1 |
Hello Zac, |
2 |
|
3 |
Please follow 72 columns standard in e-mail. :-) |
4 |
|
5 |
Regarding mdev, it was not even taken into account when writing Dracut. |
6 |
It is very possible that issues arise in such a combination. |
7 |
|
8 |
I haven't reviewed your code yet, but from what you've described, seems |
9 |
to be great work! When I find some more time, I'll review the code and |
10 |
think how it's best to import it. Please also contact Alexey Shvetsov |
11 |
<alexxy@g.o>. He's working on bringing ZFS to Gentoo. I'm |
12 |
CC'ing this reply to him. (That's why I'm top-posting, don't beat me |
13 |
guys!) |
14 |
|
15 |
Thank you for sharing your work and such a good summary! |
16 |
|
17 |
|
18 |
Cheers, |
19 |
aidecoe |
20 |
|
21 |
|
22 |
Excerpts from Zachary Bedell's message of Sun Jul 03 22:37:18 +0200 2011: |
23 |
> Alas, here I am self-replying a few hours later... |
24 |
> |
25 |
> I figured out how to get everything running properly with the Dracut |
26 |
> initramfs, though I'm afraid I'm stuck at the 6th stage of debugging |
27 |
> [1] on this one... My system didn't have udev set to start at any |
28 |
> runlevel. Adding it to boot made everything work properly with both |
29 |
> the Dracut and classic initramfs. I haven't the slightest idea how it |
30 |
> worked with the classic Genkernel initramfs, but I guess I'll take |
31 |
> it... |
32 |
> |
33 |
> So all told, I have some genkernel patches and some ZFS patches (all |
34 |
> for the Dracut module) to contribute. I'll pass along the ZFS changes |
35 |
> upstream, though I'm not sure if that's something that Dracut would |
36 |
> eventually want to pull in? I think for the time being given that |
37 |
> ZFSonLinux is rather actively evolving, it might make sense to leave |
38 |
> the modules over there until things quiet down a bit. As it is, if |
39 |
> you install the Dracut branch of Genkernel, Dracut itself, and the git |
40 |
> version of ZFS, all of the files end up in the right places so |
41 |
> everything works together. Merging them into Dracut probably doesn't |
42 |
> buy anything right now. |
43 |
> |
44 |
> Patches are attached, and everything's in a public git repo if some |
45 |
> other format is desired. For the record, this is all released under |
46 |
> GPL-2 or any later version at your option. |
47 |
> |
48 |
> As far as a summary of changes, here's what I ended up doing: |
49 |
> |
50 |
> https://github.com/pendor/gentoo-zfs-overlay/, branch master |
51 |
> ebuild: (Versus the stock -9999 ebuild in portage) |
52 |
> * Change git to dracut branch of forked repo |
53 |
> * Add dep for Dracut |
54 |
> * Add use flag based dep for ZFS (explicit on -9999 version for now |
55 |
> as there's been no release with the necessary changes) |
56 |
> |
57 |
> |
58 |
> https://github.com/pendor/genkernel-zfs, branch dracut |
59 |
> genkernel: (Versus dracut branch from git.overlays.gentoo.org) |
60 |
> * Remove -nonet flag from xsltproc call (couldn't build with it set) |
61 |
> * Add 'zfs' to list of Dracut modules in gen_dracut.sh |
62 |
> * Add '--zfs' flag to command line & ZFS setting to conf file, update |
63 |
> docs to match. |
64 |
> * Pass log level to Dracut based on Genkernel's log level |
65 |
> * Try to catch Dracut failures and die rather than reporting success |
66 |
> after broken initramfs. |
67 |
> |
68 |
> |
69 |
> I also made some minor changes to the zfs-9999 ebuild originally taken |
70 |
> from the Science overlay. The most important is the creation of |
71 |
> /etc/hostid if it doesn't exist as Gentoo's hostid binary doesn't |
72 |
> appear to write that file if it doesn't exist. |
73 |
> |
74 |
> Best regards, |
75 |
> Zac Bedell |
76 |
> |
77 |
> |
78 |
> [1] http://mark.aufflick.com/blog/2005/11/10/six-stages-of-debugging |
79 |
> |
80 |
> On Jul 3, 2011, at 12:33 PM, Zachary Bedell wrote: |
81 |
> > I've managed to make a bit of progress with getting |
82 |
> > Genkernel/Dracut/ZFS to all play nicely together. I'm running into |
83 |
> > some trouble with devices after the initramfs is actually done, and |
84 |
> > I'm not sure if I'm just hitting the current limitations of the |
85 |
> > Genkernel/Dracut branch or if there's something silly I might be |
86 |
> > doing. My first suspicion is that Dracut's using udev while the |
87 |
> > rest of my system is setup for mdev is probably the source of my |
88 |
> > issues, but my knowledge of the various device manager packages is |
89 |
> > pretty limited at this point. |