ext4: add sanity checking to count_overhead()
[pandora-kernel.git] / fs / ext4 / super.c
index cc386b2..422ed79 100644 (file)
@@ -463,9 +463,13 @@ static void ext4_handle_error(struct super_block *sb)
                ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
                sb->s_flags |= MS_RDONLY;
        }
-       if (test_opt(sb, ERRORS_PANIC))
+       if (test_opt(sb, ERRORS_PANIC)) {
+               if (EXT4_SB(sb)->s_journal &&
+                 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
+                       return;
                panic("EXT4-fs (device %s): panic forced after error\n",
                        sb->s_id);
+       }
 }
 
 void __ext4_error(struct super_block *sb, const char *function,
@@ -628,8 +632,12 @@ void __ext4_abort(struct super_block *sb, const char *function,
                        jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
                save_error_info(sb, function, line);
        }
-       if (test_opt(sb, ERRORS_PANIC))
+       if (test_opt(sb, ERRORS_PANIC)) {
+               if (EXT4_SB(sb)->s_journal &&
+                 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
+                       return;
                panic("EXT4-fs panic from previous error\n");
+       }
 }
 
 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
@@ -819,7 +827,7 @@ static void ext4_put_super(struct super_block *sb)
                        ext4_abort(sb, "Couldn't clean up the journal");
        }
 
-       del_timer(&sbi->s_err_report);
+       del_timer_sync(&sbi->s_err_report);
        ext4_release_system_zone(sb);
        ext4_mb_release(sb);
        ext4_ext_release(sb);
@@ -857,6 +865,7 @@ static void ext4_put_super(struct super_block *sb)
                dump_orphan_list(sb, sbi);
        J_ASSERT(list_empty(&sbi->s_orphan));
 
+       sync_blockdev(sb->s_bdev);
        invalidate_bdev(sb->s_bdev);
        if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
                /*
@@ -1017,10 +1026,10 @@ static inline void ext4_show_quota_options(struct seq_file *seq,
        }
 
        if (sbi->s_qf_names[USRQUOTA])
-               seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
+               seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
 
        if (sbi->s_qf_names[GRPQUOTA])
-               seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
+               seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
 
        if (test_opt(sb, USRQUOTA))
                seq_puts(seq, ",usrquota");
@@ -1185,7 +1194,7 @@ static struct inode *ext4_nfs_get_inode(struct super_block *sb,
         * Currently we don't know the generation for parent directory, so
         * a generation of 0 means "accept any"
         */
-       inode = ext4_iget(sb, ino);
+       inode = ext4_iget_normal(sb, ino);
        if (IS_ERR(inode))
                return ERR_CAST(inode);
        if (generation && inode->i_generation != generation) {
@@ -1663,8 +1672,6 @@ static int parse_options(char *options, struct super_block *sb,
                                return 0;
                        if (option < 0)
                                return 0;
-                       if (option == 0)
-                               option = EXT4_DEF_MAX_BATCH_TIME;
                        sbi->s_max_batch_time = option;
                        break;
                case Opt_min_batch_time:
@@ -1933,13 +1940,6 @@ set_qf_format:
                                        "not specified");
                        return 0;
                }
-       } else {
-               if (sbi->s_jquota_fmt) {
-                       ext4_msg(sb, KERN_ERR, "journaled quota format "
-                                       "specified with no journaling "
-                                       "enabled");
-                       return 0;
-               }
        }
 #endif
        if (test_opt(sb, DIOREAD_NOLOCK)) {
@@ -2097,6 +2097,7 @@ int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
 
 /* Called at mount-time, super-block is locked */
 static int ext4_check_descriptors(struct super_block *sb,
+                                 ext4_fsblk_t sb_block,
                                  ext4_group_t *first_not_zeroed)
 {
        struct ext4_sb_info *sbi = EXT4_SB(sb);
@@ -2127,6 +2128,11 @@ static int ext4_check_descriptors(struct super_block *sb,
                        grp = i;
 
                block_bitmap = ext4_block_bitmap(sb, gdp);
+               if (block_bitmap == sb_block) {
+                       ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
+                                "Block bitmap for group %u overlaps "
+                                "superblock", i);
+               }
                if (block_bitmap < first_block || block_bitmap > last_block) {
                        ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
                               "Block bitmap for group %u not in group "
@@ -2134,6 +2140,11 @@ static int ext4_check_descriptors(struct super_block *sb,
                        return 0;
                }
                inode_bitmap = ext4_inode_bitmap(sb, gdp);
+               if (inode_bitmap == sb_block) {
+                       ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
+                                "Inode bitmap for group %u overlaps "
+                                "superblock", i);
+               }
                if (inode_bitmap < first_block || inode_bitmap > last_block) {
                        ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
                               "Inode bitmap for group %u not in group "
@@ -2141,6 +2152,11 @@ static int ext4_check_descriptors(struct super_block *sb,
                        return 0;
                }
                inode_table = ext4_inode_table(sb, gdp);
+               if (inode_table == sb_block) {
+                       ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
+                                "Inode table for group %u overlaps "
+                                "superblock", i);
+               }
                if (inode_table < first_block ||
                    inode_table + sbi->s_itb_per_group - 1 > last_block) {
                        ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
@@ -2246,6 +2262,16 @@ static void ext4_orphan_cleanup(struct super_block *sb,
        while (es->s_last_orphan) {
                struct inode *inode;
 
+               /*
+                * We may have encountered an error during cleanup; if
+                * so, skip the rest.
+                */
+               if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
+                       jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
+                       es->s_last_orphan = 0;
+                       break;
+               }
+
                inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
                if (IS_ERR(inode)) {
                        es->s_last_orphan = 0;
@@ -2260,7 +2286,9 @@ static void ext4_orphan_cleanup(struct super_block *sb,
                                __func__, inode->i_ino, inode->i_size);
                        jbd_debug(2, "truncating inode %lu to %lld bytes\n",
                                  inode->i_ino, inode->i_size);
+                       mutex_lock(&inode->i_mutex);
                        ext4_truncate(inode);
+                       mutex_unlock(&inode->i_mutex);
                        nr_truncates++;
                } else {
                        ext4_msg(sb, KERN_DEBUG,
@@ -2724,10 +2752,11 @@ static void print_daily_error_info(unsigned long arg)
        es = sbi->s_es;
 
        if (es->s_error_count)
-               ext4_msg(sb, KERN_NOTICE, "error count: %u",
+               /* fsck newer than v1.41.13 is needed to clean this condition. */
+               ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
                         le32_to_cpu(es->s_error_count));
        if (es->s_first_error_time) {
-               printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
+               printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
                       sb->s_id, le32_to_cpu(es->s_first_error_time),
                       (int) sizeof(es->s_first_error_func),
                       es->s_first_error_func,
@@ -2741,7 +2770,7 @@ static void print_daily_error_info(unsigned long arg)
                printk("\n");
        }
        if (es->s_last_error_time) {
-               printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
+               printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
                       sb->s_id, le32_to_cpu(es->s_last_error_time),
                       (int) sizeof(es->s_last_error_func),
                       es->s_last_error_func,
@@ -3151,10 +3180,15 @@ static int count_overhead(struct super_block *sb, ext4_group_t grp,
                        ext4_set_bit(s++, buf);
                        count++;
                }
-               for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
-                       ext4_set_bit(EXT4_B2C(sbi, s++), buf);
-                       count++;
+               j = ext4_bg_num_gdb(sb, grp);
+               if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
+                       ext4_error(sb, "Invalid number of block group "
+                                  "descriptor blocks: %d", j);
+                       j = EXT4_BLOCKS_PER_GROUP(sb) - s;
                }
+               count += j;
+               for (; j > 0; j--)
+                       ext4_set_bit(EXT4_B2C(sbi, s++), buf);
        }
        if (!count)
                return 0;
@@ -3211,7 +3245,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
        char *orig_data = kstrdup(data, GFP_KERNEL);
        struct buffer_head *bh;
        struct ext4_super_block *es = NULL;
-       struct ext4_sb_info *sbi;
+       struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
        ext4_fsblk_t block;
        ext4_fsblk_t sb_block = get_sb_block(&data);
        ext4_fsblk_t logical_sb_block;
@@ -3231,16 +3265,14 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
        unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
        ext4_group_t first_not_zeroed;
 
-       sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
-       if (!sbi)
-               goto out_free_orig;
+       if ((data && !orig_data) || !sbi)
+               goto out_free_base;
 
        sbi->s_blockgroup_lock =
                kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
-       if (!sbi->s_blockgroup_lock) {
-               kfree(sbi);
-               goto out_free_orig;
-       }
+       if (!sbi->s_blockgroup_lock)
+               goto out_free_base;
+
        sb->s_fs_info = sbi;
        sbi->s_mount_opt = 0;
        sbi->s_resuid = EXT4_DEF_RESUID;
@@ -3349,11 +3381,19 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
         */
        sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
 
-       if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
-                          &journal_devnum, &journal_ioprio, NULL, 0)) {
-               ext4_msg(sb, KERN_WARNING,
-                        "failed to parse options in superblock: %s",
-                        sbi->s_es->s_mount_opts);
+       if (sbi->s_es->s_mount_opts[0]) {
+               char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
+                                             sizeof(sbi->s_es->s_mount_opts),
+                                             GFP_KERNEL);
+               if (!s_mount_opts)
+                       goto failed_mount;
+               if (!parse_options(s_mount_opts, sb, &journal_devnum,
+                                  &journal_ioprio, NULL, 0)) {
+                       ext4_msg(sb, KERN_WARNING,
+                                "failed to parse options in superblock: %s",
+                                s_mount_opts);
+               }
+               kfree(s_mount_opts);
        }
        if (!parse_options((char *) data, sb, &journal_devnum,
                           &journal_ioprio, NULL, 0))
@@ -3370,7 +3410,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
                }
                if (test_opt(sb, DIOREAD_NOLOCK)) {
                        ext4_msg(sb, KERN_ERR, "can't mount with "
-                                "both data=journal and delalloc");
+                                "both data=journal and dioread_nolock");
                        goto failed_mount;
                }
                if (test_opt(sb, DELALLOC))
@@ -3422,7 +3462,22 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
        if (blocksize < EXT4_MIN_BLOCK_SIZE ||
            blocksize > EXT4_MAX_BLOCK_SIZE) {
                ext4_msg(sb, KERN_ERR,
-                      "Unsupported filesystem blocksize %d", blocksize);
+                      "Unsupported filesystem blocksize %d (%d log_block_size)",
+                        blocksize, le32_to_cpu(es->s_log_block_size));
+               goto failed_mount;
+       }
+       if (le32_to_cpu(es->s_log_block_size) >
+           (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
+               ext4_msg(sb, KERN_ERR,
+                        "Invalid log block size: %u",
+                        le32_to_cpu(es->s_log_block_size));
+               goto failed_mount;
+       }
+
+       if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
+               ext4_msg(sb, KERN_ERR,
+                        "Number of reserved GDT blocks insanely large: %d",
+                        le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
                goto failed_mount;
        }
 
@@ -3491,12 +3546,16 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 
        sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
        sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
-       if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
-               goto cantfind_ext4;
 
        sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
        if (sbi->s_inodes_per_block == 0)
                goto cantfind_ext4;
+       if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
+           sbi->s_inodes_per_group > blocksize * 8) {
+               ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
+                        sbi->s_blocks_per_group);
+               goto failed_mount;
+       }
        sbi->s_itb_per_group = sbi->s_inodes_per_group /
                                        sbi->s_inodes_per_block;
        sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
@@ -3532,6 +3591,13 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
                                 "block size (%d)", clustersize, blocksize);
                        goto failed_mount;
                }
+               if (le32_to_cpu(es->s_log_cluster_size) >
+                   (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
+                       ext4_msg(sb, KERN_ERR,
+                                "Invalid log cluster size: %u",
+                                le32_to_cpu(es->s_log_cluster_size));
+                       goto failed_mount;
+               }
                sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
                        le32_to_cpu(es->s_log_block_size);
                sbi->s_clusters_per_group =
@@ -3568,13 +3634,6 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
        }
        sbi->s_cluster_ratio = clustersize / blocksize;
 
-       if (sbi->s_inodes_per_group > blocksize * 8) {
-               ext4_msg(sb, KERN_ERR,
-                      "#inodes per group too big: %lu",
-                      sbi->s_inodes_per_group);
-               goto failed_mount;
-       }
-
        /*
         * Test whether we have more sectors than will fit in sector_t,
         * and whether the max offset is addressable by the page cache.
@@ -3654,7 +3713,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
                        goto failed_mount2;
                }
        }
-       if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
+       if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
                ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
                goto failed_mount2;
        }
@@ -3927,7 +3986,9 @@ no_journal:
                descr = "out journal";
 
        ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
-                "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
+                "Opts: %.*s%s%s", descr,
+                (int) sizeof(sbi->s_es->s_mount_opts),
+                sbi->s_es->s_mount_opts,
                 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
 
        if (es->s_error_count)
@@ -3959,7 +4020,7 @@ failed_mount_wq:
                sbi->s_journal = NULL;
        }
 failed_mount3:
-       del_timer(&sbi->s_err_report);
+       del_timer_sync(&sbi->s_err_report);
        if (sbi->s_flex_groups)
                ext4_kvfree(sbi->s_flex_groups);
        percpu_counter_destroy(&sbi->s_freeclusters_counter);
@@ -3985,8 +4046,8 @@ failed_mount:
 out_fail:
        sb->s_fs_info = NULL;
        kfree(sbi->s_blockgroup_lock);
+out_free_base:
        kfree(sbi);
-out_free_orig:
        kfree(orig_data);
        return ret;
 }
@@ -4537,6 +4598,21 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data)
                goto restore_opts;
        }
 
+       if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
+               if (test_opt2(sb, EXPLICIT_DELALLOC)) {
+                       ext4_msg(sb, KERN_ERR, "can't mount with "
+                                "both data=journal and delalloc");
+                       err = -EINVAL;
+                       goto restore_opts;
+               }
+               if (test_opt(sb, DIOREAD_NOLOCK)) {
+                       ext4_msg(sb, KERN_ERR, "can't mount with "
+                                "both data=journal and dioread_nolock");
+                       err = -EINVAL;
+                       goto restore_opts;
+               }
+       }
+
        if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
                ext4_abort(sb, "Abort forced by user");