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