shithub: riscv

ref: 65bd53b9b89b0a84d0f16db4adc3f53fa6ab3a3a
dir: /sys/doc/port.ms/

.HTML "The Various Ports
.TL
The Various Ports
.PP
This document collects comments about the various
architectures supported by Plan 9.
The system tries to hide most of the differences between machines,
so the machines as seen by a Plan 9
user look different from how they are perceived through commercial software.
Also, because we are a small group, we couldn't do everything:
exploit every optimization, support every model,
drive every device.
This document records what we
.I have
done.
The first section discusses the compiler/assembler/loader suite for each machine.
The second talks about
the operating system implemented on each of the various
machines.
.SH
The Motorola MC68020 compiler
.PP
This is the oldest compiler of the bunch.  Relative to its
competitors\(emcommercial compilers for the same machine\(emit generates
quite good code.
It assumes at least a 68020 architecture: some of the addressing
modes it generates are not on the 68000 or 68010.
.PP
We also use this compiler for the 68040.  Except for a few
instructions and registers available only from assembly language,
the only user-visible difference between these machines is in
floating point.  Our 68020s all have 68881 or 68882 floating
point units attached, so to execute floating point programs we
depend on there being appropriate hardware.
Unfortunately, the 68040 is not quite so thorough in its implementation
of the IEEE 754 standard or in its provision of built-in instructions
for the
transcendental functions.  The latter was easy to get around: we
don't use them on the 68020 either, but we do have a library,
.CW -l68881 ,
that you can use if you need the performance (which can be
substantial:
.CW astro
runs twice as fast).
We don't use this library by default because we want to run the same
binaries on both machines and don't want to emulate
.CW FCOSH
in the operating system.
.PP
The problem with IEEE is nastier.  We didn't really want to deal
with gradual underflow and all that, especially since we had
half a dozen machines we'd need to do it on, so on the 68040
we implement non-trapping underflow as truncation to zero and
do nothing about denormalized numbers and not-a-numbers.
This means the 68020
and the 68040 are not precisely compatible.
.SH
The Motorola MC68000 compiler
.PP
This compiler is a stripped-down version of the MC68020 compiler
built for an abortive port to the Dragonball processor on the Palm Pilot.
It generates position-independent code whose overall quality is much
poorer than the code for the MC68020.
.SH
The MIPS compiler
.PP
This compiler generates code for the R2000, R3000, and R4000 machines configured
to be big-endians.  The compiler generates no R4000-specific instructions
although the assembler and loader support the new user-mode instructions.
There are options to generate code for little-endian machines.
Considering its speed, the Plan 9 compiler generates good code,
but the commercial
MIPS compiler with all the stops pulled out consistently beats it
by 20% or so, sometimes more.  Since ours compiles about 10 times
faster and we spend most of our time compiling anyway,
we are content with the tradeoff.
.PP
The compiler is solid: we've used it for several big projects and, of course,
all our applications run under it.
The behavior of floating-point programs is much like on the 68040:
the operating system emulates where necessary to get past non-trapping
underflow and overflow, but does not handle gradual underflow or
denormalized numbers or not-a-numbers.
.SH
The SPARC compiler
.PP
The SPARC compiler is also solid and fast, although we haven't
used it for a few years, due to a lack of current hardware.  We have seen it do
much better than GCC with all the optimizations, but on average
it is probably about the same.
.PP
We used to run some old SPARC machines with no multiply or divide instructions,
so the compiler
does not produce them by default.
Instead it calls internal subroutines.
A loader flag,
.CW -M ,
causes the instructions to be emitted.  The operating system has
trap code to emulate them if necessary, but the traps are slower than
emulating them in user mode.
In any modern lab, in which SPARCS have the instructions, it would be worth enabling the
.CW -M
flag by default.
.PP
The floating point story is the same as on the MIPS.
.SH
The Intel i386 compiler
.PP
This is really an
.I x 86
compiler, for
.I x >2.
It works only
if the machine is in 32-bit protected mode.
It is solid and generates tolerable code; it is our main compiler these days.
.PP
Floating point is well-behaved, but the compiler assumes i387-compatible
hardware to execute
the instructions.  With 387 hardware,
the system does the full IEEE 754 job, just like
the MC68881.  By default, the libraries don't use the 387 built-ins for
transcendentals.
If you want them,
build the code in
.CW /sys/src/libc/386/387 .
.SH
The Intel i960 compiler
.PP
This compiler was built as a weekend hack to let us get the Cyclone
boards running.  It has only been used to run one program\(emthe on-board
code in the Cyclone\(emand is therefore likely to be buggy.
There are a number of obvious optimizations to the code that have
never been attempted.
For example, the compiler does not support pipelining.
The code runs in little-endian mode.
.SH
The DEC Alpha compiler
.PP
The Alpha compiler is based on a port done by David Hogan while
studying at the Basser Department of Computer Science, University of Sydney.
It has been used to build a running version of the operating system, but has
not been stressed as much as some of the other compilers.
.PP
Although the Alpha is a 64-bit architecture, this compiler treats
.CW int s,
.CW long s
and pointers as 32 bits.  Access to the 64-bit operations is available through the
.CW vlong
type, as with the other architectures.
.PP
The compiler assumes that the target CPU supports the optional byte and
word memory operations (the ``BWX'' extension).
If you have an old system, you can generate code without using the extension
by passing the loader the
.CW -x
option.
.PP
There are a number of optimizations that the Alpha Architecture Handbook
recommends, but this compiler does not do.  In particular, there is currently
no support for the code alignment and code scheduling optimizations.
.PP
The compiler tries to conform to IEEE, but some Alpha CPUs do not implement
all of the rounding and trapping modes in silicon.  Fixing this problem requires
some software emulation code in the kernel; to date, this has not been attempted.
.SH
The PowerPC compiler
.PP
The PowerPC compiler supports the 32-bit PowerPC architecture only;
it does not support either the 64-bit extensions or the POWER compatibility instructions.
It has been used for production operating system work on the 603, 603e, 604e, 821, 823, and 860,
and experimental work on the 405, 440 and 450.
On the 8xx floating-point instructions must be emulated.
Instruction scheduling is not implemented; otherwise the code generated
is similar to that for the other load-store architectures.
The compiler makes little or no use of unusual PowerPC features such as the
counter register, several condition code registers, and multiply-accumulate
instructions, but they are sometimes
used by assembly language routines in the libraries.
.SH
The ARM compiler
.PP
The ARM compiler is fairly solid; it has been used for some production
operating system work including Inferno and the Plan 9 kernel
for the iPAQ, which uses a StrongArm SA1, and the Sheevaplug,
Guruplug, Dreamplug and others.
The compiler supports the ARMv4 architecture;
it does not support the Thumb instruction sets.
It has been used on ARM7500FE, ARM926 and Cortex-A8 processors
and the Strongarm SA1 core machines.
The compiler generates instructions for
ARM 7500 FPA floating-point coprocessor 1,
but probably should instead generate VFP 3+ instructions
for coprocessors 10 and 11.
.SH
The AMD 29000 compiler
.PP
This compiler was used to port an operating system to an AMD 29240 processor.
The project is long abandoned, but the compiler lives on.
.SH
The Carrera operating system
.PP
We used to have a number of MIPS R4400 PC-like devices called Carreras,
with custom-built frame buffers, that we used as terminals.
They're almost all decommissioned now, but we're including the source as a reference
in case someone wants to get another MIPS-based system running.
.SH
The IBM PC operating system
.PP
The PC version of Plan 9 can boot either from MS-DOS
or directly from a disk created by the
.CW format
command; see
.I prep (8).
Plan 9 runs in 32-bit mode\(emwhich requires a 386 or later model x86 processor\(emand
has an interrupt-driven I/O system, so it does not
use the BIOS (except for a small portion of the boot program and floppy boot block).
This helps performance but limits the set of I/O devices that it can support without
special code.
.PP
Plan 9 supports the ISA, EISA, and PCI buses as well as PCMCIA and PC card devices.
It is infeasible to list all the supported machines, because
the PC-clone marketplace is too volatile and there is
no guarantee that the machine you buy today will contain the
same components as the one you bought yesterday.
(For our lab, we buy components and assemble the machines
ourselves in an attempt to lessen this effect.)
Both IDE/ATA and SCSI disks are supported, and
there is support for large ATA drives.
CD-ROMs are supported two ways, either on the SCSI bus, or as ATA(PI) devices.
The SCSI adapter must be a member of the Mylex Multimaster (old Buslogic BT-*) series
or the Symbios 53C8XX series.
Supported Ethernet cards include the
AMD79C790,
3COM Etherlink III and 3C589 series,
Lucent Wavelan and compatibles,
NE2000,
WD8003,
WD8013,
SMC Elite and Elite Ultra,
Linksys Combo EthernetCard and EtherFast 10/100,
and a variety of controllers based on the
Intel i8255[789] and Digital (now Intel) 21114x chips.
We mostly use Etherlink III, i8255[789], and 21114x, so those drivers may be more robust.
There must be an explicit Plan 9 driver for peripherals;
it cannot use DOS or Windows drivers.
Also,
Plan 9 cannot exploit special hardware-related features that fall outside of the
IBM PC model,
such as power management,
unless architecture-dependent code is added to the kernel.
For more details see
.I plan9.ini (8).
.PP
Over the years,
Plan 9 has run on a number of VGA cards.
Recent changes to the graphics system have not been
tested on most of the older cards; some effort may be needed to get them working again.
In our lab, most of our machines use the ATI Mach64, S3 ViRGE, or S3 Savage chips,
so such devices are probably
the most reliable.
We also use a few Matrox and TNT cards.
The system requires a hardware cursor.
For more details see
.I vgadb (6)
and
.I vga (8).
The wiki
.CW http://plan9.bell-labs.com/wiki/plan9 ) (
contains the definitive list of cards that are known to work; see the ``supported PC hardware''
page.
.PP
For audio, Plan 9 supports the Sound Blaster 16 and compatibles.
(Note that audio doesn't work under Plan 9 with 8-bit Sound Blasters.)
There is also user-level support for USB audio devices; see 
.I usb (4).
.PP
Finally, it's important to have a three-button mouse with Plan 9.
The system currently works only with mice on the PS/2 port or USB.
Serial mouse support should return before long.
.PP
Once you have Plan 9 installed (see the wiki's installation document)
run the program
.CW ld
from DOS
or use a boot disk.  See
.I booting (8),
.I 9load (8),
and
.I prep (8)
for more information.
.SH
The Alpha PC operating system
.PP
Plan 9 runs on the Alpha PC 164.
The Alpha port has not been used as much as the others,
and should be considered a preliminary release.
.PP
The port uses the OSF/1 flavor
of PALcode, and should be booted from the SRM firmware (booting
from ARC is not supported).
Supported devices are a subset of the PC ones; currently
this includes DECchip 2114x-based ethernet cards, S3 VGA cards,
Sound Blaster 16-compatible audio, floppy drives, and ATA hard disks.
.PP
The system has to be booted via tftp.
See
.I booting (8)
for details.
.SH
The PowerPC operating system
.PP
We have a version of the system that runs on the PowerPC
on a home-grown machine called Viaduct.
The Viaduct minibrick is a small (12x9x3 cm) low-cost embedded
computer consisting of a 50Mhz MPC850, 16MB sdram, 2MB flash,
and two 10Mb Ethernet ports.  It is designed for home/SOHO
networking applications such as VPN, firewalls, NAT, etc.
.PP
The kernel has also been ported to the Motorola MTX embedded motherboard;
that port is included in the distribution.
The port only works with a 604e processor (the 603e is substantially different)
and at present only a single CPU is permitted.
.SH
The Compaq iPAQ operating system
.PP
Plan 9 was ported to Compaq's iPAQ Pocket PC,
which uses the StrongArm SA1 processor.
The model we have is a 3630; neighboring models also work.
The kernel can drive a PCMCIA sleeve with a WaveLAN card, but no other PCMCIA
devices have been ported yet.
.PP
The iPAQ runs
.CW rio
with a small keyboard application that allows Palm-style handwriting
input as well as typing with the stylus on a miniature keyboard.
.PP
Fco. J. Ballesteros
.CW nemo@plan9.escet.urjc.es ) (
added support for hibernation, but we haven't been able to
get that to work again in the new kernel; the code is there, however,
for volunteers to play with.
See the file
.CW /sys/src/9/bitsy/Booting101
for information about installing Plan 9 on the iPAQ.
.SH
The Marvell Kirkwood operating system
.PP
This is an ARM kernel for the ARM926EJ-S processor
and it emulates floating-point and
CAS (compare-and-swap) instructions.
It is known to run on the Sheevaplug, Guruplug, Dreamplug
and Openrd-client boards.
It is derived from a port of native Inferno to the Sheevaplug
by Salva Peir\f(Jpó\fP and Mechiel Lukkien.
There are many features of the Kirkwood system-on-a-chip
that it does not exploit.
There are currently drivers for up to two
Gigabit Ethernet interfaces,
USB and the console serial port;
we hope to add crypto acceleration, and a video driver for the Openrd-client.
.SH
The Marvell PXA168 operating system
.PP
This is an ARM kernel for the ARM-v5-architecture processor in the
Marvell PXA168 system-on-a-chip
and it emulates floating-point and
CAS (compare-and-swap) instructions.
It is known to run on the Guruplug Display.
There are many features of the system-on-a-chip
that it does not exploit.
There are currently drivers for
a Fast Ethernet interface,
.\" USB
and the console serial port;
we hope to add crypto acceleration, and a video driver.
.SH
The TI OMAP35 operating system
.PP
This is an ARM kernel for the Cortex-A8 processor
and it emulates pre-VFPv3 floating-point and
CAS (compare-and-swap) instructions.
It is known to run on the IGEPv2 board and the Gumstix Overo,
and might eventually run on the Beagleboard, once USB is working.
There are many features of the OMAP system-on-a-chip that it does not exploit.
Initially, there are drivers for the SMSC 9221 100Mb/s Ethernet
interface in the IGEPv2 and Overo,
and the console serial port;
we hope to add USB, flash memory and video drivers.
.
.
.ig
.SH
The file server
.PP
The file server runs on only a handful of distinct machines.
It is a stand-alone program, distantly related to the CPU server
code, that runs no user code: all it does is serve files on
network connections.
It supports only SCSI disks, which can be interleaved for
faster throughput.
A DOS file on
an IDE drive can hold the configuration information.
See
.I fsconfig (8)
for an explanation of how
to configure a file server.
.PP
To boot a file server, follow the directions for booting a CPU server
using the file name
.CW 9\f2machtype\fPfs
where
.I machtype
is
.CW pc ,
etc. as appropriate.
We are releasing only the PC version.
.SH
The IBM PC file server
.PP
Except for the restriction to SCSI disks,
the PC file server has the same hardware requirements as
the regular PC operating system.
However, only a subset of the supported SCSI (Adaptec 1542, Mylex Multimaster,
and Symbios 53C8XX) and Ethernet (Digital 2114x,
Intel 8255x, and 3Com) controllers
may be
used.
Any of the boot methods described in
.I 9load (8)
will work.
.PP
To boot any PC, the file
.CW 9load
must reside on a MS-DOS formatted floppy, IDE disk,
or SCSI disk.
However, PCs have no non-volatile RAM in which the
file server can store its configuration information, so the system
stores it in a file on an MS-DOS file system instead.
This file, however, cannot live on a SCSI disk, only a floppy or IDE.
(This restriction avoids a lot of duplicated interfaces in the
system.)
Thus the file server cannot be all-SCSI.
See
.I plan9.ini (8)
for details about the
.I nvr
variable and specifying the console device.
.SH
Backup
.PP
Our main file server is unlikely to be much like yours.
It is a PC with 128 megabytes
of cache memory, 56 gigabytes of SCSI magnetic
disk, and a Hewlett-Packard SureStore Optical 1200ex
magneto-optical jukebox, with 1.2 terabytes of storage.
This driver runs the SCSI standard jukebox protocol.
We also have a driver for a (non-standard)
SONY WDA-610
Writable Disk Auto Changer (WORM),
which stores almost 350 gigabytes of data.
.PP
The WORM is actually the prime storage; the SCSI disk is just
a cache to improve performance.
Early each morning the system constructs on WORM an image of
the entire system as it appears that day.  Our backup system
is therefore just a file server that lets
you look at yesterday's (or last year's) file system.
.PP
If you don't have a magneto-optical jukebox,
you might consider attaching a CD-R jukebox or even just
using a single WORM drive and managing the dumps a little less
automatically.  This is just a long way of saying that the
system as distributed has no explicit method of backup other
than through the WORM jukebox.
.PP
Not everyone can invest in such expensive hardware, however.
Although it wouldn't be as luxurious,
it would be possible to use
.I mkfs (8)
to build regular file system archives and use
.I scuzz (8)
to stream them to a SCSI 8mm tape drive.
.CW Mkext
could then extract them.
Another alternative is to use
.I dump9660
(see
.I mk9660 (8)),
which stores incremental backups on CD images
in the form of a dump hierarchy.
.PP
It is also possible to treat a regular disk, or even a part of a disk,
as a fake WORM, which can then be streamed to tape when it fills.
This is a bad idea for a production system but a good way to
learn about the WORM software.
Again, see
.I fsconfig (8)
for details.
..