kernel/fs/bootfs/bootfs_vnops.c

   1 /*
   2  * This file and its contents are supplied under the terms of the
   3  * Common Development and Distribution License ("CDDL"), version 1.0.
   4  * You may only use this file in accordance with the terms of version
   5  * 1.0 of the CDDL.
   6  *
   7  * A full copy of the text of the CDDL should have accompanied this
   8  * source.  A copy of the CDDL is also available via the Internet at
   9  * http://www.illumos.org/license/CDDL.
  10  */
  11
  12 /*
  13  * Copyright (c) 2015 Joyent, Inc.  All rights reserved.
  14  */
  15
  16 /*
  17  * bootfs vnode operations
  18  */
  19
  20 #include <sys/types.h>
  21 #include <sys/uio.h>
  22 #include <sys/sunddi.h>
  23 #include <sys/errno.h>
  24 #include <sys/vfs.h>
  25 #include <sys/vnode.h>
  26 #include <sys/pathname.h>
  27 #include <sys/mman.h>
  28 #include <sys/fs_subr.h>
  29 #include <sys/policy.h>
  30 #include <sys/sysmacros.h>
  31 #include <sys/dirent.h>
  32 #include <sys/uio.h>
  33 #include <vm/pvn.h>
  34 #include <vm/hat.h>
  35 #include <vm/seg_map.h>
  36 #include <vm/seg_vn.h>
  37 #include <sys/vmsystm.h>
  38
  39 #include <sys/fs/bootfs_impl.h>
  40
  41 /*ARGSUSED*/
  42 static int
  43 bootfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct)
  44 {
  45         return (0);
  46 }
  47
  48 /*ARGSUSED*/
  49 static int
  50 bootfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr,
  51     caller_context_t *ct)
  52 {
  53         return (0);
  54 }
  55
  56 /*ARGSUSED*/
  57 static int
  58 bootfs_read(vnode_t *vp, struct uio *uiop, int ioflag, cred_t *cr,
  59     caller_context_t *ct)
  60 {
  61         int err;
  62         ssize_t sres = uiop->uio_resid;
  63         bootfs_node_t *bnp = vp->v_data;
  64
  65         if (vp->v_type == VDIR)
  66                 return (EISDIR);
  67
  68         if (vp->v_type != VREG)
  69                 return (EINVAL);
  70
  71         if (uiop->uio_loffset < 0)
  72                 return (EINVAL);
  73
  74         if (uiop->uio_loffset >= bnp->bvn_size)
  75                 return (0);
  76
  77         err = 0;
  78         while (uiop->uio_resid != 0) {
  79                 caddr_t base;
  80                 long offset, frem;
  81                 ulong_t poff, segoff;
  82                 size_t bytes;
  83                 int relerr;
  84
  85                 offset = uiop->uio_loffset;
  86                 poff = offset & PAGEOFFSET;
  87                 bytes = MIN(PAGESIZE - poff, uiop->uio_resid);
  88
  89                 frem = bnp->bvn_size - offset;
  90                 if (frem <= 0) {
  91                         err = 0;
  92                         break;
  93                 }
  94
  95                 /* Don't read past EOF */
  96                 bytes = MIN(bytes, frem);
  97
  98                 /*
  99                  * Segmaps are likely larger than our page size, so make sure we
 100                  * have the proper offfset into the resulting segmap data.
 101                  */
 102                 segoff = (offset & PAGEMASK) & MAXBOFFSET;
 103
 104                 base = segmap_getmapflt(segkmap, vp, offset & MAXBMASK, bytes,
 105                     1, S_READ);
 106
 107                 err = uiomove(base + segoff + poff, bytes, UIO_READ, uiop);
 108                 relerr = segmap_release(segkmap, base, 0);
 109
 110                 if (err == 0)
 111                         err = relerr;
 112
 113                 if (err != 0)
 114                         break;
 115         }
 116
 117         /* Even if we had an error in a partial read, return success */
 118         if (uiop->uio_resid > sres)
 119                 err = 0;
 120
 121         gethrestime(&bnp->bvn_attr.va_atime);
 122
 123         return (err);
 124 }
 125
 126 /*ARGSUSED*/
 127 static int
 128 bootfs_ioctl(vnode_t *vp, int cmd, intptr_t data, int flag,
 129     cred_t *cr, int *rvalp, caller_context_t *ct)
 130 {
 131         return (ENOTTY);
 132 }
 133
 134 /*ARGSUSED*/
 135 static int
 136 bootfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
 137     caller_context_t *ct)
 138 {
 139         uint32_t mask;
 140         bootfs_node_t *bpn = (bootfs_node_t *)vp->v_data;
 141
 142         mask = vap->va_mask;
 143         bcopy(&bpn->bvn_attr, vap, sizeof (vattr_t));
 144         vap->va_mask = mask;
 145         return (0);
 146 }
 147
 148 /*ARGSUSED*/
 149 static int
 150 bootfs_access(vnode_t *vp, int mode, int flags, cred_t *cr,
 151     caller_context_t *ct)
 152 {
 153         int shift = 0;
 154         bootfs_node_t *bpn = (bootfs_node_t *)vp->v_data;
 155
 156         if (crgetuid(cr) != bpn->bvn_attr.va_uid) {
 157                 shift += 3;
 158                 if (groupmember(bpn->bvn_attr.va_gid, cr) == 0)
 159                         shift += 3;
 160         }
 161
 162         return (secpolicy_vnode_access2(cr, vp, bpn->bvn_attr.va_uid,
 163             bpn->bvn_attr.va_mode << shift, mode));
 164 }
 165
 166 /*ARGSUSED*/
 167 static int
 168 bootfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct pathname *pnp,
 169     int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
 170     int *direntflags, pathname_t *realpnp)
 171 {
 172         avl_index_t where;
 173         bootfs_node_t sn, *bnp;
 174         bootfs_node_t *bpp = (bootfs_node_t *)dvp->v_data;
 175
 176         if (flags & LOOKUP_XATTR)
 177                 return (EINVAL);
 178
 179         if (bpp->bvn_attr.va_type != VDIR)
 180                 return (ENOTDIR);
 181
 182         if (*nm == '\0' || strcmp(nm, ".") == 0) {
 183                 VN_HOLD(dvp);
 184                 *vpp = dvp;
 185                 return (0);
 186         }
 187
 188         if (strcmp(nm, "..") == 0) {
 189                 VN_HOLD(bpp->bvn_parent->bvn_vnp);
 190                 *vpp = bpp->bvn_parent->bvn_vnp;
 191                 return (0);
 192         }
 193
 194         sn.bvn_name = nm;
 195         bnp = avl_find(&bpp->bvn_dir, &sn, &where);
 196         if (bnp == NULL)
 197                 return (ENOENT);
 198
 199         VN_HOLD(bnp->bvn_vnp);
 200         *vpp = bnp->bvn_vnp;
 201         return (0);
 202 }
 203
 204 /*ARGSUSED*/
 205 static int
 206 bootfs_readdir(vnode_t *vp, struct uio *uiop, cred_t *cr, int *eofp,
 207     caller_context_t *ct, int flags)
 208 {
 209         bootfs_node_t *bnp = (bootfs_node_t *)vp->v_data;
 210         dirent64_t *dp;
 211         void *buf;
 212         ulong_t bsize, brem;
 213         offset_t coff, roff;
 214         int dlen, ret;
 215         bootfs_node_t *dnp;
 216         boolean_t first = B_TRUE;
 217
 218         if (uiop->uio_loffset >= MAXOFF_T) {
 219                 if (eofp != NULL)
 220                         *eofp = 1;
 221                 return (0);
 222         }
 223
 224         if (uiop->uio_iovcnt != 1)
 225                 return (EINVAL);
 226
 227         if (!(uiop->uio_iov->iov_len > 0))
 228                 return (EINVAL);
 229
 230         if (vp->v_type != VDIR)
 231                 return (ENOTDIR);
 232
 233         roff = uiop->uio_loffset;
 234         coff = 0;
 235         brem = bsize = uiop->uio_iov->iov_len;
 236         buf = kmem_alloc(bsize, KM_SLEEP);
 237         dp = buf;
 238
 239         /*
 240          * Recall that offsets here are done based on the name of the dirent
 241          * excluding the null terminator. Therefore `.` is always at 0, `..` is
 242          * always at 1, and then the first real dirent is at 3. This offset is
 243          * what's actually stored when we update the offset in the structure.
 244          */
 245         if (roff == 0) {
 246                 dlen = DIRENT64_RECLEN(1);
 247                 if (first == B_TRUE) {
 248                         if (dlen > brem) {
 249                                 kmem_free(buf, bsize);
 250                                 return (EINVAL);
 251                         }
 252                         first = B_FALSE;
 253                 }
 254                 dp->d_ino = (ino64_t)bnp->bvn_attr.va_nodeid;
 255                 dp->d_off = 0;
 256                 dp->d_reclen = (ushort_t)dlen;
 257                 (void) strncpy(dp->d_name, ".", DIRENT64_NAMELEN(dlen));
 258                 dp = (struct dirent64 *)((uintptr_t)dp + dp->d_reclen);
 259                 brem -= dlen;
 260         }
 261
 262         if (roff <= 1) {
 263                 dlen = DIRENT64_RECLEN(2);
 264                 if (first == B_TRUE) {
 265                         if (dlen > brem) {
 266                                 kmem_free(buf, bsize);
 267                                 return (EINVAL);
 268                         }
 269                         first = B_FALSE;
 270                 }
 271                 dp->d_ino = (ino64_t)bnp->bvn_parent->bvn_attr.va_nodeid;
 272                 dp->d_off = 1;
 273                 dp->d_reclen = (ushort_t)dlen;
 274                 (void) strncpy(dp->d_name, "..", DIRENT64_NAMELEN(dlen));
 275                 dp = (struct dirent64 *)((uintptr_t)dp + dp->d_reclen);
 276                 brem -= dlen;
 277         }
 278
 279         coff = 3;
 280         for (dnp = avl_first(&bnp->bvn_dir); dnp != NULL;
 281             dnp = AVL_NEXT(&bnp->bvn_dir, dnp)) {
 282                 size_t nlen = strlen(dnp->bvn_name);
 283
 284                 if (roff > coff) {
 285                         coff += nlen;
 286                         continue;
 287                 }
 288
 289                 dlen = DIRENT64_RECLEN(nlen);
 290                 if (dlen > brem) {
 291                         if (first == B_TRUE) {
 292                                 kmem_free(buf, bsize);
 293                                 return (EINVAL);
 294                         }
 295                         break;
 296                 }
 297                 first = B_FALSE;
 298
 299                 dp->d_ino = (ino64_t)dnp->bvn_attr.va_nodeid;
 300                 dp->d_off = coff;
 301                 dp->d_reclen = (ushort_t)dlen;
 302                 (void) strncpy(dp->d_name, dnp->bvn_name,
 303                     DIRENT64_NAMELEN(dlen));
 304                 dp = (struct dirent64 *)((uintptr_t)dp + dp->d_reclen);
 305                 brem -= dlen;
 306                 coff += nlen;
 307         }
 308
 309         ret = uiomove(buf, (bsize - brem), UIO_READ, uiop);
 310
 311         if (ret == 0) {
 312                 if (dnp == NULL) {
 313                         coff++;
 314                         if (eofp != NULL)
 315                                 *eofp = 1;
 316                 } else if (eofp != NULL) {
 317                         *eofp = 0;
 318                 }
 319                 uiop->uio_loffset = coff;
 320         }
 321         gethrestime(&bnp->bvn_attr.va_atime);
 322         kmem_free(buf, bsize);
 323         return (ret);
 324 }
 325
 326 /*ARGSUSED*/
 327 static void
 328 bootfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
 329 {
 330 }
 331
 332 /*ARGSUSED*/
 333 static int
 334 bootfs_rwlock(vnode_t *vp, int write_lock, caller_context_t *ct)
 335 {
 336         if (write_lock != 0)
 337                 return (EINVAL);
 338         return (0);
 339 }
 340
 341 /*ARGSUSED*/
 342 static void
 343 bootfs_rwunlock(vnode_t *vp, int write_lock, caller_context_t *ct)
 344 {
 345 }
 346
 347 /*ARGSUSED*/
 348 static int
 349 bootfs_seek(vnode_t *vp, offset_t ooff, offset_t *noffp,
 350     caller_context_t *ct)
 351 {
 352         bootfs_node_t *bnp = (bootfs_node_t *)vp->v_data;
 353         if (vp->v_type == VDIR)
 354                 return (0);
 355         return ((*noffp < 0 || *noffp > bnp->bvn_size ? EINVAL : 0));
 356 }
 357
 358 /*
 359  * We need to fill in a single page of a vnode's memory based on the actual data
 360  * from the kernel. We'll use this node's sliding window into physical memory
 361  * and update one page at a time.
 362  */
 363 /*ARGSUSED*/
 364 static int
 365 bootfs_getapage(vnode_t *vp, uoff_t off, size_t len, uint_t *protp,
 366     page_t *pl[], size_t plsz, struct seg *seg, caddr_t addr, enum seg_rw rw,
 367     cred_t *cr)
 368 {
 369         bootfs_node_t *bnp = vp->v_data;
 370         page_t *pp, *fpp;
 371         pfn_t pfn;
 372
 373         for (;;) {
 374                 /* Easy case where the page exists */
 375                 pp = page_lookup(&vp->v_object, off,
 376                                  rw == S_CREATE ? SE_EXCL : SE_SHARED);
 377                 if (pp != NULL) {
 378                         if (pl != NULL) {
 379                                 pl[0] = pp;
 380                                 pl[1] = NULL;
 381                         } else {
 382                                 page_unlock(pp);
 383                         }
 384                         return (0);
 385                 }
 386
 387                 pp = page_create_va(&vp->v_object, off, PAGESIZE,
 388                                     PG_EXCL | PG_WAIT, seg, addr);
 389
 390                 /*
 391                  * If we didn't get the page, that means someone else beat us to
 392                  * creating this so we need to try again.
 393                  */
 394                 if (pp != NULL)
 395                         break;
 396         }
 397
 398         pfn = btop((bnp->bvn_addr + off) & PAGEMASK);
 399         fpp = page_numtopp_nolock(pfn);
 400
 401         if (ppcopy(fpp, pp) == 0) {
 402                 pvn_read_done(pp, B_ERROR);
 403                 return (EIO);
 404         }
 405
 406         if (pl != NULL) {
 407                 pvn_plist_init(pp, pl, plsz, off, PAGESIZE, rw);
 408         } else {
 409                 pvn_io_done(pp);
 410         }
 411
 412         return (0);
 413 }
 414
 415 /*ARGSUSED*/
 416 static int
 417 bootfs_getpage(vnode_t *vp, offset_t off, size_t len, uint_t *protp,
 418     page_t *pl[], size_t plsz, struct seg *seg, caddr_t addr, enum seg_rw rw,
 419     cred_t *cr, caller_context_t *ct)
 420 {
 421         int err;
 422         bootfs_node_t *bnp = vp->v_data;
 423
 424         if (off + len > bnp->bvn_size + PAGEOFFSET)
 425                 return (EFAULT);
 426
 427         if (protp != NULL)
 428                 *protp = PROT_ALL;
 429
 430         if (len <= PAGESIZE)
 431                 err = bootfs_getapage(vp, (uoff_t)off, len, protp, pl,
 432                     plsz, seg, addr, rw, cr);
 433         else
 434                 err = pvn_getpages(bootfs_getapage, vp, (uoff_t)off, len,
 435                     protp, pl, plsz, seg, addr, rw, cr);
 436
 437         return (err);
 438 }
 439
 440 /*ARGSUSED*/
 441 static int
 442 bootfs_map(vnode_t *vp, offset_t off, struct as *as, caddr_t *addrp,
 443     size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr,
 444     caller_context_t *ct)
 445 {
 446         int ret;
 447         segvn_crargs_t vn_a;
 448
 449 #ifdef  _ILP32
 450         if (len > MAXOFF_T)
 451                 return (ENOMEM);
 452 #endif
 453
 454         if (vp->v_flag & VNOMAP)
 455                 return (ENOSYS);
 456
 457         if (off < 0 || off > MAXOFFSET_T - off)
 458                 return (ENXIO);
 459
 460         if (vp->v_type != VREG)
 461                 return (ENODEV);
 462
 463         if ((prot & PROT_WRITE) && (flags & MAP_SHARED))
 464                 return (ENOTSUP);
 465
 466         as_rangelock(as);
 467         ret = choose_addr(as, addrp, len, off, ADDR_VACALIGN, flags);
 468         if (ret != 0) {
 469                 as_rangeunlock(as);
 470                 return (ret);
 471         }
 472
 473         vn_a.vp = vp;
 474         vn_a.offset = (uoff_t)off;
 475         vn_a.type = flags & MAP_TYPE;
 476         vn_a.prot = prot;
 477         vn_a.maxprot = maxprot;
 478         vn_a.cred = cr;
 479         vn_a.amp = NULL;
 480         vn_a.flags = flags & ~MAP_TYPE;
 481         vn_a.szc = 0;
 482         vn_a.lgrp_mem_policy_flags = 0;
 483
 484         ret = as_map(as, *addrp, len, segvn_create, &vn_a);
 485
 486         as_rangeunlock(as);
 487         return (ret);
 488
 489 }
 490
 491 /*ARGSUSED*/
 492 static int
 493 bootfs_addmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr,
 494     size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr,
 495     caller_context_t *ct)
 496 {
 497         return (0);
 498 }
 499
 500 /*ARGSUSED*/
 501 static int
 502 bootfs_delmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr,
 503     size_t len, uint_t prot, uint_t maxprot, uint_t flags, cred_t *cr,
 504     caller_context_t *ct)
 505 {
 506         return (0);
 507 }
 508
 509 static int
 510 bootfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr,
 511     caller_context_t *ct)
 512 {
 513         int ret;
 514
 515         switch (cmd) {
 516         case _PC_TIMESTAMP_RESOLUTION:
 517                 *valp = 1L;
 518                 ret = 0;
 519                 break;
 520         default:
 521                 ret = fs_pathconf(vp, cmd, valp, cr, ct);
 522         }
 523
 524         return (ret);
 525 }
 526
 527 const struct vnodeops bootfs_vnodeops = {
 528         .vnop_name = "bootfs",
 529         .vop_open = bootfs_open,
 530         .vop_close = bootfs_close,
 531         .vop_read = bootfs_read,
 532         .vop_ioctl = bootfs_ioctl,
 533         .vop_getattr = bootfs_getattr,
 534         .vop_access = bootfs_access,
 535         .vop_lookup = bootfs_lookup,
 536         .vop_readdir = bootfs_readdir,
 537         .vop_inactive = bootfs_inactive,
 538         .vop_rwlock = bootfs_rwlock,
 539         .vop_rwunlock = bootfs_rwunlock,
 540         .vop_seek = bootfs_seek,
 541         .vop_getpage = bootfs_getpage,
 542         .vop_map = bootfs_map,
 543         .vop_addmap = bootfs_addmap,
 544         .vop_delmap = bootfs_delmap,
 545         .vop_pathconf = bootfs_pathconf,
 546 };