[linux-2.6-block.git] / mm / usercopy.c

/*
 * This implements the various checks for CONFIG_HARDENED_USERCOPY*,
 * which are designed to protect kernel memory from needless exposure
 * and overwrite under many unintended conditions. This code is based
 * on PAX_USERCOPY, which is:
 *
 * Copyright (C) 2001-2016 PaX Team, Bradley Spengler, Open Source
 * Security Inc.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 *
 */
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/sched.h>
#include <linux/sched/task.h>
#include <linux/sched/task_stack.h>
#include <linux/thread_info.h>
#include <asm/sections.h>

/*
 * Checks if a given pointer and length is contained by the current
 * stack frame (if possible).
 *
 * Returns:
 *	NOT_STACK: not at all on the stack
 *	GOOD_FRAME: fully within a valid stack frame
 *	GOOD_STACK: fully on the stack (when can't do frame-checking)
 *	BAD_STACK: error condition (invalid stack position or bad stack frame)
 */
static noinline int check_stack_object(const void *obj, unsigned long len)
{
	const void * const stack = task_stack_page(current);
	const void * const stackend = stack + THREAD_SIZE;
	int ret;

	/* Object is not on the stack at all. */
	if (obj + len <= stack || stackend <= obj)
		return NOT_STACK;

	/*
	 * Reject: object partially overlaps the stack (passing the
	 * the check above means at least one end is within the stack,
	 * so if this check fails, the other end is outside the stack).
	 */
	if (obj < stack || stackend < obj + len)
		return BAD_STACK;

	/* Check if object is safely within a valid frame. */
	ret = arch_within_stack_frames(stack, stackend, obj, len);
	if (ret)
		return ret;

	return GOOD_STACK;
}

/*
 * If this function is reached, then CONFIG_HARDENED_USERCOPY has found an
 * unexpected state during a copy_from_user() or copy_to_user() call.
 * There are several checks being performed on the buffer by the
 * __check_object_size() function. Normal stack buffer usage should never
 * trip the checks, and kernel text addressing will always trip the check.
 * For cache objects, copies must be within the object size.
 */
void __noreturn usercopy_abort(const char *name, const char *detail,
			       bool to_user, unsigned long offset,
			       unsigned long len)
{
	pr_emerg("Kernel memory %s attempt detected %s %s%s%s%s (offset %lu, size %lu)!\n",
		 to_user ? "exposure" : "overwrite",
		 to_user ? "from" : "to",
		 name ? : "unknown?!",
		 detail ? " '" : "", detail ? : "", detail ? "'" : "",
		 offset, len);

	/*
	 * For greater effect, it would be nice to do do_group_exit(),
	 * but BUG() actually hooks all the lock-breaking and per-arch
	 * Oops code, so that is used here instead.
	 */
	BUG();
}

/* Returns true if any portion of [ptr,ptr+n) over laps with [low,high). */
static bool overlaps(const void *ptr, unsigned long n, unsigned long low,
		     unsigned long high)
{
	unsigned long check_low = (uintptr_t)ptr;
	unsigned long check_high = check_low + n;

	/* Does not overlap if entirely above or entirely below. */
	if (check_low >= high || check_high <= low)
		return false;

	return true;
}

/* Is this address range in the kernel text area? */
static inline const char *check_kernel_text_object(const void *ptr,
						   unsigned long n)
{
	unsigned long textlow = (unsigned long)_stext;
	unsigned long texthigh = (unsigned long)_etext;
	unsigned long textlow_linear, texthigh_linear;

	if (overlaps(ptr, n, textlow, texthigh))
		return "<kernel text>";

	/*
	 * Some architectures have virtual memory mappings with a secondary
	 * mapping of the kernel text, i.e. there is more than one virtual
	 * kernel address that points to the kernel image. It is usually
	 * when there is a separate linear physical memory mapping, in that
	 * __pa() is not just the reverse of __va(). This can be detected
	 * and checked:
	 */
	textlow_linear = (unsigned long)lm_alias(textlow);
	/* No different mapping: we're done. */
	if (textlow_linear == textlow)
		return NULL;

	/* Check the secondary mapping... */
	texthigh_linear = (unsigned long)lm_alias(texthigh);
	if (overlaps(ptr, n, textlow_linear, texthigh_linear))
		return "<linear kernel text>";

	return NULL;
}

static inline const char *check_bogus_address(const void *ptr, unsigned long n)
{
	/* Reject if object wraps past end of memory. */
	if ((unsigned long)ptr + n < (unsigned long)ptr)
		return "<wrapped address>";

	/* Reject if NULL or ZERO-allocation. */
	if (ZERO_OR_NULL_PTR(ptr))
		return "<null>";

	return NULL;
}

/* Checks for allocs that are marked in some way as spanning multiple pages. */
static inline const char *check_page_span(const void *ptr, unsigned long n,
					  struct page *page, bool to_user)
{
#ifdef CONFIG_HARDENED_USERCOPY_PAGESPAN
	const void *end = ptr + n - 1;
	struct page *endpage;
	bool is_reserved, is_cma;

	/*
	 * Sometimes the kernel data regions are not marked Reserved (see
	 * check below). And sometimes [_sdata,_edata) does not cover
	 * rodata and/or bss, so check each range explicitly.
	 */

	/* Allow reads of kernel rodata region (if not marked as Reserved). */
	if (ptr >= (const void *)__start_rodata &&
	    end <= (const void *)__end_rodata) {
		if (!to_user)
			return "<rodata>";
		return NULL;
	}

	/* Allow kernel data region (if not marked as Reserved). */
	if (ptr >= (const void *)_sdata && end <= (const void *)_edata)
		return NULL;

	/* Allow kernel bss region (if not marked as Reserved). */
	if (ptr >= (const void *)__bss_start &&
	    end <= (const void *)__bss_stop)
		return NULL;

	/* Is the object wholly within one base page? */
	if (likely(((unsigned long)ptr & (unsigned long)PAGE_MASK) ==
		   ((unsigned long)end & (unsigned long)PAGE_MASK)))
		return NULL;

	/* Allow if fully inside the same compound (__GFP_COMP) page. */
	endpage = virt_to_head_page(end);
	if (likely(endpage == page))
		return NULL;

	/*
	 * Reject if range is entirely either Reserved (i.e. special or
	 * device memory), or CMA. Otherwise, reject since the object spans
	 * several independently allocated pages.
	 */
	is_reserved = PageReserved(page);
	is_cma = is_migrate_cma_page(page);
	if (!is_reserved && !is_cma)
		return "<spans multiple pages>";

	for (ptr += PAGE_SIZE; ptr <= end; ptr += PAGE_SIZE) {
		page = virt_to_head_page(ptr);
		if (is_reserved && !PageReserved(page))
			return "<spans Reserved and non-Reserved pages>";
		if (is_cma && !is_migrate_cma_page(page))
			return "<spans CMA and non-CMA pages>";
	}
#endif

	return NULL;
}

static inline const char *check_heap_object(const void *ptr, unsigned long n,
					    bool to_user)
{
	struct page *page;

	if (!virt_addr_valid(ptr))
		return NULL;

	page = virt_to_head_page(ptr);

	/* Check slab allocator for flags and size. */
	if (PageSlab(page))
		return __check_heap_object(ptr, n, page);

	/* Verify object does not incorrectly span multiple pages. */
	return check_page_span(ptr, n, page, to_user);
}

/*
 * Validates that the given object is:
 * - not bogus address
 * - known-safe heap or stack object
 * - not in kernel text
 */
void __check_object_size(const void *ptr, unsigned long n, bool to_user)
{
	const char *err;

	/* Skip all tests if size is zero. */
	if (!n)
		return;

	/* Check for invalid addresses. */
	err = check_bogus_address(ptr, n);
	if (err)
		goto report;

	/* Check for bad heap object. */
	err = check_heap_object(ptr, n, to_user);
	if (err)
		goto report;

	/* Check for bad stack object. */
	switch (check_stack_object(ptr, n)) {
	case NOT_STACK:
		/* Object is not touching the current process stack. */
		break;
	case GOOD_FRAME:
	case GOOD_STACK:
		/*
		 * Object is either in the correct frame (when it
		 * is possible to check) or just generally on the
		 * process stack (when frame checking not available).
		 */
		return;
	default:
		err = "<process stack>";
		goto report;
	}

	/* Check for object in kernel to avoid text exposure. */
	err = check_kernel_text_object(ptr, n);
	if (!err)
		return;

report:
	usercopy_abort(err, NULL, to_user, 0, n);
}
EXPORT_SYMBOL(__check_object_size);
Commit	Line	Data
f5509cc1 KC	1	/*
	2	* This implements the various checks for CONFIG_HARDENED_USERCOPY*,
	3	* which are designed to protect kernel memory from needless exposure
	4	* and overwrite under many unintended conditions. This code is based
	5	* on PAX_USERCOPY, which is:
	6	*
	7	* Copyright (C) 2001-2016 PaX Team, Bradley Spengler, Open Source
	8	* Security Inc.
	9	*
	10	* This program is free software; you can redistribute it and/or modify
	11	* it under the terms of the GNU General Public License version 2 as
	12	* published by the Free Software Foundation.
	13	*
	14	*/
	15	#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
	16
	17	#include <linux/mm.h>
	18	#include <linux/slab.h>
5b825c3a	19	#include <linux/sched.h>
29930025 IM	20	#include <linux/sched/task.h>
29930025 IM	21	#include <linux/sched/task_stack.h>
96dc4f9f	22	#include <linux/thread_info.h>
f5509cc1 KC	23	#include <asm/sections.h>
f5509cc1 KC	24
f5509cc1 KC	25	/*
	26	* Checks if a given pointer and length is contained by the current
	27	* stack frame (if possible).
	28	*
	29	* Returns:
	30	* NOT_STACK: not at all on the stack
	31	* GOOD_FRAME: fully within a valid stack frame
	32	* GOOD_STACK: fully on the stack (when can't do frame-checking)
	33	* BAD_STACK: error condition (invalid stack position or bad stack frame)
	34	*/
	35	static noinline int check_stack_object(const void *obj, unsigned long len)
	36	{
	37	const void * const stack = task_stack_page(current);
	38	const void * const stackend = stack + THREAD_SIZE;
	39	int ret;
	40
	41	/* Object is not on the stack at all. */
	42	if (obj + len <= stack \|\| stackend <= obj)
	43	return NOT_STACK;
	44
	45	/*
	46	* Reject: object partially overlaps the stack (passing the
	47	* the check above means at least one end is within the stack,
	48	* so if this check fails, the other end is outside the stack).
	49	*/
	50	if (obj < stack \|\| stackend < obj + len)
	51	return BAD_STACK;
	52
	53	/* Check if object is safely within a valid frame. */
	54	ret = arch_within_stack_frames(stack, stackend, obj, len);
	55	if (ret)
	56	return ret;
	57
	58	return GOOD_STACK;
	59	}
	60
b394d468 KC	61	/*
	62	* If this function is reached, then CONFIG_HARDENED_USERCOPY has found an
	63	* unexpected state during a copy_from_user() or copy_to_user() call.
	64	* There are several checks being performed on the buffer by the
	65	* __check_object_size() function. Normal stack buffer usage should never
	66	* trip the checks, and kernel text addressing will always trip the check.
	67	* For cache objects, copies must be within the object size.
	68	*/
	69	void __noreturn usercopy_abort(const char name, const char detail,
	70	bool to_user, unsigned long offset,
	71	unsigned long len)
f5509cc1	72	{
b394d468 KC	73	pr_emerg("Kernel memory %s attempt detected %s %s%s%s%s (offset %lu, size %lu)!\n",
	74	to_user ? "exposure" : "overwrite",
	75	to_user ? "from" : "to",
	76	name ? : "unknown?!",
	77	detail ? " '" : "", detail ? : "", detail ? "'" : "",
	78	offset, len);
	79
f5509cc1 KC	80	/*
	81	* For greater effect, it would be nice to do do_group_exit(),
	82	* but BUG() actually hooks all the lock-breaking and per-arch
	83	* Oops code, so that is used here instead.
	84	*/
	85	BUG();
	86	}
	87
	88	/* Returns true if any portion of [ptr,ptr+n) over laps with [low,high). */
	89	static bool overlaps(const void *ptr, unsigned long n, unsigned long low,
	90	unsigned long high)
	91	{
	92	unsigned long check_low = (uintptr_t)ptr;
	93	unsigned long check_high = check_low + n;
	94
	95	/* Does not overlap if entirely above or entirely below. */
94cd97af	96	if (check_low >= high \|\| check_high <= low)
f5509cc1 KC	97	return false;
	98
	99	return true;
	100	}
	101
	102	/* Is this address range in the kernel text area? */
	103	static inline const char check_kernel_text_object(const void ptr,
	104	unsigned long n)
	105	{
	106	unsigned long textlow = (unsigned long)_stext;
	107	unsigned long texthigh = (unsigned long)_etext;
	108	unsigned long textlow_linear, texthigh_linear;
	109
	110	if (overlaps(ptr, n, textlow, texthigh))
	111	return "<kernel text>";
	112
	113	/*
	114	* Some architectures have virtual memory mappings with a secondary
	115	* mapping of the kernel text, i.e. there is more than one virtual
	116	* kernel address that points to the kernel image. It is usually
	117	* when there is a separate linear physical memory mapping, in that
	118	* __pa() is not just the reverse of __va(). This can be detected
	119	* and checked:
	120	*/
46f6236a	121	textlow_linear = (unsigned long)lm_alias(textlow);
f5509cc1 KC	122	/* No different mapping: we're done. */
	123	if (textlow_linear == textlow)
	124	return NULL;
	125
	126	/* Check the secondary mapping... */
46f6236a	127	texthigh_linear = (unsigned long)lm_alias(texthigh);
f5509cc1 KC	128	if (overlaps(ptr, n, textlow_linear, texthigh_linear))
	129	return "<linear kernel text>";
	130
	131	return NULL;
	132	}
	133
	134	static inline const char check_bogus_address(const void ptr, unsigned long n)
	135	{
	136	/* Reject if object wraps past end of memory. */
7329a655	137	if ((unsigned long)ptr + n < (unsigned long)ptr)
f5509cc1 KC	138	return "<wrapped address>";
	139
	140	/* Reject if NULL or ZERO-allocation. */
	141	if (ZERO_OR_NULL_PTR(ptr))
	142	return "<null>";
	143
	144	return NULL;
	145	}
	146
8e1f74ea KC	147	/* Checks for allocs that are marked in some way as spanning multiple pages. */
	148	static inline const char check_page_span(const void ptr, unsigned long n,
	149	struct page *page, bool to_user)
f5509cc1	150	{
8e1f74ea	151	#ifdef CONFIG_HARDENED_USERCOPY_PAGESPAN
f5509cc1	152	const void *end = ptr + n - 1;
8e1f74ea	153	struct page *endpage;
f5509cc1 KC	154	bool is_reserved, is_cma;
f5509cc1 KC	155
f5509cc1 KC	156	/*
	157	* Sometimes the kernel data regions are not marked Reserved (see
	158	* check below). And sometimes [_sdata,_edata) does not cover
	159	* rodata and/or bss, so check each range explicitly.
	160	*/
	161
	162	/* Allow reads of kernel rodata region (if not marked as Reserved). */
	163	if (ptr >= (const void *)__start_rodata &&
	164	end <= (const void *)__end_rodata) {
	165	if (!to_user)
	166	return "<rodata>";
	167	return NULL;
	168	}
	169
	170	/* Allow kernel data region (if not marked as Reserved). */
	171	if (ptr >= (const void )_sdata && end <= (const void )_edata)
	172	return NULL;
	173
	174	/* Allow kernel bss region (if not marked as Reserved). */
	175	if (ptr >= (const void *)__bss_start &&
	176	end <= (const void *)__bss_stop)
	177	return NULL;
	178
	179	/* Is the object wholly within one base page? */
	180	if (likely(((unsigned long)ptr & (unsigned long)PAGE_MASK) ==
	181	((unsigned long)end & (unsigned long)PAGE_MASK)))
	182	return NULL;
	183
8e1f74ea	184	/* Allow if fully inside the same compound (__GFP_COMP) page. */
f5509cc1 KC	185	endpage = virt_to_head_page(end);
	186	if (likely(endpage == page))
	187	return NULL;
	188
	189	/*
	190	* Reject if range is entirely either Reserved (i.e. special or
	191	* device memory), or CMA. Otherwise, reject since the object spans
	192	* several independently allocated pages.
	193	*/
	194	is_reserved = PageReserved(page);
	195	is_cma = is_migrate_cma_page(page);
	196	if (!is_reserved && !is_cma)
8e1f74ea	197	return "<spans multiple pages>";
f5509cc1 KC	198
	199	for (ptr += PAGE_SIZE; ptr <= end; ptr += PAGE_SIZE) {
	200	page = virt_to_head_page(ptr);
	201	if (is_reserved && !PageReserved(page))
8e1f74ea	202	return "<spans Reserved and non-Reserved pages>";
f5509cc1	203	if (is_cma && !is_migrate_cma_page(page))
8e1f74ea	204	return "<spans CMA and non-CMA pages>";
f5509cc1	205	}
8e1f74ea	206	#endif
f5509cc1 KC	207
f5509cc1 KC	208	return NULL;
8e1f74ea KC	209	}
	210
	211	static inline const char check_heap_object(const void ptr, unsigned long n,
	212	bool to_user)
	213	{
	214	struct page *page;
	215
8e1f74ea KC	216	if (!virt_addr_valid(ptr))
	217	return NULL;
	218
	219	page = virt_to_head_page(ptr);
	220
	221	/* Check slab allocator for flags and size. */
	222	if (PageSlab(page))
	223	return __check_heap_object(ptr, n, page);
f5509cc1	224
8e1f74ea KC	225	/* Verify object does not incorrectly span multiple pages. */
8e1f74ea KC	226	return check_page_span(ptr, n, page, to_user);
f5509cc1 KC	227	}
	228
	229	/*
	230	* Validates that the given object is:
	231	* - not bogus address
	232	* - known-safe heap or stack object
	233	* - not in kernel text
	234	*/
	235	void __check_object_size(const void *ptr, unsigned long n, bool to_user)
	236	{
	237	const char *err;
	238
	239	/* Skip all tests if size is zero. */
	240	if (!n)
	241	return;
	242
	243	/* Check for invalid addresses. */
	244	err = check_bogus_address(ptr, n);
	245	if (err)
	246	goto report;
	247
	248	/* Check for bad heap object. */
	249	err = check_heap_object(ptr, n, to_user);
	250	if (err)
	251	goto report;
	252
	253	/* Check for bad stack object. */
	254	switch (check_stack_object(ptr, n)) {
	255	case NOT_STACK:
	256	/* Object is not touching the current process stack. */
	257	break;
	258	case GOOD_FRAME:
	259	case GOOD_STACK:
	260	/*
	261	* Object is either in the correct frame (when it
	262	* is possible to check) or just generally on the
	263	* process stack (when frame checking not available).
	264	*/
	265	return;
	266	default:
	267	err = "<process stack>";
	268	goto report;
	269	}
	270
	271	/* Check for object in kernel to avoid text exposure. */
	272	err = check_kernel_text_object(ptr, n);
	273	if (!err)
	274	return;
	275
	276	report:
b394d468	277	usercopy_abort(err, NULL, to_user, 0, n);
f5509cc1 KC	278	}
f5509cc1 KC	279	EXPORT_SYMBOL(__check_object_size);