Hacker News Reader

Comment by nilsb | Hacker News Reader

The article mentions that "Unlike most SBCs, the Pinephone contains a rechargeable battery intended to power the device. Correct configuration of the charging circuits, including various safety features such as thermal protection will not be enabled by the current OpenBSD kernel as of the time of writing."

Is that really the case? If so, that seems like an unnecessarily dangerous design. I would expect safety features not to rely on (user-replaceable) software.

View on HN

There was a lengthy HN thread discussing the question (software-based thermal protection):

https://news.ycombinator.com/item?id=24596248 ("Let's talk about safety of Pinephone (xnux.eu)") (2020)

I'd rather have features rely on user-controlled software than hardware-vendor-controlled software.

Do you think thermal protection is an optional feature? Because that's what would happen in some cases if it's only defined through software.

Every electronics project working with Li-Ion batteries in my opinion should use an overcharge/discharge protection circuit, it's basic safety.

I don't want software to be able to control whether or not my phone turns into a bomb

Edit: from this discussion https://news.ycombinator.com/item?id=24596248 it looks like it is using http://files.pine64.org/doc/datasheet/pine64/AXP803_Datashee... as a separate IC

> You think lack of thermal protection is a feature?

I read that as: vendor-controlled software sitting in hardware will be buggy and those bugs won't get fixed. Thermal protection implemented in user-controlled software can, over the long term, provide better safety because bugs can be fixed.

Then stick it in a user-flashable chip, but still ship it pre-flashed. The point it that failing to init something at runtime should never be a safety problem.

How is a user flashable chip any different than shipping it preconfigured to be safe in software?

The idea that there's more safety with physical separation but the same level of configurability seems incorrect at a glance.

People will misconfigure either way. Freedom is always safer given a large enough window, just maybe not always safer in the short term.

The point is that if you run a new OS on the hardware (say, OpenBSD...) and it doesn't set things up, that failure to configure the running system should not result in a safety problem. Now one obvious way to do this is to do it in hardware, but the point was made that then you have to hope that the hardcoded safety measures are bug-free. So, the obvious fix is to do it in firmware, and allow updating the firmware, but do ship with it pre-flashed in persistent memory. Thus, doing nothing is safe, if a bug is found you can fix it, and yes a user could flash unsafe firmware but they'd have to go out of their way to do so; the default is to be safe regardless of what the main system OS does.

So much energy being wasted in this thread...

That's already how it works.

The PMIC can be configured once, and it retains its configuration across reboots until OS reconfigures it.

A user has to go out of their way to flash an OS too, I'm not sure I buy the argument that open architecture is going to surprise people who also know how to flash an OS to a very new and unsupported device. No one is trivially flashing different OSes to pinephone and thus they'd need a good deal of domain knowledge to even achieve this goal.

Effectively that means no one is flashing an OS to the pinephone, getting it to work but somehow is ignorant to the fact that pinephone has a battery and that battery could present danger if handled improperly.

It seems like a safety measure that increases complexity which could potentially decrease safety just as likely as it is to increase safety.

Someone could accidentally flash an invalid image to a phone. Happens all the time. Or maybe they flash the latest Linux kernel dev branch and it crashes. That shouldn't disable battery safety.

okay so if faulty flashing is the problem, how does flashing two parts of the device make flashing better?

They're independent of each other, so flashing the OS can never damage the other chip. Isn't it obvious?

We are literally having this conversation on an article about a case where it would be useful; if battery control was on a separate controller that was flashed separately, then you could install OpenBSD and not worry about starting a fire.

What do you mean? Developing a chip with contained firmware isn't free or equivalent to software definitions, if that feature was necessary you may not even yet be in a position where you have a pinephone the even flash a new OS onto.

Further the article has domain knowledge and knew about this issue. Making my point about the type of person capable of flashing another OS as being someone capable of understanding the risks.

I'm not suggesting that it's not FOSS firmware, just that it runs on its own controller and is flashed independently of the main system. Kind of like how the Pinephone modem runs its own Linux, but without the blobs that that uses (modems are more proprietary than power controllers need to be)

Understanding the risks is what makes a person want a design that eliminates them!

As stated before, moving the risk somewhere else does not eliminate it.

You can and should eliminate this particular risk when flashing the OS.

Because reflashing the chip that stores the OS can't affect the battery. It's simple physical isolation. What if the entire chip with the OS physically fails? The code with the protection should still run.

>You think lack of thermal protection is a feature?

Where in my comment did I say that?

Fair enough, I've reworded my post to more accurately convey my point

When dealing with Lithium cells, the less software the better. If there's need to change dis/charging parameters from software, then that must be done behind a hardware protection with higher priority, so that the users still have control but cannot play outside safety limits. Not implementing a hardware safety seems really a bad deign choice to me.

I'd rather such things be a purely hardware implementation.

Knowing BSDs, this is codeword for "we don't know how to do it, nor do we care to."

Arch user right?

Correct

Reply on news.ycombinator.com