转载自：https://www.anquanke.com/post/id/83682

0x0  漏洞信息

影响所有Nexus手机和部分Android手机的漏洞,Google于2016/03/18发布了公告修复,具体请看链接.

http://www.cvedetails.com/cve-details.php?t=1&cve_id=cve-2015-1805X

http://source.android.com/security/advisory/2016-03-18.html

0x1  漏洞描述

在linux 内核3.16版本之前的fs/pipe.c当中,由于pipe_read和pipe_write没有考虑到拷贝过程中数据没有同步的一些临界情况,造成了拷贝越界的问题,因此有可能导致系统crash以及系统权限提升.这种漏洞又称之为” I/O vector array overrun”

0x2  代码分析

//摘自fs/pipe.c:

static ssize_t
pipe_read(struct kiocb *iocb, const struct iovec *_iov,
       unsigned long nr_segs, loff_t pos)
{
    struct file *filp = iocb->ki_filp;
    struct pipe_inode_info *pipe = filp->private_data;
    int do_wakeup;
    ssize_t ret;
    struct iovec *iov = (struct iovec *)_iov;
    size_t total_len;
    total_len = iov_length(iov, nr_segs);
    /* Null read succeeds. */
    if (unlikely(total_len == 0))
        return 0;
    do_wakeup = 0;
    ret = 0;
    __pipe_lock(pipe);
    for (;;) {
        int bufs = pipe->nrbufs;
        if (bufs) {
             int curbuf = pipe->curbuf;
             struct pipe_buffer *buf = pipe->bufs + curbuf;
             const struct pipe_buf_operations *ops = buf->ops;
             void *addr;
             size_t chars = buf->len;
             int error, atomic;
             if (chars > total_len)
                 chars = total_len;
             error = ops->confirm(pipe, buf);
             if (error) {
                 if (!ret)
                     ret = error;
                 break;
             }
             //(1)
             atomic = !iov_fault_in_pages_write(iov, chars);
redo:
             addr = ops->map(pipe, buf, atomic);
             //(2)
             error = pipe_iov_copy_to_user(iov, addr + buf->offset, chars, atomic);
             ops->unmap(pipe, buf, addr);
             if (unlikely(error)) {
                 /*
                  * Just retry with the slow path if we failed.
                  */
                 //(3)
                 if (atomic) {
                     atomic = 0;
                     goto redo;
                 }
                 if (!ret)
                     ret = error;
                 break;
             }
             ret += chars;
             buf->offset += chars;
             buf->len -= chars;
             /* Was it a packet buffer? Clean up and exit */
             if (buf->flags & PIPE_BUF_FLAG_PACKET) {
                 total_len = chars;
                 buf->len = 0;
             }
             if (!buf->len) {
                 buf->ops = NULL;
                 ops->release(pipe, buf);
                 curbuf = (curbuf + 1) & (pipe->buffers - 1);
                 pipe->curbuf = curbuf;
                 pipe->nrbufs = --bufs;
                 do_wakeup = 1;
             }

             (5)//在这里更新total_len

            

total_len -= chars;
             if (!total_len)
                 break;  /* common path: read succeeded */
        }
        if (bufs)    /* More to do? */
             continue;
        if (!pipe->writers)
             break;
        if (!pipe->waiting_writers) {
             /* syscall merging: Usually we must not sleep
              * if O_NONBLOCK is set, or if we got some data.
              * But if a writer sleeps in kernel space, then
              * we can wait for that data without violating POSIX.
              */
             if (ret)
                 break;
             if (filp->f_flags & O_NONBLOCK) {
                 ret = -EAGAIN;
                 break;
             }
        }
        if (signal_pending(current)) {
             if (!ret)
                 ret = -ERESTARTSYS;
             break;
        }
        if (do_wakeup) {
             wake_up_interruptible_sync_poll(&pipe->wait, POLLOUT | POLLWRNORM);
             kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
        }
        pipe_wait(pipe);
    }
    __pipe_unlock(pipe);
    /* Signal writers asynchronously that there is more room. */
    if (do_wakeup) {
        wake_up_interruptible_sync_poll(&pipe->wait, POLLOUT | POLLWRNORM);
        kill_fasync(&pipe->fasync_writers, SIGIO, POLL_OUT);
    }
    if (ret > 0)
        file_accessed(filp);
    return ret;
}

(1).首先pipe_read()函数会先循环读取iovec结构,并且通过iov_fault_in_pages_write()函数判断iov->len是否大于0,且iov->base指向的地址是否可写且处于用户态,之后返回atomic.

(2)如果atomic=1,则pipe_iov_copy_to_user -> __copy_to_user_inatomic ->

__copy_to_user_nocheck;如果atomic=0,则pipe_iov_copy_to_user -> copy_to_user -> access_ok.

(3).如果atomic为1,pipe_iov_copy_to_user拷贝出现错误,会进入redo的逻辑,将再次调用pipe_iov_copy_to_user函数进行拷贝,且将atomic置为0.但是pipe_iov_copy_to_user的第三个参数chars并没有更新,还是会拷贝total_len大小的数据

static int
pipe_iov_copy_to_user(struct iovec *iov, const void *from, unsigned long len,
              int atomic)
{
    unsigned long copy;
    while (len > 0)
    {
        while (!iov->iov_len)
             iov++;
        copy = min_t(unsigned long, len, iov->iov_len);
        if (atomic)
        {
             if (__copy_to_user_inatomic(iov->iov_base, from, copy))
                 //(4)
                 return -EFAULT;
        }
        else
        {
             if (copy_to_user(iov->iov_base, from, copy))
                 //(4)
                 return -EFAULT;
        }
        from += copy;
        len -= copy;
        iov->iov_base += copy;
        //每次对iov->iov_len进行更新
        iov->iov_len -= copy;
    }
    return 0;
}

4. 如果copy到某种情况出错返回,已经copy成功的iov->len会被减去但总长度total_len并不会同步减去.也就是说如果total_len是0x100,第一次消耗掉了x;再次进入redo逻辑后还是0x100,然而实际已经被消耗掉了x.

0x3  具体探究

假设有一个iov结构,total_len为0x40,len为0x20.

iov[0]: iov_base = 0xdead0000 iov_len = 0x10

iov[1]: iov_base = 0xdead1000 iov_len = 0x10

iov[2]: iov_base = 0xdead2000 iov_len = 0x10

iov[3]: iov_base = 0xdead3000 iov_len = 0x10

如果iov[1].iov_base的地址被设置成不可写入.那么第一次pipe_iov_copy_to_user()会返回失败.而iov->iov_base += copy,iov->iov_len -= copy.

iov[0]: iov_base = 0xdead0010 iov_len = 0

iov[1]: iov_base = 0xdead1000 iov_len = 0x10

iov[2]: iov_base = 0xdead2000 iov_len = 0x10

iov[3]: iov_base = 0xdead3000 iov_len = 0x10

现在,redo的逻辑发生在0xdead0010,它以某种方式被设置成可写,并且len仍未0x20.那么iov[1]和iov[2]都将被用掉.

iov[0]: iov_base = 0xdead0010 iov_len = 0

iov[1]: iov_base = 0xdead1010 iov_len = 0

iov[2]: iov_base = 0xdead2010 iov_len = 0

iov[3]: iov_base = 0xdead3000 iov_len = 0x10

在注释(5)中,根据total_len -= chars;那么total_len的大小就被设置为0x20(0x40 -0x20).如果total_len变为了0x20,可我们iov[3]的大小只有0x10.这就会导致pipe_iov_copy_to_user()函数有可能读取到一个未知的iov[4].具体来查看下代码

static int iov_fault_in_pages_write(struct iovec *iov, unsigned long len)
{
    //(6)
    while (!iov->iov_len)
        iov++;
    while (len > 0) {
        unsigned long this_len;
        this_len = min_t(unsigned long, len, iov->iov_len);
        if (fault_in_pages_writeable(iov->iov_base, this_len))
             break;
        len -= this_len;
        iov++;
    }
    return len;
}
static inline int fault_in_pages_writeable(char __user *uaddr, int size)
{
        int ret;
        if (unlikely(size == 0))
                return 0;
        /*
         * Writing zeroes into userspace here is OK, because we know that if
         * the zero gets there, we'll be overwriting it.
        */
        ret = __put_user(0, uaddr);
        if (ret == 0) {
                char __user *end = uaddr + size - 1;
                /*
                * If the page was already mapped, this will get a cache miss
                 * for sure, so try to avoid doing it.
                 */
                if (((unsigned long)uaddr & PAGE_MASK) !=
                                ((unsigned long)end & PAGE_MASK))
                        ret = __put_user(0, end);
       }
        return ret;
}

在iov_fault_in_pages_write()函数中的注释(6),也就意味着iov[0],iov[1],iov[2]都会被跳过,iov[3]被用掉.之后len -= this_len;len被设置为0x10.iov的指针将指向一块未知的内存区域.iov[4].iov_base将被__put_user使用.

0x4  如何利用

核心的思路就是想办法触发redo的逻辑,之后精心构造一个readv()调用.把payload结构定义在已经被校验过的iov数组后,让它成为__put_user()等函数调用的目标地址.如果我们再以某种方式让构造的slab结构在iov数组后包含一个函数指针,让它指向要写的内核地址.

1.第一次循环要保证pipe_iov_copy_to_user()函数失败,这样会进入redo逻辑

2.第二次要保证pipe_iov_copy_to_user()成功,但是不能在这里overrun,否则会走向copy_to_user,要校验地址,所以还是无法写内核地址

3.当iov->len走完之后,total_len还有剩余,所以第三次循环的时候,atomic=1.可以overrun触发

4.第一次要保证失败,也就是说需要把iov_base的地址设置成不可写,第二次要成功,就要保证iov_base的地址有效.所以这里可以通过创建竞争关系的线程,调用mmap/munmap等函数来实现.

0x5  POC

我测试的Nexus 6p  6.0.1系统会crash掉.

Talk is cheap,show me the code…

漏洞验证资源：https://download.csdn.net/download/qq_35559358/10420702

漏洞验证代码