So 4.14 continues to be a somewhat painful release, and I'm starting
to at least partly blame the fact that it's meant to be an LTS
release.
The last LTS release we had (4.9) resulted in one of the biggest
kernel releases we ever had because everybody wanted in; the 4.14
release doesn't seem to be as large, but it does seem to result in
some late work happening because people want to prep for 4.14, knowing
it will be LTS.
But who knows. Some of this may just be pure coincidence too. But I
already know of two more pull requests that are still pending that
will also probably want to be pushed into 4.14.
Anyway, on to the actual rc3 changes.. Most of them are the normal
small fixes, but a few things do stand out:
- some x86 FPU state handling fixes
- fixed some crypto problems in our internal key handling
- some smp/hotplug cleanups
and all of them are bigger than I would have wished for at this stage,
but all of them had fine reasons for going in now. They all had one
thing in common, in that they also came with cleanups in order to fix
the underlying problem (so often the actual commit that _fixes_ it is
pretty small, but there's a series of cleanups that makes that fix
possible).
The two issues that I know as potentially still pending are some of
the same kind: a writeback fix and some watchdog fixes, both with the
majority being cleanups in order to fix things.
Anyway, this all has the common thread that I'd have loved to get that
code during the merge window as "obviously good changes", but I'm not
thrilled to get it during the rc stages.
Oh well. Enough of the "Woe is me".
Things don't actually look *bad*. Yes, it's more changes than I would
have wished for at this stage, but at the same time none of it looks
like it's really fundamentally problematic for the 4.14 release. Most
of the x86 FPU state cleanups had already been around for a while just
because they were needed cleanup, for example, it's just that the bug
fixes made them get merged at a less than optimal time.
The various changes do end up making the diffstat look somewhat
unusual: driver fixes that usually dominate are just a quarter of the
haul this rc around, with arch fixes (almost all of which are x86) are
another quarter. The rest is core kernel (much of it the smp/hotplug
updates), security (the key handling changes) and tooling (much of it
perf, but also more selftests). Some fs fixes (btrfs and xfs, some
misc) accounts for the rest.
It's still early enough in the rc release that I don't know if this
will impact timing. Right now it still feels like we're fine with the
usual schedule (ie rc7 being the last rc), but we'll just have to see
how this release cycle continues.
Do go out and test, please.
Linus
