How can I read environment variables in the early stages?

Question

Use LD_PRELOAD to load shared objects with the initfirst flag. Calling the getenv() function from a function with __attribute__((constructor)) returns NULL. I think this is probably because the function is being called before someone (libc?) initializes __environ.

// foo.c
#include <stdio.h>
#include <stdlib.h>

void setup(void) __attribute__((constructor));

void
setup(void)
{
    fprintf(stderr, "FOO=\"%s\"\n", getenv("FOO"));
}

$ cc -o foo.so -fPIC -shared foo.c -Wl,-z,initfirst
$ LD_PRELOAD=./foo.so FOO=123 ./hello
FOO="(null)"

Removing the initfirst flag produced the expected result.

Questions:

Is there a way to read environment variables under these conditions? (It's fine if it only works on Linux, amd64, gcc, and glibc)
Who initializes __environ? The loader, libc, crt0, or the kernel?

Environment: Ubuntu 24.04.1 LTS, amd64, gcc 13.3.0, glibc 2.39

github.com/bminor/glibc/blob/glibc-2.39/csu/libc-start.c __libc_start_main initializes it (some interesting comments in github.com/bminor/glibc/blob/glibc-2.39/sysdeps/x86_64/start.S). — Mat
– Mat, Commented Jul 19 at 16:23
Aside: does your implementation support int main(int argc, char *argv[], char *env[])? — Weather Vane
– Weather Vane, Commented Jul 19 at 17:40

o11c · Accepted Answer · 2025-07-19 19:11:12Z

5

Funtions marked with __attribute__((constructor)) take the same arguments as main, including the envp argument. Note that envp is a bad idea to use within main since it breaks if setenv is called, but here it is useful since environ has not yet been initialized.

// foo.c
#include <stdio.h>
#include <stdlib.h>

void setup(int argc, char **argv, char **envp) __attribute__((constructor));

void
setup(int argc, char **argv, char **envp)
{
    fprintf(stderr, "argc: %d\n", argc);
    for (int i = 0; i < argc; ++i)
    {
        fprintf(stderr, "argv[%d]: %p\n", i, argv[i]);
    }
    for (char **p = envp; *p; ++p)
    {
        fprintf(stderr, "env: %s\n", *p);
    }
}

Since this question relies on dynamic linking, we can ignore all libc implementations that don't support that. Among those that do:

broken: MUSL
works: GLIBC
works: all BSDs (not tested)
not investigated: Mac OS X, Solaris, Cygwin, etc.

edited Jul 19 at 19:11

answered Jul 19 at 18:49

o11c

16.5k5 gold badges57 silver badges80 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

kakkoko Jul 20 at 7:12

Thanks to this excellent answer, I learned that arguments are passed to each function in .init_array, so I checked the glibc source code and found that three arguments were indeed passed. I couldn't find any mention of this issue in the gcc or glibc documentation.

Collectives™ on Stack Overflow

How can I read environment variables in the early stages?

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related