1 /* $NetBSD: linux_file64.c,v 1.62 2019/08/23 07:53:36 maxv Exp $ */ 2 3 /*- 4 * Copyright (c) 1995, 1998, 2000, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Frank van der Linden and Eric Haszlakiewicz. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /* 33 * Linux 64bit filesystem calls. Used on 32bit archs, not used on 64bit ones. 34 */ 35 36 #include <sys/cdefs.h> 37 __KERNEL_RCSID(0, "$NetBSD: linux_file64.c,v 1.62 2019/08/23 07:53:36 maxv Exp $"); 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/namei.h> 42 #include <sys/proc.h> 43 #include <sys/dirent.h> 44 #include <sys/file.h> 45 #include <sys/stat.h> 46 #include <sys/filedesc.h> 47 #include <sys/ioctl.h> 48 #include <sys/kernel.h> 49 #include <sys/mount.h> 50 #include <sys/malloc.h> 51 #include <sys/namei.h> 52 #include <sys/vfs_syscalls.h> 53 #include <sys/vnode.h> 54 #include <sys/tty.h> 55 #include <sys/conf.h> 56 57 #include <sys/syscallargs.h> 58 59 #include <compat/linux/common/linux_types.h> 60 #include <compat/linux/common/linux_signal.h> 61 #include <compat/linux/common/linux_fcntl.h> 62 #include <compat/linux/common/linux_util.h> 63 #include <compat/linux/common/linux_machdep.h> 64 #include <compat/linux/common/linux_dirent.h> 65 #include <compat/linux/common/linux_ipc.h> 66 #include <compat/linux/common/linux_sem.h> 67 68 #include <compat/linux/linux_syscallargs.h> 69 70 static void bsd_to_linux_stat(struct stat *, struct linux_stat64 *); 71 72 /* 73 * Convert a NetBSD stat structure to a Linux stat structure. 74 * Only the order of the fields and the padding in the structure 75 * is different. linux_fakedev is a machine-dependent function 76 * which optionally converts device driver major/minor numbers 77 * (XXX horrible, but what can you do against code that compares 78 * things against constant major device numbers? sigh) 79 */ 80 static void 81 bsd_to_linux_stat(struct stat *bsp, struct linux_stat64 *lsp) 82 { 83 memset(lsp, 0, sizeof(*lsp)); 84 lsp->lst_dev = linux_fakedev(bsp->st_dev, 0); 85 lsp->lst_ino = bsp->st_ino; 86 lsp->lst_mode = (linux_mode_t)bsp->st_mode; 87 if (bsp->st_nlink >= (1 << 15)) 88 lsp->lst_nlink = (1 << 15) - 1; 89 else 90 lsp->lst_nlink = (linux_nlink_t)bsp->st_nlink; 91 lsp->lst_uid = bsp->st_uid; 92 lsp->lst_gid = bsp->st_gid; 93 lsp->lst_rdev = linux_fakedev(bsp->st_rdev, 1); 94 lsp->lst_size = bsp->st_size; 95 lsp->lst_blksize = bsp->st_blksize; 96 lsp->lst_blocks = bsp->st_blocks; 97 lsp->lst_atime = bsp->st_atime; 98 lsp->lst_mtime = bsp->st_mtime; 99 lsp->lst_ctime = bsp->st_ctime; 100 # ifdef LINUX_STAT64_HAS_NSEC 101 lsp->lst_atime_nsec = bsp->st_atimensec; 102 lsp->lst_mtime_nsec = bsp->st_mtimensec; 103 lsp->lst_ctime_nsec = bsp->st_ctimensec; 104 # endif 105 # if LINUX_STAT64_HAS_BROKEN_ST_INO 106 lsp->__lst_ino = (linux_ino_t) bsp->st_ino; 107 # endif 108 } 109 110 /* 111 * The stat functions below are plain sailing. stat and lstat are handled 112 * by one function to avoid code duplication. 113 */ 114 int 115 linux_sys_fstat64(struct lwp *l, const struct linux_sys_fstat64_args *uap, register_t *retval) 116 { 117 /* { 118 syscallarg(int) fd; 119 syscallarg(struct linux_stat64 *) sp; 120 } */ 121 struct linux_stat64 tmplst; 122 struct stat tmpst; 123 int error; 124 125 error = do_sys_fstat(SCARG(uap, fd), &tmpst); 126 if (error != 0) 127 return error; 128 129 bsd_to_linux_stat(&tmpst, &tmplst); 130 131 return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst); 132 } 133 134 static int 135 linux_do_stat64(struct lwp *l, const struct linux_sys_stat64_args *uap, register_t *retval, int flags) 136 { 137 struct linux_stat64 tmplst; 138 struct stat tmpst; 139 int error; 140 141 error = do_sys_stat(SCARG(uap, path), flags, &tmpst); 142 if (error != 0) 143 return error; 144 145 bsd_to_linux_stat(&tmpst, &tmplst); 146 147 return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst); 148 } 149 150 int 151 linux_sys_stat64(struct lwp *l, const struct linux_sys_stat64_args *uap, register_t *retval) 152 { 153 /* { 154 syscallarg(const char *) path; 155 syscallarg(struct linux_stat64 *) sp; 156 } */ 157 158 return linux_do_stat64(l, uap, retval, FOLLOW); 159 } 160 161 int 162 linux_sys_lstat64(struct lwp *l, const struct linux_sys_lstat64_args *uap, register_t *retval) 163 { 164 /* { 165 syscallarg(const char *) path; 166 syscallarg(struct linux_stat64 *) sp; 167 } */ 168 169 return linux_do_stat64(l, (const void *)uap, retval, NOFOLLOW); 170 } 171 172 int 173 linux_sys_fstatat64(struct lwp *l, const struct linux_sys_fstatat64_args *uap, register_t *retval) 174 { 175 /* { 176 syscallarg(int) fd; 177 syscallarg(const char *) path; 178 syscallarg(struct linux_stat64 *) sp; 179 syscallarg(int) flag; 180 } */ 181 struct linux_stat64 tmplst; 182 struct stat tmpst; 183 int error, nd_flag; 184 185 if (SCARG(uap, flag) & LINUX_AT_SYMLINK_NOFOLLOW) 186 nd_flag = NOFOLLOW; 187 else 188 nd_flag = FOLLOW; 189 190 error = do_sys_statat(l, SCARG(uap, fd), SCARG(uap, path), nd_flag, &tmpst); 191 if (error != 0) 192 return error; 193 194 bsd_to_linux_stat(&tmpst, &tmplst); 195 196 return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst); 197 } 198 199 #ifndef __alpha__ 200 int 201 linux_sys_truncate64(struct lwp *l, const struct linux_sys_truncate64_args *uap, register_t *retval) 202 { 203 /* { 204 syscallarg(const char *) path; 205 syscallarg(off_t) length; 206 } */ 207 struct sys_truncate_args ta; 208 209 /* Linux doesn't have the 'pad' pseudo-parameter */ 210 SCARG(&ta, path) = SCARG(uap, path); 211 SCARG(&ta, PAD) = 0; 212 SCARG(&ta, length) = SCARG(uap, length); 213 214 return sys_truncate(l, &ta, retval); 215 } 216 217 int 218 linux_sys_ftruncate64(struct lwp *l, const struct linux_sys_ftruncate64_args *uap, register_t *retval) 219 { 220 /* { 221 syscallarg(unsigned int) fd; 222 syscallarg(off_t) length; 223 } */ 224 struct sys_ftruncate_args ta; 225 226 /* Linux doesn't have the 'pad' pseudo-parameter */ 227 SCARG(&ta, fd) = SCARG(uap, fd); 228 SCARG(&ta, PAD) = 0; 229 SCARG(&ta, length) = SCARG(uap, length); 230 231 return sys_ftruncate(l, &ta, retval); 232 } 233 #endif /* __alpha__ */ 234 235 /* 236 * Linux 'readdir' call. This code is mostly taken from the 237 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 238 * an attempt has been made to keep it a little cleaner. 239 * 240 * The d_off field contains the offset of the next valid entry, 241 * unless the older Linux getdents(2), which used to have it set 242 * to the offset of the entry itself. This function also doesn't 243 * need to deal with the old count == 1 glibc problem. 244 * 245 * Read in BSD-style entries, convert them, and copy them out. 246 * 247 * Note that this doesn't handle union-mounted filesystems. 248 */ 249 int 250 linux_sys_getdents64(struct lwp *l, const struct linux_sys_getdents64_args *uap, register_t *retval) 251 { 252 /* { 253 syscallarg(int) fd; 254 syscallarg(struct linux_dirent64 *) dent; 255 syscallarg(unsigned int) count; 256 } */ 257 struct dirent *bdp; 258 struct vnode *vp; 259 char *inp, *tbuf; /* BSD-format */ 260 int len, reclen; /* BSD-format */ 261 char *outp; /* Linux-format */ 262 int resid, linux_reclen = 0; /* Linux-format */ 263 file_t *fp; 264 struct uio auio; 265 struct iovec aiov; 266 struct linux_dirent64 idb; 267 off_t off; /* true file offset */ 268 int buflen, error, eofflag, nbytes; 269 struct vattr va; 270 off_t *cookiebuf = NULL, *cookie; 271 int ncookies; 272 273 /* fd_getvnode() will use the descriptor for us */ 274 if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0) 275 return (error); 276 277 if ((fp->f_flag & FREAD) == 0) { 278 error = EBADF; 279 goto out1; 280 } 281 282 vp = (struct vnode *)fp->f_data; 283 if (vp->v_type != VDIR) { 284 error = ENOTDIR; 285 goto out1; 286 } 287 288 vn_lock(vp, LK_SHARED | LK_RETRY); 289 error = VOP_GETATTR(vp, &va, l->l_cred); 290 VOP_UNLOCK(vp); 291 if (error) 292 goto out1; 293 294 nbytes = SCARG(uap, count); 295 buflen = uimin(MAXBSIZE, nbytes); 296 if (buflen < va.va_blocksize) 297 buflen = va.va_blocksize; 298 tbuf = malloc(buflen, M_TEMP, M_WAITOK); 299 300 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 301 off = fp->f_offset; 302 again: 303 aiov.iov_base = tbuf; 304 aiov.iov_len = buflen; 305 auio.uio_iov = &aiov; 306 auio.uio_iovcnt = 1; 307 auio.uio_rw = UIO_READ; 308 auio.uio_resid = buflen; 309 auio.uio_offset = off; 310 UIO_SETUP_SYSSPACE(&auio); 311 /* 312 * First we read into the malloc'ed buffer, then 313 * we massage it into user space, one record at a time. 314 */ 315 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 316 &ncookies); 317 if (error) 318 goto out; 319 320 inp = tbuf; 321 outp = (void *)SCARG(uap, dent); 322 resid = nbytes; 323 if ((len = buflen - auio.uio_resid) == 0) 324 goto eof; 325 326 for (cookie = cookiebuf; len > 0; len -= reclen) { 327 bdp = (struct dirent *)inp; 328 reclen = bdp->d_reclen; 329 if (reclen & 3) { 330 error = EIO; 331 goto out; 332 } 333 if (bdp->d_fileno == 0) { 334 inp += reclen; /* it is a hole; squish it out */ 335 if (cookie) 336 off = *cookie++; 337 else 338 off += reclen; 339 continue; 340 } 341 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 342 if (reclen > len || resid < linux_reclen) { 343 /* entry too big for buffer, so just stop */ 344 outp++; 345 break; 346 } 347 if (cookie) 348 off = *cookie++; /* each entry points to next */ 349 else 350 off += reclen; 351 /* 352 * Massage in place to make a Linux-shaped dirent (otherwise 353 * we have to worry about touching user memory outside of 354 * the copyout() call). 355 */ 356 memset(&idb, 0, sizeof(idb)); 357 idb.d_ino = bdp->d_fileno; 358 idb.d_type = bdp->d_type; 359 idb.d_off = off; 360 idb.d_reclen = (u_short)linux_reclen; 361 memcpy(idb.d_name, bdp->d_name, MIN(sizeof(idb.d_name), 362 bdp->d_namlen + 1)); 363 if ((error = copyout((void *)&idb, outp, linux_reclen))) 364 goto out; 365 /* advance past this real entry */ 366 inp += reclen; 367 /* advance output past Linux-shaped entry */ 368 outp += linux_reclen; 369 resid -= linux_reclen; 370 } 371 372 /* if we squished out the whole block, try again */ 373 if (outp == (void *)SCARG(uap, dent)) { 374 if (cookiebuf) 375 free(cookiebuf, M_TEMP); 376 cookiebuf = NULL; 377 goto again; 378 } 379 fp->f_offset = off; /* update the vnode offset */ 380 381 eof: 382 *retval = nbytes - resid; 383 out: 384 VOP_UNLOCK(vp); 385 if (cookiebuf) 386 free(cookiebuf, M_TEMP); 387 free(tbuf, M_TEMP); 388 out1: 389 fd_putfile(SCARG(uap, fd)); 390 return error; 391 } 392