Apple M1 Patches For The Linux Kernel Sent Out A Third Time

[AdrianBc]

how is awareness of the problem going to help?

For example, if you want to have an Apple computer with M1, you can wait a few more months before buying one, until this and other reported software problems are solved.

If you already have one, you could shut it down and avoid using it until the next OS update.
The wear is really excessive for an SSD that cannot be replaced.

[CochainComplex]

No, the SSD wear story appears to be true.

There was a discussion thread about this on HackerNews, and many Apple M1 owners and also owners of older Apple computers shared the values reported for the amount of writes to the SSD since purchase.

The values for older Apple computers were normal, but the values for all new Apple M1 computers were more than an order of magnitude higher than expected, i.e. the amount of writes after a month were about what would have been expected after a year or two.

It is likely that whatever bug causes this will be corrected soon in some future version of Mac OS, but until then it is good to be aware of this problem.

Good for apple. Then they can sell a new replacement board or a new computer. because the ssd is hard soldered (?) to the board. .... it is not by fault it is intended I would guess.

[brunosalezze]

I would buy an M1 Mac mini if linux was an option. Move to a smaller case and keep the low power computer headless all the time

[name99]

That's why I have an additional internal HDD instead of DVD drive, keeping most caches (excluding few database files that slow down eg. Spotify startup), logs, games and documents on it instead of a SSD.

But oh wait, recent Apple's hardware doesn't have such possibility.

Exactly what part of your “solution”, should you feel it important to use, demands an INTERNAL disk?
You do realize Macs still support external disks, even external HDs, just like they always have?

If you’re going to complain about Macs, at least attack something that corresponds to a real problem.

[name99]

I would buy an M1 Mac mini if linux was an option. Move to a smaller case and keep the low power computer headless all the time

You do realize Linux is an option today? Just use a VM.
If that’s not acceptable, you might want to consider why exactly? Afraid OSX will give you cooties even if it is just running in the background?

[name99]

No, the SSD wear story appears to be true.

There was a discussion thread about this on HackerNews, and many Apple M1 owners and also owners of older Apple computers shared the values reported for the amount of writes to the SSD since purchase.

The values for older Apple computers were normal, but the values for all new Apple M1 computers were more than an order of magnitude higher than expected, i.e. the amount of writes after a month were about what would have been expected after a year or two.

It is likely that whatever bug causes this will be corrected soon in some future version of Mac OS, but until then it is good to be aware of this problem.

What is true is that a particular app reports particular numbers. Anything beyond that is speculation.
Does a different (OSX based) app like, eg, DriveDX report similar problematic numbers?
Do the numbers represent a bug in the driver, or in the reporting hardware?
Do they represent something different on an M1 Mac (like data volume through the crypto system, or whatever, that are being misinterpreted?

The lack of interest in following up on these obvious questions, IMHO, tells you everything you need to know about the people peddling this story. DriveDX, for example, is hardly an obscure app — but it has played no role in this story. Strange, no?

[name99]

Must be the 'rona. I was working at DEC (later Compaq) in the mid to late 90's. I remember when porting Linux to the Alpha processor was just getting going, I think that was around 1995. DEC, being first and foremost an engineering company, gladly provided all the hardware specifications needed to do the port. That was when John "maddog" Hall was still working there. Of course once they saw the potential, they started pouring resources into tuning and refining Linux on the Alpha, eventually closing a number of large supercomputer deals in the late 90's with Linux as the OS. They saw no issue with selling Linux on Alpha vs. their own Tru64 UNIX, and why should they? They both drove hardware sales, services, and support revenue streams. It was a win-win.

Compare that to the arrogant pricks in Cupertino today. You want what?? You mean you aren't buying a Mac? No? Kthxbye. Honestly, Apple is to hardware what Microsoft is to operating systems. A bunch of closed proprietary crap that has no intention of inter-operating with anything else.

YOU want Apple to port Linux to their hardware but it’s the “pricks in Cupertino” that are arrogant? OK, then...

[AdrianBc]

What is true is that a particular app reports particular numbers. Anything beyond that is speculation.
Does a different (OSX based) app like, eg, DriveDX report similar problematic numbers?
Do the numbers represent a bug in the driver, or in the reporting hardware?
Do they represent something different on an M1 Mac (like data volume through the crypto system, or whatever, that are being misinterpreted?

The lack of interest in following up on these obvious questions, IMHO, tells you everything you need to know about the people peddling this story. DriveDX, for example, is hardly an obscure app — but it has played no role in this story. Strange, no?

The statistics for any SSD are reported in a standard way, according to the SMART specification, by the firmware of the SSD.

It is quite difficult for a an application that reads the SMART counters to show wrong values. Even if that were true, such wrong values would then be shown for any SSD.

It is very unlikely that whatever application shows the SMART counters on the new Apple computers with M1 has any changes in the way how the counters are read and displayed compared to the versions from the older Apple computers, which showed correct values.

The only way that your supposition could be true, i.e. that the high quantity of writes would be just apparent, not real, would be if the bug would be inside the firmware of the SSD, in the implementation of the SMART statistics.

For a standard SSD, it would have been almost impossible for such a bug to exist. However, it is said that Apple have their own custom SSD controller, so there is a small chance that you are right and the Apple SSD controller has a bug in the SSD statistics.

In any case, there is a bug introduced by Apple in their own software. The bug could be indeed in the SSD statistics, but that is very unlikely, because this is just an extremely simple function. It is far more likely that there is a more subtle bug in some other part of the Apple operating system, where they might write more data than necessary, maybe because there was not enough experience with real usage of the computer, to see how large the quantity of written data will become.


Even if the large quantity of written data would be just apparent and not real, it would not matter much, because the operating system will decide at some point to make the SSD read-only, waiting for a replacement that can be done only in a service center, by soldering.

Even if the SSD would not be really worn out, the operating system cannot know that, so it will decide that the SSD cannot be used any more, based on the false statistics.

But, like I said it is hard to believe that such a bug could occur in the easiest and simplest code, instead of being in some complex code with hard-to-predict effects.

Such a simple bug would have been already corrected until now, after the first reports.

[name99]

The statistics for any SSD are reported in a standard way, according to the SMART specification, by the firmware of the SSD.

It is quite difficult for a an application that reads the SMART counters to show wrong values. Even if that were true, such wrong values would then be shown for any SSD.

It is very unlikely that whatever application shows the SMART counters on the new Apple computers with M1 has any changes in the way how the counters are read and displayed compared to the versions from the older Apple computers, which showed correct values.

The only way that your supposition could be true, i.e. that the high quantity of writes would be just apparent, not real, would be if the bug would be inside the firmware of the SSD, in the implementation of the SMART statistics.

For a standard SSD, it would have been almost impossible for such a bug to exist. However, it is said that Apple have their own custom SSD controller, so there is a small chance that you are right and the Apple SSD controller has a bug in the SSD statistics.

In any case, there is a bug introduced by Apple in their own software. The bug could be indeed in the SSD statistics, but that is very unlikely, because this is just an extremely simple function. It is far more likely that there is a more subtle bug in some other part of the Apple operating system, where they might write more data than necessary, maybe because there was not enough experience with real usage of the computer, to see how large the quantity of written data will become.


Even if the large quantity of written data would be just apparent and not real, it would not matter much, because the operating system will decide at some point to make the SSD read-only, waiting for a replacement that can be done only in a service center, by soldering.

Even if the SSD would not be really worn out, the operating system cannot know that, so it will decide that the SSD cannot be used any more, based on the false statistics.

But, like I said it is hard to believe that such a bug could occur in the easiest and simplest code, instead of being in some complex code with hard-to-predict effects.

Such a simple bug would have been already corrected until now, after the first reports.

All I can say is that the few references I have seen to DriveDx' reporting of SSD usage values seem pretty much inline with general usage across Intel Macs.

https://markellisreviews.com/is-there-a-problem-with-the-ssds-on-m1-macs/

I see at least two possibilities:
- hysterical response to extremely unusual usage models (or perhaps defective hardware -- when you ship millions of units, one or two defects are common)
- the reporting SW is (for whatever reason; I'm not going to speculate on the chain of possible error locations between a Linux app, through the Linux OS, through a virtualized driver, the hypervisor, MacOS, the Apple driver, and the Apple HW/FW.

What I do find tragic is how exactly the same people who are cautious to advise "don't extrapolate from a single example; anecdote!=data; etc" are so willing to leap to an extreme (and unlikely) conclusion just as soon as it conveniently matches their prejudices. In the battle of "training as a rational skeptic" vs "tribal loyalty", tribal loyalty will win 95% of the time.

[lyamc]

I think the moral of the story is: buy a board with soldered memory/storage, get ready to get screwed