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 /* ioctls for /dev/userfaultfd */ 16 #define USERFAULTFD_IOC 0xAA 17 #define USERFAULTFD_IOC_NEW _IO(USERFAULTFD_IOC, 0x00) 18 19 /* 20 * If the UFFDIO_API is upgraded someday, the UFFDIO_UNREGISTER and 21 * UFFDIO_WAKE ioctls should be defined as _IOW and not as _IOR. In 22 * userfaultfd.h we assumed the kernel was reading (instead _IOC_READ 23 * means the userland is reading). 24 */ 25 #define UFFD_API ((__u64)0xAA) 26 #define UFFD_API_REGISTER_MODES (UFFDIO_REGISTER_MODE_MISSING | \ 27 UFFDIO_REGISTER_MODE_WP | \ 28 UFFDIO_REGISTER_MODE_MINOR) 29 #define UFFD_API_FEATURES (UFFD_FEATURE_PAGEFAULT_FLAG_WP | \ 30 UFFD_FEATURE_EVENT_FORK | \ 31 UFFD_FEATURE_EVENT_REMAP | \ 32 UFFD_FEATURE_EVENT_REMOVE | \ 33 UFFD_FEATURE_EVENT_UNMAP | \ 34 UFFD_FEATURE_MISSING_HUGETLBFS | \ 35 UFFD_FEATURE_MISSING_SHMEM | \ 36 UFFD_FEATURE_SIGBUS | \ 37 UFFD_FEATURE_THREAD_ID | \ 38 UFFD_FEATURE_MINOR_HUGETLBFS | \ 39 UFFD_FEATURE_MINOR_SHMEM | \ 40 UFFD_FEATURE_EXACT_ADDRESS | \ 41 UFFD_FEATURE_WP_HUGETLBFS_SHMEM | \ 42 UFFD_FEATURE_WP_UNPOPULATED | \ 43 UFFD_FEATURE_POISON | \ 44 UFFD_FEATURE_WP_ASYNC) 45 #define UFFD_API_IOCTLS \ 46 ((__u64)1 << _UFFDIO_REGISTER | \ 47 (__u64)1 << _UFFDIO_UNREGISTER | \ 48 (__u64)1 << _UFFDIO_API) 49 #define UFFD_API_RANGE_IOCTLS \ 50 ((__u64)1 << _UFFDIO_WAKE | \ 51 (__u64)1 << _UFFDIO_COPY | \ 52 (__u64)1 << _UFFDIO_ZEROPAGE | \ 53 (__u64)1 << _UFFDIO_WRITEPROTECT | \ 54 (__u64)1 << _UFFDIO_CONTINUE | \ 55 (__u64)1 << _UFFDIO_POISON) 56 #define UFFD_API_RANGE_IOCTLS_BASIC \ 57 ((__u64)1 << _UFFDIO_WAKE | \ 58 (__u64)1 << _UFFDIO_COPY | \ 59 (__u64)1 << _UFFDIO_WRITEPROTECT | \ 60 (__u64)1 << _UFFDIO_CONTINUE | \ 61 (__u64)1 << _UFFDIO_POISON) 62 63 /* 64 * Valid ioctl command number range with this API is from 0x00 to 65 * 0x3F. UFFDIO_API is the fixed number, everything else can be 66 * changed by implementing a different UFFD_API. If sticking to the 67 * same UFFD_API more ioctl can be added and userland will be aware of 68 * which ioctl the running kernel implements through the ioctl command 69 * bitmask written by the UFFDIO_API. 70 */ 71 #define _UFFDIO_REGISTER (0x00) 72 #define _UFFDIO_UNREGISTER (0x01) 73 #define _UFFDIO_WAKE (0x02) 74 #define _UFFDIO_COPY (0x03) 75 #define _UFFDIO_ZEROPAGE (0x04) 76 #define _UFFDIO_WRITEPROTECT (0x06) 77 #define _UFFDIO_CONTINUE (0x07) 78 #define _UFFDIO_POISON (0x08) 79 #define _UFFDIO_API (0x3F) 80 81 /* userfaultfd ioctl ids */ 82 #define UFFDIO 0xAA 83 #define UFFDIO_API _IOWR(UFFDIO, _UFFDIO_API, \ 84 struct uffdio_api) 85 #define UFFDIO_REGISTER _IOWR(UFFDIO, _UFFDIO_REGISTER, \ 86 struct uffdio_register) 87 #define UFFDIO_UNREGISTER _IOR(UFFDIO, _UFFDIO_UNREGISTER, \ 88 struct uffdio_range) 89 #define UFFDIO_WAKE _IOR(UFFDIO, _UFFDIO_WAKE, \ 90 struct uffdio_range) 91 #define UFFDIO_COPY _IOWR(UFFDIO, _UFFDIO_COPY, \ 92 struct uffdio_copy) 93 #define UFFDIO_ZEROPAGE _IOWR(UFFDIO, _UFFDIO_ZEROPAGE, \ 94 struct uffdio_zeropage) 95 #define UFFDIO_WRITEPROTECT _IOWR(UFFDIO, _UFFDIO_WRITEPROTECT, \ 96 struct uffdio_writeprotect) 97 #define UFFDIO_CONTINUE _IOWR(UFFDIO, _UFFDIO_CONTINUE, \ 98 struct uffdio_continue) 99 #define UFFDIO_POISON _IOWR(UFFDIO, _UFFDIO_POISON, \ 100 struct uffdio_poison) 101 102 /* read() structure */ 103 struct uffd_msg { 104 __u8 event; 105 106 __u8 reserved1; 107 __u16 reserved2; 108 __u32 reserved3; 109 110 union { 111 struct { 112 __u64 flags; 113 __u64 address; 114 union { 115 __u32 ptid; 116 } feat; 117 } pagefault; 118 119 struct { 120 __u32 ufd; 121 } fork; 122 123 struct { 124 __u64 from; 125 __u64 to; 126 __u64 len; 127 } remap; 128 129 struct { 130 __u64 start; 131 __u64 end; 132 } remove; 133 134 struct { 135 /* unused reserved fields */ 136 __u64 reserved1; 137 __u64 reserved2; 138 __u64 reserved3; 139 } reserved; 140 } arg; 141 } __attribute__((packed)); 142 143 /* 144 * Start at 0x12 and not at 0 to be more strict against bugs. 145 */ 146 #define UFFD_EVENT_PAGEFAULT 0x12 147 #define UFFD_EVENT_FORK 0x13 148 #define UFFD_EVENT_REMAP 0x14 149 #define UFFD_EVENT_REMOVE 0x15 150 #define UFFD_EVENT_UNMAP 0x16 151 152 /* flags for UFFD_EVENT_PAGEFAULT */ 153 #define UFFD_PAGEFAULT_FLAG_WRITE (1<<0) /* If this was a write fault */ 154 #define UFFD_PAGEFAULT_FLAG_WP (1<<1) /* If reason is VM_UFFD_WP */ 155 #define UFFD_PAGEFAULT_FLAG_MINOR (1<<2) /* If reason is VM_UFFD_MINOR */ 156 157 struct uffdio_api { 158 /* userland asks for an API number and the features to enable */ 159 __u64 api; 160 /* 161 * Kernel answers below with the all available features for 162 * the API, this notifies userland of which events and/or 163 * which flags for each event are enabled in the current 164 * kernel. 165 * 166 * Note: UFFD_EVENT_PAGEFAULT and UFFD_PAGEFAULT_FLAG_WRITE 167 * are to be considered implicitly always enabled in all kernels as 168 * long as the uffdio_api.api requested matches UFFD_API. 169 * 170 * UFFD_FEATURE_MISSING_HUGETLBFS means an UFFDIO_REGISTER 171 * with UFFDIO_REGISTER_MODE_MISSING mode will succeed on 172 * hugetlbfs virtual memory ranges. Adding or not adding 173 * UFFD_FEATURE_MISSING_HUGETLBFS to uffdio_api.features has 174 * no real functional effect after UFFDIO_API returns, but 175 * it's only useful for an initial feature set probe at 176 * UFFDIO_API time. There are two ways to use it: 177 * 178 * 1) by adding UFFD_FEATURE_MISSING_HUGETLBFS to the 179 * uffdio_api.features before calling UFFDIO_API, an error 180 * will be returned by UFFDIO_API on a kernel without 181 * hugetlbfs missing support 182 * 183 * 2) the UFFD_FEATURE_MISSING_HUGETLBFS can not be added in 184 * uffdio_api.features and instead it will be set by the 185 * kernel in the uffdio_api.features if the kernel supports 186 * it, so userland can later check if the feature flag is 187 * present in uffdio_api.features after UFFDIO_API 188 * succeeded. 189 * 190 * UFFD_FEATURE_MISSING_SHMEM works the same as 191 * UFFD_FEATURE_MISSING_HUGETLBFS, but it applies to shmem 192 * (i.e. tmpfs and other shmem based APIs). 193 * 194 * UFFD_FEATURE_SIGBUS feature means no page-fault 195 * (UFFD_EVENT_PAGEFAULT) event will be delivered, instead 196 * a SIGBUS signal will be sent to the faulting process. 197 * 198 * UFFD_FEATURE_THREAD_ID pid of the page faulted task_struct will 199 * be returned, if feature is not requested 0 will be returned. 200 * 201 * UFFD_FEATURE_MINOR_HUGETLBFS indicates that minor faults 202 * can be intercepted (via REGISTER_MODE_MINOR) for 203 * hugetlbfs-backed pages. 204 * 205 * UFFD_FEATURE_MINOR_SHMEM indicates the same support as 206 * UFFD_FEATURE_MINOR_HUGETLBFS, but for shmem-backed pages instead. 207 * 208 * UFFD_FEATURE_EXACT_ADDRESS indicates that the exact address of page 209 * faults would be provided and the offset within the page would not be 210 * masked. 211 * 212 * UFFD_FEATURE_WP_HUGETLBFS_SHMEM indicates that userfaultfd 213 * write-protection mode is supported on both shmem and hugetlbfs. 214 * 215 * UFFD_FEATURE_WP_UNPOPULATED indicates that userfaultfd 216 * write-protection mode will always apply to unpopulated pages 217 * (i.e. empty ptes). This will be the default behavior for shmem 218 * & hugetlbfs, so this flag only affects anonymous memory behavior 219 * when userfault write-protection mode is registered. 220 * 221 * UFFD_FEATURE_WP_ASYNC indicates that userfaultfd write-protection 222 * asynchronous mode is supported in which the write fault is 223 * automatically resolved and write-protection is un-set. 224 * It implies UFFD_FEATURE_WP_UNPOPULATED. 225 */ 226 #define UFFD_FEATURE_PAGEFAULT_FLAG_WP (1<<0) 227 #define UFFD_FEATURE_EVENT_FORK (1<<1) 228 #define UFFD_FEATURE_EVENT_REMAP (1<<2) 229 #define UFFD_FEATURE_EVENT_REMOVE (1<<3) 230 #define UFFD_FEATURE_MISSING_HUGETLBFS (1<<4) 231 #define UFFD_FEATURE_MISSING_SHMEM (1<<5) 232 #define UFFD_FEATURE_EVENT_UNMAP (1<<6) 233 #define UFFD_FEATURE_SIGBUS (1<<7) 234 #define UFFD_FEATURE_THREAD_ID (1<<8) 235 #define UFFD_FEATURE_MINOR_HUGETLBFS (1<<9) 236 #define UFFD_FEATURE_MINOR_SHMEM (1<<10) 237 #define UFFD_FEATURE_EXACT_ADDRESS (1<<11) 238 #define UFFD_FEATURE_WP_HUGETLBFS_SHMEM (1<<12) 239 #define UFFD_FEATURE_WP_UNPOPULATED (1<<13) 240 #define UFFD_FEATURE_POISON (1<<14) 241 #define UFFD_FEATURE_WP_ASYNC (1<<15) 242 __u64 features; 243 244 __u64 ioctls; 245 }; 246 247 struct uffdio_range { 248 __u64 start; 249 __u64 len; 250 }; 251 252 struct uffdio_register { 253 struct uffdio_range range; 254 #define UFFDIO_REGISTER_MODE_MISSING ((__u64)1<<0) 255 #define UFFDIO_REGISTER_MODE_WP ((__u64)1<<1) 256 #define UFFDIO_REGISTER_MODE_MINOR ((__u64)1<<2) 257 __u64 mode; 258 259 /* 260 * kernel answers which ioctl commands are available for the 261 * range, keep at the end as the last 8 bytes aren't read. 262 */ 263 __u64 ioctls; 264 }; 265 266 struct uffdio_copy { 267 __u64 dst; 268 __u64 src; 269 __u64 len; 270 #define UFFDIO_COPY_MODE_DONTWAKE ((__u64)1<<0) 271 /* 272 * UFFDIO_COPY_MODE_WP will map the page write protected on 273 * the fly. UFFDIO_COPY_MODE_WP is available only if the 274 * write protected ioctl is implemented for the range 275 * according to the uffdio_register.ioctls. 276 */ 277 #define UFFDIO_COPY_MODE_WP ((__u64)1<<1) 278 __u64 mode; 279 280 /* 281 * "copy" is written by the ioctl and must be at the end: the 282 * copy_from_user will not read the last 8 bytes. 283 */ 284 __s64 copy; 285 }; 286 287 struct uffdio_zeropage { 288 struct uffdio_range range; 289 #define UFFDIO_ZEROPAGE_MODE_DONTWAKE ((__u64)1<<0) 290 __u64 mode; 291 292 /* 293 * "zeropage" is written by the ioctl and must be at the end: 294 * the copy_from_user will not read the last 8 bytes. 295 */ 296 __s64 zeropage; 297 }; 298 299 struct uffdio_writeprotect { 300 struct uffdio_range range; 301 /* 302 * UFFDIO_WRITEPROTECT_MODE_WP: set the flag to write protect a range, 303 * unset the flag to undo protection of a range which was previously 304 * write protected. 305 * 306 * UFFDIO_WRITEPROTECT_MODE_DONTWAKE: set the flag to avoid waking up 307 * any wait thread after the operation succeeds. 308 * 309 * NOTE: Write protecting a region (WP=1) is unrelated to page faults, 310 * therefore DONTWAKE flag is meaningless with WP=1. Removing write 311 * protection (WP=0) in response to a page fault wakes the faulting 312 * task unless DONTWAKE is set. 313 */ 314 #define UFFDIO_WRITEPROTECT_MODE_WP ((__u64)1<<0) 315 #define UFFDIO_WRITEPROTECT_MODE_DONTWAKE ((__u64)1<<1) 316 __u64 mode; 317 }; 318 319 struct uffdio_continue { 320 struct uffdio_range range; 321 #define UFFDIO_CONTINUE_MODE_DONTWAKE ((__u64)1<<0) 322 /* 323 * UFFDIO_CONTINUE_MODE_WP will map the page write protected on 324 * the fly. UFFDIO_CONTINUE_MODE_WP is available only if the 325 * write protected ioctl is implemented for the range 326 * according to the uffdio_register.ioctls. 327 */ 328 #define UFFDIO_CONTINUE_MODE_WP ((__u64)1<<1) 329 __u64 mode; 330 331 /* 332 * Fields below here are written by the ioctl and must be at the end: 333 * the copy_from_user will not read past here. 334 */ 335 __s64 mapped; 336 }; 337 338 struct uffdio_poison { 339 struct uffdio_range range; 340 #define UFFDIO_POISON_MODE_DONTWAKE ((__u64)1<<0) 341 __u64 mode; 342 343 /* 344 * Fields below here are written by the ioctl and must be at the end: 345 * the copy_from_user will not read past here. 346 */ 347 __s64 updated; 348 }; 349 350 /* 351 * Flags for the userfaultfd(2) system call itself. 352 */ 353 354 /* 355 * Create a userfaultfd that can handle page faults only in user mode. 356 */ 357 #define UFFD_USER_MODE_ONLY 1 358 359 #endif /* _LINUX_USERFAULTFD_H */ 360