Slashdot Mirror
Linux 2.6.34 Released
diegocg writes "Linux 2.6.34 has been released. This version adds two new filesystem, the distributed filesystem Ceph and LogFS, a filesystem for flash devices. Other features are a driver for almost-native KVM network performance, the VMware balloon driver, the 'kprobes jump' optimization for dynamic probes, new perf features (the 'perf lock' tool, cross-platform analysis support), several Btrfs improvements, RCU lockdep, Generalized TTL Security Mechanism (RFC 5082) and private VLAN proxy arp (RFC 3069) support, asynchronous suspend/resume, several new drivers and many other small improvements. See the full changelog here."
> a filesystem for flash devices
here we go again, unless we stop supporting flash, Apple has refused to distribute dual-boot Linux enabled iPads
Other features are a driver for almost-native KVM network performance
KVM is fantastic virtualization technology, yet Xen gets all the hype these days. Why? Paravirtualization is pretty cool stuff, but seriously, what CPU's are made without some type of hardware-assisted virtualization support?
Is the RT2500-based chipset working reliably now?
The developers switched to a new driver model because it's "better".
If "better" means once-working wifi chipset becomes grossly unstable, previous drivers are considered "legacy" hence will not compile on kernels later that 2.6.29 and current drivers are as stable as a "one-legged man playing football".
A few years later and 2.6.34 is released - is it working yet?
Considering the RT2500 chipset is present many wifi products the current state of "stability" is woefully inadequate.
(and don't get me started on f***ed up i845 drivers for xorg! - worked fine under previous kernels & xorg an update later by both - graphics performance royally screwed and many crashes)
Apart from that - happy Linux user for over 10 years!
With releases like these, it's no wonder M$ is getting worried. Been running this kernel a while now on our production servers (even from before it it was tagged release, I like running bleeding edge in order to get the most performance from my company's hardware investment) and save from a few data corruptions issues, it's been rock stable! I have to play with the new KVM support later on one of the servers with the least amount of customers on it (couple of hundreds), looks nice!
Sadly... it looks like my company is looking at going with Windoze for a few important servers because of a few outtages. I know it was because of faulty hardware, because I had just compiled a custom kernel for those servers with just the right flags needed (I want to get the most performance!) but this must have triggered a hardware bug because the kernel worked fine on my work laptop. Sigh...
Anyway, keep up the good work!
> Some laptops have two GPUs, a low-power and inefficient GPU and a high-power and powerful GPU. Users should be able to switch to one or another at runtime. In this version, Linux adds support for this feature. You need to restart X, though.
How do you restart X without affecting all your GUI apps? If you can't restart X without bringing down your GUI apps, I don't see the point for the target audience.
For some people, "only having to restart X" will only save a bit of time over rebooting the whole laptop, reconfiguring bios etc.
that this kernel already got device IDs for next years Intel hardware. This is something completely new, since Intel so far had a much more closed policy and wouldn't have told device IDs prior to the chipset release.
Now there is a really good chance that driver code will make it into the distribution kernels until the new hardware will be released for mass production. So the chances that brand new hardware will work without any flaws in 2011 are higher than ever before.
Thanks to Intel for this change in their policy. This was a small step for Intel (since everybody "knows" that they will release new chips every year) but a giant leap for providing Linux hardware compatibility right "out-of-the-box".
I'm always amused by at least one strange juxtaposition of the big-serious-enterprise-server stuff that the corporate devs are most interested in and the oddball hobby projects that can get included as well, so long as they follow the kernel process.
In this case, I think it was all the "multi-petabyte scaleable filesystem, esoteric btrfs improvements, kernel virtualization networking stuff, gamecon: add rumble support for N64 pads" that did it.
AOL 9.5 is far more advanced.
They feared that it could be used to suppress protest or support unpopular rule.
There's experimental support for 'hotswitching' called 'PRIME' (for obvious reasons :) ).
See here: http://airlied.livejournal.com/71734.html
When I last used KDE years ago, that didn't work so well. For instance, my ssh connections weren't restored. And even for KDE apps the session saving thing didn't save everything.
You may ask "why should SSH connections be restored?" and I'll reply: why should my apps and connections go down in the first place just because X goes down?
Heck even in windows when I kill "explorer.exe" my apps still keep running. I know it's not the same thing, but who cares when all the fanatics keep saying Linux is so stable, but when X goes down, I lose stuff.
In my experience a cli-based Linux system may be more stable than windows XP, but X can be less stable than Windows XP SP2/SP3. I've got X to lock up or abend a fair number of times, and have had to restart X, this causes me to lose unsaved work. For a Desktop User, it matters little if the rest of the O/S is still running merrily when X goes down and takes the GUI apps along with it.
Yes I could use stuff like "screen" instead and treat X as a glorified interface to multiple screen sessions, but that's pathetic.