[Sebastien Tandel <sebastien@xxxxxxxxx>]

Here is the code from postfix, just as an example ;)

/*

     * Look up the uid/gid before entering the jail, and save them so they

     * can't be clobbered. Set up the primary and secondary groups.

     */

    if (user_name != 0) {

        if ((pwd = getpwnam(user_name)) == 0)

            msg_fatal("unknown user: %s", user_name);

        uid = pwd->pw_uid;

        gid = pwd->pw_gid;

        if (setgid(gid) < 0)

            msg_fatal("setgid(%ld): %m", (long) gid);

        if (initgroups(user_name, gid) < 0)

            msg_fatal("initgroups: %m");

    }

    /*

     * Enter the jail.

     */

    if (root_dir) {

        if (chroot(root_dir))

            msg_fatal("chroot(%s): %m", root_dir);

        if (chdir("/"))

            msg_fatal("chdir(/): %m");

    }

    /*

     * Drop the user privileges.

     */

    if (user_name != 0)

        if (setuid(uid) < 0)

            msg_fatal("setuid(%ld): %m", (long) uid);

On Aug 15, 2007, at 4:10 PM, Guy Harris wrote:

setuid() *should*, at least according to the Single UNIX Specification, 

set the real, effective, and set-user ID if the process is running with 

appropriate privileges.  The same applies, *mutatis mutandis*, to setgid().

In Mac OS X 10.4 and later, a process can, in effect, belong to a bigger 

group set than just the group set that fits in the credentials (checks 

whether a process's group set includes a given group are done by sending 

a message to memberd).  As I remember, the right way to change the group 

set of a process is to call initgroups(), passing it the user name of 

the user whose group set you want the process to pick up and that user's 

primary group ID.  initgroups() is a BSDism, and has been picked up by 

other OSes, so it's not OS X-only; we should probably use it if available.

Should the group set be changed *before* setting the effective user ID?

_______________________________________________

Wireshark-dev mailing list

Wireshark-dev@xxxxxxxxxxxxx

http://www.wireshark.org/mailman/listinfo/wireshark-dev

Regards,

Sebastien Tandel