C compilers may silently discard some wraparound checks
Vulnerability Note VU#162289
Original Release Date: 2008-04-04 | Last Revised: 2008-10-08
Overview
Some C compilers optimize away pointer arithmetic overflow tests that depend on undefined behavior without providing a diagnostic (a warning). Applications containing these tests may be vulnerable to buffer overflows if compiled with these compilers.
Description
In the C language, given the following types:
char *buf; int len;
some C compilers will assume that buf+len >= buf. As a result, code that performs wrapping checks similar to the following:
are optimized out by these compilers; no object code to perform the check will appear in the resulting executable program. In the case where the wrap test expression is optimized out, a subsequent manipulation of lencould cause an overflow. As a result, applications that perform such checks may be vulnerable to buffer overflows.
Wrapping checks that use methods similar to the one described above depend on undefined behavior. Conforming implementations are permitted to perform the optimization by the ISO/IEC 9899:1999 C specification (§6.5.6p8) as undefined behavior. Even if a conforming implementation currently generates object code for an undefined behavior, future versions of the compiler are not obligated to do the same; this behavior may be viewed as an opportunity for further optimization. To ensure that such changes to the compiler do not invalidate assumptions, developers should follow the recommendations described in CERT C Secure Coding recommendation MSC15-C and rule ARR38-C to avoid this error. Furthermore, compilers are not required to issue diagnostics for undefined behavior, so there is frequently no easy way to identify undefined behavior in code, particularly during manual code audits.
Note that this issue does not strictly constitute a vulnerability in the compilers themselves. Rather, this behavior may introduce vulnerabilities in applications that include similar code and are compiled with affected compiler implementations. Existing code that relies on the undefined behavior in the wrapping check is particularly susceptible to this behavior.
Multiple implementations are known to perform this optimization. This optimization may be affected by the setting of the optimization level as well as other flags. Additional information about affected implementations can be found in the Systems Affected section of this document.
Impact
An application that performs wrapping checks based on an expression such as the one described above may be vulnerable to buffer overflow if compiled with affected compiler implementations. The nature of the resulting vulnerability would be specific to the application and depends on how the affected code is used.
Solution
Use casts Cast objects of type char* to uintptr_t before comparison. The faulty wrapping check listed above would be written
Alternatively, developers can use size_t on platforms that do not provide the uintptr_t type. Developers should also follow the recommendations described in CERT C Secure Coding rule ARR38-C.
Avoid affected compiler implementations Application developers and vendors of large codebases that cannot be audited for use of the defective wrapping checks are urged to avoid using compiler implementations that perform the offending optimization. Vendors and developers should carefully evaluate the conditions under which their compiler may perform the offending optimization. In some cases, downgrading the version of the compiler in use or sticking with versions of the compiler that do not perform the offending optimization may mitigate resulting vulnerabilities in applications.
The behaviour of pointer overflow has now changed as of the following (as yet unreleased) versions:
gcc 4.2.4 gcc 4.3.1 gcc 4.4.0
and all subsequent versions (4.2.x where x >= 4, 4.3.y where y >= 1, 4.z where z >= 4).
The optimization under discussion is for comparisons between P + V1 and P + V2, where P is the same pointer and V1 and V2 are variables of some integer type. The C/C++ language standards permit this to be reduced to a comparison between V1 and V2. However, if V1 or V2 are such that the sum with P overflows, then the comparison of V1 and V2 will not yield the same result as actually computing P + V1 and P + V2 and comparing the sums.
The new behaviour as of the above releases is that this optimization is performed by default at -O2 and above, including -Os. It is not performed by default at -O1 or (of course) -O0. The optimization may be enabled for -O1 with the -fstrict-overflow option. The optimization may be disabled for -O2 and above with the -fno-strict-overflow option.
When the optimization is enabled, cases where it occurs may be detected by using -Wstrict-overflow=N where N >= 3. Note that using this warning option is likely to yield a number of false positive reports--cases where this or other overflow optimizations are being applied, but where there is no actual problem.
Please see the gcc manual for more information about these options.
Vendor Information
We are not aware of further vendor information regarding this vulnerability.
Addendum
Vendors and developers using the GNU C compiler should consider using versions of GCC that do not perform the optimization such as those listed above or GCC 4.1.2 or earlier versions (but NOT 4.1.0).
If you have feedback, comments, or additional information about this vulnerability, please send us email.
I tried the given program with our compiler on IA32 linux (Where int and ptr are both 32 bits). I couldn't ever get it to optimize the test away. It may be possible under some circumstances to get this to happen, but the test case sent doesn't show that for me. I tried icc -O1, -O2, and -O3 levels of optimization, and none of them showed the problem.
Vendor Information
We are not aware of further vendor information regarding this vulnerability.
We have run the provided test case on various versions of Sun's C compilers with various levels of optimization and we do not believe that they are impacted by this issue.
Vendor Information
We are not aware of further vendor information regarding this vulnerability.
Felix von Leitner originally published information about this behavior in the gcc compiler. Russ Cox later noticed this behavior as well and provided additional information and assistance.
This document was written by Chad R Dougherty and Robert C Seacord.