Hi everyone,
After a break, we are again having a cross distribution collaboration session[1]. The event is on wednesday 12.10pm PST (aka 19.10 UTC). The session will be broadcast on google hangouts[2] for remote participation.
If you have any great ideas you think other ARM distributions would benefit as well, wish to see something standardized, or have some gripes about supporting ARM - now is your chance to get heard. If you can't make it on wednesday (or just want to start discussing already!), feel free to email the cross-distro mailing listing and I can get it added to the agenda anyways.
Riku
[1] http://lcu-13.zerista.com/event/member/85125 [2] https://plus.google.com/events/c73vv18k4b74g8ku7frnp31h4jc?partnerid=gplp0et...
On 10/28/2013 01:01 PM, Riku Voipio wrote:
Hi everyone,
After a break, we are again having a cross distribution collaboration session[1]. The event is on wednesday 12.10pm PST (aka 19.10 UTC). The session will be broadcast on google hangouts[2] for remote participation.
If you have any great ideas you think other ARM distributions would benefit as well, wish to see something standardized, or have some gripes about supporting ARM - now is your chance to get heard. If you can't make it on wednesday (or just want to start discussing already!), feel free to email the cross-distro mailing listing and I can get it added to the agenda anyways.
I don't rememeber if this came up the last time we discussed bootloader standardization, but there's a spec:
http://www.freedesktop.org/wiki/Specifications/BootLoaderSpec/
... which sounds like it'd be interesting if all bootloaders going forward implemented, and all distros assumed. Pengutronix had a nice demo of this at their booth at ELCE.
Of course, there's also extlinux (syslinux/...?) that Dennis Gilmore proposed the last time this was discussed here, which is likely quite similar.
Perhaps all the distros can just pick one of these options, mandate it in order for a board to be supported by atrbitrary distros in a standard out-of-the-box way (as opposed to people wanting to set things up in a custom fashion, which could still be allowed if they do the work themselves locally), and then we can move forward getting it implementing in relevant bootloaders?
On Tue, Oct 29, 2013 at 04:11:53PM -0600, Stephen Warren wrote:
I don't rememeber if this came up the last time we discussed bootloader standardization, but there's a spec:
http://www.freedesktop.org/wiki/Specifications/BootLoaderSpec/
... which sounds like it'd be interesting if all bootloaders going forward implemented, and all distros assumed. Pengutronix had a nice demo of this at their booth at ELCE.
Interesting, thanks!
I'm not sure that this is quite relevant to the current question, though. The spec presupposes UEFI, which itself is an example of the spec that I'm after. But, AIUI, we can't quite do this today (though I appreciate that support is in the works - thank you Leif and Roy).
In the case of shipping an Ubuntu cloud guest image, for example, I think it's safe to assume that this image won't dual-boot with another distribution.
But what I do want to know is that I can ship a disk image with a GPT and UEFI grub installed, that grub configured to use a particular kernel and initrd from inside the disk image, and that every guest (eg. OpenStack, libvirt) on every OS, and Xen, will be able to boot it. If everything in the stack we need is ready in time (v7, v8, qemu mach-virt if we settle on that, etc), then we can use that. Otherwise, we'll need to do something else for cloud guest images, since right now OpenStack, libvirt and qemu are the targets.
For this goal, I don't think managing multiple distributions on the same disk is in scope.
Robie
On 29 October 2013 23:00, Robie Basak robie.basak@canonical.com wrote:
If everything in the stack we need is ready in time (v7, v8, qemu mach-virt if we settle on that, etc), then we can use that.
What's your definition of "in time" here? (I'm guessing you maybe have a particular release/deadline you're hoping to hit...)
thanks -- PMM
On Tue, Oct 29, 2013 at 11:55:31PM +0000, Peter Maydell wrote:
On 29 October 2013 23:00, Robie Basak robie.basak@canonical.com wrote:
If everything in the stack we need is ready in time (v7, v8, qemu mach-virt if we settle on that, etc), then we can use that.
What's your definition of "in time" here? (I'm guessing you maybe have a particular release/deadline you're hoping to hit...)
Exact timing isn't clear right now. But we have the need to make OpenStack work on ARM (both v7 and v8) which AIUI is a LEG roadmap item.
Ubuntu 14.04 LTS will be our next release, and feature freeze for that is in February.
I expect that we'll have something that we can land by that time. I hope it'll be the final arrangement.
On 10/29/2013 05:00 PM, Robie Basak wrote:
On Tue, Oct 29, 2013 at 04:11:53PM -0600, Stephen Warren wrote:
I don't rememeber if this came up the last time we discussed bootloader standardization, but there's a spec:
http://www.freedesktop.org/wiki/Specifications/BootLoaderSpec/
... which sounds like it'd be interesting if all bootloaders going forward implemented, and all distros assumed. Pengutronix had a nice demo of this at their booth at ELCE.
Interesting, thanks!
I'm not sure that this is quite relevant to the current question, though. The spec presupposes UEFI, which itself is an example of the spec that I'm after. But, AIUI, we can't quite do this today (though I appreciate that support is in the works - thank you Leif and Roy).
Perhaps it's worded assuming UEFI, but I'm pretty sure it either doesn't actually require UEFI in practice, or that there's an extension to that spec being pushed that adapts it to non-UEFI environments. As I mentioned, Pengutronix had a demo of it running at ELCE, and IIRC they were using Barebox not UEFI as the bootloader.
The following is where I got the link to the spec from http://mindlinux.wordpress.com/2013/10/25/barebox-and-the-bootloader-specifi...
In the case of shipping an Ubuntu cloud guest image, for example, I think it's safe to assume that this image won't dual-boot with another distribution.
For me, the interesting part of that spec isn't the multi-distribution-co-existence aspect, but simply the fact that it's a bootloader-independant format for specifying a list of kernels/... to boot. While the spec supports the co-existence stuff, I guess there's no specific need to distros to care about this; they just install the appropriate config files for their own kernels/filesystems, and are done.
On 10/30/2013 11:47 AM, Stephen Warren wrote:
On 10/29/2013 05:00 PM, Robie Basak wrote:
On Tue, Oct 29, 2013 at 04:11:53PM -0600, Stephen Warren wrote:
I don't rememeber if this came up the last time we discussed bootloader standardization, but there's a spec:
http://www.freedesktop.org/wiki/Specifications/BootLoaderSpec/
... which sounds like it'd be interesting if all bootloaders going forward implemented, and all distros assumed. Pengutronix had a nice demo of this at their booth at ELCE.
Interesting, thanks!
I'm not sure that this is quite relevant to the current question, though. The spec presupposes UEFI, which itself is an example of the spec that I'm after. But, AIUI, we can't quite do this today (though I appreciate that support is in the works - thank you Leif and Roy).
Perhaps it's worded assuming UEFI, but I'm pretty sure it either doesn't actually require UEFI in practice, or that there's an extension to that spec being pushed that adapts it to non-UEFI environments. As I mentioned, Pengutronix had a demo of it running at ELCE, and IIRC they were using Barebox not UEFI as the bootloader.
The following is where I got the link to the spec from http://mindlinux.wordpress.com/2013/10/25/barebox-and-the-bootloader-specifi...
In the case of shipping an Ubuntu cloud guest image, for example, I think it's safe to assume that this image won't dual-boot with another distribution.
For me, the interesting part of that spec isn't the multi-distribution-co-existence aspect, but simply the fact that it's a bootloader-independant format for specifying a list of kernels/... to boot. While the spec supports the co-existence stuff, I guess there's no specific need to distros to care about this; they just install the appropriate config files for their own kernels/filesystems, and are done.
The spec, largely, is just the "regular" syslinux format. U-Boot supports this as well, but not in obvious ways (you need to pass the menu in via sysboot or pxe commands).
So.. yes, more distros saying "Do this and we'll support you" would be helpful here.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
El Wed, 30 Oct 2013 09:47:04 -0600 Stephen Warren swarren@wwwdotorg.org escribió:
On 10/29/2013 05:00 PM, Robie Basak wrote:
On Tue, Oct 29, 2013 at 04:11:53PM -0600, Stephen Warren wrote:
I don't rememeber if this came up the last time we discussed bootloader standardization, but there's a spec:
http://www.freedesktop.org/wiki/Specifications/BootLoaderSpec/
... which sounds like it'd be interesting if all bootloaders going forward implemented, and all distros assumed. Pengutronix had a nice demo of this at their booth at ELCE.
Interesting, thanks!
I'm not sure that this is quite relevant to the current question, though. The spec presupposes UEFI, which itself is an example of the spec that I'm after. But, AIUI, we can't quite do this today (though I appreciate that support is in the works - thank you Leif and Roy).
Perhaps it's worded assuming UEFI, but I'm pretty sure it either doesn't actually require UEFI in practice, or that there's an extension to that spec being pushed that adapts it to non-UEFI environments. As I mentioned, Pengutronix had a demo of it running at ELCE, and IIRC they were using Barebox not UEFI as the bootloader.
The examples i believe are using UEFI but it is not a requirement. the implementation is basically extlinux with some set paths. with the sysboot command in u-boot it should be trivial to support. setting up the environment will be the hardest bit. I am working on getting the distro generic u-boot setting i talked about earlier this year implemented. Fedora's wandboard u-boots are using it and it is working well.
The following is where I got the link to the spec from http://mindlinux.wordpress.com/2013/10/25/barebox-and-the-bootloader-specifi...
In the case of shipping an Ubuntu cloud guest image, for example, I think it's safe to assume that this image won't dual-boot with another distribution.
For me, the interesting part of that spec isn't the multi-distribution-co-existence aspect, but simply the fact that it's a bootloader-independant format for specifying a list of kernels/... to boot. While the spec supports the co-existence stuff, I guess there's no specific need to distros to care about this; they just install the appropriate config files for their own kernels/filesystems, and are done.
the filesystem for booting from is shared. so distros will need to make sure not to break other distros. But the idea is a simple unified location for booting from
Dennis
Dennis
-
Dennis Gilmore -
Peter Maydell -
Riku Voipio -
Robie Basak -
Stephen Warren -
Tom Rini