LWN: Comments on "LinuxCon: x86 platform drivers"

LinuxCon: x86 platform drivers

cathectic — Wed, 07 Sep 2011 08:38:08 +0000

Although that seems to have failed as well - the original WMI spec suggested
Microsoft would standarise some of the WMI GUID's for common operations (GUID's are then mapped to a particular ACPI method, but the Linux WMI driver handles that conversion for you), which would have been useful here e.g. one standard GUID and arguments for querying a battery, a standard for wireless devices, etc.

Unfortunately, Microsoft have never done any such standardisation for WMI GUID calls, so it seems to have just become yet another way to put more vendor specific rubbish into the firmware in new and exciting ways.

LinuxCon: x86 platform drivers

josh — Sat, 27 Aug 2011 00:35:56 +0000

It's the second worst platform interface ever created. Everything else is tied for first. :)

LinuxCon: x86 platform drivers

agrover — Fri, 26 Aug 2011 06:04:45 +0000

Wow, "heaven" and "ACPI" used in the same sentence. *boggle*

LinuxCon: x86 platform drivers

smoogen — Thu, 25 Aug 2011 03:03:45 +0000

Actually in the world of laptops.. this is a sign of moving away from the ARM model. Back in the scary days of 1997-2001, you could order a laptop from the same manufacturer once a month and never get the same hardware twice. Each one would have slightly different takes on stuff and you would have to guess at where they had stuck some part of the hardware this time. I am not a kernel developer but from a tech support side.. this is heaven with ACPI.

LinuxCon: x86 platform drivers

Cyberax — Thu, 25 Aug 2011 02:49:30 +0000

Well, actually kernel-level API in Windows is incredibly stable. It's possible to compile NT4-era drivers for Windows 7 with only minor modifications.

And writing an input device driver is not hard - no community to slap your wrist if you try dirty hacks, no worry about code maintainability. Any half-trained code monkey can do it. And unfortunately many do, as witnessed by various crapware platform drivers.

LinuxCon: x86 platform drivers

PaulWay — Wed, 24 Aug 2011 23:44:08 +0000

I don't know, but I suspect it's a lot easier for him to keep his sanity than it is for driver writers for Windows.

With Linux you have a developer community that prefers well-written, comprehensive solutions to badly thought out, slapdash ones. If it isn't working or is obviously a bad thing, it doesn't get included. We have an open, accountable process to show those decisions and improve things. If something needs to be fixed or improved in the kernel to make it easier for a class of drivers to work, then it usually is.

With proprietary driver writers, everything is opaque. You can't change anything outside your own little cubicle. Decisions about how you write your code and when it's ready are arbitrary and driven mostly by time to market rather than quality. Every once in a while Microsoft throws the old system in the trash, declares that .Net 17.5 is going to be the new way of doing things, it's half incompatible with what you've just written, and you have absolutely no say in this. No wonder they spend all that time writing their own wifi handler software and come up with bulky, kludgy drivers.

I'd be keeping him away from any hardware manufacturers and all sharp objects for the foreseeable future, though :-)

Have fun,

Paul

LinuxCon: x86 platform drivers

mjg59 — Wed, 24 Aug 2011 23:33:22 +0000

All the standard PC functionality is still pretty consistent between machines, but in other areas it's pretty platform dependent, yes. The main difference between this and the ARM world is that the firmware usually provides an ACPI representation of the device, meaning you can build a generic kernel.

LinuxCon: x86 platform drivers

neilbrown — Wed, 24 Aug 2011 23:08:51 +0000

So is this a case of the PC platform becoming more like the ARM platform - i.e. every board is subtly different in inconsistent ways?

LinuxCon: x86 platform drivers

flewellyn — Wed, 24 Aug 2011 22:42:25 +0000

I believe Microsoft's WMI is just such an attempt. Except, of course, for compressing the MOF code using a proprietary tool. That part is just vendor lockout.

LinuxCon: x86 platform drivers

paravoid — Wed, 24 Aug 2011 22:38:41 +0000

This may sound foolishly naive, but has anyone ever tried contacting the vendors to figure out a way for things to become more sane in the future? Granted, some of those things were done on purpose and some vendors (like Apple) won't want to help but I'm wondering about the rest some of them are even selling their computers with Linux preinstalled.

LinuxCon: x86 platform drivers

mjg59 — Wed, 24 Aug 2011 19:26:38 +0000

It's mostly only in tightly integrated systems. Some Sony and Acer all in one systems have similar features, but so far they've just reused the code from their laptops so everything pretty much just works.

In most cases these features aren't things that prevent booting (people still expect most hardware to install unmodified versions of Windows), there'll just be some extra hardware that won't work.

LinuxCon: x86 platform drivers

Tester — Wed, 24 Aug 2011 19:16:16 +0000

You assume he was ever sane...

LinuxCon: x86 platform drivers

cortana — Wed, 24 Aug 2011 19:06:46 +0000

Does this craziness tend to only affect laptop (and Apple Desktop) systems, or is it creeping into regular desktop systems too? And from there, how long will it be until I can't buy a decent consumer-facing motherboard expecting to be able to boot Linux without a lot of fiddling about?

LinuxCon: x86 platform drivers

Cyberax — Wed, 24 Aug 2011 18:54:50 +0000

Wow.

A question to Matthew Garrett, how do you manage to keep your sanity?