linux - What does the usage of mprotect() as an ASM syscall look like with respect to its third argument? -

in i386 architecture linux know can build syscall loading identity of syscall eax , arguments ebx, ecx, etc.

i confused third argument mprotect in case; assuming wanted make memory segment within binary's allocated memory executable, how encoding work prot_exec argument (arg 3)? i'm aware first 2 arguments (1) pointer start of allocated block, , (2) length of allocated block relatively easy conceptualize in asm (as hexadecimal addresses in memory).

how format third argument mprotect() interrupt-issued syscall in i386 assembly on linux?

thanks.

tl;dr: an integer passed third parameter.

now let's answer question in comments. if open mman-common.h, should sit in /usr/include/asm-generic, find these values.

#define prot_read       0x1             /* page can read */ #define prot_write      0x2             /* page can written */ #define prot_exec       0x4             /* page can executed */ #define prot_sem        0x8             /* page may used atomic ops */ 

just before compiling, preprocessor replaces parameters numbers above. if had call:

mprotect(myaddress, 256, prot_read | prot_write); 

it replaced code:

mprotect(myaddress, 256, 0x1 | 0x2); 

now @ values different parameters can take: haven’t been chosen randomly, powers of two, in binary notation represented 1 one digit , zeros.

prot_read  = 0x1 =   00000001 prot_write = 0x2 =   00000010 prot_exec  = 0x4 =   00000100 

choosing powers of 2 handy because when use binary or, number obtain combine 2 previous values, both information contained or-ed number.

prot_write | prot_exec =      00000010   | 00000100 =   00000110 

so our call:

if had called mprotect(myaddress, 256, prot_read | prot_write), have happened prot_read | prot_write have been combined 0x1 | 0x2, 0x3.

now on kernel side, suppose prot_read | prot_write written user. kernel receives argument 0x3, , wants check whether prot_read written. way write this:

if (prot_read & uservalue) { } 

it works because uservalue contains combined version of prot_read , prot_write, in binary:

prot_read & uservalue =      00000001 &&  00000011  =  00000001 

if flag set, number non-zero, kernel knows flagged passed.

hope helps.