You are not logged in.
Running Daedalus on an old MSI/Phenom/WD spinners box:
I ran into this before and like then I cannot nail down the cause of the problem. It comes up on the above MSI board but NOT on another Asus board using ssd's.
It starts on boot with a message about having to use another version of fsck and to continue with Cntrl-D under a warning flag!
from /var/log:
Log of fsck -C -T -M -A -a
Sat Mar 15 05:55:14 2025
/dev/sda22: clean, 3130/37707776 files, 3233819/150799633 blocks
/dev/sda3 is mounted.
e2fsck: Cannot continue, aborting.
fsck exited with status code 8
Sat Mar 15 05:55:14 2025
It says /dev/sda3 (swap) is mounted. Yet nerar the end of dmesg there's this:
[ 53.024988] Adding 6291452k swap on /dev/sda3. Priority:-2 extents:1 across:6291452k FS
[ 53.101048] EXT4-fs (sda12): re-mounted. Quota mode: none.
[ 219.760724] EXT4-fs (sda22): mounted filesystem with ordered data mode. Quota mode: none.
[ 219.760882] /dev/sda3: Can't open blockdev
/etc/fstab is:
UUID=37fd95d7-fd2c-4020-ad07-66be4b0d9bd4 / ext4 errors=remount-ro 0 1
UUID=3b768db7-14e1-41f3-b9f8-297a63096a2a swap swap nofail 0 0
UUID=2430d815-0e57-4ac5-bc89-febccd445a70 /0/data ext4 data=ordered,nofail 0 2
UUID=3b768db7-14e1-41f3-b9f8-297a63096a2a /0/xdata ext4 data=ordered,nofail 0 2
which jives with blkids:
/dev/sda3: UUID="3b768db7-14e1-41f3-b9f8-297a63096a2a" TYPE="swap" PARTLABEL="Linux swap" PARTUUID="9b30f2c0-637e-4b51-8713-721a863256e9"
/dev/sda12: UUID="37fd95d7-fd2c-4020-ad07-66be4b0d9bd4" BLOCK_SIZE="4096" TYPE="ext4" PARTLABEL="p12-dev" PARTUUID="22054203-802a-4f37-aa13-cbc5bce21f08"
/dev/sda22: UUID="2430d815-0e57-4ac5-bc89-febccd445a70" BLOCK_SIZE="4096" TYPE="ext4" PARTLABEL="Linux filesystem" PARTUUID="e12abb80-a7ac-4535-bf80-deb5bfc1704d"
/dev/sdb1: UUID="89eddb53-303f-4bb1-807e-f0883c64d62a" BLOCK_SIZE="4096" TYPE="ext4" PARTLABEL="Linux filesystem" PARTUUID="aa6cc4f9-dfe6-4f7d-8261-30ae46549fdc"
NB. It's a multi OS box I'm trying to prepare for resale preloaded with Linux and I can't really let it go with such a hangup.
Any idea what the actual problem is and how to fix it?
Who, has loved us more?
Offline
The actual problem is that fsck wants to scan an FS already mounted. You can unmount swap and have it checked, but as you're reselling the machine, just re-install Linux afresh and be done with it.
Offline
It is a relatively fresh install, wouldn't code normally unmount before fsck or fsck before the first mount? And how come the same Daedalus on my other machine doesn'ty have this problem? I'n totally lost here. Booted into Tumbleweed on that same MSI board fsck won't do anything
# umount /dev/sda3
umount: /dev/sda3: not mounted
..even though it is in /etc/fstab as swap in that system too. CAN you do fsck on a swap partition at all?
I even thoiught of reformatting it for TS but since 10 distros use that same sawp partition that would also mean having to edit 10 fstabs due to a new UUID :-(
Who, has loved us more?
Offline
Why are you going through all this effort to have a multi-OS setup when reselling the machine anyway? Let it be the next owners issue, IMO.
Offline
10 distros use that same sawp partition that would also mean having to edit 10 fstabs due to a new UUID :-(
Time for you to use the LABEL=???? method of mounting partitions, I would go with SWAP possibly for the ????. Then you need change nothing in the /etc/fstab if needing to reformat you just need to ensure you set the label every time done with the swaplabel command.. Now I check the man page online it says you can just set the UUID using it too, so nothing need change.
Offline
Why are you going through all this effort to have a multi-OS setup when reselling the machine anyway? Let it be the next owners issue, IMO.
Because I get twice the going price for it; "if you always wanted to try Linux here's the winner for you! 16 TB's in mobile trays and you can still install windows if you like AND keep Linux...etc." It's a server tower with 20k hour spinners, I'd be lucky to get $300cad for it but with 10 Linux distros it's double that. Not my 1st one either. Seeing that I'm upgrading my own box after it, it will be the same story so I try to do everything like a bulletproof copy-paste skel.
Who, has loved us more?
Offline
Time for you to use the LABEL=???? method of mounting partitions, I would go with SWAP possibly for the ????. Then you need change nothing in the /etc/fstab if needing to reformat you just need to ensure you set the label every time done with the swaplabel command.. Now I check the man page online it says you can just set the UUID using it too, so nothing need change.
THAT's a 5-star idea, thanks, almost as good as device-names used to be. I owe you a beer so I'll have one and ask the bartender to say a mass for your soul :-) Tempted to 'label' it /dev/sda3 just out of spite!
Who, has loved us more?
Offline
Tempted to 'label' it /dev/sda3 just out of spite!
You are welcome, yes I can see why give it a couple of more years, they will find some other new shinny to chase and you will get to learn another new system of doing it. Been like that for the twenty-five plus years I have been using it, I have every confidence they will do it yet again.
Offline
I've always wondered which center of higher education the engineeres went to to get PhD's in stupidity? I mean there are HARD-WIRED sata channels but the pukes couldn't work out a persistant dev-name method?
Who, has loved us more?
Offline
Some people may want to be able to move a disk to another slot and still have the system to boot.
Offline
How many?
Who, has loved us more?
Offline
I've always wondered which center of higher education the engineeres went to to get PhD's in stupidity? I mean there are HARD-WIRED sata channels but the pukes couldn't work out a persistant dev-name method?
They did (/dev/disk/by-[foo] symlinks), but for some unfathomable reason you are clinging to /dev/[s,h]d[x] kernel naming instead of using it.
Please stop perpetuating this nonsense. Kernel device names are not consistently mapped to physical ports, and they never have been. They are simply assigned in the order the hardware is discovered.
"HARD-WIRED" means nothing here. The driver probes the hardware and whoever answers first gets the first device name, what bus or connector it's on is irrelevant.
Some hardware always comes up in the same order, some does not. The myth that these names "used to be consistent but someone did stupid" is largely an artifact of the relatively predictable initialisation order of old PATA controllers with fixed primary/secondary and master/slave channels.
The only "PhD in stupidity" here is complaining about inconsistent naming behaviour from an interface that makes no attempt to provide consistent naming *by design*.
How many?
Me, for one, since I don't need to know or care what goes in which bay when dealing with 20+ drives. If I do want to know which physical drive is where, I use by-id or by-path, which map to model and serial or bus and physical port respectively.
Last edited by steve_v (2025-03-16 12:10:00)
Once is happenstance. Twice is coincidence. Three times is enemy action. Four times is Official GNOME Policy.
Offline