New install missing kernel makefile

nottledim · n00b Joined: 15 Feb 2018 Posts: 18 Location: Edinburgh

I'm having trouble installing kernel on new system. I'm following the handbook by rote as I'm not that familiar with Gentoo - starting with minimal install on x86 32bit system.

I emerged sys-kernel/gentoo-sources but it failed without giving a reason. The /usr/src directory looks correct but there's no symlink to linux and there's no Makefile.

I tried emerging vanilla-sources and that installed without error but again there's no Makefile so it can't be compiled.

eselect kernel list says no symling targets found.

The running system is 4.9.76-gentoo-r1 gentoo-sources is the same, vanilla-sources is 4.15.3.

I don't know what to do. I'd appreciate some help.

NeddySeagoon · Posted: Fri Feb 16, 2018 6:22 pm Post subject:

nottledim,

The /usr/src/linux symlink is used by tools to find the the kernel to build and by packages that install out of kernel modules to find the kernel to build against.
Sometimes its used by the user too.

It plays no part in the operation of the system, so in many cases its not required.

If you have newer kernels in the same series (vanillia-sources, gentoo-sources ..) and you run

nottledim · n00b Joined: 15 Feb 2018 Posts: 18 Location: Edinburgh

Thanks for your help however the problem is there are no Makefiles. When I emerge
sys-kernel/gentoo-sources or vanilla-sources no symlink to /usr/src/linux is created and also the
directory containing kernel source has no Makefile.

The documentation says a symling will be created. It isn't. The documentation says run make menuconfig but there's no Makefile.

I've done nothing to remove anything and the system seems blissfully unaware that it installed anything.

This is the log of the emerge sys-kernel/gentoo-sources

NeddySeagoon · Posted: Sat Feb 17, 2018 2:01 pm Post subject:

nottledim,

We need to see the emerge log.

Below the message you quoted is some text about reporting a bug. It tells the path to the build log.
Use wgetpaste to put it onto a pastebin and post the link.

If you want a wild guess ... the kernel is a large number of small files and you are out of inodes on your filesystem.
What does

nottledim · n00b Joined: 15 Feb 2018 Posts: 18 Location: Edinburgh

If it worked the build file is: https://paste.pound-python.org/show/GZ3j0uRcWOeZaJn6WU7U/

I assume you meant the build log rather than the /var/log/emerge.log but just in case:
https://paste.pound-python.org/show/dQlhGkDNvMgbDYuknpYF/

Hu · Moderator Joined: 06 Mar 2007 Posts: 21642

According to that build log, you already have substantial portions of that kernel installed (maybe all of it, maybe not), but none of the installed files are owned by the package that is supposed to own them, so Portage considers it an error that the package you are installing tries to overwrite them. This is a rather unusual error. One possible cause is that you previously installed the kernel with a different instance of the Portage package database, so the files you have now are owned by a package in the other /var/db/pkg.

If you just want to proceed without determining how you entered this state, move aside the offending directory and retry. You will likely delete the directory at some point, but you should preserve it for now so that it is easy to undo.

If you want to understand how this happened, there is still more information to collect. If we do not know how this happened, it is possible that there are other as-yet undiscovered problems with the same root cause. During the install, did you do the entire install in a single sitting or did you stop at some point, then return and resume later? If the latter, are you certain that you mounted all the supporting filesystems at the same paths in each sitting? What is the output of cat -n /etc/fstab ; echo ; cat -n /proc/self/mountinfo ; echo ; file -s /dev/sd*, as seen inside the chroot (or live system if you have the system fully operational)? That may not be enough to solve this, but it is a good first step.

nottledim · n00b Joined: 15 Feb 2018 Posts: 18 Location: Edinburgh

I've installed the kernels a few times both overwriting and by deleting the directory first. Yes the system has been rebooted a couple of times but I think the problem I'm seeing started before it was rebooted.
I can start over again if that's a viable solution (it's only 1.2GHz Via computer - it takes forever :-)

)

/usr/src/* is owned by root as is /var/db.pkg and, it seems, so is everything else. Is that a problem? What is supposed to own them?

Output requested:

NeddySeagoon · Posted: Sat Feb 17, 2018 6:01 pm Post subject:

nottledim,

By owner, Hu was referring to the installed package that owns a file.

When you remove something without using portage, portage doesn't know and still has its record of all the files the package installed.

Your log contains ...

nottledim · n00b Joined: 15 Feb 2018 Posts: 18 Location: Edinburgh

NeddySeagoon · Posted: Sat Feb 17, 2018 6:55 pm Post subject:

nottledim,

I don't yet understand how portage believes the package is installed, which gives rise to the file collisions, yet can't find it to remove.

The can't find it has nothing to do with the content of /usr/src
Portage keeps track of things in /var/db/pkg/<category>/<package>-<version>/
Looking is safe but changing things is not.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

nottledim · n00b Joined: 15 Feb 2018 Posts: 18 Location: Edinburgh

I tried again to install with no success so I started the install again into a fresh partition. The result is exactly the same: emerge failed after cleanup phase, no Makefiles in /usr/src and /var/db/pkg/sys-kernel contains a directory -MERGING-gentoo-sources-4.9.76-r1 which I assume is named like that to indicate install process is incomplete. There is also a directory called linux-headers-4.13 which was installed by stage3 tarball. The version number puzzles me. And to confirm eselect kernel list shows empty..

How do I install a different version of the kernel?

I can't believe nobody else has had this problem. Could it be related to installing x86 32 bit? It's not so common now.

Could it be a USE flag I'm not setting? I've just taken the defaults.

NeddySeagoon · Posted: Mon Feb 19, 2018 12:26 pm Post subject:

nottledim,

linux-headers is used by things in userspace that need a stable set of kernel headers to build against.
These headers are not provided by the kernel you happen be be running or building today.

I'm not surprised you have the same error in a new install. Its probably the same problem we still don't understand.
I've not seen this issue before. Similar things happen when you run out of i-nodes or disk space.

Lets try fetching a kernel outside of portage. I expect it to fail too.

nottledim · n00b Joined: 15 Feb 2018 Posts: 18 Location: Edinburgh

That worked fine and the directory contains Makefile, README and .git whatever as I would expect.

I can compile that manually but portage won't know about it.

Can I use genkernel? Should I use genkernel?

NeddySeagoon · Posted: Mon Feb 19, 2018 1:12 pm Post subject:

nottledim,

We still don't know what the issue is. I expect other packages will be affected too.

genkernel will build you a kernel with the same settings as the liveCD uses.
Configuring the kernel by hand is probably the hardest part of a Gentoo install. genkernel will allow you to delay that learning step.
However, you may find that genkernel alone will not set up your video or wifi. You can fix that later.
That's a limitation imposed by hardware manufacturers. A one size fits all is not possible as kernel drivers conflict with one another.

You will need to make the linux symlink by hand, then you can use genkernel.

This is only a work around, not a fix. Put the full portage generated kernel install log onto a pastebin site if you still have it, or do it again if you don't.
Post

nottledim · n00b Joined: 15 Feb 2018 Posts: 18 Location: Edinburgh

build log: https://paste.pound-python.org/show/XFLCWvvnliqGiJ1ZKDG0/

snippet from emerge.log:

NeddySeagoon · Posted: Mon Feb 19, 2018 1:48 pm Post subject:

nottledim,

nottledim · n00b Joined: 15 Feb 2018 Posts: 18 Location: Edinburgh

I've no idea either.

20G should be plenty. It's only a little server. I can make the partition bigger once it's booted if I want. I was really using it as a means to get to know Gentoo. Now ArchLinux is not supporting 32bit I have to find something else for these old motherboards I have.

genkernel doesn't work. It also has some error. I can't work out what it means but it fails.. So I'm doing a manual compile which so far is going OK. I imagine it'll be several hours.

Thanks again for your help. No doubt you'll hear from me again if/when the kernel is compiled.

NeddySeagoon · Posted: Mon Feb 19, 2018 3:19 pm Post subject:

nottledim,

I run KVM servers in 20G, as you say it will be fine.
How much RAM and swap do you have?
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

nottledim · n00b Joined: 15 Feb 2018 Posts: 18 Location: Edinburgh

It looks like 512M ram. I thought it was 1G. No doubt the 1G stick was borrowed. It's only using 60..80M + buffers.

No swap configured. I'll use lvm swap once it's running.

Compile is going OK - so far so good.

NeddySeagoon · Posted: Mon Feb 19, 2018 7:06 pm Post subject:

nottledim,

Keep an eye on things. That emerge *-sources fails is a pointer to something.
If you run out of RAM, the OOM will kick in an leave messages in dmesg.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

nottledim · n00b Joined: 15 Feb 2018 Posts: 18 Location: Edinburgh

NeddySeagoon · Posted: Mon Feb 19, 2018 10:49 pm Post subject:

nottledim,

:)

Add some swap. Not having swap does not prevent the kernel swapping. It reduces it choices for swapping.
Swap is only ever used for dynamically allocated memory. That's thing that do not have a permanent home on the HDD.
When you run a program, its not loaded into memory. Its mapped into virtual memory. When a piece tries to execute, you get a page fault and that bit is loaded into physical RAM.
When there is not enough RAM, previously loaded code can be dropped and the process repeated next time the code is needed. Its swapping.

This mechanism allows Linux to run programs that are too big to fit into RAM. Its OK for several areas of virtual memory to map to the same physical RAM.
Only one can occupy the physical RAM at a time.

Not having any swap forces the kernel to keep all dynamically allocated RAM in RAM.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

nottledim · n00b Joined: 15 Feb 2018 Posts: 18 Location: Edinburgh

I added swap. I'll change the memory tomorrow when the compile has finished.

Funny, your location is very similar to mine. Not a contributor to edlug by any chance?

NeddySeagoon · Posted: Mon Feb 19, 2018 11:04 pm Post subject:

nottledim,

I subscribe to the edlug list but I've never made it to a meeting.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

nottledim · n00b Joined: 15 Feb 2018 Posts: 18 Location: Edinburgh

Sorted out the memory and it's all working fine including emerge sys-kernel/gentoo-sources

Learning experience. Thanks for your invaluable.