That Open, Upgradeable ARM Dev Board Is Trying To Make A Comeback

[robclark]

i evaluated these mmm.... 6-8 months ago because yes i was really excited and hopeful, but ultimately discovered that they've provided the most absolute cursory of minimal information - absolutely none of it useful to create your own PCB.

let me re-check ok?

here's the documents:

Qualcomm® Snapdragon™ 410 Processor documents:

• Snapdragon 410 Device Specification
• Snapdragon 410 GPIO Pin Assignment
• Hardware Register Description for Qualcomm Snapdragon 410 (APQ8016)

Qualcomm Snapdragon is a product of Qualcomm Technologies, Inc.





nice that they tell us that it's a product, gets some good keywords in there, right?

lp80-0p436-7_b_apq8016 - just gone over it, it's pretty basic, it would *just* about be enough to use ...

GPIO pin assignments i'm getting a "redirect error" on qualcomm's web site...

there's also a section with schematics: schematics look good... no PCB file though: that means about 3-6 months of work instead of 3-6 weeks...

ok we *need* the datasheet for the PM8916 though. otherwise it'd be necessary to take a risk using more well-known PMICs (easier to get hold of), which means kernel development as well because the linux kernel sources are intimately tied to the board bring-up process... huh, whaddyaknow, a google search "PM8916 datasheet" actually comes up with the goods! i have a feeling i've been over this before... yes i have, now i think about it.

tbh, last time I looked, the 96boards site just sent you to a qcom page which had more docs, including pmic. I'm not sure why they switched to just linking a subset of the docs directly from the 96boards site.

so that *might* do the trick. we're looking at 3-6 months of full-time work and maybe a couple of $2k PCB production runs to get things right.

next thing: are samples available? can we get through to qualcomm without requiring an NDA because everything is reaasonably available without one? that's the real question. i mean you _could_ go buying dragonboards and try to use a desoldering station but... don't hold your breath we could get onto taobao and see if there's some "fell off the back of a lorry" PMICs and 410 SoCs available...

I believe you can from arrow..

disclaimer: I don't speak for qcom/arrow/96boards/etc... (and really more of a sw/driver guy, never really done any board design).. but I think qcom's move into the "catalogue" market is relatively young, and rather than building up a large support org like TI/freescale, they seem to prefer go through a few 3rd parties (arrow, einfochips, inforce, etc). So you might do better to talk to one of them, than qcom directly.

but before committing $2.5k per month's worth of engineering time @ 3-6 months plus $2k per PCB production run i'd *REALLY* like to know that the SoCs and PMICs are gonna be available under contract at the end of that, y'know what i'm saying? look up what happened when hardkernel created a raspberry pi clone - how did that work out for them, ehn?

I believe it is planned that the handful of SoC's available for industrial are supported for 5-10yr lifespan. Ie. this is why it is just a small subset of all the different parts they make for mobile. It also seems to correlate with which parts there has been upstream kernel focus.

so it's hard... doable, not inconceivable from the information that's available... but it's definitely harder than it needs to be. if you know of someone who's got contacts inside qualcomm, or can think of a route in that would make things easier, that would be great and it could be added to the list.



ahhh not quite. the reason why the devicetree patches are there is because of the open nature of the db600c. linaro will be helping out there (as long as qualcomm continue to pay the $USD 1 million a year membership fees).

so we will see more of qualcomm's SoCs going forward... *if* qualcomm gets with the programme of cooperating with the largest software project on the planet (the linux kernel), and releases open boards.

the main linux kernel developers really haven't got time to waste on boards (or phones) that are entirely proprietary, which is why you haven't seen any mainline linux qualcomm SoCs added up until the db600c.

thanks for encouraging me to go over this, robclark.

[starshipeleven]

given that people keep complaining about the lack of free MALI drivers and then DON'T PAY HIM ANY MONEY, we're left wondering "what gives?".

I've yet to see any crowdfunding campaign for that.
Sorry for libv, but whining on IRC or forums isn't exactly the most effective strategy.

it's an odd situation, the software libre world.

No it isn't, it's just different from usual commercial licensing so most people don't know how to handle this (and I talk of developers).

The basic fact is that someone still has to pay the developers, and most libre devs are of the more altruistic kind that will gladly work for free for an ideal.
While this may sound good, it is ultimately stupid and leads to failure.

Libre devs must secure money first, THEN do the work.

I'm citing Krita as an example, that is at their third yearly highly successful kickstarter (38 thousand bucks or so for the last one) but there are others that launch crowdfundings and get funded.

Of course large projects have foundations and stuff, I'm talking of more small-scale teams than Debian or Libreoffice.

this btw is why i'm developing hardware, now, because hardware requires that people pay for it!

Kinda. Consider that this is a open hardware design, so any chinese manufacturer can just go and make clones if they so wish. It may not be an issue for now, but it's something that must be taken into account for later.

[pal666]

( i assume you mean 3D graphics? we're not targetting high-end games or power users remember).

all graphics is 3d nowadays

what i'm saying is that if you set certain parameters

what i'm trying to say is that if you set such parameters that you think you have to use components which make unusable laptop, don't make laptop. there other kinds of computers and some of them don't need graphics

[lkcl]

Awesome!

I work on a new, social, community driven OS that I'm hoping to unveil soon. If it ever takes off, I'd love to get into talks with you about proper hardware.
It would definitely need more "oomph" than the current reference design, though. My DE specifically requires hardware acceleration.

well... that's tricky to fulfil, but depending on how heavy the hardware acceleration is, you *might* be able to get away with an ingenic jz4775. don't hold your breath though! it's a 1.2ghz single-core 32-bit MIPS and it has an extremely fast SIMD vector pipeline that's good enough to use for 3D acceleration simply by compiling up mesagl (with some reaally bizarre macros! ingenic seeriously need some help here...)

think "altivec engine" which was in things like the PS3, but put something like that into an ultra-low-power (0.01mW per Mhz i think it is) RISC processor.

whilst that may sound strange it's *literally* the only SoC IN THE WORLD that can handle *ANY* level of 3D graphics acceleration whilst still respecting Software Freedom and ethical business practices.

... i'll let that sink in as a "wtf???" moment for a while...

in its favour the jz4775 is $3 (!)

we can't use the dual-core jz4780 because it uses PowerVR... eugh.

we can't use iMX6 despite it having etnaviv because it's a Cortex A9, uses too much power, and has a proprietary VPU. i've just learned recently that there's some proprietary VPU libraries out there which have MAJOR security vulnerabilities.... i can't give too much detail yet...

the list of "can't" just goes on and on... kinda confusing, so basically i'm reviving the "Towards and FSF-Endorseable Processor" initiative, so keep in touch ok?

[UPDATE: robclark has been helping review the qualcomm 410, it looks like that would qualify! and is available!]

[lkcl]

what i'm trying to say is that if you set such parameters that you think you have to use components which make unusable laptop, don't make laptop. there other kinds of computers and some of them don't need graphics

ah ha! right! thank you saying this, because this allows me to illustrate why i created EOMA68. what you've said "don't make laptop if makes an unusable laptop" would be absolutely true IF AND ONLY IN THE CONTEXT OF A SINGLE-BOARD MONOLITHIC DESIGN.

... yeah?

but i've not *made* a single-board monolithic design, have i? ehh?

so i've applied the "divide and conquer" approach, by making an *upgradeable* chassis that *in the future* you can get a better Computer Card... and more than that, the one that was in there when you first bought it can be RE-PURPOSED - it would make a fantastic router engine for example (doing encryption in a WIFI access point for example, where the normal router ICs are so low-speed they couldn't handle that kind of task).

the "hermetically-sealed single-board computer" mindset is so deeply entrenched, and people have so given up and expect that the mass-volume suppliers will always have total control over what we buy that it's really really hard to get over to people exactly what i've achieved here with the EOMA68 standard.

to the monolithic single-board manufacturers i have only one thing to say: i reject that reality [and substitute my own] https://www.youtube.com/watch?v=LlRVxzHUSNc

[lkcl]

I believe you can [buy] from arrow..

disclaimer: I don't speak for qcom/arrow/96boards/etc... (and really more of a sw/driver guy, never really done any board design).. but I think qcom's move into the "catalogue" market is relatively young, and rather than building up a large support org like TI/freescale, they seem to prefer go through a few 3rd parties (arrow, einfochips, inforce, etc). So you might do better to talk to one of them, than qcom directly.



I believe it is planned that the handful of SoC's available for industrial are supported for 5-10yr lifespan. Ie. this is why it is just a small subset of all the different parts they make for mobile. It also seems to correlate with which parts there has been upstream kernel focus.

iiinteresting. perhaps they're starting to get it, after all. now all they need to do is bring out SoCs that don't have gaping security vulnerabilities at the hardware level - https://slashdot.org/story/16/05/05/...roid-user-data

althouugggh... that might just be the case that you simply don't use that vulnerable hardware engine and it's fine...

darn it, robclark, you're convincing me to investigate the 410 properly mind you, i do have to say, it's still going to take around $20k and 4-6 months because they don't have PCB CAD Reference Designs available to work from. and i have to get through this campaign first, but yeah - a 5-10 year lifespan is worthwhile pursuing.

so, thank you! really grateful to you for finding that the SoCs and PMICs are available on arrow, that's a definite "tick" on hte checklist. and it's got a reverse-engineered 3D GPU, that's *really* good.

[starshipeleven]

ah ha! right! thank you saying this, because this allows me to illustrate why i created EOMA68. what you've said "don't make laptop if makes an unusable laptop" would be absolutely true IF AND ONLY IN THE CONTEXT OF A SINGLE-BOARD MONOLITHIC DESIGN.

The A20 is a tablet SoC so the "multipurpose" argument falls a bit flat.
It has all kinds of features that could make it a decent tablet or 2-in-1 laptop, but it lacks basic features to be anything else.

Like say multiple Sata controllers (for a raid maybe), USB 3.0 ports, PCIe, wireless connectivity, or a gigabit ethernet port.

[pal666]

now, about MALI: if you know what you're doing and are happy to do it, and you really *need* 3D graphics

i don't need closed graphics. what is the point of open hardware with closed drivers?

[pal666]

On other hand, Linux Sunxi community is really nice. I can't readily remember any other HW platform around with comparable community.

you don't remember any other platform with enough community to produce videodriver? your memory is not very strong

Most qualcomms are cell modems and everything else is really secondary.

hard choice between "it doesn't work at all" and "it works, but has cell modem". let me think...

So those insisting on qualcomm crap should stop bothering about all-open, RYF and somesuch anyway. If you're okay with huge proprietary firmware, it just would not do.

i'm even less okay with huge proprietary drivers. so those insisting on mali crap should stop bothering about all-open, RYF and somesuch anyway.

[lkcl]

hiya SystemCrasher, thanks for joining in here, really appreciated.

Freedreno implies Qualcomm and before expressing your valuable opinion it is really wise to check system architecture of praised SoC.

yehhh... the checklist is just enormous and extremely unlikely to be fulfilled, and it's down to the cost-savings of the "fully-integrated" bit that comes with "SoC" - System On A Chip...

If you think you could do it better than Allwinner based system, feel free to do that. But you can count I will show no mercy if I'll find any fucking blob or flaw in YOUR system. Speaking for myself I could boot Completely Deblobbed System on Allwinner. With absolutely no stinky blobs whatsoever. U-boot takes control early, no calls to resident evil firmwares are made since this point. Device runs under my full control. Using mainline kernel/uboot/etc. Sure, some options are still missing. But damn, it just usual Debian Testing and it rocks :P.

i hear ya the nice thing about the A20 is, they didn't bother to put in the usual Treacherous DRM at bootloader time or put in some on-board NAND for a bootloader which is then LOCKED IRREVOCABLY with an e-fuse. amazingly they actually did this for business reasons. recognising that factories in China have ABSOLUTELY ZERO software experience and basically "clone whatever they can get their hands on", they reasoned that to support their customers they would need an "unbrickable" SoC.

so they designed the A10 (and A20) so that the SD/MMC 0 interface could double-up as a JTAG / UART port, and could also boot unquestioningly from that very same SD/MMC right from power-on. it loads 16k of "stuff" from SD/MMC directly into memory if there's a "signature" on it (a simple 4-letter text phrase) and off we go.

This said, there is at least one really stupid fail in this design. Allwinner A20 comes with native SATA (AHCI) and real gigabit Ethernet. And things like laptops are really better off using these. One could use e.g. fairly fast SSD attached to SATA for OS and so on. One could enjoy at least quick gigabit links. These are nice options for advanced users like system admins and devs who value trusted environment over glamourous graphics, etc.

*sigh* yeah this is one of those really challenging things that i reluctantly had to deal with, 3-4 years ago. bear in mind that i've been developing and checking and re-evaluating the EOMA68 standard for *FIVE* years. i answer in-depth about this on http://rhombus-tech.net/crowdsupply/#questions, "I'm missing ethernet and SATA", but basically i had to replace SATA and Ethernet with an "up to USB 3.1" option. the reason is: without doing that i would LITERALLY have had only ONE processor that could have gone into the standard in a cost-effective way.

Sure, we all want:
- Non-locked HW without backdoors and misfeatures.
- Cooperative vendor.
- Fully-fledged opensouce drivers, including GPU part.
- Good powerful hardware.
- Low Power.
- Fairly cheap.

But damn, it is not as simple as that. I'm afraid there is no even single SoC or CPU on the planet matching all these wishes at once. If one could live without 3D graphics, allwinner is not worst option around. Just because community did really good job. It isn't allwinner achievement at all, but come on, show us better SoCs!

i'm with you here! and i've been looking constantly for five years! allwinner breaks their SoCs down into two types: those with a maximum of 1280x800 and MALI, and those with PowerVR and much higher resolutions. WHAT GIVES??? come oooon. iMX6 has vivante but has a proprietary VPU. TI uses PowerVR (argh)...

so yeah ... i think it's time to get an SoC made.