Power Management Bugs Hold Up Some Linux Laptops Due To Regulatory Requirements

[clippy]

Wow, this is ACPI S5 so it means the "fully off" state, except for standby power for BMCs.
Geez I do wonder why my electronic devices stay consuming energy even in "powered off" mode. Almost like it's doing somthing when off, can you imagine that? Hahahaha.

Probably a consequence of the "smart sleep" Microsoft is pushing? It's sleep mode except your device stays connected to the internet and can act on it's own accord for updates, notifications, etc.

Laptops don't have BMCs, that's server hardware. There are perfectly non-paranoid reasons for your hardware to consume energy in S5. The most obvious one is timekeeping, unless you want your system clock to reset to 01/01/1970 on every restart. Many laptops also have advanced battery features like charge thresholds or USB power delivery while the machine is off. Issues such as the one outlined in the article mostly result from firmware vendors being criminally incompetent and relying on weird hacks instead of implementing things to spec. In this regard, MS has actually been a help for Linux as they mandate many firmware features to be implemented in a sane manner before certifying devices as Windows compatible. (Although they would no doubt love to have internet access to your device while it's off.)

[caligula]

It's a perfectly fine thing, and used a certain way S0 can have exactly the same power cost as S3. On Windows, it is down to the specific background tasks being run. LTT thinks it's the policy for running windows update in S0 when plugged in, persisting after the charger is removed (because they're running windows update, but not whatever implements the power management policy).

Um how? I've seen multiple comparisons and those numbers don't show anything like that.

[markpearson]

I'm the technical lead for the Lenovo Linux team. Some notes:

- This is not related to sleep (S0ix/modern standby). This is related to when your system is powered off.

- We do energy certifications for our Linux certified platforms. There are, I think, 6 different certifications worldwide, but energy star is the main one. This is a "good thing" in that it ensures that Linux on our preloads is working well (as an aside, not all vendors do energy certification, but it has definitely helped us find things that need improving in Linux).

- The way energy star works is there are four different measurements at different states that have different weightings. The off measurement has the highest weighting - so your power usage has to be very low (it's not zero). On one particular platform with Linux we were 0.16W higher than Windows and that was enough to be a big problem. It was quite hard to debug a system when turned off - and quite confusing that it was OS dependent. The testing can be affected by many other things too, so it's not always obvious where the problem is (we've failed a test due to a bad power supply before...that was frustrating once we figured out what was going on)

- Kudos should go to the AMD team who have helped us drill down into the root cause, it was not easy. As noted in Michael's article it's support that is missing from Linux and when it's all done will give better power off numbers all round (which is great!).

- As a side note, for the platforms in question we are working on a FW workaround, as it will take some time to deliver the proper fix upstream, especially with the holidays on us It was really important to understand root cause though.

Mark

[loganj]

i had an acer nitro 16 (7840hs) that when i enable Wake on Lan in bios (and only in bios not the linux) the system when turned off consumed 7-9W (from wall). but when i turn it off from windows the power consumtion was gone.
every times the system manage to wake on lan even though i didnt manange to set anything in windows since there was no option to "allow device to wake up blah blah" option in network card and even if there was no specific setting done in linux to enable WoL.
but the power consumtion when turned off from linux was the same as when the system was on and idle.

do notice i said turned off not hibernate on both OS.

[Syfer]

I'm the technical lead for the Lenovo Linux team. Some notes:

- This is not related to sleep (S0ix/modern standby). This is related to when your system is powered off.

- We do energy certifications for our Linux certified platforms. There are, I think, 6 different certifications worldwide, but energy star is the main one. This is a "good thing" in that it ensures that Linux on our preloads is working well (as an aside, not all vendors do energy certification, but it has definitely helped us find things that need improving in Linux).

- The way energy star works is there are four different measurements at different states that have different weightings. The off measurement has the highest weighting - so your power usage has to be very low (it's not zero). On one particular platform with Linux we were 0.16W higher than Windows and that was enough to be a big problem. It was quite hard to debug a system when turned off - and quite confusing that it was OS dependent. The testing can be affected by many other things too, so it's not always obvious where the problem is (we've failed a test due to a bad power supply before...that was frustrating once we figured out what was going on)

- Kudos should go to the AMD team who have helped us drill down into the root cause, it was not easy. As noted in Michael's article it's support that is missing from Linux and when it's all done will give better power off numbers all round (which is great!).

- As a side note, for the platforms in question we are working on a FW workaround, as it will take some time to deliver the proper fix upstream, especially with the holidays on us It was really important to understand root cause though.

Mark

Appreciate your work - decreasing power waste is great

[OmniNegro]

Absolute nonsense. Try doing at least 5 minutes of research next time. At least in the US, emission regulations have had a transformative effect on pollution and air quality. Read a high level summary at https://www.epa.gov/transportation-a...transportation

Has it ever occurred to you that the EPA lies? Clearly you believe those lies, so fine. Enjoy the false benefits the EPA claims. But if you want to see a perfect example of their lies, look at the page you linked in. The image from New York in 1973 and 2013 is a perfect example. Those two images are not set in the same location. One is over the ocean, the next is over a field. So side by side images are lies if the EPA is the group that produced it.

Or perhaps you want to argue that New York City is not big enough that they could find a picture from 1973 and 2013 in the same location?

While you clearly want to presume I am an idiot, and then insult me as well, I am just going to presume you believe everything the EPA ever said, and are now going to argue that I misquoted it. Go right ahead. Others can see for themselves.

And in case you did not mean to insult me and instead thought I could benefit from your prospective, thanks. But I am old and set in my ways. I do not believe as you do. I wish you well nonetheless.

[oiaohm]

The image from New York in 1973 and 2013 is a perfect example. Those two images are not set in the same location. One is over the ocean, the next is over a field. So side by side images are lies if the EPA is the group that produced it.

That one of those one what makes you curse because you know a person going to use that mistake to throw the baby out with the bath water. They quote where they got the pictures from and then you wake up it set of over 800 photo pairs taken from exactly matched up locations. Yes the person making the EPA website just happens to use a miss match. The match up pairs show the same air quality change.

There is a lot of cases of USA EPA staff producing website and documentation that is pure incompetence when it comes to handling of images. Charts are tables from the USA EPA are a lot safer bets.

Fun was 500 page document with 20 photos from the USA EPA and all the correct photos were in the document just all in the wrong places not one in the right place fun of playing copy the pictures and paste them into the right location so that description in fact matched photo. Yes this included author photo being some high rise building instead of the authors photo. It took them 8 editions to get the photos in the right place. So this is a minor level of USA EPA photo mess up. Does make you wonder who does proof checking at the USA EPA at times. Yes 6 editions of that document managed to move the photos around and still not get one in the right place. Know of a library that has all 8 editions of that report on shelf for a laugh.

[OmniNegro]

That one of those one what makes you curse because you know a person going to use that mistake to throw the baby out with the bath water. They quote where they got the pictures from and then you wake up it set of over 800 photo pairs taken from exactly matched up locations. Yes the person making the EPA website just happens to use a miss match. The match up pairs show the same air quality change.

There is a lot of cases of USA EPA staff producing website and documentation that is pure incompetence when it comes to handling of images. Charts are tables from the USA EPA are a lot safer bets.

Fun was 500 page document with 20 photos from the USA EPA and all the correct photos were in the document just all in the wrong places not one in the right place fun of playing copy the pictures and paste them into the right location so that description in fact matched photo. Yes this included author photo being some high rise building instead of the authors photo. It took them 8 editions to get the photos in the right place. So this is a minor level of USA EPA photo mess up. Does make you wonder who does proof checking at the USA EPA at times. Yes 6 editions of that document managed to move the photos around and still not get one in the right place. Know of a library that has all 8 editions of that report on shelf for a laugh.

Mistakes happen. But even if this is merely a mistake, they can edit their own website. They choose not to. The only reason I can think of that they do not is that the mistake favors the opinion they want you to believe. I never saw the 500 page document you are speaking of. And I have no intention of looking it up. I was only making the point that while I do not believe the EPA, I have reasons not to. And none of those were ever mentioned here. They are not important. You clearly want to believe the EPA. I do not have a choice. I simply cannot trust them. Again, the reasons are not being mentioned.

This article is about Laptops being forced to be certified by a governmental branch for power usage standards. That is fine. But when I heard that the new "standard" MS forced upon the world makes the system do new things without asking for permission, all I can see, hear and think of are big red flags. Not worth mentioning, but there you have it.

[agd5f]

This is basically a firmware issue that showed up because two different OSes have slightly different behaviors with respect to S5. I'm not sure if it's a bug or not. We'd need to check whether the ACPI spec says anything with respect to D states when entering S5. Windows puts devices into d3cold before entering S5, Linux does not. Windows is the larger market so it gets a lot more testing. The issue was root caused. The solutions are to either change the firmware or update Linux to align with Windows. Arguably the latter is probably a better solution because there will always be hardware that is either not tested with Linux or tested much less (just due to the relative market sizes), so the more aligned both OSes can be in this area, the better the overall user experience.

[archkde]

OK. I'll say that some of the more frequent problems I've had are to do with shutdowns. I've had systems go into tight cpu loops with fans screaming until either hard powered-off or the batteries run out, systems that sit waiting indefinitely for network hardware, or graphics hardware, or unable to work out how to dismount /var that was on a separate, encrypted partition.

Ok, I see now where the misunderstanding lies. You're talking about the shutdown process getting stuck and the system remaining visibly on, which I've certainly experienced my fair share of. But here the system looks like it's actually turned off properly, but somehow internally isn't and continues consuming significant power.