|
2349
|
1 The Level III modifications are Copyright 1996 by Alan T. DeKok, all rights
|
|
|
2 reserved.
|
|
|
3
|
|
|
4
|
|
|
5 Level III
|
|
|
6 =========
|
|
|
7
|
|
|
8 It's what everybody's been asking for, so here goes.
|
|
|
9
|
|
|
10 One warning: Level III is VERY dependant on the order of modules in the
|
|
|
11 OS9Boot file. If you do NOT follow the directions below exactly, your
|
|
|
12 system will NOT BOOT. I'm willing to answer questions and perform support,
|
|
|
13 but Level III is not for the faint of heart or the OS-9 beginner.
|
|
|
14
|
|
|
15 When you install Level III, you MUST do it on a system using NitrOS-9.
|
|
|
16
|
|
|
17 When you build a Level III OS9Boot, be aware that you WILL NOT be able
|
|
|
18 to cobbler a disk once LIII is booted. You must copy the OS9Boot by hand,
|
|
|
19 and use 'ezgen' to update LSN0, or use OS9Gen.
|
|
|
20
|
|
|
21 ******************************************************************************
|
|
|
22 ** DO NOT COBBLER DISKS WITH LEVEL III! IT WILL COBBLER BUT WILL NOT BOOT! **
|
|
|
23 ******************************************************************************
|
|
|
24
|
|
|
25 Your OS9Boot file MUST be magically formatted as follows. The reasons for
|
|
|
26 this are long and detailed. I'll just say that if it isn't done this way,
|
|
|
27 Level III will NOT work.
|
|
|
28
|
|
|
29 ****** start of the OS9Boot file.
|
|
|
30 NitrOS9 (from the NitrOS9.L3 file.)
|
|
|
31 SCF
|
|
|
32 (other SCF modules, i.e. SACIA, Printer, VRN)
|
|
|
33 ( NOT CC3IO or WindInt or VDGInt. Never!)
|
|
|
34 (up to a limit of 15k of modules)
|
|
|
35 _end <--- specifies the end of SCF information, from the _end.L3 file.
|
|
|
36 RBF
|
|
|
37 (other RBF modules, i.e. CC3Disk, Rammer, RamPak, CCHDisk, etc.)
|
|
|
38 _end <--- specifies the end of RBF information
|
|
|
39 (everything else. Order does not matter here.)
|
|
|
40 (ALL descriptors, Clock, IOMan, OS9p2, Init, CC3IO, Windint,
|
|
|
41 PipeMan, etc.)
|
|
|
42 ****** End of the OS9Boot file.
|
|
|
43
|
|
|
44 Late tests indicate it is possible to put CC3IO inside the SCF local memory.
|
|
|
45 Whether or not it works is dependent on local system configurations. Either
|
|
|
46 way, there's plenty of system memory, and it doesn't appear to make any
|
|
|
47 difference to anything.
|
|
|
48
|
|
|
49 Similarly, PipeMan may be put in SCF local memory between SCF and the first
|
|
|
50 _end module. Note that PipeMan may NOT be put in with RBF!
|
|
|
51
|
|
|
52 An 'ident -s' of my Level III OS9Boot file (with comments) looks like this:
|
|
|
53
|
|
|
54 1 $C7 $389B28 . NitrOS9
|
|
|
55 16 $D1 $E2B0EB . SCF \
|
|
|
56 12 $E1 $82D6D3 . SACIA | SCF and SCF device drivers
|
|
|
57 1 $E1 $329D69 . VRN |
|
|
|
58 13 $E1 $3389A4 . PRINTER /
|
|
|
59 1 $C7 $BB7259 . _end
|
|
|
60 30 $D1 $753C86 . RBF \
|
|
|
61 4 $E1 $3AB3AA . Rammer | RBF and RBF device drivers
|
|
|
62 2 $E1 $BCE29D . RamPak |
|
|
|
63 11 $E1 $CFAA38 . CC3Disk /
|
|
|
64 1 $C7 $BB7259 . _end
|
|
|
65 1 $C1 $BFF958 . OS9p3 everything else
|
|
|
66 67 $C0 $680456 . Init
|
|
|
67 82 $F1 $7EBD63 . DD DEVICE DESCRIPTORS GO HERE
|
|
|
68 82 $F1 $93C195 . D0
|
|
|
69 82 $F1 $F09A9D . D1
|
|
|
70 82 $F1 $323E4F . D2
|
|
|
71 82 $F1 $092A49 . MD
|
|
|
72 82 $F1 $B31A99 . R0
|
|
|
73 114 $F1 $FD590E . T2
|
|
|
74 83 $F1 $C9B7C1 . P
|
|
|
75 19 $C1 $67D5A0 . WindInt OUTSIDE SCF local memory!
|
|
|
76 20 $E1 $B4F1E2 . CC3IO perhaps here, perhaps in SCF local memory
|
|
|
77 83 $F1 $17FBEF . Term
|
|
|
78 83 $F1 $F8D0CB . W
|
|
|
79 83 $F1 $5EA5AA . W1 WINDOW DESCRIPTORS HERE!
|
|
|
80 83 $F1 $691BAA . W2
|
|
|
81 83 $F1 $FB498B . W3
|
|
|
82 83 $F1 $05BFAA . W4
|
|
|
83 83 $F1 $97ED8B . W5
|
|
|
84 83 $F1 $E30B59 . W6
|
|
|
85 83 $F1 $710B78 . W7
|
|
|
86 83 $F1 $9F0B78 . W8
|
|
|
87 83 $F1 $0D0B59 . W9
|
|
|
88 18 $C0 $826A54 . OS9p2 All these modules MUST come after the '_end'
|
|
|
89 12 $C1 $1B5CA4 . IOMan module, or your system will crash.
|
|
|
90 9 $C1 $402FE2 . Clock
|
|
|
91 4 $D1 $ED94CC . PipeMan
|
|
|
92 2 $E1 $ADB22E . Piper
|
|
|
93 80 $F1 $CC06AF . Pipe
|
|
|
94 1 $F1 $B2004F . Nil
|
|
|
95
|
|
|
96
|
|
|
97 OK, now the theory behind the magic.
|
|
|
98
|
|
|
99 When booting OS-9, the startup routine in OS9p1 calls the Boot module to
|
|
|
100 load in the OS9Boot file. OS9p1 then looks for a module named 'NitrOS9
|
|
|
101 as the FIRST module of the OS9Boot file. If it exists, that module is called
|
|
|
102 BEFORE verifying the OS9Boot file.
|
|
|
103
|
|
|
104 The NitrOS9 module allocates memory, and moves the SCF modules (up to
|
|
|
105 the _end module) into 'SCF local memory', and verifies the modules. The
|
|
|
106 same procedure is followed for RBF modules, and when I get the chance later,
|
|
|
107 for PipeMan and SBF modules.
|
|
|
108
|
|
|
109 Extra memory is allocated in 16k chunks, and mapped into address $2000
|
|
|
110 to $5FFF of the system map. An 'smap' will show you that this memory is
|
|
|
111 allocated.
|
|
|
112
|
|
|
113 IOMan intercepts EVERY system call, and EVERY request/return system call.
|
|
|
114 Any module inside 'local memory' asking for system RAM gets allocated
|
|
|
115 RAM from that local memory. Any module in 'system global memory' asking
|
|
|
116 for system RAM gets allocated RAM from the global memory pool.
|
|
|
117
|
|
|
118 The net result of this is that you lose 16k of memory from the bottom
|
|
|
119 of the system memory map, but you move about 16k of modules into system local
|
|
|
120 memory. On my setup, I had about 14K free after booting, and now 'smap'
|
|
|
121 usually gives me 11k of _global_ memory free. And I still have 8k of SCF
|
|
|
122 memory, and 9k of RBF memory left!
|
|
|
123
|
|
|
124 The ONLY things that take memory out of 'system global memory' are process
|
|
|
125 descriptors (.5k each) and path descriptors (.0625K, 64 bytes). 11K is
|
|
|
126 room for another 20 processes, each with STDIN, STDOUT, and STDERR going
|
|
|
127 to a window. And 8k of SCF memory is just about enough memory for SCF
|
|
|
128 buffers for each of those 20 processes.
|
|
|
129
|
|
|
130 i.e. Level III enables you to have about 25 _different_ processes
|
|
|
131 running in different windows, all at the same time! The only way that this
|
|
|
132 is possible on a Level II system is to strip your OS9Boot file down to almost
|
|
|
133 nothing. Level III allows you to keep all of your favorite modules in your
|
|
|
134 OS9Boot.
|
|
|
135
|
|
|
136 VDG windows come out of system global memory, too. That's why I added the
|
|
|
137 'largest free block' display to the 'sfree' command, below. If you have
|
|
|
138 less than 6k free in one block, you can't run many old Coco2 games.
|
|
|
139
|
|
|
140 The local 16k memory map is saved and restored across every system call,
|
|
|
141 and changed whenever a call to SCF/RBF is done. Because only IOMan knows
|
|
|
142 about the changes, only it (and clock) have to be modified. All of the
|
|
|
143 old file managers and drivers work perfectly. This includes programs
|
|
|
144 like 'winfo' and 'wdir'.
|
|
|
145
|
|
|
146 I got the idea from thinking that it's perfectly possible to boot OS-9
|
|
|
147 with NO SCF modules, and likewise with NO RBF modules. If that's possible,
|
|
|
148 why not map SCF and RBF into the same area of memory, and switch between
|
|
|
149 them on the fly? So far, I would say it's taken me the equivalent of 4 weeks
|
|
|
150 full-time work, as I ran into HORRIBLE problems. Level III is sufficiently
|
|
|
151 different from Level II that you can't really think the same way about it.
|
|
|
152
|
|
|
153 Anyways, here's an 'sfree' of my current system, with OSTerm running
|
|
|
154 in another window, and 3 'shell' windows:
|
|
|
155
|
|
|
156 ----- Level III System Memory ----- System memory:
|
|
|
157 37 free pages, largest block 37 pages.
|
|
|
158 9K of free RAM.
|
|
|
159
|
|
|
160 SCF local memory:
|
|
|
161 21 free pages, largest block 21 pages.
|
|
|
162 5K of free RAM.
|
|
|
163
|
|
|
164 RBF local memory:
|
|
|
165 31 free pages, largest block 31 pages.
|
|
|
166 7K of free RAM.
|
|
|
167
|
|
|
168 ----
|
|
|
169 That's 89 pages (22K) of free RAM, and my OS9Boot file is $73E2 bytes long!
|
|
|
170 Even with a 4K SACia buffer open, 8 processes, and 4 windows! I like it...
|
|
|
171
|
|
|
172 Here's an 'smap':
|
|
|
173
|
|
|
174 0 1 2 3 4 5 6 7 8 9 A B C D E F
|
|
|
175 # = = = = = = = = = = = = = = = =
|
|
|
176 0 U U U U U U U U U U U U U U U U
|
|
|
177 1 U U U U U U U U U U U U U U U U
|
|
|
178 2 U U U U U U U U U U U U U U U U
|
|
|
179 3 U U U U U U U U U U U U U U U U
|
|
|
180 4 U U U U U U U U U U U U U U U U
|
|
|
181 5 U U U U U U U U U U U U U U U U
|
|
|
182 6 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
|
|
|
183 7 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
|
|
|
184 8 _ _ _ _ _ U U U U U U U U U U U
|
|
|
185 9 U U U U U U U U U U U U U U U U
|
|
|
186 A U U U U U U U U U U U U U U U U
|
|
|
187 B U U U U U U U U U U U U U U U U
|
|
|
188 C U U U U U U U U U U U U U U U U
|
|
|
189 D U U U U U U U U U U U U U U U U
|
|
|
190 E U U U U U U U U U U U U U U U U
|
|
|
191 F U U U U U U U U U U U U U U U .
|
|
|
192
|
|
|
193 Number of Free Pages: 37
|
|
|
194 Ram Free in KBytes: 9
|
|
|
195
|
|
|
196
|
|
|
197 ---
|
|
|
198 This is enough room to format a disk (greater than 7K), even with many
|
|
|
199 processes and OSTerm running!
|
|
|
200
|
|
|
201 I've been running Level III for just under a year, and I've had it in
|
|
|
202 beta test on 10+ systems for 8 months. One system was running a BBS, and
|
|
|
203 at one report was up and running for 2 months without a reboot. I think
|
|
|
204 it's stable enough to release to the general public.
|
|
|
205
|
|
|
206 Have fun!
|
|
|
207
|
|
|
208 Alan DeKok.
|
