can't wait to upgrade..
eh.. no.. 2.6.32 works well for my laptop and desktop (both 3+ years old)
The Linux kernel has been updated, again. On Sunday, Linux Lord Linus Torvalds announced version 3.16 of the kernel is now good to go. Torvalds says “3.16 looked a bit iffy for a while” but “things cleared up nicely, and there was no reason to do extra release candidates like I feared just a couple of weeks ago.” You'll be …
Just curious, but why in the world would you keep yourself on an ancient version of the kernel? I mean really, this December that version will be five years old. While I sometimes agree with the "not broke, don't fix" mindset, I can't call not upgrading for that long anywhere near a best or safe practice.
Mainly because of patch porting.
In general an embedded system needs a few custom drivers (or driver customisations). These are often hard to port because kernel interfaces etc change. Once you have a stable platform it is often hard to convince management (or customers) to refresh their kernel at great expense when they see little or no benefit.
As a result there are many embedded systems still using 2.6.x and in active development. I was modifying a driver in a 2.6.x system just last week. There are still quite a few using 2.4.x, but they will be thinner on the ground now.
The latest changes have been a hard time to go through. The changes to device tree have been painful, particularly as some drivers were migrated and others not (or had been partially migrated) and device tree syntax has been a moving target. Now we are through that some benefits are to be had.
Mainly because of patch porting.In general an embedded system needs a few custom drivers (or driver customisations). These are often hard to port because kernel interfaces etc change. Once you have a stable platform it is often hard to convince management (or customers) to refresh their kernel at great expense when they see little or no benefit.
Bearing in mind the original author specifically mentioned laptop and desktop, not embedded…
There's a good number of SoCs that are actually quite well supported in the Linux kernel. Take this embedded device for example:
http://wiki.embeddedarm.com/wiki/TS-7670
It ships with kernel 2.6.35. It took me a couple of days to come up with this:
http://bne.vrt.com.au/technologicsys/
And that supports nearly everything on the device.
Technologic Systems based their kernel on what Freescale shipped, so it's not their fault that the kernel was ancient. Freescale have a habit of doing this, they did it with the i.MX27 too, and even there, I remember getting about 80% of the thing going on the latest kernel of the day (and before DeviceTree too).
The support for the i.MX286 used in this device is so complete in the mainline kernel, I have to ask whether maintaining a separate branch is any less resource intensive then perhaps writing the one or two drivers that might be missing?
The beauty of Linux these days is you can just update the kernel, keep the rest more or less the same, and roll back easily if things go pear shaped. Kernel didn't work out? Just reboot, choose the older one in the boot menu then uninstall the newer one when you're booted.
It's the Windows equivalent of running a Windows 8.1 kernel with a Windows 7 UI, with the ability to roll back and forward kernel updates with just a simple reboot.
There are lots of up to date BSPs for ARM SoCs that are still using 2.6 series kernels. Mainly vendors like Marvell that have a bunch of hacked up drivers that were impossible to mainline.
The device tree stuff has also meant that a lot of ARM stuff is still on earlier 3 series kernels because of breakage resulting from the uptake of DT in more recent kernels.
This post has been deleted by its author
Linus might have kicked off with Linux V0.x on a 386 SX in his bedroom, but Linux has been predominantly an ARM kernel for at least 5 or 6 years now.
For every x86 running Linux there are probably a hundred ARMs in everything from phones to building control and sewing machines. It has been like that for years.
Predominantly arm and predominantly working on arm is a different story.
The interesting announcement is proper Exynos support. Presently the only thing that works on these are kernels from vendor repositories such as 3.4.0 (yes, 0) for the ARM Chromebook. They have breakage across the board in various key subsystems which are by default disabled in the original crhomebook kernel including basic stuff like NFS. Other SoCs are not much better. If you look f.e. at kernels used in various Android phones, etc you will see a mix of 2.6.33 and an occasional 3.0. Off the top of my head I cannot think of a phone or tablet that is beyond 3.4 and which will work properly using the generic kernel off kernel.org without a raft of vendor patches. There are in fact only a handful of SoCs which are "open" enough to run with a stock kernel and for some of them there is no way in hell to get the actual kernel mods (in blatant violation of the GPL).
So if we have a stable non-3.4 kernel that finally has decent Exynos support, that will be quite interesting. Ditto for other SoCs.