Partition can't be mounted at boot but can be mounted by....

[Bluey_the_dog]

Dear all,
I'm in the process of building a new machine. The partitions are formatted jfs. I initially installed the latest gentoo-kernel-bin, I set up fstab and a whole lot of other stuff, rebooted the machine and all was good. All the partitions mounted etc.

I personally don't like initramfs so installed the latest source kernel, dropped the .config from the -bin kernel into the source kernel dir and went through converting from modules to builtin. After the drudgery of doing this, I recompiled, took the bzImage, dropped it into my /efi dir and rebooted. Refind sees the kernel and I can select it and it boots. Now though, during the boot, /home is no longer mounted, an error is thrown. The boot completes but the contents of /home are not there. The error is, and it's the same as to whether it be via booting or as root and issuing a mount command:

[pingtoo]

Bluey_the_dog,

Can you share output of

[grknight]

[Bluey_the_dog]

pingtoo,
Please see the following:

[Bluey_the_dog]

grknight
I've placed rootfstype=jfs in the kernel command line, via refind, and hey presto, that error has now disappeared. This tip probably also explains why my older machine, also both partitions jfs, would have:

Older machine dmesg

[pingtoo]

Bluey_the_dog,

Have you try with just mount without type? i.e.

[Bluey_the_dog]

Pingtoo,
With respect to the mount command, as I mentioned in the original post, I've tried all three and they give the same result - refer above for the error message.

For fsck:

[Goverp]

[pingtoo]

So can you test mount right after fsck?

please test mount with strace and share the output.

if you don't have strace installed, please try emerge -1a dev-debug/strace

[jburns]

The kernel can mount the root partition because it has the code for mounting it. But it only mounts the root partition.
The gentoo-kernel-bin has an initramfs that contains mount that supports mounting with jfs by the initramfs.
A kernel without an initramfs only mounts the root partition, the other partitions must be mounted by the OS's init process. This needs sys-fs/jfsutils to be installed.

[Bluey_the_dog]

jburns
jsfutils has been installed on the machine for 6 months ie this machine has not been a top priority...

[Bluey_the_dog]

People,
Thank you for your help and suggestions, especially grknight with the tip to kill off the error in dmesg, but in my googling to try and find a solution, I've come across a few things[1] that make me thing it's time to move on. I've been using jfs for probably at least 17 years, I was doing an Engineering Masters/PhD at the time and it had benefits for the amount of data I had, but the last comment in the linked bug makes me a bit wary.

So once again, thanks

Andrew

p.s. If someone does know what the problem, please let me know, it would add "closure"


[1] https://bugs.gentoo.org/863905

[pingtoo]

Bluey_the_dog,

It is your prerogative to call it stop, I respect that.

However your last post lead me to think you may misunderstanding something. the jfsutils have nothing to do with fail to mount a jfs file system.

jfs for linux does not require a mount helper to support mounting, so the jfsutils package does not include for example "mount.jfs" binary.

jfs in linux kernel is still supported and I see from code update in past six month.

the bug you see is refer to compiling jfsutils with LTO it is IMO really unnecessary since nobody run fsck/mkfs for jfs all daylong so why bother to optimize to the extreme?

I suspect in your case it is possible kernel missing a code page that is necessary for your environment. But without seeing evidence I cannot conclusive say what is actually happen.

It is very possible even using strace I will not be able to tell exactly where is the problem, but this is one easy step to help understand, if strace step not helping I would suggest you turn on CONFIG_JFS_DEBUG as next step. because the error condition did trapped, so something should be in dmesg but may be it is one that is deep in the code that only when debug message turn on so we can see it.

[Bluey_the_dog]

pingtoo,
I've finally worked out how to use strace, you needed to add the mount command after what you provided:

[pingtoo]

Bluey_the_dog,

Thanks for providing these information.

[Bluey_the_dog]

pingtoo,
I've created the partition multiple ways. The first time would have been as per The Handbook whilst doing the initial install. I probably did a couple of reformats from the command line subsequently. As well I've done a few command line formats from sysrescue & gentoo sticks and just tonight, Gparted as per the graphical environment in sysrescue.

I think I will grab a few live sticks, sysrescue, gentoo, Arch etc and see if they can all mount the partition. If they can it would say to me that the actual partition is sound and that it is something to do with my config of the kernel.

2:30AM Perth Australia time so I'm off to bed, regards,
Andrew

[pingtoo]

Bluey_the_dog,

No worries take you time. Have a good night sleep

I was just about ask you kernel. is this is on a X86 box or something else? Do your kernel have different page size than 4096?

Yes, it is good idea to use a different livecd to test if that partition work under.

From the source code, it seems to me the meta data content on disk is not right, seems to out of range. so if there is no problem in partition table (no overlaps) and when you create the file system without special arrange, i.e. external journal device or different block size, etc than there is always a chance that part of disk are bad too.

[pingtoo]

Bluey_the_dog,

Just for your information.

I tested on my RPI5 (pagesize 16k)