VIA Embedded has so far two Freescale i.MX6Q based products, the VAB-820 board and the matching AMOS-820 system. We have started to move from LTIB-based BSP images to Yocto recently for multiple reasons.
The first, very practical one is that Freescale and the community around it has moved onto Yocto for quite a while now. They are core supporters of the project, and we can see that that they are putting a lot of effort into it.
Second, connected reason is that we see the value very much as well, and want to make sure that both we and our customers can take advantage of the improvements. Yocto enables very interesting use cases, and should change quite a bit how companies are thinking about BSPs. It makes possible separating machine support and bootable image generation. A new product added to the existing code can take advantages of (hopefully) tried and tested OS images for specific use cases. Updating machine support (meaning mostly the kernel and u-boot) does not have to wait until a new version of the OS image is released, but can be developed parallel.
The good things also come with a steeper learning curve, though, both for us and for those who want to customize the generated images. To start learning, we have just released the beta version of the new BSP. It's still under development, but should be a good starting point.
What's your experience with Embedded BSPs? Do you use Yocto, or found something else more suitable for your workflow?
Some quick ideas: Probably the quickest would be to create your own working kernel git repo based on the available kernel and adding all the patches you'd like, and use that repo's address in the recipes to create your kernel. So make a repo that is Freescale Kernel + VIA patches + driver backports. This is as opposed to Freescale kernel + VIA patches alone as it is in the current recipes.
Also, a new BSP with 3.14 should be come out not too far in the future, and in that case no backporting is needed. There's an effort to upstream the VIA patches for even more compatibility, but there's no set timeline for that, just work in progress.
And just to note, in case of significant volume of hardware, we also provide ARM software engineering services for exactly these kinds of situations.