1 /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2 /*
3  *  include/linux/userfaultfd.h
4  *
5  *  Copyright (C) 2007  Davide Libenzi <davidel@xmailserver.org>
6  *  Copyright (C) 2015  Red Hat, Inc.
7  *
8  */
9 
10 #ifndef _LINUX_USERFAULTFD_H
11 #define _LINUX_USERFAULTFD_H
12 
13 #include <linux/types.h>
14 
15 /*
16  * If the UFFDIO_API is upgraded someday, the UFFDIO_UNREGISTER and
17  * UFFDIO_WAKE ioctls should be defined as _IOW and not as _IOR.  In
18  * userfaultfd.h we assumed the kernel was reading (instead _IOC_READ
19  * means the userland is reading).
20  */
21 #define UFFD_API ((__u64)0xAA)
22 #define UFFD_API_REGISTER_MODES (UFFDIO_REGISTER_MODE_MISSING |	\
23 				 UFFDIO_REGISTER_MODE_WP |	\
24 				 UFFDIO_REGISTER_MODE_MINOR)
25 #define UFFD_API_FEATURES (UFFD_FEATURE_PAGEFAULT_FLAG_WP |	\
26 			   UFFD_FEATURE_EVENT_FORK |		\
27 			   UFFD_FEATURE_EVENT_REMAP |		\
28 			   UFFD_FEATURE_EVENT_REMOVE |		\
29 			   UFFD_FEATURE_EVENT_UNMAP |		\
30 			   UFFD_FEATURE_MISSING_HUGETLBFS |	\
31 			   UFFD_FEATURE_MISSING_SHMEM |		\
32 			   UFFD_FEATURE_SIGBUS |		\
33 			   UFFD_FEATURE_THREAD_ID |		\
34 			   UFFD_FEATURE_MINOR_HUGETLBFS |	\
35 			   UFFD_FEATURE_MINOR_SHMEM |		\
36 			   UFFD_FEATURE_EXACT_ADDRESS)
37 #define UFFD_API_IOCTLS				\
38 	((__u64)1 << _UFFDIO_REGISTER |		\
39 	 (__u64)1 << _UFFDIO_UNREGISTER |	\
40 	 (__u64)1 << _UFFDIO_API)
41 #define UFFD_API_RANGE_IOCTLS			\
42 	((__u64)1 << _UFFDIO_WAKE |		\
43 	 (__u64)1 << _UFFDIO_COPY |		\
44 	 (__u64)1 << _UFFDIO_ZEROPAGE |		\
45 	 (__u64)1 << _UFFDIO_WRITEPROTECT |	\
46 	 (__u64)1 << _UFFDIO_CONTINUE)
47 #define UFFD_API_RANGE_IOCTLS_BASIC		\
48 	((__u64)1 << _UFFDIO_WAKE |		\
49 	 (__u64)1 << _UFFDIO_COPY |		\
50 	 (__u64)1 << _UFFDIO_CONTINUE)
51 
52 /*
53  * Valid ioctl command number range with this API is from 0x00 to
54  * 0x3F.  UFFDIO_API is the fixed number, everything else can be
55  * changed by implementing a different UFFD_API. If sticking to the
56  * same UFFD_API more ioctl can be added and userland will be aware of
57  * which ioctl the running kernel implements through the ioctl command
58  * bitmask written by the UFFDIO_API.
59  */
60 #define _UFFDIO_REGISTER		(0x00)
61 #define _UFFDIO_UNREGISTER		(0x01)
62 #define _UFFDIO_WAKE			(0x02)
63 #define _UFFDIO_COPY			(0x03)
64 #define _UFFDIO_ZEROPAGE		(0x04)
65 #define _UFFDIO_WRITEPROTECT		(0x06)
66 #define _UFFDIO_CONTINUE		(0x07)
67 #define _UFFDIO_API			(0x3F)
68 
69 /* userfaultfd ioctl ids */
70 #define UFFDIO 0xAA
71 #define UFFDIO_API		_IOWR(UFFDIO, _UFFDIO_API,	\
72 				      struct uffdio_api)
73 #define UFFDIO_REGISTER		_IOWR(UFFDIO, _UFFDIO_REGISTER, \
74 				      struct uffdio_register)
75 #define UFFDIO_UNREGISTER	_IOR(UFFDIO, _UFFDIO_UNREGISTER,	\
76 				     struct uffdio_range)
77 #define UFFDIO_WAKE		_IOR(UFFDIO, _UFFDIO_WAKE,	\
78 				     struct uffdio_range)
79 #define UFFDIO_COPY		_IOWR(UFFDIO, _UFFDIO_COPY,	\
80 				      struct uffdio_copy)
81 #define UFFDIO_ZEROPAGE		_IOWR(UFFDIO, _UFFDIO_ZEROPAGE,	\
82 				      struct uffdio_zeropage)
83 #define UFFDIO_WRITEPROTECT	_IOWR(UFFDIO, _UFFDIO_WRITEPROTECT, \
84 				      struct uffdio_writeprotect)
85 #define UFFDIO_CONTINUE		_IOWR(UFFDIO, _UFFDIO_CONTINUE,	\
86 				      struct uffdio_continue)
87 
88 /* read() structure */
89 struct uffd_msg {
90 	__u8	event;
91 
92 	__u8	reserved1;
93 	__u16	reserved2;
94 	__u32	reserved3;
95 
96 	union {
97 		struct {
98 			__u64	flags;
99 			__u64	address;
100 			union {
101 				__u32 ptid;
102 			} feat;
103 		} pagefault;
104 
105 		struct {
106 			__u32	ufd;
107 		} fork;
108 
109 		struct {
110 			__u64	from;
111 			__u64	to;
112 			__u64	len;
113 		} remap;
114 
115 		struct {
116 			__u64	start;
117 			__u64	end;
118 		} remove;
119 
120 		struct {
121 			/* unused reserved fields */
122 			__u64	reserved1;
123 			__u64	reserved2;
124 			__u64	reserved3;
125 		} reserved;
126 	} arg;
127 } __attribute__((packed));
128 
129 /*
130  * Start at 0x12 and not at 0 to be more strict against bugs.
131  */
132 #define UFFD_EVENT_PAGEFAULT	0x12
133 #define UFFD_EVENT_FORK		0x13
134 #define UFFD_EVENT_REMAP	0x14
135 #define UFFD_EVENT_REMOVE	0x15
136 #define UFFD_EVENT_UNMAP	0x16
137 
138 /* flags for UFFD_EVENT_PAGEFAULT */
139 #define UFFD_PAGEFAULT_FLAG_WRITE	(1<<0)	/* If this was a write fault */
140 #define UFFD_PAGEFAULT_FLAG_WP		(1<<1)	/* If reason is VM_UFFD_WP */
141 #define UFFD_PAGEFAULT_FLAG_MINOR	(1<<2)	/* If reason is VM_UFFD_MINOR */
142 
143 struct uffdio_api {
144 	/* userland asks for an API number and the features to enable */
145 	__u64 api;
146 	/*
147 	 * Kernel answers below with the all available features for
148 	 * the API, this notifies userland of which events and/or
149 	 * which flags for each event are enabled in the current
150 	 * kernel.
151 	 *
152 	 * Note: UFFD_EVENT_PAGEFAULT and UFFD_PAGEFAULT_FLAG_WRITE
153 	 * are to be considered implicitly always enabled in all kernels as
154 	 * long as the uffdio_api.api requested matches UFFD_API.
155 	 *
156 	 * UFFD_FEATURE_MISSING_HUGETLBFS means an UFFDIO_REGISTER
157 	 * with UFFDIO_REGISTER_MODE_MISSING mode will succeed on
158 	 * hugetlbfs virtual memory ranges. Adding or not adding
159 	 * UFFD_FEATURE_MISSING_HUGETLBFS to uffdio_api.features has
160 	 * no real functional effect after UFFDIO_API returns, but
161 	 * it's only useful for an initial feature set probe at
162 	 * UFFDIO_API time. There are two ways to use it:
163 	 *
164 	 * 1) by adding UFFD_FEATURE_MISSING_HUGETLBFS to the
165 	 *    uffdio_api.features before calling UFFDIO_API, an error
166 	 *    will be returned by UFFDIO_API on a kernel without
167 	 *    hugetlbfs missing support
168 	 *
169 	 * 2) the UFFD_FEATURE_MISSING_HUGETLBFS can not be added in
170 	 *    uffdio_api.features and instead it will be set by the
171 	 *    kernel in the uffdio_api.features if the kernel supports
172 	 *    it, so userland can later check if the feature flag is
173 	 *    present in uffdio_api.features after UFFDIO_API
174 	 *    succeeded.
175 	 *
176 	 * UFFD_FEATURE_MISSING_SHMEM works the same as
177 	 * UFFD_FEATURE_MISSING_HUGETLBFS, but it applies to shmem
178 	 * (i.e. tmpfs and other shmem based APIs).
179 	 *
180 	 * UFFD_FEATURE_SIGBUS feature means no page-fault
181 	 * (UFFD_EVENT_PAGEFAULT) event will be delivered, instead
182 	 * a SIGBUS signal will be sent to the faulting process.
183 	 *
184 	 * UFFD_FEATURE_THREAD_ID pid of the page faulted task_struct will
185 	 * be returned, if feature is not requested 0 will be returned.
186 	 *
187 	 * UFFD_FEATURE_MINOR_HUGETLBFS indicates that minor faults
188 	 * can be intercepted (via REGISTER_MODE_MINOR) for
189 	 * hugetlbfs-backed pages.
190 	 *
191 	 * UFFD_FEATURE_MINOR_SHMEM indicates the same support as
192 	 * UFFD_FEATURE_MINOR_HUGETLBFS, but for shmem-backed pages instead.
193 	 *
194 	 * UFFD_FEATURE_EXACT_ADDRESS indicates that the exact address of page
195 	 * faults would be provided and the offset within the page would not be
196 	 * masked.
197 	 */
198 #define UFFD_FEATURE_PAGEFAULT_FLAG_WP		(1<<0)
199 #define UFFD_FEATURE_EVENT_FORK			(1<<1)
200 #define UFFD_FEATURE_EVENT_REMAP		(1<<2)
201 #define UFFD_FEATURE_EVENT_REMOVE		(1<<3)
202 #define UFFD_FEATURE_MISSING_HUGETLBFS		(1<<4)
203 #define UFFD_FEATURE_MISSING_SHMEM		(1<<5)
204 #define UFFD_FEATURE_EVENT_UNMAP		(1<<6)
205 #define UFFD_FEATURE_SIGBUS			(1<<7)
206 #define UFFD_FEATURE_THREAD_ID			(1<<8)
207 #define UFFD_FEATURE_MINOR_HUGETLBFS		(1<<9)
208 #define UFFD_FEATURE_MINOR_SHMEM		(1<<10)
209 #define UFFD_FEATURE_EXACT_ADDRESS		(1<<11)
210 	__u64 features;
211 
212 	__u64 ioctls;
213 };
214 
215 struct uffdio_range {
216 	__u64 start;
217 	__u64 len;
218 };
219 
220 struct uffdio_register {
221 	struct uffdio_range range;
222 #define UFFDIO_REGISTER_MODE_MISSING	((__u64)1<<0)
223 #define UFFDIO_REGISTER_MODE_WP		((__u64)1<<1)
224 #define UFFDIO_REGISTER_MODE_MINOR	((__u64)1<<2)
225 	__u64 mode;
226 
227 	/*
228 	 * kernel answers which ioctl commands are available for the
229 	 * range, keep at the end as the last 8 bytes aren't read.
230 	 */
231 	__u64 ioctls;
232 };
233 
234 struct uffdio_copy {
235 	__u64 dst;
236 	__u64 src;
237 	__u64 len;
238 #define UFFDIO_COPY_MODE_DONTWAKE		((__u64)1<<0)
239 	/*
240 	 * UFFDIO_COPY_MODE_WP will map the page write protected on
241 	 * the fly.  UFFDIO_COPY_MODE_WP is available only if the
242 	 * write protected ioctl is implemented for the range
243 	 * according to the uffdio_register.ioctls.
244 	 */
245 #define UFFDIO_COPY_MODE_WP			((__u64)1<<1)
246 	__u64 mode;
247 
248 	/*
249 	 * "copy" is written by the ioctl and must be at the end: the
250 	 * copy_from_user will not read the last 8 bytes.
251 	 */
252 	__s64 copy;
253 };
254 
255 struct uffdio_zeropage {
256 	struct uffdio_range range;
257 #define UFFDIO_ZEROPAGE_MODE_DONTWAKE		((__u64)1<<0)
258 	__u64 mode;
259 
260 	/*
261 	 * "zeropage" is written by the ioctl and must be at the end:
262 	 * the copy_from_user will not read the last 8 bytes.
263 	 */
264 	__s64 zeropage;
265 };
266 
267 struct uffdio_writeprotect {
268 	struct uffdio_range range;
269 /*
270  * UFFDIO_WRITEPROTECT_MODE_WP: set the flag to write protect a range,
271  * unset the flag to undo protection of a range which was previously
272  * write protected.
273  *
274  * UFFDIO_WRITEPROTECT_MODE_DONTWAKE: set the flag to avoid waking up
275  * any wait thread after the operation succeeds.
276  *
277  * NOTE: Write protecting a region (WP=1) is unrelated to page faults,
278  * therefore DONTWAKE flag is meaningless with WP=1.  Removing write
279  * protection (WP=0) in response to a page fault wakes the faulting
280  * task unless DONTWAKE is set.
281  */
282 #define UFFDIO_WRITEPROTECT_MODE_WP		((__u64)1<<0)
283 #define UFFDIO_WRITEPROTECT_MODE_DONTWAKE	((__u64)1<<1)
284 	__u64 mode;
285 };
286 
287 struct uffdio_continue {
288 	struct uffdio_range range;
289 #define UFFDIO_CONTINUE_MODE_DONTWAKE		((__u64)1<<0)
290 	__u64 mode;
291 
292 	/*
293 	 * Fields below here are written by the ioctl and must be at the end:
294 	 * the copy_from_user will not read past here.
295 	 */
296 	__s64 mapped;
297 };
298 
299 /*
300  * Flags for the userfaultfd(2) system call itself.
301  */
302 
303 /*
304  * Create a userfaultfd that can handle page faults only in user mode.
305  */
306 #define UFFD_USER_MODE_ONLY 1
307 
308 #endif /* _LINUX_USERFAULTFD_H */
309