I have a daemon which gets started as root (so it can bind to low ports). After initialisation I'd very much like to have it drop root privileges for safety reasons.
Can anyone point me at a known correct piece of code in C which will do this?
I've read the man pages, I've looked at various implementations of this in different applications, and they're all different, and some of them are really complex. This is security-related code, and I really don't want to reinvent the same mistakes that other people are making. What I'm looking for is a best practice, known good, portable library function that I can use in the knowledge that it's going to get it right. Does such a thing exist?
For reference: I'm starting as root; I need to change to run under a different uid and gid; I need to have the supplementary groups set up correctly; I don't need to change back to root privileges afterwards.
The first of two functions, spc_drop_privileges( ) drops any extra group or user privileges either permanently or temporarily, depending on the value of its only argument. If a nonzero value is passed, privileges will be dropped permanently; otherwise, the privilege drop is temporary.
Root privileges are the powers that the root account has on the system. The root account is the most privileged on the system and has absolute power over it (i.e., complete access to all files and commands).
Edit /etc/passwd for the particular user. Change the user's UID and GID to '0'. This will give root permissions to user.
SETUID 0 or root means that when the program is run it is as if root ran it - which leaves you open to all the usual security risks of running something at high permission levels - it could potentially do anything to your system - so generally the number of applications that require SETUID should be minimised on a Unix ...
In order to drop all privileges (user and group), you need to drop the group before the user. Given that userid
and groupid
contains the IDs of the user and the group you want to drop to, and assuming that the effective IDs are also root, this is accomplished by calling setuid() and setgid():
if (getuid() == 0) { /* process is running as root, drop privileges */ if (setgid(groupid) != 0) fatal("setgid: Unable to drop group privileges: %s", strerror(errno)); if (setuid(userid) != 0) fatal("setuid: Unable to drop user privileges: %S", strerror(errno)); }
If you are paranoid, you can try to get your root privileges back, which should fail. If it doesn't fail, you bailout:
if (setuid(0) != -1) fatal("ERROR: Managed to regain root privileges?");
Also, if you are still paranoid, you may want to seteuid() and setegid() too, but it shouldn't be necessary, since setuid() and setgid() already set all the IDs if the process is owned by root.
The supplementary group list is a problem, because there is no POSIX function to set supplementary groups (there is getgroups(), but no setgroups()). There is a BSD and Linux extension setgroups() that you can use, it this concerns you.
You should also chdir("/")
or to any other directory, so that the process doesn't remain in a root-owned directory.
Since your question is about Unix in general, this is the very general approach. Note that in Linux this is no longer the preferred approach. In current Linux versions, you should set the CAP_NET_BIND_SERVICE
capability on the executable, and run it as a normal user. No root access is needed.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With