blob: c860689271aba7b41e7d6f69b132c7bff80b86f5 (
plain) (
blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
|
Stack address is returned from function translate_one
The vulnerable version is qemu-2.8.0, and the vulnerable function is in "target-s390x/translate.c".
The code snippet is as following.
static ExitStatus translate_one(CPUS390XState *env, DisasContext *s)
{
const DisasInsn *insn;
ExitStatus ret = NO_EXIT;
DisasFields f;
...
s->fields = &f;
...
s->pc = s->next_pc;
return ret;
}
A stack address, i.e. the address of local variable "f" is returned from current function through the output parameter "s->fields" as a side effect.
This issue is one kind of undefined behaviors, according the C Standard, 6.2.4 [ISO/IEC 9899:2011] (https://www.securecoding.cert.org/confluence/display/c/DCL30-C.+Declare+objects+with+appropriate+storage+durations)
This dangerous defect may lead to an exploitable vulnerability.
We suggest sanitizing "s->fields" as null before return.
Note that this issue is reported by shqking and Zhenwei Zou together.
The calling function never uses "->fields", so I do not see a real vulnerability here, is there? Did you use a code analyser for this, or how did you come across this issue?
Thanks for your reply.
Inspired by this issue in apache httpd (https://bz.apache.org/bugzilla/show_bug.cgi?id=59844#c0),
we customized a checker based on the Clang Static Analyzer to detect such undefined behavior.
Yes.
After examining the code carefully, we didn't find any place where the "->fields" is accessed, either. However, we think this kind of defect seems like a 'time bomb' and we'd better fix it just to be on the safe side.
I've finally posted a patch for this:
https://lists.gnu.org/archive/html/qemu-devel/2020-01/msg05204.html
Fixed here:
https://git.qemu.org/?p=qemu.git;a=commitdiff;h=344a7f656e8d211cdd6e
|