Re: Kernel for Spectre and Meltdown
- Date: Mon, 29 Jan 2018 11:13:50 -0500
- From: Greg Wooledge <wooledg@xxxxxxxxxxx>
- Subject: Re: Kernel for Spectre and Meltdown
On Mon, Jan 29, 2018 at 10:47:57AM -0500, rhkramer@xxxxxxxxx wrote:
> Again, checking / confirming my understanding, if you download a kernel image
> (which is normal for me), there is no need for me to have any version of GCC
> as the image is pre-compiled.
> On the other hand, if I download kernel source, I would need GCC, and a
> version that is sufficient for the code.
All correct. (Plus several additional development packages, not just gcc.)
> I have only compiled the kernel a few times, all a long time ago (12 to 15
> years?), on the advice of members of my local LUG, and maybe as a learning
> experience. It is far from necessary for most of us. (Some members of the
> LUG seemed to think it was imperative, and maybe it is for older smaller
> machines or maybe to squeeze the very last little bit of efficiency out of the
Before Linux 2.6 (or thereabouts), compiling one's own kernel was a much
more common event. Certainly it wasn't required for ordinary use, but
hardware was much less powerful back then, so a leaner kernel tuned
exactly for the target system was sometimes desirable.
With Linux 2.6, things started to change. The Linux developers
acknowledged that the source code they were releasing wasn't really
"stable" in the sense that end users expected; the distributions (Red Hat,
Debian, et al.) were the ones doing the final stabilization and patching.
Also, the number of configuration questions one had to answer before
compiling a kernel started to balloon out of control.
This was also the time when initramfs/initrd images started to be used,
at least by Debian. My understanding of this is only partial, but it
seems that the initrd allows some adjustments of the kernel for the
target system (installation of driver modules, firmware) which may
previously have required a reconfiguration and recompilation.
Hardware was also becoming more powerful, as one would expect. More RAM
meant less pressure to produce minimalist kernel images.
All of these things put together meant that for most users, Debian's
kernel images were good enough that they didn't feel a need to build
their own kernels.