scrub/phase1.c

   1 /*
   2  * Copyright (C) 2018 Oracle.  All Rights Reserved.
   3  *
   4  * Author: Darrick J. Wong <darrick.wong@oracle.com>
   5  *
   6  * This program is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU General Public License
   8  * as published by the Free Software Foundation; either version 2
   9  * of the License, or (at your option) any later version.
  10  *
  11  * This program is distributed in the hope that it would be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14  * GNU General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * along with this program; if not, write the Free Software Foundation,
  18  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
  19  */
  20 #include "xfs.h"
  21 #include <unistd.h>
  22 #include <sys/types.h>
  23 #include <sys/time.h>
  24 #include <sys/resource.h>
  25 #include <sys/statvfs.h>
  26 #include <fcntl.h>
  27 #include <dirent.h>
  28 #include <stdint.h>
  29 #include <pthread.h>
  30 #include "libfrog.h"
  31 #include "workqueue.h"
  32 #include "input.h"
  33 #include "path.h"
  34 #include "handle.h"
  35 #include "bitops.h"
  36 #include "avl64.h"
  37 #include "list.h"
  38 #include "xfs_scrub.h"
  39 #include "common.h"
  40 #include "disk.h"
  41 #include "scrub.h"
  42
  43 /* Phase 1: Find filesystem geometry (and clean up after) */
  44
  45 /* Shut down the filesystem. */
  46 void
  47 xfs_shutdown_fs(
  48         struct scrub_ctx                *ctx)
  49 {
  50         int                             flag;
  51
  52         flag = XFS_FSOP_GOING_FLAGS_LOGFLUSH;
  53         str_info(ctx, ctx->mntpoint, _("Shutting down filesystem!"));
  54         if (ioctl(ctx->mnt_fd, XFS_IOC_GOINGDOWN, &flag))
  55                 str_errno(ctx, ctx->mntpoint);
  56 }
  57
  58 /* Clean up the XFS-specific state data. */
  59 bool
  60 xfs_cleanup_fs(
  61         struct scrub_ctx        *ctx)
  62 {
  63         int                     error;
  64
  65         if (ctx->fshandle)
  66                 free_handle(ctx->fshandle, ctx->fshandle_len);
  67         if (ctx->rtdev)
  68                 disk_close(ctx->rtdev);
  69         if (ctx->logdev)
  70                 disk_close(ctx->logdev);
  71         if (ctx->datadev)
  72                 disk_close(ctx->datadev);
  73         fshandle_destroy();
  74         error = close(ctx->mnt_fd);
  75         if (error)
  76                 str_errno(ctx, _("closing mountpoint fd"));
  77         fs_table_destroy();
  78
  79         return true;
  80 }
  81
  82 /*
  83  * Bind to the mountpoint, read the XFS geometry, bind to the block devices.
  84  * Anything we've already built will be cleaned up by xfs_cleanup_fs.
  85  */
  86 bool
  87 xfs_setup_fs(
  88         struct scrub_ctx                *ctx)
  89 {
  90         int                             error;
  91
  92         /*
  93          * Open the directory with O_NOATIME.  For mountpoints owned
  94          * by root, this should be sufficient to ensure that we have
  95          * CAP_SYS_ADMIN, which we probably need to do anything fancy
  96          * with the (XFS driver) kernel.
  97          */
  98         ctx->mnt_fd = open(ctx->mntpoint, O_RDONLY | O_NOATIME | O_DIRECTORY);
  99         if (ctx->mnt_fd < 0) {
 100                 if (errno == EPERM)
 101                         str_info(ctx, ctx->mntpoint,
 102 _("Must be root to run scrub."));
 103                 else
 104                         str_errno(ctx, ctx->mntpoint);
 105                 return false;
 106         }
 107
 108         error = fstat(ctx->mnt_fd, &ctx->mnt_sb);
 109         if (error) {
 110                 str_errno(ctx, ctx->mntpoint);
 111                 return false;
 112         }
 113         error = fstatvfs(ctx->mnt_fd, &ctx->mnt_sv);
 114         if (error) {
 115                 str_errno(ctx, ctx->mntpoint);
 116                 return false;
 117         }
 118         error = fstatfs(ctx->mnt_fd, &ctx->mnt_sf);
 119         if (error) {
 120                 str_errno(ctx, ctx->mntpoint);
 121                 return false;
 122         }
 123
 124         ctx->nr_io_threads = nproc;
 125         if (verbose) {
 126                 fprintf(stdout, _("%s: using %d threads to scrub.\n"),
 127                                 ctx->mntpoint, scrub_nproc(ctx));
 128                 fflush(stdout);
 129         }
 130
 131         if (!platform_test_xfs_fd(ctx->mnt_fd)) {
 132                 str_info(ctx, ctx->mntpoint,
 133 _("Does not appear to be an XFS filesystem!"));
 134                 return false;
 135         }
 136
 137         /*
 138          * Flush everything out to disk before we start checking.
 139          * This seems to reduce the incidence of stale file handle
 140          * errors when we open things by handle.
 141          */
 142         error = syncfs(ctx->mnt_fd);
 143         if (error) {
 144                 str_errno(ctx, ctx->mntpoint);
 145                 return false;
 146         }
 147
 148         /* Retrieve XFS geometry. */
 149         error = ioctl(ctx->mnt_fd, XFS_IOC_FSGEOMETRY, &ctx->geo);
 150         if (error) {
 151                 str_errno(ctx, ctx->mntpoint);
 152                 return false;
 153         }
 154
 155         ctx->agblklog = log2_roundup(ctx->geo.agblocks);
 156         ctx->blocklog = highbit32(ctx->geo.blocksize);
 157         ctx->inodelog = highbit32(ctx->geo.inodesize);
 158         ctx->inopblog = ctx->blocklog - ctx->inodelog;
 159
 160         error = path_to_fshandle(ctx->mntpoint, &ctx->fshandle,
 161                         &ctx->fshandle_len);
 162         if (error) {
 163                 str_errno(ctx, _("getting fshandle"));
 164                 return false;
 165         }
 166
 167         /* Do we have kernel-assisted metadata scrubbing? */
 168         if (!xfs_can_scrub_fs_metadata(ctx) || !xfs_can_scrub_inode(ctx) ||
 169             !xfs_can_scrub_bmap(ctx) || !xfs_can_scrub_dir(ctx) ||
 170             !xfs_can_scrub_attr(ctx) || !xfs_can_scrub_symlink(ctx) ||
 171             !xfs_can_scrub_parent(ctx)) {
 172                 str_info(ctx, ctx->mntpoint,
 173 _("Kernel metadata scrubbing facility is not available."));
 174                 return false;
 175         }
 176
 177         /* Do we need kernel-assisted metadata repair? */
 178         if (ctx->mode != SCRUB_MODE_DRY_RUN && !xfs_can_repair(ctx)) {
 179                 str_info(ctx, ctx->mntpoint,
 180 _("Kernel metadata repair facility is not available.  Use -n to scrub."));
 181                 return false;
 182         }
 183
 184         /* Did we find the log and rt devices, if they're present? */
 185         if (ctx->geo.logstart == 0 && ctx->fsinfo.fs_log == NULL) {
 186                 str_info(ctx, ctx->mntpoint,
 187 _("Unable to find log device path."));
 188                 return false;
 189         }
 190         if (ctx->geo.rtblocks && ctx->fsinfo.fs_rt == NULL) {
 191                 str_info(ctx, ctx->mntpoint,
 192 _("Unable to find realtime device path."));
 193                 return false;
 194         }
 195
 196         /* Open the raw devices. */
 197         ctx->datadev = disk_open(ctx->fsinfo.fs_name);
 198         if (error) {
 199                 str_errno(ctx, ctx->fsinfo.fs_name);
 200                 return false;
 201         }
 202
 203         if (ctx->fsinfo.fs_log) {
 204                 ctx->logdev = disk_open(ctx->fsinfo.fs_log);
 205                 if (error) {
 206                         str_errno(ctx, ctx->fsinfo.fs_name);
 207                         return false;
 208                 }
 209         }
 210         if (ctx->fsinfo.fs_rt) {
 211                 ctx->rtdev = disk_open(ctx->fsinfo.fs_rt);
 212                 if (error) {
 213                         str_errno(ctx, ctx->fsinfo.fs_name);
 214                         return false;
 215                 }
 216         }
 217
 218         /*
 219          * Everything's set up, which means any failures recorded after
 220          * this point are most probably corruption errors (as opposed to
 221          * purely setup errors).
 222          */
 223         log_info(ctx, _("Invoking online scrub."), ctx);
 224         ctx->scrub_setup_succeeded = true;
 225         return true;
 226 }