Dear Ktha, The problem is different: uml_net shouldn't be suid. http://rhn.redhat.com/errata/RHSA-2003-056.html Details: An updated kernel-utils package is available that removes the setuid bits incorrectly assigned to the uml_net binary. --Saturday, May 24, 2003, 5:10:07 AM, you wrote to bugtraq@securityfocus.com: K> There is a vulnerability in uml_net. The latest version is vulnerable too. K> The problem is the lack of bounds checking in uml_net.c from uml_utilities, K> A possible attack could lead to root compromise on some systems since for K> example uml_net comes suided root in RH 8.0 by default. K> Let's look over the code: K> * int n = 3, v; K> we observe how "v" is declared, as a signed integer. K> with a little bit of "luck" will end up here, knowing that "v" will get K> the value of the first parameter (an integer) given to the program: K> * if(v > CURRENT_VERSION) K> so... if "v" is smaller than 0, the test will be passed but "v" will not K> be between 0 and CURRENT_VERSION as the author would expected to. K> after some tests on the second parameter of the program... we should K> arrive to this code: K> * if(handlers[v] != NULL) (*handlers[v])(argc - n, &argv[n]); K> here it is called the function located at handlers[v]. by suplying a large K> negative integer for "v", handlers[v] can point to the stack which can be K> controlled by the user. so, the program can be fooled to call a user K> controlled "function". K> since the root privileges are not dropped till this point... the user code K> will be executed with super-user privileges. K> Suggested patch: uml_net.c K> - if(v > CURRENT_VERSION){ K> + if ((v > CURRENT_VERSION) || (v < 0)) { K> Contact: ktha@hushmail.com -- ~/ZARAZA You know my name - look up my number (Beatles)
