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 <stdio.h>
  21 #include <mntent.h>
  22 #include <unistd.h>
  23 #include <sys/types.h>
  24 #include <sys/stat.h>
  25 #include <sys/time.h>
  26 #include <sys/resource.h>
  27 #include <sys/statvfs.h>
  28 #include <sys/vfs.h>
  29 #include <fcntl.h>
  30 #include <dirent.h>
  31 #include <stdint.h>
  32 #include <stdbool.h>
  33 #include <pthread.h>
  34 #include <errno.h>
  35 #include <linux/fs.h>
  36 #include "libfrog.h"
  37 #include "workqueue.h"
  38 #include "input.h"
  39 #include "path.h"
  40 #include "handle.h"
  41 #include "bitops.h"
  42 #include "xfs_arch.h"
  43 #include "xfs_format.h"
  44 #include "avl64.h"
  45 #include "list.h"
  46 #include "xfs_scrub.h"
  47 #include "common.h"
  48 #include "disk.h"
  49 #include "scrub.h"
  50
  51 /* Phase 1: Find filesystem geometry (and clean up after) */
  52
  53 /* Shut down the filesystem. */
  54 void
  55 xfs_shutdown_fs(
  56         struct scrub_ctx                *ctx)
  57 {
  58         int                             flag;
  59
  60         flag = XFS_FSOP_GOING_FLAGS_LOGFLUSH;
  61         str_info(ctx, ctx->mntpoint, _("Shutting down filesystem!"));
  62         if (ioctl(ctx->mnt_fd, XFS_IOC_GOINGDOWN, &flag))
  63                 str_errno(ctx, ctx->mntpoint);
  64 }
  65
  66 /* Clean up the XFS-specific state data. */
  67 bool
  68 xfs_cleanup_fs(
  69         struct scrub_ctx        *ctx)
  70 {
  71         if (ctx->fshandle)
  72                 free_handle(ctx->fshandle, ctx->fshandle_len);
  73         if (ctx->rtdev)
  74                 disk_close(ctx->rtdev);
  75         if (ctx->logdev)
  76                 disk_close(ctx->logdev);
  77         if (ctx->datadev)
  78                 disk_close(ctx->datadev);
  79         fshandle_destroy();
  80         close(ctx->mnt_fd);
  81         fs_table_destroy();
  82
  83         return true;
  84 }
  85
  86 /*
  87  * Bind to the mountpoint, read the XFS geometry, bind to the block devices.
  88  * Anything we've already built will be cleaned up by xfs_cleanup_fs.
  89  */
  90 bool
  91 xfs_setup_fs(
  92         struct scrub_ctx                *ctx)
  93 {
  94         struct fs_path                  *fsp;
  95         int                             error;
  96
  97         /*
  98          * Open the directory with O_NOATIME.  For mountpoints owned
  99          * by root, this should be sufficient to ensure that we have
 100          * CAP_SYS_ADMIN, which we probably need to do anything fancy
 101          * with the (XFS driver) kernel.
 102          */
 103         ctx->mnt_fd = open(ctx->mntpoint, O_RDONLY | O_NOATIME | O_DIRECTORY);
 104         if (ctx->mnt_fd < 0) {
 105                 if (errno == EPERM)
 106                         str_info(ctx, ctx->mntpoint,
 107 _("Must be root to run scrub."));
 108                 else
 109                         str_errno(ctx, ctx->mntpoint);
 110                 return false;
 111         }
 112
 113         error = fstat(ctx->mnt_fd, &ctx->mnt_sb);
 114         if (error) {
 115                 str_errno(ctx, ctx->mntpoint);
 116                 return false;
 117         }
 118         error = fstatvfs(ctx->mnt_fd, &ctx->mnt_sv);
 119         if (error) {
 120                 str_errno(ctx, ctx->mntpoint);
 121                 return false;
 122         }
 123         error = fstatfs(ctx->mnt_fd, &ctx->mnt_sf);
 124         if (error) {
 125                 str_errno(ctx, ctx->mntpoint);
 126                 return false;
 127         }
 128
 129         ctx->nr_io_threads = nproc;
 130         if (verbose) {
 131                 fprintf(stdout, _("%s: using %d threads to scrub.\n"),
 132                                 ctx->mntpoint, scrub_nproc(ctx));
 133                 fflush(stdout);
 134         }
 135
 136         if (!platform_test_xfs_fd(ctx->mnt_fd)) {
 137                 str_info(ctx, ctx->mntpoint,
 138 _("Does not appear to be an XFS filesystem!"));
 139                 return false;
 140         }
 141
 142         /*
 143          * Flush everything out to disk before we start checking.
 144          * This seems to reduce the incidence of stale file handle
 145          * errors when we open things by handle.
 146          */
 147         error = syncfs(ctx->mnt_fd);
 148         if (error) {
 149                 str_errno(ctx, ctx->mntpoint);
 150                 return false;
 151         }
 152
 153         /* Retrieve XFS geometry. */
 154         error = ioctl(ctx->mnt_fd, XFS_IOC_FSGEOMETRY, &ctx->geo);
 155         if (error) {
 156                 str_errno(ctx, ctx->mntpoint);
 157                 return false;
 158         }
 159
 160         ctx->agblklog = log2_roundup(ctx->geo.agblocks);
 161         ctx->blocklog = highbit32(ctx->geo.blocksize);
 162         ctx->inodelog = highbit32(ctx->geo.inodesize);
 163         ctx->inopblog = ctx->blocklog - ctx->inodelog;
 164
 165         error = path_to_fshandle(ctx->mntpoint, &ctx->fshandle,
 166                         &ctx->fshandle_len);
 167         if (error) {
 168                 str_errno(ctx, _("getting fshandle"));
 169                 return false;
 170         }
 171
 172         /* Do we have kernel-assisted metadata scrubbing? */
 173         if (!xfs_can_scrub_fs_metadata(ctx) || !xfs_can_scrub_inode(ctx) ||
 174             !xfs_can_scrub_bmap(ctx) || !xfs_can_scrub_dir(ctx) ||
 175             !xfs_can_scrub_attr(ctx) || !xfs_can_scrub_symlink(ctx) ||
 176             !xfs_can_scrub_parent(ctx)) {
 177                 str_info(ctx, ctx->mntpoint,
 178 _("Kernel metadata scrubbing facility is not available."));
 179                 return false;
 180         }
 181
 182         /* Do we need kernel-assisted metadata repair? */
 183         if (ctx->mode != SCRUB_MODE_DRY_RUN && !xfs_can_repair(ctx)) {
 184                 str_info(ctx, ctx->mntpoint,
 185 _("Kernel metadata repair facility is not available.  Use -n to scrub."));
 186                 return false;
 187         }
 188
 189         /* Go find the XFS devices if we have a usable fsmap. */
 190         fs_table_initialise(0, NULL, 0, NULL);
 191         errno = 0;
 192         fsp = fs_table_lookup(ctx->mntpoint, FS_MOUNT_POINT);
 193         if (!fsp) {
 194                 str_info(ctx, ctx->mntpoint,
 195 _("Unable to find XFS information."));
 196                 return false;
 197         }
 198         memcpy(&ctx->fsinfo, fsp, sizeof(struct fs_path));
 199
 200         /* Did we find the log and rt devices, if they're present? */
 201         if (ctx->geo.logstart == 0 && ctx->fsinfo.fs_log == NULL) {
 202                 str_info(ctx, ctx->mntpoint,
 203 _("Unable to find log device path."));
 204                 return false;
 205         }
 206         if (ctx->geo.rtblocks && ctx->fsinfo.fs_rt == NULL) {
 207                 str_info(ctx, ctx->mntpoint,
 208 _("Unable to find realtime device path."));
 209                 return false;
 210         }
 211
 212         /* Open the raw devices. */
 213         ctx->datadev = disk_open(ctx->fsinfo.fs_name);
 214         if (error) {
 215                 str_errno(ctx, ctx->fsinfo.fs_name);
 216                 return false;
 217         }
 218
 219         if (ctx->fsinfo.fs_log) {
 220                 ctx->logdev = disk_open(ctx->fsinfo.fs_log);
 221                 if (error) {
 222                         str_errno(ctx, ctx->fsinfo.fs_name);
 223                         return false;
 224                 }
 225         }
 226         if (ctx->fsinfo.fs_rt) {
 227                 ctx->rtdev = disk_open(ctx->fsinfo.fs_rt);
 228                 if (error) {
 229                         str_errno(ctx, ctx->fsinfo.fs_name);
 230                         return false;
 231                 }
 232         }
 233
 234         /*
 235          * Everything's set up, which means any failures recorded after
 236          * this point are most probably corruption errors (as opposed to
 237          * purely setup errors).
 238          */
 239         log_info(ctx, _("Invoking online scrub."), ctx);
 240         ctx->scrub_setup_succeeded = true;
 241         return true;
 242 }