The officially official Devuan Forum!

You are not logged in.

#1 Re: Desktop and Multimedia » What AMD graphics card (< US$250) works on Beowulf? » 2021-03-10 04:06:47

It appears when I installed the backports 5.10 kernel that I used the meta package.  This is what 'dpkg -l' says for it:

ii  linux-image-amd64                       5.10.13-1~bpo10+1                   amd64        Linux for 64-bit PCs (meta-package)

Again, thank you both for the very helpful guidance.

#2 Re: Desktop and Multimedia » What AMD graphics card (< US$250) works on Beowulf? » 2021-03-09 19:13:29

Thanks for the great info about the meta package.  During installation, I just followed the default path.  Now that I know about the meta package structure, I'll make sure to install the meta package from backports.

#4 Re: Desktop and Multimedia » What AMD graphics card (< US$250) works on Beowulf? » 2021-03-09 01:44:15

That's a very good point.  Thank you for suggesting the 5.10 LTS branch from the beowulf-backports repository.  So far, today's current 5.10 kernel from beowulf-backports appears to be working well.  I partitioned the big HDDs and created the RAIDs.  The big RAID is doing its initial sync as I type (and for the next ~4+ hours).

I did a web search or two about how to add a backports repository.  One thing was not clear from the relevant web pages those searches found.  Does apt-get pick up updated kernel packages from backports automagically?  When I tried "apt-get -d -t beowulf-backports update", it wanted to update an enormous boatload of things, which I immediately rejected.  (Packages from the stable, non-backports repositories had already been fully updated.)

#5 Re: Desktop and Multimedia » What AMD graphics card (< US$250) works on Beowulf? » 2021-03-07 01:11:17

Thank you both for those very helpful answers.  In case it might help someone else who might happen upon this thread:

After a long wait for the CPU (and some problems with a now-EOL SAS/SATA card) the machine is running.  (I had wanted a 5900X CPU, but it's still not in stock anywhere at a price even close to list.)  With a temporary (2200G) CPU, I upgraded the motherboard's BIOS to a version that supports the 5800X CPU.  That temporary CPU also allowed testing of some other issues.  I installed Beowulf with the temporary CPU.

For the kernel, I downloaded the 5.11.2 source from kernel.org to my old machine, ran "make oldconfig" based on the config file from the stock Beowulf kernel (accepting the defult for all questions), cleared the value of CONFIG__SYSTEM_TRUSTED_KEYS, ran "make -j 6 bindeb-pkg", waited an hour for the build to finish, installed three DEBs (not the one with "dbg" in its name), installed the 5800X CPU, and booted it up.  So far, so good on the kernel.

Getting the GPU up to speed is still pending--one step at a time.

#6 Desktop and Multimedia » What AMD graphics card (< US$250) works on Beowulf? » 2020-12-20 03:38:26

rmrichesjr
Replies: 11

What AMD graphics card (up to US$250 price) will work with Beowulf?

I'm preparing to build a new system with an Asus Pro WS X570-ACE motherboard and either a 9 5900X or a 7 5800X CPU, with Devuan Beowulf as its first OS.  I plan to use open-source drivers, so I'm avoiding the 'PRO' cards.  After considerable searching, I have yet to find a cross-reference that says what AMD GPUs work with what version of kernel, xorg, etc.  Any pointers or suggestions?

Thanks,

Robert

#7 Other Issues » package for libvirtd init script » 2020-11-05 04:58:31

rmrichesjr
Replies: 2

On a Devuan Ascii system, what is the package that installs the /etc/init.d script for libvirtd?  Is it libvirt-daemon-system, or is it libvirt-daemon-system-sysv?

Thanks,

Robert

#8 Re: Other Issues » [SOLVED]Package updates in ascii? » 2019-03-31 04:11:02

I found the reason for my case of not getting updates.  My /etc/apt/sources.list had originally only listed the cdrom (well, actually DVD to be more precise) as sources.  That's likely a side effect of not configuring the network in the installer.  When I noticed a Raspberry Pi 3 running Devuan Ascii was getting updates that my main X86 system was not getting, investigation yielded that the Raspberry Pi's /etc/apt/sources.list had separate lines for "ascii", "ascii-updates", and "ascii-security" (along with a commented-out line for "ascii-backports").  After changing my /etc/apt/sources.list to mimic the one on the Raspberry Pi, I had ~164MB of updates to download, including Firefox 60.6.1esr.

#9 Re: Hardware & System Configuration » Raspberry Pi 3B+ network not working at boot with Ascii » 2019-02-16 22:50:32

Thanks for that pointer.  The case described there is for when "... eth0 is not connected to a network,"  In my case, my eth0 _is_ connected to a network, the same cable, switch/router port, and such where the X86 machine running Devuan Ascii boots with_out_ any DCPDISCOVER delay.  The only relevant hits found by a couple of searches were the 2660 and 1688 threads already discussed.  The 2660 thread has one solution of removing all mention of eth0 from /etc/network/interfaces, which would leave that NIC entirely unconfigured.  That thread also points to the 1668 thread.  In the 1668 thread, post #2 says to, "Replace 'allow-hotplug' with 'auto'.  That ('auto') is what I have.  Post #18 also says to remove all mention of eth0 from /etc/network/interfaces, but again that would leave that NIC entirely unconfigured, which is not the desired state when eth0 is connected to a network.

#10 Re: Hardware & System Configuration » Raspberry Pi 3B+ network not working at boot with Ascii » 2019-02-16 22:20:15

Thanks for the reply.  I read through both of those threads and the Debian bug report referenced in the latter thread.  Here's a little more detail on my situation:

I'm using only eth0 (well, and lo, of course), not wlan.

For essentially all practical purposes, I always have the network cable plugged in when I boot the Raspberry Pi.

I do not have Network Manager installed.

I boot to text console, not graphical login manager.

My directory /etc/network/interfaces.d is empty, and I have these four non-comment, non-empty lines in /etc/network/interfaces:

  auto lo
  iface lo inet loopback 
  auto eth0
  iface eth0 inet dhcp

Those four lines are identical to the corresponding lines in a 64-bit X86 machine that also runs Devuan Ascii.  The X86 machine does not have any delays.

Those four lines are also exactly how the datacenter network gurus where I work have always configured Debian-based systems for dynamic IP address on eth0.  Following the examples of said gurus, that's the way I have configured Debian-based systems I have worked with.  Except for cases whether the DHCP server (run by corporate IT) is slower than slow, I have not seen similar delays on those machines--only on the Raspberry Pi.

The X86 machine I mentioned and Raspberry Pi were tested with the same network cable, same port on the switch and router, etc.

When I boot the Raspberry Pi, the LEDs on the RJ45 jack behave curiously.  Initially, both are off.  Then, the yellow/amber LED on the right blinks on a couple of times and then goes back off--doing this once shortly before the DHCPDISCOVER line appears on the console and then again shortly after that line appears.  After the DHCPDISCOVER line appears on the console, and I think after the second round of yellow/amber blinks, the tree LED on the left comes on and stays on.

My hunch is there is something different about how the Raspberry Pi's NIC hardware, firmware, and/or drivers are initialized (relative to most X86 desktops and servers) that causes the initial dhclient process invoked by /etc/init.d/networking invoking "ifup -a" to fail/hang.

Is anyone acquainted well enough with the Raspberry Pi 3B+ NIC hardware/firmware/drivers to shed more light on why the Raspberry Pi shows the hang while the X86 machine does not?

I have a hypothesis that one possible workaround would be to remove the 'auto eth0' line from /etc/network/interfaces and add an "ifup eth0" or perhaps "ifup -a" command to /etc/rd.local.  Any second opinions on whether this would likely be reliable?

#11 Re: Other Issues » [SOLVED]Package updates in ascii? » 2019-02-16 21:45:41

Thank you, golinux, for your #22 reply.  Bug fixes and security updates are all I was looking for.  The Firefox ESR updates in question are security updates and bug fixes.  Interestingly, even though I had done nothing to my apt configuration, this morning there were over a hundred package upgrades available.  Perhaps, the mirrors might not have been syncing.

#12 Hardware & System Configuration » Raspberry Pi 3B+ network not working at boot with Ascii » 2019-02-14 04:19:19

rmrichesjr
Replies: 4

When booting a Raspberry Pi 3B+ running Devuan Ascii from image devuan_ascii_2.0.0_arm64_raspi3.img.xz plus package upgrades plus installing a bunch of non-default packages, the system tries to get an IP address from DHCP, but DHCP does not usually succeed.  About 3/4 of the time, the boot process pauses after printing one DHCPDISCOVER line, then a few/several dozen seconds later several more DHCPDISCOVER lines plus one "No DHCPOFFERS received" line print, and the boot process goes on without eth0 being up.  Later, after everything else has booted, a simple (as root) ifdown eth0 followed by ifup eth0 works within a second or two.  About 1/4 of all boots, DHCP gets an IP address in a few/several seconds, and the system boots with eth0 having a functional IP address.

Is this a known issue with DHCP and/or the Ethernet adapter driver?  Is there a known solution or workaround?  Is there a kernel module that should be listed to be loaded earlier in the boot process?

Thanks.

#13 Re: Other Issues » [SOLVED]Package updates in ascii? » 2019-02-09 21:59:51

It has been at least two or three weeks again since I have seen an ascii update.  For instance, IIRC, Firefox ESR has two updates released (60.4.0 and 60.5.0) but not yet visible through Devuan ascii (60.3.0esr).  Or, did I miss a memo that ascii has been EOLed?

#14 Other Issues » Seeking a replacement for xmacroplay » 2019-01-02 04:37:26

rmrichesjr
Replies: 1

Is there a replacement for xmacroplay?  It's a marvelously useful utility that uses the XTest facility and allows scripting of control of GUIs.  It can move the cursor to specified pixel positions, pause for specified durations, press and release buttons and/or keys, and enter strings.  It has served me well for many years, including on Debian 7/Wheezy and Slackware 14.2.  However, on a new Devuan Ascii installation, the 32-bit binary (with the necessary 32-bit library packages installed) often behaves slightly erratically.  When entering a string, occasionally it omits or repeats characters, sometimes repeating multiple characters in a row (turning "blog" into "blloog", for example).  I tried recompiling it, but the code is so old the modern compilers choke on it (include <iostream.h>, using a string constant as a (non-const) char array, and at least one more serious issue.

Any suggestions?

Thanks in advance.

#15 Hardware & System Configuration » scanimage + libusb = "Error during device I/O" » 2019-01-02 04:24:43

rmrichesjr
Replies: 1

On a new Devuan Ascii installation, my HP ScanJet 6300C is seen by sane-find-scanner at some "libusb:00N:00M".  If I do "scanimage --help -d hp:libusb:00N:00M" (same numbers returned by sane-find-scanner), I get a list of scanner-specific options that appear to be reasonable.  However, if I attempt "scanimage -d hp:libusb:00N:00M" (same numbers returned by sane-find-scanner), I get "Error during device I/O".  The scanner worked fine with Debian 7/Wheezy.  Slackware 14.2 gave me the "Error during device I/O", appearing to be the same symptom as with Devuan Ascii.  The scanner is between 15 and 20 years old, but it works fine from a Raspberry Pi (itself and its OS a few years old) with essentially the same commands as give the "Error during device I/O" on Slackware 14.2 and Devuan Ascii.  My hunch is bit rot has caused a bug to creep into libusb and/or scanimage when working with hardware that old.

Any suggestions?

Thanks in advance.

#16 Re: Installation » Devuan ASCII installer shows incorrect existing partition(s) » 2018-12-22 01:15:24

kuleszdl wrote:

Hi and welcome. I think it could help the devs the reproduce the issue if you could attach your actual MBR with which the failure occurs.

Yes, that would be helpful.  I'll be happy to provide that.  How many bytes from the start of the disk image are needed?  Does this BB have a mechanism to attach binary data--or--in what form should it be attached?

#17 Re: Installation » Devuan ASCII installer shows incorrect existing partition(s) » 2018-12-20 18:06:18

A workaround is to uninstall SBM (Smart Boot Manager), install Devuan (installing Devuan's GRUB in the /boot partition), then install and configure SBM again.

#18 Re: Installation » Devuan running very slow on Raspberry PI 3 b+ » 2018-12-18 19:35:28

A couple of suggestions until somebody who really knows what they're talking about shows up:

1) Use "top" to see whether something is hogging the CPUs.

2) Use "vmstat 3" to see whether something is hogging disk bandwidth and to see what disk bandwidth you are getting.

#19 Installation » Devuan ASCII installer shows incorrect existing partition(s) » 2018-12-18 19:32:54

rmrichesjr
Replies: 3

(First time Dev1Galaxy poster; 2-3-time refugee from systemd; reasonably experienced with several Linux distributions;)

While attempting to install Devuan ASCII on a test VM, the 'Partition disks' step shows an incorrect existing partition and does not allow me to use the correct existing partitions.  Eventually, I need to install to a physical machine with existing MBR-style partitions (and mdadm-style per-partition RAID10) with an existing installation of SBM (smart boot manager).  The test VM is a simplified mock-up of the eventual physical machine.

The symptoms are independent of whether I use the graphical installer, non-graphical installer, or Advanced Options -> Expert Install.  Symptoms are also independent (for the latter case) of whether I load additional components cfdisk-udeb, mbr-udeb, and parted-udeb.  When I get to the 'Partition disks' step, I choose manual partitioning (tried other options, but none lead anywhere useful).  The installer shows "Virtual disk 1 (vda) - 21.5GB Virtio Block Device" and "#1 21.5 GB fat16".  That appears to indicate the installer mistakenly thinks there is one 21.5 GB FAT16 partition on virtio block device /dev/vda.

In reality, there are four MBR-style partitions on the device: three marked ID 83 (Linux), and one marked ID 82 (Linux Swap / Solaris).  In the expert installer, if I execute a shell, both /proc/partitions and "fdisk -l /dev/vda" show the correct four partitions.  However, the installer still insists it sees only one 21.5GB partition formatted FAT16 (which I'm not sure is even possible given FAT16 size limitations).

Any suggestions to force the 'Partitions disks' installer stage to see the correct existing four MBR-style partitions?

Thanks,

Robert

Board footer

Forum Software