Re: which book to start with...?

Frank Kotler wrote:
naunetr wrote:

...
okay thanks. luckily linux API is simple to use and resembles the c standard lib routines somewhat. the man page descriptions are written for c programming. but its easy to translate to assembler so far. but what about syscalls that take more than 5 parametres? are the others pushed onto stack like in c (but not in reverse order)?

If there are more than 5 parameters (6 in newer sys_calls), they're arranged in a structure, and the address of the structure is passed in ebx. (usually... AFAIK). This often *is* on the stack, but doesn't have to be. I guess "fastcall" can put some parameters in registers, and "extras" on the stack, and I think the 64-bit calling conventions do it, but I think 32-bit Linux is always either/or.

okay this is sensible. only problem is the alignment of structure fields. is there any specification for alignment requirements for structures and stuff? both for kernel and clib? something taht says for example that char vars are padded to occupy four bytes rather than just one etc?

...
can we define the standard sections (.text, .data and .bss) sevaral times? will they be merged when object file is created?

Yeah.

section .text
_start:
mov eax, 4
section .data
msg db "hello "
section .text
mov ebx, 1
section .data
db "world"
section .text
mov ecx, msg
section .data
db "!"
section .text
mov edx, 13
section .data
db 10
section .text
int 80h
mov eax, 1
int 80h

Something like that ought to work (untested). We can also get ld to "merge" or "combine" or "overlap" sections, I think, but I don't know how that works.

i think Jeff Duntemann's book explains this. havent got to that chapter yet though. just installed nasm 16 bit and 32 bit bins under dosemu. also got djgpp to work under dosemu.


also is there any chanvce that .data can be put in read-only memory?

Yeah, as long as we don't need to alter it. We can put "constant data" in .text, or Nasm knows .rodata. The default attributes of an arbitrarily-named section are right for "constant data" too, I think.

so any vars that we need to modify at runtime must be placed only in ..bss or a custom defined section, because .data may or may not be modifyable?

this is a problem when we want to have vars that we want to give initial values but want to modify at runtime too. .bss wont accept initialisations while .data will but no garantee for modification at runtime.


because i use it to define buffers also... or should i be safe and put them in .bss as nasm advises?

Buffers - not initialized - can go in .bss, where they won't add anything to the length of the file. Or we could put 'em in .data... but they *will* add to the length of the file. We need the memory at runtime either way, of course.

okay. so .data is modifiable? sorry for the confusion...


section .bss
buffer resb 1000h

or:

section .data
buffer resb 1000h

Nasm will complain about that, but do the right thing. The "right" way to do it - in .data - would be:

buffer times 1000h db 0

You wouldn't want a buffer in .text, since we'll be writing to it. Section .bss is nominally "uninitialized" data, but is in fact cleard to zero. An exception to this is a dos .com file, where you can write something to .bss, exit the program, restart it and still find our data there. If something happens between runs to cause us to be loaded at a different address, this won't work, of course...

okay interesting...

...
so if we have compared two signed values with cmp we should probably use jumps like je, jg, jl, js etc. and for unsigned compares jz, ja, jb etc. am i on the right track.

Right.

but if we are sure that the result of a cmp between two signed values will give only a positive result then we can use ja, jb and similar?

I guess it depends on what you mean by "positive result". We can mix 'em a little, but be aware that a "large" unsigned number can be a "small" (less than zero) signed number.

okay. i think once twos complement is understood then the whole thing becomes logical.


mov al, 0FFh ; could mean 255 or -1
mov bl, 1
cmp al, bl
ja someplace ; taken 255 greater than 1 (unsigned)
jg someplace ; not taken -1 not greater than 1 (signed)


so both these jumps are "correct" depending on what type of comparison you want. is that right?

Right. We'd probably write "mov al, -1" for clarity, if we intended it to be treated as signed, but it's the same bit pattern.

...
even for binary files i think read will return 0 when the end of file is reached.

Yes. Read returns the number of bytes read (in eax). But our "getc" wants to return the character we read, not the number of bytes. If read returns zero - EOF - we haven't *got* a byte to return. We could just exit right there - *this* program is finished when that happens. But we might want to do more, so we return some value that couldn't be a valid character. If we returned zero from getc when read returned zero, we couldn't distinguish between a valid character (zero is unlikely in a text file, but...) and EOF.

okay. this is api design stuff. im thinking that sometime later i can write a small code library so that i dont have to cut and paste evertime. i still dont know enough about global, extern and stuff but later... also somehow i dont like macros but i really like routines. but nasm seems to have a lot of facility for macros. but macros tend to confuse me :(


...
basically i thought preopening stdin, stdout and stderr are only c behaviour and since we are not using the c library at all, we should do an open for them. in windows we have to do GetStdHandle before we can write to console. but stdin, stdout and stderr have no pathnames associated for them so open is not possible.

Might be able to open "/dev/tty?" or something...

i think this might be variable. it could be /dev/pty or something else. in anycase we may not have permissions also.


this is the docs installed by nasm v2 installation from source. it installed in /usr/doc/nasm and the html contents file (nasmdoc0.html) doesnt have a link to the instruction refarence but i have to manually load the nasmdocb.html file in browser.

There shouldn't be any nasmdocb.html in 2.00 at all. Shouldn't be any reference to it in the contents, or the index, or anywhere. I wouldn't want to guarantee that it's completely "clean", though.

there is a nasmdocb.html in the archive i dl'ed from sourceforge two days ago. it was a .tar.gz source archive. but it's not linked into the table of contents file. why remove it? the info is still correct if also incomplete...


http://home.comcast.net/~fbui/intel.html

So...I think it's "too bad" that the section was removed from the Nasm manual, but it isn't a total loss...


okay i'll use that then. i used nasm's one because intel's manuals are too big and detailed for me yet :) also i hate pdf files...

Painful Document Format? Agreed! (but I really should read that Intel manual some day...)

Best,
Frank

yes. dont know how pdf became the most popular doc format on the net and for everyone it seems. i prefer html first and then plain text. all other formats are too heavy and lack good programs to view them.


thanks again.
.

Relevant Pages

  • Re: Buffers in Assembly (NASM)
    ... can create a zero filled buffer in the bss section in NASM with this: ... I am unfamiliar with NASM and I think it also depends on the object ... mov ebx, esp;save the start point of the buffer ...
    (comp.lang.asm.x86)
  • Re: Buffers in Assembly (NASM)
    ... can create a zero filled buffer in the bss section in NASM with this: ... it doesn't know what ".bss" means. ... In an uninitialized section, there's "nothing there", so it would be "conceptually impossible" for Nasm to zero it. ... mov al, 'N' ...
    (comp.lang.asm.x86)
  • Re: debug and the newbe
    ... 0DF1:0102 mov dx, 107 ... 0DF1:0105 int 21 ... Here is a more appropriate example, written in NASM. ... AL is set to zero and that returns a code of ...
    (alt.lang.asm)
  • [NASM V.0.98.38] Defining a Stack
    ... define a Stack Segment using NASM V.0.98.38. ... I tried [SEGMENT .bss] but, ... mov ax, my_segment_name ...
    (alt.lang.asm)
  • Re: which book to start with...?
    ... mov eax, 4 ... buffer resb 1000h ... Section .bss is nominally "uninitialized" data, but is in fact cleard to zero. ... i used nasm's one because intel's manuals are too big and detailed for me yet:) also i hate pdf files... ...
    (alt.lang.asm)