4.3BSD-tahoe release
Keith Bostic
4.3BSD for the tahoe processors. Attached is a brief summary
of the information mailed to all 4.2 and 4.3BSD licensees.
This mailing contains all necessary ordering information;
if you have not received it by July 5th, please contact our
distribution office at +1-415-642-7780.
Mike Karels
Kirk McKusick
Keith Bostic
========================
We are happy to send you information about our June
1988 revision of the Fourth Berkeley Software Distribution,
hereafter referred to as the ``4.3BSD tahoe'' distribution.
The purpose of this release is to provide 4.3BSD support for
the ``tahoe'' processor and to get feedback on some new
features and facilities that we expect to include in future
Berkeley releases. This distribution is not a standard
Berkeley release; it is an interim release intended for
testing and evaluation rather than production use by naive
users. The distribution is available to users with licenses
for the April 1986 4.3BSD release for the VAX.
What is the 4.3BSD Tahoe Release?
The distribution consists of one 6250 BPI or three 1600
BPI magnetic tapes, certain updated manual pages, and a copy
of ``Installing and Operating 4.3BSD on the Tahoe.'' The
magnetic tape(s) contain copies of source code and documen-
tation for both the VAX and the tahoe, and binaries for the
tahoe. On the three-tape 1600 BPI set, parts of the user-
contributed software are shipped as compressed archives to
save space.
The primary purpose of this release is to provide sup-
port for the ``tahoe'' processor, the CPU used by Computer
Consoles, Inc. (CCI Power 6/32, 6/32SX), and high end lines
of Harris (HCX-7 and HCX-9), Unisys (7000/40), and ICL (Clan
7). Support for this processor is derived from the 4.2BSD
system done by CCI. Support for new DEC equipment has also
been added, including support for the 8250 BI-based CPU and
the KDB-50 BI disk controller from Chris Torek, and the QVSS
and QDSS display drivers for the MicroVAX II, contributed by
Digital Equipment Corporation. We expect to provide support
for both the VAX and the tahoe processors in future
releases.
The major new software systems include the following
features:
+ The TCP and IP code is the same as that recently made
available via the ARPANET and Usenet. Several new
algorithms are used in TCP, in particular Van
Jacobson's slow start and dynamic window size selection
algorithms and Phil Karn's modification to the round-
trip timing algorithm. These changes increase
throughput and reduce congestion and retransmission.
Several fixes were made in the handling of IP options
and other gateway support.
+ The file system has been generalized to remove the lim-
its on the maximum number of inodes per cylinder group,
cylinders per cylinder group, and number of dis-
tinguished rotational positions. The kernel and file
system utilities operate normally on both new and old
format file systems; old kernels treat the new file
systems as read-only. This change allows better utili-
zation of newer disks with larger numbers of sectors
per cylinder.
+ The system has full support for disk labels that con-
tain disk geometry information and partition layout on
each disk. Labels are used on disks using the hp or
uda drivers on the VAX (hp or ra disks) and all sup-
ported disks on the tahoe. The utility to read and
write disk labels can be used with the system running
multiuser; the labels are used and updated as appropri-
ate by newfs, bad144, the kernel and the bootstrap pro-
grams. Basic file system parameters are stored in the
label so that fsck can locate alternate superblocks.
The filesystem and newfs use additional information
about the disk geometry that is now available, such as
track-to-track skew.
+ A new general-purpose dynamic memory allocator has been
written that can be used by all the kernel subsystems.
The design of this allocator takes advantage of known
memory usage patterns in the UNIX kernel, using a
hybrid strategy that is time-efficient for small allo-
cations and space-efficient for large allocations.
This allocator replaces most of the previous memory
allocation interfaces with a single easy-to-program
interface, results in more efficient use of global
memory by eliminating partitioned and specialized
memory pools, and is quick enough that no performance
loss is observed relative to the previous implementa-
tions. Most of the kernel memory allocation is now
done using this allocator.
+ The source code has been reorganized to ease support of
multiple machine types.
+ The Olson/Harris/Elz timezone implementation has been
added to the system.
+ Numerous bug fixes and enhancements have been made
throughout the system, including new versions of the
nameserver named and the routing daemon routed.
Several years' backlog of bug fixes have been applied.
+ We have started a process of identifying the code in
the 4.3BSD distribution that is not derived from AT&T
code. The copyrights in this code have been changed to
indicate that it may be freely distributed if the copy-
right notice is retained and that due credit for its
origin is given to The Regents of the University of
California. Over 1000 files have been identified in
this distribution.