1 |
On 2013-08-19 12:04, Alan McKinnon wrote: |
2 |
|
3 |
> It's not that separate /usr is broken - it's not. |
4 |
|
5 |
I know. |
6 |
|
7 |
> The issue is a separate /usr without an initramfs. And the issue ONLY |
8 |
> occurs at early-boot time. |
9 |
|
10 |
It is broken for *some* systems. |
11 |
|
12 |
> The problem is that with modern hardware much code that was |
13 |
> traditionally stored in /usr may be needed early in the boot sequence, |
14 |
> before /usr is mounted. The obvious case is firmware and drivers, and |
15 |
> the usual example cited is bluetooth keyboards. If you need keyboard |
16 |
> input at this time, you need to have the bluetooth daemon running, which |
17 |
> is on /usr, which is not mounted. |
18 |
|
19 |
Yes, bluetooth... the very thing that should not have come to pass. It |
20 |
is broken by design. "Wireless" is fine but the way bluetooth works... |
21 |
Back to the drawing board, please! |
22 |
|
23 |
> The solution is to use an initramfs, and on a technical level it's not |
24 |
> any different to needing a way to get the ext4 module off disk so you |
25 |
> can mount /. |
26 |
|
27 |
Yes, that is one way of solving it... But I question the sanity by |
28 |
having ext4 as a module if you know you are going to use it on your |
29 |
system; it's not as if you are going to use ext4 one day and reiserfs |
30 |
the next day and XFS the day after that, or? The only ones that benefits |
31 |
from that kind of setup is binary distros that can compile everything as |
32 |
module and probe as they load. |
33 |
I do however have some things compiled as modules (that I only load when |
34 |
needed) but those things are not needed at boot. So for my case it's not |
35 |
needed. |
36 |
|
37 |
> Some may argue that bluetooth keyboards are a rarity and that's tough. |
38 |
> Well, there's Macbook hardware, and phones which have soft keyboards. |
39 |
> But many scenarios could exist, all due to the fact that hot-pluggable |
40 |
> hardware can in theory run any arbitrary code to get itself up and |
41 |
> running, and if that code is on a volume that is not mounted... The |
42 |
> solution is obvious - all that code should be on / somewhere, or should |
43 |
> be mountable using an initramfs. |
44 |
|
45 |
Yes, *should* be. Quite optional. As it has "always" been. Just because |
46 |
people are using bluetooth devices and/or want the computer to sing and |
47 |
dance while booting should not impose restrictions to those who don't |
48 |
want that, which is why I'm protesting. |
49 |
|
50 |
> Do you see that although you and I can deal with this with relative |
51 |
> ease, Aunt Tillie probably couldn't and the junior sysadmins I have to |
52 |
> deal with certainly can't? |
53 |
|
54 |
Yes. But have Gentoo ever been a distro for Aunt Tillie or junior |
55 |
sysadmins? I don't want to discourage them to try it out of course but I |
56 |
don't want to put restrictions on myself (or others) either... |
57 |
Flexibility is the keyword here. |
58 |
|
59 |
> Personally, I think that splitting / and /usr is a daft idea: |
60 |
|
61 |
That's fine. I, respectfully, disagree. If I could break the system down |
62 |
into bits and put each bit on a separate "harddrive" with a massive I/O |
63 |
connection I would (yes, I exaggerate but I'm sure you get the idea). |
64 |
|
65 |
> a. I have multi-TB hard disks, completely unlike the 5M monsters that |
66 |
> Thomson had to deal with in the 70s |
67 |
|
68 |
Haven't you heard? Size does not matter... ;-) |
69 |
|
70 |
> b. I haven't had /usr break on me during boot requiring busybox in |
71 |
> maintenance mode for at least 5 years. Every startup failure in that |
72 |
> time required a rescue cd anyway, and I always have one of those handy |
73 |
|
74 |
I haven't had /usr break either for at least that time even though I've |
75 |
always had it separate. To me, I like to keep things organised in |
76 |
different compartments using, perhaps somewhat arbitrary, rules. |
77 |
Therefore keeping system administration tools in /sbin, user accessible |
78 |
tools in /usr/bin etc. makes perfect sense (I know you think it's |
79 |
arbitrary and I agree but it works, for me at least). There is no *real* |
80 |
need to keep /usr separate for normal users it's just that I think it's |
81 |
flexible and I want it that way. There is no right or wrong here, merely |
82 |
philosophical differences. How you solve the different problems are |
83 |
technical however. I do have a rescue USB stick handy as well though but |
84 |
since I rarely use it I tend to forget to update it... |
85 |
|
86 |
> c. it IS useful for terminal servers, but those tend to have experienced |
87 |
> sysadmins, and they really should be OK with an initramfs (or their |
88 |
> vendor should ship one) |
89 |
|
90 |
Using an initramfs means you duplicate parts of your OS and copy them |
91 |
into the kernel or using a tool (like dracut or genkernel). If you need |
92 |
it from a technical point of view (bluetooth keyboard), that's fine but |
93 |
if I don't have any hardware that requires it then why use an initramfs? |
94 |
I guess it's a matter of taste (or "philosophy" if you will)... An |
95 |
initramfs seems like bandaid to me (and it is). |
96 |
|
97 |
> I'm often at the front of the Lennart-bashing parade, and what he says |
98 |
> often makes sense but only in his narrow view of the world, but in |
99 |
> *this* case, I can't help but admit he does have a point. |
100 |
|
101 |
I don't really see it... I don't really care what Lennart does as long |
102 |
as it doesn't affect me (and he may be the greatest person that ever |
103 |
lived) but here we are... I choose to run Gentoo because it suits me |
104 |
best of all the operating systems out there. If I didn't care about how |
105 |
things works I would run Windows (or maybe MacOS). |
106 |
|
107 |
Best regards |
108 |
|
109 |
Peter K |