Linus Torvalds (
Wed, 3 Jul 1996 19:25:19 +0300 (EET DST)

Ok, I just made the first patch available for 2.0, and it's a lot larger than
I'd like. A _lot_ larger. Oh, well..

Note that despite of the size of this patch, 2.0.0 seems to have been a
reasonably stable release, certainly no worse than expected. Keep the
reports coming in, and please tell of any problems with the new kernel.

Developers: I will NOT accept large patches any more. If you find bugs,
go for the one-liner obvious bugfix rather than a rewrite. Due to me
being abroad I couldn't work as closely with people on 2.0.1 patches as I
need to, but I'm back, and this kind of mega-patch will NOT happen again..

The patches do:
- ISDN subsystem updates.
- Sound driver patches. These are actually the largest part of the patches,
happily a lot of it is just comment changes and code re-organization.
- alpha patches: IRQ handling fixes and a updated QLogic ISP driver.
This kernel is stable on the AlphaStation 600. Also some signal
handler patches for POSIX.1 conformance.
- POSIX.1 conformance patches. This is part of my integration of the POSIX.1
certification patches from Unifix. Thanks to the Unifix people for
supplying patches, and especially to Heiko for some additional conformance
The POSIX patches are mainly details for signal handling and some
error returns. For example, the SIGCHLD semantics that have been
discussed on the kernel list have changed subtly (and "wait()" doesn't
mask out SIGCHLD signals any more, for example).
- loop and md driver updates
- baycom and de4x5 drivers updated
- SCSI tape update + various minor SCSI changes
- NFS client retry fixes
- load average counting should not result in the bogus loads of 1.0 on
idle machines..
- you shouldn't be able to crash the machine with the client/server
program on this list..
- multicast one-liner
- masquerading fixes

NOTE! The POSIX.1 patches in 2.0.1 is _not_ the complete set of patches,
and I still have some stuff to integrate yet. As such we can't say that
2.0.1 is POSIX.1 conformant yet, but the patches really aren't very large
(the original POSIX patches were less then 2000 lines, and around half of
them are integrated in one form or another).

Also note that the POSIX requirements for "rename()" meant that I had to
change the VFS interface to the low-level devices. That's rather unfortunate,
as it means that binary kernel modules will mostly _not_ work across the
2.0.0->2.0.1 upgrade. However, the new interface should be stable now, and
I'd rather have break it now rather than have problems with POSIX.1..

Anything else?