How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
The ease of writing drivers; Linux drivers are at normally one-third smaller than Windows drivers or other operating system drivers. We have all the examples there, so it's trivial to write a new one if you have new hardware, usually because you can copy the code and go. We maintain them for forever, so the old ones don't disappear and we run on every single processor out there. I mean Linux is 80% of the world's top 500 super computers right now and we're also the number one embedded operating system today."
Posted Nov 3, 2008 14:53 UTC (Mon)
by nix (subscriber, #2304)
[Link] (6 responses)
(I think I'd have liked the content if it wasn't such a straight transcription!)
Posted Nov 3, 2008 19:15 UTC (Mon)
by chromatic (guest, #26207)
[Link] (5 responses)
Posted Nov 3, 2008 23:15 UTC (Mon)
by nix (subscriber, #2304)
[Link]
Nice questions, nicely-run interview, nice webpage, shame about the
-- N., ran a school newspaper fifteen years ago so knows everything
Posted Nov 4, 2008 1:31 UTC (Tue)
by ballombe (subscriber, #9523)
[Link]
Posted Nov 4, 2008 3:54 UTC (Tue)
by BrucePerens (guest, #2510)
[Link]
Posted Nov 4, 2008 10:17 UTC (Tue)
by ekj (guest, #1524)
[Link] (1 responses)
Posted Nov 4, 2008 19:15 UTC (Tue)
by chromatic (guest, #26207)
[Link]
Posted Nov 3, 2008 15:42 UTC (Mon)
by JoeF (guest, #4486)
[Link] (11 responses)
Well, not quite. I have a framegrabber card that is not supported anymore, and the code actually was removed from the V4L subsystem.
Posted Nov 3, 2008 16:17 UTC (Mon)
by alex (subscriber, #1355)
[Link] (1 responses)
What card was it?
Posted Nov 3, 2008 22:13 UTC (Mon)
by JoeF (guest, #4486)
[Link]
Posted Nov 3, 2008 16:41 UTC (Mon)
by iabervon (subscriber, #722)
[Link] (8 responses)
Posted Nov 3, 2008 20:55 UTC (Mon)
by avik (guest, #704)
[Link] (7 responses)
Posted Nov 3, 2008 22:16 UTC (Mon)
by iabervon (subscriber, #722)
[Link]
Posted Nov 3, 2008 22:20 UTC (Mon)
by JoeF (guest, #4486)
[Link] (5 responses)
Indeed.
In my case, there are still people who have cards with the chip in question, I have submitted a patch, but even so, the V4L maintainers apparently don't care...
Posted Nov 4, 2008 14:00 UTC (Tue)
by nhippi (subscriber, #34640)
[Link] (4 responses)
Looking at that thread, you were just not persistent enough. It's not really fair to complain that others don't care when you don't seemingly care enough to keep reposting the patch/driver to v4l list until you get a reaction out of them ;)
If the subsystem maintainer is not reacting your patches, ask others for help. Greg's Linux Driver Project or linux-staging are good places to start at. If you feel your patch is already in good condition, you can ask Andrew Morton to include it in his tree.
Posted Nov 4, 2008 16:43 UTC (Tue)
by JoeF (guest, #4486)
[Link]
Posted Nov 4, 2008 17:14 UTC (Tue)
by JoeF (guest, #4486)
[Link] (2 responses)
Posted Nov 5, 2008 10:18 UTC (Wed)
by nhippi (subscriber, #34640)
[Link] (1 responses)
from the original thread:
- already been marked as BROKEN in 2.6.0 three years ago and
Without warning eh?
> Here, I was really just commenting on Greg's claim that Linux keeps drivers forever, which clearly isn't true.
Well lets refine that. Linux keeps any driver forever, as long there is atleast one person who finds enough time at least once in three years to keep the driver functioning ;)
Vendors have a clear interest to write drivers for their current hardware, but for legacy drivers Linux depends on volunteers on their limited free time. The mythical "someone else" has her own itches to scratch, and unless you are very lucky, it isn't "cleaning up drivers for esoteric hardware I don't own".
Posted Nov 5, 2008 15:56 UTC (Wed)
by JoeF (guest, #4486)
[Link]
Well lets refine that. Linux keeps any driver forever, as long there is atleast one person who finds enough time at least once in three years to keep the driver functioning ;)
I did. I originally submitted my patches to the person listed as maintainer of the code.
Posted Nov 3, 2008 15:54 UTC (Mon)
by mrshiny (guest, #4266)
[Link] (5 responses)
Posted Nov 3, 2008 16:15 UTC (Mon)
by alex (subscriber, #1355)
[Link] (1 responses)
It's a shame your having issues with your WiFi card. WiFi has been one
Posted Nov 3, 2008 20:22 UTC (Mon)
by mrshiny (guest, #4266)
[Link]
As for out-of-tree drivers being sub-optimal: they "break" because the kernel devs don't maintain a stable API (nevermind ABI) for any length of time and thus nobody even bothers to try to ship out-of-tree drivers. GKH has his stable-api-nonsense doc but frankly I'm not convinced. I have issues with my hardware, under Linux, all the time because the drivers have to be shipped with the kernel; I can demonstrate actual harm caused by this policy.
It boils down to this: I HAD a working WiFi driver. However in order to get security updates and other updated drivers, I updated my kernel and WiFi stopped working. Of every 2.6.25 release Fedora 8 had, only ONE worked for me, and none of the later ones have worked at all. I don't mess around with kernel compilation anymore (isn't Linux supposed to just work now?) but I can install a driver if I have to... lots of regular users know how to install a driver; in the Linux world you just can't do that, you need to upgrade everything. Changing thousands of lines of code when I only need a few changed is simply bad risk management and bad engineering.
Posted Nov 3, 2008 16:30 UTC (Mon)
by drag (guest, #31333)
[Link]
For example this is my situation:
I am using Debian unstable on my laptop. I bought my 1420n laptop a while ago and it came with Ubuntu pre-installed on it. At that point everything 'just worked', including my wireless. My wireless card is Intel 3945ABG.
When I installed Debian over Ubuntu wireless became unreliable. Eventually I found that compiling the out of kernel code from the Iwlwifi project made my card work quite reliable.. however there were still Intel firmware bugs that caused my card to fail when connecting and getting disconnected from some specific routers. It would just stop working and sometimes rmmod and modprobe made it work again, but not always.
However since the Iwlwifi stuff has been incorporated into the kernel and intel update it's firmware... it's been dead stable.
---------------------
Keep in mind that I've run into XP machines that have had VERY unreliable wireless cards due to Intel firmware bugs. Now they 'just worked' in Linux because Linux was much more capable of dealing with intermittent hardware failures gracefully.
Posted Nov 4, 2008 3:57 UTC (Tue)
by BrucePerens (guest, #2510)
[Link] (1 responses)
Posted Nov 4, 2008 6:19 UTC (Tue)
by jamesh (guest, #1159)
[Link]
That said, the problem code that was writing to the firmware (dynamic ftrace) could easily have been stomping over other drivers' IO space in a way that causes malfunctions but not firmware destruction. There were reports of ipw3945 hangs during boot on beta versions of Ubuntu 8.10 that stopped occuring when the dynamic ftrace code was disabled, so that is certainly something to keep in mind.
Posted Nov 3, 2008 20:02 UTC (Mon)
by jwb (guest, #15467)
[Link] (9 responses)
Posted Nov 3, 2008 20:28 UTC (Mon)
by ca9mbu (guest, #11098)
[Link] (4 responses)
Maybe what you meant was that the full 3D capabilities of the card are not supported in upstream Linux releases? That's a far cry from the card itself not working/being supported at all which is how I read your original comment.
Such support is available via binary Nvidia-supplied drivers. The lack of upstream support for the full Nvidia GPU functionality is noone's fault other than Nvidia's themselves (or, possibly, more specifically, their lawyers).
I'm not sure that anything remotely like the following is possible to get an answer for, but I think a much fairer question would be "of those devices that we have documentation for or are otherwise able to figure out how they work, how many are a) upstream or b) otherwise available as a source tarball".
Posted Nov 4, 2008 1:27 UTC (Tue)
by JoeBuck (subscriber, #2330)
[Link] (3 responses)
Posted Nov 4, 2008 4:00 UTC (Tue)
by BrucePerens (guest, #2510)
[Link] (1 responses)
Posted Nov 4, 2008 19:50 UTC (Tue)
by pr1268 (guest, #24648)
[Link]
Wow! I wasn't even aware that nVidia is having reliability issues with their chips. The graphics card currently in my home PC is an nVidia 6600GT, purchased over a year ago, with a different brand name (EVGA) and a sticker on the box proclaiming a "Lifetime Warranty". I've never had any hardware issues with an nVidia card other than the occasional on-card fan breaking. (Disclosure: I use the proprietary binary driver in order to utilize all the capabilities of the graphics card. I do wish nVidia would be more Open-Source friendly, but their current "support" for Linux is about as good as I can expect for now.) You make a good point, Bruce, about nVidia possibly owing millions to the other hardware vendors... they may be a takeover target soon, but with the economy in the pan I'm a little skeptical for now.
Posted Nov 4, 2008 9:30 UTC (Tue)
by ca9mbu (guest, #11098)
[Link]
Posted Nov 4, 2008 6:30 UTC (Tue)
by alankila (guest, #47141)
[Link] (1 responses)
Posted Nov 4, 2008 8:49 UTC (Tue)
by Los__D (guest, #15263)
[Link]
Posted Nov 4, 2008 12:49 UTC (Tue)
by alecs1 (guest, #46699)
[Link] (1 responses)
Fglrx doesn't do that for this computer, never did for the other one. Neither radeonhd. At this time I am very unhappy for falling in the marketing trap of radeonhd and buying a AMD/ATI card.
Listening to this interview made me feel good for the duration of the audio recording. But in reality I am very unhappy with my laptop support. Actually, I will be happy when my bug reports get some sort of attention (Debian and ALSA bts, fixing the problems myself is well beyond my capabilities, although I am programmer myself).
Posted Nov 4, 2008 15:55 UTC (Tue)
by drag (guest, #31333)
[Link]
So between the both of them you have the ability to the majority of video hardware under linux. Majority by a large margin.
Like Wifi cards the ability for Linux to support these things is largely dependent on the existing subsystems in the kernel... Without good code base that driver developers can take from and depend on making video card drivers is exceptionally difficult. Just like trying to support Wireless devices without a proper 802.11 protocol stack to depend on.
So by having at least two different types of hardware, three if you include the Nouveau project, the Linux devs should be able to have the facilities and non-video card specific video card code to get supporting new devices much easier.
Also you have to depend on X.org to keep up with Linux and work with Linux developers. So that adds a entirely new twist on everything.
With things like GEM, KMS, Wayland, DRI2. Gallium, etc.. Linux may be in pretty good shape in another year. OSS may even be competitive in this department with hardware that has been out for a while.
Posted Nov 4, 2008 4:05 UTC (Tue)
by BrucePerens (guest, #2510)
[Link] (3 responses)
Posted Nov 4, 2008 20:01 UTC (Tue)
by JoeF (guest, #4486)
[Link] (2 responses)
Indeed. And taking code for legacy devices out of the kernel doesn't help.
There are still tons of scanners not in SANE yet
Yeah, another sore point of mine. My Brother MFC printer/scanner is not supported by SANE...
Posted Nov 6, 2008 23:56 UTC (Thu)
by hppnq (guest, #14462)
[Link] (1 responses)
Posted Nov 7, 2008 19:15 UTC (Fri)
by jonasj (guest, #44344)
[Link]
Posted Nov 5, 2008 10:13 UTC (Wed)
by renox (guest, #23785)
[Link] (2 responses)
Because even though Windows itself doesn't have a driver doesn't mean that the HW is unusable under Windows, it just mean that you have to install the manufacturer provided CD which is usually very simple..
What matter to users isn't so much whether the OS includes the driver or not but whether he can have his HW working (without too much trouble) or not.
Posted Nov 5, 2008 10:43 UTC (Wed)
by tzafrir (subscriber, #11501)
[Link] (1 responses)
With windows with only works if I carry the drivers disk around with me. Do I really need to carry it around with me? Does it have the latest version?
Posted Nov 5, 2008 16:34 UTC (Wed)
by renox (guest, #23785)
[Link]
Usually for Windows you don't carry around your driver disk, you download the driver from the manufacturer so that in one step you have the latest version.
Look, I don't dispute that having drivers integrated on the OS is very convenient, I'd just like to know what 'support' means here 'integrated support within the OS', yes in this case Linux has probably more drivers than Windows, if it's 'support' as in 'work' then I'm more surprised.
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
and several areas which were almost incomprehensible (as in, you needed to
know Greg's opinion on a subject in order to know what he was saying).
This is normal in any verbal interview: generally one *does* in fact clean
these things up. (If you're concerned, the thing to do is run it past Greg
again and let him OK the changed version.)
transcription is all :(
(because people who are almost totally ignorant in a subject are always
know-it-alls!)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
Despite this claim, I doubt Greg ever uttered "thoguht" and "awht".
Oh. Perhaps we should suggest that people you interview in the future use email, then. Most are just not capable of producing text-quality material when speaking. I just hate when direct transcriptions of my off-the-cuff speech go to print.
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
Deleting drivers
Deleting drivers
Admittedly, the driver was broken and the maintainer didn't seem to be interested in it anymore. I had sent her my patches to make it work.
It is for the Zoran 36120 chip, btw.
http://marc.info/?l=linux-video&m=116448212417204&...
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
That's a relatively short definition of 'forever'.
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
I have neither the time nor the inclination to play all the political games to get something included in the kernel. I play with the kernel in my limited spare time...
Here, I was really just commenting on Greg's claim that Linux keeps drivers forever, which clearly isn't true.
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
Cleaning it up I could have understood, but taking it out? And pretty much without warning. A simple question on the V4L list like "anybody still using this" would have helped. Indeed, there have been such questions before, and I had chimed in at that point.
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
- is still marked as BROKEN.
As I mentioned, I had sent my patches to the maintainer, which I think is the right thing to do.How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
Now, the *maintainer* was the person who didn't maintain the code anymore.
And it *was* known to the V4L list that there are people who still use this chip.
There was *no* communication on the list about "we want to remove this, unless somebody is fixing it". So, I maintain that the *removal* came without warning.
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
Out of tree doesn't work
been shown to be suboptimal. They break because there isn't a stable
internal API and that is a design decision in the kernel. If you want
a stable driver ABI that allows that I suggest you try Solaris,
assuming it supports your device of course.
of the late developers for Linux support for a myriad of reasons.
Hopefully things like the new regulatory framework stuff will make it
better. I assume you've raised bug reports for your breakage? Is it
Fedora breakage or mainline kernel breakage?
Out of tree doesn't work
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
Not supporting 3d means not supporting the vast majority of the hardware in a video card, and not being able to use the most computationally powerful part of your machine. Too bad nVidia won't play nice, but please don't pretend this isn't a concern.
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
nVidia chip quality declining?
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
Legacy devices are of diminishing importance - unless you have one. We still have lots of problems with wifi cards that require use of ndiswrapper, etc. There are still tons of scanners not in SANE yet (pun intended). So perhaps we should not celebrate victory until we can say we have more devices when the list is weighted by the number of devices sold and still existing, etc.
More of today's devices?
Legacy devices are of diminishing importance - unless you have one.
More of today's devices?
Are you sure? My Brother is in sane, has been for a while.
More of today's devices?
My Brother is in sane, has been for a while.
I hope he gets better soon ;-)
More of today's devices?
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)
How Linux Supports More Devices Than Any Other OS, Ever (O'Reilly)