[Bug sanitizer/105084] New: ASAN false positive on global-buffer-overflow

[Bug sanitizer/105084] New: ASAN false positive on global-buffer-overflow

[shaohua.li at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105084

            Bug ID: 105084
           Summary: ASAN false positive on global-buffer-overflow
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: sanitizer
          Assignee: unassigned at gcc dot gnu.org
          Reporter: shaohua.li at inf dot ethz.ch
                CC: dodji at gcc dot gnu.org, dvyukov at gcc dot gnu.org,
                    jakub at gcc dot gnu.org, kcc at gcc dot gnu.org, marxin at gcc dot gnu.org
  Target Milestone: ---

Hi,

ASAN seems to falsely report global-buffer-overflow with "-O1" and above. I
tested with gcc trunk, 11.0, 10.0, and 9.0. They all reported the same issue. I
wonder if this behaviour is intended or not. For clang's ASAN, it does not
report anything on the program.

$
$cat a.c
int a[] = {3};
int b = 7;

main() {
    unsigned int *c = &b;
    *c = a[1];
}
$
$gcc -fsanitize=address -w -O0; ./a.out
$
$gcc -fsanitize=address -w -O1; ./a.out
==292501==ERROR: AddressSanitizer: global-buffer-overflow on address
0x55aff123e040 at pc 0x55aff123b218 bp 0x7ffd41151fd0 sp 0x7ffd41151fc0
READ of size 4 at 0x55aff123e040 thread T0
    #0 0x55aff123b217 in main (/shared/oob/debug_wrong3/a.out+0x2217)
    #1 0x7f3933a360b2 in __libc_start_main
(/lib/x86_64-linux-gnu/libc.so.6+0x240b2)
    #2 0x55aff123b10d in _start (/shared/oob/debug_wrong3/a.out+0x210d)

0x55aff123e040 is located 32 bytes to the left of global variable 'ak' defined
in 'simple.c:26:16' (0x55aff123e060) of size 4
0x55aff123e040 is located 28 bytes to the right of global variable 'al' defined
in 'simple.c:27:16' (0x55aff123e020) of size 4
SUMMARY: AddressSanitizer: global-buffer-overflow
(/shared/oob/debug_wrong3/a.out+0x2217) in main
Shadow bytes around the buggy address:
  0x0ab67e23fbb0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  0x0ab67e23fbc0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  0x0ab67e23fbd0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  0x0ab67e23fbe0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  0x0ab67e23fbf0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
=>0x0ab67e23fc00: 00 00 00 00 04 f9 f9 f9[f9]f9 f9 f9 04 f9 f9 f9
  0x0ab67e23fc10: f9 f9 f9 f9 00 00 00 00 00 00 00 00 00 00 00 00
  0x0ab67e23fc20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  0x0ab67e23fc30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  0x0ab67e23fc40: 00 00 00 00 00 00 00 00 00 00 00 f9 f9 f9 f9 f9
  0x0ab67e23fc50: 04 f9 f9 f9 f9 f9 f9 f9 00 00 00 00 00 00 00 00
Shadow byte legend (one shadow byte represents 8 application bytes):

[Bug sanitizer/105084] ASAN false positive on global-buffer-overflow

[shaohua.li at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105084

--- Comment #1 from Li Shaohua <shaohua.li at inf dot ethz.ch> ---
Sorry, the triggering program should be the following:

$cat a.c
int a[] = {3};
int b = 7;

main() {
    unsigned int *c = &b;
    *c = a[-1];
}

[Bug sanitizer/105084] ASAN false positive on global-buffer-overflow

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105084

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Neither is false positive, both are real bugs in your test programs.
The fact that on the second testcase a violation isn't reported at -O0 and is
at -O2 depends on the ordering of the global variables (we emit a then b for
-O0 and b then a for -O2).
We only emit the red zones after each variable, not both before and after each
var, because that would waste too much memory especially in programs that use a
lot of translation units and very few global vars in each.
If all the translation units in a program (or shared library) are compiled with
-fsanitize=address, the net effect is that underflows for the first variable
(probably in each section variables are present) aren't detected, when mixing
-fsanitize=address and non-sanitized objects it can result in more underflows
not being detected.

[Bug sanitizer/105084] ASAN false positive on global-buffer-overflow

[shaohua.li at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105084

--- Comment #3 from Li Shaohua <shaohua.li at inf dot ethz.ch> ---
Thanks a lot for your explanation. That makes sense to me :).

[Bug sanitizer/105084] ASAN false positive on global-buffer-overflow

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105084

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |INVALID
             Status|UNCONFIRMED                 |RESOLVED

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
.