Is Devuan Ryzen?

ghaverla · 2017-12-08 00:44:09

I seen the
Ascii installer image? by mskala
thread, and I have the Devuan-Live and refracto ISOs downloading. I also have the full Debian-Buster alpha2 installer downloading. But if I have to install that, it may be a while before I can upgrade to Devuan. :-)

I just finished assembling a new machine:
Asus x370 Prime Pro
Ryzen 1600X
32 GB RAM (G.Skill)
RX-560 GPU
boot device is a partition on a NVMe M.2 500GB "SSD"
4 Seagate 2TB rust drives (to be set up as a btrfs RAID-10 for /home)
Currently using a BluRay optical, will be changed to an ordinary DVD after install (plus a couple of weeks)

It powered up, and I could see all the fancy things in the BIOS. I could select the optical device (as SATA) and boot both the Devuan install DVD and the Devuan netinst CD. You get to the splash screen, and the keyboard works to let you select expert install. First set of menu's to set the language, everything locks up. Or at least appears to. Nothing on the keyboard seems to work. I never tried getting to another text console to see if the logs or dmesg is showing anything.

Hopefully one of those two ISO's will work. And then we can close this thread.

Gord

golinux · 2017-12-08 02:52:04

ascii is not buster.

beowulf will mirror buster

ascii mirrors stretch

So it is not currently possible to upgrade from debian buster to anything devuan except maybe ceres (sid)

ghaverla · 2017-12-08 03:28:18

I kind of thought that was the situation. That if I had to use the Debian-Buster alpha2 installer to put linux on this machine, that I would have to wait for a while before I could upgrade it to Devuan.

Of course, seamonkey got cut off for too long in downloads, and so I had to use wget to finish things. I have the Devuan-Live and Refracta ISOs downloaded. I will burn and try them, starting with the refracta as if I can format filesystems from the Live disk, I want to have a Linux-4.x kernel to make the btrfs filesystems.

I am hoping I don't have to use this Debian-Buster thing.

golinux · 2017-12-08 03:58:38

One of the Devuan devs reported this a few weeks ago:

Ran a refracta ascii snapshot with 4.12 on an AMD Ryzen7 bench test system with ECC ram and ECC works properly!!!

fsmithred is providing some unofficial ascii isos. Hopefully he'll leave some urls for you here.

ghaverla · 2017-12-08 04:51:14

I tried to get to an text terminal with the NET ISO, and nothing happened, so I am guessing that nothing keyboard or mouse was working.

Just now, I burned the refracta to a DVD and booted it. It got through the boot fine, but once everything settled there was no response to keyboard or mouse. Just like the NET ISO and DVD install images.

It's getting late, time to go to sleep.

golinux · 2017-12-08 06:35:18

@ghaverta . . . The ascii isos are only unofficially available atm. It's unlikely the Jessie ones will be compatible with newer hardware. Maybe with a backported 4.9 kernel . . .

fsmithred · 2017-12-08 12:59:44

You need the ascii-backports kernel. (4.10 or newer, so plain ascii won't do it.) I did put an iso up, but it was meant for one person. It was only there for a few hours. Been thinking about putting up another one. (One that doesn't have a copyrighted desktop background) But I wanted to wait until after the regular ascii isos were up before I do the bpo. Guess that's not helpful for all you razor-edge ryzenistas.

Lemme figure out which VM has the right system, whether or not 'less' is installed, add the bpo kernel, and upload it. I guess it'll happen later today.

ghaverla · 2017-12-08 13:02:21

Okay, I will check back later today.

The Debian-Buster alpha2 install DVD was finished downloading when I woke up (and no coffee yet), so I burned it and tried it. It boots and lets me choose a language, country, UTF-8 locale (and then I stopped). So perhaps that information is useful?

Have a great day!

ghaverla · 2017-12-08 13:06:26

Razor edge? Some of these parts have been sitting on the shelf for 6 months, waiting for winter to allow time to put things together. :-)

But, if you want me to try a bunch of things, I can try to do that. I shouldn't run out of disk space with 4TB of completely empty RAID-10 disk (rust) space.

fsmithred · 2017-12-08 17:01:51

I'm still in the neolithic age...

model name	: AMD Athlon(tm) 64 X2 Dual Core Processor 3800+

(I actually have something a lot newer than that for building isos.)

Here's an ascii live iso with 4.12 backports kernel. There's not a lot installed, but it should be enough to get you going. I'd hate for anyone to be forced to use debian when they don't want to.

I'm sure others will see this and want it, so here are some important notes...

First of all, this is not an official release (of devuan or refracta.) See that it's in the EXPERIMENTAL folder. If something doesn't work or goes wrong, don't be surprised. Feel free to ask for help, but there should be no whining or complaining.

The backports repo is commented out in sources.list. There are also a few wireless firmware packages installed from backports. Regular ascii firmware packages can be found in /firmware. Also, contrib and non-free are enabled.

The installers should see your nvme disks and handle it correctly. If you use the cli installer, you'll have to type in the device name when you select a partition for installation. If you prefer, you can use this live iso to do a debootstrap install of devuan. I forgot to include the instructions for that.

There's no display manager. You should not need to log into a live session, but in case you do,

login name: user
password: user.
Root password is root.
Run 'startx' to get into an openbox session.

grub-efi-amd64 is installed. If you are installing to legacy bios, I recommend using the cli installer (run 'refractainstaller' in a root terminal). It will find the grub-pc packages in the root of the filesystem and do the right thing.

http://distro.ibiblio.org/refracta/file … 8_1525.iso

There's no gpg signature, but here's the sha256sum of the iso.

$ sha256sum snapshot-ascii-bpo-20171208_1525.iso
22f4bd23457c9a3619243967824bee14625777cc60e586b5c892bf1da069f8e3  snapshot-ascii-bpo-20171208_1525.iso

ghaverla · 2017-12-08 19:41:32

Thanks a lot. I think I should be able to figure it out. I started with Linux with the 1.2.13 kernel (Yggdrasil, or something like that). Me and command lines go way back.

Some place in my old hardware pile, I have a pair of AMD CPUs that double as heaters. I think they are still on a motherboard, I don't remember if it was working when I replaced them, but it is likely they still work.

golinux · 2017-12-08 21:27:29

@ghaverla . . . looking forward to your report of how it went.

fsmithred to the rescue yet again!!

ghaverla · 2017-12-09 00:31:09

I went part way through the Debian install, thinking I could maybe get the NVME "SSD" and the 4 spinning rust disks partitioned. But the Debian installer doesn't really look like it is set up to do GPT disks "nicely". So, I bailed without writing any GPT partitions.

Oh, this was my first NVMe device. I had no idea what it would be called. There is a /dev/nvme and then there is a /dev/nvme0n1. One is not a symlink to the other, the first is a character device and the second (nvme0n1) is a block device, just like a disk. It is this second one that I partitioned.

I downloaded the latest System Rescue CD (because it has GPT fdisk tools from Rod Smith), and I booted that just fine. I put GPT partitions on all 5 devices, with 201MB EFI System partitions in position 1. The NVMe got a /boot (2), root (3), usr (4), var (5), usr/local (6), var/log (7), opt (8) and a scratch (9) partition. For 2 of the spinning rust disks, position 2 is swap (16G) and for the other 2 it is to be a RAID-0 (striped) /tmp. But those are all the same size (16GiB). The third partition is everything remaining, except for about 48 GiB, which works out to about 1.8TiB.

I then used the system rescue CD to mkswap -c on the two swap spaces.

I then switched over to fsmithred's DVD and it too booted just fine, to a GUI. I thought I was just going to have the text console on F1. It starts up an xterm of some kind as user. User doesn't have permission to run gdisk (which is in /sbin), so I 'su root' to get a root shell, and gdisk can find the 5 devices just fine. So it looks good to install, if I can remember how to do these things. :-)

It is probably going to be easier just to install everything into a single partition, which is what that scratch partition is for. I can then put a btrfs partition on the root, and set up the directories needed to mount other partitions on. Next I probably need to setup the RAID0 /tmp, and I typically use ext2 for /tmp. The /usr, /var, /usr/local, /var and /opt partitions are all 32GiB in size (probably too big), and I will format them with ext4. Which leaves setting up /home as a btrfs RAID10, which is a completely new process to me (I've only ever used single partition btrfs for /home).

Anyway, what I've done for 2 machines with updated hardware since summer ended, is install to the scratch space, do an archive copy to all the mounted partitions, chroot to the new root, dpkg-reconfigure the grub packages, adjust /etc/fstab and some of the apt configuration in /etc to make use of my local apt-cache setup (I think it is actually apt-cacheng), and then I think I can boot into the new system. But maybe I missed a step or two in there.

(The two previous machines were running Debian/stable (Jessie). They were upgraded to Devuan/Jessie (which is still a 3. kernel), upgraded to Devuan/Ascii, had a 4. kernel installed and btrfs utilities, formatted all the various filesystems, and then copied from a scratch partition to the set of mounted partitions. I have one more machine to upgrade after this, the server for the LAN. And then everything on the LAN is Devuan. For now - I have a bunch of RPi coming in the future, and my routers run OpenWRT.)

It does look optimistic at this point.

ghaverla · 2017-12-09 01:35:26

This machine (and most of my machines) connect to the OpenWRT router via an unmanaged switch. But, when I looked at the switch before plugging in this new machine, all the ports are full.

Some of the disks I have booted, have attempted to configure networking, and been unsuccessful. I am beginning to think that unmanaged switches don't like to have old hardware removed and new hardware installed.

The GUI does have a menu item to use the refracta (I hope I spelled that correct) installer. I ran that, and it allowed me to pick the scratch partition, and it seemed to install stuff. But, a reboot would not get there, but maybe I need to mess with the BIOS on that. I then booted the system rescue CD, and attempted to have it boot into an OS on the machine. No joy there. But maybe it doesn't look in /dev/nvme* for OS's?

The router should only have WAN and a single LAN connection (to the switch). So I should be able to plug the network cable into the router, and maybe then dhclient (or whatever) will be able to get an IP, and be able to connect to the internet.

That should allow me using the system rescue CD, to boot into what refracta installed.

I do have another unmanaged switch here (never used) with 8 ports (more than the 5 the current switch has), because my network is growing. So I could invoke that.

But the ISO that fsmithred produced still looks like a good start. He doesn't want complaints, can I send teases? Emacs is not present. So I will guess that fsmithred is a vi person. :-) I seen a virc file from (I believe) Tom Christensen (one of the gods of Perl), and it was huge. So I have no doubt that vi works well for some people. One explanation I've seen on vi/emacs, is how long your fingers are. Sure. I don't think mine are that much longer than everyone elses. And I've never written any extension in LISP. So I guess I am not a _real_ emacs user.

Everyone can have problems. I think I am a long ways along the process of getting Devuan working on this "razor edge" machine. :-)

I am just missing a few details.

I think I like penguinista more than ryzenista. But, I had never run across the Ryzen version before.

Live long and prosper.
Gord

fsmithred · 2017-12-09 04:03:06

No emacs here. Just vi, which I do not like. I can tolerate vim, but plain old vi always screws me up. I inevitably hit an arrow key to move the cursor, and it spits out inappropriate characters. I would put vim on my isos, except for the fact that it takes up too much space, and I try to keep the isos down to CD size. But I'm perfectly happy with nano. If I forget how to use it, the instructions are right there in front of my face.

When you tried to boot after installing, did you get anything? A grub boot menu or even a grub prompt? If so, it should be possible to boot with the right commands.

Did you look at the efi partition to see if anything got copied there?
There should be /boot/efi/EFI/<devuan or refracta>/grubx64.efi

ghaverla · 2017-12-09 17:14:12

Good morning (or at least it is morning here).

I can attempt to boot the new installation from the SystemRescue CD (which I believe is still a 3. kernel?), and it complains about no /sbin/init. I think it is looking for some different /sbin/init than what it is finding (it is a binary executable on the NVMe). If I attempt to boot the NVMe from the BIOS, it just silently fails.

Or, it silently failed after today's messing around. And it might be that I messed things up.

I chrooted into that scratch partition, and ran dpkg-reconfigure on the grub packages. Three do nothing, and the one that does something complains about the efi being bad. I also ran update-initramfs -u

Just booting the SystemRescue CD, it will not bring the NIC up. lspci identifies it as a I211 that is using igb as a driver.

There are 4 directories under /boot/efi/EFI. They all contain a grubx64.efi file that is the same size.

I guess the next thing to do, is to reboot your disk, and reinstall from refracta, and maybe that undoes something I screwed up.

I once got asked to stand in and teach a UNIX class at a college. They insisted that everyone know vi. I figured that as long as they could find and use an editor, that was all that was needed. nano works fine.

ghaverla · 2017-12-09 18:20:13

Back again.

I tried running the install again. The refract error log has an error with grub-install of
/usr/lib/grub/i386-pc/modinfo.sh
doesn't exist. specify target or directory.

There is a /usr/lib/grub/x86_64-efi directory, and it has a modinfo.sh file.

The install doesn't use the uefi installer, as /sys/firmware/efi doesn't exist.

I tried looking in the refracta shell scripts and the config file for this /usr/lib/grub subdirectory name, and if it is there, I missed it.

But, from the BIOS, if I choose the NVMe to boot, it is just a silent failure. No grub menu or grub shell is ever displayed.

The refracta environment is also loading the igb driver for the network. I'm beginning to think the cable might be bad, so maybe I should check that.

fsmithred · 2017-12-09 22:11:48

Sounds like you're in legacy mode. If you boot in uefi mode, there should be a /sys/firmware/efi. Check your bios settings.

When you boot the refracta iso, do you get a grub boot menu or an isolinux boot menu? (with grub, 'e' will get you to edit the boot command, with isolinux, TAB will get you to edit the boot command.)

ghaverla · 2017-12-09 23:20:18

The boot menu I had been seeing, was the isolinux one.

I did fiddle with the BIOS, and told it to try UEFI first. So the boot menu I seen there (for yet another install :-) ) was grub. And the install was slightly different form before (as /sys/firmware/efi existed), and at the end I did the dpkg -i grub*deb in the chroot. grub complained about a missing BIOS boot block. I've been going by Rod's Books when using gdisk, and the EFI partition (#1) is 200MB in size at the beginning of the device. Maybe I misunderstood something about the setup? I thought this BIOS boot block thing was about hybrid disks, and this system is Linux only.

I then attempted to boot from the NVMe. I got a couple of messages before the mode change, and then basically nothing for a long time (20 minutes?). I've booted the refracta CD again, and mounted the NVMe partition in question, and /var/log/messages caught the boot sequence in some respect. I haven't read through it yet.

ghaverla · 2017-12-09 23:44:15

This is the stock BIOS on the motherboard, and I've had the motherboard for a while. Overclockers has a long thread on the BIOS of this card, and there have been many BIOS updates. At the moment, all the settings (except for the UEFI) are stock.

fsmithred · 2017-12-10 13:38:41

And the install was slightly different form before (as /sys/firmware/efi existed), and at the end I did the dpkg -i grub*deb in the chroot. grub complained about a missing BIOS boot block.

Yeah, that makes sense. You tried to install grub-pc in a uefi system. You only need to install the grub packages if the wrong grub is installed in the live system. In this case, grub-efi is installed, so you should boot in efi mode and let the installer handle to bootloader. You might not be able to use legacy bios with gpt partitions. And that uefi readme in the user's home is probably wrong and needs to be updated.

It might be better to use the cli installer, because that's a newer version, and I know it works well in a uefi install. Run 'refractainstaller' in a root terminal, or even better, (if you can stand to look at additional and distracting output on the screen) 'refractainstaller -d' to get a more verbose error log. One drawback is that you'll have to type in /dev/nvme-whatever to tell it which partition to install.

ghaverla · 2017-12-10 17:14:55

Okay, I can try the cli installer. I am getting okay at typing nvme0n1p9 anyway.

The /var/log/messages file had 2 booting sessions in it. I think the first is the booting of the VM, and the second is the (attempted) booting of the NVMe. The end of the messages file is a bunch of nulls. I don't see any obvious errors in it.

The NIC is getting the igp driver, but for whatever reason it never comes up. There seems to be problems with bad non volatile memory in the I211 NIC, but I am not seeing that in the log file.

Rebooting the VM, I see nothing in the log file or .xsession-errors about the NIC. WICD tries to do something, but I don't know why it is failing.

I downloaded lubuntu-17.10, and booted it to live-cd. It too uses igp for a driver, and this gets an IP. So I now know that the cable works. I've got some stuff to do outside, and then I'll come back and try to figure out what lubuntu is doing differently.

And try the cli installer.

ghaverla · 2017-12-10 19:02:59

running refractainstaller-uefi from cli

Everything goes fine and I get the message about chrooting. In the chroot I do a dpkg -l | grep grub, and it shows the grub-efi packages are installed. So I choose the second set of options.

Running grub-install, I get an error message that EFI directory doesn't exist. There is a /boot/efi/EFI directory, which has 4 subdirectories in it. Running find / -name EFI -print shows that this is the only EFI on the system. I do see that grub-install has an option to set the path to this directory.

Is grub-install expecting that nvme0n1p1 (the 200M EFI System partition) be mounted someplace? And so I need to indicate the mount point? I wonder what kind of formatting that partition has? Did gdisk format it?

Hmmm, things to research.

ghaverla · 2017-12-10 19:30:50

Researched a little. I put a FAT filesystem on /dev/vnme0n1p1 (200M partition), and mounted it on /mnt.

I then reran grub-install with the /mnt directory in the option to install the EFI stuff. It ran without error. I then ran update-grub, which also went without error.

I rebooted the machine, stopping in the BIOS. I picked to boot off the NVME, and it is just sitting there, with a big cursor blinking in the upper left corner. I'll let it sit a bit, and then get back to the BIOS. Maybe it is not treating the NVME as UEFI? I will look.

ghaverla · 2017-12-10 19:42:30

It seems to have been a UEFI issue again. Now that device is presented with 2 options in the BIOS boot menu, by itself, and as a parenthesissed string after "devuan". The second one causes a UEFI boot.

I see a grub menu, and choose the default (top) option. The boot starts, and then ends with a message about kvm being disabled by bios (at about 8 seconds?). And it just sits there.

So, go dig into BIOS and see if there is some kind of KVM setting.

The officially official Devuan Forum!

#1 2017-12-08 00:44:09

Is Devuan Ryzen?

#2 2017-12-08 02:52:04

Re: Is Devuan Ryzen?

#3 2017-12-08 03:28:18

Re: Is Devuan Ryzen?

#4 2017-12-08 03:58:38

Re: Is Devuan Ryzen?

#5 2017-12-08 04:51:14

Re: Is Devuan Ryzen?

#6 2017-12-08 06:35:18

Re: Is Devuan Ryzen?

#7 2017-12-08 12:59:44

Re: Is Devuan Ryzen?

#8 2017-12-08 13:02:21

Re: Is Devuan Ryzen?

#9 2017-12-08 13:06:26

Re: Is Devuan Ryzen?

#10 2017-12-08 17:01:51

Re: Is Devuan Ryzen?

#11 2017-12-08 19:41:32

Re: Is Devuan Ryzen?

#12 2017-12-08 21:27:29

Re: Is Devuan Ryzen?

#13 2017-12-09 00:31:09

Re: Is Devuan Ryzen?

#14 2017-12-09 01:35:26

Re: Is Devuan Ryzen?

#15 2017-12-09 04:03:06

Re: Is Devuan Ryzen?

#16 2017-12-09 17:14:12

Re: Is Devuan Ryzen?

#17 2017-12-09 18:20:13

Re: Is Devuan Ryzen?

#18 2017-12-09 22:11:48

Re: Is Devuan Ryzen?

#19 2017-12-09 23:20:18

Re: Is Devuan Ryzen?

#20 2017-12-09 23:44:15

Re: Is Devuan Ryzen?

#21 2017-12-10 13:38:41

Re: Is Devuan Ryzen?

#22 2017-12-10 17:14:55

Re: Is Devuan Ryzen?

#23 2017-12-10 19:02:59

Re: Is Devuan Ryzen?

#24 2017-12-10 19:30:50

Re: Is Devuan Ryzen?

#25 2017-12-10 19:42:30

Re: Is Devuan Ryzen?

Board footer