Yes, I agree with you: liveusb images are just a hack and are not reliable. They're useful for quick QA tests and for developers. But they have serious drawbacks.
Since the activation of SMACK recently, there's no more liveusb able to boot because the initramfs (a cpio archive) doesn't contain any SMACK label. When the kernel loads the initrd, even if it contains systemd and the initial services, nothing happens: the system is locked by SMACK because the files in ramfs don't have the appropriate labels. We could try to workaround this by creating a first service that would apply the labels on the ramfs... It's still hacky ! If we can distribute Tizen:Generic with another format, it's a better idea. The format has just to be as convenient as liveusb: easy to spread, easy to flash, bootable directly etc. As Tizen:Generic is quite new, we're ready to evaluate alternatives to liveusb: adopting your method seems a good idea. I'll dig on this tomorrow. If you have more detailed instructions, please share a wiki page. Thanks Artem ! -- Stéphane Desneux Intel OTC - Vannes/FR gpg:1CA35726/DFA9B0232EF80493AF2891FA24E3A2841CA35726 2014/1/13 Artem Bityutskiy <[email protected]> > On Mon, 2014-01-13 at 17:18 +0100, Stéphane Desneux wrote: > > - liveusb doesn't boot (SMACK labels not set in initramfs) > > This is my personal opinion, but I would not recommend you to use > liveusb images at all. They are really complex and in my opinion very > difficult to maintain. Simply because most of the logic about how it is > build is coded in MIC (all the special initrd and squashfs wrapping). > > So not only the wrapping is an unnecessary complexity, but changing this > is very difficult. Indeed, you need to change MIC, then make sure it is > released (you may wait a month here), then make sure it is deployed. And > repeat this if you want to change it again. > > I recommend you to try the approach we taken in Tizen IVI. > > Raw images are just fine, and their beauty is their simplicity. > > Flashing them is fast if you use bmap-tools. They are based on the basic > idea which is cool and elegant technology. I did not invent it, just for > reference. It was actually used in production line for a real phone. > > (BTW, the technology may be very useful for Tizen phone, in a bit > different form to bmap-tools. If someone is interested, we could discuss > separately too. I already tried to point at this to Łukasz, but I think > he may not understand me correctly and he quickly dismissed the entire > idea). > > The gap was - installing to the internal disk. > > I've covered this gap with the 'setup-ivi' project. Now we can chose > whether we boot from the USB stick or we want to install to the SSD/HDD. > I did not deploy this yet, because I am waiting for the Tizen IVI > release. When it happens, I'll deploy that. > > We can generalize this a bit and make a 'setup-general' package or > something which will use the same. > > This way you have the gap covered. > > Really, if you think about this, liveusb stuff is kind of a hack, while > raw image + bmap + setup-ivi is a cleaner approach. > > -- > Best Regards, > Artem Bityutskiy > >
_______________________________________________ Dev mailing list [email protected] https://lists.tizen.org/listinfo/dev
