SSL decryption doesn't work because of p / q parameter swap

This issue was migrated from bug 7388 in our old bug tracker.

Original bug information:

Reporter: Sylvain Munaut
Status: RESOLVED FIXED
Product: Wireshark
Component: GTK+ UI
OS: All
Platform: All
Version: unspecified

Sylvain Munaut said:

Build Information: TShark 1.9.0 (SVN Rev 43405 from /trunk)

Copyright 1998-2012 Gerald Combs <gerald@‍wireshark.org> and contributors. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Compiled (64-bit) with GLib 2.30.3, with libpcap, with libz 1.2.5.1, with POSIX capabilities (Linux), without SMI, without c-ares, without ADNS, with Lua 5.1, without Python, with GnuTLS 2.12.18, with Gcrypt 1.4.6, without Kerberos, without GeoIP.

Running on Linux 3.2.12-gentoo, without locale, with libpcap version 1.1.1, with libz 1.2.5.1.

Built using gcc 4.5.3.ssl_privkey_to_sexp

If you look at packet-ssl-utils.c you can find in ssl_privkey_to_sexp :

---
    /*
     * note: openssl and gnutls use 'p' and 'q' with opposite meaning:
     * our 'p' must be equal to 'q' as provided from openssl and viceversa
     */

#if (LIBGNUTLS_VERSION_MAJOR>2)||((LIBGNUTLS_VERSION_MAJOR==2)&&(LIBGNUTLS_VERSION_MINOR>=5))
    p_idx = 3; q_idx = 4;
#else /* versions 2.4.x and older need 'p' and 'q' swapped */
    p_idx = 4; q_idx = 3;
#endif
----


and a bit later another bit :

---
    ssl_get_version(&major, &minor, &patch);

    /* certain versions of gnutls require swap of rsa params 'p' and 'q' */
    if ((major <= 1) && (minor <= 0) && (patch <= 13))
    {
        gcry_mpi_t tmp;
        ssl_debug_printf("ssl_load_key: swapping p and q parameters\n");
        tmp = rsa_params[4];
        rsa_params[4] = rsa_params[3];
        rsa_params[3] = tmp;
    }
---


Despite all those supposed "fixes", on my laptop, despite using a recent GnuTLS 2.12.18 and Gcrypt 1.4.6, p&q are not in the right order, resulting in bad SSL decryption ... (well, no decryption at all). If I manually force them, then it works fine.

added uigtk label

Christopher Maynard said:

Sylvain, is this bug still relevant? There have been a number of changes in the SSL dissector since then. In particular, the "note" you reference was removed with r45465, so perhaps this has been fixed already?

Sylvain Munaut said:

Huh ... honestly I'm not sure. I'll try to give it a shot during the week with a trunk build.

Balint Reczey said:

Sylvain, I believe the bug has been fixed with r45465. Please reopen the bug if you find it still present in latest master.

closed