| View previous topic :: View next topic |
| Author |
Message |
BoneKracker
Veteran
Joined: 14 Mar 2006
Posts: 1499
Location: U.S.A.
|
 Posted: Fri Feb 08, 2013 10:23 pm Post subject: What does udev's "hwdb" actually do? |
 |
|
I'm looking for a fairly detailed explanation that can serve as a basis for determining if one wants it or not. At present, I have chosen to disable it.
_________________
Oldthinkers unbellyfeel INGSOC.
-- Headline of a document on Winston Smith's terminal in his cubicle at the Ministry of Truth, seen briefly in the background in one scene of the movie rendition of Nineteen Eighty-Four. |
|
| Back to top |
|
 |
sikpuppy
n00b
Joined: 12 Jun 2012
Posts: 25
Location: Central Coast, NSW
|
| Posted: Sun Feb 10, 2013 5:22 am Post subject: |
|
|
| Isn't it just a (human readable) list of manufacturers, devices, device IDS etc? So if you plug your iFoobar in it recognises it as an iFoobar. |
|
| Back to top |
|
|
BoneKracker
Veteran
Joined: 14 Mar 2006
Posts: 1499
Location: U.S.A.
|
| Posted: Sun Feb 10, 2013 5:50 am Post subject: |
|
|
That would be a logical assumption, but where is it documented? Who creates it? What's in it, and what's not, and why?
Is it not the kernel which recognizes hardware and populates sysfs with such information? If so, then how does this hwdb relate to that, and why is it not part of that functionality of the kernel?
My hardware seems to work fine, so how do I know if I want to use it?
_________________
Oldthinkers unbellyfeel INGSOC.
-- Headline of a document on Winston Smith's terminal in his cubicle at the Ministry of Truth, seen briefly in the background in one scene of the movie rendition of Nineteen Eighty-Four. |
|
| Back to top |
|
|
sikpuppy
n00b
Joined: 12 Jun 2012
Posts: 25
Location: Central Coast, NSW
|
| Posted: Sun Feb 10, 2013 6:01 am Post subject: |
|
|
| BoneKracker wrote: | That would be a logical assumption, but where is it documented? Who creates it? What's in it, and what's not, and why?
Is it not the kernel which recognizes hardware and populates sysfs with such information? If so, then how does this hwdb relate to that, and why is it not part of that functionality of the kernel?
My hardware seems to work fine, so how do I know if I want to use it? |
http://pci-ids.ucw.cz/
That explains a bit. It's mainly to name devices, otherwise I don't see the need for it.
http://pci-ids.ucw.cz/v2.2/pci.ids
Text file with device to device naming. |
|
| Back to top |
|
|
BoneKracker
Veteran
Joined: 14 Mar 2006
Posts: 1499
Location: U.S.A.
|
| Posted: Sun Feb 10, 2013 6:03 am Post subject: |
|
|
Okay, that's what you meant by "human readable". It still seems redundant with the kernel, which has a database of vendor names and associated codes, and a data model which drivers are supposed to conform to which allows devices and drivers to be identifiable in sysfs.
I'm not saying it shouldn't exist; I'm just saying I don't understand why it does.
_________________
Oldthinkers unbellyfeel INGSOC.
-- Headline of a document on Winston Smith's terminal in his cubicle at the Ministry of Truth, seen briefly in the background in one scene of the movie rendition of Nineteen Eighty-Four.
Last edited by BoneKracker on Sun Feb 10, 2013 6:08 am; edited 1 time in total |
|
| Back to top |
|
|
sikpuppy
n00b
Joined: 12 Jun 2012
Posts: 25
Location: Central Coast, NSW
|
| Posted: Sun Feb 10, 2013 6:07 am Post subject: |
|
|
| BoneKracker wrote: | Okay, that's what you meant by "human readable". It still seems redundant with the kernel, which has a database of vendor names, device names, and associated hardware codes.
I'm not saying it shouldn't exist; I'm just saying I don't understand why it does. |
But does the kernel pass the name along with the hardware codes? I get the impression it doesn't and the names are not resolved properly until udev is started. |
|
| Back to top |
|
|
BoneKracker
Veteran
Joined: 14 Mar 2006
Posts: 1499
Location: U.S.A.
|
| Posted: Sun Feb 10, 2013 6:17 am Post subject: |
|
|
Maybe it doesn't, or doesn't always, but it seems to me that if there's a basic function already being performed to identify devices and tabulate their various characteristics, why should a completely separate function be required merely to layer on top of that some additional information.
I suppose it must be a kernelspace vs userspace thing. The kernel doesn't need to know the human-readable names, but it's useful in userspace. Maybe I'll shut up and first read what you've provided though. 
Edit: okay I looked at it. Now I'm left wondering why create something like this just for PCI? Wouldn't it be best to ingrate some information model across all the buses and classes of devices and so on?
Ah well, I suppose I don't really need to know.
_________________
Oldthinkers unbellyfeel INGSOC.
-- Headline of a document on Winston Smith's terminal in his cubicle at the Ministry of Truth, seen briefly in the background in one scene of the movie rendition of Nineteen Eighty-Four. |
|
| Back to top |
|
|
sikpuppy
n00b
Joined: 12 Jun 2012
Posts: 25
Location: Central Coast, NSW
|
| Posted: Sun Feb 10, 2013 6:25 am Post subject: |
|
|
| BoneKracker wrote: | Maybe it doesn't, or doesn't always, but it seems to me that if there's a basic function already being performed to identify devices and tabulate their various characteristics, why should a completely separate function be required merely to layer on top of that some additional information.
I suppose it must be a kernelspace vs userspace thing. The kernel doesn't need to know the human-readable names, but it's useful in userspace. Maybe I'll shut up and first read what you've provided though. |
It doesn't seem to be very useful if all the peripherals work and you don't change your hardware configuration much. I do seem to recall that the parsing of the database while certain packages are compiling is very slow, even on a fast processor, to the point I thought the computer had crashed. Anyway, this isn't useful, so good luck, hope someone else has more definitive advice than I.  |
|
| Back to top |
|
|
BoneKracker
Veteran
Joined: 14 Mar 2006
Posts: 1499
Location: U.S.A.
|
| Posted: Sun Feb 10, 2013 6:28 am Post subject: |
|
|
Your answer was helpful. Thanks.
_________________
Oldthinkers unbellyfeel INGSOC.
-- Headline of a document on Winston Smith's terminal in his cubicle at the Ministry of Truth, seen briefly in the background in one scene of the movie rendition of Nineteen Eighty-Four. |
|
| Back to top |
|
|
Darth Marley
Tux's lil' helper
Joined: 25 Jan 2007
Posts: 105
|
| Posted: Tue Feb 12, 2013 3:24 pm Post subject: |
|
|
| Would it be useful in implementing a live CD? |
|
| Back to top |
|
|
RobPearce
n00b
Joined: 12 Jul 2005
Posts: 42
Location: Cambridge, England
|
| Posted: Tue Apr 30, 2013 7:40 am Post subject: |
|
|
| sikpuppy wrote: |
But does the kernel pass the name along with the hardware codes? I get the impression it doesn't and the names are not resolved properly until udev is started. |
The kernel "passes" a default name. Actually, since udev-197, the kernel creates the device node via the devtmpfs support. Udev then does nothing with that except add symlinks. Any attempt to rename it is met with an error message.
This is, of course, all in COMPLETE CONTRADICTION to the original stated aim of UDEV that there should be no naming policy in the kernel. UDEV is now completely broken in this regard.
Now, if you're saying that hwdb might actually allow me to get round the catastrophic brokenness then it's worth investigating. |
|
| Back to top |
|
|
Prenj
n00b
Joined: 20 Nov 2011
Posts: 8
Location: Mostar, BiH
|
| Posted: Tue Apr 30, 2013 8:37 am Post subject: |
|
|
| is it part of systemd? |
|
| Back to top |
|
|
sikpuppy
n00b
Joined: 12 Jun 2012
Posts: 25
Location: Central Coast, NSW
|
| Posted: Tue Apr 30, 2013 8:56 am Post subject: |
|
|
| Prenj wrote: | | is it part of systemd? |
No. |
|
| Back to top |
|
|
BoneKracker
Veteran
Joined: 14 Mar 2006
Posts: 1499
Location: U.S.A.
|
| Posted: Tue Apr 30, 2013 2:07 pm Post subject: |
|
|
Everything is part of systemd. If you drew a Venn digram, you could make a big circle and label it "Everything", then draw an even bigger one around it, and label that "systemd".
Here are the only things systemd is missing:
- kernel
- ntpd
- sshd
- MS Office
- iTunes
_________________
Oldthinkers unbellyfeel INGSOC.
-- Headline of a document on Winston Smith's terminal in his cubicle at the Ministry of Truth, seen briefly in the background in one scene of the movie rendition of Nineteen Eighty-Four. |
|
| Back to top |
|
|
Prenj
n00b
Joined: 20 Nov 2011
Posts: 8
Location: Mostar, BiH
|
| Posted: Tue Apr 30, 2013 2:26 pm Post subject: |
|
|
| BoneKracker wrote: | Everything is part of systemd. If you drew a Venn digram, you could make a big circle and label it "Everything", then draw an even bigger one around it, and label that "systemd".
Here are the only things systemd is missing:
- kernel
- ntpd
- sshd
- MS Office
- iTunes |
Lets not forget Lotus Notes! |
|
| Back to top |
|
|
BoneKracker
Veteran
Joined: 14 Mar 2006
Posts: 1499
Location: U.S.A.
|
| Posted: Tue Apr 30, 2013 6:10 pm Post subject: |
|
|
Yeah. "Speaking of bloatware." The main thing I remember about Lotus Notes is that it was complicated and butt-ugly. One of my former employers used it, and had tens of thousands of "databases".
_________________
Oldthinkers unbellyfeel INGSOC.
-- Headline of a document on Winston Smith's terminal in his cubicle at the Ministry of Truth, seen briefly in the background in one scene of the movie rendition of Nineteen Eighty-Four. |
|
| Back to top |
|
|
|
Off the Wall
