IRC channel logs

<AidenIsik>damo22: I looked into using AHCI as suggested yesterday to fix my boot problem.

<AidenIsik>My board's southbridge (Intel ICH7) doesn't support AHCI.

<damo22>sneek: later tell AidenIsik if you want to be sure if your southbridge supports AHCI or not, run this command: sudo setpci -s 0000:00:1f.0 e4.b

<sneek>Will do.

<damo22>#define FDVCT 0xe4

<damo22>#define AHCI_UNSUPPORTED (1 << 3)

<hwpplayer1>which file ? which line ?

<hwpplayer1>damo22: ^

<damo22>its from coreboot/src/southbridge/intel/i82801gx/i82801gx.h

<hwpplayer1>Thanks

<damo22>if the return value has bit 3 set, it does not support AHCI

<hwpplayer1>where is the source code of sneek bot of #hurd channel ?

<AidenIsik>damo22: It definitely doesn't support AHCI (AHCI was never supported on the base ICH7, and the result of that command does in fact have bit 3 set (0xa8)). Surely IDE can be made to work?

<sneek>Welcome back AidenIsik, you have 1 message!

<sneek>AidenIsik, damo22 says: if you want to be sure if your southbridge supports AHCI or not, run this command: sudo setpci -s 0000:00:1f.0 e4.b

<damo22>okay

<damo22>as far as i know, piixide wasnt fully working or had some issues

<damo22>but we should fix it

<damo22>do you have an error log?

<AidenIsik>nope, I'm attempting to boot a live environment

<AidenIsik>I can try to record the errors but I'm aware that's not an ideal log format at all

<damo22>was it lost interrupt?

<AidenIsik>yes

<damo22>hmm

<damo22>are you booting the latest version??

<damo22>s/??/?

<AidenIsik>Whatever the Debian nightly from a few days ago is using

<AidenIsik>I can check the exact version

<damo22>maybe you should try debian stable

<damo22>as in the latest released version

<damo22>the daily cd images are not always working

<AidenIsik>Sure, I can try that

<damo22>there is no livecd

<damo22>there is an installer or a disk image

<AidenIsik>yes that's what I meant

<AidenIsik>It's the installer

<damo22>did you burn an actual cd?

<AidenIsik>no, I flashed it to a drive and plugged it into a SATA port

<AidenIsik>I don't have a CD drive

<AidenIsik>And from what I understand USB booting does not work

<damo22>if you are doing that, it might fail, you may as well flash the disk image directly and boot it

<AidenIsik>I will try the disk image after I try the stable installer then

<AidenIsik>But this seems like a kernel issue so I'm not sure it will help

<damo22>no its not kernel

<damo22>disk driver is in userspace

<AidenIsik>yeah, microkernel whoops

<AidenIsik>wrong terminology

<AidenIsik>driver issue

<damo22>do you have a way to plug the raw disk into a different host

<AidenIsik>i have some Toshiba Satellite I could probably plug it into

<damo22>i mean how did you "flash" it

<AidenIsik>dd

<damo22>from where

<AidenIsik>Linux

<damo22>so you have linux installed already on the target machine?

<AidenIsik>yes

<damo22>and a second physical disk inside?

<AidenIsik>yes, the main drive is an SSD and I plug in the HDD next to it

<AidenIsik>HDD has the installer on it

<damo22>dont bother with the installer, just flash the disk image

<AidenIsik>alright

<damo22>the .img.tar.gz download needs to be extracted to .img

<damo22>then you can flash it to disk with dd

<AidenIsik>yeah just extracting and flashing it now

<AidenIsik>sudo dd if=debian-hurd-amd64-20260314.img of=/dev/sdb status=progress && sync

<AidenIsik>bs=1M too

<damo22>careful with raw devices

<damo22>make sure its the right one obv

<AidenIsik>yeah it definitely is

<AidenIsik>i'm not new to dealing with GNU/Linux, just Hurd :p

<damo22>:D

<damo22>that should just boot

<damo22>what hardware is it, ICH7 is quite dated

<AidenIsik>It's a Corebooted GA-G41M-ES2L

<damo22>haha, i ported that

<AidenIsik>Core2Quad Q9650, GT1030, 8GB DDR2

<AidenIsik>Damn really??

<damo22>yeah the G41 chipset

<AidenIsik>oh nice

<AidenIsik>small world

<AidenIsik>still waiting for dd to finish

<damo22>the fan still not working on that board?

<AidenIsik>nah my fan works

<AidenIsik>system and cpu fan

<damo22>but it spins at full tilt?

<AidenIsik>if it is it is a very quiet fan

<AidenIsik>i'm unsure

<AidenIsik>well dd is finished now

<AidenIsik>time to test this

<AidenIsik>booting now...

<AidenIsik>hopefully this works

<AidenIsik>ext2fs: part:2:device:wd0: Input/output error

<AidenIsik>://

<AidenIsik>will try again

<AidenIsik>Yeah no it doesn't boot at all

<AidenIsik>only drive in the system

<AidenIsik>i can send a picture of the end of the logs if that is useful at all

<rrq>wouldnt sdb be wd1 ?

<AidenIsik>currently the only drive in the system is the hdd

<AidenIsik>I unplugged the ssd

<youpi>AidenIsik: more information is never useless ;)

<rrq>ok.. I'm not on top of the device naming; I believe wd0 is first sata disk .. is the hdd sata or plain ata (hd0) ?

<AidenIsik>it's sata but the board runs it in ide mode

<AidenIsik>so plain ata is probably correct

<damo22>i believe rumpdisk uses wdX for both sata or ide

<youpi>yes

<damo22>but if you have two disks it may not be 0

<AidenIsik>uploading this image to my server right now then i'll send the link

<rrq>or maybe then numbering is also by it's position on the cable

<damo22>that is possible

<rrq>(in case): the wd0 choice is in /boot/grub/grub.conf

<rrq>on the multiboot gnumach .. line

<rrq>(lines)

<AidenIsik> https://files.aidenisik.scot/images/hurd-fail-boot-ioerr.jpg

<youpi>AidenIsik: your disk image does have partition 2, right?

<AidenIsik>let me reboot and check

<AidenIsik>2 zero-indexed or 1-indexed?

<youpi>1-indexed

<damo22>piixide1 is the second ide controller

<AidenIsik>Yes partition 2 exists

<AidenIsik>It's the rootfs

<damo22>hmm maybe not

<AidenIsik>\/boot is on the same partition too

<AidenIsik>1st partition is just swap

<damo22>are you certain you only have one disk plugged in and its the right one?

<AidenIsik>yes

<AidenIsik>i disconnect the Linux disk and plug this one in in its place

<damo22>the error message is consistent with the wrong disk being plugged

<AidenIsik>SATA port 0

<rrq>wd0 is a 698 GB disk

<damo22>do you have a sata port as well?

<AidenIsik>yep that's the size of the disk hurd is flashed to

<AidenIsik>yes i'm plugged into a sata port

<AidenIsik>it just runs in ide mode

<damo22>okay

<damo22>can you boot into grub> shell and run "ls"

<AidenIsik>I can try

<AidenIsik>don't know if my usb keyboard works though :p

<AidenIsik>might need a ps2 one

<damo22>grub should support usb keybd

<AidenIsik>hang on need to modify the grub config to bring up the menu

<damo22>hmm the disk image should be modified to set GRUB_TERMINAL=console

<Alicia>I think usually you can just keep pressing a key on boot to get the menu even if it isn't set to show it by default. usually I press down

<AidenIsik>i tried esc and shift, never heard of using down though

<AidenIsik>i'm booted into a linux-based system right now anyways so ill just edit the config

<damo22>AidenIsik: while you are still in linux, can you run sensors and check if the cpu temp works?

<AidenIsik>sure

<damo22>im curious about that board

<AidenIsik>they do

<damo22>wow

<damo22>someone must have fixed it in coreboot

<AidenIsik>Core 0: +34.0C, Core 1: +31.0C, Core 2: +28.0C, Core 3: +36.0C

<AidenIsik>:D

<AidenIsik>sorry this is taking me a bit i need to manually edit grub.cfg because i don't think i can regenerate the grub config from my linux host

<AidenIsik>and i have no idea how to enable the menu/console here :p

<damo22>remove anything that says insmod all_video or vga/vbe

<rrq>you might want to try using part:1:device:wd0 ... once, just in case

<AidenIsik>damo22: will that not just disable the video drivers grub has?

<damo22>i think inserting the video modules makes it load the display in gfx mode

<AidenIsik>set default=0

<AidenIsik>i wonder if that does what i want by disabling autoboot

<damo22>set timeout=5

<damo22>that will give you 5 seconds to drop to shell

<AidenIsik>okay rebooting now

<AidenIsik>well that did not work

<AidenIsik>i'll just edit the menu entries manually from there

<AidenIsik>should I try hd0 instead?

<rrq>the log suggests gnumach finds the disk as wd0

<AidenIsik>oh okay

<AidenIsik>just because I saw it was hardcoded in grub I thought maybe that is a problem

<rrq>grub names disks in it's own way (too)

<rrq>I think it's possibly to make a story on why to call it partition 1 rather than 2 .. so that's one test at least

<AidenIsik>yes but in this case it is specifying the name mach calls it

<AidenIsik>but yeah ill try that

<rrq>another test is to skip the readonly argument for the ext2fs module

<damo22>its a very unusual problem, the wd0 is the correct name according to the rumpdisk log

<rrq>there's sometimes a need of fsck which won't happen if readonly

<AidenIsik>i can also try fsck on the linux install

<damo22>yes i would definitely fsck partition 2

<rrq>that might work... though the filesystem is an "-o hurd" and not "-o linux"

<rrq>(larger inodes)

<damo22>i dont think fsck cares

<AidenIsik>fsck says the partition is clean

<AidenIsik>i will try to unset readonly first

<AidenIsik>unsetting readonly did nothing

<AidenIsik>will now try partition 1

<damo22>its probably a problem with DMA

<AidenIsik>hm

<AidenIsik>specific to rumpdisk or in general

<AidenIsik>because i wonder if i can just delete the lines which tell mach to use it

<damo22>specific to the piixide rump driver that we are using

<AidenIsik>i'll try disabling rumpdisk then and renaming the disk to hd0

<damo22>you need to remove the noide option from gnumach commandline and use hd0 most likely

<AidenIsik>noide isn't passed by grub at all

<AidenIsik>module /hurd/rumpdisk.static rumpdisk \

<AidenIsik> \

<AidenIsik> --next-task='${fs-task}' \

<AidenIsik> '$(disk-task=task-create)'

<AidenIsik>is though

<damo22>multiboot gnumach ... noide ?

<AidenIsik>nope

<AidenIsik>no instance of noide anywhere in the file

<damo22>oh maybe linux drivers are not built into hurd-amd64 gnumach?

<AidenIsik>hmm, maybe

<damo22>darn, we need to fix this

<AidenIsik>i will try i386

<rrq>damo22: is there a way to make it use DMA mode 2 (rather than the 5 it opted for?) .. a commandline arg maybe?

<damo22>unfortunately not, i think its compiled in, i might be wrong

<AidenIsik>i just checked the i386 config and it uses hd0 by default

<AidenIsik>so probably the linux 2.6 ide driver

<AidenIsik>will test now

<AidenIsik>ext2fs: part:2:device:hd0: No such file or directory

<AidenIsik>It still wants wd0

<AidenIsik>still using rumpdisk

<AidenIsik>noide is not set

<azert>AidenIsik: can you install and boot NetBSD?

<AidenIsik>i will check

<AidenIsik>that's a good point, rumpdisk is a netbsd driver

<azert>exactly, either it is a bug in the netbsd driver or in our pci/acpi configuration

<AidenIsik>does HDD vs SSD affect this?

<azert>I don’t think so

<AidenIsik>I could install onto the SSD but I've been trying to boot from an HDD so didn't want to pollute the test

<AidenIsik>but if it doesnt matter i'll do the ssd for speed's sake

<damo22>dont ruin a linux host install

<AidenIsik>the linux was literally just installed to flash coreboot

<rrq>"Ultra DMA" is apparently very sensitive to cable flaw.. resulting in crc errors

<damo22>do you have a usb stick?

<AidenIsik>yes

<damo22>try downloading netbsd installer and flash it to the usb stick and install that to the hdd

<AidenIsik>alright

<AidenIsik>also, Linux does actually take quite a while to boot up on this system

<AidenIsik>i wonder if that is related

<damo22>then you still keep linux as a backup

<damo22>or just use the netbsd installer to drop to a shell and check if the hdd is working

<damo22>no need to fully install it

<azert>you could also check dmesg in Linux if you suspect the disk is damaged

<solid_black>afternoon!

<AidenIsik>o/

<AidenIsik>hang on coreboot won't boot my usb lol

<AidenIsik>will try a different image

<AidenIsik>(well, seabios)

<damo22>bsds should boot with seabios

<AidenIsik>yeah i think i'm just using the wrong image

<AidenIsik>give me a bit lol i'm still struggling to get seabios to boot it

<AidenIsik>i think the lack of an isohybrid image is the problem

<damo22>its on the front page of netbsd.org

<damo22>USB image

<damo22>dont use the CD iso

<AidenIsik>USB image also doesn't work

<AidenIsik>but I can boot Linux from USB

<AidenIsik>nothing but problems with this computer lmao

<AidenIsik>well i'll go check linux dmesg for now

<youpi>damo22: we can't enable the linux drivers in gnumach, it's really not 64b-ready

<damo22>AidenIsik: $ qemu-system-x86_64 -drive if=none,id=usbstick,format=raw,file=../iso/NetBSD-10.1-amd64-install.img -device qemu-xhci -device usb-storage,drive=usbstick

<damo22>that works

<damo22>i assume you gunzipped the image before writing to stick?

<AidenIsik>yes

<AidenIsik>okay so in dmesg:

<AidenIsik>ata3: SRST failed (errno=-16)

<AidenIsik>ata3: softreset failed

<AidenIsik>ata3: reset failed, giving up

<AidenIsik>I don't know if that's because I hotplugged the drive to bypass the boot order or an actual issue with it

<damo22>check the stick integrity

<AidenIsik>those are logs from the hdd not the usb stick sorry

<damo22>what happens when you try to boot the stick?

<damo22>i know

<AidenIsik>SeaBIOS says it's not bootable

<AidenIsik>I've not tried that qemu command yet though

<AidenIsik>I assume that is just to make sure the image works

<damo22>yes

<damo22>the image works from a usb stick using seabios

<AidenIsik>yep boots under qemu

<damo22>something must be wrong with your stick

<AidenIsik>maybe it is the stick

<AidenIsik>yeah

<AidenIsik>I'll try flashing it to the one with Linux on it

<azert>solid_black: hello!

<AidenIsik>I have flashed NetBSD to the USB stick and reading back with strings /dev/sdb | grep NetBSD does show NetBSD there, but booting it up and I get greeted with the linux install

<AidenIsik>Partitions seen by lsblk are also still the old ones from the linux install

<AidenIsik>I have never run into such strange issues ???

<damo22>eh? did you mix up your devices

<AidenIsik>nope

<AidenIsik>I'm doing this on my laptop now, flashing to /dev/sda

<AidenIsik>the internal drive is nvme0n1

<AidenIsik>well, /dev/sdb or /dev/sda depending on where the kernel registers it

<damo22>fdisk -l

<AidenIsik>huh, sda/sdb isn't even listed

<damo22>maybe with sudo

<AidenIsik>nah I think I found the issue lmao

<AidenIsik>one time it mounted as /dev/sdb, I accidentally dded to /dev/sda (nothing there)

<AidenIsik>I'm pretty sure that created /dev/sda as a regular file and screwed stuff up

<AidenIsik>ughh that was a ridiculous mistake

<AidenIsik>well it is flashing slower now so i am pretty sure it is actually flashing

<damo22>(20:23:14) damo22: careful with raw devices , make sure its the right one obv

<AidenIsik>yeah lsblk was picking up the "fake" /dev/sda as a disk

<AidenIsik>and the kernel mounted it nowhere

<damo22>i always check dmesg | tail to see what actually mounted

<damo22>probed*

<AidenIsik>maybe I should start doing that because that was a really stupid mistake

<AidenIsik>funny now though lol

<AidenIsik>booting netbsd now

<AidenIsik>it crashes on the gpu driver.........

<damo22>nv?

<AidenIsik>well i think it is safe to say the gpu driver comes after the ide driver, and that it worked?

<AidenIsik>i915

<AidenIsik>the integrated gpu

<damo22>well you need to save a log

<AidenIsik>does netbsd write kernel logs to disk>

<damo22>i think you have a physical serial port on that board

<AidenIsik>i do but no cable to interface with it

<damo22>ok

<AidenIsik>i can look up a pinout and try to connect to it later today

<azert>you can probably disable the gpu driver

<damo22>maybe its a bit too much to get hurd 64 bit running on a board with ICH7 for now

<azert>do you see something on the screen before the crash?

<AidenIsik>yeah i get kernel logs

<azert>this means that the text mode console works

<AidenIsik>yeah it should be going through the gt1030 which is installed in here

<azert>so you can disable or delete the gpu drivers for the moment, to fix just one thing at a time

<AidenIsik>wait i think i have a few frames in a video where I can see the disk driver

<damo22>boot -c, then type disable nouveau to disable the NVIDIA driver, and disable i915drmkms to disable the i915 driver, followed by quit to continue booting

<AidenIsik>it's reading the disk successfully in NetBSD

<AidenIsik>dk0 at wd0: "EFI System", 262144 blocks at 2048, type: msdos

<damo22>did you mount it with ext2?

<AidenIsik>i have no shell right now

<AidenIsik>other than some kernel shell (?)

<AidenIsik>but i found that line in the logs playing back a video i recorded of them

<azert>now the trick is to understand what goes wrong in rumpdisk xD

<azert>that doesn’t affect NetBSD proper

<damo22>my only suggestion at the moment is to try i386 version of hurd on this old hardware and use linux driver built into gnumach

<azert>the surface, at least, is much reduces

<AidenIsik>yeah I tried i386 but it still wants to use rumpdisk, even without noide

<damo22>i think its a DMA issue

<AidenIsik>i need to look into the kernel args I can pass to mach

<AidenIsik>to tell it to use the linux driver and see if that works

<AidenIsik>but that is for later, i need to do things for now

<damo22>i gtg

<AidenIsik>thanks for entertaining me lol

<AidenIsik>cya

<AidenIsik>For future reference, here are the disk-related logs from NetBSD, in case they are useful to track down this bug: https://files.aidenisik.scot/pastes/2026/07/20/rumpdisk-netbsd.log

<AidenIsik>The only Hurd logs I could get are up the chat

<AidenIsik>More info to add to the pile: it does the same thing on a Toshiba Satellite T110-107 (ICH9)

<AidenIsik>My guess is it is something about this drive tripping it up

<AidenIsik>Input/output error on both machines

<youpi>AidenIsik: which image did you try?

<AidenIsik>latest Debian hurd images, both i386 and amd64

<youpi>"latest" doesn't mean anything

<youpi>it could be stable, latest/, daily

<AidenIsik>latest

<youpi>you meant latest/ ?

<youpi>as in : in the "latest" directory?

<AidenIsik>yes it is in the latest subdir

<youpi>ok

<AidenIsik>stable also misbehaves

<youpi>have you tried daily?

<AidenIsik>nope i will try that if you want in about an hour or so

<youpi>while it can easily fail to install a system, it will have one of the latest gnumach kernels

<youpi>note that there isn't a preinstalled image in that case, but you can try to boot the installer and see if it manages to access the disk

<AidenIsik>accessing the installer may be a problem because I don't have a CD/DVD drive

<AidenIsik>I can try flash it to the HDD but unless it's isohybrid idk if it would work

<AidenIsik>Anyways, I'm off to go eat

<youpi>it'll be bootable from usb

<AidenIsik>oh is usb boot supported now

<youpi>err, since like a decade

<AidenIsik>oh lol

<youpi>the problem is getting files from usb

<youpi>that one is relatively recent, and I don't remember if it's finished integrating in the debian installer

<AidenIsik>ohhh does it load it as a ramdisk or something to sidestep the lack of usb mass storage support

<youpi>but at least you can boot it

<youpi>of course

<youpi>like all OS

<youpi>initially there is no driver

<youpi>so you load an initrd

<AidenIsik>I'm not seeing any "daily" or anything like that on Debian's servers for the Hurd

<AidenIsik>could you point me to where these images are?

<youpi> https://people.debian.org/~sthibault/hurd-amd64/installer/cdimage/daily/

<AidenIsik>thanks

<AidenIsik>sneek: tell youpi i tested the daily installer, and while it can boot on both systems it can read the disk on neither. even on ahci mode on the laptop, i am unable to access the disk. i am however able to boot a prebuilt image in ahci mode on the laptop (but that is no good, because that isn't my target)

<sneek>youpi, AidenIsik says: i tested the daily installer, and while it can boot on both systems it can read the disk on neither. even on ahci mode on the laptop, i am unable to access the disk. i am however able to boot a prebuilt image in ahci mode on the laptop (but that is no good, because that isn't my target)

<AidenIsik>sneek: tell youpi when possible i tested the daily installer, and while it can boot on both systems it can read the disk on neither. even on ahci mode on the laptop, i am unable to access the disk. i am however able to boot a prebuilt image in ahci mode on the laptop (but that is no good, because that isn't my target)

<sneek>youpi, AidenIsik says: when possible i tested the daily installer, and while it can boot on both systems it can read the disk on neither. even on ahci mode on the laptop, i am unable to access the disk. i am however able to boot a prebuilt image in ahci mode on the laptop (but that is no good, because that isn't my target)

<AidenIsik>well im not using that bot right

<AidenIsik>sneek: later tell youpi i tested the daily installer, and while it can boot on both systems it can read the disk on neither. even on ahci mode on the laptop, i am unable to access the disk. i am however able to boot a prebuilt image in ahci mode on the laptop (but that is no good, because that isn't my target)

<sneek>Got it.

<AidenIsik>finally

<AidenIsik>It also seems to not work with the SSD either.

<azert>solid_black: I’m thinking that by the reasoning i was using yesterday, one would try to move the pflocal server in glibc. Which is clearly not the design choice that was made for the Hurd

<youpi>pflocal cannot be inside a process, we need it for named pipes etc.

<sneek>youpi, you have 1 message!

<azert>although I suspect one could think of a complex design to increase performances

<azert>youpi: glibc could serve named pipes and the namespace is the filesystem server. I think that the only limit is the coupling between the task lifetime and the socket lifetime that would be problematic

<youpi>named / anonymous sockets are often passed between processes in odd ways

<youpi>through passing rights over a socket etc.

<youpi>(while for an eventfd, I have never seen that)

<azert>that could be probably handled by glibc exactly like

<azert>pflocal do

<azert>but I’m not sure, can you have more than one pflocal running on the Hurd?

<azert>would they be fine with each others?

<youpi>you can bind several on different paths, you just need to use remap/chroot to get processes to use it

<azert>ok but can you pass one right handled by a server over a socket of another server? I guess so

<youpi>but I really don't see how you manage to reasonably manage passing an anonymous pipe through a socket, and then abort()

<youpi>sure you can

<youpi>once it's open, it'll just work

<youpi>you don't need any more rendez-vous

<azert>so it’s doable. The problem is the life time. If you abort they would just stop working if in glibc

<youpi>(abort, or _exit, or whatever odd thing)

<azert>exec() is probably the most odd thing that would interfere with this

<azert>since it would keep the receive rights but lose the thread polling in them

<azert>youpi: afaik kqueue, instead of epoll, relies on sockets. And could be implemented as part of pflocal instead of duplicating functions

<azert>the MacOS version, somehow, even handles raw mach receive rights

<azert>but probably also epoll could be put there

IRC channel logs

2026-07-20.log