diff options
Diffstat (limited to 'fs/incfs')
-rw-r--r-- | fs/incfs/data_mgmt.c | 690 | ||||
-rw-r--r-- | fs/incfs/data_mgmt.h | 130 | ||||
-rw-r--r-- | fs/incfs/format.c | 126 | ||||
-rw-r--r-- | fs/incfs/format.h | 67 | ||||
-rw-r--r-- | fs/incfs/integrity.c | 183 | ||||
-rw-r--r-- | fs/incfs/integrity.h | 20 | ||||
-rw-r--r-- | fs/incfs/vfs.c | 552 |
7 files changed, 1035 insertions, 733 deletions
diff --git a/fs/incfs/data_mgmt.c b/fs/incfs/data_mgmt.c index 4698f14bbdf7..d9c43d5cca19 100644 --- a/fs/incfs/data_mgmt.c +++ b/fs/incfs/data_mgmt.c @@ -27,28 +27,19 @@ struct mount_info *incfs_alloc_mount_info(struct super_block *sb, return ERR_PTR(-ENOMEM); mi->mi_sb = sb; - mi->mi_options = *options; mi->mi_backing_dir_path = *backing_dir_path; mi->mi_owner = get_current_cred(); path_get(&mi->mi_backing_dir_path); mutex_init(&mi->mi_dir_struct_mutex); mutex_init(&mi->mi_pending_reads_mutex); init_waitqueue_head(&mi->mi_pending_reads_notif_wq); + init_waitqueue_head(&mi->mi_log.ml_notif_wq); + spin_lock_init(&mi->mi_log.rl_lock); INIT_LIST_HEAD(&mi->mi_reads_list_head); - if (options->read_log_pages != 0) { - size_t buf_size = PAGE_SIZE * options->read_log_pages; - - spin_lock_init(&mi->mi_log.rl_writer_lock); - init_waitqueue_head(&mi->mi_log.ml_notif_wq); - - mi->mi_log.rl_size = buf_size / sizeof(*mi->mi_log.rl_ring_buf); - mi->mi_log.rl_ring_buf = kzalloc(buf_size, GFP_NOFS); - if (!mi->mi_log.rl_ring_buf) { - error = -ENOMEM; - goto err; - } - } + error = incfs_realloc_mount_info(mi, options); + if (error) + goto err; return mi; @@ -57,6 +48,47 @@ err: return ERR_PTR(error); } +int incfs_realloc_mount_info(struct mount_info *mi, + struct mount_options *options) +{ + void *new_buffer = NULL; + void *old_buffer; + size_t new_buffer_size = 0; + + if (options->read_log_pages != mi->mi_options.read_log_pages) { + struct read_log_state log_state; + /* + * Even though having two buffers allocated at once isn't + * usually good, allocating a multipage buffer under a spinlock + * is even worse, so let's optimize for the shorter lock + * duration. It's not end of the world if we fail to increase + * the buffer size anyway. + */ + if (options->read_log_pages > 0) { + new_buffer_size = PAGE_SIZE * options->read_log_pages; + new_buffer = kzalloc(new_buffer_size, GFP_NOFS); + if (!new_buffer) + return -ENOMEM; + } + + spin_lock(&mi->mi_log.rl_lock); + old_buffer = mi->mi_log.rl_ring_buf; + mi->mi_log.rl_ring_buf = new_buffer; + mi->mi_log.rl_size = new_buffer_size; + log_state = (struct read_log_state){ + .generation_id = mi->mi_log.rl_head.generation_id + 1, + }; + mi->mi_log.rl_head = log_state; + mi->mi_log.rl_tail = log_state; + spin_unlock(&mi->mi_log.rl_lock); + + kfree(old_buffer); + } + + mi->mi_options = *options; + return 0; +} + void incfs_free_mount_info(struct mount_info *mi) { if (!mi) @@ -68,6 +100,8 @@ void incfs_free_mount_info(struct mount_info *mi) mutex_destroy(&mi->mi_pending_reads_mutex); put_cred(mi->mi_owner); kfree(mi->mi_log.rl_ring_buf); + kfree(mi->log_xattr); + kfree(mi->pending_read_xattr); kfree(mi); } @@ -85,11 +119,11 @@ static void data_file_segment_destroy(struct data_file_segment *segment) struct data_file *incfs_open_data_file(struct mount_info *mi, struct file *bf) { - struct data_file *df; - struct backing_file_context *bfc; + struct data_file *df = NULL; + struct backing_file_context *bfc = NULL; int md_records; u64 size; - int error; + int error = 0; int i; if (!bf || !mi) @@ -116,8 +150,8 @@ struct data_file *incfs_open_data_file(struct mount_info *mi, struct file *bf) error = mutex_lock_interruptible(&bfc->bc_mutex); if (error) goto out; - error = incfs_read_file_header(bfc, &df->df_metadata_off, - &df->df_id, &size); + error = incfs_read_file_header(bfc, &df->df_metadata_off, &df->df_id, + &size, &df->df_header_flags); mutex_unlock(&bfc->bc_mutex); if (error) @@ -125,7 +159,7 @@ struct data_file *incfs_open_data_file(struct mount_info *mi, struct file *bf) df->df_size = size; if (size > 0) - df->df_block_count = get_blocks_count_for_size(size); + df->df_data_block_count = get_blocks_count_for_size(size); md_records = incfs_scan_metadata_chain(df); if (md_records < 0) @@ -160,7 +194,7 @@ int make_inode_ready_for_data_ops(struct mount_info *mi, struct file *backing_file) { struct inode_info *node = get_incfs_node(inode); - struct data_file *df; + struct data_file *df = NULL; int err = 0; inode_lock(inode); @@ -181,7 +215,7 @@ int make_inode_ready_for_data_ops(struct mount_info *mi, struct dir_file *incfs_open_dir_file(struct mount_info *mi, struct file *bf) { - struct dir_file *dir; + struct dir_file *dir = NULL; if (!S_ISDIR(bf->f_inode->i_mode)) return ERR_PTR(-EBADF); @@ -214,33 +248,120 @@ static ssize_t decompress(struct mem_range src, struct mem_range dst) return result; } +static void log_read_one_record(struct read_log *rl, struct read_log_state *rs) +{ + union log_record *record = + (union log_record *)((u8 *)rl->rl_ring_buf + rs->next_offset); + size_t record_size; + + switch (record->full_record.type) { + case FULL: + rs->base_record = record->full_record; + record_size = sizeof(record->full_record); + break; + + case SAME_FILE: + rs->base_record.block_index = + record->same_file_record.block_index; + rs->base_record.absolute_ts_us += + record->same_file_record.relative_ts_us; + record_size = sizeof(record->same_file_record); + break; + + case SAME_FILE_NEXT_BLOCK: + ++rs->base_record.block_index; + rs->base_record.absolute_ts_us += + record->same_file_next_block.relative_ts_us; + record_size = sizeof(record->same_file_next_block); + break; + + case SAME_FILE_NEXT_BLOCK_SHORT: + ++rs->base_record.block_index; + rs->base_record.absolute_ts_us += + record->same_file_next_block_short.relative_ts_us; + record_size = sizeof(record->same_file_next_block_short); + break; + } + + rs->next_offset += record_size; + if (rs->next_offset > rl->rl_size - sizeof(*record)) { + rs->next_offset = 0; + ++rs->current_pass_no; + } + ++rs->current_record_no; +} + static void log_block_read(struct mount_info *mi, incfs_uuid_t *id, - int block_index, bool timed_out) + int block_index) { struct read_log *log = &mi->mi_log; - struct read_log_state state; + struct read_log_state *head, *tail; s64 now_us = ktime_to_us(ktime_get()); - struct read_log_record record = { - .file_id = *id, - .timestamp_us = now_us - }; - - set_block_index(&record, block_index); - set_timed_out(&record, timed_out); + s64 relative_us; + union log_record record; + size_t record_size; - if (log->rl_size == 0) + spin_lock(&log->rl_lock); + if (log->rl_size == 0) { + spin_unlock(&log->rl_lock); return; + } - spin_lock(&log->rl_writer_lock); - state = READ_ONCE(log->rl_state); - log->rl_ring_buf[state.next_index] = record; - if (++state.next_index == log->rl_size) { - state.next_index = 0; - ++state.current_pass_no; + head = &log->rl_head; + tail = &log->rl_tail; + relative_us = now_us - head->base_record.absolute_ts_us; + + if (memcmp(id, &head->base_record.file_id, sizeof(incfs_uuid_t)) || + relative_us >= 1ll << 32) { + record.full_record = (struct full_record){ + .type = FULL, + .block_index = block_index, + .file_id = *id, + .absolute_ts_us = now_us, + }; + record_size = sizeof(struct full_record); + } else if (block_index != head->base_record.block_index + 1 || + relative_us >= 1 << 30) { + record.same_file_record = (struct same_file_record){ + .type = SAME_FILE, + .block_index = block_index, + .relative_ts_us = relative_us, + }; + record_size = sizeof(struct same_file_record); + } else if (relative_us >= 1 << 14) { + record.same_file_next_block = (struct same_file_next_block){ + .type = SAME_FILE_NEXT_BLOCK, + .relative_ts_us = relative_us, + }; + record_size = sizeof(struct same_file_next_block); + } else { + record.same_file_next_block_short = + (struct same_file_next_block_short){ + .type = SAME_FILE_NEXT_BLOCK_SHORT, + .relative_ts_us = relative_us, + }; + record_size = sizeof(struct same_file_next_block_short); } - WRITE_ONCE(log->rl_state, state); - spin_unlock(&log->rl_writer_lock); + head->base_record.file_id = *id; + head->base_record.block_index = block_index; + head->base_record.absolute_ts_us = now_us; + + /* Advance tail beyond area we are going to overwrite */ + while (tail->current_pass_no < head->current_pass_no && + tail->next_offset < head->next_offset + record_size) + log_read_one_record(log, tail); + + memcpy(((u8 *)log->rl_ring_buf) + head->next_offset, &record, + record_size); + head->next_offset += record_size; + if (head->next_offset > log->rl_size - sizeof(record)) { + head->next_offset = 0; + ++head->current_pass_no; + } + ++head->current_record_no; + + spin_unlock(&log->rl_lock); wake_up_all(&log->ml_notif_wq); } @@ -249,7 +370,7 @@ static int validate_hash_tree(struct file *bf, struct data_file *df, { u8 digest[INCFS_MAX_HASH_SIZE] = {}; struct mtree *tree = NULL; - struct ondisk_signature *sig = NULL; + struct incfs_df_signature *sig = NULL; struct mem_range calc_digest_rng; struct mem_range saved_digest_rng; struct mem_range root_hash_rng; @@ -272,8 +393,8 @@ static int validate_hash_tree(struct file *bf, struct data_file *df, return res; for (lvl = 0; lvl < tree->depth; lvl++) { - loff_t lvl_off = tree->hash_level_suboffset[lvl] + - sig->mtree_offset; + loff_t lvl_off = + tree->hash_level_suboffset[lvl] + sig->hash_offset; loff_t hash_block_off = lvl_off + round_down(hash_block_index * digest_size, INCFS_DATA_FILE_BLOCK_SIZE); @@ -321,72 +442,6 @@ static int validate_hash_tree(struct file *bf, struct data_file *df, return 0; } -static int revalidate_signature(struct file *bf, struct data_file *df) -{ - struct ondisk_signature *sig = df->df_signature; - struct mem_range root_hash = {}; - int result = 0; - u8 *sig_buf = NULL; - u8 *add_data_buf = NULL; - ssize_t read_res; - - /* File has no signature. */ - if (!sig || !df->df_hash_tree || sig->sig_size == 0) - return 0; - - /* Signature has already been validated. */ - if (df->df_signature_validated) - return 0; - - add_data_buf = kzalloc(sig->add_data_size, GFP_NOFS); - if (!add_data_buf) { - result = -ENOMEM; - goto out; - } - - read_res = incfs_kread(bf, add_data_buf, sig->add_data_size, - sig->add_data_offset); - if (read_res < 0) { - result = read_res; - goto out; - } - if (read_res != sig->add_data_size) { - result = -EIO; - goto out; - } - - sig_buf = kzalloc(sig->sig_size, GFP_NOFS); - if (!sig_buf) { - result = -ENOMEM; - goto out; - } - - read_res = incfs_kread(bf, sig_buf, sig->sig_size, sig->sig_offset); - if (read_res < 0) { - result = read_res; - goto out; - } - if (read_res != sig->sig_size) { - result = -EIO; - goto out; - } - - root_hash = range(df->df_hash_tree->root_hash, - df->df_hash_tree->alg->digest_size); - - result = incfs_validate_pkcs7_signature( - range(sig_buf, sig->sig_size), - root_hash, - range(add_data_buf, sig->add_data_size)); - - if (result == 0) - df->df_signature_validated = true; -out: - kfree(sig_buf); - kfree(add_data_buf); - return result; -} - static struct data_file_segment *get_file_segment(struct data_file *df, int block_index) { @@ -401,13 +456,28 @@ static bool is_data_block_present(struct data_file_block *block) (block->db_stored_size != 0); } +static void convert_data_file_block(struct incfs_blockmap_entry *bme, + struct data_file_block *res_block) +{ + u16 flags = le16_to_cpu(bme->me_flags); + + res_block->db_backing_file_data_offset = + le16_to_cpu(bme->me_data_offset_hi); + res_block->db_backing_file_data_offset <<= 32; + res_block->db_backing_file_data_offset |= + le32_to_cpu(bme->me_data_offset_lo); + res_block->db_stored_size = le16_to_cpu(bme->me_data_size); + res_block->db_comp_alg = (flags & INCFS_BLOCK_COMPRESSED_LZ4) ? + COMPRESSION_LZ4 : + COMPRESSION_NONE; +} + static int get_data_file_block(struct data_file *df, int index, struct data_file_block *res_block) { struct incfs_blockmap_entry bme = {}; struct backing_file_context *bfc = NULL; loff_t blockmap_off = 0; - u16 flags = 0; int error = 0; if (!df || !res_block) @@ -416,26 +486,184 @@ static int get_data_file_block(struct data_file *df, int index, blockmap_off = df->df_blockmap_off; bfc = df->df_backing_file_context; - if (index < 0 || index >= df->df_block_count || blockmap_off == 0) + if (index < 0 || blockmap_off == 0) return -EINVAL; error = incfs_read_blockmap_entry(bfc, index, blockmap_off, &bme); if (error) return error; - flags = le16_to_cpu(bme.me_flags); - res_block->db_backing_file_data_offset = - le16_to_cpu(bme.me_data_offset_hi); - res_block->db_backing_file_data_offset <<= 32; - res_block->db_backing_file_data_offset |= - le32_to_cpu(bme.me_data_offset_lo); - res_block->db_stored_size = le16_to_cpu(bme.me_data_size); - res_block->db_comp_alg = (flags & INCFS_BLOCK_COMPRESSED_LZ4) ? - COMPRESSION_LZ4 : - COMPRESSION_NONE; + convert_data_file_block(&bme, res_block); return 0; } +static int check_room_for_one_range(u32 size, u32 size_out) +{ + if (size_out + sizeof(struct incfs_filled_range) > size) + return -ERANGE; + return 0; +} + +static int copy_one_range(struct incfs_filled_range *range, void __user *buffer, + u32 size, u32 *size_out) +{ + int error = check_room_for_one_range(size, *size_out); + if (error) + return error; + + if (copy_to_user(((char __user *)buffer) + *size_out, range, + sizeof(*range))) + return -EFAULT; + + *size_out += sizeof(*range); + return 0; +} + +static int update_file_header_flags(struct data_file *df, u32 bits_to_reset, + u32 bits_to_set) +{ + int result; + u32 new_flags; + struct backing_file_context *bfc; + + if (!df) + return -EFAULT; + bfc = df->df_backing_file_context; + if (!bfc) + return -EFAULT; + + result = mutex_lock_interruptible(&bfc->bc_mutex); + if (result) + return result; + + new_flags = (df->df_header_flags & ~bits_to_reset) | bits_to_set; + if (new_flags != df->df_header_flags) { + df->df_header_flags = new_flags; + result = incfs_write_file_header_flags(bfc, new_flags); + } + + mutex_unlock(&bfc->bc_mutex); + + return result; +} + +#define READ_BLOCKMAP_ENTRIES 512 +int incfs_get_filled_blocks(struct data_file *df, + struct incfs_get_filled_blocks_args *arg) +{ + int error = 0; + bool in_range = false; + struct incfs_filled_range range; + void __user *buffer = u64_to_user_ptr(arg->range_buffer); + u32 size = arg->range_buffer_size; + u32 end_index = + arg->end_index ? arg->end_index : df->df_total_block_count; + u32 *size_out = &arg->range_buffer_size_out; + int i = READ_BLOCKMAP_ENTRIES - 1; + int entries_read = 0; + struct incfs_blockmap_entry *bme; + + *size_out = 0; + if (end_index > df->df_total_block_count) + end_index = df->df_total_block_count; + arg->total_blocks_out = df->df_total_block_count; + arg->data_blocks_out = df->df_data_block_count; + + if (df->df_header_flags & INCFS_FILE_COMPLETE) { + pr_debug("File marked full, fast get_filled_blocks"); + if (arg->start_index > end_index) { + arg->index_out = arg->start_index; + return 0; + } + arg->index_out = arg->start_index; + + error = check_room_for_one_range(size, *size_out); + if (error) + return error; + + range = (struct incfs_filled_range){ + .begin = arg->start_index, + .end = end_index, + }; + + error = copy_one_range(&range, buffer, size, size_out); + if (error) + return error; + arg->index_out = end_index; + return 0; + } + + bme = kzalloc(sizeof(*bme) * READ_BLOCKMAP_ENTRIES, + GFP_NOFS | __GFP_COMP); + if (!bme) + return -ENOMEM; + + for (arg->index_out = arg->start_index; arg->index_out < end_index; + ++arg->index_out) { + struct data_file_block dfb; + + if (++i == READ_BLOCKMAP_ENTRIES) { + entries_read = incfs_read_blockmap_entries( + df->df_backing_file_context, bme, + arg->index_out, READ_BLOCKMAP_ENTRIES, + df->df_blockmap_off); + if (entries_read < 0) { + error = entries_read; + break; + } + + i = 0; + } + + if (i >= entries_read) { + error = -EIO; + break; + } + + convert_data_file_block(bme + i, &dfb); + + if (is_data_block_present(&dfb) == in_range) + continue; + + if (!in_range) { + error = check_room_for_one_range(size, *size_out); + if (error) + break; + in_range = true; + range.begin = arg->index_out; + } else { + range.end = arg->index_out; + error = copy_one_range(&range, buffer, size, size_out); + if (error) { + /* there will be another try out of the loop, + * it will reset the index_out if it fails too + */ + break; + } + in_range = false; + } + } + + if (in_range) { + range.end = arg->index_out; + error = copy_one_range(&range, buffer, size, size_out); + if (error) + arg->index_out = range.begin; + } + + if (!error && in_range && arg->start_index == 0 && + end_index == df->df_total_block_count && + *size_out == sizeof(struct incfs_filled_range)) { + int result = + update_file_header_flags(df, 0, INCFS_FILE_COMPLETE); + /* Log failure only, since it's just a failed optimization */ + pr_debug("Marked file full with result %d", result); + } + + kfree(bme); + return error; +} + static bool is_read_done(struct pending_read *read) { return atomic_read_acquire(&read->done) != 0; @@ -535,7 +763,7 @@ static int wait_for_data_block(struct data_file *df, int block_index, if (!df || !res_block) return -EFAULT; - if (block_index < 0 || block_index >= df->df_block_count) + if (block_index < 0 || block_index >= df->df_data_block_count) return -EINVAL; if (df->df_blockmap_off <= 0) @@ -566,8 +794,7 @@ static int wait_for_data_block(struct data_file *df, int block_index, mi = df->df_mount_info; if (timeout_ms == 0) { - log_block_read(mi, &df->df_id, block_index, - true /*timed out*/); + log_block_read(mi, &df->df_id, block_index); return -ETIME; } @@ -586,8 +813,7 @@ static int wait_for_data_block(struct data_file *df, int block_index, if (wait_res == 0) { /* Wait has timed out */ - log_block_read(mi, &df->df_id, block_index, - true /*timed out*/); + log_block_read(mi, &df->df_id, block_index); return -ETIME; } if (wait_res < 0) { @@ -682,22 +908,15 @@ ssize_t incfs_read_data_file_block(struct mem_range dst, struct data_file *df, result = err; } - if (result > 0) { - int err = revalidate_signature(bf, df); - - if (err < 0) - result = err; - } - if (result >= 0) - log_block_read(mi, &df->df_id, index, false /*timed out*/); + log_block_read(mi, &df->df_id, index); out: return result; } int incfs_process_new_data_block(struct data_file *df, - struct incfs_new_data_block *block, u8 *data) + struct incfs_fill_block *block, u8 *data) { struct mount_info *mi = NULL; struct backing_file_context *bfc = NULL; @@ -712,7 +931,7 @@ int incfs_process_new_data_block(struct data_file *df, bfc = df->df_backing_file_context; mi = df->df_mount_info; - if (block->block_index >= df->df_block_count) + if (block->block_index >= df->df_data_block_count) return -ERANGE; segment = get_file_segment(df, block->block_index); @@ -754,7 +973,7 @@ unlock: int incfs_read_file_signature(struct data_file *df, struct mem_range dst) { struct file *bf = df->df_backing_file_context->bc_file; - struct ondisk_signature *sig; + struct incfs_df_signature *sig; int read_res = 0; if (!dst.data) @@ -779,12 +998,12 @@ int incfs_read_file_signature(struct data_file *df, struct mem_range dst) } int incfs_process_new_hash_block(struct data_file *df, - struct incfs_new_data_block *block, u8 *data) + struct incfs_fill_block *block, u8 *data) { struct backing_file_context *bfc = NULL; struct mount_info *mi = NULL; struct mtree *hash_tree = NULL; - struct ondisk_signature *sig = NULL; + struct incfs_df_signature *sig = NULL; loff_t hash_area_base = 0; loff_t hash_area_size = 0; int error = 0; @@ -803,11 +1022,11 @@ int incfs_process_new_hash_block(struct data_file *df, hash_tree = df->df_hash_tree; sig = df->df_signature; - if (!hash_tree || !sig || sig->mtree_offset == 0) + if (!hash_tree || !sig || sig->hash_offset == 0) return -ENOTSUPP; - hash_area_base = sig->mtree_offset; - hash_area_size = sig->mtree_size; + hash_area_base = sig->hash_offset; + hash_area_size = sig->hash_size; if (hash_area_size < block->block_index * INCFS_DATA_FILE_BLOCK_SIZE + block->data_len) { /* Hash block goes beyond dedicated hash area of this file. */ @@ -818,7 +1037,7 @@ int incfs_process_new_hash_block(struct data_file *df, if (!error) error = incfs_write_hash_block_to_backing_file( bfc, range(data, block->data_len), block->block_index, - hash_area_base); + hash_area_base, df->df_blockmap_off, df->df_size); mutex_unlock(&bfc->bc_mutex); return error; } @@ -834,9 +1053,10 @@ static int process_blockmap_md(struct incfs_blockmap *bm, if (!df) return -EFAULT; - if (df->df_block_count != block_count) + if (df->df_data_block_count > block_count) return -EBADMSG; + df->df_total_block_count = block_count; df->df_blockmap_off = base_off; return error; } @@ -865,58 +1085,69 @@ static int process_file_signature_md(struct incfs_file_signature *sg, { struct data_file *df = handler->context; struct mtree *hash_tree = NULL; - struct ondisk_signature *signature = NULL; int error = 0; - loff_t base_tree_off = le64_to_cpu(sg->sg_hash_tree_offset); - u32 tree_size = le32_to_cpu(sg->sg_hash_tree_size); - loff_t sig_off = le64_to_cpu(sg->sg_sig_offset); - u32 sig_size = le32_to_cpu(sg->sg_sig_size); - loff_t add_data_off = le64_to_cpu(sg->sg_add_data_offset); - u32 add_data_size = le32_to_cpu(sg->sg_add_data_size); + struct incfs_df_signature *signature = + kzalloc(sizeof(*signature), GFP_NOFS); + void *buf = NULL; + ssize_t read; + + if (!df || !df->df_backing_file_context || + !df->df_backing_file_context->bc_file) { + error = -ENOENT; + goto out; + } - if (!df) - return -ENOENT; + signature->hash_offset = le64_to_cpu(sg->sg_hash_tree_offset); + signature->hash_size = le32_to_cpu(sg->sg_hash_tree_size); + signature->sig_offset = le64_to_cpu(sg->sg_sig_offset); + signature->sig_size = le32_to_cpu(sg->sg_sig_size); - signature = kzalloc(sizeof(*signature), GFP_NOFS); - if (!signature) { + buf = kzalloc(signature->sig_size, GFP_NOFS); + if (!buf) { error = -ENOMEM; goto out; } - signature->add_data_offset = add_data_off; - signature->add_data_size = add_data_size; - signature->sig_offset = sig_off; - signature->sig_size = sig_size; - signature->mtree_offset = base_tree_off; - signature->mtree_size = tree_size; + read = incfs_kread(df->df_backing_file_context->bc_file, buf, + signature->sig_size, signature->sig_offset); + if (read < 0) { + error = read; + goto out; + } - hash_tree = incfs_alloc_mtree(sg->sg_hash_alg, df->df_block_count, - range(sg->sg_root_hash, sizeof(sg->sg_root_hash))); + if (read != signature->sig_size) { + error = -EINVAL; + goto out; + } + + hash_tree = incfs_alloc_mtree(range(buf, signature->sig_size), + df->df_data_block_count); if (IS_ERR(hash_tree)) { error = PTR_ERR(hash_tree); hash_tree = NULL; goto out; } - if (hash_tree->hash_tree_area_size != tree_size) { + if (hash_tree->hash_tree_area_size != signature->hash_size) { error = -EINVAL; goto out; } - if (tree_size > 0 && handler->md_record_offset <= base_tree_off) { + if (signature->hash_size > 0 && + handler->md_record_offset <= signature->hash_offset) { error = -EINVAL; goto out; } - if (handler->md_record_offset <= signature->add_data_offset || - handler->md_record_offset <= signature->sig_offset) { + if (handler->md_record_offset <= signature->sig_offset) { error = -EINVAL; goto out; } df->df_hash_tree = hash_tree; + hash_tree = NULL; df->df_signature = signature; + signature = NULL; out: - if (error) { - incfs_free_mtree(hash_tree); - kfree(signature); - } + incfs_free_mtree(hash_tree); + kfree(signature); + kfree(buf); return error; } @@ -972,6 +1203,17 @@ int incfs_scan_metadata_chain(struct data_file *df) result = records_count; } mutex_unlock(&bfc->bc_mutex); + + if (df->df_hash_tree) { + int hash_block_count = get_blocks_count_for_size( + df->df_hash_tree->hash_tree_area_size); + + if (df->df_data_block_count + hash_block_count != + df->df_total_block_count) + result = -EINVAL; + } else if (df->df_data_block_count != df->df_total_block_count) + result = -EINVAL; + out: kfree(handler); return result; @@ -1037,36 +1279,29 @@ struct read_log_state incfs_get_log_state(struct mount_info *mi) struct read_log *log = &mi->mi_log; struct read_log_state result; - spin_lock(&log->rl_writer_lock); - result = READ_ONCE(log->rl_state); - spin_unlock(&log->rl_writer_lock); + spin_lock(&log->rl_lock); + result = log->rl_head; + spin_unlock(&log->rl_lock); return result; } -static u64 calc_record_count(const struct read_log_state *state, int rl_size) -{ - return state->current_pass_no * (u64)rl_size + state->next_index; -} - int incfs_get_uncollected_logs_count(struct mount_info *mi, - struct read_log_state state) + const struct read_log_state *state) { struct read_log *log = &mi->mi_log; - - u64 count = calc_record_count(&log->rl_state, log->rl_size) - - calc_record_count(&state, log->rl_size); - return min_t(int, count, log->rl_size); -} - -static void fill_pending_read_from_log_record( - struct incfs_pending_read_info *dest, const struct read_log_record *src, - struct read_log_state *state, u64 log_size) -{ - dest->file_id = src->file_id; - dest->block_index = get_block_index(src); - dest->serial_number = - state->current_pass_no * log_size + state->next_index; - dest->timestamp_us = src->timestamp_us; + u32 generation; + u64 head_no, tail_no; + + spin_lock(&log->rl_lock); + tail_no = log->rl_tail.current_record_no; + head_no = log->rl_head.current_record_no; + generation = log->rl_head.generation_id; + spin_unlock(&log->rl_lock); + + if (generation != state->generation_id) + return head_no - tail_no; + else + return head_no - max_t(u64, tail_no, state->current_record_no); } int incfs_collect_logged_reads(struct mount_info *mi, @@ -1074,58 +1309,47 @@ int incfs_collect_logged_reads(struct mount_info *mi, struct incfs_pending_read_info *reads, int reads_size) { - struct read_log *log = &mi->mi_log; - struct read_log_state live_state = incfs_get_log_state(mi); - u64 read_count = calc_record_count(reader_state, log->rl_size); - u64 written_count = calc_record_count(&live_state, log->rl_size); int dst_idx; + struct read_log *log = &mi->mi_log; + struct read_log_state *head, *tail; - if (reader_state->next_index >= log->rl_size || - read_count > written_count) - return -ERANGE; + spin_lock(&log->rl_lock); + head = &log->rl_head; + tail = &log->rl_tail; - if (read_count == written_count) - return 0; + if (reader_state->generation_id != head->generation_id) { + pr_debug("read ptr is wrong generation: %u/%u", + reader_state->generation_id, head->generation_id); - if (read_count > written_count) { - /* This reader is somehow ahead of the writer. */ - pr_debug("incfs: Log reader is ahead of writer\n"); - *reader_state = live_state; + *reader_state = (struct read_log_state){ + .generation_id = head->generation_id, + }; } - if (written_count - read_count > log->rl_size) { - /* - * Reading pointer is too far behind, - * start from the record following the write pointer. - */ - pr_debug("incfs: read pointer is behind, moving: %u/%u -> %u/%u / %u\n", - (u32)reader_state->next_index, - (u32)reader_state->current_pass_no, - (u32)live_state.next_index, - (u32)live_state.current_pass_no - 1, (u32)log->rl_size); + if (reader_state->current_record_no < tail->current_record_no) { + pr_debug("read ptr is behind, moving: %u/%u -> %u/%u\n", + (u32)reader_state->next_offset, + (u32)reader_state->current_pass_no, + (u32)tail->next_offset, (u32)tail->current_pass_no); - *reader_state = (struct read_log_state){ - .next_index = live_state.next_index, - .current_pass_no = live_state.current_pass_no - 1, - }; + *reader_state = *tail; } for (dst_idx = 0; dst_idx < reads_size; dst_idx++) { - if (reader_state->next_index == live_state.next_index && - reader_state->current_pass_no == live_state.current_pass_no) + if (reader_state->current_record_no == head->current_record_no) break; - fill_pending_read_from_log_record( - &reads[dst_idx], - &log->rl_ring_buf[reader_state->next_index], - reader_state, log->rl_size); + log_read_one_record(log, reader_state); - reader_state->next_index++; - if (reader_state->next_index == log->rl_size) { - reader_state->next_index = 0; - reader_state->current_pass_no++; - } + reads[dst_idx] = (struct incfs_pending_read_info){ + .file_id = reader_state->base_record.file_id, + .block_index = reader_state->base_record.block_index, + .serial_number = reader_state->current_record_no, + .timestamp_us = reader_state->base_record.absolute_ts_us + }; } + + spin_unlock(&log->rl_lock); return dst_idx; } diff --git a/fs/incfs/data_mgmt.h b/fs/incfs/data_mgmt.h index 6722cef1608c..b7aecdd5bf4a 100644 --- a/fs/incfs/data_mgmt.h +++ b/fs/incfs/data_mgmt.h @@ -20,63 +20,74 @@ #define SEGMENTS_PER_FILE 3 -struct read_log_record { - u32 bitfield; - - u64 timestamp_us; +enum LOG_RECORD_TYPE { + FULL, + SAME_FILE, + SAME_FILE_NEXT_BLOCK, + SAME_FILE_NEXT_BLOCK_SHORT, +}; +struct full_record { + enum LOG_RECORD_TYPE type : 2; /* FULL */ + u32 block_index : 30; incfs_uuid_t file_id; -} __packed; - -#define RLR_BLOCK_INDEX_MASK 0x7fff -#define RLR_TIMED_OUT_MASK 0x8000 - -static inline u32 get_block_index(const struct read_log_record *rlr) -{ - return rlr->bitfield & RLR_BLOCK_INDEX_MASK; -} + u64 absolute_ts_us; +} __packed; /* 28 bytes */ + +struct same_file_record { + enum LOG_RECORD_TYPE type : 2; /* SAME_FILE */ + u32 block_index : 30; + u32 relative_ts_us; /* max 2^32 us ~= 1 hour (1:11:30) */ +} __packed; /* 12 bytes */ + +struct same_file_next_block { + enum LOG_RECORD_TYPE type : 2; /* SAME_FILE_NEXT_BLOCK */ + u32 relative_ts_us : 30; /* max 2^30 us ~= 15 min (17:50) */ +} __packed; /* 4 bytes */ + +struct same_file_next_block_short { + enum LOG_RECORD_TYPE type : 2; /* SAME_FILE_NEXT_BLOCK_SHORT */ + u16 relative_ts_us : 14; /* max 2^14 us ~= 16 ms */ +} __packed; /* 2 bytes */ + +union log_record { + struct full_record full_record; + struct same_file_record same_file_record; + struct same_file_next_block same_file_next_block; + struct same_file_next_block_short same_file_next_block_short; +}; -static inline void set_block_index(struct read_log_record *rlr, - u32 block_index) -{ - rlr->bitfield = (rlr->bitfield & ~RLR_BLOCK_INDEX_MASK) - | (block_index & RLR_BLOCK_INDEX_MASK); -} +struct read_log_state { + /* Log buffer generation id, incremented on configuration changes */ + u32 generation_id; -static inline bool get_timed_out(const struct read_log_record *rlr) -{ - return (rlr->bitfield & RLR_TIMED_OUT_MASK) == RLR_TIMED_OUT_MASK; -} + /* Offset in rl_ring_buf to write into. */ + u32 next_offset; -static inline void set_timed_out(struct read_log_record *rlr, bool timed_out) -{ - if (timed_out) - rlr->bitfield |= RLR_TIMED_OUT_MASK; - else - rlr->bitfield &= ~RLR_TIMED_OUT_MASK; -} + /* Current number of writer passes over rl_ring_buf */ + u32 current_pass_no; -struct read_log_state { - /* Next slot in rl_ring_buf to write to. */ - u32 next_index; + /* Current full_record to diff against */ + struct full_record base_record; - /* Current number of writer pass over rl_ring_buf */ - u32 current_pass_no; + /* Current record number counting from configuration change */ + u64 current_record_no; }; /* A ring buffer to save records about data blocks which were recently read. */ struct read_log { - struct read_log_record *rl_ring_buf; + void *rl_ring_buf; - struct read_log_state rl_state; + int rl_size; - spinlock_t rl_writer_lock; + struct read_log_state rl_head; - int rl_size; + struct read_log_state rl_tail; - /* - * A queue of waiters who want to be notified about reads. - */ + /* A lock to protect the above fields */ + spinlock_t rl_lock; + + /* A queue of waiters who want to be notified about reads */ wait_queue_head_t ml_notif_wq; }; @@ -131,6 +142,12 @@ struct mount_info { /* Temporary buffer for read logger. */ struct read_log mi_log; + + void *log_xattr; + size_t log_xattr_size; + + void *pending_read_xattr; + size_t pending_read_xattr_size; }; struct data_file_block { @@ -203,16 +220,20 @@ struct data_file { /* File size in bytes */ loff_t df_size; - int df_block_count; /* File size in DATA_FILE_BLOCK_SIZE blocks */ + /* File header flags */ + u32 df_header_flags; + + /* File size in DATA_FILE_BLOCK_SIZE blocks */ + int df_data_block_count; + + /* Total number of blocks, data + hash */ + int df_total_block_count; struct file_attr n_attr; struct mtree *df_hash_tree; - struct ondisk_signature *df_signature; - - /* True, if file signature has already been validated. */ - bool df_signature_validated; + struct incfs_df_signature *df_signature; }; struct dir_file { @@ -239,6 +260,9 @@ struct mount_info *incfs_alloc_mount_info(struct super_block *sb, struct mount_options *options, struct path *backing_dir_path); +int incfs_realloc_mount_info(struct mount_info *mi, + struct mount_options *options); + void incfs_free_mount_info(struct mount_info *mi); struct data_file *incfs_open_data_file(struct mount_info *mi, struct file *bf); @@ -253,14 +277,16 @@ ssize_t incfs_read_data_file_block(struct mem_range dst, struct data_file *df, int index, int timeout_ms, struct mem_range tmp); +int incfs_get_filled_blocks(struct data_file *df, + struct incfs_get_filled_blocks_args *arg); + int incfs_read_file_signature(struct data_file *df, struct mem_range dst); int incfs_process_new_data_block(struct data_file *df, - struct incfs_new_data_block *block, u8 *data); + struct incfs_fill_block *block, u8 *data); int incfs_process_new_hash_block(struct data_file *df, - struct incfs_new_data_block *block, u8 *data); - + struct incfs_fill_block *block, u8 *data); bool incfs_fresh_pending_reads_exist(struct mount_info *mi, int last_number); @@ -279,7 +305,7 @@ int incfs_collect_logged_reads(struct mount_info *mi, int reads_size); struct read_log_state incfs_get_log_state(struct mount_info *mi); int incfs_get_uncollected_logs_count(struct mount_info *mi, - struct read_log_state state); + const struct read_log_state *state); static inline struct inode_info *get_incfs_node(struct inode *inode) { @@ -297,7 +323,7 @@ static inline struct inode_info *get_incfs_node(struct inode *inode) static inline struct data_file *get_incfs_data_file(struct file *f) { - struct inode_info *node; + struct inode_info *node = NULL; if (!f) return NULL; diff --git a/fs/incfs/format.c b/fs/incfs/format.c index db71f527cf36..c56e559b6893 100644 --- a/fs/incfs/format.c +++ b/fs/incfs/format.c @@ -13,6 +13,7 @@ #include <linux/kernel.h> #include "format.h" +#include "data_mgmt.h" struct backing_file_context *incfs_alloc_bfc(struct file *backing_file) { @@ -93,7 +94,6 @@ static int append_zeros(struct backing_file_context *bfc, size_t len) { loff_t file_size = 0; loff_t new_last_byte_offset = 0; - int res = 0; if (!bfc) return -EFAULT; @@ -110,28 +110,18 @@ static int append_zeros(struct backing_file_context *bfc, size_t len) */ file_size = incfs_get_end_offset(bfc->bc_file); new_last_byte_offset = file_size + len - 1; - res = vfs_fallocate(bfc->bc_file, 0, new_last_byte_offset, 1); - if (res) - return res; - - res = vfs_fsync_range(bfc->bc_file, file_size, file_size + len, 1); - return res; + return vfs_fallocate(bfc->bc_file, 0, new_last_byte_offset, 1); } static int write_to_bf(struct backing_file_context *bfc, const void *buf, - size_t count, loff_t pos, bool sync) + size_t count, loff_t pos) { - ssize_t res = 0; + ssize_t res = incfs_kwrite(bfc->bc_file, buf, count, pos); - res = incfs_kwrite(bfc->bc_file, buf, count, pos); if (res < 0) return res; if (res != count) return -EIO; - - if (sync) - return vfs_fsync_range(bfc->bc_file, pos, pos + count, 1); - return 0; } @@ -185,7 +175,7 @@ static int append_md_to_backing_file(struct backing_file_context *bfc, /* Write the metadata record to the end of the backing file */ record_offset = file_pos; new_md_offset = cpu_to_le64(record_offset); - result = write_to_bf(bfc, record, record_size, file_pos, true); + result = write_to_bf(bfc, record, record_size, file_pos); if (result) return result; @@ -206,7 +196,7 @@ static int append_md_to_backing_file(struct backing_file_context *bfc, fh_first_md_offset); } result = write_to_bf(bfc, &new_md_offset, sizeof(new_md_offset), - file_pos, true); + file_pos); if (result) return result; @@ -214,12 +204,22 @@ static int append_md_to_backing_file(struct backing_file_context *bfc, return result; } +int incfs_write_file_header_flags(struct backing_file_context *bfc, u32 flags) +{ + if (!bfc) + return -EFAULT; + + return write_to_bf(bfc, &flags, sizeof(flags), + offsetof(struct incfs_file_header, + fh_file_header_flags)); +} + /* * Reserve 0-filled space for the blockmap body, and append * incfs_blockmap metadata record pointing to it. */ int incfs_write_blockmap_to_backing_file(struct backing_file_context *bfc, - u32 block_count, loff_t *map_base_off) + u32 block_count) { struct incfs_blockmap blockmap = {}; int result = 0; @@ -245,12 +245,9 @@ int incfs_write_blockmap_to_backing_file(struct backing_file_context *bfc, /* Write blockmap metadata record pointing to the body written above. */ blockmap.m_base_offset = cpu_to_le64(file_end); result = append_md_to_backing_file(bfc, &blockmap.m_header); - if (result) { + if (result) /* Error, rollback file changes */ truncate_backing_file(bfc, file_end); - } else if (map_base_off) { - *map_base_off = file_end; - } return result; } @@ -283,7 +280,7 @@ int incfs_write_file_attr_to_backing_file(struct backing_file_context *bfc, file_attr.fa_offset = cpu_to_le64(value_offset); file_attr.fa_crc = cpu_to_le32(crc); - result = write_to_bf(bfc, value.data, value.len, value_offset, true); + result = write_to_bf(bfc, value.data, value.len, value_offset); if (result) return result; @@ -299,9 +296,7 @@ int incfs_write_file_attr_to_backing_file(struct backing_file_context *bfc, } int incfs_write_signature_to_backing_file(struct backing_file_context *bfc, - u8 hash_alg, u32 tree_size, - struct mem_range root_hash, struct mem_range add_data, - struct mem_range sig) + struct mem_range sig, u32 tree_size) { struct incfs_file_signature sg = {}; int result = 0; @@ -311,8 +306,6 @@ int incfs_write_signature_to_backing_file(struct backing_file_context *bfc, if (!bfc) return -EFAULT; - if (root_hash.len > sizeof(sg.sg_root_hash)) - return -E2BIG; LOCK_REQUIRED(bfc->bc_mutex); @@ -321,32 +314,19 @@ int incfs_write_signature_to_backing_file(struct backing_file_context *bfc, sg.sg_header.h_md_entry_type = INCFS_MD_SIGNATURE; sg.sg_header.h_record_size = cpu_to_le16(sizeof(sg)); sg.sg_header.h_next_md_offset = cpu_to_le64(0); - sg.sg_hash_alg = hash_alg; if (sig.data != NULL && sig.len > 0) { loff_t pos = incfs_get_end_offset(bfc->bc_file); sg.sg_sig_size = cpu_to_le32(sig.len); sg.sg_sig_offset = cpu_to_le64(pos); - result = write_to_bf(bfc, sig.data, sig.len, pos, false); - if (result) - goto err; - } - - if (add_data.len > 0) { - loff_t pos = incfs_get_end_offset(bfc->bc_file); - - sg.sg_add_data_size = cpu_to_le32(add_data.len); - sg.sg_add_data_offset = cpu_to_le64(pos); - - result = write_to_bf(bfc, add_data.data, - add_data.len, pos, false); + result = write_to_bf(bfc, sig.data, sig.len, pos); if (result) goto err; } tree_area_pos = incfs_get_end_offset(bfc->bc_file); - if (hash_alg && tree_size > 0) { + if (tree_size > 0) { if (tree_size > 5 * INCFS_DATA_FILE_BLOCK_SIZE) { /* * If hash tree is big enough, it makes sense to @@ -369,15 +349,13 @@ int incfs_write_signature_to_backing_file(struct backing_file_context *bfc, sg.sg_hash_tree_size = cpu_to_le32(tree_size); sg.sg_hash_tree_offset = cpu_to_le64(tree_area_pos); } - memcpy(sg.sg_root_hash, root_hash.data, root_hash.len); /* Write a hash tree metadata record pointing to the hash tree above. */ result = append_md_to_backing_file(bfc, &sg.sg_header); err: - if (result) { + if (result) /* Error, rollback file changes */ truncate_backing_file(bfc, rollback_pos); - } return result; } @@ -411,7 +389,7 @@ int incfs_write_fh_to_backing_file(struct backing_file_context *bfc, if (file_pos != 0) return -EEXIST; - return write_to_bf(bfc, &fh, sizeof(fh), file_pos, true); + return write_to_bf(bfc, &fh, sizeof(fh), file_pos); } /* Write a given data block and update file's blockmap to point it. */ @@ -440,7 +418,7 @@ int incfs_write_data_block_to_backing_file(struct backing_file_context *bfc, } /* Write the block data at the end of the backing file. */ - result = write_to_bf(bfc, block.data, block.len, data_offset, false); + result = write_to_bf(bfc, block.data, block.len, data_offset); if (result) return result; @@ -450,18 +428,25 @@ int incfs_write_data_block_to_backing_file(struct backing_file_context *bfc, bm_entry.me_data_size = cpu_to_le16((u16)block.len); bm_entry.me_flags = cpu_to_le16(flags); - result = write_to_bf(bfc, &bm_entry, sizeof(bm_entry), - bm_entry_off, false); - return result; + return write_to_bf(bfc, &bm_entry, sizeof(bm_entry), + bm_entry_off); } int incfs_write_hash_block_to_backing_file(struct backing_file_context *bfc, - struct mem_range block, - int block_index, loff_t hash_area_off) + struct mem_range block, + int block_index, + loff_t hash_area_off, + loff_t bm_base_off, + loff_t file_size) { + struct incfs_blockmap_entry bm_entry = {}; + int result; loff_t data_offset = 0; loff_t file_end = 0; - + loff_t bm_entry_off = + bm_base_off + + sizeof(struct incfs_blockmap_entry) * + (block_index + get_blocks_count_for_size(file_size)); if (!bfc) return -EFAULT; @@ -475,7 +460,16 @@ int incfs_write_hash_block_to_backing_file(struct backing_file_context *bfc, return -EINVAL; } - return write_to_bf(bfc, block.data, block.len, data_offset, false); + result = write_to_bf(bfc, block.data, block.len, data_offset); + if (result) + return result; + + bm_entry.me_data_offset_lo = cpu_to_le32((u32)data_offset); + bm_entry.me_data_offset_hi = cpu_to_le16((u16)(data_offset >> 32)); + bm_entry.me_data_size = cpu_to_le16(INCFS_DATA_FILE_BLOCK_SIZE); + bm_entry.me_flags = cpu_to_le16(INCFS_BLOCK_HASH); + + return write_to_bf(bfc, &bm_entry, sizeof(bm_entry), bm_entry_off); } /* Initialize a new image in a given backing file. */ @@ -505,8 +499,19 @@ int incfs_read_blockmap_entry(struct backing_file_context *bfc, int block_index, loff_t bm_base_off, struct incfs_blockmap_entry *bm_entry) { - return incfs_read_blockmap_entries(bfc, bm_entry, block_index, 1, - bm_base_off); + int error = incfs_read_blockmap_entries(bfc, bm_entry, block_index, 1, + bm_base_off); + + if (error < 0) + return error; + + if (error == 0) + return -EIO; + + if (error != 1) + return -EFAULT; + + return 0; } int incfs_read_blockmap_entries(struct backing_file_context *bfc, @@ -530,15 +535,12 @@ int incfs_read_blockmap_entries(struct backing_file_context *bfc, bm_entry_off); if (result < 0) return result; - if (result < bytes_to_read) - return -EIO; - return 0; + return result / sizeof(*entries); } - int incfs_read_file_header(struct backing_file_context *bfc, loff_t *first_md_off, incfs_uuid_t *uuid, - u64 *file_size) + u64 *file_size, u32 *flags) { ssize_t bytes_read = 0; struct incfs_file_header fh = {}; @@ -572,6 +574,8 @@ int incfs_read_file_header(struct backing_file_context *bfc, *uuid = fh.fh_uuid; if (file_size) *file_size = le64_to_cpu(fh.fh_file_size); + if (flags) + *flags = le32_to_cpu(fh.fh_file_header_flags); return 0; } diff --git a/fs/incfs/format.h b/fs/incfs/format.h index a86881482e19..1a83349bb2eb 100644 --- a/fs/incfs/format.h +++ b/fs/incfs/format.h @@ -121,6 +121,10 @@ enum incfs_metadata_type { INCFS_MD_SIGNATURE = 3 }; +enum incfs_file_header_flags { + INCFS_FILE_COMPLETE = 1 << 0, +}; + /* Header included at the beginning of all metadata records on the disk. */ struct incfs_md_header { __u8 h_md_entry_type; @@ -159,8 +163,8 @@ struct incfs_file_header { /* INCFS_DATA_FILE_BLOCK_SIZE */ __le16 fh_data_block_size; - /* Padding, also reserved for future use. */ - __le32 fh_dummy; + /* File flags, from incfs_file_header_flags */ + __le32 fh_file_header_flags; /* Offset of the first metadata record */ __le64 fh_first_md_offset; @@ -178,6 +182,7 @@ struct incfs_file_header { enum incfs_block_map_entry_flags { INCFS_BLOCK_COMPRESSED_LZ4 = (1 << 0), + INCFS_BLOCK_HASH = (1 << 1), }; /* Block map entry pointing to an actual location of the data block. */ @@ -217,27 +222,27 @@ struct incfs_file_attr { __le32 fa_crc; } __packed; -/* Metadata record for file attribute. Type = INCFS_MD_SIGNATURE */ +/* Metadata record for file signature. Type = INCFS_MD_SIGNATURE */ struct incfs_file_signature { struct incfs_md_header sg_header; - __u8 sg_hash_alg; /* Value from incfs_hash_tree_algorithm */ + __le32 sg_sig_size; /* The size of the signature. */ + + __le64 sg_sig_offset; /* Signature's offset in the backing file */ __le32 sg_hash_tree_size; /* The size of the hash tree. */ __le64 sg_hash_tree_offset; /* Hash tree offset in the backing file */ - - __u8 sg_root_hash[INCFS_MAX_HASH_SIZE]; - - __le32 sg_sig_size; /* The size of the pkcs7 signature. */ - - __le64 sg_sig_offset; /* pkcs7 signature's offset in the backing file */ - - __le32 sg_add_data_size; /* The size of the additional data. */ - - __le64 sg_add_data_offset; /* Additional data's offset */ } __packed; +/* In memory version of above */ +struct incfs_df_signature { + u32 sig_size; + u64 sig_offset; + u32 hash_size; + u64 hash_offset; +}; + /* State of the backing file. */ struct backing_file_context { /* Protects writes to bc_file */ @@ -253,23 +258,6 @@ struct backing_file_context { loff_t bc_last_md_record_offset; }; - -/* Backing file locations of things required for signature validation. */ -struct ondisk_signature { - - loff_t add_data_offset; /* Additional data's offset */ - - loff_t sig_offset; /* pkcs7 signature's offset in the backing file */ - - loff_t mtree_offset; /* Backing file offset of the hash tree. */ - - u32 add_data_size; /* The size of the additional data. */ - - u32 sig_size; /* The size of the pkcs7 signature. */ - - u32 mtree_size; /* The size of the hash tree. */ -}; - struct metadata_handler { loff_t md_record_offset; loff_t md_prev_record_offset; @@ -301,7 +289,7 @@ void incfs_free_bfc(struct backing_file_context *bfc); /* Writing stuff */ int incfs_write_blockmap_to_backing_file(struct backing_file_context *bfc, - u32 block_count, loff_t *map_base_off); + u32 block_count); int incfs_write_fh_to_backing_file(struct backing_file_context *bfc, incfs_uuid_t *uuid, u64 file_size); @@ -312,16 +300,19 @@ int incfs_write_data_block_to_backing_file(struct backing_file_context *bfc, u16 flags); int incfs_write_hash_block_to_backing_file(struct backing_file_context *bfc, - struct mem_range block, - int block_index, loff_t hash_area_off); + struct mem_range block, + int block_index, + loff_t hash_area_off, + loff_t bm_base_off, + loff_t file_size); int incfs_write_file_attr_to_backing_file(struct backing_file_context *bfc, struct mem_range value, struct incfs_file_attr *attr); int incfs_write_signature_to_backing_file(struct backing_file_context *bfc, - u8 hash_alg, u32 tree_size, - struct mem_range root_hash, struct mem_range add_data, - struct mem_range sig); + struct mem_range sig, u32 tree_size); + +int incfs_write_file_header_flags(struct backing_file_context *bfc, u32 flags); int incfs_make_empty_backing_file(struct backing_file_context *bfc, incfs_uuid_t *uuid, u64 file_size); @@ -329,7 +320,7 @@ int incfs_make_empty_backing_file(struct backing_file_context *bfc, /* Reading stuff */ int incfs_read_file_header(struct backing_file_context *bfc, loff_t *first_md_off, incfs_uuid_t *uuid, - u64 *file_size); + u64 *file_size, u32 *flags); int incfs_read_blockmap_entry(struct backing_file_context *bfc, int block_index, loff_t bm_base_off, diff --git a/fs/incfs/integrity.c b/fs/incfs/integrity.c index feb212c38945..d049988ef037 100644 --- a/fs/incfs/integrity.c +++ b/fs/incfs/integrity.c @@ -10,70 +10,6 @@ #include "integrity.h" -int incfs_validate_pkcs7_signature(struct mem_range pkcs7_blob, - struct mem_range root_hash, struct mem_range add_data) -{ - struct pkcs7_message *pkcs7 = NULL; - const void *data = NULL; - size_t data_len = 0; - const char *p; - int err; - - pkcs7 = pkcs7_parse_message(pkcs7_blob.data, pkcs7_blob.len); - if (IS_ERR(pkcs7)) { - pr_debug("PKCS#7 parsing error. ptr=%p size=%ld err=%ld\n", - pkcs7_blob.data, pkcs7_blob.len, -PTR_ERR(pkcs7)); - return PTR_ERR(pkcs7); - } - - err = pkcs7_get_content_data(pkcs7, &data, &data_len, NULL); - if (err || data_len == 0 || data == NULL) { - pr_debug("PKCS#7 message does not contain data\n"); - err = -EBADMSG; - goto out; - } - - if (root_hash.len == 0) { - pr_debug("Root hash is empty.\n"); - err = -EBADMSG; - goto out; - } - - if (data_len != root_hash.len + add_data.len) { - pr_debug("PKCS#7 data size doesn't match arguments.\n"); - err = -EKEYREJECTED; - goto out; - } - - p = data; - if (memcmp(p, root_hash.data, root_hash.len) != 0) { - pr_debug("Root hash mismatch.\n"); - err = -EKEYREJECTED; - goto out; - } - p += root_hash.len; - if (memcmp(p, add_data.data, add_data.len) != 0) { - pr_debug("Additional data mismatch.\n"); - err = -EKEYREJECTED; - goto out; - } - - err = pkcs7_verify(pkcs7, VERIFYING_UNSPECIFIED_SIGNATURE); - if (err) - pr_debug("PKCS#7 signature verification error: %d\n", -err); - - /* - * RSA signature verification sometimes returns unexpected error codes - * when signature doesn't match. - */ - if (err == -ERANGE || err == -EINVAL) - err = -EBADMSG; - -out: - pkcs7_free_message(pkcs7); - return err; -} - struct incfs_hash_alg *incfs_get_hash_alg(enum incfs_hash_tree_algorithm id) { static struct incfs_hash_alg sha256 = { @@ -113,11 +49,90 @@ struct incfs_hash_alg *incfs_get_hash_alg(enum incfs_hash_tree_algorithm id) return result; } +struct signature_info { + u32 version; + enum incfs_hash_tree_algorithm hash_algorithm; + u8 log2_blocksize; + struct mem_range salt; + struct mem_range root_hash; +}; + +static bool read_u32(u8 **p, u8 *top, u32 *result) +{ + if (*p + sizeof(u32) > top) + return false; + + *result = le32_to_cpu(*(__le32 *)*p); + *p += sizeof(u32); + return true; +} + +static bool read_u8(u8 **p, u8 *top, u8 *result) +{ + if (*p + sizeof(u8) > top) + return false; + + *result = *(u8 *)*p; + *p += sizeof(u8); + return true; +} + +static bool read_mem_range(u8 **p, u8 *top, struct mem_range *range) +{ + u32 len; + + if (!read_u32(p, top, &len) || *p + len > top) + return false; + + range->len = len; + range->data = *p; + *p += len; + return true; +} -struct mtree *incfs_alloc_mtree(enum incfs_hash_tree_algorithm id, - int data_block_count, - struct mem_range root_hash) +static int incfs_parse_signature(struct mem_range signature, + struct signature_info *si) { + u8 *p = signature.data; + u8 *top = signature.data + signature.len; + u32 hash_section_size; + + if (signature.len > INCFS_MAX_SIGNATURE_SIZE) + return -EINVAL; + + if (!read_u32(&p, top, &si->version) || + si->version != INCFS_SIGNATURE_VERSION) + return -EINVAL; + + if (!read_u32(&p, top, &hash_section_size) || + p + hash_section_size > top) + return -EINVAL; + top = p + hash_section_size; + + if (!read_u32(&p, top, &si->hash_algorithm) || + si->hash_algorithm != INCFS_HASH_TREE_SHA256) + return -EINVAL; + + if (!read_u8(&p, top, &si->log2_blocksize) || si->log2_blocksize != 12) + return -EINVAL; + + if (!read_mem_range(&p, top, &si->salt)) + return -EINVAL; + + if (!read_mem_range(&p, top, &si->root_hash)) + return -EINVAL; + + if (p != top) + return -EINVAL; + + return 0; +} + +struct mtree *incfs_alloc_mtree(struct mem_range signature, + int data_block_count) +{ + int error; + struct signature_info si; struct mtree *result = NULL; struct incfs_hash_alg *hash_alg = NULL; int hash_per_block; @@ -129,11 +144,15 @@ struct mtree *incfs_alloc_mtree(enum incfs_hash_tree_algorithm id, if (data_block_count <= 0) return ERR_PTR(-EINVAL); - hash_alg = incfs_get_hash_alg(id); + error = incfs_parse_signature(signature, &si); + if (error) + return ERR_PTR(error); + + hash_alg = incfs_get_hash_alg(si.hash_algorithm); if (IS_ERR(hash_alg)) return ERR_PTR(PTR_ERR(hash_alg)); - if (root_hash.len < hash_alg->digest_size) + if (si.root_hash.len < hash_alg->digest_size) return ERR_PTR(-EINVAL); result = kzalloc(sizeof(*result), GFP_NOFS); @@ -173,7 +192,7 @@ struct mtree *incfs_alloc_mtree(enum incfs_hash_tree_algorithm id, } /* Root hash is stored separately from the rest of the tree. */ - memcpy(result->root_hash, root_hash.data, hash_alg->digest_size); + memcpy(result->root_hash, si.root_hash.data, hash_alg->digest_size); return result; err: @@ -198,16 +217,20 @@ int incfs_calc_digest(struct incfs_hash_alg *alg, struct mem_range data, return -EINVAL; desc->tfm = alg->shash; - return crypto_shash_digest(desc, data.data, data.len, digest.data); -} -void incfs_free_signature_info(struct signature_info *si) -{ - if (!si) - return; - kfree(si->root_hash.data); - kfree(si->additional_data.data); - kfree(si->signature.data); - kfree(si); + if (data.len < INCFS_DATA_FILE_BLOCK_SIZE) { + int err; + void *buf = kzalloc(INCFS_DATA_FILE_BLOCK_SIZE, GFP_NOFS); + + if (!buf) + return -ENOMEM; + + memcpy(buf, data.data, data.len); + err = crypto_shash_digest(desc, buf, INCFS_DATA_FILE_BLOCK_SIZE, + digest.data); + kfree(buf); + return err; + } + return crypto_shash_digest(desc, data.data, data.len, digest.data); } diff --git a/fs/incfs/integrity.h b/fs/incfs/integrity.h index da1c38486b2f..cf79b64da736 100644 --- a/fs/incfs/integrity.h +++ b/fs/incfs/integrity.h @@ -38,21 +38,10 @@ struct mtree { int depth; }; -struct signature_info { - struct mem_range root_hash; - - struct mem_range additional_data; - - struct mem_range signature; - - enum incfs_hash_tree_algorithm hash_alg; -}; - struct incfs_hash_alg *incfs_get_hash_alg(enum incfs_hash_tree_algorithm id); -struct mtree *incfs_alloc_mtree(enum incfs_hash_tree_algorithm id, - int data_block_count, - struct mem_range root_hash); +struct mtree *incfs_alloc_mtree(struct mem_range signature, + int data_block_count); void incfs_free_mtree(struct mtree *tree); @@ -64,9 +53,4 @@ size_t incfs_get_mtree_hash_count(enum incfs_hash_tree_algorithm alg, int incfs_calc_digest(struct incfs_hash_alg *alg, struct mem_range data, struct mem_range digest); -int incfs_validate_pkcs7_signature(struct mem_range pkcs7_blob, - struct mem_range root_hash, struct mem_range add_data); - -void incfs_free_signature_info(struct signature_info *si); - #endif /* _INCFS_INTEGRITY_H */ diff --git a/fs/incfs/vfs.c b/fs/incfs/vfs.c index aebd2b02bd83..ec028fc96303 100644 --- a/fs/incfs/vfs.c +++ b/fs/incfs/vfs.c @@ -52,8 +52,6 @@ static int dir_rename(struct inode *old_dir, struct dentry *old_dentry, static int file_open(struct inode *inode, struct file *file); static int file_release(struct inode *inode, struct file *file); -static ssize_t file_write(struct file *f, const char __user *buf, - size_t size, loff_t *offset); static int read_single_page(struct file *f, struct page *page); static long dispatch_ioctl(struct file *f, unsigned int req, unsigned long arg); @@ -75,6 +73,8 @@ static void evict_inode(struct inode *inode); static ssize_t incfs_getxattr(struct dentry *d, const char *name, void *value, size_t size); +static ssize_t incfs_setxattr(struct dentry *d, const char *name, + const void *value, size_t size, int flags); static ssize_t incfs_listxattr(struct dentry *d, char *list, size_t size); static int show_options(struct seq_file *, struct dentry *); @@ -127,7 +127,6 @@ static const struct address_space_operations incfs_address_space_ops = { static const struct file_operations incfs_file_ops = { .open = file_open, .release = file_release, - .write = file_write, .read_iter = generic_file_read_iter, .mmap = generic_file_mmap, .splice_read = generic_file_splice_read, @@ -136,6 +135,11 @@ static const struct file_operations incfs_file_ops = { .compat_ioctl = dispatch_ioctl }; +enum FILL_PERMISSION { + CANT_FILL = 0, + CAN_FILL = 1, +}; + static const struct file_operations incfs_pending_read_file_ops = { .read = pending_reads_read, .poll = pending_reads_poll, @@ -169,9 +173,18 @@ static int incfs_handler_getxattr(const struct xattr_handler *xh, return incfs_getxattr(d, name, buffer, size); } +static int incfs_handler_setxattr(const struct xattr_handler *xh, + struct dentry *d, struct inode *inode, + const char *name, const void *buffer, + size_t size, int flags) +{ + return incfs_setxattr(d, name, buffer, size, flags); +} + static const struct xattr_handler incfs_xattr_handler = { .prefix = "", /* AKA all attributes */ .get = incfs_handler_getxattr, + .set = incfs_handler_setxattr, }; static const struct xattr_handler *incfs_xattr_ops[] = { @@ -336,8 +349,8 @@ static int inode_test(struct inode *inode, void *opaque) return (node->n_backing_inode == backing_inode) && inode->i_ino == search->ino; - } - return 1; + } else + return inode->i_ino == search->ino; } static int inode_set(struct inode *inode, void *opaque) @@ -454,9 +467,6 @@ static ssize_t pending_reads_read(struct file *f, char __user *buf, size_t len, ssize_t result = 0; int i = 0; - if (!access_ok(VERIFY_WRITE, buf, len)) - return -EFAULT; - if (!incfs_fresh_pending_reads_exist(mi, last_known_read_sn)) return 0; @@ -574,22 +584,27 @@ static ssize_t log_read(struct file *f, char __user *buf, size_t len, { struct log_file_state *log_state = f->private_data; struct mount_info *mi = get_mount_info(file_superblock(f)); - struct incfs_pending_read_info *reads_buf = - (struct incfs_pending_read_info *)__get_free_page(GFP_NOFS); - size_t reads_to_collect = len / sizeof(*reads_buf); - size_t reads_per_page = PAGE_SIZE / sizeof(*reads_buf); int total_reads_collected = 0; + int rl_size; ssize_t result = 0; + struct incfs_pending_read_info *reads_buf; + ssize_t reads_to_collect = len / sizeof(*reads_buf); + ssize_t reads_per_page = PAGE_SIZE / sizeof(*reads_buf); + + rl_size = READ_ONCE(mi->mi_log.rl_size); + if (rl_size == 0) + return 0; + reads_buf = (struct incfs_pending_read_info *)__get_free_page(GFP_NOFS); if (!reads_buf) return -ENOMEM; - reads_to_collect = min_t(size_t, mi->mi_log.rl_size, reads_to_collect); + reads_to_collect = min_t(ssize_t, rl_size, reads_to_collect); while (reads_to_collect > 0) { struct read_log_state next_state = READ_ONCE(log_state->state); int reads_collected = incfs_collect_logged_reads( mi, &next_state, reads_buf, - min_t(size_t, reads_to_collect, reads_per_page)); + min_t(ssize_t, reads_to_collect, reads_per_page)); if (reads_collected <= 0) { result = total_reads_collected ? total_reads_collected * @@ -628,7 +643,7 @@ static __poll_t log_poll(struct file *file, poll_table *wait) __poll_t ret = 0; poll_wait(file, &mi->mi_log.ml_notif_wq, wait); - count = incfs_get_uncollected_logs_count(mi, log_state->state); + count = incfs_get_uncollected_logs_count(mi, &log_state->state); if (count >= mi->mi_options.read_log_wakeup_count) ret = EPOLLIN | EPOLLRDNORM; @@ -789,9 +804,6 @@ static int read_single_page(struct file *f, struct page *page) size = df->df_size; timeout_ms = df->df_mount_info->mi_options.read_timeout_ms; - pr_debug("incfs: %s %s %lld\n", __func__, - f->f_path.dentry->d_name.name, offset); - if (offset < size) { struct mem_range tmp = { .len = 2 * INCFS_DATA_FILE_BLOCK_SIZE @@ -838,107 +850,39 @@ static char *file_id_to_str(incfs_uuid_t id) return result; } -static struct signature_info *incfs_copy_signature_info_from_user( - struct incfs_file_signature_info __user *original) +static struct mem_range incfs_copy_signature_info_from_user(u8 __user *original, + u64 size) { - struct incfs_file_signature_info usr_si; - struct signature_info *result; - int error; + u8 *result; if (!original) - return NULL; + return range(NULL, 0); - if (!access_ok(VERIFY_READ, original, sizeof(usr_si))) - return ERR_PTR(-EFAULT); + if (size > INCFS_MAX_SIGNATURE_SIZE) + return range(ERR_PTR(-EFAULT), 0); - if (copy_from_user(&usr_si, original, sizeof(usr_si)) > 0) - return ERR_PTR(-EFAULT); - - result = kzalloc(sizeof(*result), GFP_NOFS); + result = kzalloc(size, GFP_NOFS | __GFP_COMP); if (!result) - return ERR_PTR(-ENOMEM); - - result->hash_alg = usr_si.hash_tree_alg; - - if (result->hash_alg) { - void *p = kzalloc(INCFS_MAX_HASH_SIZE, GFP_NOFS); - - if (!p) { - error = -ENOMEM; - goto err; - } + return range(ERR_PTR(-ENOMEM), 0); - /* TODO this sets the root_hash length to MAX_HASH_SIZE not - * the actual size. Fix, then set INCFS_MAX_HASH_SIZE back - * to 64 - */ - result->root_hash = range(p, INCFS_MAX_HASH_SIZE); - if (copy_from_user(p, u64_to_user_ptr(usr_si.root_hash), - result->root_hash.len) > 0) { - error = -EFAULT; - goto err; - } - } - - if (usr_si.additional_data_size > INCFS_MAX_FILE_ATTR_SIZE) { - error = -E2BIG; - goto err; + if (copy_from_user(result, original, size)) { + kfree(result); + return range(ERR_PTR(-EFAULT), 0); } - if (usr_si.additional_data && usr_si.additional_data_size) { - void *p = kzalloc(usr_si.additional_data_size, GFP_NOFS); - - if (!p) { - error = -ENOMEM; - goto err; - } - result->additional_data = range(p, - usr_si.additional_data_size); - if (copy_from_user(p, u64_to_user_ptr(usr_si.additional_data), - result->additional_data.len) > 0) { - error = -EFAULT; - goto err; - } - } - - if (usr_si.signature_size > INCFS_MAX_SIGNATURE_SIZE) { - error = -E2BIG; - goto err; - } - - if (usr_si.signature && usr_si.signature_size) { - void *p = kzalloc(usr_si.signature_size, GFP_NOFS); - - if (!p) { - error = -ENOMEM; - goto err; - } - result->signature = range(p, usr_si.signature_size); - if (copy_from_user(p, u64_to_user_ptr(usr_si.signature), - result->signature.len) > 0) { - error = -EFAULT; - goto err; - } - } - - return result; - -err: - incfs_free_signature_info(result); - return ERR_PTR(-error); + return range(result, size); } static int init_new_file(struct mount_info *mi, struct dentry *dentry, - incfs_uuid_t *uuid, u64 size, struct mem_range attr, - struct incfs_file_signature_info __user *fsi) + incfs_uuid_t *uuid, u64 size, struct mem_range attr, + u8 __user *user_signature_info, u64 signature_size) { struct path path = {}; struct file *new_file; int error = 0; struct backing_file_context *bfc = NULL; u32 block_count; - struct mem_range mem_range = {NULL}; - struct signature_info *si = NULL; + struct mem_range raw_signature = { NULL }; struct mtree *hash_tree = NULL; if (!mi || !dentry || !uuid) @@ -957,6 +901,7 @@ static int init_new_file(struct mount_info *mi, struct dentry *dentry, } bfc = incfs_alloc_bfc(new_file); + fput(new_file); if (IS_ERR(bfc)) { error = PTR_ERR(bfc); bfc = NULL; @@ -968,19 +913,6 @@ static int init_new_file(struct mount_info *mi, struct dentry *dentry, if (error) goto out; - block_count = (u32)get_blocks_count_for_size(size); - error = incfs_write_blockmap_to_backing_file(bfc, block_count, NULL); - if (error) - goto out; - - /* This fill has data, reserve space for the block map. */ - if (block_count > 0) { - error = incfs_write_blockmap_to_backing_file( - bfc, block_count, NULL); - if (error) - goto out; - } - if (attr.data && attr.len) { error = incfs_write_file_attr_to_backing_file(bfc, attr, NULL); @@ -988,54 +920,46 @@ static int init_new_file(struct mount_info *mi, struct dentry *dentry, goto out; } - if (fsi) { - si = incfs_copy_signature_info_from_user(fsi); + block_count = (u32)get_blocks_count_for_size(size); + + if (user_signature_info) { + raw_signature = incfs_copy_signature_info_from_user( + user_signature_info, signature_size); - if (IS_ERR(si)) { - error = PTR_ERR(si); - si = NULL; + if (IS_ERR(raw_signature.data)) { + error = PTR_ERR(raw_signature.data); + raw_signature.data = NULL; goto out; } - if (si->hash_alg) { - hash_tree = incfs_alloc_mtree(si->hash_alg, block_count, - si->root_hash); - if (IS_ERR(hash_tree)) { - error = PTR_ERR(hash_tree); - hash_tree = NULL; - goto out; - } - - /* TODO This code seems wrong when len is zero - we - * should error out?? - */ - if (si->signature.len > 0) - error = incfs_validate_pkcs7_signature( - si->signature, - si->root_hash, - si->additional_data); - if (error) - goto out; + hash_tree = incfs_alloc_mtree(raw_signature, block_count); + if (IS_ERR(hash_tree)) { + error = PTR_ERR(hash_tree); + hash_tree = NULL; + goto out; + } - error = incfs_write_signature_to_backing_file(bfc, - si->hash_alg, - hash_tree->hash_tree_area_size, - si->root_hash, si->additional_data, - si->signature); + error = incfs_write_signature_to_backing_file( + bfc, raw_signature, hash_tree->hash_tree_area_size); + if (error) + goto out; - if (error) - goto out; - } + block_count += get_blocks_count_for_size( + hash_tree->hash_tree_area_size); } + if (block_count) + error = incfs_write_blockmap_to_backing_file(bfc, block_count); + + if (error) + goto out; out: if (bfc) { mutex_unlock(&bfc->bc_mutex); incfs_free_bfc(bfc); } incfs_free_mtree(hash_tree); - incfs_free_signature_info(si); - kfree(mem_range.data); + kfree(raw_signature.data); if (error) pr_debug("incfs: %s error: %d\n", __func__, error); @@ -1180,10 +1104,7 @@ static long ioctl_create_file(struct mount_info *mi, error = -EFAULT; goto out; } - if (!access_ok(VERIFY_READ, usr_args, sizeof(args))) { - error = -EFAULT; - goto out; - } + if (copy_from_user(&args, usr_args, sizeof(args)) > 0) { error = -EFAULT; goto out; @@ -1296,7 +1217,7 @@ static long ioctl_create_file(struct mount_info *mi, goto delete_index_file; } - /* Save the file's attrubute as an xattr */ + /* Save the file's attribute as an xattr */ if (args.file_attr_len && args.file_attr) { if (args.file_attr_len > INCFS_MAX_FILE_ATTR_SIZE) { error = -E2BIG; @@ -1309,12 +1230,6 @@ static long ioctl_create_file(struct mount_info *mi, goto delete_index_file; } - if (!access_ok(VERIFY_READ, u64_to_user_ptr(args.file_attr), - args.file_attr_len)) { - error = -EFAULT; - goto delete_index_file; - } - if (copy_from_user(attr_value, u64_to_user_ptr(args.file_attr), args.file_attr_len) > 0) { @@ -1333,9 +1248,9 @@ static long ioctl_create_file(struct mount_info *mi, /* Initializing a newly created file. */ error = init_new_file(mi, index_file_dentry, &args.file_id, args.size, - range(attr_value, args.file_attr_len), - (struct incfs_file_signature_info __user *) - args.signature_info); + range(attr_value, args.file_attr_len), + (u8 __user *)args.signature_info, + args.signature_size); if (error) goto delete_index_file; @@ -1363,6 +1278,123 @@ out: return error; } +static long ioctl_fill_blocks(struct file *f, void __user *arg) +{ + struct incfs_fill_blocks __user *usr_fill_blocks = arg; + struct incfs_fill_blocks fill_blocks; + struct incfs_fill_block __user *usr_fill_block_array; + struct data_file *df = get_incfs_data_file(f); + const ssize_t data_buf_size = 2 * INCFS_DATA_FILE_BLOCK_SIZE; + u8 *data_buf = NULL; + ssize_t error = 0; + int i = 0; + + if (!df) + return -EBADF; + + if ((uintptr_t)f->private_data != CAN_FILL) + return -EPERM; + + if (copy_from_user(&fill_blocks, usr_fill_blocks, sizeof(fill_blocks))) + return -EFAULT; + + usr_fill_block_array = u64_to_user_ptr(fill_blocks.fill_blocks); + data_buf = (u8 *)__get_free_pages(GFP_NOFS | __GFP_COMP, + get_order(data_buf_size)); + if (!data_buf) + return -ENOMEM; + + for (i = 0; i < fill_blocks.count; i++) { + struct incfs_fill_block fill_block = {}; + + if (copy_from_user(&fill_block, &usr_fill_block_array[i], + sizeof(fill_block)) > 0) { + error = -EFAULT; + break; + } + + if (fill_block.data_len > data_buf_size) { + error = -E2BIG; + break; + } + + if (copy_from_user(data_buf, u64_to_user_ptr(fill_block.data), + fill_block.data_len) > 0) { + error = -EFAULT; + break; + } + fill_block.data = 0; /* To make sure nobody uses it. */ + if (fill_block.flags & INCFS_BLOCK_FLAGS_HASH) { + error = incfs_process_new_hash_block(df, &fill_block, + data_buf); + } else { + error = incfs_process_new_data_block(df, &fill_block, + data_buf); + } + if (error) + break; + } + + if (data_buf) + free_pages((unsigned long)data_buf, get_order(data_buf_size)); + + /* + * Only report the error if no records were processed, otherwise + * just return how many were processed successfully. + */ + if (i == 0) + return error; + + return i; +} + +static long ioctl_permit_fill(struct file *f, void __user *arg) +{ + struct incfs_permit_fill __user *usr_permit_fill = arg; + struct incfs_permit_fill permit_fill; + long error = 0; + struct file *file = NULL; + + if (f->f_op != &incfs_pending_read_file_ops) + return -EPERM; + + if (copy_from_user(&permit_fill, usr_permit_fill, sizeof(permit_fill))) + return -EFAULT; + + file = fget(permit_fill.file_descriptor); + if (IS_ERR(file)) + return PTR_ERR(file); + + if (file->f_op != &incfs_file_ops) { + error = -EPERM; + goto out; + } + + if (file->f_inode->i_sb != f->f_inode->i_sb) { + error = -EPERM; + goto out; + } + + switch ((uintptr_t)file->private_data) { + case CANT_FILL: + file->private_data = (void *)CAN_FILL; + break; + + case CAN_FILL: + pr_debug("CAN_FILL already set"); + break; + + default: + pr_warn("Invalid file private data"); + error = -EFAULT; + goto out; + } + +out: + fput(file); + return error; +} + static long ioctl_read_file_signature(struct file *f, void __user *arg) { struct incfs_get_file_sig_args __user *args_usr_ptr = arg; @@ -1376,20 +1408,14 @@ static long ioctl_read_file_signature(struct file *f, void __user *arg) if (!df) return -EINVAL; - if (!access_ok(VERIFY_READ, args_usr_ptr, sizeof(args))) - return -EFAULT; if (copy_from_user(&args, args_usr_ptr, sizeof(args)) > 0) return -EINVAL; - if (!access_ok(VERIFY_WRITE, u64_to_user_ptr(args.file_signature), - args.file_signature_buf_size)) - return -EFAULT; - sig_buf_size = args.file_signature_buf_size; if (sig_buf_size > INCFS_MAX_SIGNATURE_SIZE) return -E2BIG; - sig_buffer = kzalloc(sig_buf_size, GFP_NOFS); + sig_buffer = kzalloc(sig_buf_size, GFP_NOFS | __GFP_COMP); if (!sig_buffer) return -ENOMEM; @@ -1417,6 +1443,30 @@ out: return error; } +static long ioctl_get_filled_blocks(struct file *f, void __user *arg) +{ + struct incfs_get_filled_blocks_args __user *args_usr_ptr = arg; + struct incfs_get_filled_blocks_args args = {}; + struct data_file *df = get_incfs_data_file(f); + int error; + + if (!df) + return -EINVAL; + + if ((uintptr_t)f->private_data != CAN_FILL) + return -EPERM; + + if (copy_from_user(&args, args_usr_ptr, sizeof(args)) > 0) + return -EINVAL; + + error = incfs_get_filled_blocks(df, &args); + + if (copy_to_user(args_usr_ptr, &args, sizeof(args))) + return -EFAULT; + + return error; +} + static long dispatch_ioctl(struct file *f, unsigned int req, unsigned long arg) { struct mount_info *mi = get_mount_info(file_superblock(f)); @@ -1424,8 +1474,14 @@ static long dispatch_ioctl(struct file *f, unsigned int req, unsigned long arg) switch (req) { case INCFS_IOC_CREATE_FILE: return ioctl_create_file(mi, (void __user *)arg); + case INCFS_IOC_FILL_BLOCKS: + return ioctl_fill_blocks(f, (void __user *)arg); + case INCFS_IOC_PERMIT_FILL: + return ioctl_permit_fill(f, (void __user *)arg); case INCFS_IOC_READ_FILE_SIGNATURE: return ioctl_read_file_signature(f, (void __user *)arg); + case INCFS_IOC_GET_FILLED_BLOCKS: + return ioctl_get_filled_blocks(f, (void __user *)arg); default: return -EINVAL; } @@ -1632,6 +1688,7 @@ static int final_file_delete(struct mount_info *mi, if (d_really_is_positive(index_file_dentry)) error = incfs_unlink(index_file_dentry); out: + dput(index_file_dentry); if (error) pr_debug("incfs: delete_file_from_index err:%d\n", error); return error; @@ -1854,9 +1911,10 @@ static int file_open(struct inode *inode, struct file *file) goto out; } - if (S_ISREG(inode->i_mode)) + if (S_ISREG(inode->i_mode)) { err = make_inode_ready_for_data_ops(mi, inode, backing_file); - else if (S_ISDIR(inode->i_mode)) { + file->private_data = (void *)CANT_FILL; + } else if (S_ISDIR(inode->i_mode)) { struct dir_file *dir = NULL; dir = incfs_open_dir_file(mi, backing_file); @@ -1891,77 +1949,6 @@ static int file_release(struct inode *inode, struct file *file) return 0; } -static ssize_t file_write(struct file *f, const char __user *buf, - size_t size, loff_t *offset) -{ - struct data_file *df = get_incfs_data_file(f); - const ssize_t data_buf_size = 2 * INCFS_DATA_FILE_BLOCK_SIZE; - size_t block_count = size / sizeof(struct incfs_new_data_block); - struct incfs_new_data_block __user *usr_blocks = - (struct incfs_new_data_block __user *)buf; - u8 *data_buf = NULL; - ssize_t error = 0; - int i = 0; - - if (!df) - return -EBADF; - - if (!access_ok(VERIFY_READ, usr_blocks, size)) - return -EFAULT; - - data_buf = (u8 *)__get_free_pages(GFP_NOFS, get_order(data_buf_size)); - if (!data_buf) - return -ENOMEM; - - for (i = 0; i < block_count; i++) { - struct incfs_new_data_block block = {}; - - if (copy_from_user(&block, &usr_blocks[i], sizeof(block)) > 0) { - error = -EFAULT; - break; - } - - if (block.data_len > data_buf_size) { - error = -E2BIG; - break; - } - if (!access_ok(VERIFY_READ, u64_to_user_ptr(block.data), - block.data_len)) { - error = -EFAULT; - break; - } - if (copy_from_user(data_buf, u64_to_user_ptr(block.data), - block.data_len) > 0) { - error = -EFAULT; - break; - } - block.data = 0; /* To make sure nobody uses it. */ - if (block.flags & INCFS_BLOCK_FLAGS_HASH) { - error = incfs_process_new_hash_block(df, &block, - data_buf); - } else { - error = incfs_process_new_data_block(df, &block, - data_buf); - } - if (error) - break; - } - - if (data_buf) - free_pages((unsigned long)data_buf, get_order(data_buf_size)); - *offset = 0; - - /* - * Only report the error if no records were processed, otherwise - * just return how many were processed successfully. - */ - if (i == 0) - return error; - - return i * sizeof(struct incfs_new_data_block); -} - - static int dentry_revalidate(struct dentry *d, unsigned int flags) { struct path backing_path = {}; @@ -2004,6 +1991,7 @@ static void dentry_release(struct dentry *d) if (di) path_put(&di->backing_path); + kfree(d->d_fsdata); d->d_fsdata = NULL; } @@ -2048,11 +2036,74 @@ static ssize_t incfs_getxattr(struct dentry *d, const char *name, void *value, size_t size) { struct dentry_info *di = get_incfs_dentry(d); + struct mount_info *mi = get_mount_info(d->d_sb); + char *stored_value; + size_t stored_size; - if (!di || !di->backing_path.dentry) + if (di && di->backing_path.dentry) + return vfs_getxattr(di->backing_path.dentry, name, value, size); + + if (strcmp(name, "security.selinux")) + return -ENODATA; + + if (!strcmp(d->d_iname, INCFS_PENDING_READS_FILENAME)) { + stored_value = mi->pending_read_xattr; + stored_size = mi->pending_read_xattr_size; + } else if (!strcmp(d->d_iname, INCFS_LOG_FILENAME)) { + stored_value = mi->log_xattr; + stored_size = mi->log_xattr_size; + } else { + return -ENODATA; + } + + if (!stored_value) + return -ENODATA; + + if (stored_size > size) + return -E2BIG; + + memcpy(value, stored_value, stored_size); + return stored_size; + +} + + +static ssize_t incfs_setxattr(struct dentry *d, const char *name, + const void *value, size_t size, int flags) +{ + struct dentry_info *di = get_incfs_dentry(d); + struct mount_info *mi = get_mount_info(d->d_sb); + void **stored_value; + size_t *stored_size; + + if (di && di->backing_path.dentry) + return vfs_setxattr(di->backing_path.dentry, name, value, size, + flags); + + if (strcmp(name, "security.selinux")) + return -ENODATA; + + if (size > INCFS_MAX_FILE_ATTR_SIZE) + return -E2BIG; + + if (!strcmp(d->d_iname, INCFS_PENDING_READS_FILENAME)) { + stored_value = &mi->pending_read_xattr; + stored_size = &mi->pending_read_xattr_size; + } else if (!strcmp(d->d_iname, INCFS_LOG_FILENAME)) { + stored_value = &mi->log_xattr; + stored_size = &mi->log_xattr_size; + } else { return -ENODATA; + } + + kfree (*stored_value); + *stored_value = kzalloc(size, GFP_NOFS); + if (!*stored_value) + return -ENOMEM; - return vfs_getxattr(di->backing_path.dentry, name, value, size); + memcpy(*stored_value, value, size); + *stored_size = size; + return 0; } static ssize_t incfs_listxattr(struct dentry *d, char *list, size_t size) @@ -2152,7 +2203,7 @@ struct dentry *incfs_mount_fs(struct file_system_type *type, int flags, path_put(&backing_dir_path); sb->s_flags |= SB_ACTIVE; - pr_debug("infs: mount\n"); + pr_debug("incfs: mount\n"); return dget(sb->s_root); err: sb->s_fs_info = NULL; @@ -2173,12 +2224,11 @@ static int incfs_remount_fs(struct super_block *sb, int *flags, char *data) if (err) return err; - if (mi->mi_options.read_timeout_ms != options.read_timeout_ms) { - mi->mi_options.read_timeout_ms = options.read_timeout_ms; - pr_debug("incfs: new timeout_ms=%d", options.read_timeout_ms); - } + err = incfs_realloc_mount_info(mi, &options); + if (err) + return err; - pr_debug("infs: remount\n"); + pr_debug("incfs: remount\n"); return 0; } @@ -2186,7 +2236,7 @@ void incfs_kill_sb(struct super_block *sb) { struct mount_info *mi = sb->s_fs_info; - pr_debug("infs: unmount\n"); + pr_debug("incfs: unmount\n"); incfs_free_mount_info(mi); generic_shutdown_super(sb); } |