Posts Tagged ‘arm’

h1

Two different Android kernel trees?

November 21, 2007

Is Google hiding the Android kernel source?

Ok, that was a bit extreme… but definitely Google/OHA is not making it easy to reach the android kernel source code. You can of course google it, and find git.android.com and find a git repository of what we can call version A of Android source, but other than the announcement in the ARM Linux Kernel Mailing List (about which I commented in a previous post) there is really no way to get to it from anywhere else (in particular from the Android website or the OHA one).

However if you go through the “fine print” in the Android website (that is, go to Download the SDK, and then go to the bottom of the page and click through the not so flashy link Android project page ) and that will take you to what I would call version B. I am giving different names because they are definitely not the same thing.

Version A last update seem to have been on November 8 (13 days of no activity by the time of this post), and version B is a static tar.gz file dated November 11 (I don’t even know if that file is being updated).

There are of course more differences than the latest modification, for instance version A clearly is focused on mach_msm (for the Qualcomm chipset I mentioned in my previous post) and has as the proposed board definition the “halibut”, it is noticeable that most of the code contributions in this one are found in the architecture section of the code (arch/arm/mach-msm).

In Contrast version B is targetted to mach_goldfish and uses the board definition for “goldfish”, in this one the specific code is equally split between the architecture section and the drivers section (this one clearly named drivers/android).

Halibut and Goldfish are two of the at least four boards that Google registered at the ARM Linux Machine Registry (being the other two Sardine and Trout), it is funny that instead of having a single repository including the two, they have this two separate versions… there is something fishy here, and I am not talking about the board names.

Probably there is much more that can be inferred about potential specifications by looking at both sources, but I will leave that for another day.

Maybe I am just being paranoid about the android here and there is some reason to all this, but I am failing to see it.

h1

Xen Summit Fall 2007

November 20, 2007

I was hoping for the material to be available already so I could link to it, but it is not yet the case.

So what is in store for Xen?

As I mentioned previously for day one some directions on Xen roadmap were shown some interesting xen extensions were shown, one I found particularly interesting is JavaGuest (Java running on top of Xen, allowing a full Java stack… more on that in a post to follow).

Day 2 started with AMD and Intel updates, and as I said it is really great that they are caring about virtualization. I am particularly excited about the features coming up with Intel VT-d which allows DMA for IO devices by means of DMA remapping (basically assigning regions of host physical memory as “protection domains”, assigning devices to this domains and remapping DMA access to host physical addresses on this domains, the remapping being done in HW).

Also very interesting was Samsung’s presentation on the ARM port of Xen, which they showed originally on the previous Xen Summit. Now it is even more advanced, they even showed Doom running in one of the domains, and following the current trend of disaggregating Dom0, they have a separate domain for drivers. Very good work, I really hope they release their work, there are a lot of great things that can be done with it.

Very interesting work for consolidating memory in Xen (the more VMs you run the more important this will become) by using Copy on Write was also shown, and a great approach at treating Xen VM as transactional applications (checkpointing the VM every few milliseconds, executing speculatively and holding any output in between checkpoints).

Finally, the recently announced OracleVM (Sun, Novell and RedHat also showed their stuff), now that Viridian is going to be released soon, pretty much all the major players have a virtualization solution in hand, this is definitely getting big.

h1

Android’s source… HTC Dream?

November 10, 2007

So the Android team at Google released the source code of the work they are doing on Linux with support of Qualcomm for the MSM7200A SURF Development board (MACH_HALIBUT). (Full announcement here)

Probably this means that the first Android enabled “gPhones” will be supporting the Qualcomm Mobile Solution Modems 7200 Chipset. So at least we know that much of the potential specifications of the future phones.

It is well known that HTC is working in the so called “Dream” phone, whose specifications have not yet been revealed, but it has been speculated that it may be the HTC Omni.

Since both are using the MSM 7200, I am tempted to say that it may be the case, and if not at least the suspicions are well funded.