Linus Torvalds is furious at Intel over its handling of the current processor security flaw

[Bit_Guardian]

9 hours ago, Drak3 said:

The tech news is that irrelevant, petty man, who created the base kernal that would be adapted and added to by multiple groups, is mad that a company has a security issue and he can't comprehend that holes can go undiscovered for years.

Which is hilarious considering it was Linus' own hand-optimised assembly code that led to Dirty-COW in Linux, discovered after 6 years...

[Arika]

if there is someone in the tech sphere that i couldn't care less about, it's Linus Torvalds. an utter twat

[Polakki]

Linus Torvalds F...O...

[Sauron]

11 hours ago, mr moose said:

Regardless of what Torvalds might be trying to get at, it's still the same thing.  To be honest I'm not sure why people don't understand the necessity of PR in situations like this. 

It's not good PR if they lie and people instantly see through it... besides it's not like they're sharing information about an alien attack, they don't need to worry about panic in the streets. The proper response would be aknowledging the problem and promising to fix it as soon as possible.

[Sauron]

3 hours ago, Bit_Guardian said:

Which is hilarious considering it was Linus' own hand-optimised assembly code that led to Dirty-COW in Linux, discovered after 6 years...

He's not complaining because the bug exists, he's complaining because the way the cpus were designed makes this hard to fix and not addressable through a microcode update, meaning it is a consequence of poor engineering (at least in his opinion). The DC exploit was patched in a matter of hours with no ripercussions, and it had been patched years earlier as well before an unrelated change reopened it. On top of that, Intel is dragging its feet. He's understandably frustrated.

[RefresherMan]

17 hours ago, potoooooooo said:

I mean he's not wrong, but why is he relevant? When's the last time linus torvalds did something actually newsworthy?

Now lol

[Bit_Guardian]

8 minutes ago, Sauron said:

It's not good PR if they lie and people instantly see through it... besides it's not like they're sharing information about an alien attack, they don't need to worry about panic in the streets. The proper response would be aknowledging the problem and promising to fix it as soon as possible.

Which they did! Hell Intel's the chief one contributing to the Linux kernel to patch it.

 

4 minutes ago, Sauron said:

He's not complaining because the bug exists, he's complaining because the way the cpus were designed makes this hard to fix and not addressable through a microcode update, meaning it is a consequence of poor engineering (at least in his opinion). The DC exploit was patched in a matter of hours with no ripercussions, and it had been patched years earlier as well before an unrelated change reopened it. On top of that, Intel is dragging its feet. He's understandably frustrated.

"poor" when he hasn't the first clue how to design high performance hardware.

 

How is Intel dragging its feet?!

[Gdog]

How is Linus irrelevant? He is "the final arbiter of all changes accepted into the Linux kernel. "

 

You know... the kernel that powers one of the most widely used operating systems in the world? The guy is pretty relevant if you ask me!

[Bit_Guardian]

4 minutes ago, Gdog said:

How is Linus irrelevant? He is "the final arbiter of all changes accepted into the Linux kernel. "

 

You know... the kernel that powers one of the most widely used operating systems in the world? The guy is pretty relevant if you ask me!

Except he isn't the final arbiter and hasn't been for years. It's a committee now, made up of many engineers from the big cloud providers.

[Sauron]

22 minutes ago, Bit_Guardian said:

Which they did! Hell Intel's the chief one contributing to the Linux kernel to patch it.

 

"poor" when he hasn't the first clue how to design high performance hardware.

 

How is Intel dragging its feet?!

Patching kernels is just a band-aid, this needs to be fixed at the hardware level starting from their next generation. However I get the feeling that this would mean delaying the launch and they don't want to do that.

 

And I'm fairly sure Torvalds knows how intel cpus work pretty well. One doesn't need to know how to design one from scratch to know how certain flaws could have been avoided.

[Bit_Guardian]

3 minutes ago, Sauron said:

Patching kernels is just a band-aid, this needs to be fixed at the hardware level starting from their next generation. However I get the feeling that this would mean delaying the launch and they don't want to do that.

 

And I'm fairly sure Torvalds knows how intel cpus work pretty well. One doesn't need to know how to design one from scratch to know how certain flaws could have been avoided.

If you think Intel's engineers haven't been on top of that since June, you're an idiot.

 

And no, conceptually this is pretty easy. The branch predictor is less than 1% of the die.

 

He understands nothing more than branch predictor accuracy for n-depth branches, cache latencies, the pipeline width, and instruction clock counts. The man is not some god when it comes to programming for bare metal. Any number of good HPC programmers out there are just as good as Linus. In fact much of the Linux Foundation wants him gone specifically because he's more trouble than he's worth these days.

[samcool55]

15 hours ago, Sauron said:

Yes and no, they could have been honest about it and said they are doing all they can to solve the issue... but are they?

Intel just got hardcore tackled by AMD on the desktop and has almost completely lost the mobile market. They are doing just fine right now, but if the big software developers start turning their back on them they'll start feeling it eventually. Linux may not be that popular for the average user, but it dominates datacenters. If those datacenters decide to go with a competitor (don't forget IBM is still alive by the way, and they don't use x86) it will be a serious hit to Intel's revenue.

Assuming datacenters won't switch quickly to IBM stuff or something else but sticking with x86, it does look this bug will give AMD a nice push in the datacenter market.

For now they don't seem to be vulnerable for the Meltdown bug, so no performance difference (in some datacenter workloads the fix for Meltdown causes a 10% drop in performance on Intel hardware) and their products and prices are competitive so they have a nice alternative.

 

This will very likely be another punch against Intel's monopoly so I hope we'll see a more competitive future because of this. I don't want Intel to die, just to be clear but it's really about time they get some slaps in the in the commercial/datacenter market. It's likely the competition from there will eventually end up being something positive for consumers as well.

[mynameisjuan]

3 minutes ago, Bit_Guardian said:

He understands nothing more than branch predictor accuracy for n-depth branches, cache latencies, the pipeline width, and instruction clock counts. The man is not some god when it comes to programming for bare metal. Any number of good HPC programmers out there are just as good as Linus. In fact much of the Linux Foundation wants him gone specifically because he's more trouble than he's worth these days.

Yeah I am not sure why people think he is this all knowing god. Just because he wrote the original kernals and now has some say in the recent ones doesnt mean shit. He knows his kernals, Sauron saying that he knows intel's CPUs inside out is just a joke. Outside of being a developer, linus for the most part is actually kind of a dumbass. Like you expect a guy that can develop a kernal can reinstall a program on linus WITHOUT having to reinstall the OS?  

[Sauron]

7 minutes ago, Bit_Guardian said:

If you think Intel's engineers haven't been on top of that since June, you're an idiot.

Nowhere have I said that Intel's engineers haven't been working on it. Please read again before you start calling people names.

 

What I said is that they are probably unwilling to delay their next product line to fix this and will rely on software workarounds.

10 minutes ago, Bit_Guardian said:

And no, conceptually this is pretty easy. The branch predictor is less than 1% of the die.

Cutting or altering part of a cpu like that without causing other problems is never easy. Judging from this statement:

Quote

Because I really see exactly two possibibilities: - Intel never intends to fix anything OR - these workarounds should have a way to disable them. Which of the two is it?

and his comment on what should be done to fix this ("why is this done with no configuration options?") I get the idea that there would need to be some fairly radical changes to the design phylosophy.

 

And of course this is partly speculation and partly exaggerating on his part, since we don't know for sure what Intel is going to end up doing, but so far they have shown little intention to admit their mistakes and do something other than having kernels gimp their cpus.

16 minutes ago, Bit_Guardian said:

He understands nothing more than branch predictor accuracy for n-depth branches, cache latencies, the pipeline width, and instruction clock counts. The man is not some god when it comes to programming for bare metal. Any number of good HPC programmers out there are just as good as Linus. In fact much of the Linux Foundation wants him gone specifically because he's more trouble than he's worth these days.

I never said he somehow knows more than everyone else doing the same things, he's just one of the people working on it closely and one who's been particularly vocal about it. As far as part the Linux foundation wanting him gone goes, that has probably more to do with the fact that he gets the last word on what ends up in the kernel than being in any way incompetent or "trouble". He can be pretty rude, so it's no wonder some people are fed up with being called idiots over a piece of code that ends up getting scrapped.

[Bit_Guardian]

2 minutes ago, Sauron said:

Nowhere have I said that Intel's engineers haven't been working on it. Please read again before you start calling people names.

 

What I said is that they are probably unwilling to delay their next product line to fix this and will rely on software workarounds.

Cutting or altering part of a cpu like that without causing other problems is never easy. Judging from this statement:

and his comment on what should be done to fix this ("why is this done with no configuration options?") I get the idea that there would need to be some fairly radical changes to the design phylosophy.

 

And of course this is partly speculation and partly exaggerating on his part, since we don't know for sure what Intel is going to end up doing, but so far they have shown little intention to admit their mistakes and do something other than having kernels gimp their cpus.

I never said he somehow knows more than everyone else doing the same things, he's just one of the people working on it closely and one who's been particularly vocal about it. As far as part the Linux foundation wanting him gone goes, that has probably more to do with the fact that he gets the last word on what ends up in the kernel than being in any way incompetent or "trouble". He can be pretty rude, so it's no wonder some people are fed up with being called idiots over a piece of code that ends up getting scrapped.

And like I said, that's less than 1% of a core. It's actually a fairly simple fix conceptually, and the logic can be written on a napkin, so it's maybe 100,000 transistors or so.

 

Certain components are more tightly coupled and difficult to maintain than others. The branch predictor is pretty independent. It'll be simpler than most other things.

 

Because they shouldn't be done with config options. He's as short-sighted as Stallman has been saying he is for years. And no, it's not a radical difference in design philosophy. It's one more small but crucial thing to look after, as in any large, complicated system. The key to good architecting, in anything, is to not stress about it. You know you can do it. The only variable is time to completion.

 

No.  Stallman, Heinrich, Stern, and Banes all want him booted because he no longer has vision and has ruined a lot of their relationships with hardware providers, such as ARM whom he used as a jab after he told the Linux community to abandon ARM the moment the Chinese took it over.

[Bit_Guardian]

32 minutes ago, samcool55 said:

Assuming datacenters won't switch quickly to IBM stuff or something else but sticking with x86, it does look this bug will give AMD a nice push in the datacenter market.

For now they don't seem to be vulnerable for the Meltdown bug, so no performance difference (in some datacenter workloads the fix for Meltdown causes a 10% drop in performance on Intel hardware) and their products and prices are competitive so they have a nice alternative.

 

This will very likely be another punch against Intel's monopoly so I hope we'll see a more competitive future because of this. I don't want Intel to die, just to be clear but it's really about time they get some slaps in the in the commercial/datacenter market. It's likely the competition from there will eventually end up being something positive for consumers as well.

The push was already coming from the likes of Qualcomm and Cavium. I think the rumours about Intel completely redoing "x86" as we know it just might be true. Sure, Xeon D and Avoton have kept ARM out of cloud computing so far, but it's also true you don't need all the power and heat of a Skylake-X core to run a basic java spring workflow webapp for a business that doesn't even run any analytics or batch jobs. You could run it off of a 1GHz Atom core or off of an A53 and you seriously wouldn't be able tell the difference vs. a 2.5GHz Skylake core up probably close to 100 simultaneous users (an app I'm putting the finishing touches on has no discernible performance difference between running on a T2 nano and an M3 medium up to 100 simultaneous requests, a scenario that will never come up in its usage anyway).

 

Exciting times are ahead for sure, but there's no way Intel's giving up the crown.

[Stefan Payne]

18 hours ago, Sauron said:

Well, can't say I disagree with Torvalds. It seems Intel is doing everything it can to downplay the issue.

Not only Intel downplays the issue, the same goes for the Intel Fans who still have no clue what this shit is all about and what's really happening. 

They don't see it as a Fukushima or Tschernobyl grade issue.

 

 

And some of them even think that everything is over - well its not!!!

 

Spectre 2 is still not fixed. It wasn't an issue with AMD, because it seems that AMD did think about security at least a bit and came to the conclusion that what Intel is doing isn't a particularly great idea.

 

I mean speculation OK, also when it was introduced in 1995 it wasn't really an issue with mostly running 16/32bit Windows like 3.1 and 9x...

And with RISC processors dominating the Server market.

 

It is an issue however with Windows Vista and later, when the user really worked as a user even with an admin account. And then it should have been fixed!

With the integration of the memory controller, that was the right time for it and also a big change. But they didn't...

Quite the opposite, they did work on this now almost 10 year old design and modified it further and further.

In the meantime AMD had THREE completely different Designs - the K10 back in the day, the failed Bulldozer and now Zen. All three have not much to do with each other...
Well, Bulldozer sadly had the K10 Northbridge and I/O Interface, wich crippled the performance further...

 

Anyway Intel was maximizing the profit with minimal innovation in the last 10 years - and now it comes back to bite them...

 

 

 

[Gdog]

1 hour ago, Bit_Guardian said:

Except he isn't the final arbiter and hasn't been for years. It's a committee now, made up of many engineers from the big cloud providers.

Yeah, I thought he still had the final say on important changes though, I could be wrong?

[Stefan Payne]

1 hour ago, Bit_Guardian said:

Which they did! Hell Intel's the chief one contributing to the Linux kernel to patch it.

This shit shouldn't have happened in the first place!
The issue here is kinda similar to the VW Exaust scandal. And should be handeled in a similar way.

 

In this case Intel sacrificed safety/security for Performance. And that is why they are better than 'the others', because they cheated, with a technic that should have never been used because its so unsafe...

 

 

Oh and another thing:
Intel knew since June or July!!

 

So what happened at Intel between June/July and now??

 

They introduced two(!!!) CPU Lines!

Skylake-X and Coffeelake while knowing about this issue!

 

 

Don't you think that this needs to be called out?!

[Drak3]

5 hours ago, Urishima said:

The man maintains the kernel of the OS that basically 'runs' the Internet. Point at any datacenter and I will almost guarantee you that it basically runs on Linux.

That ultimately means very little. He just okays changes to code, and only to the latest versions. Everything is is delegated to other people.

 

5 hours ago, Urishima said:

Yes, the guy is a big deal.

No, he's not. He's an irrelevant, petty asshole, who suffers illusions of grandeur.

[Drak3]

1 hour ago, Gdog said:

How is Linus irrelevant? He is "the final arbiter of all changes accepted into the Linux kernel. "

 

You know... the kernel that powers one of the most widely used operating systems in the world? The guy is pretty relevant if you ask me!

Torvalds is a jackass that does nothing more than say yay or nay on what gets into the official versions of the latest kernal. Anything even slightly older, and the responcibility is delegated.

 

Not that that means much, distro maintainers/writers can easily use forks of the official distro and circumvent Torvalds entirely.

[Stefan Payne]

58 minutes ago, Drak3 said:

No, he's not. He's an irrelevant, petty asshole, who suffers illusions of grandeur.

You have never heard him Talk, have you?!
 

Here the full lenght nVidia Video:

 

He doesn't seem to be an ass. Quite the contary.
He is just very direct but seems rather nice.

[JohnMiller92]

10 minutes ago, Stefan Payne said:

You have never heard him Talk, have you?!
 

Here the full lenght nVidia Video:

 

He doesn't seem to be an ass. Quite the contary.
He is just very direct but seems rather nice.

"I use strong language on the internet, to the point where some people feel offended. And that's their problem"

 

"He doesn't seem to be an ass"

 

Pick one.

[Stefan Payne]

3 minutes ago, JohnMiller92 said:

"I use strong language on the internet, to the point where some people feel offended. And that's their problem"

 

"He doesn't seem to be an ass"

 

Pick one.

Yes, and?!

I use strong language sometimes too because if you are friendly to the other one you get ignored.

So I can totally understand where he's coming from and why he does it.

 

 

So to make it clear:
If you are nice and friendly either nobody takes you serious or you get ignored.

If you use strong language, its more likely that people will react to you.

[straight_stewie]

20 hours ago, D13H4RD2L1V3 said:

Linus Torvalds, creator and principal developer of the Linux kernel, is reportedly unhappy

When is linus torvalds ever happy though?