Facebook Joins Linaro Linux-on-ARM Effort

dgharmon writes "It has been more than two years since Freescale Semiconductor, IBM, Samsung, ST-Ericsson, and Texas Instruments formed a non-profit software company called Linaro to help focus the disparate efforts to get Linux running well on ARM processors and system-on-chip designs. A slew of companies, some new to the ARM racket, have joined the Linaro effort – and as of Thursday afternoon, so has social media juggernaut Facebook."

Comments

Makes sense.

[cultiv8]

http://mobile.slashdot.org/story/12/05/28/1632226/is-facebook-working-on-a-smartphone

Re:Makes sense.

[hendridm]

FTFA: "Facebook and AMD are joining the Linaro Enterprise Group, which was formed to focus on "the development of foundational software for ARM server Linux," as the announcement put it."

Re:Makes sense.

[rtfa-troll]

There are about three linux based distros for the openmoko, all with custom UI front ends.

Most "distros" that you see are exactly that; ways of experimenting with different UI, usability and system administration concepts. The guys that do them don't want to do much of the low level plumbing. Think of Linaro as being similar to Mer which is the underlying developer whilst Plasma Active and Nemo the consumer distros built on top of it.

In the case of Linaro, though, they are trying to support completely embedded customers. These users often don't end using a proper distro at all. For this reason Linaro is less building a distro that people build on top and more building examples of working systems people can copy from. Let's say you are an SOC or sensor vendor that wants to sell to Android vendors in future. Your first place to start would be cooperating (paying) Linaro to get it working. Once that's done, everybody can can see how it works and is confident that they can also get it working in their own build of Android. At the same time you get access to the massive embedded Linux market.

This is one of the ways in which modern community oriented FOSS software works a bit differently from proprietary software. You see all of these experiments and development versions out in the open where each of them would be hidden somewhere deep inside Microsoft's R&D. For the hardware company this is a great way to address all the different Linux distros via one place. For the Linaro guys this is great because they continue to get the latest hardware to develop on all the time and they don't need to stop working and integrate to other distros. For the developer this is a great opportunity to understand which hardware will be best supported in future. For the normal user the message is "don't worry be happy".

Re:Um...

[godrik]

Of course, that's not what Linaro is about. They are looking forward to stop the explosion of code and architecture within the ARM familly. No two ARM machine boots the same. No two ARM processors expose component the same way. You did not read Linus saying "what about stoping the ARM crap?"

http://linux.slashdot.org/story/11/08/18/1728227/arm-is-a-promising-platform-but-needs-to-learn-from-the-pc

" to get Linux running well on ARM processors"

[Gaygirlie]

I guess it depends on what you mean with "running well": I have a Pandaboard and, well, it has been a major clusterfuck all the way from the beginning, what with constant breakage of features, on some releases of the software the features disappear completely, and then there's the constant crashes in the kernel. Imagine my surprise when I installed the latest stable Texas Instruments - release of the kernel only to find that networking is completely broken and the kernel goes to a hard lock-up after being on for 5-10 minutes, whether or not it just sits idle this whole time.

If all I want out of it is unaccelerated X or just console applications then yes, it runs very well, and it works great as a low-foothold server for all kinds of things. I'm just saying that I sure have no high expectations for these guys and their efforts.

Re:" to get Linux running well on ARM processors"

[rdnetto]

The very reason it's a clusterfuck is because of the fragmentation these guys are trying to address. Each device has a different kernel, even those that use the same SoC (because the GPIOs, etc. are hardcoded). That means that the developers efforts are fragmented - only a small number of people see the bugs and put in the effort to fix them, which undermines Linus' law. 3.7 will help, but there's still a lot to be done before ARM has the kind of compatibility that x86 does.

tldr: These guys are doing useful, important work that will vastly improve the state of Linux ARM.

Re:Stealing Credit

[Elbereth]

According to the Wikipedia article, these guys are simply trying to simplify, optimize, and reduce fragmentation in the ARM/Linux world. They're not trying to claim anything except that their tools and validation suite make your life easier.

It's sort of like the Linux Standard Base, if you remember that initiative. The LSB was invented to address concerns of fragmentation and difficulty in porting applications to Linux, because the distributions were so radically different from each other. While it didn't work out as well as hoped, it did manage to reduce the idiosyncrasies.

Re:Android?

[Gaygirlie]

Isn't Android itself a Linux-derived OS? I thought Torvalds said that Android will be converging back to Linux in the future.

Yes, Android runs on a slightly modified Linux-kernel. Most of the modifications have now made their way into the mainline kernel, so in the future it may well be possible that there is no need for modifying the kernel at all. The userland on Android and any Linux-distro is entirely different, but I s'spose you knew that already.

Re:Stealing Credit

[Anonymous Coward]

For that matter Linaro has been making ARM patchsets for various hardware for over 10 years (maybe even longer!). Probably 3/4 of the ARM Linux hardware you'll find (be it routers, IPCams, print servers, etc) will be running some kernel with a -linaro in it, and even the ones that aren't usually have a -linaro in the gcc/binutils buildversion.

They're just trying to reduce their own costs by consolidating the majority of necessary changes into patchsets that will be accepted into the kernel, so that minimal additional, potentially conflicting, code will end up outside of the kernel in the long run.

Can't fault them for wanting to make both our and their lives easier.

ARM servers

[symbolset]

It's long past time for this. China is doing MIPS servers too.

Not About Mobile

[fm6]

Even if Facebook is planning a smartphone (huh?) they don't have a lot of motivation to improve core Linux on ARM. I think it more likely that they're seriously looking at transitioning their (huge) data centers to ARM in order to save energy costs.

I've poo-pooed ARM-server speculation in the past, but this goes way beyond speculation.

Re:Downsides to running ARM servers?

[imsabbel]

Server chips make less than 12% of AMDs revenue, tieing with chipsets, of all thing. Graphic chips are more than twice that.

Re:While you are at it

[rubycodez]

no, it has no MMU

Re:Stealing Credit

[rdnetto]

For that matter Linaro has been making ARM patchsets for various hardware for over 10 years (maybe even longer!).

Citation needed. According to wikipedia, Linaro was only founded 2 years ago.

Re:Downsides to running ARM servers?

[drinkypoo]

It remains to be seen if ARM will ever offer the same single-thread performance as x86, but it offers better performance per watt now, so embarrassingly parallelizable tasks are already better done on ARM so long as the servers don't cost so much that you don't save any money using less power. Power is pretty expensive these days though, which is one big reason you're seeing so much interest in ARM servers, and intel isn't exactly known for offering their chips at unbelievably low prices.