I know we don't support tablets. That's been pretty comprehensively covered on the list already, eg:
https://lists.fedoraproject.org/pipermail/arm/2012-April/003107.html
But! Suddenly a tablet appears which is both much cheaper and (in some respects) far more powerful than the Trim Slices and BeagleBoards that we do support.
Is it possible to support Tegra2/3 tablets, at least in a headless configuration until there are graphics drivers?
Rich.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
El Tue, 3 Jul 2012 15:13:09 +0100 "Richard W.M. Jones" rjones@redhat.com escribió:
I know we don't support tablets. That's been pretty comprehensively covered on the list already, eg:
https://lists.fedoraproject.org/pipermail/arm/2012-April/003107.html
But! Suddenly a tablet appears which is both much cheaper and (in some respects) far more powerful than the Trim Slices and BeagleBoards that we do support.
Is it possible to support Tegra2/3 tablets, at least in a headless configuration until there are graphics drivers?
Rich.
It should be doable, aboot that android devices use is different to uboot. on my ac100 ive had troubles getting an initrd loaded. i get errors about the initramfs being corrupted. its also space constrained so you have to make a host specific initramfs to fit in the 8mb limits it places on kernel and initramfs. you could possibly also just use the android kernel though im not sure if that is sufficient. I know KDE has been working on a touch interface, pretty sure gnome has also. I do expect at some point soon we will see tablet spins show up for arm and x86.
Dennis
On 07/03/2012 03:23 PM, Dennis Gilmore wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
El Tue, 3 Jul 2012 15:13:09 +0100 "Richard W.M. Jones"rjones@redhat.com escribió:
I know we don't support tablets. That's been pretty comprehensively covered on the list already, eg:
https://lists.fedoraproject.org/pipermail/arm/2012-April/003107.html
But! Suddenly a tablet appears which is both much cheaper and (in some respects) far more powerful than the Trim Slices and BeagleBoards that we do support.
Is it possible to support Tegra2/3 tablets, at least in a headless configuration until there are graphics drivers?
Rich.
It should be doable, aboot that android devices use is different to uboot. on my ac100 ive had troubles getting an initrd loaded. i get errors about the initramfs being corrupted.
The only time I have seen that problem was when I forgot that I excluded all compression methods except xz from the kernel, but I compressed the initrd using bzip2. Easy mistake to make and an easy thing to forget sufficiently after a few months that it has you chasing your tail for a while. Might be worth checking if that is your problem, too. :)
its also space constrained so you have to make a host specific initramfs to fit in the 8mb limits it places on kernel and initramfs.
8MB for the primary boot image. 5MB for the backup boot image if you are running the Android 2.1 boot loader. Very handy if you break your primary boot image. :)
5MB isn't too unreasonable, once you cut down what you are including. I know dracut has a habit of including everything plus the kitchen sink by default, and spewing out a 14MB (after compression) initramfs image, but that's just perverse. You can cut that down in half easily, and xz will get you under the 5MB limit.
you could possibly also just use the android kernel though im not sure if that is sufficient. I know KDE has been working on a touch interface, pretty sure gnome has also. I do expect at some point soon we will see tablet spins show up for arm and x86.
The OP was asking for headless, so GUI seems like very much an optional extra. I, too am quite interested in using these, headless or otherwise since they would make handy compile boxes, especially at the price they are selling at.
Gordan
On 07/03/2012 03:13 PM, Richard W.M. Jones wrote:
I know we don't support tablets. That's been pretty comprehensively covered on the list already, eg:
https://lists.fedoraproject.org/pipermail/arm/2012-April/003107.html
But! Suddenly a tablet appears which is both much cheaper and (in some respects) far more powerful than the Trim Slices and BeagleBoards that we do support.
Is it possible to support Tegra2/3 tablets, at least in a headless configuration until there are graphics drivers?
Is there any particular reason why you cannot do something like: 1) root the device 2) dump the rootfs in a subdirectory 3) chroot /sbin/init ?
Assuming you enabled sshd, you should at that point be able to ssh into the tablet. Or is there something in the Android kernel patches that stops this approach?
Gordan
On Tue, Jul 3, 2012 at 3:23 PM, Gordan Bobic gordan@bobich.net wrote:
On 07/03/2012 03:13 PM, Richard W.M. Jones wrote:
I know we don't support tablets. That's been pretty comprehensively covered on the list already, eg:
https://lists.fedoraproject.org/pipermail/arm/2012-April/003107.html
But! Suddenly a tablet appears which is both much cheaper and (in some respects) far more powerful than the Trim Slices and BeagleBoards that we do support.
Is it possible to support Tegra2/3 tablets, at least in a headless configuration until there are graphics drivers?
Is there any particular reason why you cannot do something like:
- root the device
- dump the rootfs in a subdirectory
- chroot /sbin/init
?
Assuming you enabled sshd, you should at that point be able to ssh into the tablet. Or is there something in the Android kernel patches that stops this approach?
Nothing in the Android kernel stops that approach (I think that's what Ubuntu does on some of theirs) but it's more the things that Fedora (primarily systemd) needs to be in the kernel such as cgroups that may cause problems but I'm not overly up on my android default configs to comment further.
Peter
On Tue, Jul 03, 2012 at 03:23:41PM +0100, Gordan Bobic wrote:
On 07/03/2012 03:13 PM, Richard W.M. Jones wrote:
I know we don't support tablets. That's been pretty comprehensively covered on the list already, eg:
https://lists.fedoraproject.org/pipermail/arm/2012-April/003107.html
But! Suddenly a tablet appears which is both much cheaper and (in some respects) far more powerful than the Trim Slices and BeagleBoards that we do support.
Is it possible to support Tegra2/3 tablets, at least in a headless configuration until there are graphics drivers?
Is there any particular reason why you cannot do something like:
- root the device
- dump the rootfs in a subdirectory
- chroot /sbin/init
?
Assuming you enabled sshd, you should at that point be able to ssh into the tablet. Or is there something in the Android kernel patches that stops this approach?
As Peter said, I didn't know the Android kernel was sufficiently compatible. I'll give it a go later on my Galaxy Tab.
Rich.
On Tue, Jul 3, 2012 at 3:13 PM, Richard W.M. Jones rjones@redhat.com wrote:
I know we don't support tablets. That's been pretty comprehensively covered on the list already, eg:
https://lists.fedoraproject.org/pipermail/arm/2012-April/003107.html
Actually that doesn't state we don't comprehensively want to support tablets, it states at the moment it's hard to support tablets. I would love to be able to support tablets and the touch stuff is improving all the time, like with F-17 we actually now have the underlying support for multitouch, now if only we had some even half decent X drivers we'd be well on the way there. There's even people working on the gnome touch stuff on ARM so it should improve even more in the F-18 cycle!
But! Suddenly a tablet appears which is both much cheaper and (in some respects) far more powerful than the Trim Slices and BeagleBoards that we do support.
Is it possible to support Tegra2/3 tablets, at least in a headless configuration until there are graphics drivers?
It's possible if the relevant kernel code is upstream. I've got one on order with the intention of looking at what would be required to make it work and to hack around with it. With 3.4 the beginnings of the tegra30 support started to appear. I'm hoping google is a lot better at pushing it upstream than say Toshiba with the AC100. There is a basic modesetting kernel side bits hanging around but still not made it upstream that would even possibly give us basic 2D X.
I'm also looking at the asus TF series as they have real keyboards as well to see what's required to support them, with luck being manufactured by the same company as the Nexus 7 and the same underlying SoC we might be able to get a whole group of devices in a single swing :-)
Peter
On 07/03/2012 03:32 PM, Peter Robinson wrote:
On Tue, Jul 3, 2012 at 3:13 PM, Richard W.M. Jonesrjones@redhat.com wrote:
I know we don't support tablets. That's been pretty comprehensively covered on the list already, eg:
https://lists.fedoraproject.org/pipermail/arm/2012-April/003107.html
Actually that doesn't state we don't comprehensively want to support tablets, it states at the moment it's hard to support tablets. I would love to be able to support tablets and the touch stuff is improving all the time, like with F-17 we actually now have the underlying support for multitouch, now if only we had some even half decent X drivers we'd be well on the way there. There's even people working on the gnome touch stuff on ARM so it should improve even more in the F-18 cycle!
But! Suddenly a tablet appears which is both much cheaper and (in some respects) far more powerful than the Trim Slices and BeagleBoards that we do support.
Is it possible to support Tegra2/3 tablets, at least in a headless configuration until there are graphics drivers?
It's possible if the relevant kernel code is upstream. I've got one on order with the intention of looking at what would be required to make it work and to hack around with it. With 3.4 the beginnings of the tegra30 support started to appear. I'm hoping google is a lot better at pushing it upstream than say Toshiba with the AC100.
This is another part that is really perverse - it shouldn't be Google/Toshiba pushing Tegra support upstream, it should be Nvidia.
I'm also looking at the asus TF series as they have real keyboards as well to see what's required to support them, with luck being manufactured by the same company as the Nexus 7 and the same underlying SoC we might be able to get a whole group of devices in a single swing :-)
We can but hope.
Gordan
On Tue, Jul 3, 2012 at 3:35 PM, Gordan Bobic gordan@bobich.net wrote:
On 07/03/2012 03:32 PM, Peter Robinson wrote:
On Tue, Jul 3, 2012 at 3:13 PM, Richard W.M. Jonesrjones@redhat.com wrote:
I know we don't support tablets. That's been pretty comprehensively covered on the list already, eg:
https://lists.fedoraproject.org/pipermail/arm/2012-April/003107.html
Actually that doesn't state we don't comprehensively want to support tablets, it states at the moment it's hard to support tablets. I would love to be able to support tablets and the touch stuff is improving all the time, like with F-17 we actually now have the underlying support for multitouch, now if only we had some even half decent X drivers we'd be well on the way there. There's even people working on the gnome touch stuff on ARM so it should improve even more in the F-18 cycle!
But! Suddenly a tablet appears which is both much cheaper and (in some respects) far more powerful than the Trim Slices and BeagleBoards that we do support.
Is it possible to support Tegra2/3 tablets, at least in a headless configuration until there are graphics drivers?
It's possible if the relevant kernel code is upstream. I've got one on order with the intention of looking at what would be required to make it work and to hack around with it. With 3.4 the beginnings of the tegra30 support started to appear. I'm hoping google is a lot better at pushing it upstream than say Toshiba with the AC100.
This is another part that is really perverse - it shouldn't be Google/Toshiba pushing Tegra support upstream, it should be Nvidia.
No, that's not correct. Nvidia is quite good at getting the core tegra SoC bits upstream as can be seen in the kernels with the tegra30 support. But each manufacturer wires up each device differently and then adds a whole raft of other stuff from compasses to wifi to sound chips and even connects them to different pins on the SoC depending on their board layout or the direction of the wind and it's this that is the problem and it is asus/toshiba/google's job to get that upstream because the SoC manufacturer has no idea how each company using their SoCs does this.
Peter
On Tue, Jul 03, 2012 at 03:32:39PM +0100, Peter Robinson wrote:
It's possible if the relevant kernel code is upstream. I've got one on order with the intention of looking at what would be required to make it work and to hack around with it.
[...]
I'm also looking at the asus TF series as they have real keyboards as well to see what's required to support them, with luck being manufactured by the same company as the Nexus 7 and the same underlying SoC we might be able to get a whole group of devices in a single swing :-)
I ordered a Nexus 7 too.
I also played with the Asus Transformer, which if it didn't have Android could just about be a usable mini-laptop (palmtop?). I think it would make an excellent Fedora laptop.
And I've got a Samsung Galaxy Tab 10.1 with the external keyboard which is also a potential mini-laptop. Apart from the larger screen it is quite similar internally to the published specs of the Nexus 7 (dual core Tegra2 vs quad core Tegra3). Ubuntu is available for it, but I haven't tried it yet.
Rich.