One reason why the platform won't succeed

linux-sunxi / u-boot-sunxi Commit c668a5439373e0d6e9c37f1f591313346e6413d1 :

sunxi: remove nand lib and nand related operation

Why should I have to go to all the work to put it back?

Personally, I decided I won't continue to waste time to support people (hipboi & hno) that pull this kind of crap and make things difficult for people who want control of their hardware. I don't like people that remove functionality for corporate reasons. Allwinner also won't release critical closed stuff like BOOT0 and BOOT 1, or help us with Mali and CedarX. That all makes improving the platform almost impossible. Contrary to the spinmeisters trying to sell hardware this platform is still more closed than many other ARM platforms such as OMAP, Tegra, or the Marvell SoCs. In the meantime Raspi now has a huge community, and has just open-sourced the rest of their closed stuff - unlike Allwinner.

I had good hope for the A10 at the beginning of this forum. The hardware had good features. But Allwinner and the linux-sunxi people are holding it back (examples - why is the uboot not yet at Denx, or the kernel at kernel.org?). Without a decent community to overcome the stagnation caused by the linux-sunxi principles (Cubie, Amery, and Henrik) the platform cannot succeed as a popular platform for running Linux with full hardware support.

Arm development is moving extremely fast. This platform will likely be in the history books before it has full support.

Please keep in mind I really don't like saying any of this. But I have to be realistic and commit my time to a Linux project that is more deserving. Thanks to Jeff and everyone else who helped to support my work. Wish I could have done more. . .

What are the corporate reasons?

I was involved in the process when i tried to solve my problems with the mele and everything seemed fine to me (but slow ;-)
Amery has finally done the job (for the sata), Tom pointed the crap on the uboot and Henrik seems to have refactor the clock stuff in uboot (not yet tested).
They didn't seem to block anything on purpose or participating to any conspiracy.

I'm just a passenger on this project and sticked on it on my spare time because i enjoy(ed) my "mele experience".
I just don't get where you want to go with all these allegations.

Everywhere you go and get involve... in a company, a community, a family; No matter what, you should never leave like that with so much disappointment and probably anger: because it's your own choice and your own responsibility.

PS: And i'm sure everyone already knows you prefer guiness more than the a10 now ... but in my opinion, you should have trade it better with one your fine irish whiskey ;-)

I think Gnexus is talking fundamentally about people wanting to make a quick buck selling products that use other peoples work BUT will not recipricate by releasing their portions of the work for other people to benefit from ....


If you step back throught the "A10 story" it goes

Pi came out and created a new market

TCubie and his UK pirate party associate jumped on the back of it (Rhombus tech) with as far as i can tell the aim of monetising some patents (using reccycled connectors to connect core processor board to daughterboards or something along those lines) and trying to get all the design, development, proto, pre production etc for an A10 based board done for zero cost to them upfront

A lot of talk then followed ... with zero outcomes

The mele A1000 came on the scene and someone found out it was so easy to hack (maybe TCubie did this ?)

A lot of talk then followed ... with zero outcomes

A lot of bitching then followed ... with zero outcomes

A lot of forum abusing then followed ... with zero outcomes

(the above three killed off any chance of a "community" developing around the A10 or mele)

We then have hno who to his credit has got and makes efforts to improve a working GIT source .. it's a bit unfair of Gnexus to call it crap as the guy has put in clearly a lot of effort to get it in to a readily usable state (with the help of a good few people such as Guillaume on the SATA) .. and i would imagine with little support rom Allwinner in terms of drivers etc

a couple of chinese maufacturers knocked out dirt cheap A10 boards people try to resell

the cubieboard comes way too late to the game, presumably after a fall out with his pal and the patents ?

So as I see it thats the "A10 charge to copy the pi" story ..

Time waits for noone .... and too many of the tin foil hat brigade spoil everything ;-)

Theres still lots of life left in the mele/similar yet IMHO as a headless server, thats fundamentally what they should focus on and dump the video driver efforts and concentrate on upgrading the kernel and particularly the ethernet drivers/associated code


Thats why Jeff's site is so good, theres only the odd serial abuser/bully .. pity really jeff's site wasnt the focus for the initial A10 mele hacking.

Hopefully jeff will get a section up and running for the imx6 soon ...


rgds

ian

Ok, it's just a giant cluster fuck... it's sad.

I still don't get why Tom is still wasting his time by selling mele platform at that low price if he is so interested to make easy money and sell properly the cubieboard.

Something just don't fit and don't make sense.
I'm not sure the view is as manichean as you say.

Agreed with you, hyena, about the headless server or appliance.
It's the only pragmatic usage to consider of the mele (and it's already a lot)...
that's why i got here the first time.

What's Jeff's website?

Quote
guillaume
What are the corporate reasons?

Obviously Allwinner does not want to release a lot of the A10 software code. I can only assume that it was either that, or laziness in Cubie's attempts to get the SPL to work with the NAND code. Either way it is a piss poor excuse for pulling NAND support.

Quote

you should have trade it better with one your fine irish whiskey ;-)

Agreed, but the cost of $7 for a single A10 only buys one pint of Guiness here. My A10 tablet is cost of one bottle of whisky.

Quote
hyena
I think Gnexus is talking fundamentally about people wanting to make a quick buck selling products that use other peoples work BUT will not recipricate by releasing their portions of the work for other people to benefit from ....

Exactly!

Ian's comments here and elsewhere in this forum are actually what started our drunken discussion which ended with deleting the site. But actually it goes back much further. After the site was completed but not yet public we very seriously considered not publishing it. The work alone was sufficient for us to be satisfied. The reason then to not publish it was that it primarily benefits the sellers of the devices and Allwinner, who are just riding the coat tails of the now very open Raspi. Allwinner has not endorsed the open source community.

Quote

The mele A1000 came on the scene and someone found out it was so easy to hack (maybe TCubie did this ?)

Cubie is the main guy that started all of this. He created the SPL which allows to boot from MMC. He then hid the fact that you could also boot from NAND. We now consider him not trustworthy. When I first started this forum I did not know the full details. He is the one who supposedly "leaked" the kernel and u-boot. But how do you you go against your employers wishes and yet remain employed by them? The code that was "leaked" is actually all GPL anyway. So Allwinner was previously violating the GPL by not releasing it. Then the NAND code was pulled. Allwinner still refuses to release BOOT1, which I can almost guarantee has GPL stuff in it. Then the Cubieboard (Beagleboard, Pandboard knockoff) suddenly appears. Now uboot-allwinner only has default support for Cubie and Olimex stuff - Cubieboard and Mele. . . It was all a bit too dodgy for us to keep supporting it unless an alternate open community arose from our site.

Quote

it's a bit unfair of Gnexus to call it crap as the guy has put in clearly a lot of effort to get it in to a readily usable state

Maybe crap was too strong a word. But lichee is better code than exists in uboot-allwinner. The more work that gets done on hno's the more it deviates from standard Denx, and the more difficult it gets to add support for NAND, SATA and USB.
Uboot-multi manages to boot from those devices by using an initrd, but native uboot support should have been there.
Marvell and Denx has support for all of that in u-boot, and it should have been in the Allwinner code from day one.

Quote

Something just don't fit and don't make sense.

Yes. That is the problem. It seems just a bit too dodgy for us. Too many unanswered questions, and a silent Allwinner. Very silent. . . No word yet on their new SoC. It's too late from Christmas now. Their old SoC is now obsolete. No market share for them this Christmas except maybe in China. They will likely be hurting badly in this economy. They are a private company. No financial details. Who are they owned by? We can't even locate that information. It is just a bit too opaque.

After looking at the Freescale IMX specs and docs, and all the available tablets and devices, I'm quite happy about selling my Mele. Much better Linux support from Freescale! This board here looks very interesting with quad-core, gigabit ethernet, and at least 1GB ram. Their Hi802 (now we also have MK802 name knockoffs!) has the similar specs and is only $89.

ARM platforms are quickly moving forward. Time to move on.

I'm already enjoying this again! Looks like I just needed to get away from Allwinner. Company was too depressing. . .

Thanks Ian!



Edited 1 time(s). Last edit at 10/27/2012 11:34PM by gnexus.

Gnexus,

for the imX6 .. you might find some useful docs in the links/posts in the powerdevelopers forum

http://www.powerdeveloper.org/

This is from a little ARM/linux US outfit called Genesi who were really well in front of the pi with a lovely little sub $50 imx5 board .. but turned chicken in releasing it and then turned alternating shades of red and green when the pi started selling in the hundreds of thousands ;-)

.. they have been promissing a cheap imx6 board for a year but i suspect they have "pivoted" now to be just ARM design consultants, but you may find lots of savvy imx6 guys in and around its powerdeveloper forums .. it was (maybe still is) very close to freescale as a developer

Samsung I've heard has a decent reputation as well for such thing as releasing/supporting linux SOC drivers and its a pretty professional organisation, its rumoured their nice quadcore SOC chip is targetting this low cost segment as well .. so this might be an option as well.

rgds

ian

Thanks for the info!

My main personal goals were to have a tablet that could support alternate OS than Android - like Meego/Mer or Tizen, and to have an low-power device that can function as a MythTV frontend. I don't need another server device like a Mele. So both devices need to have working drivers for X. imX6 looks promising as supposedly there are now or soon open Vivante drivers.

My goal for the community is to get something together that highlights ARM platforms with decent GNU/Linux support. Then we would not need spend time hunting that information. A10 could of course be part of that, but with a few '*'s as advisories. . .

The CNX site is very good, as is Jeff's. But we need something more community-oriented than CNX, and less spartan than Jeff's, that offers information, tutorials, and basic things to get people going for each platform. That is currently sorely lacking. The cost of energy keeps rising, and ARM platforms keep getting more powerful. There will be huge demand. But the general public needs an alternative to Android, which is the main ARM OS. It is just too limited. Such a site would be very popular if it was properly balanced, and had the right information and the right community behind it.

i.MX6 according to the XBMC guys:

Quote

the GPU is over twice as fast as NVidia Tegra 2 (90MT/s) and over 5 times faster than Mali 400 (30 Mt/s). 5 times is exactly the performance improvement announced earlier for Mali T-604

Looks like the video codecs will be available too! Closed source of course. But they will be available for GNU userspace.

another quote from that thread:

Quote

it's better to wait to see real code and real linux tests (someting learnt from the a10 experience).

Sounds promising. But for now I may just go with an Alien to replace my Mele as a MythTV player. I know it would do the job ;)

Sorry gnexus that you feel this way about my work on u-boot

The nand driver was removed by Tom (hipboi) from lichee-dev-mmc (what is nowdays u-boot-sunxi) precisely because it deviates completely from upstream u-boot, done completely wrong way and in bad style. And tryign to fix that up was not inline with the goal of making an u-boot that can boot directly from MMC/SD.

My work on u-boot is primarily on the SPL (what makes booting u-boot directly from MMC possible), and of my own interest only. There is no corporate or financial interests.

I have not added nand driver back on uboot-sunxi because

* For a long time the licensing of that driver was unclear. Thankfully Tom have cleared that up lately.

* My interest have been on SPL to avoid the need of proprietary boot0/boot1/boot.axf botloaders. Have not really needed NAND support.

* Second goal is to be able to upstream the changes to mainline u-boot. The NAND driver and the other parts that was removed have no chance of getting accepted upstream.

* My time is very limited, doing Allwinner u-boot stuff on my spare time with a full time job and 6 kinds to care for.

* Don't at all like the hackish Allwinner NAND implementation in lichee-dev. There is nothing done right in the u-boot integration there.

* Lack of documentation of the NAND controller, and no working code suitable to be used as basis for an NAND SPL u-boot loader.

I have always been clear to anyone who bothers asking that if you want NAND support then for now use the lichee-dev tree where there is a working NAND driver.

We are now slowly looking into the NAND controller with primary goals of using it as an MTD. But it is not a high priority. SD is there and works (plus a lot faster).

Quote

SD is there and works (plus a lot faster).

SD is much slower. . . That is common knowledge and shows either your ignorance or your deception.

Quote

The NAND driver and the other parts that was removed have no chance of getting accepted upstream.

Basically all the other SoC manufacturers have their code hosted at Denx as custodian trees. If the Allwinner NAND drivers are so bad that it cannot be hosted there then the code is probably stolen from another manufacturer or software developer.

Allwinner was purported to be an open-source supporter. But we find that they are actually GPL violators, and probably also have stolen valuable intellectual property from other SoC manufacturers. The entire A1X platform is now seeming like a joke.

Quote
gnexus
Basically all the other SoC manufacturers have their code hosted at Denx as custodian trees. If the Allwinner NAND drivers are so bad that it cannot be hosted there then the code is probably stolen from another manufacturer or software developer.

The NAND controller and driver is Allwinner creation.

The driver consists of

1. A platform independent driver with origins in their earlier RTOS for F20 and earlier chips (same is used for boot1 & boot.axf, boot.axf is running that same RTOS from wahat I understand).

2. A simple platform layer to match u-boot functions

3. Cudely hacked into the mtd nand layer of u-boot, despite allwinner nand driver being an emulated block device ontop of nand and not at all mtd.

4. Also hacked into fat layer a bit.

Quote

The NAND controller and driver is Allwinner creation.

Hmm. Okay then, let me get this straight. Please correct me if I'm mistaken. It is Allwinner software, but can't be at Denx as:

Quote

have no chance of getting accepted upstream

That means then that it is either proprietary non-GPL stuff, "leaked" by Cubie without Allwinner permission, that has no place in my open source project, but that I mistakenly used because I cloned your "open source" git, and that now I could possibly get sued by Allwinner for using and distributing it?

or just that it is a "hackish Allwinner NAND implementation" that is "Cudely hacked into the mtd nand layer of u-boot" that is still GPL, but code that is so bad nobody wants to touch it ?

or both D:<

If it's redone as a proper block device in u-boot, or even better as an mtd driver with the Allwinner logical drive layer ontop then it can be discussed.

The current implementation, no, it's just to ugily done in how it interfaces with u-boot.

There is no known GPL issues lichee-dev.

As mentioned earlier lichee-dev-ICS lacks proper license in NAND driver and some other parts, but it's confirmed by Allwinner via Email that the Android SDK sources uncluding u-boot and kernel is free software.

And at the same time Tom with Allwinner official hat on cleared up the license doubts regarding the NAND driver in lichee-dev.