3

I am trying to figure out if I'm having an idiot moment or if there really is a memory leak in libpcap. I'm running Ubuntu 17.10 and libpcap 1.8.1-5ubuntu1. It seems unlikely that such a mature library would have a leak.

I've cut out everything to make a MVCE, so as a consequence, this code doesn't really do much of anything except demonstrate the leak:

#include <stdio.h>
#include <stdlib.h>
#include <pcap.h>

int main(int argc, char **argv)
{
        char errbuf[PCAP_ERRBUF_SIZE];

        pcap_t *fd = pcap_open_offline(argv[1], errbuf);

        if (!fd) {
                printf("error: %s\n", errbuf);
        }

        free(fd); fd = 0;

        return 0;
}

The valgrind report (emphasis added):

==6871==
==6871== HEAP SUMMARY:
==6871==     in use at exit: 262,696 bytes in 2 blocks
==6871==   total heap usage: 4 allocs, 2 frees, 267,432 bytes allocated
==6871==
==6871== Searching for pointers to 2 not-freed blocks
==6871== Checked 73,072 bytes
==6871==
==6871== 262,144 bytes in 1 blocks are definitely lost in loss record 2 of 2
==6871==    at 0x4C2FB0F: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==6871==    by 0x4E5B89F: ??? (in /usr/lib/x86_64-linux-gnu/libpcap.so.1.8.1)
==6871==    by 0x4E5AE5C: pcap_fopen_offline_with_tstamp_precision (in /usr/lib/x86_64-linux-gnu/libpcap.so.1.8.1)
==6871==    by 0x4E5B05D: pcap_open_offline_with_tstamp_precision (in /usr/lib/x86_64-linux-gnu/libpcap.so.1.8.1)
==6871==    by 0x1087A0: main (test.c:9)
==6871==
==6871== LEAK SUMMARY:
==6871==    definitely lost: 262,144 bytes in 1 blocks
==6871==    indirectly lost: 0 bytes in 0 blocks
==6871==      possibly lost: 0 bytes in 0 blocks
==6871==    still reachable: 552 bytes in 1 blocks
==6871==         suppressed: 0 bytes in 0 blocks
==6871== Reachable blocks (those to which a pointer was found) are not shown.
==6871== To see them, rerun with: --leak-check=full --show-leak-kinds=all
==6871==
==6871== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)
==6871== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)
Greg Schmit
  • 4,275
  • 2
  • 21
  • 36

1 Answers1

4

From the libpcap manpage, edited slightly:

pcap_fopen_offline() return[s] a pointer to a pcap_t, which is the handle used for reading packets… To close a handle, use pcap_close().

free(fd) just frees a single block of memory, since free() knows nothing about the internals of a pcap_t. In order to properly dispose of allocated resources, you need to use pcap_close(fd) instead, as the documentation indicates.

rici
  • 234,347
  • 28
  • 237
  • 341
  • Ah, thanks! I was looking at the man page of `pcap_open_offline()` (https://www.tcpdump.org/manpages/pcap_open_offline.3pcap.html) and it didn't mention that. – Greg Schmit Nov 16 '18 at 00:26