Power-Programmierung/CD1.mdf/assemblr/library/sampler0/resource.doc

Text File | 1986-01-28 | 26KB | 772 lines

        RESOURCE

    A disassembler for 8080 programs
    by Ward Christensen

    CP/M U.G. 1/80

Suggestions?  Call me eve's at [phone]
        ----------------
RESOURCE commands are inconsistent at best. - RESOURCE is a
kludge based on years of disassembler experience and hacking,
and was never "planned" - just coded sitting at a tube, and
modified over 2 years before being contributed to the CP/M UG.

For example, to kill a symbol: k.label   but to kill a
control value: caddr,k and to kill a comment: ;addr,
but RESOURCE does the job like no other I have seen.
        ----------------
N-O-T-E: Pardon the editorial, but I feel hardware without
good software is useless to 99% of us.  Most good software
has to be paid for.  I strongly support the legitimate purchase
of licensed software.  I do not regularly use any programs
which I have not purchased.  (Yes, I do occasionally "try"
one, but then buy it if I plan on using it).  I have been
asked by software businesses to NOT distribute RESOURCE -
because of it's ability to produce good .asm source quickly.
But, there are so many disassemblers out, why not a good,
conversational one?   Please use it in the spirit in which it
was contributed: to enlarge your understanding of the micro-
computer world around you, and to allow you to customize
programs which you legitimately own, for your own use.

    "Semper non rippus offus"
----
NOTE: any command taking a hex address (Dnnnn, etc)
may take a value in the form .label but arithmetic
may not be performed.  (i.e. d.start is ok, but d.start+8 not)
        ----------------
    Overall structure of RESOURCE:  It is a .COM file,
which runs at 100H.  It goes thru 1700 or so, then the stack.
At 1800 is a 512 entry table for control commands.  Each is
3 bytes long, and controls the format of the re-sourced list,
i.e. is it DB, DS, DW, instructions, etc.
    At 1E00 is the start of the symbol table.  It has no
defined length as such.  If it is not used, it occupies only
2 bytes.
    If you want to re-source something which is in memory,
such as a PROM, a program previously loaded in high memory,
"CP/M itself", or whatever, you can just do so.
    However, typically you want to disassemble a program
which runs at 100H, which is were RESOURCE runs.   Bob
Van Valzah would have solved that by making resource
relocatable and moving itself up under BDOS.  I wasn't that
industrious.
    Instead, RESOURCE uses the concept of an "invisible"
OFFSET.  After all, what do you care where it is as long
as it LOOKS like it's at 100h?
    So, you set an offset.  O2F00 sets it to 2F00 Hex.
Reading a .COM file (RFOO.COM) causes it to come into 3000 on.
If you say D100 or L100 it dumps or lists what LOOKS like your
program.  Internally, RESOURCE is adding the offset to the
D and L addresses.
    What should you set the offset to?  Well, that depends
upon how many symbols you will use.  O1F00 will load the
program at 2000, thus allowing only 1E00-1FFF for symbols,
i.e. 512 bytes or about 50-60 labels.  If you didn't leave
enough space, then used B to build a default symbol table,
the table could run into and clobber your .com file!  (easy
recovery, however: just change the offset to being higher,
and read in the .COM file again)  Each entry takes 3 bytes
+ the symbol length, and if you like 7 byte labels like I
do, that means 10 bytes/label.  An offset of 2F00 should
be adequate.
    If you want to put comments into the disassembled
program, you will have to designate an area to Use for the
comments.  The U command (e.g. U4000) specifies what area
is to be used.
    Before issuing the O (offset) command, do:
    L5 7
which will show the JMP to BDOS, which is the highest memory
you may use.  (Note if you have, for example, an empty 4K
memory board in high memory, you can Use THAT for comments).
    Let's take an example:  You have an 8K file, FOO.COM
which you want to disassemble.  It will have about 300 labels.
300 x 10 is 3000, or call it 4K (what's a K unless your
tight).  The symbol table starts at 1E00.  4K more is 2E00.
Let's load the .COM at 2E00, so since it normally starts
at 100H, the offset is 2D00.  O2D00 is the command.
We then RFOO.COM to read it in.  It says 4E00 2100 which
means it came into actual memory to 4E00, but 2100 if we
are talking with respect to loading at 100.
    Thus, we could set our comments table up after
the .COM program - say at 5000:  U5000
    The ? command shows the memory utilization for
control, symbol, and comments entries.  (no, I never put
in anything to keep track of the .COM - you'll just have
to do that yourself).
    If you ever want to dump real memory, you'll have
to reset the offset to 0: O0  but then set it back.
If you are not sure what it is, typing O will tell the
current offset.

    Hoo, boy!  Hope this kludge of documentation
is enough to get you going - hmmm, better give you
some of the gotcha's I've discovered...

        ---- WATCH FOR ----

*    Symbols overflowing into the .COM.
    (Use ? command to see how full symbol table is)

*    Control entries overflowing into .SYM (altho I
    can't believe anyone will have a program with
    more than 512 control entries!!!)

*    Comments overflowing into BDOS (ug!!)

*    Using an offset which is not in free memory
    and overlaying BDOS or whatever.

*    The B(uild) command gobbling up too much when building
    a DB:  "B" will take a DB 'GOBBELDY GOOK' followed
    by LXI H,FOO and take the LXI as a '!' (21H) so
    you'll have to manually stick a new "I" control
    entry in at the address of the LXI.  You might
    also delete the incorrect "I" entry which RESOURCE
    stuck in (typically at the second byte of the LXI)

*    Trying to dump real memory without setting the
    offset back to 0.  (then forgetting to set it back
    to its proper value)

*    Forgetting how big the .COM file you are disassembling
    was.

*    Using RESOURCE to rip off software (yes, I know, you
    heard that before, but only 3 in 100 needed to be
    told, and 2 in 100 needs to be told again, and 1 in
    100 doesn't give a rat's fuzzy behind anyway!!)

*    Forgetting to take checkpoints when disassembling
    large files.  You may even want to rotate the names
    under which things are saved:

    STEMP1.SYM
    STEMP1.CTL
    STEMP1.DOC

*    Missing a label:  Suppose you have a control entry
    for a DW, resulting in:

    DFLT:        ;172C
        DW    100H

    but somewhere in the program, the following exists:

        LDA     172DH

    Even if you did a B and have a label L172D, it won't
    show up since it's in the middle of a DW.  Instead,
    do this:

    K.l172d        kill the old label
    e172d,.dflt+1    put in the new label as a displacement
            off the beginning.

*    improperly disassembling DW's (see previous item).
    You might be tempted to make DFLT a DB so that

    DFLT:        ;172C
        DB    0
    L172D:        ;172D
        DB    1

    Note that while this disassembles and reassembles
    properly, it is not "as correct" as the technique
    used in the previous item.

*    Having the "B" command overlay your "E" control entry.
    What?  Well, "B"uild is pretty dumb.  If he finds 8
    DB type characters in a row, he fires off a DB from
    then on until he runs out of those characters.  Suppose
    your program was 200 long (ended at 3FF), and you
    had zeroed (aha!  Nice DB candidates) memory there
    (there meaning at your offset address + whatever).

    Then you QB100,400 and viola!!  RESOURCE overlaid
    your "E" control with a "B".
        ----------------

    RESOURCE is relatively complete.  (well, actually,
    the phrase "rampant featureitis" has been "mentioned").
    ...But there's always another day, and another K...

        SO... Here's my "wish list"

    ..it might save you telling me YOU think such-and-such
    would be nice...

*    Targets of LHLD, SHLD should automatically be flagged
    as type DW in the control table.  Ditto LDA and STA
    as DB or as second half of DW.  Ditto targets of LXI
    as DB (?).

*        E5C,.FCB
    followed by
        E6C,.FCB+
    should automatically calculate the appropriate
    displacement, and put it in the symbol table.

*    The comments facility should be enhanced to allow
    total SUBSTITUTION of entire line(s) of the code,
    i.e. at address such-and-such, replace the next 3
    bytes with the following arbitrary line.  This
    would help those "how do I explain what was being
    done" cases such as:    LXI    H,BUFFER AND 0FF00H

*    Add the ability to, in one instruction, rename
    a default (LXXXX) label to a meaningful name.

        ----------------
RESOURCE types an "*" prompt when it is loaded.  You may
then enter any of the following commands.  Each command
is a single letter followed by operands.  Commas are shown
as the delimiter, but a space will also work.

        ----------------
        ----------------
;    Put comments into the program.
    (must execute 'u' command first, to assign area
    for comments to be placed)

    ;addr,comment    enter a comment
    ;addr        lists existing comment
    ;        lists entire comments table
    ;addr,        deletes existing comment

    note that '\' is treated as a new line, i.e.
    \test\ will be formatted:

    ;
    ;TEST
    ;
        ----------------
Attempt to find DB's while listing the program.
-    This command works just like 'L', but attempts
    to find DB's of 8 chars or longer.

    (see 'L' command for operand formats)
        ----------------
Build default sym tbl (LXXXX) labels for each
-    2 byte operand encountered.  Note 'B' is
    identical to 'L' except labels are built.

    (see 'L' command for operand formats)
        ----------------
Control table usage:
-    c        dump ctl tbl
    cnnnn        dump from starting
    cnnnn,x        define format from nnnn
            to next entry.  values of x:

            B = DB (attempts ASCII
                printable, 0DH, 0AH, 0)
            W = DW (attempts label)
            S = DW to next ctl entry
            I = instructions
            K = kill this ctl entry
            E = end of disassembly

    NOTE every control entry causes a "control break"
    (NO, RESOURCE was NOT written in RPG) which means
    a new line will be started.  Thus if you have a
    string in memory which disassembles as:

    DB    'Invalid operand',0DH
    DB    0AH

    You might want to change it putting the 0DH,0AH
    together on the second line - just enter a "B"
    control entry for the address of the 0DH.

    The same technique could be used to make

    DB    'TYPESAVEDIR ERA REN '

    appear as

    DB    'TYPE'
    DB    'SAVE'
    DB    'DIR '
    DB    'ERA '
    DB    'REN '

        ----------------
dump:
-    dxxxx        Dumps 80H from xxxx on
    daaaa,bbbb    Dumps from aaaa thru bbbb
    d,bbbb        Continues, thru bbbb
    d        Continues, 80H more

    NOTE 80H is the default dump length.  If you have
    a larger display, you can change the default via:

    d=nn        nn is the HEX new default.

    For example, a 24 line tube could display 100H:

    d=100        or..
    d=100,200    Defaults to 100, dumps 200-2ff

        ----------------
enter symbol:
-    ennnn,.symbol     symbol may be of any length,
            and contain any char A-Z or 0-9,
            or "+" or "-".  This allows:
            E5D,.FCB+1.  Note the "+" is not
            checked, i.e. E5D,.FCB+2 would be
            wrong (assuming FCB is at 5C) but
            would be allowed to be entered.

    Note if you enter two symbols for the same address,
    whichever one is first alphabetically will show up
    on the disassembled listing.  If you have a label
    which has the wrong address, you need not explicitly
    kill the old one before entering the new.  A label
    which is spelled exactly the same as an existing one
    will replace the existing one even if the addresses
    are different.
        ----------------
Find occurrence of address or label.  Note this function
-    runs until interrupted (press any key).
    fnnnn,ssss    find address nnnn in memory.  Start
            the search at ssss.  Runs forever.
            Press any key to stop.
    f        continue previous find command
      fnnnn        find nnnn starting at address you last
            stopped at in the f command
        ----------------
kill symbol from table
-    k.symbol
        ----------------
list (disassemble).  This command is used to list the
-    file, or to list it to disk after enabling
    the .ASM file save via 'SFILENAME.ASM' command

    l        lists 10 lines from prev pc
    lssss,eeee    lists from ssss to eeee
    l,eeee        lists from current pc to eeee
    lssss        lists 10 lines at ssss

    Note that if you have a control 'e' entry, then the
    list will stop when that address is found.  This allows
    you to 'lstart,ffff'.

    The 10 line default may be changed via:

    L=nn        where nn is a HEX line count, e.g.
    L=14        set to 20 lines/screen

    You can change the default and list, e.g.

    L=9,100        Dflt to 9 lines, list at 100.

    NOTE when using L to list the .ASM program to disk,
    you should either list the entire program at once
    using: Lssss,eeee or, you can list small pieces
    at a time.  As long as you list again without
    specifying a starting address, (L or L,nnnn) then
    the output file will continue uninterrupted.

    You may do dump commands, and others, without
    affecting what is being written to disk.
        ----------------
offset for disassembly
-    o        print current offset
    onnnn        establish new offset

    (note the offset is always added to any
    address specified in an a, b, d, or l command.
    to dump real memory, the offset must be reset to
    0 (O0) before the dump.)
        ----------------
prolog generation - this routine generates an
-    ORG instruction, and equates for any label
    outside of a given low-hi address pair.
    (the start and end addresses of your program).
    e.g. if disassembling from 100 to 3ff, it will
    generate 'fcb equ 5ch' if FCB is in the symbol
    table.  In typical use, you would 'sfilename.asm'
    then use the P command to write the prolog, then
    the L command to write the program itself.

    Pstart addr,end addr

quiet command: any command which is preceeded by a q
-    will be done 'quietly'.  For example, to save
    a .asm program, you could just do:

    ql100,3ff    or    ql100,ffff if you have
    set the 'e' control in the control table.

    Another use is to build a default symbol table
    by taking a pass thru the program: QB100,xxxx
        ----------------
read .com, .ctl, .sym, or .doc file
-    rfilename.com    reads in at offset+100h
    rfilename.ctl    loads the ctl table
    rfilename.sym    loads the sym file
    rfilename.doc    loads the comments table (note
            'u' command must have been issued)
        ----------------
save .asm, .ctl, .sym, or .doc file
-    sfilename.asm    use 'l' command to write, z to end
    sfilename.CTL    saves the CTL table
    stablename.sym    saves the sym file
    sfilename.doc    saves the comments table
        ----------------
use area of memory for comments table
-    unnnn        such as ud000 if you had an
            open board at 0d000h
        ----------------
purge sym tbl and CTL tbl
    x
    -
        ----------------
close .asm file (note that a preferred way to close the
    .asm file is to have specified a control entry
    for the end address (e.g. c1ff,e))
    z
    -
--------------------------------------------
--------------------------------------------
Here is a sample of the RESOURCE usage.

Given:    a COM file (lets say test.com) which runs at 100
    (as any good COM file should), and goes thru 2FF.

lines preceeded with ---> are typed by you.

---> RESOURCE
---> o2200    set the offset to 2200, which means the
        program will read into 2200 + 100 = 2300.

---> rtest.com    reads the com file into memory. system says:
        2500 0300 which is the actual hi load addr,
        (2500) and the original hi load addr (300)

        REMEMBER this address (300) because you might
        want to put a "E" (end of assembly) control
        entry there.

<<<<note>>>> that all 'L' (disassembly list) and 'D' (dump)
    commands work with the offset added.  Thus, you
    should learn to forget that the disassembler is
    in memory, and think of it as if your program were
    actually at 100.  D100 will dump your program.

    also note:  if the program being "RESOURCEd" will
    have a fairly large symbol table, then you will
    have to set the offset higher: o2f00 or some such.
    (the ? command will show symbol table usage: if your
    symbol table is nearing the .com file, then just
    set a new offset (higher) and re-load the .com)

    if you want to dump r-e-a-l memory, you would have
    to reset the offset to 0: o0  (but don't forget to
    reset it to 1f00 before continuing with your program.)

If you are disassembling something which is in memory at
it's correct address (such as looking at ccp) then don't
set the offset.  It defaults to 0 when dis is first loaded.

--->    l100    list your program - lists "about" 10 lines.

--->    d100    do a dump of your program.

NOTE that typically here are the steps to disassembling
    a program which has just been read into memory:

Use the dump command to find the ASCII DB areas.

Note that the 'a' command may be used to automatically
find the db's, but you must then check them to insure
that they don't extend too far.  All printable characters,
0dh, 0ah, and 00h are considered candidates for ascii db's.

At least 8 characters in a row must be found to make sure
that long sequences of mov instructions won't be taken
as db's.

Use the cnnnn,k command to kill erronious entries put
in the control table by the a command, but then immediately
put in the right address, such as via cnnnn,i

if you wanted to scan the program for ascii db's yourself,
use the 'c' (control) command to set the beginning and
    end of ascii areas.  For example, a program
    which starts out:
0100        jmp    start
0103        db    'copyright .....'
0117    start    .....

would show up in the dump as:

0100  c3170144 4f50xxxx  xxxxxxxx xxxxxxxx  *...copyr ight....*
0110  xxxxxxxx xxxxxxxx  xxxxxxxx xxxxxxxx  *xxxxxxxx ........*


thus you would want to instruct the disassembler to switch
to db mode at 103, and back to instruction mode at 117, thus:
    c103,b
    c117,i

Continue doing this, bracketing every ascii db which is in
the middle of instructions, by a b control instruction
and an i control instruction.  Note that multiple db's in
a row need not have separate cnnnn,b instructions, but that
these do cause a 'line break', i.e. if you have a table
of ascii commands, for example:

02e5    db    'load'
02e9    db    'save'

the disassembler would disassemble these as:
02e4    db    'loadsave'

you could put in an additional control entry: c2e9,b, which
would cause the disassembler to generate:

02e4    db    'load'
02e8    db    'save'

which is much more readable and realistic.

Note that before generating each byte of a db, a symbol
table lookup is done to determine if there is a label
at that location, and if so, a new line is started.

Thus if 'loadlit' and 'savelit' were in the symbol table,
as the labels on the 'load' and 'save' above, no separate
'b' control instruction would be required as the label
would cause the break.

<<<<NOTE>>>> that at this time the automatic label checking
is n-o-t done for ds instructions.  Make sure that each
ds instrucion references only up to the next label.  This
means that multiple ds's in a row must each be explicitly
entered into the control table.  Presence of a label is
not sufficient.
            ----
steps, continued:

After building the control entries with cnnnn,b and cnnnn,i
put in a control entry cnnnn,e which defines the address of
the end of your program.  The l command will then automatically
stop there, and in addition, if you are in 'save xxx.asm'
mode, the output .asm file will be closed.  If you do
mot define a control 'e' entry, then you will have to use
the break facility to stop the l command (don't use control-c
as that will re-boot cp/m).  If you were writing an .asm
file, you would have to user the z command to close the
file.

Next, you would list your program to determine how it looks.
when you recognize a routine by it's function, insert a label.

For example, if you saw that location 7ef was a character
out routine (type) then enter a label into the symbol table:
    E7EF,.TYPE

NOTE that all symbols start with a '.', so as to be dis-
tinguished from hex data.

NOTE that if you want the disassembler to make default labels
for you, use b (for build labels) instead of l (for list
program).  The b commands causes lnnnn default labels to
be inserted in the symbol table for every 2 byte operand
encountered (LXI, SHLD, JMP, etc).  It will undoubtedly
make some you don't want, such as L0000.  You will have to:

    K.L0000        kill label L0000 from the table.

When you encounter data reference instructions, try
to determine what type of area the instruction points to.
Typically,, LXI instructions may point to a work area
which should be defined as a DS, or to an ASCII string,
in which case we will have already made it a 'b' control
instruction.  Operands of LHLD and SHLD instructions
should be made DW instructions.  For example if you
encounter LHLD 0534H, then issue a control instruction:

    C534,W

NOTE that whatever mode you are last in will remain in
effect.  Therefore if 534,w is the last entry in the
control table, all data from there on will be taken to be
DW's.

Suppose that you determine that address 7cf is a 128 byte
buffer for disk I/O.  You want it to disassemble to:

DKBUF        ;07CF
    DS    80H

You do this as follows:

    C7CF,S        to start the DS
    C84F,B        to define it's end, and
    E7CF,.DKBUF    to put the symbol in the table.

Continue, iteratively using the 'l' command and the 'c'
and 'e' commands until you have the listing in a nice
format.  You will then probably want to save the control
symbol, and comments tables.  Or, you could have been
saving them at checkpoint times (so if you make a
major mistake you could go back to a previous one).

To save a control file:
    sfilename.CTL    (any filename, may include a: or b:)

To save a symbol file:
    sfilename.sym

To save a comments file:
    sfilename.doc    (not ".com" of course)

NOTE that the filetypes must be used as shown, but
that any legal filename (or disk:filename such as b:xxxx.CTL)
may be used.

You could now control-c to return to CP/M, and come back
later to resume your disassembly:

    RESOURCE
    o2200
    rtemp.com
    rtemp.sym
    rtemp.ctl
    uxxxx        (such as u4000)
    rtemp.doc

This will take you back exactly where you left off.

If you want to save a .asm file out to disk, do the following:

Make sure that there is a control entry defining the end
of the program (such as c200,e) or else you will have to
specify the ending address and manually type a z command to
close the file.

    sfilename.asm

A message will indicate that the file is opened.  Any
subsequent a, b, or l command will have whatever is listed
written to disk.  Encountering a 'e' control, or typing a z
command will then close the .asm file.  The listing may
be interrupted, and continued.  Since the l command types
only 10 lines, use laddr,ffff to list thru the end of the
assembly.

If this is the 'final' save of the .asm file, you will
probably want to put an 'org' at the beginning of the
output file, as well as generate equ instructions for
any references outside of the program.  For example, a
typical cp/m program will have references to:
    bdos    at 5
    fcb    at 5ch
    tbuff    at 80h

the 'p' (for prologue) command generates the org, then
scans the symbol table and generates equates:
    BDOS    EQU    05H
    FCB    EQU    05CH    (etc.)


If you have a "e" control entry in your file, you can
list as follows:  laddr,ffff - the listing will continue
until the "e" control entry is found


additional commands:

if you entered a label in the symbol table but now want
to get rid of it:
    k.symbol

note to rename a symbol, such as when you had a system-
assigned lnnnn label but now want to make it meaningful:

    k.l0334
    e334,.type

you could even:
    e.l0334,.type
    k.l0334

but that takes more typing.



here are some more commands:

?        prints statistics on symbol and control table
        usage, etc.

c        prints the entire control table

cnnnn        prints the control table starting
        at address nnnn

ds        dumps the symbol table.  Interrupt it
        by typing any key.

ds.symbol    starts dumping at the specified symbol,
        or the nearest symbol.  thus "ds.f" starts
        the dump at the first label starting
        with the letter 'f'.

....have fun, and let me know of any problems or
    suggested improvements
    ------------------------

        RESOURCE
    "Quick" command summary

Any address may be replaced by .symbol i.e. D.START

;addr,comment        Enter a comment
;addr            Lists existing comment
;            Lists entire comments table
;addr,            Deletes existing comment
A(see "L" for operands)    Attempt to find DB's
B(see "L" for operands)    Build default sym tbl (Lxxxx)
C            Dump ctl tbl
Cnnnn            Dump ctl starting at nnnn
Cnnnn,x            Define format from nnnn (B,E,I,S,W)
Dxxxx            Dumps 80H from xxxx on
Daaaa,bbbb        Dumps from aaaa thru bbbb
D,bbbb            Dump thru bbbb
D            Dump 80H more
D=nn            nn= Hex dump size default.
Ds            Dumps the symbol table.
Ds.symbol        Sym dump starting at .symbol
Ennnn,.symbol        Enter symbol into table
Fnnnn,ssss        Find address nnnn starting at ssss
F            Continue previous find command
Fnnnn            Find nnnn
K.symbol        Kill symbol from symbol table
L            Lists 10 lines from prev pc
Lssss,eeee        Lists from ssss to eeee
L,eeee            Lists from current pc to eeee
Lssss            Lists 10 lines at ssss
L=nn            nn is hex list default # of lines
O            Print current offset
Onnnn            Establish new offset
Pstart addr,end addr    Generate program prolog
Q            Before any command suppresses
            console output: QB100,200
Rfilename.COM        Reads in at offset+100h
Rfilename.CTL        Loads the ctl table
Rfilename.SYM        Loads the sym file
Rfilename.DOC        Loads the comments table (note
Sfilename.ASM        Save .ASM file.  Write w/L, Z to end
Sfilename.CTL        Saves the CTL table
Sfilename.SYM        Saves the sym file
Sfilename.DOS        Saves the comments table
Unnnn            Use nnnn for comments table
X            Purge all symbols and control
Z            Write eof to .ASM file (
?            Prints statistics (sym, ctl, comments)