The SoC industry is a cesspool of secretive, proprietary non-standards, that makes Microsoft and Apple look like the Free Software Foundation by comparison.
Help us out here amateur psychologists: Linus Torvalds has just unleashed his second shouty rant in as many days. Do we need to worry, or is the moon in a particular phase that makes this kind of thing more likely? Has Portland's water supply taken a turn for the worse? Or are we simply seeing a frustrated middle aged man …
After trying to debug why my arm Exynos notebook wakes up immediately after going to sleep last week I am going wholeheartedly agree with Linus on this one. Every SoC is different, in order to do something as basic as going to sleep the kernel has to write a block of magic numbers into registers the size of average firmware. Basic functions like mmc, usb, ethernet are one-offs imlemented slightly differently by each SoC designer. One architecture my a***
The worst bit - it will only get worse from there on. Arm is there. MIPS has been there for ages (we all know how entertaining development of openwrt and friends gets because of that). Power recently joined the club and Intel joined yesterday by announcing Quark. In a couple of years time everything will be a f*** SoC and every second one of them will have at least one feature added by a person which fits Linuses description.
"While his technical opinion might be valid, his comment was tasteless and out of order, so wiping out any moral high-ground he might have been trying for."
Well, he wasn't trying for any moral high ground at all, so that's okay. I'm not sure why anyone would think that technical failures can be defended on moral grounds in the first place, though.
he makes the point and gets it noticed by press due to the 'childish' nature of the comments.
if he'd politely complained nobody would have really noticed, cared, or reposted it, and the whole issue would be easier to sweep back under a rug.
sometimes measures like this are needed to make a point, even if it offends people. sometimes people need offending.
Not convinced he is completely right.
If you want discoverable buses, that all adds transistors, taking more power, and probably starts trampling over some other company's intellectual property, etc.
For ARM SOCs, power is king, and no chip builder is ever going to put in a whole bunch of additional transistors that then makes their chip look bad from a power consumption point of view. Nor can they afford to pinch someone else's intellectual property; someone out there is bound to be holding patents on discoverable buses like PCI.
Linus wants these things so that Linux doesn't need to be manually configured and built for each individual SOC design from every individual manufacturer. But if that's what we're going to be stuck with, how about making it possible for the manufacturers to easily contribute a single 'config.sys' file (for want of a better phrase) for their SOC that is then automatically available to everyone downloading the Linux source code? That would at least mean that the work gets done only once.
I don't know enough about the Linux ARM source code base to know if that makes sense, but something along the lines ought to be possible. For all I know it may even already be there but the manufacturers aren't playing ball, which would be a pity.
Yes, but if they are undescoverable then they must be clearly documented. If there was a proper manual, API, feature documentation etc then it wouldn't be such a huge issue. But most of it is an undocumented nightmare - in this day and age I believe this sort of practice deserves to be *very* heavily vilified.
Suggesting killing them is a bit OTT, but heavy vilification and disapproval are required.
How can you write for shit like this if you have to almost reverse engineer how each one works?
Um, actually I think that publicly advocating murder is more along the lines of totally illegal. As in get-arrested-and-risk-jail illegal.
On the books anyway, and IANAL, of course, but if he got a "courtesy" visit from the local Sheriff accompanied by a night in the cooler, he could only have himself to blame.
Not that I mind a good shouty rant from this guy. He has a way of cutting through civilized behavior like a surgeon that I like reading about.
But actually suggesting that people go out and create deadly accident conditions ? Sorry, Torvalds, you've largely overstepped your (considerable) notoriety this time. Stick to calling people names and comparing them to cockroaches or something.
"Um, actually I think that publicly advocating murder is more along the lines of totally illegal."
Not in the United States it's not. I don't know about the UK, but here, courts have generally recognized the existence of satire, exaggeration for effect, and other extraordinarily common writing techniques which have apparently evaded your understanding.
Were Torvalds referring the president of the US rather than unspecified SoC designers, he might find himself visited by some guys in dark glasses. A few years back one of my buddies managed that by dint of an ill-advised op-ed piece in a college newspaper - but the suits, far from throwing him in prison, talked to him for a few minutes and determined the obvious, They warned him not to put them in the position of wasting their time on due diligence again, and that was that.
But for obviously comic 'threats' against an unspecified group of people? Please. Anyone who advocates jail time for something like that should be hauled out and shot at dawn.
(Come and get me!)
> Not in the United States it's not. I don't know about the UK,
To learn about the legality of this in the UK, google for "Robin Hood Airport Tweet".
I believe Torvalds is coming to the UK next month.
With luck, someone at ARM will have spoken to the police and he will be arrested on arrival.
They are documented - often in great detail. But the documentation is frequently under NDA, and most of the code is contributed by the SoC vendors.
The Linux Kernel mainline is very, very far from being the mainline for any ARM SoCs anyway, all Linus will achieve is more forking.
I'm _guessing_ that perhaps Linus's annoyance with undiscoverable cpu features isn't that they're not documented but that he can't use a small number of methods or functions to identify them and instead needs unique code for each one. I'll swiftly add that this isn't due to lazyness but because of the increased amount of code that needs to be maintained, and with more code you have a greater likelyhood of errors.
In coding terms it's a bit like having a line of code for every data record you need to process instead of processing all data records in a single loop.
However, this is all nothing to do with what the article was about, which was, essentially, is he getting too stressed out?
Well, there's no doubt that he's in a _very_ stressful position: the Linux kernel runs on a wider range of h/w than any other other bit of s/w that's ever existed and trying to coordinate, collate and implement the correspondingly wide amount and variety of code modules submitted by an equally wide variety of code submitters, from individuals to global corporations, can be nothing but stressful.
So is he getting too stressed out? No, what he said was was not to be taken literally and was clearly a tongue-in-cheek gesture to express his annoyance, the severity of the gesture indicating his degree of annoyance. This is something that every human being does from time to time: someone does something a bit stupid or thoughtless and the person who is annoyed by it replies with a deliberately irrational, out of proportion and over the top response. It's no different to that young chap who threatened to blow up an airport because of repeated delays due to the volcanic ash cloud or Jeremy Clarkson suggesting that certain people who had annoyed him should be shot in front of their families.
Personally, I'm more concerned with the mental wellbeing of people who appear to be either incapable of discerning the real meaning of these gestures or, even more worryingly, use them as the basis for an attack on someone. Why even more worringly? Because if you feel entitled to launch an attack on someone whom you don't know because of something they said to someone else whom you don't know then you must be _really_ fscked up.
I don't understand how this is really any different to the publicly undocumented stuff on all 3D graphics chips.
The documentation is usually available only under NDA because a) the registers are there for hardware hacks and you can potentially do an inadvertent HCF by writing the wrong thing or even just doing so in the wrong order and b) documenting it would reveal stuff about how it works which is useful to the competition.
With the pace of SoC development, registers for fudge factors and such will always be present in these devices as there just aren't going to be enough respins of a part to iron them out. Therefore SoC's in general don't make for great general compute platforms. (Says an OpenWRT hacker. :-)
Not enough extra transistors to make any difference, compared to the 100 million plus transistors per SoC. It could be done with a standardised ROM containing a list of device IDs (64 bits, to avoid ever running out of IDs) and a base address (64 bits). Some address decode logic, and 128 bits (128 transistors?) per device.
Any patent on a ROM containing a look-up table surely expired in the 1970s.
No reason a ROM consumes any power at all, except when it's being read. But even ignoring that, a thousand transistors compared to many million is under 0.1%. That's way below manufacturing variability.
1. Never underestimate the vagueness of Patents, especially in the US.
2. He is behaving like a spoilt toddler, throwing his teddy out of the pram when he doesn't get what he wants.
These things are possible, but in the low value/high volume world of ARM changes like this are a much more significant cost than can be estimated, especially given the current volume of ARM chips out there and being made.
Additionally, IF ARM or another supplier chose to add this functionality, while Linux still supports processors without these features, he and others will still have to do the work.
So it's a rant to nothing.
When the mkIII and the mkIIIa and the mk111a2 all vary by substantial amounts, it's a problem.
Linux has a point and being shouty about it means his point gets circulated.
My sympathies to the BSD guys btw. There's a lot of things which can (and should) be done better in the linux kernel - and with sufficient feedback might well be. Then again there are the odd bits I can't stand in the BSD kernel too. :)
No only is he technically right, in my opinion he's right to use strong language. In a world where any regular person's comments are drowned out by well funded 'PR' blogging. He doesn't have the funds of many of the Linux contributors so it's cute trick to make sure his opinion is heard.
Quite. I invariably find his unabashed hyperbole and hammed up indignation both amusing and refreshingly honest. Shirley I'm not the only one to realise he's NOT really trying to incite murder? So why these waves of Daily Mail damnation? Are we striving to promote a more usual workplace culture of oleaginous insincere correctness, cliques, plotting and treachery? Are we jealous of an environment of open expression and honesty in which everyone knows exactly where they stand, what's expected of them and potential problems are swiftly and transparently excised?
Whenever there's a story about some poor pleb getting into hot water over some quip about blowing up an airport or destroying America or whatever the Commenterderate Soviet is always quick to defend. So why is Torvalds any different? Personal jealousy? Someone that famous, successful and influential "needs to be taken down a peg or two"? A spot of group iconoclasty?
Please tell. I'm genuinely puzzled and a little dismayed by this phenomenon.
The movies portray nerds that way because it's often accurate. How many IT administrators become jumped up little twerps because of the power they have? Linus is hardly unique, except that most nerds don't get into the public view until they've had some of the rough edges knocked off through having to work with other people in order to become important.
And I can't believe the Reg is suggesting we all hew to the standards of Human Resources.
Tell me, has any company with a HR department ever been successful?
I thought the natural development line for companies was
1) Get set up by brilliant ideas people
2) Grow to the point where they acquire and HR department
3) Go down the tubes rapidly.
Isn't this how things normally work?
Biting the hand that feeds IT © 1998–2019