fs/xfs/linux-2.6/xfs_file.c

   1 /*
   2  * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or modify it
   5  * under the terms of version 2 of the GNU General Public License as
   6  * published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it would be useful, but
   9  * WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
  11  *
  12  * Further, this software is distributed without any warranty that it is
  13  * free of the rightful claim of any third person regarding infringement
  14  * or the like.  Any license provided herein, whether implied or
  15  * otherwise, applies only to this software file.  Patent licenses, if
  16  * any, provided herein do not apply to combinations of this program with
  17  * other software, or any other product whatsoever.
  18  *
  19  * You should have received a copy of the GNU General Public License along
  20  * with this program; if not, write the Free Software Foundation, Inc., 59
  21  * Temple Place - Suite 330, Boston MA 02111-1307, USA.
  22  *
  23  * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
  24  * Mountain View, CA  94043, or:
  25  *
  26  * http://www.sgi.com
  27  *
  28  * For further information regarding this notice, see:
  29  *
  30  * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
  31  */
  32
  33 #include "xfs.h"
  34 #include "xfs_inum.h"
  35 #include "xfs_log.h"
  36 #include "xfs_sb.h"
  37 #include "xfs_dir.h"
  38 #include "xfs_dir2.h"
  39 #include "xfs_trans.h"
  40 #include "xfs_dmapi.h"
  41 #include "xfs_mount.h"
  42 #include "xfs_bmap_btree.h"
  43 #include "xfs_alloc_btree.h"
  44 #include "xfs_ialloc_btree.h"
  45 #include "xfs_alloc.h"
  46 #include "xfs_btree.h"
  47 #include "xfs_attr_sf.h"
  48 #include "xfs_dir_sf.h"
  49 #include "xfs_dir2_sf.h"
  50 #include "xfs_dinode.h"
  51 #include "xfs_inode.h"
  52 #include "xfs_error.h"
  53 #include "xfs_rw.h"
  54 #include "xfs_ioctl32.h"
  55
  56 #include <linux/dcache.h>
  57 #include <linux/smp_lock.h>
  58
  59 static struct vm_operations_struct linvfs_file_vm_ops;
  60 #ifdef CONFIG_XFS_DMAPI
  61 static struct vm_operations_struct linvfs_dmapi_file_vm_ops;
  62 #endif
  63
  64 STATIC inline ssize_t
  65 __linvfs_read(
  66         struct kiocb            *iocb,
  67         char                    __user *buf,
  68         int                     ioflags,
  69         size_t                  count,
  70         loff_t                  pos)
  71 {
  72         struct iovec            iov = {buf, count};
  73         struct file             *file = iocb->ki_filp;
  74         vnode_t                 *vp = LINVFS_GET_VP(file->f_dentry->d_inode);
  75         ssize_t                 rval;
  76
  77         BUG_ON(iocb->ki_pos != pos);
  78
  79         if (unlikely(file->f_flags & O_DIRECT))
  80                 ioflags |= IO_ISDIRECT;
  81         VOP_READ(vp, iocb, &iov, 1, &iocb->ki_pos, ioflags, NULL, rval);
  82         return rval;
  83 }
  84
  85
  86 STATIC ssize_t
  87 linvfs_aio_read(
  88         struct kiocb            *iocb,
  89         char                    __user *buf,
  90         size_t                  count,
  91         loff_t                  pos)
  92 {
  93         return __linvfs_read(iocb, buf, IO_ISAIO, count, pos);
  94 }
  95
  96 STATIC ssize_t
  97 linvfs_aio_read_invis(
  98         struct kiocb            *iocb,
  99         char                    __user *buf,
 100         size_t                  count,
 101         loff_t                  pos)
 102 {
 103         return __linvfs_read(iocb, buf, IO_ISAIO|IO_INVIS, count, pos);
 104 }
 105
 106
 107 STATIC inline ssize_t
 108 __linvfs_write(
 109         struct kiocb    *iocb,
 110         const char      __user *buf,
 111         int             ioflags,
 112         size_t          count,
 113         loff_t          pos)
 114 {
 115         struct iovec    iov = {(void __user *)buf, count};
 116         struct file     *file = iocb->ki_filp;
 117         struct inode    *inode = file->f_mapping->host;
 118         vnode_t         *vp = LINVFS_GET_VP(inode);
 119         ssize_t         rval;
 120
 121         BUG_ON(iocb->ki_pos != pos);
 122         if (unlikely(file->f_flags & O_DIRECT))
 123                 ioflags |= IO_ISDIRECT;
 124
 125         VOP_WRITE(vp, iocb, &iov, 1, &iocb->ki_pos, ioflags, NULL, rval);
 126         return rval;
 127 }
 128
 129
 130 STATIC ssize_t
 131 linvfs_aio_write(
 132         struct kiocb            *iocb,
 133         const char              __user *buf,
 134         size_t                  count,
 135         loff_t                  pos)
 136 {
 137         return __linvfs_write(iocb, buf, IO_ISAIO, count, pos);
 138 }
 139
 140 STATIC ssize_t
 141 linvfs_aio_write_invis(
 142         struct kiocb            *iocb,
 143         const char              __user *buf,
 144         size_t                  count,
 145         loff_t                  pos)
 146 {
 147         return __linvfs_write(iocb, buf, IO_ISAIO|IO_INVIS, count, pos);
 148 }
 149
 150
 151 STATIC inline ssize_t
 152 __linvfs_readv(
 153         struct file             *file,
 154         const struct iovec      *iov,
 155         int                     ioflags,
 156         unsigned long           nr_segs,
 157         loff_t                  *ppos)
 158 {
 159         struct inode    *inode = file->f_mapping->host;
 160         vnode_t         *vp = LINVFS_GET_VP(inode);
 161         struct          kiocb kiocb;
 162         ssize_t         rval;
 163
 164         init_sync_kiocb(&kiocb, file);
 165         kiocb.ki_pos = *ppos;
 166
 167         if (unlikely(file->f_flags & O_DIRECT))
 168                 ioflags |= IO_ISDIRECT;
 169         VOP_READ(vp, &kiocb, iov, nr_segs, &kiocb.ki_pos, ioflags, NULL, rval);
 170
 171         *ppos = kiocb.ki_pos;
 172         return rval;
 173 }
 174
 175 STATIC ssize_t
 176 linvfs_readv(
 177         struct file             *file,
 178         const struct iovec      *iov,
 179         unsigned long           nr_segs,
 180         loff_t                  *ppos)
 181 {
 182         return __linvfs_readv(file, iov, 0, nr_segs, ppos);
 183 }
 184
 185 STATIC ssize_t
 186 linvfs_readv_invis(
 187         struct file             *file,
 188         const struct iovec      *iov,
 189         unsigned long           nr_segs,
 190         loff_t                  *ppos)
 191 {
 192         return __linvfs_readv(file, iov, IO_INVIS, nr_segs, ppos);
 193 }
 194
 195
 196 STATIC inline ssize_t
 197 __linvfs_writev(
 198         struct file             *file,
 199         const struct iovec      *iov,
 200         int                     ioflags,
 201         unsigned long           nr_segs,
 202         loff_t                  *ppos)
 203 {
 204         struct inode    *inode = file->f_mapping->host;
 205         vnode_t         *vp = LINVFS_GET_VP(inode);
 206         struct          kiocb kiocb;
 207         ssize_t         rval;
 208
 209         init_sync_kiocb(&kiocb, file);
 210         kiocb.ki_pos = *ppos;
 211         if (unlikely(file->f_flags & O_DIRECT))
 212                 ioflags |= IO_ISDIRECT;
 213
 214         VOP_WRITE(vp, &kiocb, iov, nr_segs, &kiocb.ki_pos, ioflags, NULL, rval);
 215
 216         *ppos = kiocb.ki_pos;
 217         return rval;
 218 }
 219
 220
 221 STATIC ssize_t
 222 linvfs_writev(
 223         struct file             *file,
 224         const struct iovec      *iov,
 225         unsigned long           nr_segs,
 226         loff_t                  *ppos)
 227 {
 228         return __linvfs_writev(file, iov, 0, nr_segs, ppos);
 229 }
 230
 231 STATIC ssize_t
 232 linvfs_writev_invis(
 233         struct file             *file,
 234         const struct iovec      *iov,
 235         unsigned long           nr_segs,
 236         loff_t                  *ppos)
 237 {
 238         return __linvfs_writev(file, iov, IO_INVIS, nr_segs, ppos);
 239 }
 240
 241 STATIC ssize_t
 242 linvfs_sendfile(
 243         struct file             *filp,
 244         loff_t                  *ppos,
 245         size_t                  count,
 246         read_actor_t            actor,
 247         void                    *target)
 248 {
 249         vnode_t                 *vp = LINVFS_GET_VP(filp->f_dentry->d_inode);
 250         ssize_t                 rval;
 251
 252         VOP_SENDFILE(vp, filp, ppos, 0, count, actor, target, NULL, rval);
 253         return rval;
 254 }
 255
 256
 257 STATIC int
 258 linvfs_open(
 259         struct inode    *inode,
 260         struct file     *filp)
 261 {
 262         vnode_t         *vp = LINVFS_GET_VP(inode);
 263         int             error;
 264
 265         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
 266                 return -EFBIG;
 267
 268         ASSERT(vp);
 269         VOP_OPEN(vp, NULL, error);
 270         return -error;
 271 }
 272
 273
 274 STATIC int
 275 linvfs_release(
 276         struct inode    *inode,
 277         struct file     *filp)
 278 {
 279         vnode_t         *vp = LINVFS_GET_VP(inode);
 280         int             error = 0;
 281
 282         if (vp)
 283                 VOP_RELEASE(vp, error);
 284         return -error;
 285 }
 286
 287
 288 STATIC int
 289 linvfs_fsync(
 290         struct file     *filp,
 291         struct dentry   *dentry,
 292         int             datasync)
 293 {
 294         struct inode    *inode = dentry->d_inode;
 295         vnode_t         *vp = LINVFS_GET_VP(inode);
 296         int             error;
 297         int             flags = FSYNC_WAIT;
 298
 299         if (datasync)
 300                 flags |= FSYNC_DATA;
 301
 302         ASSERT(vp);
 303         VOP_FSYNC(vp, flags, NULL, (xfs_off_t)0, (xfs_off_t)-1, error);
 304         return -error;
 305 }
 306
 307 /*
 308  * linvfs_readdir maps to VOP_READDIR().
 309  * We need to build a uio, cred, ...
 310  */
 311
 312 #define nextdp(dp)      ((struct xfs_dirent *)((char *)(dp) + (dp)->d_reclen))
 313
 314 STATIC int
 315 linvfs_readdir(
 316         struct file     *filp,
 317         void            *dirent,
 318         filldir_t       filldir)
 319 {
 320         int             error = 0;
 321         vnode_t         *vp;
 322         uio_t           uio;
 323         iovec_t         iov;
 324         int             eof = 0;
 325         caddr_t         read_buf;
 326         int             namelen, size = 0;
 327         size_t          rlen = PAGE_CACHE_SIZE;
 328         xfs_off_t       start_offset, curr_offset;
 329         xfs_dirent_t    *dbp = NULL;
 330
 331         vp = LINVFS_GET_VP(filp->f_dentry->d_inode);
 332         ASSERT(vp);
 333
 334         /* Try fairly hard to get memory */
 335         do {
 336                 if ((read_buf = (caddr_t)kmalloc(rlen, GFP_KERNEL)))
 337                         break;
 338                 rlen >>= 1;
 339         } while (rlen >= 1024);
 340
 341         if (read_buf == NULL)
 342                 return -ENOMEM;
 343
 344         uio.uio_iov = &iov;
 345         uio.uio_segflg = UIO_SYSSPACE;
 346         curr_offset = filp->f_pos;
 347         if (filp->f_pos != 0x7fffffff)
 348                 uio.uio_offset = filp->f_pos;
 349         else
 350                 uio.uio_offset = 0xffffffff;
 351
 352         while (!eof) {
 353                 uio.uio_resid = iov.iov_len = rlen;
 354                 iov.iov_base = read_buf;
 355                 uio.uio_iovcnt = 1;
 356
 357                 start_offset = uio.uio_offset;
 358
 359                 VOP_READDIR(vp, &uio, NULL, &eof, error);
 360                 if ((uio.uio_offset == start_offset) || error) {
 361                         size = 0;
 362                         break;
 363                 }
 364
 365                 size = rlen - uio.uio_resid;
 366                 dbp = (xfs_dirent_t *)read_buf;
 367                 while (size > 0) {
 368                         namelen = strlen(dbp->d_name);
 369
 370                         if (filldir(dirent, dbp->d_name, namelen,
 371                                         (loff_t) curr_offset & 0x7fffffff,
 372                                         (ino_t) dbp->d_ino,
 373                                         DT_UNKNOWN)) {
 374                                 goto done;
 375                         }
 376                         size -= dbp->d_reclen;
 377                         curr_offset = (loff_t)dbp->d_off /* & 0x7fffffff */;
 378                         dbp = nextdp(dbp);
 379                 }
 380         }
 381 done:
 382         if (!error) {
 383                 if (size == 0)
 384                         filp->f_pos = uio.uio_offset & 0x7fffffff;
 385                 else if (dbp)
 386                         filp->f_pos = curr_offset;
 387         }
 388
 389         kfree(read_buf);
 390         return -error;
 391 }
 392
 393 #ifdef CONFIG_XFS_DMAPI
 394 STATIC void
 395 linvfs_mmap_close(
 396         struct vm_area_struct   *vma)
 397 {
 398         xfs_dm_mm_put(vma);
 399 }
 400 #endif /* CONFIG_XFS_DMAPI */
 401
 402 STATIC int
 403 linvfs_file_mmap(
 404         struct file     *filp,
 405         struct vm_area_struct *vma)
 406 {
 407         struct inode    *ip = filp->f_dentry->d_inode;
 408         vnode_t         *vp = LINVFS_GET_VP(ip);
 409         vattr_t         va = { .va_mask = XFS_AT_UPDATIME };
 410         int             error;
 411
 412         vma->vm_ops = &linvfs_file_vm_ops;
 413
 414         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 415                 xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 416
 417                 error = -XFS_SEND_MMAP(mp, vma, 0);
 418                 if (error)
 419                         return error;
 420 #ifdef CONFIG_XFS_DMAPI
 421                 vma->vm_ops = &linvfs_dmapi_file_vm_ops;
 422 #endif
 423         }
 424
 425         VOP_SETATTR(vp, &va, XFS_AT_UPDATIME, NULL, error);
 426         if (!error)
 427                 vn_revalidate(vp);      /* update Linux inode flags */
 428         return 0;
 429 }
 430
 431
 432 STATIC long
 433 linvfs_ioctl(
 434         struct file     *filp,
 435         unsigned int    cmd,
 436         unsigned long   arg)
 437 {
 438         int             error;
 439         struct inode *inode = filp->f_dentry->d_inode;
 440         vnode_t         *vp = LINVFS_GET_VP(inode);
 441
 442         VOP_IOCTL(vp, inode, filp, 0, cmd, (void __user *)arg, error);
 443         VMODIFY(vp);
 444
 445         /* NOTE:  some of the ioctl's return positive #'s as a
 446          *        byte count indicating success, such as
 447          *        readlink_by_handle.  So we don't "sign flip"
 448          *        like most other routines.  This means true
 449          *        errors need to be returned as a negative value.
 450          */
 451         return error;
 452 }
 453
 454 STATIC long
 455 linvfs_ioctl_invis(
 456         struct file     *filp,
 457         unsigned int    cmd,
 458         unsigned long   arg)
 459 {
 460         int             error;
 461         struct inode *inode = filp->f_dentry->d_inode;
 462         vnode_t         *vp = LINVFS_GET_VP(inode);
 463
 464         ASSERT(vp);
 465         VOP_IOCTL(vp, inode, filp, IO_INVIS, cmd, (void __user *)arg, error);
 466         VMODIFY(vp);
 467
 468         /* NOTE:  some of the ioctl's return positive #'s as a
 469          *        byte count indicating success, such as
 470          *        readlink_by_handle.  So we don't "sign flip"
 471          *        like most other routines.  This means true
 472          *        errors need to be returned as a negative value.
 473          */
 474         return error;
 475 }
 476
 477 #ifdef HAVE_VMOP_MPROTECT
 478 STATIC int
 479 linvfs_mprotect(
 480         struct vm_area_struct *vma,
 481         unsigned int    newflags)
 482 {
 483         vnode_t         *vp = LINVFS_GET_VP(vma->vm_file->f_dentry->d_inode);
 484         int             error = 0;
 485
 486         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 487                 if ((vma->vm_flags & VM_MAYSHARE) &&
 488                     (newflags & VM_WRITE) && !(vma->vm_flags & VM_WRITE)) {
 489                         xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 490
 491                         error = XFS_SEND_MMAP(mp, vma, VM_WRITE);
 492                     }
 493         }
 494         return error;
 495 }
 496 #endif /* HAVE_VMOP_MPROTECT */
 497
 498 #ifdef HAVE_FOP_OPEN_EXEC
 499 /* If the user is attempting to execute a file that is offline then
 500  * we have to trigger a DMAPI READ event before the file is marked as busy
 501  * otherwise the invisible I/O will not be able to write to the file to bring
 502  * it back online.
 503  */
 504 STATIC int
 505 linvfs_open_exec(
 506         struct inode    *inode)
 507 {
 508         vnode_t         *vp = LINVFS_GET_VP(inode);
 509         xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 510         int             error = 0;
 511         bhv_desc_t      *bdp;
 512         xfs_inode_t     *ip;
 513
 514         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 515                 bdp = vn_bhv_lookup(VN_BHV_HEAD(vp), &xfs_vnodeops);
 516                 if (!bdp) {
 517                         error = -EINVAL;
 518                         goto open_exec_out;
 519                 }
 520                 ip = XFS_BHVTOI(bdp);
 521                 if (DM_EVENT_ENABLED(vp->v_vfsp, ip, DM_EVENT_READ)) {
 522                         error = -XFS_SEND_DATA(mp, DM_EVENT_READ, vp,
 523                                                0, 0, 0, NULL);
 524                 }
 525         }
 526 open_exec_out:
 527         return error;
 528 }
 529 #endif /* HAVE_FOP_OPEN_EXEC */
 530
 531 /*
 532  * Temporary workaround to the AIO direct IO write problem.
 533  * This code can go and we can revert to do_sync_write once
 534  * the writepage(s) rework is merged.
 535  */
 536 STATIC ssize_t
 537 linvfs_write(
 538         struct file     *filp,
 539         const char      __user *buf,
 540         size_t          len,
 541         loff_t          *ppos)
 542 {
 543         struct kiocb    kiocb;
 544         ssize_t         ret;
 545
 546         init_sync_kiocb(&kiocb, filp);
 547         kiocb.ki_pos = *ppos;
 548         ret = __linvfs_write(&kiocb, buf, 0, len, kiocb.ki_pos);
 549         *ppos = kiocb.ki_pos;
 550         return ret;
 551 }
 552 STATIC ssize_t
 553 linvfs_write_invis(
 554         struct file     *filp,
 555         const char      __user *buf,
 556         size_t          len,
 557         loff_t          *ppos)
 558 {
 559         struct kiocb    kiocb;
 560         ssize_t         ret;
 561
 562         init_sync_kiocb(&kiocb, filp);
 563         kiocb.ki_pos = *ppos;
 564         ret = __linvfs_write(&kiocb, buf, IO_INVIS, len, kiocb.ki_pos);
 565         *ppos = kiocb.ki_pos;
 566         return ret;
 567 }
 568
 569
 570 struct file_operations linvfs_file_operations = {
 571         .llseek         = generic_file_llseek,
 572         .read           = do_sync_read,
 573         .write          = linvfs_write,
 574         .readv          = linvfs_readv,
 575         .writev         = linvfs_writev,
 576         .aio_read       = linvfs_aio_read,
 577         .aio_write      = linvfs_aio_write,
 578         .sendfile       = linvfs_sendfile,
 579         .unlocked_ioctl = linvfs_ioctl,
 580 #ifdef CONFIG_COMPAT
 581         .compat_ioctl   = linvfs_compat_ioctl,
 582 #endif
 583         .mmap           = linvfs_file_mmap,
 584         .open           = linvfs_open,
 585         .release        = linvfs_release,
 586         .fsync          = linvfs_fsync,
 587 #ifdef HAVE_FOP_OPEN_EXEC
 588         .open_exec      = linvfs_open_exec,
 589 #endif
 590 };
 591
 592 struct file_operations linvfs_invis_file_operations = {
 593         .llseek         = generic_file_llseek,
 594         .read           = do_sync_read,
 595         .write          = linvfs_write_invis,
 596         .readv          = linvfs_readv_invis,
 597         .writev         = linvfs_writev_invis,
 598         .aio_read       = linvfs_aio_read_invis,
 599         .aio_write      = linvfs_aio_write_invis,
 600         .sendfile       = linvfs_sendfile,
 601         .unlocked_ioctl = linvfs_ioctl_invis,
 602 #ifdef CONFIG_COMPAT
 603         .compat_ioctl   = linvfs_compat_invis_ioctl,
 604 #endif
 605         .mmap           = linvfs_file_mmap,
 606         .open           = linvfs_open,
 607         .release        = linvfs_release,
 608         .fsync          = linvfs_fsync,
 609 };
 610
 611
 612 struct file_operations linvfs_dir_operations = {
 613         .read           = generic_read_dir,
 614         .readdir        = linvfs_readdir,
 615         .unlocked_ioctl = linvfs_ioctl,
 616 #ifdef CONFIG_COMPAT
 617         .compat_ioctl   = linvfs_compat_ioctl,
 618 #endif
 619         .fsync          = linvfs_fsync,
 620 };
 621
 622 static struct vm_operations_struct linvfs_file_vm_ops = {
 623         .nopage         = filemap_nopage,
 624         .populate       = filemap_populate,
 625 };
 626
 627 #ifdef CONFIG_XFS_DMAPI
 628 static struct vm_operations_struct linvfs_dmapi_file_vm_ops = {
 629         .close          = linvfs_mmap_close,
 630         .nopage         = filemap_nopage,
 631         .populate       = filemap_populate,
 632 #ifdef HAVE_VMOP_MPROTECT
 633         .mprotect       = linvfs_mprotect,
 634 #endif
 635 };
 636 #endif /* CONFIG_XFS_DMAPI */