in the kernel
18:20
<
Denis >
anyway, it is booting and in a minute should be up
18:20
<
davidlt[m] >
EFI is available within the kernel for a long time now
18:21
<
tekkamanninja >
sorry , typo , I mean 5.4.61
18:21
<
davidlt[m] >
Yeah, that should be too old.
18:21
<
tekkamanninja >
5.14.16 can boot with grub
18:22
<
Denis >
1GB RAM it is
18:22
<
tekkamanninja >
but I didn't put 5.14.16 into image , so ... that maybe your own knernel I guess
18:23
<
Denis >
hmm, something isn't right, because the kernel says it is 5.4.61
18:23
<
Denis >
let me watch the boot closely
18:23
<
tekkamanninja >
stop uboot, by v
18:23
<
tekkamanninja >
then , "run boot_grub"
18:23
<
Denis >
yes, that's exactly what I did, let me check the scrollback
18:24
<
tekkamanninja >
you can see grub menu
18:26
<
Denis >
as it turns out, GRUB started, displayed the menu, started to load the kernel, then it was a hardware reset
18:26
<
Denis >
which then defaulted to extlinux and the old 5.4.61 kernel
18:27
<
tekkamanninja >
ha , my friend have checked the image , "run boot_grub" should work , I guess, can you try it again ?
18:29
<
tekkamanninja >
I have rebased the kernel to 5.16.0-rc3 , works well for now , you can build it by nezha_fedora_defconfig
18:31
<
Denis >
as it turns out, grub.cfg has "devicetree /aw_nezha_d1_1G.dtb" from one of my previous attempts
18:32
<
tekkamanninja >
ah , you are using wrong dtb
18:32
<
Denis >
let me set it to the original filename and try again
18:32
<
tekkamanninja >
please try devicetree /dtb-5.16.0-rc2/allwinner/sun20i-d1-nezha-1g.dtb
18:34
<
Denis >
I am using "/dtb-5.16.0-rc2/allwinner/sun20i-d1-nezha-1g.dtb" to match the filesystem layout
18:34
<
tekkamanninja >
yes
18:35
<
Denis >
I meant "/dtbs/5.16.0-rc2+/allwinner/sun20i-d1-nezha-1g.dtb"
18:36
<
tekkamanninja >
yes
18:36
<
tekkamanninja >
sorry , that is my test PATH,
18:36
<
tekkamanninja >
you are right
18:36
<
tekkamanninja >
. /dtbs/5.16.0-rc2+/allwinner/sun20i-d1-nezha-1g.dtb is right
18:37
<
Denis >
looks like this time it managed to start booting instead of resetting
18:38
<
Denis >
I am capturing a log of that as well, in case you need it
18:38
<
Denis >
in fact, it should not be difficult to provide you with serial console access, if it helps
18:39
<
tekkamanninja >
have you got into login:
18:40
<
Denis >
yes. I was delayed by the fact it now has a different MAC address and a different DHCP lease
18:40
<
Denis >
it is kernel 5.16.0-rc2+ and it sees 1GB of RAM, which is success
18:41
<
Denis >
also load average is below 1.0, which is useful
18:41
<
tekkamanninja >
:-)
18:42
<
Denis >
tekkamanninja: thank you for the help, is there any other testing to do before you are confident about the bugfix?
18:43
<
Denis >
I can try booting with sun20i-d1-nezha.dtb if it helps
18:43
<
Denis >
it does not say the amount of RAM in the filename, is it because it is supposed to auto-detect, or because 2GB is the default?
18:44
<
tekkamanninja >
no, I think I should update uboot/linux repo in my github, Great thanks for letting me kwon there is a bug in my wiki and image :-)
18:45
<
tekkamanninja >
sun20i-d1-nezha.dtb is the one without ddr size info
18:46
<
Denis >
is it supposed to just work?
18:46
<
tekkamanninja >
so this info is decided by uboot in build time
18:47
<
tekkamanninja >
so for booting kernel , sun20i-d1-nezha.dtb should work
18:47
<
Denis >
let me test whilst everything is hot
18:47
<
tekkamanninja >
I did not notice that , thanks to you to let me know :-)
18:48
<
tekkamanninja >
sure , thanks :-)
18:49
<
tekkamanninja >
sun20i-d1-nezha.dtb == sun20i-d1-nezha-1g.dtb == sun20i-d1-nezha-2g.dtb == sun20i-d1-nezha-512m.dtb
18:49
<
tekkamanninja >
those should be the same
18:49
<
tekkamanninja >
I only tested 2G , my bad :-(
18:50
<
Denis >
devicetree sun20i-d1-nezha.dtb boots OK and the amount of RAM is 1GB
18:51
<
tekkamanninja >
Denis: I need to sleep for RISC-V open hour in 8 hours, sorry, ttyl :-) Great thanks for your testing :-)
18:51
<
tekkamanninja >
3am for me now
18:52
<
Denis >
contents is different in all four .dtb files, as far as sha1sum sees it
18:52
<
Denis >
size is the same for -1g, -2g and -512m
18:53
<
tekkamanninja >
yes, you can see the code in my uboot repo
18:53
<
Denis >
tekkamanninja: thank you very much, when would be a good time to talk later?
18:54
<
tekkamanninja >
or slack
18:55
<
Denis >
alright, talk to you later
18:55
<
tekkamanninja >
Wei Fu on slack
18:58
<
Esmil >
tekkamanninja: speaking of u-boot. i was preparing an sd-card for the starlight board yesterday, and i noticed that having an EFI partition with an efi-executable at the standard location /EFI/BOOT/BOOTRISCV64.EFI doesn't seem to work.
18:59
<
Esmil >
the Fedora image has grub at /EFI/fedora/grubriscv64.efi, but it would be great if the standard path worked too
19:08
<
tekkamanninja >
Esmil: Great to see you here
19:08
<
tekkamanninja >
which image you are using ?
19:09
<
tekkamanninja >
for starlight board, we are alway using grub to boot kernel
19:09
<
Esmil >
Right, but not everyone is using the fedora image. So it would be nice if u-boot would just default to load the efi-executable at the standard location
19:10
<
Esmil >
i just created the efi and root partitions myself and bootstrapped my own root filesystem
19:11
<
tekkamanninja >
ah, I got your point, this depends on Image boot(partition)/boot/uEnv.txt
19:12
<
tekkamanninja >
if the image is for debian/gentoo/arch ..... , they should modify uEnv.txt in their own image
19:13
<
Esmil >
hmm.. bit this file is not in the latest fedora image, so there must be some default built into u-boot
19:13
<
Esmil >
i just think this built-in default should be the distro-neutral efi-defined standard path
19:13
<
tekkamanninja >
yes, in a config file
19:14
<
tekkamanninja >
they can change it by setevn and saveenv in to flash
19:15
<
Esmil >
right it's definitely possible to change. again i just think the default should be what efi-defined path
19:15
<
tekkamanninja >
partition info for uEnv.txt is set in uboot, but I will fix it ASAP, I will scearch the file instead of hard-coding the partition
19:16
<
Esmil >
cool, thanks
19:16
<
tekkamanninja >
yes, you are right , I should fix it in uboot header file
19:17
<
Esmil >
yes, i think that's the best place
*thumbs up*
19:21
<
tekkamanninja >
include/configs/starfive-jh7100.h
19:21
<
tekkamanninja >
#define STARLIGHT_FEDORA_BOOTENV \
19:21
<
tekkamanninja >
"bootdir=/boot\0" \
19:21
<
tekkamanninja >
"bootenv=uEnv.txt\0" \
19:21
<
tekkamanninja >
"mmcdev=0\0" \
19:21
<
tekkamanninja >
"mmcpart=3\0"
19:21
<
tekkamanninja >
there is the place I should fix
19:22
<
tekkamanninja >
I will fix ASAP , will let you kown
19:22
<
tekkamanninja >
Esmil: :-)
19:22
<
tekkamanninja >
Esmil: :-)
19:23
<
davidlt[m] >
I wonder how it works, because I don't see anything Fedora specific.
19:24
<
tekkamanninja >
davidlt[m]: it import uEnv.txt frist ,
19:24
<
tekkamanninja >
bootcmd is in uEnv.txt
19:25
<
davidlt[m] >
I mean most boards don't support uEnv.txt
19:25
<
tekkamanninja >
so you can do whatever you want in uEnv.txt
19:25
<
tekkamanninja >
sysboot?
19:25
<
tekkamanninja >
some board can not load file from storage ???
19:26
<
davidlt[m] >
no, Unleashed, Unmatched, etc. don't have uEnv.txt support
19:26
<
tekkamanninja >
uEnv.txt just a file or a env storage
19:27
<
tekkamanninja >
davidlt[m]: we may can update the uboot to make it supported
19:27
<
davidlt[m] >
Yes, but there is no generic "uEnv.txt" support in U-Boot
19:27
<
davidlt[m] >
Those are board specific and not all the boards support it (most don't)
19:27
<
tekkamanninja >
Unmatched uboot even support ssd in M.2
19:28
<
davidlt[m] >
Yes, but that's a different thing.
19:29
<
davidlt[m] >
But I wonder in general how it works
19:29
<
davidlt[m] >
I don't see anything distro specific in U-Boot and there aren't any Fedora specific patches.
19:29
<
tekkamanninja >
we can import env from SD or other storage , because uboot(firmware) is board specific , right ?
19:29
<
Esmil >
tekkamanninja: oh, i see now what you mean. that file hardcodes that u-boot looks at mmcblk0p3 for the uEnv.txt
19:30
<
tekkamanninja >
Esmil: yes
19:30
<
davidlt[m] >
Typically uEnv.txt is the same as extlinux.conf, which tends to be mmc 0:0
19:30
<
tekkamanninja >
Esmil: that is wrong , I should scearch uEnv.txt file in SD file instead of hard-coding
19:30
<
Esmil >
I guess that's fine, but it it's not found then it would be great if it would fall by to just trying the standard /EFI/BOOT/BOOTRISCV64.EFI
19:31
<
Esmil >
tekkamanninja: yes, that would be good too
19:31
<
Esmil >
*fall back to
19:31
<
tekkamanninja >
davidlt[m]: yes
19:32
<
davidlt[m] >
Distroboot command will do: "efi/boot/"BOOTEFI_NAME"
19:32
<
davidlt[m] >
#define BOOTEFI_NAME "bootriscv64.efi"
19:32
<
tekkamanninja >
where is the "uEnv.txt" is board specific , I guess
19:32
<
Esmil >
davidlt[m]: yeah, so I guess I'm just asking for jh7100 u-boot to be more distroboot-like ;)
19:33
<
tekkamanninja >
Esmil: yes,
19:34
<
davidlt[m] >
I wonder how BOOTX64.EFI and gets to grubx64.efi
19:34
<
tekkamanninja >
davidlt[m]: ah, yes, you are right, let me fix it and let you and Esmil know this ASAP
19:34
<
davidlt[m] >
I wonder if bootriscv64.efi is kinda a shim that later loads grub efi or something
19:35
<
tekkamanninja >
I should use Distroboot
19:35
<
Esmil >
tekkamanninja: that be great if you can make that work, yes
19:35
<
tekkamanninja >
I knew there is a command "Distroboot"
19:36
<
tekkamanninja >
Esmil: yes, I will use that instead of my ugly hack :-)
19:36
<
tekkamanninja >
davidlt[m]: thanks for your info , that is helpfull
19:37
<
davidlt[m] >
ah, on x86_64 /boot/efi/EFI/BOOT/BOOTX64.EFI is a shim
19:38
<
tekkamanninja >
davidlt[m]: Esmil: see you guys in riscv open hours soon, I need to go to bad now
19:39
<
tekkamanninja >
glad to see you are all here :-)
19:39
<
Esmil >
tekkamanninja: oh yeah. if you're in china it must be past midnight. sleep well ;)
19:39
<
tekkamanninja >
3:40am
19:40
<
davidlt[m] >
go get some sleep :)
19:42
jcajka has quit [Quit: Leaving]
19:42
<
tekkamanninja >
davidlt[m]: going to sleep now, Great thanks for your works(rocks) BTW
19:49
bkeys has quit [Remote host closed the connection]
20:01
bkeys has joined #fedora-riscv
20:04
tekkamanninja has quit [Ping timeout: 256 seconds]
20:10
tekkamanninja has joined #fedora-riscv
20:52
drewfustini has quit []
20:52
drewfustini has joined #fedora-riscv
23:20
ahs3 has quit [Remote host closed the connection]
23:53
nb has quit [Read error: Connection reset by peer]
23:54
nb has joined #fedora-riscv