Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

"Pointer is not a function pointer" #686

Closed
weaversa opened this issue Apr 23, 2020 · 10 comments
Closed

"Pointer is not a function pointer" #686

weaversa opened this issue Apr 23, 2020 · 10 comments
Assignees
Labels
subsystem: crucible-llvm Issues related to LLVM bitcode verification with crucible-llvm topics: error-messages Issues involving the messages SAW produces on error usability An issue that impedes efficient understanding and use
Milestone

Comments

@weaversa
Copy link
Contributor

#include <stdint.h>
#include <string.h>
#include <stdlib.h>

uint32_t f() {
  uint32_t *x	= malloc(1 * sizeof(uint32_t));
  memset(x, 0, 1 * sizeof(uint32_t));
  uint32_t y = x[0];
  free(x);
  return y;
}
let f_spec = do {
  crucible_execute_func [];
  crucible_return (crucible_term {{ 0 : [32] }});
};

test_bc <- llvm_load_module "test.bc";

f_result <- crucible_llvm_verify test_bc "f" [] true f_spec z3;
$ clang -g -O0 test.c -o test.bc -emit-llvm -c && saw test.saw
[00:39:45.262] Loading file "test.saw"
[00:39:45.287] Verifying f ...
[00:39:45.287] Simulating f ...
[00:39:45.290] "crucible_llvm_verify" (test.saw:8:13-8:33):
Symbolic execution failed.
Abort due to false assumption:
  Pointer is not a function pointer:
(4, 0x0:[64])
  in f at test.c:7:3
Stack frame
  Allocations:
    HeapAlloc 9 0x4:[64] Mutable 16-byte-aligned test.c:6:17
    StackAlloc 8 0x4:[64] Mutable 4-byte-aligned "test.c:8:12"
    StackAlloc 7 0x8:[64] Mutable 8-byte-aligned "test.c:6:13"
  Writes:
    
Base memory
  Allocations:
    GlobalAlloc 6 0x0:[64] Immutable 1-byte-aligned f
    GlobalAlloc 5 0x0:[64] Immutable 1-byte-aligned free
    GlobalAlloc 4 0x0:[64] Immutable 1-byte-aligned llvm.objectsize.i64.p0i8
    GlobalAlloc 3 0x0:[64] Immutable 1-byte-aligned __memset_chk
    GlobalAlloc 2 0x0:[64] Immutable 1-byte-aligned malloc
    GlobalAlloc 1 0x0:[64] Immutable 1-byte-aligned llvm.dbg.declare
  Writes:

What function pointer?

I've only gotten this error when using memset, and only when passing in an array to f or mallocing an array (as I did above). If I give a stack allocated array uint32_t x[1] to memset then saw does not complain.

Also, there is a printing inconsistency --- HeapAlloc does not put quotes around the filename whereas StackAlloc does, and the stuff at the beginning uses parens around the filename.

@atomb
Copy link
Contributor

atomb commented Apr 23, 2020

We should really improve the error message here. What it should be is something like call to undefined external function llvm.objectsize.i64.p0i8. In the meantime, the trick to tracking this down is that the pointer (4, 0x0:[64]) means "offset 0 from allocation ID 4". Down in the "Allocations" section below you can see that GlobalAlloc 4 is the global variable containing the pointer to the function llvm.objectsize.i64.p0i8. I'm wondering whether that LLVM intrinsic has changed types again in recent LLVM versions. We have two different types for it in our LLVM support modules already, because it changed types once in the past.

@weaversa
Copy link
Contributor Author

I'm having the same problem with memcpy.

@atomb
Copy link
Contributor

atomb commented Apr 24, 2020

Which version of Clang are you using?

@weaversa
Copy link
Contributor Author

I just updated to the latest xcode...so you're probably right that the latest LLVM has some tweaks that break things.

$ clang --version
Apple clang version 11.0.3 (clang-1103.0.32.59)
Target: x86_64-apple-darwin19.4.0
Thread model: posix
InstalledDir: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin

@atomb
Copy link
Contributor

atomb commented Apr 27, 2020

I've just confirmed that I can replicate this using the latest Xcode. I think it should be easy to fix.

@atomb atomb self-assigned this Apr 27, 2020
@langston-barrett langston-barrett added the subsystem: crucible-llvm Issues related to LLVM bitcode verification with crucible-llvm label Apr 27, 2020
@atomb
Copy link
Contributor

atomb commented Apr 28, 2020

The specific example listed here works now, but the error message is still misleading.

@atomb atomb added the usability An issue that impedes efficient understanding and use label Apr 28, 2020
@atomb atomb added this to the 0.6 milestone May 5, 2020
@robdockins robdockins self-assigned this May 15, 2020
@atomb atomb added the topics: error-messages Issues involving the messages SAW produces on error label May 15, 2020
@robdockins
Copy link
Contributor

The error message should get the metadata about the pointer, so we can at least see what the symbol we tried to call was.

@atomb atomb modified the milestones: 0.6, 0.7 Aug 31, 2020
@robdockins
Copy link
Contributor

Current status: the specific problem that started this thread has been solved. If I deliberately call an external function that does not have an override, I get a message saying "no implementation or override found", and a useful source location, but it still doesn't tell me what function it tried to call.

@robdockins
Copy link
Contributor

Some fixes to the pretty printers in GaloisInc/crucible#574 should fix this bug.

@atomb
Copy link
Contributor

atomb commented Dec 14, 2020

The error message seems good now.

@atomb atomb closed this as completed Dec 14, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
subsystem: crucible-llvm Issues related to LLVM bitcode verification with crucible-llvm topics: error-messages Issues involving the messages SAW produces on error usability An issue that impedes efficient understanding and use
Projects
None yet
Development

No branches or pull requests

4 participants