Upgrade to 4.4.50-rt62
[kvmfornfv.git] / kernel / fs / ext4 / super.c
index c9ab67d..b405a7b 100644 (file)
@@ -958,6 +958,7 @@ static void init_once(void *foo)
        INIT_LIST_HEAD(&ei->i_orphan);
        init_rwsem(&ei->xattr_sem);
        init_rwsem(&ei->i_data_sem);
+       init_rwsem(&ei->i_mmap_sem);
        inode_init_once(&ei->vfs_inode);
 }
 
@@ -1292,9 +1293,9 @@ static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
                return -1;
        }
        if (ext4_has_feature_quota(sb)) {
-               ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
-                        "when QUOTA feature is enabled");
-               return -1;
+               ext4_msg(sb, KERN_INFO, "Journaled quota options "
+                        "ignored when QUOTA feature is enabled");
+               return 1;
        }
        qname = match_strdup(args);
        if (!qname) {
@@ -1657,10 +1658,10 @@ static int handle_mount_opt(struct super_block *sb, char *opt, int token,
                        return -1;
                }
                if (ext4_has_feature_quota(sb)) {
-                       ext4_msg(sb, KERN_ERR,
-                                "Cannot set journaled quota options "
+                       ext4_msg(sb, KERN_INFO,
+                                "Quota format mount options ignored "
                                 "when QUOTA feature is enabled");
-                       return -1;
+                       return 1;
                }
                sbi->s_jquota_fmt = m->mount_opt;
 #endif
@@ -1721,11 +1722,11 @@ static int parse_options(char *options, struct super_block *sb,
 #ifdef CONFIG_QUOTA
        if (ext4_has_feature_quota(sb) &&
            (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
-               ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
-                        "feature is enabled");
-               return 0;
-       }
-       if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
+               ext4_msg(sb, KERN_INFO, "Quota feature enabled, usrquota and grpquota "
+                        "mount options ignored.");
+               clear_opt(sb, USRQUOTA);
+               clear_opt(sb, GRPQUOTA);
+       } else if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
                if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
                        clear_opt(sb, USRQUOTA);
 
@@ -2029,23 +2030,25 @@ failed:
 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
                                   struct ext4_group_desc *gdp)
 {
-       int offset;
+       int offset = offsetof(struct ext4_group_desc, bg_checksum);
        __u16 crc = 0;
        __le32 le_group = cpu_to_le32(block_group);
        struct ext4_sb_info *sbi = EXT4_SB(sb);
 
        if (ext4_has_metadata_csum(sbi->s_sb)) {
                /* Use new metadata_csum algorithm */
-               __le16 save_csum;
                __u32 csum32;
+               __u16 dummy_csum = 0;
 
-               save_csum = gdp->bg_checksum;
-               gdp->bg_checksum = 0;
                csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
                                     sizeof(le_group));
-               csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
-                                    sbi->s_desc_size);
-               gdp->bg_checksum = save_csum;
+               csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
+               csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
+                                    sizeof(dummy_csum));
+               offset += sizeof(dummy_csum);
+               if (offset < sbi->s_desc_size)
+                       csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
+                                            sbi->s_desc_size - offset);
 
                crc = csum32 & 0xFFFF;
                goto out;
@@ -2055,8 +2058,6 @@ static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
        if (!ext4_has_feature_gdt_csum(sb))
                return 0;
 
-       offset = offsetof(struct ext4_group_desc, bg_checksum);
-
        crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
        crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
        crc = crc16(crc, (__u8 *)gdp, offset);
@@ -2092,6 +2093,7 @@ void ext4_group_desc_csum_set(struct super_block *sb, __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);
@@ -2122,6 +2124,11 @@ 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 "
@@ -2129,6 +2136,11 @@ 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 "
@@ -2136,6 +2148,11 @@ 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: "
@@ -2239,6 +2256,16 @@ 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;
@@ -3010,10 +3037,15 @@ 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;
@@ -3103,7 +3135,7 @@ 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;
@@ -3122,16 +3154,14 @@ 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_sb = sb;
        sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
@@ -3277,11 +3307,19 @@ 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, 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, 0)) {
+                       ext4_msg(sb, KERN_WARNING,
+                                "failed to parse options in superblock: %s",
+                                s_mount_opts);
+               }
+               kfree(s_mount_opts);
        }
        sbi->s_def_mount_opt = sbi->s_mount_opt;
        if (!parse_options((char *) data, sb, &journal_devnum,
@@ -3307,6 +3345,11 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
                                 "both data=journal and dax");
                        goto failed_mount;
                }
+               if (ext4_has_feature_encrypt(sb)) {
+                       ext4_msg(sb, KERN_WARNING,
+                                "encrypted files will use data=ordered "
+                                "instead of data journaling mode");
+               }
                if (test_opt(sb, DELALLOC))
                        clear_opt(sb, DELALLOC);
        } else {
@@ -3367,7 +3410,22 @@ 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;
        }
 
@@ -3454,12 +3512,16 @@ 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);
@@ -3499,6 +3561,13 @@ 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 =
@@ -3535,13 +3604,6 @@ 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;
-       }
-
        /* Do we have standard group size of clustersize * 8 blocks ? */
        if (sbi->s_blocks_per_group == clustersize << 3)
                set_opt2(sb, STD_GROUP_SIZE);
@@ -3601,6 +3663,15 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
                        (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
        db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
                   EXT4_DESC_PER_BLOCK(sb);
+       if (ext4_has_feature_meta_bg(sb)) {
+               if (le32_to_cpu(es->s_first_meta_bg) >= db_count) {
+                       ext4_msg(sb, KERN_WARNING,
+                                "first meta block group too large: %u "
+                                "(group descriptor block count %u)",
+                                le32_to_cpu(es->s_first_meta_bg), db_count);
+                       goto failed_mount;
+               }
+       }
        sbi->s_group_desc = ext4_kvmalloc(db_count *
                                          sizeof(struct buffer_head *),
                                          GFP_KERNEL);
@@ -3622,7 +3693,7 @@ 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!");
                ret = -EFSCORRUPTED;
                goto failed_mount2;
@@ -3945,7 +4016,9 @@ no_journal:
 
        if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
                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)
@@ -4015,8 +4088,8 @@ 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 err ? err : ret;
 }
@@ -4936,6 +5009,20 @@ static int ext4_quota_on_mount(struct super_block *sb, int type)
                                        EXT4_SB(sb)->s_jquota_fmt, type);
 }
 
+static void lockdep_set_quota_inode(struct inode *inode, int subclass)
+{
+       struct ext4_inode_info *ei = EXT4_I(inode);
+
+       /* The first argument of lockdep_set_subclass has to be
+        * *exactly* the same as the argument to init_rwsem() --- in
+        * this case, in init_once() --- or lockdep gets unhappy
+        * because the name of the lock is set using the
+        * stringification of the argument to init_rwsem().
+        */
+       (void) ei;      /* shut up clang warning if !CONFIG_LOCKDEP */
+       lockdep_set_subclass(&ei->i_data_sem, subclass);
+}
+
 /*
  * Standard function to be called on quota_on
  */
@@ -4975,8 +5062,12 @@ static int ext4_quota_on(struct super_block *sb, int type, int format_id,
                if (err)
                        return err;
        }
-
-       return dquot_quota_on(sb, type, format_id, path);
+       lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
+       err = dquot_quota_on(sb, type, format_id, path);
+       if (err)
+               lockdep_set_quota_inode(path->dentry->d_inode,
+                                            I_DATA_SEM_NORMAL);
+       return err;
 }
 
 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
@@ -5002,8 +5093,11 @@ static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
 
        /* Don't account quota for quota files to avoid recursion */
        qf_inode->i_flags |= S_NOQUOTA;
+       lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
        err = dquot_enable(qf_inode, type, format_id, flags);
        iput(qf_inode);
+       if (err)
+               lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
 
        return err;
 }