Gentoo Archives: gentoo-user

From: Jack <ostroffjh@×××××××××××××××××.net>
To: gentoo-user@l.g.o
Subject: Re: [gentoo-user] new genkernel problem
Date: Sat, 06 Jun 2020 17:21:00
Message-Id: UT6ADSKN.DZKXKH44.5EDAMNQN@72B56744.SUEYPPU4.WAUD4QAV
In Reply to: Re: [gentoo-user] new genkernel problem by Dale
1 On 2020.06.06 12:34, Dale wrote:
2 > Neil Bothwick wrote:
3 > > On Sat, 6 Jun 2020 10:13:56 -0500, Dale wrote:
4 > >
5 > >>>> If you do copy yours manually to /boot, what command do you use
6 > for
7 > >>>> dracut?  Maybe I'm doing it a hard way or something and you have
8 > a
9 > >>>> easier method. 
10 > >>> cd /usr/src/linux
11 > >>> make all modules_install install
12 > >>> dracut --kver=$(cat include/config/kernel.release)
13 > >>>
14 > >>> It doesn't get much easier ;-)
15 > >
16 > >> From what I've read, I like my way better.  I did have to change
17 > the
18 > >> names from bzimage* to kernel* but other than that, I can use the
19 > naming
20 > >> method I've used for years and keep the good kernels I want.
21 > > make install names the kernels vmlinuz-$VERSION, and updates a
22 > symlink to
23 > > vmlinuz if one exists.
24 >
25 > But sometimes I have more than one of the same version.  I add a -1,
26 > -2,
27 > -3 etc to them as I go.  From my understanding, make install doesn't
28 > do
29 > that.  I do and with good reason. It seems make install won't
30 > accomplish
31 > what I do by hand.
32 That's what to use CONFIG_LOCALVERSION for. It gets appended to the
33 kernel file name, and to the /usr/lib directory created by
34 modules_install. The hardest part of my figuring out how to use
35 genkernel was that it overrides that value in .config, so I had to rig
36 up a wrapper script so I could set that on the command line. That's
37 actually better for me, since I don't need to edit .config to rerun
38 genkernel if something else changed.
39 >
40 > >> On
41 > >> occasion I remove outdated ones I no longer plan to use.  I still
42 > wish I
43 > >> didn't need the init thingy but still.
44 > > make install doesn't remove anything, that's your job!
45 >
46 > Neither does my method.  When /boot starts getting fullish, I use
47 > uprecords to help determine what kernels to keep along with what
48 > kernels
49 > would still be compatible with system changes.  In my opinion, the
50 > admin
51 > should do those manually. 
52 >
53 >
54 > >  
55 > >> My biggest problem, getting the dracut command options right.  If I
56 > >> didn't need dracut, I'd be in heaven. 
57 > > If you have a plain setup, dracut shouldn't need any options.
58 > >
59 > >
60 >
61 > I don't have a plain setup tho nor do I really want that setup.  I
62 > like
63 > having backup kernels and my own numbering system.  It has worked for
64 > me
65 > for decades, ever since I started using Gentoo and building my own
66 > kernels.  I don't see any point in changing what works unless I can
67 > streamline what I'm already getting with the results I expect.  If I
68 > could get rid of the init thingy, I would have zero issues with my
69 > method.  It's dracut that causes the issues. We all know how much I
70 > dislike init thingys tho.  ;-)  That said, dracut hasn't failed me in
71 > a
72 > while.  If it can't build correctly, it does spit out it failed.  It's
73 > been a while since the init thingy it creates has failed as well.  So,
74 > at least there is that. 
75 Once I did get genkernel fully configured to my liking (especially
76 getting it to append a string of my choosing to the names of all the
77 files it creates in /boot and to the /lib/modules directory) it hasn't
78 failed, unless the actual kernel compile failed (as for gcc-10, not
79 fixed) or the failure to compile something to go into the initramfs
80 that started this thread. I imagine dracut will produce a very similar
81 initramfs, and maybe it wont try to pull in an outdated package, but.
82 like you, I prefer to stick with what works, unless I have a good
83 reason to change, such as simplifying my process, even if it does take
84 some effort to get there.
85 >
86 > Dale
87 >
88 > :-)  :-) 
89 >
90 > P. S. Second drive passed the long selftest.  Copying files over as I
91 > type.  I think this one might live.  I suspect at least one of them
92 > was
93 > broken.  I just got the two mixed up.  They are identical in brand and
94 > size.  Now to go rebuild the mineral site for the deer.  Picked up 150
95 > lbs of goodness yesterday. :-D
96 >