1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Copyright (c) 2014 The FreeBSD Foundation 13 * 14 * Portions of this software were developed by Konstantin Belousov 15 * under sponsorship from the FreeBSD Foundation. 16 * 17 * Redistribution and use in source and binary forms, with or without 18 * modification, are permitted provided that the following conditions 19 * are met: 20 * 1. Redistributions of source code must retain the above copyright 21 * notice, this list of conditions and the following disclaimer. 22 * 2. Redistributions in binary form must reproduce the above copyright 23 * notice, this list of conditions and the following disclaimer in the 24 * documentation and/or other materials provided with the distribution. 25 * 3. Neither the name of the University nor the names of its contributors 26 * may be used to endorse or promote products derived from this software 27 * without specific prior written permission. 28 * 29 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 30 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 31 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 32 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 33 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 34 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 35 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 36 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 38 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 39 * SUCH DAMAGE. 40 * 41 * @(#)kern_subr.c 8.3 (Berkeley) 1/21/94 42 */ 43 44 #include <sys/cdefs.h> 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/kernel.h> 48 #include <sys/limits.h> 49 #include <sys/lock.h> 50 #include <sys/mman.h> 51 #include <sys/proc.h> 52 #include <sys/resourcevar.h> 53 #include <sys/rwlock.h> 54 #include <sys/sched.h> 55 #include <sys/sysctl.h> 56 #include <sys/vnode.h> 57 58 #include <vm/vm.h> 59 #include <vm/vm_param.h> 60 #include <vm/vm_extern.h> 61 #include <vm/vm_page.h> 62 #include <vm/vm_pageout.h> 63 #include <vm/vm_map.h> 64 65 #include <machine/bus.h> 66 67 SYSCTL_INT(_kern, KERN_IOV_MAX, iov_max, CTLFLAG_RD, SYSCTL_NULL_INT_PTR, UIO_MAXIOV, 68 "Maximum number of elements in an I/O vector; sysconf(_SC_IOV_MAX)"); 69 70 static int uiomove_faultflag(void *cp, int n, struct uio *uio, int nofault); 71 72 int 73 copyin_nofault(const void *udaddr, void *kaddr, size_t len) 74 { 75 int error, save; 76 77 save = vm_fault_disable_pagefaults(); 78 error = copyin(udaddr, kaddr, len); 79 vm_fault_enable_pagefaults(save); 80 return (error); 81 } 82 83 int 84 copyout_nofault(const void *kaddr, void *udaddr, size_t len) 85 { 86 int error, save; 87 88 save = vm_fault_disable_pagefaults(); 89 error = copyout(kaddr, udaddr, len); 90 vm_fault_enable_pagefaults(save); 91 return (error); 92 } 93 94 #define PHYS_PAGE_COUNT(len) (howmany(len, PAGE_SIZE) + 1) 95 96 int 97 physcopyin(void *src, vm_paddr_t dst, size_t len) 98 { 99 vm_page_t m[PHYS_PAGE_COUNT(len)]; 100 struct iovec iov[1]; 101 struct uio uio; 102 int i; 103 104 iov[0].iov_base = src; 105 iov[0].iov_len = len; 106 uio.uio_iov = iov; 107 uio.uio_iovcnt = 1; 108 uio.uio_offset = 0; 109 uio.uio_resid = len; 110 uio.uio_segflg = UIO_SYSSPACE; 111 uio.uio_rw = UIO_WRITE; 112 for (i = 0; i < PHYS_PAGE_COUNT(len); i++, dst += PAGE_SIZE) 113 m[i] = PHYS_TO_VM_PAGE(dst); 114 return (uiomove_fromphys(m, dst & PAGE_MASK, len, &uio)); 115 } 116 117 int 118 physcopyout(vm_paddr_t src, void *dst, size_t len) 119 { 120 vm_page_t m[PHYS_PAGE_COUNT(len)]; 121 struct iovec iov[1]; 122 struct uio uio; 123 int i; 124 125 iov[0].iov_base = dst; 126 iov[0].iov_len = len; 127 uio.uio_iov = iov; 128 uio.uio_iovcnt = 1; 129 uio.uio_offset = 0; 130 uio.uio_resid = len; 131 uio.uio_segflg = UIO_SYSSPACE; 132 uio.uio_rw = UIO_READ; 133 for (i = 0; i < PHYS_PAGE_COUNT(len); i++, src += PAGE_SIZE) 134 m[i] = PHYS_TO_VM_PAGE(src); 135 return (uiomove_fromphys(m, src & PAGE_MASK, len, &uio)); 136 } 137 138 #undef PHYS_PAGE_COUNT 139 140 int 141 physcopyin_vlist(bus_dma_segment_t *src, off_t offset, vm_paddr_t dst, 142 size_t len) 143 { 144 size_t seg_len; 145 int error; 146 147 error = 0; 148 while (offset >= src->ds_len) { 149 offset -= src->ds_len; 150 src++; 151 } 152 153 while (len > 0 && error == 0) { 154 seg_len = MIN(src->ds_len - offset, len); 155 error = physcopyin((void *)(uintptr_t)(src->ds_addr + offset), 156 dst, seg_len); 157 offset = 0; 158 src++; 159 len -= seg_len; 160 dst += seg_len; 161 } 162 163 return (error); 164 } 165 166 int 167 physcopyout_vlist(vm_paddr_t src, bus_dma_segment_t *dst, off_t offset, 168 size_t len) 169 { 170 size_t seg_len; 171 int error; 172 173 error = 0; 174 while (offset >= dst->ds_len) { 175 offset -= dst->ds_len; 176 dst++; 177 } 178 179 while (len > 0 && error == 0) { 180 seg_len = MIN(dst->ds_len - offset, len); 181 error = physcopyout(src, (void *)(uintptr_t)(dst->ds_addr + 182 offset), seg_len); 183 offset = 0; 184 dst++; 185 len -= seg_len; 186 src += seg_len; 187 } 188 189 return (error); 190 } 191 192 int 193 uiomove(void *cp, int n, struct uio *uio) 194 { 195 196 return (uiomove_faultflag(cp, n, uio, 0)); 197 } 198 199 int 200 uiomove_nofault(void *cp, int n, struct uio *uio) 201 { 202 203 return (uiomove_faultflag(cp, n, uio, 1)); 204 } 205 206 static int 207 uiomove_faultflag(void *cp, int n, struct uio *uio, int nofault) 208 { 209 struct iovec *iov; 210 size_t cnt; 211 int error, newflags, save; 212 213 save = error = 0; 214 215 KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE, 216 ("uiomove: mode")); 217 KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread, 218 ("uiomove proc")); 219 220 if (uio->uio_segflg == UIO_USERSPACE) { 221 newflags = TDP_DEADLKTREAT; 222 if (nofault) { 223 /* 224 * Fail if a non-spurious page fault occurs. 225 */ 226 newflags |= TDP_NOFAULTING | TDP_RESETSPUR; 227 } else { 228 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, 229 "Calling uiomove()"); 230 } 231 save = curthread_pflags_set(newflags); 232 } else { 233 KASSERT(nofault == 0, ("uiomove: nofault")); 234 } 235 236 while (n > 0 && uio->uio_resid) { 237 iov = uio->uio_iov; 238 cnt = iov->iov_len; 239 if (cnt == 0) { 240 uio->uio_iov++; 241 uio->uio_iovcnt--; 242 continue; 243 } 244 if (cnt > n) 245 cnt = n; 246 247 switch (uio->uio_segflg) { 248 case UIO_USERSPACE: 249 maybe_yield(); 250 if (uio->uio_rw == UIO_READ) 251 error = copyout(cp, iov->iov_base, cnt); 252 else 253 error = copyin(iov->iov_base, cp, cnt); 254 if (error) 255 goto out; 256 break; 257 258 case UIO_SYSSPACE: 259 if (uio->uio_rw == UIO_READ) 260 bcopy(cp, iov->iov_base, cnt); 261 else 262 bcopy(iov->iov_base, cp, cnt); 263 break; 264 case UIO_NOCOPY: 265 break; 266 } 267 iov->iov_base = (char *)iov->iov_base + cnt; 268 iov->iov_len -= cnt; 269 uio->uio_resid -= cnt; 270 uio->uio_offset += cnt; 271 cp = (char *)cp + cnt; 272 n -= cnt; 273 } 274 out: 275 if (save) 276 curthread_pflags_restore(save); 277 return (error); 278 } 279 280 /* 281 * Wrapper for uiomove() that validates the arguments against a known-good 282 * kernel buffer. Currently, uiomove accepts a signed (n) argument, which 283 * is almost definitely a bad thing, so we catch that here as well. We 284 * return a runtime failure, but it might be desirable to generate a runtime 285 * assertion failure instead. 286 */ 287 int 288 uiomove_frombuf(void *buf, int buflen, struct uio *uio) 289 { 290 size_t offset, n; 291 292 if (uio->uio_offset < 0 || uio->uio_resid < 0 || 293 (offset = uio->uio_offset) != uio->uio_offset) 294 return (EINVAL); 295 if (buflen <= 0 || offset >= buflen) 296 return (0); 297 if ((n = buflen - offset) > IOSIZE_MAX) 298 return (EINVAL); 299 return (uiomove((char *)buf + offset, n, uio)); 300 } 301 302 /* 303 * Give next character to user as result of read. 304 */ 305 int 306 ureadc(int c, struct uio *uio) 307 { 308 struct iovec *iov; 309 char *iov_base; 310 311 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, 312 "Calling ureadc()"); 313 314 again: 315 if (uio->uio_iovcnt == 0 || uio->uio_resid == 0) 316 panic("ureadc"); 317 iov = uio->uio_iov; 318 if (iov->iov_len == 0) { 319 uio->uio_iovcnt--; 320 uio->uio_iov++; 321 goto again; 322 } 323 switch (uio->uio_segflg) { 324 case UIO_USERSPACE: 325 if (subyte(iov->iov_base, c) < 0) 326 return (EFAULT); 327 break; 328 329 case UIO_SYSSPACE: 330 iov_base = iov->iov_base; 331 *iov_base = c; 332 break; 333 334 case UIO_NOCOPY: 335 break; 336 } 337 iov->iov_base = (char *)iov->iov_base + 1; 338 iov->iov_len--; 339 uio->uio_resid--; 340 uio->uio_offset++; 341 return (0); 342 } 343 344 int 345 copyiniov(const struct iovec *iovp, u_int iovcnt, struct iovec **iov, int error) 346 { 347 u_int iovlen; 348 349 *iov = NULL; 350 if (iovcnt > UIO_MAXIOV) 351 return (error); 352 iovlen = iovcnt * sizeof (struct iovec); 353 *iov = malloc(iovlen, M_IOV, M_WAITOK); 354 error = copyin(iovp, *iov, iovlen); 355 if (error) { 356 free(*iov, M_IOV); 357 *iov = NULL; 358 } 359 return (error); 360 } 361 362 int 363 copyinuio(const struct iovec *iovp, u_int iovcnt, struct uio **uiop) 364 { 365 struct iovec *iov; 366 struct uio *uio; 367 u_int iovlen; 368 int error, i; 369 370 *uiop = NULL; 371 if (iovcnt > UIO_MAXIOV) 372 return (EINVAL); 373 iovlen = iovcnt * sizeof (struct iovec); 374 uio = malloc(iovlen + sizeof *uio, M_IOV, M_WAITOK); 375 iov = (struct iovec *)(uio + 1); 376 error = copyin(iovp, iov, iovlen); 377 if (error) { 378 free(uio, M_IOV); 379 return (error); 380 } 381 uio->uio_iov = iov; 382 uio->uio_iovcnt = iovcnt; 383 uio->uio_segflg = UIO_USERSPACE; 384 uio->uio_offset = -1; 385 uio->uio_resid = 0; 386 for (i = 0; i < iovcnt; i++) { 387 if (iov->iov_len > IOSIZE_MAX - uio->uio_resid) { 388 free(uio, M_IOV); 389 return (EINVAL); 390 } 391 uio->uio_resid += iov->iov_len; 392 iov++; 393 } 394 *uiop = uio; 395 return (0); 396 } 397 398 struct uio * 399 cloneuio(struct uio *uiop) 400 { 401 struct uio *uio; 402 int iovlen; 403 404 iovlen = uiop->uio_iovcnt * sizeof (struct iovec); 405 uio = malloc(iovlen + sizeof *uio, M_IOV, M_WAITOK); 406 *uio = *uiop; 407 uio->uio_iov = (struct iovec *)(uio + 1); 408 bcopy(uiop->uio_iov, uio->uio_iov, iovlen); 409 return (uio); 410 } 411 412 /* 413 * Map some anonymous memory in user space of size sz, rounded up to the page 414 * boundary. 415 */ 416 int 417 copyout_map(struct thread *td, vm_offset_t *addr, size_t sz) 418 { 419 struct vmspace *vms; 420 int error; 421 vm_size_t size; 422 423 vms = td->td_proc->p_vmspace; 424 425 /* 426 * Map somewhere after heap in process memory. 427 */ 428 *addr = round_page((vm_offset_t)vms->vm_daddr + 429 lim_max(td, RLIMIT_DATA)); 430 431 /* round size up to page boundary */ 432 size = (vm_size_t)round_page(sz); 433 if (size == 0) 434 return (EINVAL); 435 error = vm_mmap_object(&vms->vm_map, addr, size, VM_PROT_READ | 436 VM_PROT_WRITE, VM_PROT_ALL, MAP_PRIVATE | MAP_ANON, NULL, 0, 437 FALSE, td); 438 return (error); 439 } 440 441 /* 442 * Unmap memory in user space. 443 */ 444 int 445 copyout_unmap(struct thread *td, vm_offset_t addr, size_t sz) 446 { 447 vm_map_t map; 448 vm_size_t size; 449 450 if (sz == 0) 451 return (0); 452 453 map = &td->td_proc->p_vmspace->vm_map; 454 size = (vm_size_t)round_page(sz); 455 456 if (vm_map_remove(map, addr, addr + size) != KERN_SUCCESS) 457 return (EINVAL); 458 459 return (0); 460 } 461 462 int32_t 463 fuword32(volatile const void *addr) 464 { 465 int rv; 466 int32_t val; 467 468 rv = fueword32(addr, &val); 469 return (rv == -1 ? -1 : val); 470 } 471 472 #ifdef _LP64 473 int64_t 474 fuword64(volatile const void *addr) 475 { 476 int rv; 477 int64_t val; 478 479 rv = fueword64(addr, &val); 480 return (rv == -1 ? -1 : val); 481 } 482 #endif /* _LP64 */ 483 484 long 485 fuword(volatile const void *addr) 486 { 487 long val; 488 int rv; 489 490 rv = fueword(addr, &val); 491 return (rv == -1 ? -1 : val); 492 } 493 494 uint32_t 495 casuword32(volatile uint32_t *addr, uint32_t old, uint32_t new) 496 { 497 int rv; 498 uint32_t val; 499 500 rv = casueword32(addr, old, &val, new); 501 return (rv == -1 ? -1 : val); 502 } 503 504 u_long 505 casuword(volatile u_long *addr, u_long old, u_long new) 506 { 507 int rv; 508 u_long val; 509 510 rv = casueword(addr, old, &val, new); 511 return (rv == -1 ? -1 : val); 512 } 513