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 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/kernel.h> 45 #include <sys/limits.h> 46 #include <sys/lock.h> 47 #include <sys/mman.h> 48 #include <sys/proc.h> 49 #include <sys/resourcevar.h> 50 #include <sys/rwlock.h> 51 #include <sys/sched.h> 52 #include <sys/sysctl.h> 53 #include <sys/vnode.h> 54 55 #include <vm/vm.h> 56 #include <vm/vm_param.h> 57 #include <vm/vm_extern.h> 58 #include <vm/vm_page.h> 59 #include <vm/vm_pageout.h> 60 #include <vm/vm_map.h> 61 62 #include <machine/bus.h> 63 64 SYSCTL_INT(_kern, KERN_IOV_MAX, iov_max, CTLFLAG_RD, SYSCTL_NULL_INT_PTR, UIO_MAXIOV, 65 "Maximum number of elements in an I/O vector; sysconf(_SC_IOV_MAX)"); 66 67 static int uiomove_faultflag(void *cp, int n, struct uio *uio, int nofault); 68 69 int 70 copyin_nofault(const void *udaddr, void *kaddr, size_t len) 71 { 72 int error, save; 73 74 save = vm_fault_disable_pagefaults(); 75 error = copyin(udaddr, kaddr, len); 76 vm_fault_enable_pagefaults(save); 77 return (error); 78 } 79 80 int 81 copyout_nofault(const void *kaddr, void *udaddr, size_t len) 82 { 83 int error, save; 84 85 save = vm_fault_disable_pagefaults(); 86 error = copyout(kaddr, udaddr, len); 87 vm_fault_enable_pagefaults(save); 88 return (error); 89 } 90 91 #define PHYS_PAGE_COUNT(len) (howmany(len, PAGE_SIZE) + 1) 92 93 int 94 physcopyin(void *src, vm_paddr_t dst, size_t len) 95 { 96 vm_page_t m[PHYS_PAGE_COUNT(len)]; 97 struct iovec iov[1]; 98 struct uio uio; 99 int i; 100 101 iov[0].iov_base = src; 102 iov[0].iov_len = len; 103 uio.uio_iov = iov; 104 uio.uio_iovcnt = 1; 105 uio.uio_offset = 0; 106 uio.uio_resid = len; 107 uio.uio_segflg = UIO_SYSSPACE; 108 uio.uio_rw = UIO_WRITE; 109 for (i = 0; i < PHYS_PAGE_COUNT(len); i++, dst += PAGE_SIZE) 110 m[i] = PHYS_TO_VM_PAGE(dst); 111 return (uiomove_fromphys(m, dst & PAGE_MASK, len, &uio)); 112 } 113 114 int 115 physcopyout(vm_paddr_t src, void *dst, size_t len) 116 { 117 vm_page_t m[PHYS_PAGE_COUNT(len)]; 118 struct iovec iov[1]; 119 struct uio uio; 120 int i; 121 122 iov[0].iov_base = dst; 123 iov[0].iov_len = len; 124 uio.uio_iov = iov; 125 uio.uio_iovcnt = 1; 126 uio.uio_offset = 0; 127 uio.uio_resid = len; 128 uio.uio_segflg = UIO_SYSSPACE; 129 uio.uio_rw = UIO_READ; 130 for (i = 0; i < PHYS_PAGE_COUNT(len); i++, src += PAGE_SIZE) 131 m[i] = PHYS_TO_VM_PAGE(src); 132 return (uiomove_fromphys(m, src & PAGE_MASK, len, &uio)); 133 } 134 135 #undef PHYS_PAGE_COUNT 136 137 int 138 physcopyin_vlist(bus_dma_segment_t *src, off_t offset, vm_paddr_t dst, 139 size_t len) 140 { 141 size_t seg_len; 142 int error; 143 144 error = 0; 145 while (offset >= src->ds_len) { 146 offset -= src->ds_len; 147 src++; 148 } 149 150 while (len > 0 && error == 0) { 151 seg_len = MIN(src->ds_len - offset, len); 152 error = physcopyin((void *)(uintptr_t)(src->ds_addr + offset), 153 dst, seg_len); 154 offset = 0; 155 src++; 156 len -= seg_len; 157 dst += seg_len; 158 } 159 160 return (error); 161 } 162 163 int 164 physcopyout_vlist(vm_paddr_t src, bus_dma_segment_t *dst, off_t offset, 165 size_t len) 166 { 167 size_t seg_len; 168 int error; 169 170 error = 0; 171 while (offset >= dst->ds_len) { 172 offset -= dst->ds_len; 173 dst++; 174 } 175 176 while (len > 0 && error == 0) { 177 seg_len = MIN(dst->ds_len - offset, len); 178 error = physcopyout(src, (void *)(uintptr_t)(dst->ds_addr + 179 offset), seg_len); 180 offset = 0; 181 dst++; 182 len -= seg_len; 183 src += seg_len; 184 } 185 186 return (error); 187 } 188 189 int 190 uiomove(void *cp, int n, struct uio *uio) 191 { 192 193 return (uiomove_faultflag(cp, n, uio, 0)); 194 } 195 196 int 197 uiomove_nofault(void *cp, int n, struct uio *uio) 198 { 199 200 return (uiomove_faultflag(cp, n, uio, 1)); 201 } 202 203 static int 204 uiomove_faultflag(void *cp, int n, struct uio *uio, int nofault) 205 { 206 struct iovec *iov; 207 size_t cnt; 208 int error, newflags, save; 209 210 save = error = 0; 211 212 KASSERT(uio->uio_rw == UIO_READ || uio->uio_rw == UIO_WRITE, 213 ("uiomove: mode")); 214 KASSERT(uio->uio_segflg != UIO_USERSPACE || uio->uio_td == curthread, 215 ("uiomove proc")); 216 KASSERT(uio->uio_resid >= 0, 217 ("%s: uio %p resid underflow", __func__, uio)); 218 219 if (uio->uio_segflg == UIO_USERSPACE) { 220 newflags = TDP_DEADLKTREAT; 221 if (nofault) { 222 /* 223 * Fail if a non-spurious page fault occurs. 224 */ 225 newflags |= TDP_NOFAULTING | TDP_RESETSPUR; 226 } else { 227 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, 228 "Calling uiomove()"); 229 } 230 save = curthread_pflags_set(newflags); 231 } else { 232 KASSERT(nofault == 0, ("uiomove: nofault")); 233 } 234 235 while (n > 0 && uio->uio_resid) { 236 KASSERT(uio->uio_iovcnt > 0, 237 ("%s: uio %p iovcnt underflow", __func__, uio)); 238 239 iov = uio->uio_iov; 240 cnt = iov->iov_len; 241 if (cnt == 0) { 242 uio->uio_iov++; 243 uio->uio_iovcnt--; 244 continue; 245 } 246 if (cnt > n) 247 cnt = n; 248 249 switch (uio->uio_segflg) { 250 case UIO_USERSPACE: 251 maybe_yield(); 252 if (uio->uio_rw == UIO_READ) 253 error = copyout(cp, iov->iov_base, cnt); 254 else 255 error = copyin(iov->iov_base, cp, cnt); 256 if (error) 257 goto out; 258 break; 259 260 case UIO_SYSSPACE: 261 if (uio->uio_rw == UIO_READ) 262 bcopy(cp, iov->iov_base, cnt); 263 else 264 bcopy(iov->iov_base, cp, cnt); 265 break; 266 case UIO_NOCOPY: 267 break; 268 } 269 iov->iov_base = (char *)iov->iov_base + cnt; 270 iov->iov_len -= cnt; 271 uio->uio_resid -= cnt; 272 uio->uio_offset += cnt; 273 cp = (char *)cp + cnt; 274 n -= cnt; 275 } 276 out: 277 if (save) 278 curthread_pflags_restore(save); 279 return (error); 280 } 281 282 /* 283 * Wrapper for uiomove() that validates the arguments against a known-good 284 * kernel buffer. Currently, uiomove accepts a signed (n) argument, which 285 * is almost definitely a bad thing, so we catch that here as well. We 286 * return a runtime failure, but it might be desirable to generate a runtime 287 * assertion failure instead. 288 */ 289 int 290 uiomove_frombuf(void *buf, int buflen, struct uio *uio) 291 { 292 size_t offset, n; 293 294 if (uio->uio_offset < 0 || uio->uio_resid < 0 || 295 (offset = uio->uio_offset) != uio->uio_offset) 296 return (EINVAL); 297 if (buflen <= 0 || offset >= buflen) 298 return (0); 299 if ((n = buflen - offset) > IOSIZE_MAX) 300 return (EINVAL); 301 return (uiomove((char *)buf + offset, n, uio)); 302 } 303 304 /* 305 * Give next character to user as result of read. 306 */ 307 int 308 ureadc(int c, struct uio *uio) 309 { 310 struct iovec *iov; 311 char *iov_base; 312 313 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, 314 "Calling ureadc()"); 315 316 again: 317 if (uio->uio_iovcnt == 0 || uio->uio_resid == 0) 318 panic("ureadc"); 319 iov = uio->uio_iov; 320 if (iov->iov_len == 0) { 321 uio->uio_iovcnt--; 322 uio->uio_iov++; 323 goto again; 324 } 325 switch (uio->uio_segflg) { 326 case UIO_USERSPACE: 327 if (subyte(iov->iov_base, c) < 0) 328 return (EFAULT); 329 break; 330 331 case UIO_SYSSPACE: 332 iov_base = iov->iov_base; 333 *iov_base = c; 334 break; 335 336 case UIO_NOCOPY: 337 break; 338 } 339 iov->iov_base = (char *)iov->iov_base + 1; 340 iov->iov_len--; 341 uio->uio_resid--; 342 uio->uio_offset++; 343 return (0); 344 } 345 346 int 347 copyiniov(const struct iovec *iovp, u_int iovcnt, struct iovec **iov, int error) 348 { 349 u_int iovlen; 350 351 *iov = NULL; 352 if (iovcnt > UIO_MAXIOV) 353 return (error); 354 iovlen = iovcnt * sizeof (struct iovec); 355 *iov = malloc(iovlen, M_IOV, M_WAITOK); 356 error = copyin(iovp, *iov, iovlen); 357 if (error) { 358 free(*iov, M_IOV); 359 *iov = NULL; 360 } 361 return (error); 362 } 363 364 int 365 copyinuio(const struct iovec *iovp, u_int iovcnt, struct uio **uiop) 366 { 367 struct iovec *iov; 368 struct uio *uio; 369 u_int iovlen; 370 int error, i; 371 372 *uiop = NULL; 373 if (iovcnt > UIO_MAXIOV) 374 return (EINVAL); 375 iovlen = iovcnt * sizeof (struct iovec); 376 uio = malloc(iovlen + sizeof *uio, M_IOV, M_WAITOK); 377 iov = (struct iovec *)(uio + 1); 378 error = copyin(iovp, iov, iovlen); 379 if (error) { 380 free(uio, M_IOV); 381 return (error); 382 } 383 uio->uio_iov = iov; 384 uio->uio_iovcnt = iovcnt; 385 uio->uio_segflg = UIO_USERSPACE; 386 uio->uio_offset = -1; 387 uio->uio_resid = 0; 388 for (i = 0; i < iovcnt; i++) { 389 if (iov->iov_len > IOSIZE_MAX - uio->uio_resid) { 390 free(uio, M_IOV); 391 return (EINVAL); 392 } 393 uio->uio_resid += iov->iov_len; 394 iov++; 395 } 396 *uiop = uio; 397 return (0); 398 } 399 400 struct uio * 401 cloneuio(struct uio *uiop) 402 { 403 struct uio *uio; 404 int iovlen; 405 406 iovlen = uiop->uio_iovcnt * sizeof (struct iovec); 407 uio = malloc(iovlen + sizeof *uio, M_IOV, M_WAITOK); 408 *uio = *uiop; 409 uio->uio_iov = (struct iovec *)(uio + 1); 410 bcopy(uiop->uio_iov, uio->uio_iov, iovlen); 411 return (uio); 412 } 413 414 /* 415 * Map some anonymous memory in user space of size sz, rounded up to the page 416 * boundary. 417 */ 418 int 419 copyout_map(struct thread *td, vm_offset_t *addr, size_t sz) 420 { 421 struct vmspace *vms; 422 int error; 423 vm_size_t size; 424 425 vms = td->td_proc->p_vmspace; 426 427 /* 428 * Map somewhere after heap in process memory. 429 */ 430 *addr = round_page((vm_offset_t)vms->vm_daddr + 431 lim_max(td, RLIMIT_DATA)); 432 433 /* round size up to page boundary */ 434 size = (vm_size_t)round_page(sz); 435 if (size == 0) 436 return (EINVAL); 437 error = vm_mmap_object(&vms->vm_map, addr, size, VM_PROT_READ | 438 VM_PROT_WRITE, VM_PROT_ALL, MAP_PRIVATE | MAP_ANON, NULL, 0, 439 FALSE, td); 440 return (error); 441 } 442 443 /* 444 * Unmap memory in user space. 445 */ 446 int 447 copyout_unmap(struct thread *td, vm_offset_t addr, size_t sz) 448 { 449 vm_map_t map; 450 vm_size_t size; 451 452 if (sz == 0) 453 return (0); 454 455 map = &td->td_proc->p_vmspace->vm_map; 456 size = (vm_size_t)round_page(sz); 457 458 if (vm_map_remove(map, addr, addr + size) != KERN_SUCCESS) 459 return (EINVAL); 460 461 return (0); 462 } 463 464 int32_t 465 fuword32(volatile const void *addr) 466 { 467 int rv; 468 int32_t val; 469 470 rv = fueword32(addr, &val); 471 return (rv == -1 ? -1 : val); 472 } 473 474 #ifdef _LP64 475 int64_t 476 fuword64(volatile const void *addr) 477 { 478 int rv; 479 int64_t val; 480 481 rv = fueword64(addr, &val); 482 return (rv == -1 ? -1 : val); 483 } 484 #endif /* _LP64 */ 485 486 long 487 fuword(volatile const void *addr) 488 { 489 long val; 490 int rv; 491 492 rv = fueword(addr, &val); 493 return (rv == -1 ? -1 : val); 494 } 495 496 uint32_t 497 casuword32(volatile uint32_t *addr, uint32_t old, uint32_t new) 498 { 499 int rv; 500 uint32_t val; 501 502 rv = casueword32(addr, old, &val, new); 503 return (rv == -1 ? -1 : val); 504 } 505 506 u_long 507 casuword(volatile u_long *addr, u_long old, u_long new) 508 { 509 int rv; 510 u_long val; 511 512 rv = casueword(addr, old, &val, new); 513 return (rv == -1 ? -1 : val); 514 } 515