1 |
BASILISK II TECHNICAL MANUAL |
2 |
============================ |
3 |
|
4 |
0. Table of Contents |
5 |
-------------------- |
6 |
|
7 |
1. Introduction |
8 |
2. Modes of operation |
9 |
3. Memory access |
10 |
4. Calling native routines from 68k mode and vice-versa |
11 |
5. Interrupts |
12 |
6. Parts of Basilisk II |
13 |
7. Porting Basilisk II |
14 |
|
15 |
1. Introduction |
16 |
--------------- |
17 |
|
18 |
Basilisk II can emulate two kind of Macs, depending on the ROM being used: |
19 |
|
20 |
1. A Mac Classic |
21 |
2. A Mac II series computer ("Mac II series" here means all 68020/30/40 |
22 |
based Macs with 32-bit clean ROMs (this excludes the original Mac II, |
23 |
the IIx/IIcx and the SE/030), except PowerBooks; in the following, |
24 |
"Mac II" is used as an abbreviation of "Mac II series computer", as |
25 |
defined above) |
26 |
|
27 |
More precisely spoken, MacOS under Basilisk II behaves like on a Mac Classic |
28 |
or Mac II because, apart from the CPU, the RAM and the ROM, absolutely no Mac |
29 |
hardware is emulated. Rather, Basilisk II provides replacements (usually in |
30 |
the form of MacOS drivers) for the parts of MacOS that access hardware. As |
31 |
there are practically no Mac applications that access hardware directly (this |
32 |
is also due to the fact that the hardware of different Mac models is sometimes |
33 |
as different as, say, the hardware of an Atari ST and an Amiga 500), both the |
34 |
compatibility and speed of this approach are very high. |
35 |
|
36 |
2. Modes of operation |
37 |
--------------------- |
38 |
|
39 |
Basilisk II is designed to run on many different hardware platforms and on |
40 |
many different operating systems. To provide optimal performance under all |
41 |
environments, it can run in three different modes, depending on the features |
42 |
of the underlying environment (the modes are selected with the REAL_ADDRESSING |
43 |
and EMULATED_68K defines in "sysdeps.h"): |
44 |
|
45 |
1. Emulated CPU, "virtual" addressing (EMULATED_68K = 1, REAL_ADDRESSING = 0): |
46 |
This mode is designed for non-68k or little-endian systems or systems that |
47 |
don't allow accessing RAM at 0x0000..0x1fff. This is also the only mode that |
48 |
allows 24-bit addressing, and thus the only mode that allows Mac Classic |
49 |
emulation. The 68k processor is emulated with the UAE CPU engine and two |
50 |
memory areas are allocated for Mac RAM and ROM. The memory map seen by the |
51 |
emulated CPU and the host CPU are different. Mac RAM starts at address 0 |
52 |
for the emulated 68k, but it may start at a different address for the host |
53 |
CPU. All memory accesses of the CPU emulation go through memory access |
54 |
functions (do_get_mem_long() etc.) that translate addresses. This slows |
55 |
down the emulator, of course. |
56 |
|
57 |
2. Emulated CPU, "real" addressing (EMULATED_68K = 1, REAL_ADDRESSING = 0): |
58 |
This mode is intended for big-endian non-68k systems that do allow access to |
59 |
RAM at 0x0000..0x1fff. As in the virtual addressing mode, the 68k processor |
60 |
is emulated with the UAE CPU engine and two areas are set up for RAM and ROM |
61 |
but the emulated CPU lives in the same address space as the host CPU. |
62 |
This means that if something is located at a certain address for the 68k, |
63 |
it is located at the exact same address for the host CPU. Mac addresses |
64 |
and host addresses are the same. The memory accesses of the CPU emulation |
65 |
still go through access functions but the address translation is no longer |
66 |
needed, and if the host CPU uses big-endian data layout and can handle |
67 |
unaligned accesses like the 68k, the memory access functions are replaced |
68 |
by direct, inlined memory accesses, making for the fastest possible speed |
69 |
of the emulator. |
70 |
A usual consequence of the real addressing mode is that the Mac RAM doesn't |
71 |
any longer begin at address 0 for the Mac and that the Mac ROM also is not |
72 |
located where it usually is on a real Mac. But as the Mac ROM is relocatable |
73 |
and the available RAM is defined for MacOS by the start of the system zone |
74 |
(which is relocated to the start of the allocated RAM area) and the MemTop |
75 |
variable (which is also set correctly) this is not a problem. There is, |
76 |
however, one RAM area that must lie in a certain address range. This area |
77 |
contains the Mac "Low Memory Globals" which (on a Mac II) are located at |
78 |
0x0000..0x1fff and which cannot be moved to a different address range. |
79 |
The Low Memory Globals constitute of many important MacOS and application |
80 |
global variables (e.g. the above mentioned "MemTop" variable which is |
81 |
located at 0x0108). For the real addressing mode to work, the host CPU |
82 |
needs access to 0x0000..0x1fff. Under most operating systems, this is a |
83 |
big problem. On some systems, patches (like PrepareEmul on the Amiga or |
84 |
the sheep_driver under BeOS) can be installed to "open up" this area. On |
85 |
other systems, it might be possible to use access exception handlers to |
86 |
emulate accesses to this area. But if the Low Memory Globals area cannot |
87 |
be made available, using the real addressing mode is not possible. |
88 |
|
89 |
3. Native CPU (EMULATED_68K = 0, this also requires REAL_ADDRESSING = 1) |
90 |
This mode is designed for systems that use a 68k (68020 or better) processor |
91 |
as host CPU and is the technically most difficult mode to handle. The Mac |
92 |
CPU is no longer emulated (the UAE CPU emulation is not needed) but MacOS |
93 |
and Mac applications run natively on the existing 68k CPU. This means that |
94 |
the emulator has its maximum possible speed (very close to that of a real |
95 |
Mac with the same CPU). As there is no control over the memory accesses of |
96 |
the CPU, real addressing mode is implied, and so the Low Memory area must |
97 |
be accessible (an MMU might be used to set up different address spaces for |
98 |
the Mac and the host, but this is not implemented in Basilisk II). The |
99 |
native CPU mode has some possible pitfalls that might make its |
100 |
implementation difficult on some systems: |
101 |
a) Implied real addressing (this also means that Mac programs that go out |
102 |
of control can crash the emulator or the whole system) |
103 |
b) MacOS and Mac applications assume that they always run in supervisor |
104 |
mode (more precisely, they assume that they can safely use certain |
105 |
priviledged instructions, mostly for interrupt control). So either |
106 |
the whole emulator has to be run in supervisor mode (which usually is |
107 |
not possible on multitasking systems) or priviledged instructions have |
108 |
to be trapped and emulated. The Amiga version of Basilisk II uses the |
109 |
latter approach (it is possible to run supervisor mode tasks under |
110 |
the AmigaOS multitasking kernel (ShapeShifter does this) but it |
111 |
requires modifying the task switcher and makes the emulator more |
112 |
unstable). |
113 |
c) On multitasking systems, interrupts can usually not be handled as on |
114 |
a real Mac (or with the UAE CPU). The interrupt levels of the host |
115 |
will not be the same as on a Mac, and the operating systems might not |
116 |
allow installing hardware interrupt handlers or the interrupt handlers |
117 |
might have different stack frames and run-time environments than 68k |
118 |
hardware interrupts. The usual solution is to use some sort of software |
119 |
interrupts or signals to interrupt the main emulation process and to |
120 |
manually call the Mac 68k interrupt handler with a faked stack frame. |
121 |
d) 68060 systems are a small problem because there is no Mac that ever |
122 |
used the 68060 and MacOS doesn't know about this processor. Basilisk II |
123 |
reports the 68060 as being a 68040 to the MacOS and patches some places |
124 |
where MacOS makes use of certain 68040-specific features such as the |
125 |
FPU state frame layout or the PTEST instruction. Also, Basilisk II |
126 |
requires that all of the Motorola support software for the 68060 to |
127 |
emulate missing FPU and integer instructions and addressing modes is |
128 |
provided by the host operating system (this also applies to the 68040). |
129 |
e) The "EMUL_OP" mechanism described below requires the interception and |
130 |
handling of certain emulator-defined instructions. |
131 |
|
132 |
3. Memory access |
133 |
---------------- |
134 |
|
135 |
There is often a need to access Mac RAM and ROM inside the emulator. As |
136 |
Basilisk II may run in "real" or "virtual" addressing mode on many different |
137 |
architectures, big-endian or little-endian, certain platform-independent |
138 |
data types and functions are provided: |
139 |
|
140 |
a) "sysdeps.h" defines the types int8, uint8, int16, uint16, int32 and uint32 |
141 |
for numeric quantities of a certain signedness and bit length |
142 |
b) "cpu_emulation.h" defines the ReadMacInt*() and WriteMacInt*() functions |
143 |
which should always be used to read from or write to Mac RAM or ROM |
144 |
c) "cpu_emulation.h" also defines the Mac2HostAddr() function that translates |
145 |
a Mac memory address to a (uint8 *) in host address space. This allows you |
146 |
to access larger chunks of Mac memory directly, without going through the |
147 |
read/write functions for every access. But doing so you have to perform |
148 |
any needed endianess conversion of the data yourself by using the ntohs() |
149 |
etc. macros which are available on most systems or defined in "sysdeps.h". |
150 |
|
151 |
4. Calling native routines from 68k mode and vice-versa |
152 |
------------------------------------------------------- |
153 |
|
154 |
An emulator like Basilisk II requires two kinds of cross-platform function |
155 |
calls: |
156 |
|
157 |
a) Calling a native routine from the Mac 68k context |
158 |
b) Calling a Mac 68k routine from the native context |
159 |
|
160 |
Situation a) arises in nearly all Basilisk drivers and system patches while |
161 |
case b) is needed for the invocation of Mac call-back or interrupt routines. |
162 |
Basilisk II tries to solve both problems in a way that provides the same |
163 |
interface whether it is running on a 68k or a non-68k system. |
164 |
|
165 |
4.1. The EMUL_OP mechanism |
166 |
-------------------------- |
167 |
|
168 |
Calling native routines from the Mac 68k context requires breaking out of the |
169 |
68k emulator or interrupting the current instruction flow and is done via |
170 |
unimplemented 68k opcodes (called "EMUL_OP" opcodes). Basilisk II uses opcodes |
171 |
of the form 0x71xx (these are invalid MOVEQ opcodes) which are defined in |
172 |
"emul_op.h". When such an opcode is encountered, whether by the emulated CPU |
173 |
or a real 68k, the execution is interrupted, all CPU registers saved and the |
174 |
EmulOp() function from "emul_op.cpp" is called. EmulOp() decides which opcode |
175 |
caused the interrupt and performs the required actions (mostly by calling other |
176 |
emulator routines). The EMUL_OP handler routines have access to nearly all of |
177 |
the 68k user mode registers (exceptions being the PC, A7 and SR). So the |
178 |
EMUL_OP opcodes can be thought of as extensions to the 68k instruction set. |
179 |
Some of these opcodes are used to implement ROM or resource patches because |
180 |
they only occupy 2 bytes and there is sometimes not more room for a patch. |
181 |
|
182 |
4.2. Execute68k() |
183 |
----------------- |
184 |
|
185 |
"cpu_emulation.h" declares the functions Execute68k() and Execute68kTrap() to |
186 |
call Mac 68k routines or MacOS system traps from inside an EMUL_OP handler |
187 |
routine. They allow setting all 68k user mode registers (except PC and SR) |
188 |
before the call and examining all register contents after the call has |
189 |
returned. EMUL_OP and Execute68k() may be nested, i.e. a routine called with |
190 |
Execute68k() may contain EMUL_OP opcodes and the EMUL_OP handlers may in turn |
191 |
call Execute68k() again. |
192 |
|
193 |
5. Interrupts |
194 |
------------- |
195 |
|
196 |
Various parts of Basilisk II (such as the Time Manager and the serial driver) |
197 |
need an interrupt facility to trigger asynchronous events. The MacOS uses |
198 |
different 68k interrupt levels for different events, but for simplicity |
199 |
Basilisk II only uses level 1 and does it's own interrupt dispatching. The |
200 |
"InterruptFlags" contains a bit mask of the pending interrupts. These are the |
201 |
currently defined interrupt sources (see main.h): |
202 |
|
203 |
INTFLAG_60HZ - MacOS 60Hz interrupt (unlike a real Mac, we also handle |
204 |
VBL interrupts, ADB events and the Time Manager here) |
205 |
INTFLAG_SERIAL - Interrupt for serial driver I/O completion |
206 |
INTFLAG_ETHER - Interrupt for Ethernet driver I/O completion and packet |
207 |
reception |
208 |
INTFLAG_AUDIO - Interrupt for audio "next block" requests |
209 |
INTFLAG_TIMER - Reserved for a future implementation of a more precise |
210 |
Time Manager (currently not used) |
211 |
|
212 |
An interrupt is triggered by calling SetInterruptFlag() with the desired |
213 |
interrupt flag constant and then TriggerInterrupt(). When the UAE 68k |
214 |
emulator is used, this will signal a hardware interrupt to the emulated 680x0. |
215 |
On a native 68k machine, some other method for interrupting the MacOS thread |
216 |
has to be used (e.g. on AmigaOS, a signal exception is used). Care has to be |
217 |
taken because with the UAE CPU, the interrupt will only occur when Basilisk II |
218 |
is executing MacOS code while on a native 68k machine, the interrupt could |
219 |
occur at any time (e.g. inside an EMUL_OP handler routine). In any case, the |
220 |
MacOS thread will eventually end up in the level 1 interrupt handler which |
221 |
contains an M68K_EMUL_OP_IRQ opcode. The opcode handler in emul_op.cpp will |
222 |
then look at InterruptFlags and decide which routines to call. |
223 |
|
224 |
6. Parts of Basilisk II |
225 |
----------------------- |
226 |
|
227 |
The conception of Basilisk II is quite modular and consists of many parts |
228 |
which are relatively independent from each other: |
229 |
|
230 |
- UAE CPU engine ("uae_cpu/*", not needed on all systems) |
231 |
- ROM patches ("rom_patches.cpp", "slot_rom.cpp" and "emul_op.cpp") |
232 |
- resource patches ("rsrc_patches.cpp" and "emul_op.cpp") |
233 |
- PRAM Utilities replacement ("xpram.cpp") |
234 |
- ADB Manager replacement ("adb.cpp") |
235 |
- Time Manager replacement ("timer.cpp") |
236 |
- SCSI Manager replacement ("scsi.cpp") |
237 |
- video driver ("video.cpp") |
238 |
- audio component ("audio.cpp") |
239 |
- floppy driver ("sony.cpp") |
240 |
- disk driver ("disk.cpp") |
241 |
- CD-ROM driver ("cdrom.cpp") |
242 |
- serial drivers ("serial.cpp") |
243 |
- Ethernet driver ("ether.cpp") |
244 |
- system-dependant device access ("sys_*.cpp") |
245 |
- user interface strings ("user_strings.cpp") |
246 |
- preferences management ("prefs.cpp" and "prefs_editor_*.cpp") |
247 |
|
248 |
Most modules consist of a platform-independant part (such as video.cpp) and a |
249 |
platform-dependant part (such as video_beos.cpp). The "dummy" directory |
250 |
contains generic "do-nothing" versions of some of the platform-dependant |
251 |
parts to aid in testing and porting. |
252 |
|
253 |
6.1. UAE CPU engine |
254 |
------------------- |
255 |
|
256 |
All files relating to the UAE 680x0 emulation are kept in the "uae_cpu" |
257 |
directory. The "cpu_emulation.h" header file defines the link between the |
258 |
UAE CPU and the rest of Basilisk II, and "basilisk_glue.cpp" implements the |
259 |
link. It should be possible to replace the UAE CPU with a different 680x0 |
260 |
emulation by creating a new "xxx_cpu" directory with an appropriate |
261 |
"cpu_emulation.h" header file (for the inlined memory access functions) and |
262 |
writing glue code between the functions declared in "cpu_emulation.h" and |
263 |
those provided by the 680x0 emulator. |
264 |
|
265 |
6.2. ROM and resource patches |
266 |
----------------------------- |
267 |
|
268 |
As described above, instead of emulating custom Mac hardware, Basilisk II |
269 |
provides replacements for certain parts of MacOS to redirect input, output |
270 |
and system control functions of the Mac hardware to the underlying operating |
271 |
systems. This is done by applying patches to the Mac ROM ("ROM patches") and |
272 |
the MacOS system file ("resource patches", because nearly all system software |
273 |
is contained in MacOS resources). Unless resources are written back to disk, |
274 |
the system file patches are not permanent (it would cause many problems if |
275 |
they were permanent, because some of the patches vary with different |
276 |
versions of Basilisk II or even every time the emulator is launched). |
277 |
|
278 |
ROM patches are contained in "rom_patches.cpp" and resource patches are |
279 |
contained in "rsrc_patches.cpp". The ROM patches are far more numerous because |
280 |
nearly all the software needed to run MacOS is contained in the Mac ROM (the |
281 |
system file itself consists mainly of ROM patches, in addition to pictures and |
282 |
text). One part of the ROM patches involves the construction of a NuBus slot |
283 |
declaration ROM (in "slot_rom.cpp") which is used to add the video and Ethernet |
284 |
drivers. Apart from the CPU emulation, the ROM and resource patches contain |
285 |
most of the "logic" of the emulator. |
286 |
|
287 |
6.3. PRAM Utilities |
288 |
------------------- |
289 |
|
290 |
MacOS stores certain nonvolatile system parameters in a 256 byte battery |
291 |
backed-up CMOS RAM area called "Parameter RAM", "PRAM" or "XPRAM" (which refers |
292 |
to "Extended PRAM" because the earliest Mac models only had 20 bytes of PRAM). |
293 |
Basilisk II patches the ClkNoMem() MacOS trap which is used to access the XPRAM |
294 |
(apart from some routines which are only used early during system startup) |
295 |
and the real-time clock. The XPRAM is emulated in a 256 byte array which is |
296 |
saved to disk to preserve the contents for the next time Basilisk is launched. |
297 |
|
298 |
6.4. ADB Manager |
299 |
---------------- |
300 |
|
301 |
For emulating a mouse and a keyboard, Basilisk II patches the ADBOp() MacOS |
302 |
trap. Platform-dependant code reports mouse and keyboard events with the |
303 |
ADBMouseDown() etc. functions which are queued and sent to MacOS inside the |
304 |
ADBInterrupt() function (which is called as a part of the 60Hz interrupt |
305 |
handler) by calling the ADB mouse and keyboard handlers with Execute68k(). |
306 |
|
307 |
6.5. Time Manager |
308 |
----------------- |
309 |
|
310 |
Basilisk II completely replaces the Time Manager (InsTime(), RmvTime(), |
311 |
PrimeTime() and Microseconds() traps). A "TMDesc" structure is associated with |
312 |
each Time Manager task, that contains additional data. The tasks are executed |
313 |
in the TimerInterrupt() function which is currently called inside the 60Hz |
314 |
interrupt handler, thus limiting the resolution of the Time Manager to 16.6ms. |
315 |
|
316 |
6.6. SCSI Manager |
317 |
----------------- |
318 |
|
319 |
The (old-style) SCSI Manager is also completely replaced and the MacOS |
320 |
SCSIDispatch() trap redirected to the routines in "scsi.cpp". Under the MacOS, |
321 |
programs have to issue multiple calls for all the different phases of a |
322 |
SCSI bus interaction (arbitration, selection, command transfer etc.). |
323 |
Basilisk II maps this API to an atomic API which is used by most modern |
324 |
operating systems. All action is deferred until the call to SCSIComplete(). |
325 |
The TIB (Transfer Instruction Block) mini-programs used by the MacOS are |
326 |
translated into a scatter/gather list of data blocks. Operating systems that |
327 |
don't support scatter/gather SCSI I/O will have to use buffering if more than |
328 |
one data block is being transmitted. Some more advanced (but rarely used) |
329 |
aspects of the SCSI Manager (like messaging and compare operations) are not |
330 |
emulated. |
331 |
|
332 |
6.7. Video driver |
333 |
----------------- |
334 |
|
335 |
The NuBus slot declaration ROM constructed in "slot_rom.cpp" contains a driver |
336 |
definition for a video driver. The Control and Status calls of this driver are |
337 |
implemented in "video.cpp". Run-time video mode and depth switching are |
338 |
currently not supported. |
339 |
|
340 |
The host-side initialization of the video system is done in VideoInit(). |
341 |
This function must provide access to a frame buffer for MacOS and supply |
342 |
its address, resolution and color depth in a video_desc structure (there |
343 |
is currently only one video_desc structure, called VideoMonitor; this is |
344 |
going to change once multiple displays are supported). In real addressing |
345 |
mode, this frame buffer must be in a MacOS compatible layout (big-endian |
346 |
and 1, 2, 4 or 8 bits paletted chunky pixels, RGB 5:5:5 or xRGB 8:8:8:8). |
347 |
In virtual addressing mode, the frame buffer is located at address |
348 |
0xa0000000 on the Mac side and you have to supply the host address, size |
349 |
and layout (BasiliskII will do an automatic pixel format conversion in |
350 |
virtual addressing mode) in the variables MacFrameBaseHost, MacFrameSize |
351 |
and MacFrameLayout. |
352 |
|
353 |
6.8. Audio component |
354 |
-------------------- |
355 |
|
356 |
Basilisk II provides a Sound Manager 3.x audio component for sound output. |
357 |
Earlier Sound Manager versions that don't use components but 'snth' resources |
358 |
are not supported. Nearly all component functions are implemented in |
359 |
"audio.cpp". The system-dependant modules ("audio_*.cpp") handle the |
360 |
initialization of the audio hardware/driver, volume controls, and the actual |
361 |
sound output. |
362 |
|
363 |
The mechanism of sound output varies depending on the platform but usually |
364 |
there will be one "streaming thread" (either a thread that continuously writes |
365 |
data buffers to the audio device or a callback function that provides the |
366 |
next data buffer) that reads blocks of sound data from the MacOS Sound Manager |
367 |
and writes them to the audio device. To request the next data buffer, the |
368 |
streaming thread triggers the INTFLAG_AUDIO interrupt which will cause the |
369 |
MacOS thread to eventually call AudioInterrupt(). Inside AudioInterrupt(), |
370 |
the next data block will be read and the streaming thread is signalled that |
371 |
new audio data is available. |
372 |
|
373 |
6.9. Floppy, disk and CD-ROM drivers |
374 |
------------------------------------ |
375 |
|
376 |
Basilisk II contains three MacOS drivers that implement floppy, disk and CD-ROM |
377 |
access ("sony.cpp", "disk.cpp" and "cdrom.cpp"). They rely heavily on the |
378 |
functionality provided by the "sys_*.cpp" module. BTW, the name ".Sony" of the |
379 |
MacOS floppy driver comes from the fact that the 3.5" floppy drive in the first |
380 |
Mac models was custom-built for Apple by Sony (this was one of the first |
381 |
applications of the 3.5" floppy format which was also invented by Sony). |
382 |
|
383 |
6.10. Serial drivers |
384 |
-------------------- |
385 |
|
386 |
Similar to the disk drivers, Basilisk II contains replacement serial drivers |
387 |
for the emulation of Mac modem and printer ports. To avoid duplicating code, |
388 |
both ports are handled by the same set of routines. The SerialPrime() etc. |
389 |
functions are mostly wrappers that determine which port is being accessed. |
390 |
All the real work is done by the "SERDPort" class which is subclassed by the |
391 |
platform-dependant code. There are two instances (for port A and B) of the |
392 |
subclasses. |
393 |
|
394 |
Unlike the disk drivers, the serial driver must be able to handle asynchronous |
395 |
operations. Calls to SerialPrime() will usually not actually transmit or receive |
396 |
data but delegate the action to an independant thread. SerialPrime() then |
397 |
returns "1" to indicate that the I/O operation is not yet completed. The |
398 |
completion of the I/O request is signalled by calling the MacOS trap "IODone". |
399 |
However, this can't be done by the I/O thread because it's not in the right |
400 |
run-time environment to call MacOS functions. Therefore it will trigger the |
401 |
INTFLAG_SERIAL interrupt which causes the MacOS thread to eventually call |
402 |
SerialInterrupt(). SerialInterrupt(), in turn, will not call IODone either but |
403 |
install a Deferred Task to do the job. The Deferred Task will be called by |
404 |
MacOS when it returns to interrupt level 0. This mechanism sounds complicated |
405 |
but is necessary to ensure stable operation of the serial driver. |
406 |
|
407 |
6.11. Ethernet driver |
408 |
--------------------- |
409 |
|
410 |
A driver for Ethernet networking is also contained in the NuBus slot ROM. |
411 |
Only one ethernet card can be handled by Basilisk II. For Ethernet to work, |
412 |
Basilisk II must be able to send and receive raw Ethernet packets, including |
413 |
the 14-byte header (destination and source address and type/length field), |
414 |
but not including the 4-byte CRC. This may not be possible on all platforms |
415 |
or it may require writing special net drivers or add-ons or running with |
416 |
superuser priviledges to get access to the raw packets. |
417 |
|
418 |
Writing packets works as in the serial drivers. The ether_write() routine may |
419 |
choose to send the packet immediately (e.g. under BeOS) and return noErr or to |
420 |
delegate the sending to a separate thread (e.g. under AmigaOS) and return "1" to |
421 |
indicate that the operation is still in progress. For the latter case, a |
422 |
Deferred Task structure is provided in the ether_data area to call IODone from |
423 |
EtherInterrupt() when the packet write is complete (see above for a description |
424 |
of the mechanism). |
425 |
|
426 |
Packet reception is a different story. First of all, there are two methods |
427 |
provided by the MacOS Ethernet driver API to read packets, one of which (ERead/ |
428 |
ERdCancel) is not supported by Basilisk II. Basilisk II only supports reading |
429 |
packets by attaching protocol handlers. This shouldn't be a problem because |
430 |
the only network code I've seen so far that uses ERead is some Apple sample |
431 |
code. AppleTalk, MacTCP, MacIPX, OpenTransport etc. all use protocol handlers. |
432 |
By attaching a protocol handler, the user of the Ethernet driver supplies a |
433 |
handler routine that should be called by the driver upon reception of Ethernet |
434 |
packets of a certain type. 802.2 packets (type/length field of 0..1500 in the |
435 |
packet header) are a bit special: there can be only one protocol handler attached |
436 |
for 802.2 packets (by specifying a packet type of "0"). The MacOS LAP Manager |
437 |
will attach a 802.2 handler upon startup and handle the distribution of 802.2 |
438 |
packets to sub-protocol handlers, but the Basilisk II Ethernet driver is not |
439 |
concerned with this. |
440 |
|
441 |
When the driver receives a packet, it has to look up the protocol handler |
442 |
installed for the respective packet type (if any has been installed at all) |
443 |
and call the packet handler routine. This must be done with Execute68k() from |
444 |
the MacOS thread, so an interrupt (INTFLAG_ETHER) is triggered upon reception |
445 |
of a packet so the EtherInterrupt() routine can call the protocol handler. |
446 |
Before calling the handler, the Ethernet packet header has to be copied to |
447 |
MacOS RAM (the "ed_RHA" field of the ether_data structure is provided for this). |
448 |
The protocol handler will read the packet data by means of the ReadPacket/ReadRest |
449 |
routines supplied by the Ethernet driver. Both routines will eventually end up |
450 |
in EtherReadPacket() which copies the data to Mac address space. EtherReadPacket() |
451 |
requires the host address and length of the packet to be loaded to a0 and d1 |
452 |
before calling the protocol handler. |
453 |
|
454 |
Does this sound complicated? You are probably right. Here is another description |
455 |
of what happens upon reception of a packet: |
456 |
1. Ethernet card receives packet and notifies some platform-dependant entity |
457 |
inside Basilisk II |
458 |
2. This entity will store the packet in some safe place and trigger the |
459 |
INTFLAG_ETHER interrupt |
460 |
3. The MacOS thread will execute the EtherInterrupt() routine and look for |
461 |
received packets |
462 |
4. If a packet was received of a type to which a protocol handler had been |
463 |
attached, the packet header is copied to ed_RHA, a0/d1 are loaded with |
464 |
the host address and length of the packet data, a3 is loaded with the |
465 |
Mac address of the first byte behing ed_RHA and a4 is loaded with the |
466 |
Mac address of the ed_ReadPacket code inside ether_data, and the protocol |
467 |
handler is called with Execute68k() |
468 |
5. The protocol handler will eventually try to read the packet data with |
469 |
a "jsr (a4)" or "jsr 2(a4)" |
470 |
6. This will execute an M68K_EMUL_OP_ETHER_READ_PACKET opcode |
471 |
7. The EtherReadPacket() opcode handling routine will copy the requested |
472 |
part of the packet data to Mac RAM using the pointer and length which are |
473 |
still in a0/d1 |
474 |
|
475 |
For a more detailed description of the Ethernet driver, see "Inside AppleTalk". |
476 |
|
477 |
6.12. System-dependant device access |
478 |
------------------------------------ |
479 |
|
480 |
The method for accessing floppy drives, hard disks, CD-ROM drives and files |
481 |
vary greatly between different operating systems. To make Basilisk II easily |
482 |
portable, all device I/O is made via the functions declared in "sys.h" and |
483 |
implemented by the (system-dependant) "sys_*.cpp" modules which provides a |
484 |
standard, Unix-like interface to all kinds of devices. |
485 |
|
486 |
6.13. User interface strings |
487 |
---------------------------- |
488 |
|
489 |
To aid in localization, all user interface strings of Basilisk II are collected |
490 |
in "user_strings.cpp" and accessed via the GetString() function. This way, |
491 |
Basilisk II may be easily translated to different languages. |
492 |
|
493 |
6.14. Preferences management |
494 |
---------------------------- |
495 |
|
496 |
The module "prefs.cpp" handles user preferences in a system-independant way. |
497 |
Preferences items are accessed with the PrefsAdd*(), PrefsReplace*() and |
498 |
PrefsFind*() functions and stored in human-readable and editable text files |
499 |
on disk. There are two lists of available preferences items. The first one, |
500 |
common_prefs_items, defines the items which are available on all systems. |
501 |
The second one, platform_prefs_items, is defined in prefs_*.cpp and lists |
502 |
the prefs items which are specific to a certain platform. |
503 |
|
504 |
The "prefs_editor_*.cpp" module provides a graphical user interface for |
505 |
setting the preferences so users won't have to edit the preferences file |
506 |
manually. |
507 |
|
508 |
7. Porting Basilisk II |
509 |
---------------------- |
510 |
|
511 |
Porting Basilisk II to a new platform should not be hard. These are the steps |
512 |
involved in the process: |
513 |
|
514 |
1. Create a new directory inside the "src" directory for your platform. If |
515 |
your platform comes in several "flavours" that require adapted files, you |
516 |
should consider creating subdirectories inside the platform directory. |
517 |
All files needed for your port must be placed inside the new directory. |
518 |
Don't scatter platform-dependant files across the "src" hierarchy. |
519 |
2. Decide in which mode (virtual addressing, real addressing or native CPU) |
520 |
Basilisk II will run. |
521 |
3. Create a "sysdeps.h" file which defines the mode and system-dependant |
522 |
data types and memory access functions. Things which are used in Basilisk |
523 |
but missing on your platform (such as endianess macros) should also be |
524 |
defined here. |
525 |
4. Implement the system-specific parts of Basilisk: |
526 |
main_*.cpp, sys_*.cpp, prefs_*.cpp, prefs_editor_*.cpp, xpram_*.cpp, |
527 |
timer_*.cpp, audio_*.cpp, video_*.cpp, serial_*.cpp, ether_*.cpp, |
528 |
scsi_*.cpp and clip_*.cpp |
529 |
You may want to take the skeleton implementations in the "dummy" directory |
530 |
as a starting point and look at the implementation for other platforms |
531 |
before writing your own. |
532 |
5. Important things to remember: |
533 |
- Use the ReadMacInt*() and WriteMacInt*() functions from "cpu_emulation.h" |
534 |
to access Mac memory |
535 |
- Use the ntohs() etc. macros to convert endianess when accessing Mac |
536 |
memory directly |
537 |
- Don't modify any source files outside of your platform directory unless |
538 |
you really, really have to. Instead of adding "#ifdef PLATFORM" blocks |
539 |
to one of the platform-independant source files, you should contact me |
540 |
so that we may find a more elegant and more portable solution. |
541 |
6. Coding style: indent -kr -ts4 |
542 |
|
543 |
|
544 |
Christian Bauer |
545 |
<Christian.Bauer@uni-mainz.de> |