Linked by Mike on Fri 11th Nov 2011 18:40 UTC
Linux "At long last, it looks like there is an adequate solution to the Active State Power Management (ASPM) problem in the Linux kernel , a.k.a. the well-known and wide-spread power regression in the Linux 2.6.38 kernel, which has been causing many laptops to go through significantly more power than they should. This is not another workaround, but rather a behavioral change in the kernel to better decide when the PCI Express ASPM support should be toggled."
Thread beginning with comment 496980
To read all comments associated with this story, please click here.
Surprising
by B. Janssen on Fri 11th Nov 2011 19:23 UTC
B. Janssen
Member since:
2006-10-11

It's surprising that Matthew Garrett would have to divine this information from a Microsoft presentation on Vista's handling of PCI Express. In his e-mail to the list Mr. Garrett also says that there is "zero public documentation" on this issue. I can believe that, often engineering specs are not public and must be bought from the standard body.

But I would have expected that at least one of the commercial GNU/Linux big shots, i. e. RH, Canonical, SUSE, had done so. Apparently not!

Reply Score: 3

RE: Surprising
by Sodki on Fri 11th Nov 2011 20:16 in reply to "Surprising"
Sodki Member since:
2005-11-10

In his e-mail to the list Mr. Garrett also says that there is "zero public documentation" on this issue. I can believe that, often engineering specs are not public and must be bought from the standard body.

But I would have expected that at least one of the commercial GNU/Linux big shots, i. e. RH, Canonical, SUSE, had done so. Apparently not!


Well, sometimes you can't buy the specs because they don't exist or aren't followed by the hardware makers. And sometimes the information cannot be disclosed publicly. Put all this together and the "zero public documentation" makes perfect sense.

Reply Parent Score: 4

RE: Surprising
by mjg59 on Fri 11th Nov 2011 21:09 in reply to "Surprising"
mjg59 Member since:
2005-10-17

The mistake you're making here is to assume that there's any external documentation at all. _OSC is well defined, but the precise semantics of how hardware vendors expect it to work isn't. The FADT bit is moderately well defined, but again the precise semantics of how hardware vendors expect it to work isn't. ASPM itself is well defined, but... you get the picture. And often even the hardware vendors don't really know how it all works, they just know that they configured a machine in a specific way with a specific set of firmware configuration options and everything worked as they expected when they booted Windows.

The only really important specification here is how Windows behaves, and I doubt that's documented anywhere other than the Windows source code. So we're left trying to infer how things are meant to work by a combination of trying different things until something seems to work for everyone.

Reply Parent Score: 6

RE: Surprising
by Neolander on Sat 12th Nov 2011 08:56 in reply to "Surprising"
Neolander Member since:
2010-03-08

/rant on

So, I am an OS developer and I want to read the latest and greatest PCI Express spec.

I go to http://www.pcisig.com/home
Click "Download specifications"
Click "PCI Express"
Click "PCIe base spec 3.0" on the right
Select the version without the change bar
End up on a login window, click "how to join"
Click "Join PCI-SIG"
And then I have to fill out a 10-pages paper form, stick it in an envelope along with a $3000 check (or equivalent mean of payment), and send everything to an address in the US.

Err... seriously ?

I understand the need to pay for a spec. I mean, it takes lots of time and money to write it and all. Sure, personally, I'd end up downloading and using it illegally anyway because I'm a hobby OSdever with no income and there's no way I can put that kind of money on a personal project, but that's not the point.

The point is, if they get $3000 per year and juridic entity who deals with them, can't they come up with something... you know... faster and more secure than putting very large amounts of paper money in the postal circuit ?

It's not as if wire transfers are something new. And for all I hate credit cards, maybe I'd trust PCI-SIG enough to give them my card information. But even though they have thought about the latter possibility, they make it extra cumbersome and insecure by asking you to call their system administrator on the phone for that. Without specifying at which hours he's actually there.

I find it ironical that computer standards body, of all things, can still live in the beginning of the 20th century.

/rant off

Reply Parent Score: 3

RE: Surprising
by Lennie on Mon 14th Nov 2011 11:03 in reply to "Surprising"
Lennie Member since:
2007-09-22

Specification and implementations are never the same. Especially if the specifications are to vague.

Reply Parent Score: 3