github / github/codeql

[C++] Assigning to function pointer in a function appears to defeat dispatch analysis

Open
#18,103 2 comments 0 reactions 0 assignees View on GitHub
question
Dominant language
CodeQL
Stars
10.1k
Forks
2.1k
Avg merge
2d 15h
Merged PRs (30d)
141

Description

Assigning to a function pointer in a function appears to defeat the points-to analysis used to resolve the dispatch of function pointers.

In the following example I would have expected to see two flows from `source()` to `target()` but only the second one from the direct assignment of the function pointer is reported.

```cpp
int source()
{
return 2;
}

int a_function()
{
return source();
}

int target(int source)
{
return source;
}

void set(int (**ptr)(), int (*ptr2)())
{
*ptr = ptr2;
}

int main(int argv, char **argc)
{
int (*fptr)();

set(&fptr, a_function);

target(fptr()); // not detected as source

fptr = a_function;

target(fptr()); // detected

return 0;
}
```

This is the complete query

```ql
import cpp
import semmle.code.cpp.dataflow.new.TaintTracking

module SourceSinkCallConfig implements DataFlow::ConfigSig {
predicate isSource(DataFlow::Node source) {
source.asExpr().(FunctionCall).getTarget().getName() = "source"
}

predicate isSink(DataFlow::Node sink) {
exists(Call call |
call.getTarget().getName() = "target" and
call.getArgument(0) = sink.asExpr()
)
}
}

module SourceSinkCallTaint = TaintTracking::Global;

from DataFlow::Node source, DataFlow::Node sink, int source_line, int sink_line
where
SourceSinkCallTaint::flow(source, sink) and
source_line = source.getLocation().getStartLine() and
sink_line = sink.getLocation().getStartLine()
select source, source_line, sink, sink_line

```

This is the output. I would have expected to also see a flow to line 28.

```
| source | source_line | sink | sink_line |
+----------------+-------------+--------------------+-----------+
| call to source | 9 | call to expression | 32 |
```

CodeQL version: 2.19.3

Contributor guide

Open the contributing guide

Research direction

Start by running the complete query from the issue against the supplied C++ example on CodeQL 2.19.3. Read the C++ data-flow and points-to behavior behind the imported TaintTracking library, then verify that the indirect assignment through set() is resolved. Done means the query reports the expected additional flow to line 28.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
devtools
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
42/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.