linux_file64.c revision 1.46
1/* $NetBSD: linux_file64.c,v 1.46 2008/03/21 21:54:58 ad 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 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the NetBSD 21 * Foundation, Inc. and its contributors. 22 * 4. Neither the name of The NetBSD Foundation nor the names of its 23 * contributors may be used to endorse or promote products derived 24 * from this software without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 27 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 28 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 29 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 30 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 31 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 32 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 33 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 34 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 35 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 36 * POSSIBILITY OF SUCH DAMAGE. 37 */ 38 39/* 40 * Linux 64bit filesystem calls. Used on 32bit archs, not used on 64bit ones. 41 */ 42 43#include <sys/cdefs.h> 44__KERNEL_RCSID(0, "$NetBSD: linux_file64.c,v 1.46 2008/03/21 21:54:58 ad Exp $"); 45 46#include <sys/param.h> 47#include <sys/systm.h> 48#include <sys/namei.h> 49#include <sys/proc.h> 50#include <sys/dirent.h> 51#include <sys/file.h> 52#include <sys/stat.h> 53#include <sys/filedesc.h> 54#include <sys/ioctl.h> 55#include <sys/kernel.h> 56#include <sys/mount.h> 57#include <sys/malloc.h> 58#include <sys/namei.h> 59#include <sys/vfs_syscalls.h> 60#include <sys/vnode.h> 61#include <sys/tty.h> 62#include <sys/conf.h> 63 64#include <sys/syscallargs.h> 65 66#include <compat/linux/common/linux_types.h> 67#include <compat/linux/common/linux_signal.h> 68#include <compat/linux/common/linux_fcntl.h> 69#include <compat/linux/common/linux_util.h> 70#include <compat/linux/common/linux_machdep.h> 71#include <compat/linux/common/linux_dirent.h> 72#include <compat/linux/common/linux_ipc.h> 73#include <compat/linux/common/linux_sem.h> 74 75#include <compat/linux/linux_syscallargs.h> 76 77#ifndef alpha 78 79static void bsd_to_linux_stat(struct stat *, struct linux_stat64 *); 80 81/* 82 * Convert a NetBSD stat structure to a Linux stat structure. 83 * Only the order of the fields and the padding in the structure 84 * is different. linux_fakedev is a machine-dependent function 85 * which optionally converts device driver major/minor numbers 86 * (XXX horrible, but what can you do against code that compares 87 * things against constant major device numbers? sigh) 88 */ 89static void 90bsd_to_linux_stat(struct stat *bsp, struct linux_stat64 *lsp) 91{ 92 lsp->lst_dev = linux_fakedev(bsp->st_dev, 0); 93 lsp->lst_ino = bsp->st_ino; 94 lsp->lst_mode = (linux_mode_t)bsp->st_mode; 95 if (bsp->st_nlink >= (1 << 15)) 96 lsp->lst_nlink = (1 << 15) - 1; 97 else 98 lsp->lst_nlink = (linux_nlink_t)bsp->st_nlink; 99 lsp->lst_uid = bsp->st_uid; 100 lsp->lst_gid = bsp->st_gid; 101 lsp->lst_rdev = linux_fakedev(bsp->st_rdev, 1); 102 lsp->lst_size = bsp->st_size; 103 lsp->lst_blksize = bsp->st_blksize; 104 lsp->lst_blocks = bsp->st_blocks; 105 lsp->lst_atime = bsp->st_atime; 106 lsp->lst_mtime = bsp->st_mtime; 107 lsp->lst_ctime = bsp->st_ctime; 108# ifdef LINUX_STAT64_HAS_NSEC 109 lsp->lst_atime_nsec = bsp->st_atimensec; 110 lsp->lst_mtime_nsec = bsp->st_mtimensec; 111 lsp->lst_ctime_nsec = bsp->st_ctimensec; 112# endif 113# if LINUX_STAT64_HAS_BROKEN_ST_INO 114 lsp->__lst_ino = (linux_ino_t) bsp->st_ino; 115# endif 116} 117 118/* 119 * The stat functions below are plain sailing. stat and lstat are handled 120 * by one function to avoid code duplication. 121 */ 122int 123linux_sys_fstat64(struct lwp *l, const struct linux_sys_fstat64_args *uap, register_t *retval) 124{ 125 /* { 126 syscallarg(int) fd; 127 syscallarg(struct linux_stat64 *) sp; 128 } */ 129 struct linux_stat64 tmplst; 130 struct stat tmpst; 131 int error; 132 133 error = do_sys_fstat(SCARG(uap, fd), &tmpst); 134 if (error != 0) 135 return error; 136 137 bsd_to_linux_stat(&tmpst, &tmplst); 138 139 return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst); 140} 141 142static int 143linux_do_stat64(struct lwp *l, const struct linux_sys_stat64_args *uap, register_t *retval, int flags) 144{ 145 struct linux_stat64 tmplst; 146 struct stat tmpst; 147 int error; 148 149 error = do_sys_stat(SCARG(uap, path), flags, &tmpst); 150 if (error != 0) 151 return error; 152 153 bsd_to_linux_stat(&tmpst, &tmplst); 154 155 return copyout(&tmplst, SCARG(uap, sp), sizeof tmplst); 156} 157 158int 159linux_sys_stat64(struct lwp *l, const struct linux_sys_stat64_args *uap, register_t *retval) 160{ 161 /* { 162 syscallarg(const char *) path; 163 syscallarg(struct linux_stat64 *) sp; 164 } */ 165 166 return linux_do_stat64(l, uap, retval, FOLLOW); 167} 168 169int 170linux_sys_lstat64(struct lwp *l, const struct linux_sys_lstat64_args *uap, register_t *retval) 171{ 172 /* { 173 syscallarg(const char *) path; 174 syscallarg(struct linux_stat64 *) sp; 175 } */ 176 177 return linux_do_stat64(l, (const void *)uap, retval, NOFOLLOW); 178} 179 180int 181linux_sys_truncate64(struct lwp *l, const struct linux_sys_truncate64_args *uap, register_t *retval) 182{ 183 /* { 184 syscallarg(const char *) path; 185 syscallarg(off_t) length; 186 } */ 187 struct sys_truncate_args ta; 188 189 /* Linux doesn't have the 'pad' pseudo-parameter */ 190 SCARG(&ta, path) = SCARG(uap, path); 191 SCARG(&ta, pad) = 0; 192 SCARG(&ta, length) = SCARG(uap, length); 193 194 return sys_truncate(l, &ta, retval); 195} 196 197int 198linux_sys_ftruncate64(struct lwp *l, const struct linux_sys_ftruncate64_args *uap, register_t *retval) 199{ 200 /* { 201 syscallarg(unsigned int) fd; 202 syscallarg(off_t) length; 203 } */ 204 struct sys_ftruncate_args ta; 205 206 /* Linux doesn't have the 'pad' pseudo-parameter */ 207 SCARG(&ta, fd) = SCARG(uap, fd); 208 SCARG(&ta, pad) = 0; 209 SCARG(&ta, length) = SCARG(uap, length); 210 211 return sys_ftruncate(l, &ta, retval); 212} 213#endif /* !alpha */ 214 215/* 216 * Linux 'readdir' call. This code is mostly taken from the 217 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 218 * an attempt has been made to keep it a little cleaner. 219 * 220 * The d_off field contains the offset of the next valid entry, 221 * unless the older Linux getdents(2), which used to have it set 222 * to the offset of the entry itself. This function also doesn't 223 * need to deal with the old count == 1 glibc problem. 224 * 225 * Read in BSD-style entries, convert them, and copy them out. 226 * 227 * Note that this doesn't handle union-mounted filesystems. 228 */ 229int 230linux_sys_getdents64(struct lwp *l, const struct linux_sys_getdents64_args *uap, register_t *retval) 231{ 232 /* { 233 syscallarg(int) fd; 234 syscallarg(struct linux_dirent64 *) dent; 235 syscallarg(unsigned int) count; 236 } */ 237 struct dirent *bdp; 238 struct vnode *vp; 239 char *inp, *tbuf; /* BSD-format */ 240 int len, reclen; /* BSD-format */ 241 char *outp; /* Linux-format */ 242 int resid, linux_reclen = 0; /* Linux-format */ 243 file_t *fp; 244 struct uio auio; 245 struct iovec aiov; 246 struct linux_dirent64 idb; 247 off_t off; /* true file offset */ 248 int buflen, error, eofflag, nbytes; 249 struct vattr va; 250 off_t *cookiebuf = NULL, *cookie; 251 int ncookies; 252 253 /* getvnode() will use the descriptor for us */ 254 if ((error = getvnode(SCARG(uap, fd), &fp)) != 0) 255 return (error); 256 257 if ((fp->f_flag & FREAD) == 0) { 258 error = EBADF; 259 goto out1; 260 } 261 262 vp = (struct vnode *)fp->f_data; 263 if (vp->v_type != VDIR) { 264 error = EINVAL; 265 goto out1; 266 } 267 268 if ((error = VOP_GETATTR(vp, &va, l->l_cred))) 269 goto out1; 270 271 nbytes = SCARG(uap, count); 272 buflen = min(MAXBSIZE, nbytes); 273 if (buflen < va.va_blocksize) 274 buflen = va.va_blocksize; 275 tbuf = malloc(buflen, M_TEMP, M_WAITOK); 276 277 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 278 off = fp->f_offset; 279again: 280 aiov.iov_base = tbuf; 281 aiov.iov_len = buflen; 282 auio.uio_iov = &aiov; 283 auio.uio_iovcnt = 1; 284 auio.uio_rw = UIO_READ; 285 auio.uio_resid = buflen; 286 auio.uio_offset = off; 287 UIO_SETUP_SYSSPACE(&auio); 288 /* 289 * First we read into the malloc'ed buffer, then 290 * we massage it into user space, one record at a time. 291 */ 292 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 293 &ncookies); 294 if (error) 295 goto out; 296 297 inp = tbuf; 298 outp = (void *)SCARG(uap, dent); 299 resid = nbytes; 300 if ((len = buflen - auio.uio_resid) == 0) 301 goto eof; 302 303 for (cookie = cookiebuf; len > 0; len -= reclen) { 304 bdp = (struct dirent *)inp; 305 reclen = bdp->d_reclen; 306 if (reclen & 3) 307 panic("linux_readdir"); 308 if (bdp->d_fileno == 0) { 309 inp += reclen; /* it is a hole; squish it out */ 310 if (cookie) 311 off = *cookie++; 312 else 313 off += reclen; 314 continue; 315 } 316 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 317 if (reclen > len || resid < linux_reclen) { 318 /* entry too big for buffer, so just stop */ 319 outp++; 320 break; 321 } 322 if (cookie) 323 off = *cookie++; /* each entry points to next */ 324 else 325 off += reclen; 326 /* 327 * Massage in place to make a Linux-shaped dirent (otherwise 328 * we have to worry about touching user memory outside of 329 * the copyout() call). 330 */ 331 idb.d_ino = bdp->d_fileno; 332 idb.d_type = bdp->d_type; 333 idb.d_off = off; 334 idb.d_reclen = (u_short)linux_reclen; 335 strcpy(idb.d_name, bdp->d_name); 336 if ((error = copyout((void *)&idb, outp, linux_reclen))) 337 goto out; 338 /* advance past this real entry */ 339 inp += reclen; 340 /* advance output past Linux-shaped entry */ 341 outp += linux_reclen; 342 resid -= linux_reclen; 343 } 344 345 /* if we squished out the whole block, try again */ 346 if (outp == (void *)SCARG(uap, dent)) 347 goto again; 348 fp->f_offset = off; /* update the vnode offset */ 349 350eof: 351 *retval = nbytes - resid; 352out: 353 VOP_UNLOCK(vp, 0); 354 if (cookiebuf) 355 free(cookiebuf, M_TEMP); 356 free(tbuf, M_TEMP); 357out1: 358 fd_putfile(SCARG(uap, fd)); 359 return error; 360} 361