Linux updates without reboot [a commercial product, not FOSS]

[TWPonKubuntu]

I found this item and my thought is that this looks too much like a route for malicious software to attack a system.

http://phys.org/news/2014-01-world-reboot-less.html

While the overall concept (no need to reboot on update) is valid (and commercial, not FOSS), there is no discussion about what security measures have been or SHOULD be used in actual implementation.

At the least, I see the commercial nature of this software will prevent it from being used by most linux and FOSS fans. However the larger business IT departments may find it desirable... If they don't shoot themselves in the foot by installing it without a lot of prior thought...

[GreyGeek]

When I was using Win3FWG I ran it from within OS/2 to reduce its crashes and to recover gracefully when it did crash. I left OS/2 + Win3 and moved to Win95 when I bought my next computer, which was on Dec 29th, 1997. Between Jan 1, 1998 and May 1, 1998 that Sony VAIO desktop crashed several times and hour. Sony had a layer between the hardware and Win95 called "Medi-Kit". It's purpose was to catch Win95 abends and recover gracefully without having to reboot. One could always tell when that happened because every thing would freeze for up to a minute. That freeze, I learned, was Medi-Kit saving the state of the machine and then doing a warm reboot without erasing the screen, leaving the impression that the machine was still running. At least 2 or 3 times an hour Medi-Kit failed to do what it was intended for and I had to reboot.

When I was running NT4 it, or one of the applications it was running crashed frequently. There was a website that you could download a tool from which would track your crashes and report back on what caused the crash, either the OS or one of the utilities. It ran for several years before it was shut down, some claim by Microsoft, which shut down the "***kMicrosoft.com* site.

When Win2K came along the crashes appeared to stop, but I noticed a screen freeze that was similar to what Medi-Kit was doing. One of the technical journals reported that Win2K was "phantom rebooting" when it or a utility crashed, and that there was a crash counter address which kept track. It reported back to MS for the first 10 crashes but didn't after the 10th. My counter was long past 10. But, Win2K was, compared to Win3, Win95 or NT4, a reasonably stable OS, but nowhere near as stable as OS/2. WinXP rarely crashed on me, and when it did it was usually hardware related. I retired before Win7 was released and haven't used it enough to comment on its stability, but I hear it is better than WinXP.

I say all that to say that I suspect what they might be doing is a warm reboot, like Medi-Kit or Win2K attempted. I can't imagine them shifting the memory location of a running Linux kernel "on the fly" without masking the interrupts to prevent the OS or an app from being called while the move was in progress.

[TWPonKubuntu]

Ah! I was an IBM rep for OS/2, back in the day. It was far better than anything MS offered then. I suspect that OS/2 would be better than what they now offer.
Sadly, IBM had it's head in a bad place and lost the "OS wars".

Hindsight tells me I should be glad IBM did not succeed, but OS/2 was still a great option.

I'm just glad I have the Linux option(s) now.

[NickStone]

Interesting article, but Linux distros can do that anyway. Updating any Linux distro can be updated without the need to re-boot (maybe except the colonel). All the guy has done is use the Linux kernel to perform updates to system files whilst still running. But you don't ned their software just use any Linux distro.

[SteveRiley]

I say all that to say that I suspect what they might be doing is a warm reboot, like Medi-Kit or Win2K attempted. I can't imagine them shifting the memory location of a running Linux kernel "on the fly" without masking the interrupts to prevent the OS or an app from being called while the move was in progress.

Shifting pointers to new memory locations is exactly how KSplice works:

"To apply a patch, Ksplice first freezes execution of a computer so it is the only program running. The system verifies that no processors were in the middle of executing functions that will be modified by the patch. Ksplice modifies the beginning of changed functions so that they instead point to new, updated versions of those functions, and modifies data and structures in memory that need to be changed. Finally, Ksplice resumes each processor running where it left off."

KSplice totally rocks. I've seen it used in production to update the kernel on a machine hosting several VMs. The VMs never noticed a thing and of course the machine had a completely new kernel running in a matter of a few seconds. Don't doubt the ability of this tool to do exactly what it claims.

Interesting article, but Linux distros can do that anyway. Updating any Linux distro can be updated without the need to re-boot (maybe except the colonel). All the guy has done is use the Linux kernel to perform updates to system files whilst still running. But you don't ned their software just use any Linux distro.

No, the article is discussing making changes to the currently-running kernel without rebooting. It's true that you can swap out the entire kernel using the built-in kexec function, but this the equivalent of rebooting while bypassing the bootloader. Files must be saved and all userland processes (including X) are restarted.