You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

1017 lines
35 KiB

* Copyright (c) 2007 Vreixo Formoso
* Copyright (c) 2009 - 2015 Thomas Schmitt
* This file is part of the libisofs project; you can redistribute it and/or
* modify it under the terms of the GNU General Public License version 2
* or later as published by the Free Software Foundation.
* See COPYING file for details.
#ifndef LIBISO_ECMA119_H_
#define LIBISO_ECMA119_H_
#include "libisofs.h"
#include "util.h"
#include "buffer.h"
#include <stdint.h>
#include <inttypes.h>
#include <pthread.h>
#define BLOCK_SIZE 2048
* Maximum file section size. Set to 4GB - 1 = 0xffffffff
#define MAX_ISO_FILE_SECTION_SIZE 0xffffffff
* When a file need to be splitted in several sections, the maximum size
* of such sections, but the last one. Set to a multiple of BLOCK_SIZE.
* Default to 4GB - 2048 = 0xFFFFF800
* The maximum number of partition images that can be registered. Depending
* on the system area type, the effectively usable number may be smaller or
* even 0.
* The cylindersize with SUN Disk Label
* (512 bytes/sector, 640 sectors/head, 1 head/cyl = 320 KiB).
* Expressed in ECMA-119 blocks of 2048 bytes/block.
#define ISO_SUN_CYL_SIZE 160
* Maximum length of a disc label text plus 1.
/* The maximum lenght of an specs violating ECMA-119 file identifier.
The theoretical limit is 254 - 34 - 28 (len of SUSP CE entry) = 192
Currently the practical limit is 254 - 34 - 96 (non-CE RR entries) - 28 (CE)
#ifdef Libisofs_with_rrip_rR
/* The theoretical maximum number of Apple Partition Map entries in the
System Area of an ISO image:
Block0 plus 63 entries with block size 512
/* The maximum number of MBR partition table entries.
/* The theoretical maximum number of GPT entries in the System Area of an
ISO image:
MBR plus GPT header block plus 248 GPT entries of 128 bytes each.
/* How many warnings to issue about writing Joliet names which cannot be
properly represented in UCS-2 and thus had to be defaultet to '_'.
* Holds the options for the image generation.
struct iso_write_opts {
int will_cancel;
int iso_level; /**< ISO level to write at. (ECMA-119, 10) */
/** Which extensions to support. */
unsigned int rockridge :1;
unsigned int joliet :1;
unsigned int iso1999 :1;
unsigned int hfsplus :1;
unsigned int fat :1;
unsigned int aaip :1; /* whether to write eventual ACL and EAs */
/* allways write timestamps in GMT */
unsigned int always_gmt :1;
* Relaxed constraints. Setting any of these to 1 break the specifications,
* but it is supposed to work on most moderns systems. Use with caution.
* Convert directory names for ECMA-119 the same way as other file names
* but do not force dots or add version numbers.
* This violates ECMA-119 by allowing one "." and especially ISO level 1
* by allowing DOS style 8.3 names rather than only 8 characters.
unsigned int allow_dir_id_ext :1;
* Omit the version number (";1") at the end of the ISO-9660 identifiers.
* Version numbers are usually not used.
* bit0= ECMA-119 and Joliet (for historical reasons)
* bit1= Joliet
unsigned int omit_version_numbers :2;
* Allow ISO-9660 directory hierarchy to be deeper than 8 levels.
unsigned int allow_deep_paths :1;
* Allow path in the ISO-9660 tree to have more than 255 characters.
unsigned int allow_longer_paths :1;
* Allow a single file or directory hierarchy to have up to 37 characters.
* This is larger than the 31 characters allowed by ISO level 2, and the
* extra space is taken from the version number, so this also forces
* omit_version_numbers.
unsigned int max_37_char_filenames :1;
* ISO-9660 forces filenames to have a ".", that separates file name from
* extension. libisofs adds it if original filename doesn't has one. Set
* this to 1 to prevent this behavior
* bit0= ECMA-119
* bit1= Joliet
unsigned int no_force_dots :2;
* Allow lowercase characters in ISO-9660 filenames. By default, only
* uppercase characters, numbers and a few other characters are allowed.
unsigned int allow_lowercase :1;
* Allow all ASCII characters to be appear on an ISO-9660 filename. Note
* that "/" and "\0" characters are never allowed, even in RR names.
unsigned int allow_full_ascii :1;
* If not allow_full_ascii is set: allow all 7 bit characters that would
* be allowed by allow_full_ascii. But still map lowercase to uppercase if
* not allow_lowercase is set to 1.
unsigned int allow_7bit_ascii :1;
* Allow all characters to be part of Volume and Volset identifiers on
* the Primary Volume Descriptor. This breaks ISO-9660 contraints, but
* should work on modern systems.
unsigned int relaxed_vol_atts :1;
* Allow paths in the Joliet tree to have more than 240 characters.
unsigned int joliet_longer_paths :1;
* Allow Joliet names up to 103 characters rather than 64.
unsigned int joliet_long_names :1;
* Use UTF-16BE rather than its subset UCS-2
unsigned int joliet_utf16 :1;
* Write Rock Ridge info as of specification RRIP-1.10 rather than
* RRIP-1.12: signature "RRIP_1991A" rather than "IEEE_1282",
* field PX without file serial number
unsigned int rrip_version_1_10 :1;
* Write field PX with file serial number even with RRIP-1.10
unsigned int rrip_1_10_px_ino :1;
* See iso_write_opts_set_hardlinks()
unsigned int hardlinks:1;
* Write AAIP as extension according to SUSP 1.10 rather than SUSP 1.12.
* I.e. without announcing it by an ER field and thus without the need
* to preceed the RRIP fields by an ES and to preceed the AA field by ES.
* This saves bytes and might avoid problems with readers which dislike
* ER fields other than the ones for RRIP.
* On the other hand, SUSP 1.12 frowns on such unannounced extensions
* and prescribes ER and ES. It does this since year 1994.
* In effect only if above flag .aaip is set to 1.
unsigned int aaip_susp_1_10 :1;
* Store as ECMA-119 Directory Record timestamp the mtime of the source
* rather than the image creation time. (The ECMA-119 prescription seems
* to expect that we do have a creation timestamp with the source.
* mkisofs writes mtimes and the result seems more suitable if mounted
* without Rock Ridge support.)
* bit0= ECMA-119, bit1= Joliet, bit2= ISO 9660:1999
unsigned int dir_rec_mtime :3;
* This describes the directory where to store Rock Ridge relocated
* directories.
* If not relaxation "allow_deep_paths" is in effect, it is necessary to
* relocate directories so that no ECMA-119 file path has more than
* 8 components. For Rock Ridge the relocated directories are linked forth
* and back to a placeholder at their original position in path level 8
* (entries CL and PL). Directories marked by entry RE are to be considered
* artefacts of relocation and shall not be read into a Rock Ridge tree.
* For plain ECMA-119, the relocation directory is just a normal directory
* which contains normal files and directories.
char *rr_reloc_dir; /* IsoNode name in root directory */
int rr_reloc_flags; /* bit0= mark auto-created rr_reloc_dir by RE
bit1= directory was auto-created
(cannot be set via API)
* Compute MD5 checksum for the whole session and record it as index 0 of
* the checksum blocks after the data area of the session. The layout and
* position of these blocks will be recorded in xattr "" of the
* root node. See see also API call iso_image_get_session_md5().
unsigned int md5_session_checksum :1;
* Compute MD5 checksums for IsoFile objects and write them to blocks
* after the data area of the session. The layout and position of these
* blocks will be recorded in xattr "" of the root node.
* The indice of the MD5 sums will be recorded with the IsoFile directory
* entries as xattr "". See also API call iso_file_get_md5().
* bit0= compute individual checksums
* bit1= pre-compute checksum and compare it with actual one.
* Raise MISHAP if mismatch.
unsigned int md5_file_checksums :2;
/** If files should be sorted based on their weight. */
unsigned int sort_files :1;
* The following options set the default values for files and directory
* permissions, gid and uid. All these take one of three values: 0, 1 or 2.
* If 0, the corresponding attribute will be kept as setted in the IsoNode.
* Unless you have changed it, it corresponds to the value on disc, so it
* is suitable for backup purposes. If set to 1, the corresponding attrib.
* will be changed by a default suitable value. Finally, if you set it to
* 2, the attrib. will be changed with the value specified in the options
* below. Note that for mode attributes, only the permissions are set, the
* file type remains unchanged.
unsigned int replace_dir_mode :2;
unsigned int replace_file_mode :2;
unsigned int replace_uid :2;
unsigned int replace_gid :2;
mode_t dir_mode; /** Mode to use on dirs when replace_dir_mode == 2. */
mode_t file_mode; /** Mode to use on files when replace_file_mode == 2. */
uid_t uid; /** uid to use when replace_uid == 2. */
gid_t gid; /** gid to use when replace_gid == 2. */
* See API call iso_write_opts_set_old_empty().
unsigned int old_empty :1;
* Extra Caution: This option breaks any assumptions about names that
* are supported by ECMA-119 specifications.
* Omit any translation which would make a file name compliant to the
* ECMA-119 rules. This includes and exceeds omit_version_numbers,
* max_37_char_filenames, no_force_dots bit0, allow_lowercase.
* The maximum name length is given by this variable.
* There is a length limit of ISO_UNTRANSLATED_NAMES_MAX characters,
* because ECMA-119 allows 254 byte in a directory record, some
* of them are occupied by ECMA-119, some more are needed for SUSP CE,
* and some are fixely occupied by libisofs Rock Ridge code.
* The default value 0 disables this feature.
unsigned int untranslated_name_len;
* 0 to use IsoNode timestamps, 1 to use recording time, 2 to use
* values from timestamp field. This has only meaning if RR extensions
* are enabled.
unsigned int replace_timestamps :2;
time_t timestamp;
* Charset for the RR filenames that will be created.
* NULL to use default charset, the locale one.
char *output_charset;
* This flags control the type of the image to create. Libisofs support
* two kind of images: stand-alone and appendable.
* A stand-alone image is an image that is valid alone, and that can be
* mounted by its own. This is the kind of image you will want to create
* in most cases. A stand-alone image can be burned in an empty CD or DVD,
* or write to an .iso file for future burning or distribution.
* On the other side, an appendable image is not self contained, it refers
* to serveral files that are stored outside the image. Its usage is for
* multisession discs, where you add data in a new session, while the
* previous session data can still be accessed. In those cases, the old
* data is not written again. Instead, the new image refers to it, and thus
* it's only valid when appended to the original. Note that in those cases
* the image will be written after the original, and thus you will want
* to use a ms_block greater than 0.
* Note that if you haven't import a previous image (by means of
* iso_image_import()), the image will always be a stand-alone image, as
* there is no previous data to refer to.
unsigned int appendable : 1;
* Start block of the image. It is supposed to be the lba where the first
* block of the image will be written on disc. All references inside the
* ISO image will take this into account, thus providing a mountable image.
* For appendable images, that are written to a new session, you should
* pass here the lba of the next writable address on disc.
* In stand alone images this is usually 0. However, you may want to
* provide a different ms_block if you don't plan to burn the image in the
* first session on disc, such as in some CD-Extra disc whether the data
* image is written in a new session after some audio tracks.
uint32_t ms_block;
* When not NULL, it should point to a buffer of at least 64KiB, where
* libisofs will write the contents that should be written at the beginning
* of a overwriteable media, to grow the image. The growing of an image is
* a way, used by first time in growisofs by Andy Polyakov, to allow the
* appending of new data to non-multisession media, such as DVD+RW, in the
* same way you append a new session to a multisession disc, i.e., without
* need to write again the contents of the previous image.
* Note that if you want this kind of image growing, you will also need to
* set appendable to "1" and provide a valid ms_block after the previous
* image.
* You should initialize the buffer either with 0s, or with the contents of
* the first blocks of the image you're growing. In most cases, 0 is good
* enought.
uint8_t *overwrite;
* Size, in number of blocks, of the FIFO buffer used between the writer
* thread and the burn_source. You have to provide at least a 32 blocks
* buffer.
size_t fifo_size;
* This is not an option setting but a value returned after the options
* were used to compute the layout of the image.
* It tells the LBA of the first plain file data block in the image.
uint32_t data_start_lba;
* If not empty: A text holding parameters "name" and "timestamp" for
* a scdbackup stream checksum tag. See scdbackup/README appendix VERIFY.
* It makes sense only for single session images which start at LBA 0.
* Such a tag may be part of a libisofs checksum tag block after the
* session tag line. It then covers the whole session up to its own start
* position.
char scdbackup_tag_parm[100];
/* If not NULL: A pointer to an application provided array with
at least 512 characters. The effectively written scdbackup tag
will be copied to this memory location.
char *scdbackup_tag_written;
* See ecma119_image : System Area related information
char *system_area_data;
int system_area_size;
int system_area_options;
/* User settable PVD time stamps */
time_t vol_creation_time;
time_t vol_modification_time;
time_t vol_expiration_time;
time_t vol_effective_time;
/* To eventually override vol_creation_time and vol_modification_time
* by unconverted string with timezone 0
char vol_uuid[17];
/* The number of unclaimed 2K blocks before start of partition 1 as of
the MBR in system area.
Must be 0 or >= 16. (Actually >= number of voldescr + checksum tag)
uint32_t partition_offset;
/* Partition table parameter: 1 to 63, 0= disabled/default */
int partition_secs_per_head;
/* 1 to 255, 0= disabled/default */
int partition_heads_per_cyl;
#ifdef Libisofs_with_libjtE
/* Parameters and state of Jigdo Template Export environment.
struct libjte_env *libjte_handle;
#endif /* Libisofs_with_libjtE */
/* A trailing padding of zero bytes which belongs to the image
uint32_t tail_blocks;
/* Eventual disk file path of a PreP partition which shall be prepended
to HFS+/FAT and IsoFileSrc areas and marked by an MBR partition entry.
char *prep_partition;
int prep_part_flag;
/* Eventual disk file path of an EFI system partition image which shall
be prepended to HFS+/FAT and IsoFileSrc areas and marked by a GPT entry.
char *efi_boot_partition;
int efi_boot_part_flag;
/* Eventual disk file paths of prepared images which shall be appended
after the ISO image and described by partiton table entries in a MBR
char *appended_partitions[ISO_MAX_PARTITIONS];
uint8_t appended_part_types[ISO_MAX_PARTITIONS];
int appended_part_flags[ISO_MAX_PARTITIONS];
/* If 1: With appended partitions: create protective MBR and mark by GPT
int appended_as_gpt;
/* If 1: With appended partitions: mark by APM partition
int appended_as_apm;
/* If 1: Obey struct el_torito_boot_image.isolinux_options bit2-7 and bit8.
I.e. mention boot image as partition in GPT and/or APM.
int part_like_isohybrid;
/* Eventual name of the non-ISO aspect of the image. E.g. SUN ASCII label.
char ascii_disc_label[ISO_DISC_LABEL_SIZE];
/* HFS+ image serial number.
* 00...00 means that it shall be generated by libisofs.
uint8_t hfsp_serial_number[8];
/* Allocation block size of HFS+ : 0= auto , 512, or 2048
int hfsp_block_size;
/* Block size of and in APM : 0= auto , 512, or 2048
int apm_block_size;
typedef struct ecma119_image Ecma119Image;
typedef struct ecma119_node Ecma119Node;
typedef struct joliet_node JolietNode;
typedef struct iso1999_node Iso1999Node;
typedef struct hfsplus_node HFSPlusNode;
typedef struct Iso_File_Src IsoFileSrc;
typedef struct Iso_Image_Writer IsoImageWriter;
struct ecma119_image
int refcount;
IsoImage *image;
Ecma119Node *root;
IsoWriteOpts *opts;
/** Whether El Torito data will be produced */
unsigned int eltorito :1;
/* The ECMA-119 directory node where to store Rock Ridge relocated
directories. (Path is in IsoWriteOpts.rr_reloc_dir)
Ecma119Node *rr_reloc_node; /* Directory node in ecma119_image */
* Mode replace. If one of these flags is set, the correspodent values are
* replaced with values below. Both get computed from IsoWriteOpts.
unsigned int replace_uid :1;
unsigned int replace_gid :1;
unsigned int replace_file_mode :1;
unsigned int replace_dir_mode :1;
unsigned int replace_timestamps :1;
/* Mode replacement values. */
uid_t uid;
gid_t gid;
mode_t file_mode;
mode_t dir_mode;
time_t timestamp;
/* Effective charsets */
char *input_charset;
char *output_charset;
time_t now; /**< Time at which writing began. */
/* Total size of the output. Counted in bytes.
* Includes ISO filesystem and appended data.
off_t total_size;
/** Size actually governed by the ISO filesystem part of the output */
uint32_t vol_space_size;
/* 1= write the total size into the PVD of the ISO,
* 0= write vol_space_size
int pvd_size_is_total_size;
/* Bytes already written to image output */
off_t bytes_written;
/* just for progress notification */
int percent_written;
* Block being processed, either during image writing or structure
* size calculation.
uint32_t curblock;
* The address to be used for the content pointer of empty data files.
uint32_t empty_file_block;
* The calculated block address after ECMA-119 tree and eventual
* tree checksum tag.
uint32_t tree_end_block;
* number of dirs in ECMA-119 tree, computed together with dir position,
* and needed for path table computation in a efficient way
size_t ndirs;
uint32_t path_table_size;
uint32_t l_path_table_pos;
uint32_t m_path_table_pos;
* Joliet related information
JolietNode *joliet_root;
size_t joliet_ndirs;
uint32_t joliet_path_table_size;
uint32_t joliet_l_path_table_pos;
uint32_t joliet_m_path_table_pos;
size_t joliet_ucs2_failures;
* HFS+ related information
* (by Vladimir Serbinenko, see libisofs/hfsplus.c)
HFSPlusNode *hfsp_leafs;
struct hfsplus_btree_level *hfsp_levels;
uint32_t hfsp_nlevels;
uint32_t hfsp_part_start;
uint32_t hfsp_nfiles;
uint32_t hfsp_ndirs;
uint32_t hfsp_cat_id;
uint32_t hfsp_allocation_blocks;
uint32_t hfsp_allocation_file_start;
uint32_t hfsp_extent_file_start;
uint32_t hfsp_catalog_file_start;
uint32_t hfsp_total_blocks;
uint32_t hfsp_allocation_size;
uint32_t hfsp_nleafs;
uint32_t hfsp_curleaf;
uint32_t hfsp_nnodes;
uint32_t hfsp_bless_id[ISO_HFSPLUS_BLESS_MAX];
uint32_t hfsp_collision_count;
* ISO 9660:1999 related information
Iso1999Node *iso1999_root;
size_t iso1999_ndirs;
uint32_t iso1999_path_table_size;
uint32_t iso1999_l_path_table_pos;
uint32_t iso1999_m_path_table_pos;
* El-Torito related information
struct el_torito_boot_catalog *catalog;
IsoFileSrc *cat; /**< location of the boot catalog in the new image */
int num_bootsrc;
IsoFileSrc **bootsrc; /* location of the boot images in the new image */
int *boot_appended_idx; /* Appended partition which serve as boot images */
uint32_t *boot_intvl_start; /* In blocks of 2048 bytes */
uint32_t *boot_intvl_size; /* In blocks of 512 bytes */
* System Area related information
/* Content of an embedded boot image. Valid if not NULL.
* In that case it must point to a memory buffer at least 32 kB.
char *system_area_data;
* bit0= Only with DOS MBR
* Make bytes 446 - 512 of the system area a partition
* table which reserves partition 1 from byte 63*512 to the
* end of the ISO image. Assumed are 63 secs/hed, 255 head/cyl.
* (GRUB protective msdos label.)
* This works with and without system_area_data.
* bit1= Only with DOS MBR
* Apply isohybrid MBR patching to the system area.
* This works only with system_area_data plus ISOLINUX boot image
* and only if not bit0 is set.
* bit2-7= System area type
* 0= DOS MBR
* 1= MIPS Big Endian Volume Header
* 2= DEC Boot Block for MIPS Little Endian
* 3= SUN Disk Label for SUN SPARC
* bit8-9= Only with DOS MBR
* Cylinder alignment mode eventually pads the image to make it
* end at a cylinder boundary.
* 0 = auto (align if bit1)
* 1 = always align to cylinder boundary
* 2 = never align to cylinder boundary
* 3 = always align, additionally pad up and align partitions
* which were appended by iso_write_opts_set_partition_img()
* bit10-13= System area sub type
* With type 0 = MBR:
* Gets overridden by bit0 and bit1.
* 0 = no particular sub type
* 1 = CHRP: A single MBR partition of type 0x96 covers the
* ISO image. Not compatible with any other feature
* which needs to have own MBR partition entries.
int system_area_options;
* Number of pad blocks that we need to write. Padding blocks are blocks
* filled by 0s that we put between the directory structures and the file
* data. These padding blocks are added by libisofs to improve the handling
* of image growing. The idea is that the first blocks in the image are
* overwritten with the volume descriptors of the new image. These first
* blocks usually correspond to the volume descriptors and directory
* structure of the old image, and can be safety overwritten. However,
* with very small images they might correspond to valid data. To ensure
* this never happens, what we do is to add padding bytes, to ensure no
* file data is written in the first 64 KiB, that are the bytes we usually
* overwrite.
uint32_t mspad_blocks;
size_t nwriters;
IsoImageWriter **writers;
/* tree of files sources */
IsoRBTree *files;
struct iso_filesrc_list_item *ecma119_hidden_list;
unsigned int checksum_idx_counter;
void *checksum_ctx;
off_t checksum_counter;
uint32_t checksum_rlsb_tag_pos;
uint32_t checksum_sb_tag_pos;
uint32_t checksum_tree_tag_pos;
uint32_t checksum_tag_pos;
char image_md5[16];
char *checksum_buffer;
uint32_t checksum_array_pos;
uint32_t checksum_range_start;
uint32_t checksum_range_size;
char *opts_overwrite; /* Points to IsoWriteOpts->overwrite.
Use only underneath ecma119_image_new()
and if not NULL*/
/* Buffer for communication between burn_source and writer thread */
IsoRingBuffer *buffer;
/* writer thread descriptor */
pthread_t wthread;
int wthread_is_running;
pthread_attr_t th_attr;
/* Effective partition table parameter: 1 to 63, 0= disabled/default */
int partition_secs_per_head;
/* 1 to 255, 0= disabled/default */
int partition_heads_per_cyl;
/* The currently applicable LBA offset. To be subtracted from any LBA
* that is mentioned in volume descriptors, trees, path tables,
* Either 0 or .partition_offset
uint32_t eff_partition_offset;
/* The second ECMA-119 directory tree and path tables */
Ecma119Node *partition_root;
uint32_t partition_l_table_pos;
uint32_t partition_m_table_pos;
/* The second Joliet directory tree and path tables */
JolietNode *j_part_root;
uint32_t j_part_l_path_table_pos;
uint32_t j_part_m_path_table_pos;
/* Memorized ELF parameters from MIPS Little Endian boot file */
uint32_t mipsel_e_entry;
uint32_t mipsel_p_offset;
uint32_t mipsel_p_vaddr;
uint32_t mipsel_p_filesz;
/* A data file of which the position and size shall be written after
a SUN Disk Label.
IsoFileSrc *sparc_core_src;
/* Counted in blocks of 2048 */
uint32_t appended_part_prepad[ISO_MAX_PARTITIONS];
uint32_t appended_part_start[ISO_MAX_PARTITIONS];
uint32_t appended_part_size[ISO_MAX_PARTITIONS];
int have_appended_partitions;