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         if (ctx->fshandle)
  64                 free_handle(ctx->fshandle, ctx->fshandle_len);
  65         if (ctx->rtdev)
  66                 disk_close(ctx->rtdev);
  67         if (ctx->logdev)
  68                 disk_close(ctx->logdev);
  69         if (ctx->datadev)
  70                 disk_close(ctx->datadev);
  71         fshandle_destroy();
  72         close(ctx->mnt_fd);
  73         fs_table_destroy();
  74
  75         return true;
  76 }
  77
  78 /*
  79  * Bind to the mountpoint, read the XFS geometry, bind to the block devices.
  80  * Anything we've already built will be cleaned up by xfs_cleanup_fs.
  81  */
  82 bool
  83 xfs_setup_fs(
  84         struct scrub_ctx                *ctx)
  85 {
  86         int                             error;
  87
  88         /*
  89          * Open the directory with O_NOATIME.  For mountpoints owned
  90          * by root, this should be sufficient to ensure that we have
  91          * CAP_SYS_ADMIN, which we probably need to do anything fancy
  92          * with the (XFS driver) kernel.
  93          */
  94         ctx->mnt_fd = open(ctx->mntpoint, O_RDONLY | O_NOATIME | O_DIRECTORY);
  95         if (ctx->mnt_fd < 0) {
  96                 if (errno == EPERM)
  97                         str_info(ctx, ctx->mntpoint,
  98 _("Must be root to run scrub."));
  99                 else
 100                         str_errno(ctx, ctx->mntpoint);
 101                 return false;
 102         }
 103
 104         error = fstat(ctx->mnt_fd, &ctx->mnt_sb);
 105         if (error) {
 106                 str_errno(ctx, ctx->mntpoint);
 107                 return false;
 108         }
 109         error = fstatvfs(ctx->mnt_fd, &ctx->mnt_sv);
 110         if (error) {
 111                 str_errno(ctx, ctx->mntpoint);
 112                 return false;
 113         }
 114         error = fstatfs(ctx->mnt_fd, &ctx->mnt_sf);
 115         if (error) {
 116                 str_errno(ctx, ctx->mntpoint);
 117                 return false;
 118         }
 119
 120         ctx->nr_io_threads = nproc;
 121         if (verbose) {
 122                 fprintf(stdout, _("%s: using %d threads to scrub.\n"),
 123                                 ctx->mntpoint, scrub_nproc(ctx));
 124                 fflush(stdout);
 125         }
 126
 127         if (!platform_test_xfs_fd(ctx->mnt_fd)) {
 128                 str_info(ctx, ctx->mntpoint,
 129 _("Does not appear to be an XFS filesystem!"));
 130                 return false;
 131         }
 132
 133         /*
 134          * Flush everything out to disk before we start checking.
 135          * This seems to reduce the incidence of stale file handle
 136          * errors when we open things by handle.
 137          */
 138         error = syncfs(ctx->mnt_fd);
 139         if (error) {
 140                 str_errno(ctx, ctx->mntpoint);
 141                 return false;
 142         }
 143
 144         /* Retrieve XFS geometry. */
 145         error = ioctl(ctx->mnt_fd, XFS_IOC_FSGEOMETRY, &ctx->geo);
 146         if (error) {
 147                 str_errno(ctx, ctx->mntpoint);
 148                 return false;
 149         }
 150
 151         ctx->agblklog = log2_roundup(ctx->geo.agblocks);
 152         ctx->blocklog = highbit32(ctx->geo.blocksize);
 153         ctx->inodelog = highbit32(ctx->geo.inodesize);
 154         ctx->inopblog = ctx->blocklog - ctx->inodelog;
 155
 156         error = path_to_fshandle(ctx->mntpoint, &ctx->fshandle,
 157                         &ctx->fshandle_len);
 158         if (error) {
 159                 str_errno(ctx, _("getting fshandle"));
 160                 return false;
 161         }
 162
 163         /* Do we have kernel-assisted metadata scrubbing? */
 164         if (!xfs_can_scrub_fs_metadata(ctx) || !xfs_can_scrub_inode(ctx) ||
 165             !xfs_can_scrub_bmap(ctx) || !xfs_can_scrub_dir(ctx) ||
 166             !xfs_can_scrub_attr(ctx) || !xfs_can_scrub_symlink(ctx) ||
 167             !xfs_can_scrub_parent(ctx)) {
 168                 str_info(ctx, ctx->mntpoint,
 169 _("Kernel metadata scrubbing facility is not available."));
 170                 return false;
 171         }
 172
 173         /* Do we need kernel-assisted metadata repair? */
 174         if (ctx->mode != SCRUB_MODE_DRY_RUN && !xfs_can_repair(ctx)) {
 175                 str_info(ctx, ctx->mntpoint,
 176 _("Kernel metadata repair facility is not available.  Use -n to scrub."));
 177                 return false;
 178         }
 179
 180         /* Did we find the log and rt devices, if they're present? */
 181         if (ctx->geo.logstart == 0 && ctx->fsinfo.fs_log == NULL) {
 182                 str_info(ctx, ctx->mntpoint,
 183 _("Unable to find log device path."));
 184                 return false;
 185         }
 186         if (ctx->geo.rtblocks && ctx->fsinfo.fs_rt == NULL) {
 187                 str_info(ctx, ctx->mntpoint,
 188 _("Unable to find realtime device path."));
 189                 return false;
 190         }
 191
 192         /* Open the raw devices. */
 193         ctx->datadev = disk_open(ctx->fsinfo.fs_name);
 194         if (error) {
 195                 str_errno(ctx, ctx->fsinfo.fs_name);
 196                 return false;
 197         }
 198
 199         if (ctx->fsinfo.fs_log) {
 200                 ctx->logdev = disk_open(ctx->fsinfo.fs_log);
 201                 if (error) {
 202                         str_errno(ctx, ctx->fsinfo.fs_name);
 203                         return false;
 204                 }
 205         }
 206         if (ctx->fsinfo.fs_rt) {
 207                 ctx->rtdev = disk_open(ctx->fsinfo.fs_rt);
 208                 if (error) {
 209                         str_errno(ctx, ctx->fsinfo.fs_name);
 210                         return false;
 211                 }
 212         }
 213
 214         /*
 215          * Everything's set up, which means any failures recorded after
 216          * this point are most probably corruption errors (as opposed to
 217          * purely setup errors).
 218          */
 219         log_info(ctx, _("Invoking online scrub."), ctx);
 220         ctx->scrub_setup_succeeded = true;
 221         return true;
 222 }