The kernel is the gooey core of the operating system. If you think of an operating system as a Tootsie pop with layers of sugary shell, it’s down at the lowest level managing the basic things that the OS needs to work, and takes more than a few licks to get to.
More than likely, your computer completely crashes out way less than it used to—or at least, way less than Windows 95. There’s a few reasons for that. A major reason, says Maximum PC Editor Maximus Will Smith, is that Apple and Microsoft have spent a lot of time moving stuff that used to run at really low level, deep in the guts of the OS, up a few layers into the user space, so an application error that would’ve crashed a whole system by borking something at the kernel level just results in an annoying program-level hang up. More simply put, OSes have been getting better at isolating and containing problems, so a bad app commits suicide, rather than suicide bombing your whole computer.
This is part of the reason drivers—the software that lets a piece of hardware, like a video card talk to your OS and other programs—are a bigger source of full-on crashes than standard apps nowadays when it comes to modern operating systems. By their nature, drivers have pretty deep access, and the kernel sits smack in the middle of that, says Flores. So if something goes wrong with a driver, it can result in some bigtime ka-blooey. Theoretically, signed (i.e., vetted) drivers help avoid some of the problems, but take graphics drivers, which were a huge problem with Vista crashes at launch: Flores says that “some of the most complex programming in the world is done by graphics device driver software writers,” and when Microsoft changed to a new driver model with Vista, it was a whole new set of rules to play by. (Obviously, stuff got screwed up.)
Another reason things crash less now is that Apple and Microsoft have metric tons of data about what causes crashes with more advanced telemetry—information the OS sends home, like system configurations, what a program was doing, the state of memory, and other in-depth details about a crash—than ever. With that information, they can do more to prevent crashes, obviously, so don’t be (too) afraid to click “send” on that error message.
In Windows 7, for instance, there’s a new fault tolerance heap—basically, a heap’s a special area of memory that’s fairly low-level—which could get corrupted easily in past versions of Windows. In Windows 7, it can tell when a crash in the heap is about to happen and take steps to isolate an application from everything else.
0 Response to “Linux Memory for Crash Kernel”