diff options
author | Martin Matuska <mm@FreeBSD.org> | 2016-05-11 10:10:11 +0000 |
---|---|---|
committer | Martin Matuska <mm@FreeBSD.org> | 2016-05-11 10:10:11 +0000 |
commit | 8c8f03ca5b7f1f0d07c23e7c354a45b7c95335f0 (patch) | |
tree | 6dbaec0a9141562742e447716c2fed6e53d97c6d /libarchive | |
parent | 43341e7cf4b9eee3ead77da04edc14370fe5b166 (diff) | |
download | src-8c8f03ca5b7f1f0d07c23e7c354a45b7c95335f0.tar.gz src-8c8f03ca5b7f1f0d07c23e7c354a45b7c95335f0.zip |
Notes
Diffstat (limited to 'libarchive')
-rw-r--r-- | libarchive/CMakeLists.txt | 193 | ||||
-rw-r--r-- | libarchive/archive_entry_copy_bhfi.c | 75 | ||||
-rw-r--r-- | libarchive/archive_read_disk_windows.c | 2296 | ||||
-rw-r--r-- | libarchive/archive_windows.c | 908 | ||||
-rw-r--r-- | libarchive/archive_windows.h | 306 | ||||
-rw-r--r-- | libarchive/archive_write_disk_windows.c | 2502 | ||||
-rw-r--r-- | libarchive/config_freebsd.h | 160 | ||||
-rw-r--r-- | libarchive/filter_fork_windows.c | 190 | ||||
-rw-r--r-- | libarchive/mtree.5 | 269 | ||||
-rw-r--r-- | libarchive/test/.cvsignore | 10 | ||||
-rw-r--r-- | libarchive/test/CMakeLists.txt | 256 |
11 files changed, 7165 insertions, 0 deletions
diff --git a/libarchive/CMakeLists.txt b/libarchive/CMakeLists.txt new file mode 100644 index 000000000000..ecb0409bd9d8 --- /dev/null +++ b/libarchive/CMakeLists.txt @@ -0,0 +1,193 @@ + +############################################ +# +# How to build libarchive +# +############################################ + +# Public headers +SET(include_HEADERS + archive.h + archive_entry.h +) + +# Sources and private headers +SET(libarchive_SOURCES + archive_acl.c + archive_check_magic.c + archive_cmdline.c + archive_cmdline_private.h + archive_crc32.h + archive_crypto.c + archive_crypto_private.h + archive_endian.h + archive_entry.c + archive_entry.h + archive_entry_copy_stat.c + archive_entry_link_resolver.c + archive_entry_locale.h + archive_entry_private.h + archive_entry_sparse.c + archive_entry_stat.c + archive_entry_strmode.c + archive_entry_xattr.c + archive_getdate.c + archive_match.c + archive_options.c + archive_options_private.h + archive_pathmatch.c + archive_pathmatch.h + archive_platform.h + archive_ppmd_private.h + archive_ppmd7.c + archive_ppmd7_private.h + archive_private.h + archive_rb.c + archive_rb.h + archive_read.c + archive_read_append_filter.c + archive_read_data_into_fd.c + archive_read_disk_entry_from_file.c + archive_read_disk_posix.c + archive_read_disk_private.h + archive_read_disk_set_standard_lookup.c + archive_read_extract.c + archive_read_open_fd.c + archive_read_open_file.c + archive_read_open_filename.c + archive_read_open_memory.c + archive_read_private.h + archive_read_set_format.c + archive_read_set_options.c + archive_read_support_filter_all.c + archive_read_support_filter_bzip2.c + archive_read_support_filter_compress.c + archive_read_support_filter_gzip.c + archive_read_support_filter_grzip.c + archive_read_support_filter_lrzip.c + archive_read_support_filter_lzop.c + archive_read_support_filter_none.c + archive_read_support_filter_program.c + archive_read_support_filter_rpm.c + archive_read_support_filter_uu.c + archive_read_support_filter_xz.c + archive_read_support_format_7zip.c + archive_read_support_format_all.c + archive_read_support_format_ar.c + archive_read_support_format_by_code.c + archive_read_support_format_cab.c + archive_read_support_format_cpio.c + archive_read_support_format_empty.c + archive_read_support_format_iso9660.c + archive_read_support_format_lha.c + archive_read_support_format_mtree.c + archive_read_support_format_rar.c + archive_read_support_format_raw.c + archive_read_support_format_tar.c + archive_read_support_format_xar.c + archive_read_support_format_zip.c + archive_string.c + archive_string.h + archive_string_composition.h + archive_string_sprintf.c + archive_util.c + archive_virtual.c + archive_write.c + archive_write_disk_acl.c + archive_write_disk_posix.c + archive_write_disk_private.h + archive_write_disk_set_standard_lookup.c + archive_write_private.h + archive_write_open_fd.c + archive_write_open_file.c + archive_write_open_filename.c + archive_write_open_memory.c + archive_write_add_filter.c + archive_write_add_filter_b64encode.c + archive_write_add_filter_by_name.c + archive_write_add_filter_bzip2.c + archive_write_add_filter_compress.c + archive_write_add_filter_grzip.c + archive_write_add_filter_gzip.c + archive_write_add_filter_lrzip.c + archive_write_add_filter_lzop.c + archive_write_add_filter_none.c + archive_write_add_filter_program.c + archive_write_add_filter_uuencode.c + archive_write_add_filter_xz.c + archive_write_set_format.c + archive_write_set_format_7zip.c + archive_write_set_format_ar.c + archive_write_set_format_by_name.c + archive_write_set_format_cpio.c + archive_write_set_format_cpio_newc.c + archive_write_set_format_gnutar.c + archive_write_set_format_iso9660.c + archive_write_set_format_mtree.c + archive_write_set_format_pax.c + archive_write_set_format_shar.c + archive_write_set_format_ustar.c + archive_write_set_format_v7tar.c + archive_write_set_format_xar.c + archive_write_set_format_zip.c + archive_write_set_options.c + filter_fork_posix.c + filter_fork.h +) + +# Man pages +SET(libarchive_MANS + archive_entry.3 + archive_entry_acl.3 + archive_entry_linkify.3 + archive_entry_paths.3 + archive_entry_perms.3 + archive_entry_stat.3 + archive_entry_time.3 + archive_read.3 + archive_read_disk.3 + archive_read_set_options.3 + archive_util.3 + archive_write.3 + archive_write_disk.3 + archive_write_set_options.3 + cpio.5 + libarchive.3 + libarchive_internals.3 + libarchive-formats.5 + mtree.5 + tar.5 +) + +IF(WIN32 AND NOT CYGWIN) + LIST(APPEND libarchive_SOURCES archive_entry_copy_bhfi.c) + LIST(APPEND libarchive_SOURCES archive_read_disk_windows.c) + LIST(APPEND libarchive_SOURCES archive_windows.c) + LIST(APPEND libarchive_SOURCES archive_windows.h) + LIST(APPEND libarchive_SOURCES archive_write_disk_windows.c) + LIST(APPEND libarchive_SOURCES filter_fork_windows.c) +ENDIF(WIN32 AND NOT CYGWIN) + +# Libarchive is a shared library +ADD_LIBRARY(archive SHARED ${libarchive_SOURCES} ${include_HEADERS}) +TARGET_LINK_LIBRARIES(archive ${ADDITIONAL_LIBS}) +SET_TARGET_PROPERTIES(archive PROPERTIES SOVERSION ${SOVERSION}) + +# archive_static is a static library +ADD_LIBRARY(archive_static STATIC ${libarchive_SOURCES} ${include_HEADERS}) +SET_TARGET_PROPERTIES(archive_static PROPERTIES COMPILE_DEFINITIONS + LIBARCHIVE_STATIC) +# On Posix systems, libarchive.so and libarchive.a can co-exist. +IF(NOT WIN32 OR CYGWIN) + SET_TARGET_PROPERTIES(archive_static PROPERTIES OUTPUT_NAME archive) +ENDIF(NOT WIN32 OR CYGWIN) + +# How to install the libraries +INSTALL(TARGETS archive archive_static + RUNTIME DESTINATION bin + LIBRARY DESTINATION lib + ARCHIVE DESTINATION lib) +INSTALL_MAN(${libarchive_MANS}) +INSTALL(FILES ${include_HEADERS} DESTINATION include) + +add_subdirectory(test) diff --git a/libarchive/archive_entry_copy_bhfi.c b/libarchive/archive_entry_copy_bhfi.c new file mode 100644 index 000000000000..77bf38e450f2 --- /dev/null +++ b/libarchive/archive_entry_copy_bhfi.c @@ -0,0 +1,75 @@ +/*- + * Copyright (c) 2003-2007 Tim Kientzle + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "archive_platform.h" +__FBSDID("$FreeBSD$"); + +#include "archive_private.h" +#include "archive_entry.h" + +#if defined(_WIN32) && !defined(__CYGWIN__) + +#define EPOC_TIME ARCHIVE_LITERAL_ULL(116444736000000000) + +__inline static void +fileTimeToUtc(const FILETIME *filetime, time_t *t, long *ns) +{ + ULARGE_INTEGER utc; + + utc.HighPart = filetime->dwHighDateTime; + utc.LowPart = filetime->dwLowDateTime; + if (utc.QuadPart >= EPOC_TIME) { + utc.QuadPart -= EPOC_TIME; + *t = (time_t)(utc.QuadPart / 10000000); /* milli seconds base */ + *ns = (long)(utc.QuadPart % 10000000) * 100;/* nano seconds base */ + } else { + *t = 0; + *ns = 0; + } +} + +void +archive_entry_copy_bhfi(struct archive_entry *entry, + BY_HANDLE_FILE_INFORMATION *bhfi) +{ + time_t secs; + long nsecs; + + fileTimeToUtc(&bhfi->ftLastAccessTime, &secs, &nsecs); + archive_entry_set_atime(entry, secs, nsecs); + fileTimeToUtc(&bhfi->ftLastWriteTime, &secs, &nsecs); + archive_entry_set_mtime(entry, secs, nsecs); + fileTimeToUtc(&bhfi->ftCreationTime, &secs, &nsecs); + archive_entry_set_birthtime(entry, secs, nsecs); + archive_entry_set_ctime(entry, secs, nsecs); + archive_entry_set_dev(entry, bhfi->dwVolumeSerialNumber); + archive_entry_set_ino64(entry, (((int64_t)bhfi->nFileIndexHigh) << 32) + + bhfi->nFileIndexLow); + archive_entry_set_nlink(entry, bhfi->nNumberOfLinks); + archive_entry_set_size(entry, (((int64_t)bhfi->nFileSizeHigh) << 32) + + bhfi->nFileSizeLow); + /* archive_entry_set_mode(entry, st->st_mode); */ +} +#endif diff --git a/libarchive/archive_read_disk_windows.c b/libarchive/archive_read_disk_windows.c new file mode 100644 index 000000000000..9c5420d80e77 --- /dev/null +++ b/libarchive/archive_read_disk_windows.c @@ -0,0 +1,2296 @@ +/*- + * Copyright (c) 2003-2009 Tim Kientzle + * Copyright (c) 2010-2012 Michihiro NAKAJIMA + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer + * in this position and unchanged. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ +#include "archive_platform.h" +__FBSDID("$FreeBSD$"); + +#if defined(_WIN32) && !defined(__CYGWIN__) + +#ifdef HAVE_ERRNO_H +#include <errno.h> +#endif +#ifdef HAVE_STDLIB_H +#include <stdlib.h> +#endif +#include <winioctl.h> + +#include "archive.h" +#include "archive_string.h" +#include "archive_entry.h" +#include "archive_private.h" +#include "archive_read_disk_private.h" + +#ifndef O_BINARY +#define O_BINARY 0 +#endif +#ifndef IO_REPARSE_TAG_SYMLINK +/* Old SDKs do not provide IO_REPARSE_TAG_SYMLINK */ +#define IO_REPARSE_TAG_SYMLINK 0xA000000CL +#endif + +/*- + * This is a new directory-walking system that addresses a number + * of problems I've had with fts(3). In particular, it has no + * pathname-length limits (other than the size of 'int'), handles + * deep logical traversals, uses considerably less memory, and has + * an opaque interface (easier to modify in the future). + * + * Internally, it keeps a single list of "tree_entry" items that + * represent filesystem objects that require further attention. + * Non-directories are not kept in memory: they are pulled from + * readdir(), returned to the client, then freed as soon as possible. + * Any directory entry to be traversed gets pushed onto the stack. + * + * There is surprisingly little information that needs to be kept for + * each item on the stack. Just the name, depth (represented here as the + * string length of the parent directory's pathname), and some markers + * indicating how to get back to the parent (via chdir("..") for a + * regular dir or via fchdir(2) for a symlink). + */ + +struct restore_time { + const wchar_t *full_path; + FILETIME lastWriteTime; + FILETIME lastAccessTime; + mode_t filetype; +}; + +struct tree_entry { + int depth; + struct tree_entry *next; + struct tree_entry *parent; + size_t full_path_dir_length; + struct archive_wstring name; + struct archive_wstring full_path; + size_t dirname_length; + int64_t dev; + int64_t ino; + int flags; + int filesystem_id; + /* How to restore time of a directory. */ + struct restore_time restore_time; +}; + +struct filesystem { + int64_t dev; + int synthetic; + int remote; + DWORD bytesPerSector; +}; + +/* Definitions for tree_entry.flags bitmap. */ +#define isDir 1 /* This entry is a regular directory. */ +#define isDirLink 2 /* This entry is a symbolic link to a directory. */ +#define needsFirstVisit 4 /* This is an initial entry. */ +#define needsDescent 8 /* This entry needs to be previsited. */ +#define needsOpen 16 /* This is a directory that needs to be opened. */ +#define needsAscent 32 /* This entry needs to be postvisited. */ + +/* + * On Windows, "first visit" is handled as a pattern to be handed to + * _findfirst(). This is consistent with Windows conventions that + * file patterns are handled within the application. On Posix, + * "first visit" is just returned to the client. + */ + +#define MAX_OVERLAPPED 8 +#define BUFFER_SIZE (1024 * 8) +#define DIRECT_IO 0/* Disabled */ +#define ASYNC_IO 1/* Enabled */ + +/* + * Local data for this package. + */ +struct tree { + struct tree_entry *stack; + struct tree_entry *current; + HANDLE d; + WIN32_FIND_DATAW _findData; + WIN32_FIND_DATAW *findData; + int flags; + int visit_type; + /* Error code from last failed operation. */ + int tree_errno; + + /* A full path with "\\?\" prefix. */ + struct archive_wstring full_path; + size_t full_path_dir_length; + /* Dynamically-sized buffer for holding path */ + struct archive_wstring path; + + /* Last path element */ + const wchar_t *basename; + /* Leading dir length */ + size_t dirname_length; + + int depth; + + BY_HANDLE_FILE_INFORMATION lst; + BY_HANDLE_FILE_INFORMATION st; + int descend; + /* How to restore time of a file. */ + struct restore_time restore_time; + + struct entry_sparse { + int64_t length; + int64_t offset; + } *sparse_list, *current_sparse; + int sparse_count; + int sparse_list_size; + + char initial_symlink_mode; + char symlink_mode; + struct filesystem *current_filesystem; + struct filesystem *filesystem_table; + int initial_filesystem_id; + int current_filesystem_id; + int max_filesystem_id; + int allocated_filesytem; + + HANDLE entry_fh; + int entry_eof; + int64_t entry_remaining_bytes; + int64_t entry_total; + + int ol_idx_doing; + int ol_idx_done; + int ol_num_doing; + int ol_num_done; + int64_t ol_remaining_bytes; + int64_t ol_total; + struct la_overlapped { + OVERLAPPED ol; + struct archive * _a; + unsigned char *buff; + size_t buff_size; + int64_t offset; + size_t bytes_expected; + size_t bytes_transferred; + } ol[MAX_OVERLAPPED]; + int direct_io; + int async_io; +}; + +#define bhfi_dev(bhfi) ((bhfi)->dwVolumeSerialNumber) +/* Treat FileIndex as i-node. We should remove a sequence number + * which is high-16-bits of nFileIndexHigh. */ +#define bhfi_ino(bhfi) \ + ((((int64_t)((bhfi)->nFileIndexHigh & 0x0000FFFFUL)) << 32) \ + + (bhfi)->nFileIndexLow) + +/* Definitions for tree.flags bitmap. */ +#define hasStat 16 /* The st entry is valid. */ +#define hasLstat 32 /* The lst entry is valid. */ +#define needsRestoreTimes 128 + +static int +tree_dir_next_windows(struct tree *t, const wchar_t *pattern); + +/* Initiate/terminate a tree traversal. */ +static struct tree *tree_open(const wchar_t *, int, int); +static struct tree *tree_reopen(struct tree *, const wchar_t *, int); +static void tree_close(struct tree *); +static void tree_free(struct tree *); +static void tree_push(struct tree *, const wchar_t *, const wchar_t *, + int, int64_t, int64_t, struct restore_time *); + +/* + * tree_next() returns Zero if there is no next entry, non-zero if + * there is. Note that directories are visited three times. + * Directories are always visited first as part of enumerating their + * parent; that is a "regular" visit. If tree_descend() is invoked at + * that time, the directory is added to a work list and will + * subsequently be visited two more times: once just after descending + * into the directory ("postdescent") and again just after ascending + * back to the parent ("postascent"). + * + * TREE_ERROR_DIR is returned if the descent failed (because the + * directory couldn't be opened, for instance). This is returned + * instead of TREE_POSTDESCENT/TREE_POSTASCENT. TREE_ERROR_DIR is not a + * fatal error, but it does imply that the relevant subtree won't be + * visited. TREE_ERROR_FATAL is returned for an error that left the + * traversal completely hosed. Right now, this is only returned for + * chdir() failures during ascent. + */ +#define TREE_REGULAR 1 +#define TREE_POSTDESCENT 2 +#define TREE_POSTASCENT 3 +#define TREE_ERROR_DIR -1 +#define TREE_ERROR_FATAL -2 + +static int tree_next(struct tree *); + +/* + * Return information about the current entry. + */ + +/* + * The current full pathname, length of the full pathname, and a name + * that can be used to access the file. Because tree does use chdir + * extensively, the access path is almost never the same as the full + * current path. + * + */ +static const wchar_t *tree_current_path(struct tree *); +static const wchar_t *tree_current_access_path(struct tree *); + +/* + * Request the lstat() or stat() data for the current path. Since the + * tree package needs to do some of this anyway, and caches the + * results, you should take advantage of it here if you need it rather + * than make a redundant stat() or lstat() call of your own. + */ +static const BY_HANDLE_FILE_INFORMATION *tree_current_stat(struct tree *); +static const BY_HANDLE_FILE_INFORMATION *tree_current_lstat(struct tree *); + +/* The following functions use tricks to avoid a certain number of + * stat()/lstat() calls. */ +/* "is_physical_dir" is equivalent to S_ISDIR(tree_current_lstat()->st_mode) */ +static int tree_current_is_physical_dir(struct tree *); +/* "is_physical_link" is equivalent to S_ISLNK(tree_current_lstat()->st_mode) */ +static int tree_current_is_physical_link(struct tree *); +/* Instead of archive_entry_copy_stat for BY_HANDLE_FILE_INFORMATION */ +static void tree_archive_entry_copy_bhfi(struct archive_entry *, + struct tree *, const BY_HANDLE_FILE_INFORMATION *); +/* "is_dir" is equivalent to S_ISDIR(tree_current_stat()->st_mode) */ +static int tree_current_is_dir(struct tree *); +static int update_current_filesystem(struct archive_read_disk *a, + int64_t dev); +static int setup_current_filesystem(struct archive_read_disk *); +static int tree_target_is_same_as_parent(struct tree *, + const BY_HANDLE_FILE_INFORMATION *); + +static int _archive_read_disk_open_w(struct archive *, const wchar_t *); +static int _archive_read_free(struct archive *); +static int _archive_read_close(struct archive *); +static int _archive_read_data_block(struct archive *, + const void **, size_t *, int64_t *); +static int _archive_read_next_header2(struct archive *, + struct archive_entry *); +static const char *trivial_lookup_gname(void *, int64_t gid); +static const char *trivial_lookup_uname(void *, int64_t uid); +static int setup_sparse(struct archive_read_disk *, struct archive_entry *); +static int close_and_restore_time(HANDLE, struct tree *, + struct restore_time *); +static int setup_sparse_from_disk(struct archive_read_disk *, + struct archive_entry *, HANDLE); + + + +static struct archive_vtable * +archive_read_disk_vtable(void) +{ + static struct archive_vtable av; + static int inited = 0; + + if (!inited) { + av.archive_free = _archive_read_free; + av.archive_close = _archive_read_close; + av.archive_read_data_block = _archive_read_data_block; + av.archive_read_next_header2 = _archive_read_next_header2; + inited = 1; + } + return (&av); +} + +const char * +archive_read_disk_gname(struct archive *_a, int64_t gid) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_read_disk_gname")) + return (NULL); + if (a->lookup_gname == NULL) + return (NULL); + return ((*a->lookup_gname)(a->lookup_gname_data, gid)); +} + +const char * +archive_read_disk_uname(struct archive *_a, int64_t uid) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + if (ARCHIVE_OK != __archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_read_disk_uname")) + return (NULL); + if (a->lookup_uname == NULL) + return (NULL); + return ((*a->lookup_uname)(a->lookup_uname_data, uid)); +} + +int +archive_read_disk_set_gname_lookup(struct archive *_a, + void *private_data, + const char * (*lookup_gname)(void *private, int64_t gid), + void (*cleanup_gname)(void *private)) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + archive_check_magic(&a->archive, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_read_disk_set_gname_lookup"); + + if (a->cleanup_gname != NULL && a->lookup_gname_data != NULL) + (a->cleanup_gname)(a->lookup_gname_data); + + a->lookup_gname = lookup_gname; + a->cleanup_gname = cleanup_gname; + a->lookup_gname_data = private_data; + return (ARCHIVE_OK); +} + +int +archive_read_disk_set_uname_lookup(struct archive *_a, + void *private_data, + const char * (*lookup_uname)(void *private, int64_t uid), + void (*cleanup_uname)(void *private)) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + archive_check_magic(&a->archive, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_read_disk_set_uname_lookup"); + + if (a->cleanup_uname != NULL && a->lookup_uname_data != NULL) + (a->cleanup_uname)(a->lookup_uname_data); + + a->lookup_uname = lookup_uname; + a->cleanup_uname = cleanup_uname; + a->lookup_uname_data = private_data; + return (ARCHIVE_OK); +} + +/* + * Create a new archive_read_disk object and initialize it with global state. + */ +struct archive * +archive_read_disk_new(void) +{ + struct archive_read_disk *a; + + a = (struct archive_read_disk *)malloc(sizeof(*a)); + if (a == NULL) + return (NULL); + memset(a, 0, sizeof(*a)); + a->archive.magic = ARCHIVE_READ_DISK_MAGIC; + a->archive.state = ARCHIVE_STATE_NEW; + a->archive.vtable = archive_read_disk_vtable(); + a->lookup_uname = trivial_lookup_uname; + a->lookup_gname = trivial_lookup_gname; + a->enable_copyfile = 1; + a->traverse_mount_points = 1; + return (&a->archive); +} + +static int +_archive_read_free(struct archive *_a) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + int r; + + if (_a == NULL) + return (ARCHIVE_OK); + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_free"); + + if (a->archive.state != ARCHIVE_STATE_CLOSED) + r = _archive_read_close(&a->archive); + else + r = ARCHIVE_OK; + + tree_free(a->tree); + if (a->cleanup_gname != NULL && a->lookup_gname_data != NULL) + (a->cleanup_gname)(a->lookup_gname_data); + if (a->cleanup_uname != NULL && a->lookup_uname_data != NULL) + (a->cleanup_uname)(a->lookup_uname_data); + archive_string_free(&a->archive.error_string); + a->archive.magic = 0; + free(a); + return (r); +} + +static int +_archive_read_close(struct archive *_a) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_read_close"); + + if (a->archive.state != ARCHIVE_STATE_FATAL) + a->archive.state = ARCHIVE_STATE_CLOSED; + + tree_close(a->tree); + + return (ARCHIVE_OK); +} + +static void +setup_symlink_mode(struct archive_read_disk *a, char symlink_mode, + int follow_symlinks) +{ + a->symlink_mode = symlink_mode; + a->follow_symlinks = follow_symlinks; + if (a->tree != NULL) { + a->tree->initial_symlink_mode = a->symlink_mode; + a->tree->symlink_mode = a->symlink_mode; + } +} + +int +archive_read_disk_set_symlink_logical(struct archive *_a) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_logical"); + setup_symlink_mode(a, 'L', 1); + return (ARCHIVE_OK); +} + +int +archive_read_disk_set_symlink_physical(struct archive *_a) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_physical"); + setup_symlink_mode(a, 'P', 0); + return (ARCHIVE_OK); +} + +int +archive_read_disk_set_symlink_hybrid(struct archive *_a) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_read_disk_set_symlink_hybrid"); + setup_symlink_mode(a, 'H', 1);/* Follow symlinks initially. */ + return (ARCHIVE_OK); +} + +int +archive_read_disk_set_atime_restored(struct archive *_a) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_read_disk_restore_atime"); + a->restore_time = 1; + if (a->tree != NULL) + a->tree->flags |= needsRestoreTimes; + return (ARCHIVE_OK); +} + +int +archive_read_disk_set_behavior(struct archive *_a, int flags) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + int r = ARCHIVE_OK; + + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_read_disk_honor_nodump"); + + if (flags & ARCHIVE_READDISK_RESTORE_ATIME) + r = archive_read_disk_set_atime_restored(_a); + else { + a->restore_time = 0; + if (a->tree != NULL) + a->tree->flags &= ~needsRestoreTimes; + } + if (flags & ARCHIVE_READDISK_HONOR_NODUMP) + a->honor_nodump = 1; + else + a->honor_nodump = 0; + if (flags & ARCHIVE_READDISK_MAC_COPYFILE) + a->enable_copyfile = 1; + else + a->enable_copyfile = 0; + if (flags & ARCHIVE_READDISK_NO_TRAVERSE_MOUNTS) + a->traverse_mount_points = 0; + else + a->traverse_mount_points = 1; + return (r); +} + +/* + * Trivial implementations of gname/uname lookup functions. + * These are normally overridden by the client, but these stub + * versions ensure that we always have something that works. + */ +static const char * +trivial_lookup_gname(void *private_data, int64_t gid) +{ + (void)private_data; /* UNUSED */ + (void)gid; /* UNUSED */ + return (NULL); +} + +static const char * +trivial_lookup_uname(void *private_data, int64_t uid) +{ + (void)private_data; /* UNUSED */ + (void)uid; /* UNUSED */ + return (NULL); +} + +static int64_t +align_num_per_sector(struct tree *t, int64_t size) +{ + int64_t surplus; + + size += t->current_filesystem->bytesPerSector -1; + surplus = size % t->current_filesystem->bytesPerSector; + size -= surplus; + return (size); +} + +static int +start_next_async_read(struct archive_read_disk *a, struct tree *t) +{ + struct la_overlapped *olp; + DWORD buffbytes, rbytes; + + if (t->ol_remaining_bytes == 0) + return (ARCHIVE_EOF); + + olp = &(t->ol[t->ol_idx_doing]); + t->ol_idx_doing = (t->ol_idx_doing + 1) % MAX_OVERLAPPED; + + /* Allocate read buffer. */ + if (olp->buff == NULL) { + void *p; + size_t s = (size_t)align_num_per_sector(t, BUFFER_SIZE); + p = VirtualAlloc(NULL, s, MEM_COMMIT, PAGE_READWRITE); + if (p == NULL) { + archive_set_error(&a->archive, ENOMEM, + "Couldn't allocate memory"); + a->archive.state = ARCHIVE_STATE_FATAL; + return (ARCHIVE_FATAL); + } + olp->buff = p; + olp->buff_size = s; + olp->_a = &a->archive; + olp->ol.hEvent = CreateEventW(NULL, TRUE, FALSE, NULL); + if (olp->ol.hEvent == NULL) { + la_dosmaperr(GetLastError()); + archive_set_error(&a->archive, errno, + "CreateEvent failed"); + a->archive.state = ARCHIVE_STATE_FATAL; + return (ARCHIVE_FATAL); + } + } else + ResetEvent(olp->ol.hEvent); + + buffbytes = (DWORD)olp->buff_size; + if (buffbytes > t->current_sparse->length) + buffbytes = (DWORD)t->current_sparse->length; + + /* Skip hole. */ + if (t->current_sparse->offset > t->ol_total) { + t->ol_remaining_bytes -= + t->current_sparse->offset - t->ol_total; + } + + olp->offset = t->current_sparse->offset; + olp->ol.Offset = (DWORD)(olp->offset & 0xffffffff); + olp->ol.OffsetHigh = (DWORD)(olp->offset >> 32); + + if (t->ol_remaining_bytes > buffbytes) { + olp->bytes_expected = buffbytes; + t->ol_remaining_bytes -= buffbytes; + } else { + olp->bytes_expected = (size_t)t->ol_remaining_bytes; + t->ol_remaining_bytes = 0; + } + olp->bytes_transferred = 0; + t->current_sparse->offset += buffbytes; + t->current_sparse->length -= buffbytes; + t->ol_total = t->current_sparse->offset; + if (t->current_sparse->length == 0 && t->ol_remaining_bytes > 0) + t->current_sparse++; + + if (!ReadFile(t->entry_fh, olp->buff, buffbytes, &rbytes, &(olp->ol))) { + DWORD lasterr; + + lasterr = GetLastError(); + if (lasterr == ERROR_HANDLE_EOF) { + archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, + "Reading file truncated"); + a->archive.state = ARCHIVE_STATE_FATAL; + return (ARCHIVE_FATAL); + } else if (lasterr != ERROR_IO_PENDING) { + if (lasterr == ERROR_NO_DATA) + errno = EAGAIN; + else if (lasterr == ERROR_ACCESS_DENIED) + errno = EBADF; + else + la_dosmaperr(lasterr); + archive_set_error(&a->archive, errno, "Read error"); + a->archive.state = ARCHIVE_STATE_FATAL; + return (ARCHIVE_FATAL); + } + } else + olp->bytes_transferred = rbytes; + t->ol_num_doing++; + + return (t->ol_remaining_bytes == 0)? ARCHIVE_EOF: ARCHIVE_OK; +} + +static void +cancel_async(struct tree *t) +{ + if (t->ol_num_doing != t->ol_num_done) { + CancelIo(t->entry_fh); + t->ol_num_doing = t->ol_num_done = 0; + } +} + +static int +_archive_read_data_block(struct archive *_a, const void **buff, + size_t *size, int64_t *offset) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + struct tree *t = a->tree; + struct la_overlapped *olp; + DWORD bytes_transferred; + int r = ARCHIVE_FATAL; + + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA, + "archive_read_data_block"); + + if (t->entry_eof || t->entry_remaining_bytes <= 0) { + r = ARCHIVE_EOF; + goto abort_read_data; + } + + /* + * Make a request to read the file in asynchronous. + */ + if (t->ol_num_doing == 0) { + do { + r = start_next_async_read(a, t); + if (r == ARCHIVE_FATAL) + goto abort_read_data; + if (!t->async_io) + break; + } while (r == ARCHIVE_OK && t->ol_num_doing < MAX_OVERLAPPED); + } else { + if (start_next_async_read(a, t) == ARCHIVE_FATAL) + goto abort_read_data; + } + + olp = &(t->ol[t->ol_idx_done]); + t->ol_idx_done = (t->ol_idx_done + 1) % MAX_OVERLAPPED; + if (olp->bytes_transferred) + bytes_transferred = (DWORD)olp->bytes_transferred; + else if (!GetOverlappedResult(t->entry_fh, &(olp->ol), + &bytes_transferred, TRUE)) { + la_dosmaperr(GetLastError()); + archive_set_error(&a->archive, errno, + "GetOverlappedResult failed"); + a->archive.state = ARCHIVE_STATE_FATAL; + r = ARCHIVE_FATAL; + goto abort_read_data; + } + t->ol_num_done++; + + if (bytes_transferred == 0 || + olp->bytes_expected != bytes_transferred) { + archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, + "Reading file truncated"); + a->archive.state = ARCHIVE_STATE_FATAL; + r = ARCHIVE_FATAL; + goto abort_read_data; + } + + *buff = olp->buff; + *size = bytes_transferred; + *offset = olp->offset; + if (olp->offset > t->entry_total) + t->entry_remaining_bytes -= olp->offset - t->entry_total; + t->entry_total = olp->offset + *size; + t->entry_remaining_bytes -= *size; + if (t->entry_remaining_bytes == 0) { + /* Close the current file descriptor */ + close_and_restore_time(t->entry_fh, t, &t->restore_time); + t->entry_fh = INVALID_HANDLE_VALUE; + t->entry_eof = 1; + } + return (ARCHIVE_OK); + +abort_read_data: + *buff = NULL; + *size = 0; + *offset = t->entry_total; + if (t->entry_fh != INVALID_HANDLE_VALUE) { + cancel_async(t); + /* Close the current file descriptor */ + close_and_restore_time(t->entry_fh, t, &t->restore_time); + t->entry_fh = INVALID_HANDLE_VALUE; + } + return (r); +} + +static int +next_entry(struct archive_read_disk *a, struct tree *t, + struct archive_entry *entry) +{ + const BY_HANDLE_FILE_INFORMATION *st; + const BY_HANDLE_FILE_INFORMATION *lst; + const char*name; + int descend, r; + + st = NULL; + lst = NULL; + t->descend = 0; + do { + switch (tree_next(t)) { + case TREE_ERROR_FATAL: + archive_set_error(&a->archive, t->tree_errno, + "%ls: Unable to continue traversing directory tree", + tree_current_path(t)); + a->archive.state = ARCHIVE_STATE_FATAL; + return (ARCHIVE_FATAL); + case TREE_ERROR_DIR: + archive_set_error(&a->archive, t->tree_errno, + "%ls: Couldn't visit directory", + tree_current_path(t)); + return (ARCHIVE_FAILED); + case 0: + return (ARCHIVE_EOF); + case TREE_POSTDESCENT: + case TREE_POSTASCENT: + break; + case TREE_REGULAR: + lst = tree_current_lstat(t); + if (lst == NULL) { + archive_set_error(&a->archive, t->tree_errno, + "%ls: Cannot stat", + tree_current_path(t)); + return (ARCHIVE_FAILED); + } + break; + } + } while (lst == NULL); + + archive_entry_copy_pathname_w(entry, tree_current_path(t)); + + /* + * Perform path matching. + */ + if (a->matching) { + r = archive_match_path_excluded(a->matching, entry); + if (r < 0) { + archive_set_error(&(a->archive), errno, + "Faild : %s", archive_error_string(a->matching)); + return (r); + } + if (r) { + if (a->excluded_cb_func) + a->excluded_cb_func(&(a->archive), + a->excluded_cb_data, entry); + return (ARCHIVE_RETRY); + } + } + + /* + * Distinguish 'L'/'P'/'H' symlink following. + */ + switch(t->symlink_mode) { + case 'H': + /* 'H': After the first item, rest like 'P'. */ + t->symlink_mode = 'P'; + /* 'H': First item (from command line) like 'L'. */ + /* FALLTHROUGH */ + case 'L': + /* 'L': Do descend through a symlink to dir. */ + descend = tree_current_is_dir(t); + /* 'L': Follow symlinks to files. */ + a->symlink_mode = 'L'; + a->follow_symlinks = 1; + /* 'L': Archive symlinks as targets, if we can. */ + st = tree_current_stat(t); + if (st != NULL && !tree_target_is_same_as_parent(t, st)) + break; + /* If stat fails, we have a broken symlink; + * in that case, don't follow the link. */ + /* FALLTHROUGH */ + default: + /* 'P': Don't descend through a symlink to dir. */ + descend = tree_current_is_physical_dir(t); + /* 'P': Don't follow symlinks to files. */ + a->symlink_mode = 'P'; + a->follow_symlinks = 0; + /* 'P': Archive symlinks as symlinks. */ + st = lst; + break; + } + + if (update_current_filesystem(a, bhfi_dev(st)) != ARCHIVE_OK) { + a->archive.state = ARCHIVE_STATE_FATAL; + return (ARCHIVE_FATAL); + } + if (t->initial_filesystem_id == -1) + t->initial_filesystem_id = t->current_filesystem_id; + if (!a->traverse_mount_points) { + if (t->initial_filesystem_id != t->current_filesystem_id) + return (ARCHIVE_RETRY); + } + t->descend = descend; + + tree_archive_entry_copy_bhfi(entry, t, st); + + /* Save the times to be restored. This must be in before + * calling archive_read_disk_descend() or any chance of it, + * especially, invokng a callback. */ + t->restore_time.lastWriteTime = st->ftLastWriteTime; + t->restore_time.lastAccessTime = st->ftLastAccessTime; + t->restore_time.filetype = archive_entry_filetype(entry); + + /* + * Perform time matching. + */ + if (a->matching) { + r = archive_match_time_excluded(a->matching, entry); + if (r < 0) { + archive_set_error(&(a->archive), errno, + "Faild : %s", archive_error_string(a->matching)); + return (r); + } + if (r) { + if (a->excluded_cb_func) + a->excluded_cb_func(&(a->archive), + a->excluded_cb_data, entry); + return (ARCHIVE_RETRY); + } + } + + /* Lookup uname/gname */ + name = archive_read_disk_uname(&(a->archive), archive_entry_uid(entry)); + if (name != NULL) + archive_entry_copy_uname(entry, name); + name = archive_read_disk_gname(&(a->archive), archive_entry_gid(entry)); + if (name != NULL) + archive_entry_copy_gname(entry, name); + + /* + * Perform owner matching. + */ + if (a->matching) { + r = archive_match_owner_excluded(a->matching, entry); + if (r < 0) { + archive_set_error(&(a->archive), errno, + "Faild : %s", archive_error_string(a->matching)); + return (r); + } + if (r) { + if (a->excluded_cb_func) + a->excluded_cb_func(&(a->archive), + a->excluded_cb_data, entry); + return (ARCHIVE_RETRY); + } + } + + /* + * Invoke a meta data filter callback. + */ + if (a->metadata_filter_func) { + if (!a->metadata_filter_func(&(a->archive), + a->metadata_filter_data, entry)) + return (ARCHIVE_RETRY); + } + + archive_entry_copy_sourcepath_w(entry, tree_current_access_path(t)); + + r = ARCHIVE_OK; + if (archive_entry_filetype(entry) == AE_IFREG && + archive_entry_size(entry) > 0) { + DWORD flags = FILE_FLAG_BACKUP_SEMANTICS; + if (t->async_io) + flags |= FILE_FLAG_OVERLAPPED; + if (t->direct_io) + flags |= FILE_FLAG_NO_BUFFERING; + else + flags |= FILE_FLAG_SEQUENTIAL_SCAN; + t->entry_fh = CreateFileW(tree_current_access_path(t), + GENERIC_READ, 0, NULL, OPEN_EXISTING, flags, NULL); + if (t->entry_fh == INVALID_HANDLE_VALUE) { + archive_set_error(&a->archive, errno, + "Couldn't open %ls", tree_current_path(a->tree)); + return (ARCHIVE_FAILED); + } + + /* Find sparse data from the disk. */ + if (archive_entry_hardlink(entry) == NULL && + (st->dwFileAttributes & FILE_ATTRIBUTE_SPARSE_FILE) != 0) + r = setup_sparse_from_disk(a, entry, t->entry_fh); + } + return (r); +} + +static int +_archive_read_next_header2(struct archive *_a, struct archive_entry *entry) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + struct tree *t; + int r; + + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, + "archive_read_next_header2"); + + t = a->tree; + if (t->entry_fh != INVALID_HANDLE_VALUE) { + cancel_async(t); + close_and_restore_time(t->entry_fh, t, &t->restore_time); + t->entry_fh = INVALID_HANDLE_VALUE; + } + + while ((r = next_entry(a, t, entry)) == ARCHIVE_RETRY) + archive_entry_clear(entry); + + /* + * EOF and FATAL are persistent at this layer. By + * modifying the state, we guarantee that future calls to + * read a header or read data will fail. + */ + switch (r) { + case ARCHIVE_EOF: + a->archive.state = ARCHIVE_STATE_EOF; + break; + case ARCHIVE_OK: + case ARCHIVE_WARN: + t->entry_total = 0; + if (archive_entry_filetype(entry) == AE_IFREG) { + t->entry_remaining_bytes = archive_entry_size(entry); + t->entry_eof = (t->entry_remaining_bytes == 0)? 1: 0; + if (!t->entry_eof && + setup_sparse(a, entry) != ARCHIVE_OK) + return (ARCHIVE_FATAL); + } else { + t->entry_remaining_bytes = 0; + t->entry_eof = 1; + } + t->ol_idx_doing = t->ol_idx_done = 0; + t->ol_num_doing = t->ol_num_done = 0; + t->ol_remaining_bytes = t->entry_remaining_bytes; + t->ol_total = 0; + a->archive.state = ARCHIVE_STATE_DATA; + break; + case ARCHIVE_RETRY: + break; + case ARCHIVE_FATAL: + a->archive.state = ARCHIVE_STATE_FATAL; + break; + } + + return (r); +} + +static int +setup_sparse(struct archive_read_disk *a, struct archive_entry *entry) +{ + struct tree *t = a->tree; + int64_t aligned, length, offset; + int i; + + t->sparse_count = archive_entry_sparse_reset(entry); + if (t->sparse_count+1 > t->sparse_list_size) { + free(t->sparse_list); + t->sparse_list_size = t->sparse_count + 1; + t->sparse_list = malloc(sizeof(t->sparse_list[0]) * + t->sparse_list_size); + if (t->sparse_list == NULL) { + t->sparse_list_size = 0; + archive_set_error(&a->archive, ENOMEM, + "Can't allocate data"); + a->archive.state = ARCHIVE_STATE_FATAL; + return (ARCHIVE_FATAL); + } + } + /* + * Get sparse list and make sure those offsets and lengths are + * aligned by a sector size. + */ + for (i = 0; i < t->sparse_count; i++) { + archive_entry_sparse_next(entry, &offset, &length); + aligned = align_num_per_sector(t, offset); + if (aligned != offset) { + aligned -= t->current_filesystem->bytesPerSector; + length += offset - aligned; + } + t->sparse_list[i].offset = aligned; + aligned = align_num_per_sector(t, length); + t->sparse_list[i].length = aligned; + } + + aligned = align_num_per_sector(t, archive_entry_size(entry)); + if (i == 0) { + t->sparse_list[i].offset = 0; + t->sparse_list[i].length = aligned; + } else { + int j, last = i; + + t->sparse_list[i].offset = aligned; + t->sparse_list[i].length = 0; + for (i = 0; i < last; i++) { + if ((t->sparse_list[i].offset + + t->sparse_list[i].length) <= + t->sparse_list[i+1].offset) + continue; + /* + * Now sparse_list[i+1] is overlapped by sparse_list[i]. + * Merge those two. + */ + length = t->sparse_list[i+1].offset - + t->sparse_list[i].offset; + t->sparse_list[i+1].offset = t->sparse_list[i].offset; + t->sparse_list[i+1].length += length; + /* Remove sparse_list[i]. */ + for (j = i; j < last; j++) { + t->sparse_list[j].offset = + t->sparse_list[j+1].offset; + t->sparse_list[j].length = + t->sparse_list[j+1].length; + } + last--; + } + } + t->current_sparse = t->sparse_list; + + return (ARCHIVE_OK); +} + +int +archive_read_disk_set_matching(struct archive *_a, struct archive *_ma, + void (*_excluded_func)(struct archive *, void *, struct archive_entry *), + void *_client_data) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_read_disk_set_matching"); + a->matching = _ma; + a->excluded_cb_func = _excluded_func; + a->excluded_cb_data = _client_data; + return (ARCHIVE_OK); +} + +int +archive_read_disk_set_metadata_filter_callback(struct archive *_a, + int (*_metadata_filter_func)(struct archive *, void *, + struct archive_entry *), void *_client_data) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_ANY, + "archive_read_disk_set_metadata_filter_callback"); + + a->metadata_filter_func = _metadata_filter_func; + a->metadata_filter_data = _client_data; + return (ARCHIVE_OK); +} + +int +archive_read_disk_can_descend(struct archive *_a) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + struct tree *t = a->tree; + + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, + "archive_read_disk_can_descend"); + + return (t->visit_type == TREE_REGULAR && t->descend); +} + +/* + * Called by the client to mark the directory just returned from + * tree_next() as needing to be visited. + */ +int +archive_read_disk_descend(struct archive *_a) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + struct tree *t = a->tree; + + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, + "archive_read_disk_descend"); + + if (t->visit_type != TREE_REGULAR || !t->descend) + return (ARCHIVE_OK); + + if (tree_current_is_physical_dir(t)) { + tree_push(t, t->basename, t->full_path.s, + t->current_filesystem_id, + bhfi_dev(&(t->lst)), bhfi_ino(&(t->lst)), + &t->restore_time); + t->stack->flags |= isDir; + } else if (tree_current_is_dir(t)) { + tree_push(t, t->basename, t->full_path.s, + t->current_filesystem_id, + bhfi_dev(&(t->st)), bhfi_ino(&(t->st)), + &t->restore_time); + t->stack->flags |= isDirLink; + } + t->descend = 0; + return (ARCHIVE_OK); +} + +int +archive_read_disk_open(struct archive *_a, const char *pathname) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + struct archive_wstring wpath; + int ret; + + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED, + "archive_read_disk_open"); + archive_clear_error(&a->archive); + + /* Make a wchar_t string from a char string. */ + archive_string_init(&wpath); + if (archive_wstring_append_from_mbs(&wpath, pathname, + strlen(pathname)) != 0) { + if (errno == ENOMEM) + archive_set_error(&a->archive, ENOMEM, + "Can't allocate memory"); + else + archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, + "Can't convert a path to a wchar_t string"); + a->archive.state = ARCHIVE_STATE_FATAL; + ret = ARCHIVE_FATAL; + } else + ret = _archive_read_disk_open_w(_a, wpath.s); + + archive_wstring_free(&wpath); + return (ret); +} + +int +archive_read_disk_open_w(struct archive *_a, const wchar_t *pathname) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, + ARCHIVE_STATE_NEW | ARCHIVE_STATE_CLOSED, + "archive_read_disk_open_w"); + archive_clear_error(&a->archive); + + return (_archive_read_disk_open_w(_a, pathname)); +} + +static int +_archive_read_disk_open_w(struct archive *_a, const wchar_t *pathname) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + + if (a->tree != NULL) + a->tree = tree_reopen(a->tree, pathname, a->restore_time); + else + a->tree = tree_open(pathname, a->symlink_mode, a->restore_time); + if (a->tree == NULL) { + archive_set_error(&a->archive, ENOMEM, + "Can't allocate directory traversal data"); + a->archive.state = ARCHIVE_STATE_FATAL; + return (ARCHIVE_FATAL); + } + a->archive.state = ARCHIVE_STATE_HEADER; + + return (ARCHIVE_OK); +} + +/* + * Return a current filesystem ID which is index of the filesystem entry + * you've visited through archive_read_disk. + */ +int +archive_read_disk_current_filesystem(struct archive *_a) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA, + "archive_read_disk_current_filesystem"); + + return (a->tree->current_filesystem_id); +} + +static int +update_current_filesystem(struct archive_read_disk *a, int64_t dev) +{ + struct tree *t = a->tree; + int i, fid; + + if (t->current_filesystem != NULL && + t->current_filesystem->dev == dev) + return (ARCHIVE_OK); + + for (i = 0; i < t->max_filesystem_id; i++) { + if (t->filesystem_table[i].dev == dev) { + /* There is the filesytem ID we've already generated. */ + t->current_filesystem_id = i; + t->current_filesystem = &(t->filesystem_table[i]); + return (ARCHIVE_OK); + } + } + + /* + * There is a new filesytem, we generate a new ID for. + */ + fid = t->max_filesystem_id++; + if (t->max_filesystem_id > t->allocated_filesytem) { + size_t s; + void *p; + + s = t->max_filesystem_id * 2; + p = realloc(t->filesystem_table, + s * sizeof(*t->filesystem_table)); + if (p == NULL) { + archive_set_error(&a->archive, ENOMEM, + "Can't allocate tar data"); + return (ARCHIVE_FATAL); + } + t->filesystem_table = (struct filesystem *)p; + t->allocated_filesytem = (int)s; + } + t->current_filesystem_id = fid; + t->current_filesystem = &(t->filesystem_table[fid]); + t->current_filesystem->dev = dev; + + return (setup_current_filesystem(a)); +} + +/* + * Returns 1 if current filesystem is generated filesystem, 0 if it is not + * or -1 if it is unknown. + */ +int +archive_read_disk_current_filesystem_is_synthetic(struct archive *_a) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA, + "archive_read_disk_current_filesystem"); + + return (a->tree->current_filesystem->synthetic); +} + +/* + * Returns 1 if current filesystem is remote filesystem, 0 if it is not + * or -1 if it is unknown. + */ +int +archive_read_disk_current_filesystem_is_remote(struct archive *_a) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + + archive_check_magic(_a, ARCHIVE_READ_DISK_MAGIC, ARCHIVE_STATE_DATA, + "archive_read_disk_current_filesystem"); + + return (a->tree->current_filesystem->remote); +} + +/* + * If symlink is broken, statfs or statvfs will fail. + * Use its directory path instead. + */ +static wchar_t * +safe_path_for_statfs(struct tree *t) +{ + const wchar_t *path; + wchar_t *cp, *p = NULL; + + path = tree_current_access_path(t); + if (tree_current_stat(t) == NULL) { + p = _wcsdup(path); + cp = wcsrchr(p, '/'); + if (cp != NULL && wcslen(cp) >= 2) { + cp[1] = '.'; + cp[2] = '\0'; + path = p; + } + } else + p = _wcsdup(path); + return (p); +} + +/* + * Get conditions of synthetic and remote on Windows + */ +static int +setup_current_filesystem(struct archive_read_disk *a) +{ + struct tree *t = a->tree; + wchar_t vol[256]; + wchar_t *path; + + t->current_filesystem->synthetic = -1;/* Not supported */ + path = safe_path_for_statfs(t); + if (!GetVolumePathNameW(path, vol, sizeof(vol)/sizeof(vol[0]))) { + free(path); + t->current_filesystem->remote = -1; + t->current_filesystem->bytesPerSector = 0; + archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, + "GetVolumePathName failed: %d", (int)GetLastError()); + return (ARCHIVE_FAILED); + } + free(path); + switch (GetDriveTypeW(vol)) { + case DRIVE_UNKNOWN: + case DRIVE_NO_ROOT_DIR: + t->current_filesystem->remote = -1; + break; + case DRIVE_REMOTE: + t->current_filesystem->remote = 1; + break; + default: + t->current_filesystem->remote = 0; + break; + } + + if (!GetDiskFreeSpaceW(vol, NULL, + &(t->current_filesystem->bytesPerSector), NULL, NULL)) { + t->current_filesystem->bytesPerSector = 0; + archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, + "GetDiskFreeSpace failed: %d", (int)GetLastError()); + return (ARCHIVE_FAILED); + } + + return (ARCHIVE_OK); +} + +static int +close_and_restore_time(HANDLE h, struct tree *t, struct restore_time *rt) +{ + HANDLE handle; + int r = 0; + + if (h == INVALID_HANDLE_VALUE && AE_IFLNK == rt->filetype) + return (0); + + /* Close a file descritor. + * It will not be used for SetFileTime() because it has been opened + * by a read only mode. + */ + if (h != INVALID_HANDLE_VALUE) + CloseHandle(h); + if ((t->flags & needsRestoreTimes) == 0) + return (r); + + handle = CreateFileW(rt->full_path, FILE_WRITE_ATTRIBUTES, + 0, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); + if (handle == INVALID_HANDLE_VALUE) { + errno = EINVAL; + return (-1); + } + + if (SetFileTime(handle, NULL, &rt->lastAccessTime, + &rt->lastWriteTime) == 0) { + errno = EINVAL; + r = -1; + } else + r = 0; + CloseHandle(handle); + return (r); +} + +/* + * Add a directory path to the current stack. + */ +static void +tree_push(struct tree *t, const wchar_t *path, const wchar_t *full_path, + int filesystem_id, int64_t dev, int64_t ino, struct restore_time *rt) +{ + struct tree_entry *te; + + te = malloc(sizeof(*te)); + memset(te, 0, sizeof(*te)); + te->next = t->stack; + te->parent = t->current; + if (te->parent) + te->depth = te->parent->depth + 1; + t->stack = te; + archive_string_init(&te->name); + archive_wstrcpy(&te->name, path); + archive_string_init(&te->full_path); + archive_wstrcpy(&te->full_path, full_path); + te->flags = needsDescent | needsOpen | needsAscent; + te->filesystem_id = filesystem_id; + te->dev = dev; + te->ino = ino; + te->dirname_length = t->dirname_length; + te->full_path_dir_length = t->full_path_dir_length; + te->restore_time.full_path = te->full_path.s; + if (rt != NULL) { + te->restore_time.lastWriteTime = rt->lastWriteTime; + te->restore_time.lastAccessTime = rt->lastAccessTime; + te->restore_time.filetype = rt->filetype; + } +} + +/* + * Append a name to the current dir path. + */ +static void +tree_append(struct tree *t, const wchar_t *name, size_t name_length) +{ + size_t size_needed; + + t->path.s[t->dirname_length] = L'\0'; + t->path.length = t->dirname_length; + /* Strip trailing '/' from name, unless entire name is "/". */ + while (name_length > 1 && name[name_length - 1] == L'/') + name_length--; + + /* Resize pathname buffer as needed. */ + size_needed = name_length + t->dirname_length + 2; + archive_wstring_ensure(&t->path, size_needed); + /* Add a separating '/' if it's needed. */ + if (t->dirname_length > 0 && + t->path.s[archive_strlen(&t->path)-1] != L'/') + archive_wstrappend_wchar(&t->path, L'/'); + t->basename = t->path.s + archive_strlen(&t->path); + archive_wstrncat(&t->path, name, name_length); + t->restore_time.full_path = t->basename; + if (t->full_path_dir_length > 0) { + t->full_path.s[t->full_path_dir_length] = L'\0'; + t->full_path.length = t->full_path_dir_length; + size_needed = name_length + t->full_path_dir_length + 2; + archive_wstring_ensure(&t->full_path, size_needed); + /* Add a separating '\' if it's needed. */ + if (t->full_path.s[archive_strlen(&t->full_path)-1] != L'\\') + archive_wstrappend_wchar(&t->full_path, L'\\'); + archive_wstrncat(&t->full_path, name, name_length); + t->restore_time.full_path = t->full_path.s; + } +} + +/* + * Open a directory tree for traversal. + */ +static struct tree * +tree_open(const wchar_t *path, int symlink_mode, int restore_time) +{ + struct tree *t; + + t = malloc(sizeof(*t)); + memset(t, 0, sizeof(*t)); + archive_string_init(&(t->full_path)); + archive_string_init(&t->path); + archive_wstring_ensure(&t->path, 15); + t->initial_symlink_mode = symlink_mode; + return (tree_reopen(t, path, restore_time)); +} + +static struct tree * +tree_reopen(struct tree *t, const wchar_t *path, int restore_time) +{ + struct archive_wstring ws; + wchar_t *pathname, *p, *base; + + t->flags = (restore_time)?needsRestoreTimes:0; + t->visit_type = 0; + t->tree_errno = 0; + t->full_path_dir_length = 0; + t->dirname_length = 0; + t->depth = 0; + t->descend = 0; + t->current = NULL; + t->d = INVALID_HANDLE_VALUE; + t->symlink_mode = t->initial_symlink_mode; + archive_string_empty(&(t->full_path)); + archive_string_empty(&t->path); + t->entry_fh = INVALID_HANDLE_VALUE; + t->entry_eof = 0; + t->entry_remaining_bytes = 0; + t->initial_filesystem_id = -1; + + /* Get wchar_t strings from char strings. */ + archive_string_init(&ws); + archive_wstrcpy(&ws, path); + pathname = ws.s; + /* Get a full-path-name. */ + p = __la_win_permissive_name_w(pathname); + if (p == NULL) + goto failed; + archive_wstrcpy(&(t->full_path), p); + free(p); + + /* Convert path separators from '\' to '/' */ + for (p = pathname; *p != L'\0'; ++p) { + if (*p == L'\\') + *p = L'/'; + } + base = pathname; + + /* First item is set up a lot like a symlink traversal. */ + /* printf("Looking for wildcard in %s\n", path); */ + if ((base[0] == L'/' && base[1] == L'/' && + base[2] == L'?' && base[3] == L'/' && + (wcschr(base+4, L'*') || wcschr(base+4, L'?'))) || + (!(base[0] == L'/' && base[1] == L'/' && + base[2] == L'?' && base[3] == L'/') && + (wcschr(base, L'*') || wcschr(base, L'?')))) { + // It has a wildcard in it... + // Separate the last element. + p = wcsrchr(base, L'/'); + if (p != NULL) { + *p = L'\0'; + tree_append(t, base, p - base); + t->dirname_length = archive_strlen(&t->path); + base = p + 1; + } + p = wcsrchr(t->full_path.s, L'\\'); + if (p != NULL) { + *p = L'\0'; + t->full_path.length = wcslen(t->full_path.s); + t->full_path_dir_length = archive_strlen(&t->full_path); + } + } + tree_push(t, base, t->full_path.s, 0, 0, 0, NULL); + archive_wstring_free(&ws); + t->stack->flags = needsFirstVisit; + /* + * Debug flag for Direct IO(No buffering) or Async IO. + * Those dependant on environment variable switches + * will be removed until next release. + */ + { + const char *e; + if ((e = getenv("LIBARCHIVE_DIRECT_IO")) != NULL) { + if (e[0] == '0') + t->direct_io = 0; + else + t->direct_io = 1; + fprintf(stderr, "LIBARCHIVE_DIRECT_IO=%s\n", + (t->direct_io)?"Enabled":"Disabled"); + } else + t->direct_io = DIRECT_IO; + if ((e = getenv("LIBARCHIVE_ASYNC_IO")) != NULL) { + if (e[0] == '0') + t->async_io = 0; + else + t->async_io = 1; + fprintf(stderr, "LIBARCHIVE_ASYNC_IO=%s\n", + (t->async_io)?"Enabled":"Disabled"); + } else + t->async_io = ASYNC_IO; + } + return (t); +failed: + archive_wstring_free(&ws); + tree_free(t); + return (NULL); +} + +static int +tree_descent(struct tree *t) +{ + t->dirname_length = archive_strlen(&t->path); + t->full_path_dir_length = archive_strlen(&t->full_path); + t->depth++; + return (0); +} + +/* + * We've finished a directory; ascend back to the parent. + */ +static int +tree_ascend(struct tree *t) +{ + struct tree_entry *te; + + te = t->stack; + t->depth--; + close_and_restore_time(INVALID_HANDLE_VALUE, t, &te->restore_time); + return (0); +} + +/* + * Pop the working stack. + */ +static void +tree_pop(struct tree *t) +{ + struct tree_entry *te; + + t->full_path.s[t->full_path_dir_length] = L'\0'; + t->full_path.length = t->full_path_dir_length; + t->path.s[t->dirname_length] = L'\0'; + t->path.length = t->dirname_length; + if (t->stack == t->current && t->current != NULL) + t->current = t->current->parent; + te = t->stack; + t->stack = te->next; + t->dirname_length = te->dirname_length; + t->basename = t->path.s + t->dirname_length; + t->full_path_dir_length = te->full_path_dir_length; + while (t->basename[0] == L'/') + t->basename++; + archive_wstring_free(&te->name); + archive_wstring_free(&te->full_path); + free(te); +} + +/* + * Get the next item in the tree traversal. + */ +static int +tree_next(struct tree *t) +{ + int r; + + while (t->stack != NULL) { + /* If there's an open dir, get the next entry from there. */ + if (t->d != INVALID_HANDLE_VALUE) { + r = tree_dir_next_windows(t, NULL); + if (r == 0) + continue; + return (r); + } + + if (t->stack->flags & needsFirstVisit) { + wchar_t *d = t->stack->name.s; + t->stack->flags &= ~needsFirstVisit; + if (!(d[0] == L'/' && d[1] == L'/' && + d[2] == L'?' && d[3] == L'/') && + (wcschr(d, L'*') || wcschr(d, L'?'))) { + r = tree_dir_next_windows(t, d); + if (r == 0) + continue; + return (r); + } else { + HANDLE h = FindFirstFileW(d, &t->_findData); + if (h == INVALID_HANDLE_VALUE) { + la_dosmaperr(GetLastError()); + t->tree_errno = errno; + t->visit_type = TREE_ERROR_DIR; + return (t->visit_type); + } + t->findData = &t->_findData; + FindClose(h); + } + /* Top stack item needs a regular visit. */ + t->current = t->stack; + tree_append(t, t->stack->name.s, + archive_strlen(&(t->stack->name))); + //t->dirname_length = t->path_length; + //tree_pop(t); + t->stack->flags &= ~needsFirstVisit; + return (t->visit_type = TREE_REGULAR); + } else if (t->stack->flags & needsDescent) { + /* Top stack item is dir to descend into. */ + t->current = t->stack; + tree_append(t, t->stack->name.s, + archive_strlen(&(t->stack->name))); + t->stack->flags &= ~needsDescent; + r = tree_descent(t); + if (r != 0) { + tree_pop(t); + t->visit_type = r; + } else + t->visit_type = TREE_POSTDESCENT; + return (t->visit_type); + } else if (t->stack->flags & needsOpen) { + t->stack->flags &= ~needsOpen; + r = tree_dir_next_windows(t, L"*"); + if (r == 0) + continue; + return (r); + } else if (t->stack->flags & needsAscent) { + /* Top stack item is dir and we're done with it. */ + r = tree_ascend(t); + tree_pop(t); + t->visit_type = r != 0 ? r : TREE_POSTASCENT; + return (t->visit_type); + } else { + /* Top item on stack is dead. */ + tree_pop(t); + t->flags &= ~hasLstat; + t->flags &= ~hasStat; + } + } + return (t->visit_type = 0); +} + +static int +tree_dir_next_windows(struct tree *t, const wchar_t *pattern) +{ + const wchar_t *name; + size_t namelen; + int r; + + for (;;) { + if (pattern != NULL) { + struct archive_wstring pt; + + archive_string_init(&pt); + archive_wstring_ensure(&pt, + archive_strlen(&(t->full_path)) + + 2 + wcslen(pattern)); + archive_wstring_copy(&pt, &(t->full_path)); + archive_wstrappend_wchar(&pt, L'\\'); + archive_wstrcat(&pt, pattern); + t->d = FindFirstFileW(pt.s, &t->_findData); + archive_wstring_free(&pt); + if (t->d == INVALID_HANDLE_VALUE) { + la_dosmaperr(GetLastError()); + t->tree_errno = errno; + r = tree_ascend(t); /* Undo "chdir" */ + tree_pop(t); + t->visit_type = r != 0 ? r : TREE_ERROR_DIR; + return (t->visit_type); + } + t->findData = &t->_findData; + pattern = NULL; + } else if (!FindNextFileW(t->d, &t->_findData)) { + FindClose(t->d); + t->d = INVALID_HANDLE_VALUE; + t->findData = NULL; + return (0); + } + name = t->findData->cFileName; + namelen = wcslen(name); + t->flags &= ~hasLstat; + t->flags &= ~hasStat; + if (name[0] == L'.' && name[1] == L'\0') + continue; + if (name[0] == L'.' && name[1] == L'.' && name[2] == L'\0') + continue; + tree_append(t, name, namelen); + return (t->visit_type = TREE_REGULAR); + } +} + +#define EPOC_TIME ARCHIVE_LITERAL_ULL(116444736000000000) +static void +fileTimeToUtc(const FILETIME *filetime, time_t *t, long *ns) +{ + ULARGE_INTEGER utc; + + utc.HighPart = filetime->dwHighDateTime; + utc.LowPart = filetime->dwLowDateTime; + if (utc.QuadPart >= EPOC_TIME) { + utc.QuadPart -= EPOC_TIME; + /* milli seconds base */ + *t = (time_t)(utc.QuadPart / 10000000); + /* nano seconds base */ + *ns = (long)(utc.QuadPart % 10000000) * 100; + } else { + *t = 0; + *ns = 0; + } +} + +static void +entry_copy_bhfi(struct archive_entry *entry, const wchar_t *path, + const WIN32_FIND_DATAW *findData, + const BY_HANDLE_FILE_INFORMATION *bhfi) +{ + time_t secs; + long nsecs; + mode_t mode; + + fileTimeToUtc(&bhfi->ftLastAccessTime, &secs, &nsecs); + archive_entry_set_atime(entry, secs, nsecs); + fileTimeToUtc(&bhfi->ftLastWriteTime, &secs, &nsecs); + archive_entry_set_mtime(entry, secs, nsecs); + fileTimeToUtc(&bhfi->ftCreationTime, &secs, &nsecs); + archive_entry_set_birthtime(entry, secs, nsecs); + archive_entry_set_ctime(entry, secs, nsecs); + archive_entry_set_dev(entry, bhfi_dev(bhfi)); + archive_entry_set_ino64(entry, bhfi_ino(bhfi)); + if (bhfi->dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) + archive_entry_set_nlink(entry, bhfi->nNumberOfLinks + 1); + else + archive_entry_set_nlink(entry, bhfi->nNumberOfLinks); + archive_entry_set_size(entry, + (((int64_t)bhfi->nFileSizeHigh) << 32) + + bhfi->nFileSizeLow); + archive_entry_set_uid(entry, 0); + archive_entry_set_gid(entry, 0); + archive_entry_set_rdev(entry, 0); + + mode = S_IRUSR | S_IRGRP | S_IROTH; + if ((bhfi->dwFileAttributes & FILE_ATTRIBUTE_READONLY) == 0) + mode |= S_IWUSR | S_IWGRP | S_IWOTH; + if ((bhfi->dwFileAttributes & FILE_ATTRIBUTE_REPARSE_POINT) && + findData != NULL && + findData->dwReserved0 == IO_REPARSE_TAG_SYMLINK) + mode |= S_IFLNK; + else if (bhfi->dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) + mode |= S_IFDIR | S_IXUSR | S_IXGRP | S_IXOTH; + else { + const wchar_t *p; + + mode |= S_IFREG; + p = wcsrchr(path, L'.'); + if (p != NULL && wcslen(p) == 4) { + switch (p[1]) { + case L'B': case L'b': + if ((p[2] == L'A' || p[2] == L'a' ) && + (p[3] == L'T' || p[3] == L't' )) + mode |= S_IXUSR | S_IXGRP | S_IXOTH; + break; + case L'C': case L'c': + if (((p[2] == L'M' || p[2] == L'm' ) && + (p[3] == L'D' || p[3] == L'd' )) || + ((p[2] == L'M' || p[2] == L'm' ) && + (p[3] == L'D' || p[3] == L'd' ))) + mode |= S_IXUSR | S_IXGRP | S_IXOTH; + break; + case L'E': case L'e': + if ((p[2] == L'X' || p[2] == L'x' ) && + (p[3] == L'E' || p[3] == L'e' )) + mode |= S_IXUSR | S_IXGRP | S_IXOTH; + break; + default: + break; + } + } + } + archive_entry_set_mode(entry, mode); +} + +static void +tree_archive_entry_copy_bhfi(struct archive_entry *entry, struct tree *t, + const BY_HANDLE_FILE_INFORMATION *bhfi) +{ + entry_copy_bhfi(entry, tree_current_path(t), t->findData, bhfi); +} + +static int +tree_current_file_information(struct tree *t, BY_HANDLE_FILE_INFORMATION *st, + int sim_lstat) +{ + HANDLE h; + int r; + DWORD flag = FILE_FLAG_BACKUP_SEMANTICS; + + if (sim_lstat && tree_current_is_physical_link(t)) + flag |= FILE_FLAG_OPEN_REPARSE_POINT; + h = CreateFileW(tree_current_access_path(t), 0, 0, NULL, + OPEN_EXISTING, flag, NULL); + if (h == INVALID_HANDLE_VALUE) { + la_dosmaperr(GetLastError()); + t->tree_errno = errno; + return (0); + } + r = GetFileInformationByHandle(h, st); + CloseHandle(h); + return (r); +} + +/* + * Get the stat() data for the entry just returned from tree_next(). + */ +static const BY_HANDLE_FILE_INFORMATION * +tree_current_stat(struct tree *t) +{ + if (!(t->flags & hasStat)) { + if (!tree_current_file_information(t, &t->st, 0)) + return NULL; + t->flags |= hasStat; + } + return (&t->st); +} + +/* + * Get the lstat() data for the entry just returned from tree_next(). + */ +static const BY_HANDLE_FILE_INFORMATION * +tree_current_lstat(struct tree *t) +{ + if (!(t->flags & hasLstat)) { + if (!tree_current_file_information(t, &t->lst, 1)) + return NULL; + t->flags |= hasLstat; + } + return (&t->lst); +} + +/* + * Test whether current entry is a dir or link to a dir. + */ +static int +tree_current_is_dir(struct tree *t) +{ + if (t->findData) + return (t->findData->dwFileAttributes + & FILE_ATTRIBUTE_DIRECTORY); + return (0); +} + +/* + * Test whether current entry is a physical directory. Usually, we + * already have at least one of stat() or lstat() in memory, so we + * use tricks to try to avoid an extra trip to the disk. + */ +static int +tree_current_is_physical_dir(struct tree *t) +{ + if (tree_current_is_physical_link(t)) + return (0); + return (tree_current_is_dir(t)); +} + +/* + * Test whether current entry is a symbolic link. + */ +static int +tree_current_is_physical_link(struct tree *t) +{ + if (t->findData) + return ((t->findData->dwFileAttributes + & FILE_ATTRIBUTE_REPARSE_POINT) && + (t->findData->dwReserved0 + == IO_REPARSE_TAG_SYMLINK)); + return (0); +} + +/* + * Test whether the same file has been in the tree as its parent. + */ +static int +tree_target_is_same_as_parent(struct tree *t, + const BY_HANDLE_FILE_INFORMATION *st) +{ + struct tree_entry *te; + int64_t dev = bhfi_dev(st); + int64_t ino = bhfi_ino(st); + + for (te = t->current->parent; te != NULL; te = te->parent) { + if (te->dev == dev && te->ino == ino) + return (1); + } + return (0); +} + +/* + * Return the access path for the entry just returned from tree_next(). + */ +static const wchar_t * +tree_current_access_path(struct tree *t) +{ + return (t->full_path.s); +} + +/* + * Return the full path for the entry just returned from tree_next(). + */ +static const wchar_t * +tree_current_path(struct tree *t) +{ + return (t->path.s); +} + +/* + * Terminate the traversal. + */ +static void +tree_close(struct tree *t) +{ + + if (t == NULL) + return; + if (t->entry_fh != INVALID_HANDLE_VALUE) { + cancel_async(t); + close_and_restore_time(t->entry_fh, t, &t->restore_time); + t->entry_fh = INVALID_HANDLE_VALUE; + } + /* Close the handle of FindFirstFileW */ + if (t->d != INVALID_HANDLE_VALUE) { + FindClose(t->d); + t->d = INVALID_HANDLE_VALUE; + t->findData = NULL; + } + /* Release anything remaining in the stack. */ + while (t->stack != NULL) + tree_pop(t); +} + +/* + * Release any resources. + */ +static void +tree_free(struct tree *t) +{ + int i; + + if (t == NULL) + return; + archive_wstring_free(&t->path); + archive_wstring_free(&t->full_path); + free(t->sparse_list); + free(t->filesystem_table); + for (i = 0; i < MAX_OVERLAPPED; i++) { + if (t->ol[i].buff) + VirtualFree(t->ol[i].buff, 0, MEM_RELEASE); + CloseHandle(t->ol[i].ol.hEvent); + } + free(t); +} + + +/* + * Populate the archive_entry with metadata from the disk. + */ +int +archive_read_disk_entry_from_file(struct archive *_a, + struct archive_entry *entry, int fd, const struct stat *st) +{ + struct archive_read_disk *a = (struct archive_read_disk *)_a; + const wchar_t *path; + const wchar_t *wname; + const char *name; + HANDLE h; + BY_HANDLE_FILE_INFORMATION bhfi; + DWORD fileAttributes = 0; + int r; + + archive_clear_error(_a); + wname = archive_entry_sourcepath_w(entry); + if (wname == NULL) + wname = archive_entry_pathname_w(entry); + if (wname == NULL) { + archive_set_error(&a->archive, EINVAL, + "Can't get a wide character version of the path"); + return (ARCHIVE_FAILED); + } + path = __la_win_permissive_name_w(wname); + + if (st == NULL) { + /* + * Get metadata through GetFileInformationByHandle(). + */ + if (fd >= 0) { + h = (HANDLE)_get_osfhandle(fd); + r = GetFileInformationByHandle(h, &bhfi); + if (r == 0) { + la_dosmaperr(GetLastError()); + archive_set_error(&a->archive, errno, + "Can't GetFileInformationByHandle"); + return (ARCHIVE_FAILED); + } + entry_copy_bhfi(entry, path, NULL, &bhfi); + } else { + WIN32_FIND_DATAW findData; + DWORD flag, desiredAccess; + + h = FindFirstFileW(path, &findData); + if (h == INVALID_HANDLE_VALUE) { + la_dosmaperr(GetLastError()); + archive_set_error(&a->archive, errno, + "Can't FindFirstFileW"); + return (ARCHIVE_FAILED); + } + FindClose(h); + + flag = FILE_FLAG_BACKUP_SEMANTICS; + if (!a->follow_symlinks && + (findData.dwFileAttributes + & FILE_ATTRIBUTE_REPARSE_POINT) && + (findData.dwReserved0 == IO_REPARSE_TAG_SYMLINK)) { + flag |= FILE_FLAG_OPEN_REPARSE_POINT; + desiredAccess = 0; + } else if (findData.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) { + desiredAccess = 0; + } else + desiredAccess = GENERIC_READ; + + h = CreateFileW(path, desiredAccess, 0, NULL, + OPEN_EXISTING, flag, NULL); + if (h == INVALID_HANDLE_VALUE) { + la_dosmaperr(GetLastError()); + archive_set_error(&a->archive, errno, + "Can't CreateFileW"); + return (ARCHIVE_FAILED); + } + r = GetFileInformationByHandle(h, &bhfi); + if (r == 0) { + la_dosmaperr(GetLastError()); + archive_set_error(&a->archive, errno, + "Can't GetFileInformationByHandle"); + CloseHandle(h); + return (ARCHIVE_FAILED); + } + entry_copy_bhfi(entry, path, &findData, &bhfi); + } + fileAttributes = bhfi.dwFileAttributes; + } else { + archive_entry_copy_stat(entry, st); + h = INVALID_HANDLE_VALUE; + } + + /* Lookup uname/gname */ + name = archive_read_disk_uname(_a, archive_entry_uid(entry)); + if (name != NULL) + archive_entry_copy_uname(entry, name); + name = archive_read_disk_gname(_a, archive_entry_gid(entry)); + if (name != NULL) + archive_entry_copy_gname(entry, name); + + /* + * Can this file be sparse file ? + */ + if (archive_entry_filetype(entry) != AE_IFREG + || archive_entry_size(entry) <= 0 + || archive_entry_hardlink(entry) != NULL) { + if (h != INVALID_HANDLE_VALUE && fd < 0) + CloseHandle(h); + return (ARCHIVE_OK); + } + + if (h == INVALID_HANDLE_VALUE) { + if (fd >= 0) { + h = (HANDLE)_get_osfhandle(fd); + } else { + h = CreateFileW(path, GENERIC_READ, 0, NULL, + OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); + if (h == INVALID_HANDLE_VALUE) { + la_dosmaperr(GetLastError()); + archive_set_error(&a->archive, errno, + "Can't CreateFileW"); + return (ARCHIVE_FAILED); + } + } + r = GetFileInformationByHandle(h, &bhfi); + if (r == 0) { + la_dosmaperr(GetLastError()); + archive_set_error(&a->archive, errno, + "Can't GetFileInformationByHandle"); + if (h != INVALID_HANDLE_VALUE && fd < 0) + CloseHandle(h); + return (ARCHIVE_FAILED); + } + fileAttributes = bhfi.dwFileAttributes; + } + + /* Sparse file must be set a mark, FILE_ATTRIBUTE_SPARSE_FILE */ + if ((fileAttributes & FILE_ATTRIBUTE_SPARSE_FILE) == 0) { + if (fd < 0) + CloseHandle(h); + return (ARCHIVE_OK); + } + + r = setup_sparse_from_disk(a, entry, h); + if (fd < 0) + CloseHandle(h); + + return (r); +} + +/* + * Windows sparse interface. + */ +#if defined(__MINGW32__) && !defined(FSCTL_QUERY_ALLOCATED_RANGES) +#define FSCTL_QUERY_ALLOCATED_RANGES 0x940CF +typedef struct { + LARGE_INTEGER FileOffset; + LARGE_INTEGER Length; +} FILE_ALLOCATED_RANGE_BUFFER; +#endif + +static int +setup_sparse_from_disk(struct archive_read_disk *a, + struct archive_entry *entry, HANDLE handle) +{ + FILE_ALLOCATED_RANGE_BUFFER range, *outranges = NULL; + size_t outranges_size; + int64_t entry_size = archive_entry_size(entry); + int exit_sts = ARCHIVE_OK; + + range.FileOffset.QuadPart = 0; + range.Length.QuadPart = entry_size; + outranges_size = 2048; + outranges = (FILE_ALLOCATED_RANGE_BUFFER *)malloc(outranges_size); + if (outranges == NULL) { + archive_set_error(&a->archive, ENOMEM, + "Couldn't allocate memory"); + exit_sts = ARCHIVE_FATAL; + goto exit_setup_sparse; + } + + for (;;) { + DWORD retbytes; + BOOL ret; + + for (;;) { + ret = DeviceIoControl(handle, + FSCTL_QUERY_ALLOCATED_RANGES, + &range, sizeof(range), outranges, + (DWORD)outranges_size, &retbytes, NULL); + if (ret == 0 && GetLastError() == ERROR_MORE_DATA) { + free(outranges); + outranges_size *= 2; + outranges = (FILE_ALLOCATED_RANGE_BUFFER *) + malloc(outranges_size); + if (outranges == NULL) { + archive_set_error(&a->archive, ENOMEM, + "Couldn't allocate memory"); + exit_sts = ARCHIVE_FATAL; + goto exit_setup_sparse; + } + continue; + } else + break; + } + if (ret != 0) { + if (retbytes > 0) { + DWORD i, n; + + n = retbytes / sizeof(outranges[0]); + if (n == 1 && + outranges[0].FileOffset.QuadPart == 0 && + outranges[0].Length.QuadPart == entry_size) + break;/* This is not sparse. */ + for (i = 0; i < n; i++) + archive_entry_sparse_add_entry(entry, + outranges[i].FileOffset.QuadPart, + outranges[i].Length.QuadPart); + range.FileOffset.QuadPart = + outranges[n-1].FileOffset.QuadPart + + outranges[n-1].Length.QuadPart; + range.Length.QuadPart = + entry_size - range.FileOffset.QuadPart; + if (range.Length.QuadPart > 0) + continue; + } else { + /* The remaining data is hole. */ + archive_entry_sparse_add_entry(entry, + range.FileOffset.QuadPart, + range.Length.QuadPart); + } + break; + } else { + la_dosmaperr(GetLastError()); + archive_set_error(&a->archive, errno, + "DeviceIoControl Failed: %lu", GetLastError()); + exit_sts = ARCHIVE_FAILED; + goto exit_setup_sparse; + } + } +exit_setup_sparse: + free(outranges); + + return (exit_sts); +} + +#endif diff --git a/libarchive/archive_windows.c b/libarchive/archive_windows.c new file mode 100644 index 000000000000..d3bf758bb39e --- /dev/null +++ b/libarchive/archive_windows.c @@ -0,0 +1,908 @@ +/*- + * Copyright (c) 2009-2011 Michihiro NAKAJIMA + * Copyright (c) 2003-2007 Kees Zeelenberg + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + * $FreeBSD$ + */ + +/* + * A set of compatibility glue for building libarchive on Windows platforms. + * + * Originally created as "libarchive-nonposix.c" by Kees Zeelenberg + * for the GnuWin32 project, trimmed significantly by Tim Kientzle. + * + * Much of the original file was unnecessary for libarchive, because + * many of the features it emulated were not strictly necessary for + * libarchive. I hope for this to shrink further as libarchive + * internals are gradually reworked to sit more naturally on both + * POSIX and Windows. Any ideas for this are greatly appreciated. + * + * The biggest remaining issue is the dev/ino emulation; libarchive + * has a couple of public APIs that rely on dev/ino uniquely + * identifying a file. This doesn't match well with Windows. I'm + * considering alternative APIs. + */ + +#if defined(_WIN32) && !defined(__CYGWIN__) + +#include "archive_platform.h" +#include "archive_private.h" +#include "archive_entry.h" +#include <ctype.h> +#include <errno.h> +#include <stddef.h> +#ifdef HAVE_SYS_UTIME_H +#include <sys/utime.h> +#endif +#include <sys/stat.h> +#include <locale.h> +#include <process.h> +#include <stdlib.h> +#include <wchar.h> +#include <windows.h> +#include <share.h> + +#define EPOC_TIME ARCHIVE_LITERAL_ULL(116444736000000000) + +#if defined(__LA_LSEEK_NEEDED) +static BOOL SetFilePointerEx_perso(HANDLE hFile, + LARGE_INTEGER liDistanceToMove, + PLARGE_INTEGER lpNewFilePointer, + DWORD dwMoveMethod) +{ + LARGE_INTEGER li; + li.QuadPart = liDistanceToMove.QuadPart; + li.LowPart = SetFilePointer( + hFile, li.LowPart, &li.HighPart, dwMoveMethod); + if(lpNewFilePointer) { + lpNewFilePointer->QuadPart = li.QuadPart; + } + return li.LowPart != -1 || GetLastError() == NO_ERROR; +} +#endif + +struct ustat { + int64_t st_atime; + uint32_t st_atime_nsec; + int64_t st_ctime; + uint32_t st_ctime_nsec; + int64_t st_mtime; + uint32_t st_mtime_nsec; + gid_t st_gid; + /* 64bits ino */ + int64_t st_ino; + mode_t st_mode; + uint32_t st_nlink; + uint64_t st_size; + uid_t st_uid; + dev_t st_dev; + dev_t st_rdev; +}; + +/* Transform 64-bits ino into 32-bits by hashing. + * You do not forget that really unique number size is 64-bits. + */ +#define INOSIZE (8*sizeof(ino_t)) /* 32 */ +static __inline ino_t +getino(struct ustat *ub) +{ + ULARGE_INTEGER ino64; + ino64.QuadPart = ub->st_ino; + /* I don't know this hashing is correct way */ + return ((ino_t)(ino64.LowPart ^ (ino64.LowPart >> INOSIZE))); +} + +/* + * Prepend "\\?\" to the path name and convert it to unicode to permit + * an extended-length path for a maximum total path length of 32767 + * characters. + * see also http://msdn.microsoft.com/en-us/library/aa365247.aspx + */ +wchar_t * +__la_win_permissive_name(const char *name) +{ + wchar_t *wn; + wchar_t *ws; + size_t ll; + + ll = strlen(name); + wn = malloc((ll + 1) * sizeof(wchar_t)); + if (wn == NULL) + return (NULL); + ll = mbstowcs(wn, name, ll); + if (ll == (size_t)-1) { + free(wn); + return (NULL); + } + wn[ll] = L'\0'; + ws = __la_win_permissive_name_w(wn); + free(wn); + return (ws); +} + +wchar_t * +__la_win_permissive_name_w(const wchar_t *wname) +{ + wchar_t *wn, *wnp; + wchar_t *ws, *wsp; + DWORD l, len, slen; + int unc; + + /* Get a full-pathname. */ + l = GetFullPathNameW(wname, 0, NULL, NULL); + if (l == 0) + return (NULL); + /* NOTE: GetFullPathNameW has a bug that if the length of the file + * name is just 1 then it returns incomplete buffer size. Thus, we + * have to add three to the size to allocate a sufficient buffer + * size for the full-pathname of the file name. */ + l += 3; + wnp = malloc(l * sizeof(wchar_t)); + if (wnp == NULL) + return (NULL); + len = GetFullPathNameW(wname, l, wnp, NULL); + wn = wnp; + + if (wnp[0] == L'\\' && wnp[1] == L'\\' && + wnp[2] == L'?' && wnp[3] == L'\\') + /* We have already a permissive name. */ + return (wn); + + if (wnp[0] == L'\\' && wnp[1] == L'\\' && + wnp[2] == L'.' && wnp[3] == L'\\') { + /* This is a device name */ + if (((wnp[4] >= L'a' && wnp[4] <= L'z') || + (wnp[4] >= L'A' && wnp[4] <= L'Z')) && + wnp[5] == L':' && wnp[6] == L'\\') + wnp[2] = L'?';/* Not device name. */ + return (wn); + } + + unc = 0; + if (wnp[0] == L'\\' && wnp[1] == L'\\' && wnp[2] != L'\\') { + wchar_t *p = &wnp[2]; + + /* Skip server-name letters. */ + while (*p != L'\\' && *p != L'\0') + ++p; + if (*p == L'\\') { + wchar_t *rp = ++p; + /* Skip share-name letters. */ + while (*p != L'\\' && *p != L'\0') + ++p; + if (*p == L'\\' && p != rp) { + /* Now, match patterns such as + * "\\server-name\share-name\" */ + wnp += 2; + len -= 2; + unc = 1; + } + } + } + + slen = 4 + (unc * 4) + len + 1; + ws = wsp = malloc(slen * sizeof(wchar_t)); + if (ws == NULL) { + free(wn); + return (NULL); + } + /* prepend "\\?\" */ + wcsncpy(wsp, L"\\\\?\\", 4); + wsp += 4; + slen -= 4; + if (unc) { + /* append "UNC\" ---> "\\?\UNC\" */ + wcsncpy(wsp, L"UNC\\", 4); + wsp += 4; + slen -= 4; + } + wcsncpy(wsp, wnp, slen); + wsp[slen - 1] = L'\0'; /* Ensure null termination. */ + free(wn); + return (ws); +} + +/* + * Create a file handle. + * This can exceed MAX_PATH limitation. + */ +static HANDLE +la_CreateFile(const char *path, DWORD dwDesiredAccess, DWORD dwShareMode, + LPSECURITY_ATTRIBUTES lpSecurityAttributes, DWORD dwCreationDisposition, + DWORD dwFlagsAndAttributes, HANDLE hTemplateFile) +{ + wchar_t *wpath; + HANDLE handle; + + handle = CreateFileA(path, dwDesiredAccess, dwShareMode, + lpSecurityAttributes, dwCreationDisposition, dwFlagsAndAttributes, + hTemplateFile); + if (handle != INVALID_HANDLE_VALUE) + return (handle); + if (GetLastError() != ERROR_PATH_NOT_FOUND) + return (handle); + wpath = __la_win_permissive_name(path); + if (wpath == NULL) + return (handle); + handle = CreateFileW(wpath, dwDesiredAccess, dwShareMode, + lpSecurityAttributes, dwCreationDisposition, dwFlagsAndAttributes, + hTemplateFile); + free(wpath); + return (handle); +} + +#if defined(__LA_LSEEK_NEEDED) +__int64 +__la_lseek(int fd, __int64 offset, int whence) +{ + LARGE_INTEGER distance; + LARGE_INTEGER newpointer; + HANDLE handle; + + if (fd < 0) { + errno = EBADF; + return (-1); + } + handle = (HANDLE)_get_osfhandle(fd); + if (GetFileType(handle) != FILE_TYPE_DISK) { + errno = EBADF; + return (-1); + } + distance.QuadPart = offset; + if (!SetFilePointerEx_perso(handle, distance, &newpointer, whence)) { + DWORD lasterr; + + lasterr = GetLastError(); + if (lasterr == ERROR_BROKEN_PIPE) + return (0); + if (lasterr == ERROR_ACCESS_DENIED) + errno = EBADF; + else + la_dosmaperr(lasterr); + return (-1); + } + return (newpointer.QuadPart); +} +#endif + +/* This can exceed MAX_PATH limitation. */ +int +__la_open(const char *path, int flags, ...) +{ + va_list ap; + wchar_t *ws; + int r, pmode; + DWORD attr; + + va_start(ap, flags); + pmode = va_arg(ap, int); + va_end(ap); + ws = NULL; + if ((flags & ~O_BINARY) == O_RDONLY) { + /* + * When we open a directory, _open function returns + * "Permission denied" error. + */ + attr = GetFileAttributesA(path); + if (attr == (DWORD)-1 && GetLastError() == ERROR_PATH_NOT_FOUND) { + ws = __la_win_permissive_name(path); + if (ws == NULL) { + errno = EINVAL; + return (-1); + } + attr = GetFileAttributesW(ws); + } + if (attr == (DWORD)-1) { + la_dosmaperr(GetLastError()); + free(ws); + return (-1); + } + if (attr & FILE_ATTRIBUTE_DIRECTORY) { + HANDLE handle; + + if (ws != NULL) + handle = CreateFileW(ws, 0, 0, NULL, + OPEN_EXISTING, + FILE_FLAG_BACKUP_SEMANTICS | + FILE_ATTRIBUTE_READONLY, + NULL); + else + handle = CreateFileA(path, 0, 0, NULL, + OPEN_EXISTING, + FILE_FLAG_BACKUP_SEMANTICS | + FILE_ATTRIBUTE_READONLY, + NULL); + free(ws); + if (handle == INVALID_HANDLE_VALUE) { + la_dosmaperr(GetLastError()); + return (-1); + } + r = _open_osfhandle((intptr_t)handle, _O_RDONLY); + return (r); + } + } + if (ws == NULL) { +#if defined(__BORLANDC__) + /* Borland has no mode argument. + TODO: Fix mode of new file. */ + r = _open(path, flags); +#else + r = _open(path, flags, pmode); +#endif + if (r < 0 && errno == EACCES && (flags & O_CREAT) != 0) { + /* Simulate other POSIX system action to pass our test suite. */ + attr = GetFileAttributesA(path); + if (attr == (DWORD)-1) + la_dosmaperr(GetLastError()); + else if (attr & FILE_ATTRIBUTE_DIRECTORY) + errno = EISDIR; + else + errno = EACCES; + return (-1); + } + if (r >= 0 || errno != ENOENT) + return (r); + ws = __la_win_permissive_name(path); + if (ws == NULL) { + errno = EINVAL; + return (-1); + } + } + r = _wopen(ws, flags, pmode); + if (r < 0 && errno == EACCES && (flags & O_CREAT) != 0) { + /* Simulate other POSIX system action to pass our test suite. */ + attr = GetFileAttributesW(ws); + if (attr == (DWORD)-1) + la_dosmaperr(GetLastError()); + else if (attr & FILE_ATTRIBUTE_DIRECTORY) + errno = EISDIR; + else + errno = EACCES; + } + free(ws); + return (r); +} + +ssize_t +__la_read(int fd, void *buf, size_t nbytes) +{ + HANDLE handle; + DWORD bytes_read, lasterr; + int r; + +#ifdef _WIN64 + if (nbytes > UINT32_MAX) + nbytes = UINT32_MAX; +#endif + if (fd < 0) { + errno = EBADF; + return (-1); + } + /* Do not pass 0 to third parameter of ReadFile(), read bytes. + * This will not return to application side. */ + if (nbytes == 0) + return (0); + handle = (HANDLE)_get_osfhandle(fd); + r = ReadFile(handle, buf, (uint32_t)nbytes, + &bytes_read, NULL); + if (r == 0) { + lasterr = GetLastError(); + if (lasterr == ERROR_NO_DATA) { + errno = EAGAIN; + return (-1); + } + if (lasterr == ERROR_BROKEN_PIPE) + return (0); + if (lasterr == ERROR_ACCESS_DENIED) + errno = EBADF; + else + la_dosmaperr(lasterr); + return (-1); + } + return ((ssize_t)bytes_read); +} + +/* Convert Windows FILETIME to UTC */ +__inline static void +fileTimeToUTC(const FILETIME *filetime, time_t *t, long *ns) +{ + ULARGE_INTEGER utc; + + utc.HighPart = filetime->dwHighDateTime; + utc.LowPart = filetime->dwLowDateTime; + if (utc.QuadPart >= EPOC_TIME) { + utc.QuadPart -= EPOC_TIME; + *t = (time_t)(utc.QuadPart / 10000000); /* milli seconds base */ + *ns = (long)(utc.QuadPart % 10000000) * 100;/* nano seconds base */ + } else { + *t = 0; + *ns = 0; + } +} + +/* Stat by handle + * Windows' stat() does not accept the path added "\\?\" especially "?" + * character. + * It means we cannot access the long name path longer than MAX_PATH. + * So I've implemented simular Windows' stat() to access the long name path. + * And I've added some feature. + * 1. set st_ino by nFileIndexHigh and nFileIndexLow of + * BY_HANDLE_FILE_INFORMATION. + * 2. set st_nlink by nNumberOfLinks of BY_HANDLE_FILE_INFORMATION. + * 3. set st_dev by dwVolumeSerialNumber by BY_HANDLE_FILE_INFORMATION. + */ +static int +__hstat(HANDLE handle, struct ustat *st) +{ + BY_HANDLE_FILE_INFORMATION info; + ULARGE_INTEGER ino64; + DWORD ftype; + mode_t mode; + time_t t; + long ns; + + switch (ftype = GetFileType(handle)) { + case FILE_TYPE_UNKNOWN: + errno = EBADF; + return (-1); + case FILE_TYPE_CHAR: + case FILE_TYPE_PIPE: + if (ftype == FILE_TYPE_CHAR) { + st->st_mode = S_IFCHR; + st->st_size = 0; + } else { + DWORD avail; + + st->st_mode = S_IFIFO; + if (PeekNamedPipe(handle, NULL, 0, NULL, &avail, NULL)) + st->st_size = avail; + else + st->st_size = 0; + } + st->st_atime = 0; + st->st_atime_nsec = 0; + st->st_mtime = 0; + st->st_mtime_nsec = 0; + st->st_ctime = 0; + st->st_ctime_nsec = 0; + st->st_ino = 0; + st->st_nlink = 1; + st->st_uid = 0; + st->st_gid = 0; + st->st_rdev = 0; + st->st_dev = 0; + return (0); + case FILE_TYPE_DISK: + break; + default: + /* This ftype is undocumented type. */ + la_dosmaperr(GetLastError()); + return (-1); + } + + ZeroMemory(&info, sizeof(info)); + if (!GetFileInformationByHandle (handle, &info)) { + la_dosmaperr(GetLastError()); + return (-1); + } + + mode = S_IRUSR | S_IRGRP | S_IROTH; + if ((info.dwFileAttributes & FILE_ATTRIBUTE_READONLY) == 0) + mode |= S_IWUSR | S_IWGRP | S_IWOTH; + if (info.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) + mode |= S_IFDIR | S_IXUSR | S_IXGRP | S_IXOTH; + else + mode |= S_IFREG; + st->st_mode = mode; + + fileTimeToUTC(&info.ftLastAccessTime, &t, &ns); + st->st_atime = t; + st->st_atime_nsec = ns; + fileTimeToUTC(&info.ftLastWriteTime, &t, &ns); + st->st_mtime = t; + st->st_mtime_nsec = ns; + fileTimeToUTC(&info.ftCreationTime, &t, &ns); + st->st_ctime = t; + st->st_ctime_nsec = ns; + st->st_size = + ((int64_t)(info.nFileSizeHigh) * ((int64_t)MAXDWORD + 1)) + + (int64_t)(info.nFileSizeLow); +#ifdef SIMULATE_WIN_STAT + st->st_ino = 0; + st->st_nlink = 1; + st->st_dev = 0; +#else + /* Getting FileIndex as i-node. We should remove a sequence which + * is high-16-bits of nFileIndexHigh. */ + ino64.HighPart = info.nFileIndexHigh & 0x0000FFFFUL; + ino64.LowPart = info.nFileIndexLow; + st->st_ino = ino64.QuadPart; + st->st_nlink = info.nNumberOfLinks; + if (info.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) + ++st->st_nlink;/* Add parent directory. */ + st->st_dev = info.dwVolumeSerialNumber; +#endif + st->st_uid = 0; + st->st_gid = 0; + st->st_rdev = 0; + return (0); +} + +static void +copy_stat(struct stat *st, struct ustat *us) +{ + st->st_atime = us->st_atime; + st->st_ctime = us->st_ctime; + st->st_mtime = us->st_mtime; + st->st_gid = us->st_gid; + st->st_ino = getino(us); + st->st_mode = us->st_mode; + st->st_nlink = us->st_nlink; + st->st_size = (off_t)us->st_size; + st->st_uid = us->st_uid; + st->st_dev = us->st_dev; + st->st_rdev = us->st_rdev; +} + +/* + * TODO: Remove a use of __la_fstat and __la_stat. + * We should use GetFileInformationByHandle in place + * where We still use the *stat functions. + */ +int +__la_fstat(int fd, struct stat *st) +{ + struct ustat u; + int ret; + + if (fd < 0) { + errno = EBADF; + return (-1); + } + ret = __hstat((HANDLE)_get_osfhandle(fd), &u); + if (ret >= 0) { + copy_stat(st, &u); + if (u.st_mode & (S_IFCHR | S_IFIFO)) { + st->st_dev = fd; + st->st_rdev = fd; + } + } + return (ret); +} + +/* This can exceed MAX_PATH limitation. */ +int +__la_stat(const char *path, struct stat *st) +{ + HANDLE handle; + struct ustat u; + int ret; + + handle = la_CreateFile(path, 0, 0, NULL, OPEN_EXISTING, + FILE_FLAG_BACKUP_SEMANTICS, + NULL); + if (handle == INVALID_HANDLE_VALUE) { + la_dosmaperr(GetLastError()); + return (-1); + } + ret = __hstat(handle, &u); + CloseHandle(handle); + if (ret >= 0) { + char *p; + + copy_stat(st, &u); + p = strrchr(path, '.'); + if (p != NULL && strlen(p) == 4) { + char exttype[4]; + + ++ p; + exttype[0] = toupper(*p++); + exttype[1] = toupper(*p++); + exttype[2] = toupper(*p++); + exttype[3] = '\0'; + if (!strcmp(exttype, "EXE") || !strcmp(exttype, "CMD") || + !strcmp(exttype, "BAT") || !strcmp(exttype, "COM")) + st->st_mode |= S_IXUSR | S_IXGRP | S_IXOTH; + } + } + return (ret); +} + +/* + * This waitpid is limited implementation. + */ +pid_t +__la_waitpid(HANDLE child, int *status, int option) +{ + DWORD cs; + + (void)option;/* UNUSED */ + do { + if (GetExitCodeProcess(child, &cs) == 0) { + CloseHandle(child); + la_dosmaperr(GetLastError()); + *status = 0; + return (-1); + } + } while (cs == STILL_ACTIVE); + + *status = (int)(cs & 0xff); + return (0); +} + +ssize_t +__la_write(int fd, const void *buf, size_t nbytes) +{ + DWORD bytes_written; + +#ifdef _WIN64 + if (nbytes > UINT32_MAX) + nbytes = UINT32_MAX; +#endif + if (fd < 0) { + errno = EBADF; + return (-1); + } + if (!WriteFile((HANDLE)_get_osfhandle(fd), buf, (uint32_t)nbytes, + &bytes_written, NULL)) { + DWORD lasterr; + + lasterr = GetLastError(); + if (lasterr == ERROR_ACCESS_DENIED) + errno = EBADF; + else + la_dosmaperr(lasterr); + return (-1); + } + return (bytes_written); +} + +/* + * Replace the Windows path separator '\' with '/'. + */ +static int +replace_pathseparator(struct archive_wstring *ws, const wchar_t *wp) +{ + wchar_t *w; + size_t path_length; + + if (wp == NULL) + return(0); + if (wcschr(wp, L'\\') == NULL) + return(0); + path_length = wcslen(wp); + if (archive_wstring_ensure(ws, path_length) == NULL) + return(-1); + archive_wstrncpy(ws, wp, path_length); + for (w = ws->s; *w; w++) { + if (*w == L'\\') + *w = L'/'; + } + return(1); +} + +static int +fix_pathseparator(struct archive_entry *entry) +{ + struct archive_wstring ws; + const wchar_t *wp; + int ret = ARCHIVE_OK; + + archive_string_init(&ws); + wp = archive_entry_pathname_w(entry); + switch (replace_pathseparator(&ws, wp)) { + case 0: /* Not replaced. */ + break; + case 1: /* Replaced. */ + archive_entry_copy_pathname_w(entry, ws.s); + break; + default: + ret = ARCHIVE_FAILED; + } + wp = archive_entry_hardlink_w(entry); + switch (replace_pathseparator(&ws, wp)) { + case 0: /* Not replaced. */ + break; + case 1: /* Replaced. */ + archive_entry_copy_hardlink_w(entry, ws.s); + break; + default: + ret = ARCHIVE_FAILED; + } + wp = archive_entry_symlink_w(entry); + switch (replace_pathseparator(&ws, wp)) { + case 0: /* Not replaced. */ + break; + case 1: /* Replaced. */ + archive_entry_copy_symlink_w(entry, ws.s); + break; + default: + ret = ARCHIVE_FAILED; + } + archive_wstring_free(&ws); + return(ret); +} + +struct archive_entry * +__la_win_entry_in_posix_pathseparator(struct archive_entry *entry) +{ + struct archive_entry *entry_main; + const wchar_t *wp; + int has_backslash = 0; + int ret; + + wp = archive_entry_pathname_w(entry); + if (wp != NULL && wcschr(wp, L'\\') != NULL) + has_backslash = 1; + if (!has_backslash) { + wp = archive_entry_hardlink_w(entry); + if (wp != NULL && wcschr(wp, L'\\') != NULL) + has_backslash = 1; + } + if (!has_backslash) { + wp = archive_entry_symlink_w(entry); + if (wp != NULL && wcschr(wp, L'\\') != NULL) + has_backslash = 1; + } + /* + * If there is no backslach chars, return the original. + */ + if (!has_backslash) + return (entry); + + /* Copy entry so we can modify it as needed. */ + entry_main = archive_entry_clone(entry); + if (entry_main == NULL) + return (NULL); + /* Replace the Windows path-separator '\' with '/'. */ + ret = fix_pathseparator(entry_main); + if (ret < ARCHIVE_WARN) { + archive_entry_free(entry_main); + return (NULL); + } + return (entry_main); +} + + +/* + * The following function was modified from PostgreSQL sources and is + * subject to the copyright below. + */ +/*------------------------------------------------------------------------- + * + * win32error.c + * Map win32 error codes to errno values + * + * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group + * + * IDENTIFICATION + * $PostgreSQL: pgsql/src/port/win32error.c,v 1.4 2008/01/01 19:46:00 momjian Exp $ + * + *------------------------------------------------------------------------- + */ +/* +PostgreSQL Database Management System +(formerly known as Postgres, then as Postgres95) + +Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group + +Portions Copyright (c) 1994, The Regents of the University of California + +Permission to use, copy, modify, and distribute this software and its +documentation for any purpose, without fee, and without a written agreement +is hereby granted, provided that the above copyright notice and this +paragraph and the following two paragraphs appear in all copies. + +IN NO EVENT SHALL THE UNIVERSITY OF CALIFORNIA BE LIABLE TO ANY PARTY FOR +DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, INCLUDING +LOST PROFITS, ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS +DOCUMENTATION, EVEN IF THE UNIVERSITY OF CALIFORNIA HAS BEEN ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. + +THE UNIVERSITY OF CALIFORNIA SPECIFICALLY DISCLAIMS ANY WARRANTIES, +INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY +AND FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS +ON AN "AS IS" BASIS, AND THE UNIVERSITY OF CALIFORNIA HAS NO OBLIGATIONS TO +PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS. +*/ + +static const struct { + DWORD winerr; + int doserr; +} doserrors[] = +{ + { ERROR_INVALID_FUNCTION, EINVAL }, + { ERROR_FILE_NOT_FOUND, ENOENT }, + { ERROR_PATH_NOT_FOUND, ENOENT }, + { ERROR_TOO_MANY_OPEN_FILES, EMFILE }, + { ERROR_ACCESS_DENIED, EACCES }, + { ERROR_INVALID_HANDLE, EBADF }, + { ERROR_ARENA_TRASHED, ENOMEM }, + { ERROR_NOT_ENOUGH_MEMORY, ENOMEM }, + { ERROR_INVALID_BLOCK, ENOMEM }, + { ERROR_BAD_ENVIRONMENT, E2BIG }, + { ERROR_BAD_FORMAT, ENOEXEC }, + { ERROR_INVALID_ACCESS, EINVAL }, + { ERROR_INVALID_DATA, EINVAL }, + { ERROR_INVALID_DRIVE, ENOENT }, + { ERROR_CURRENT_DIRECTORY, EACCES }, + { ERROR_NOT_SAME_DEVICE, EXDEV }, + { ERROR_NO_MORE_FILES, ENOENT }, + { ERROR_LOCK_VIOLATION, EACCES }, + { ERROR_SHARING_VIOLATION, EACCES }, + { ERROR_BAD_NETPATH, ENOENT }, + { ERROR_NETWORK_ACCESS_DENIED, EACCES }, + { ERROR_BAD_NET_NAME, ENOENT }, + { ERROR_FILE_EXISTS, EEXIST }, + { ERROR_CANNOT_MAKE, EACCES }, + { ERROR_FAIL_I24, EACCES }, + { ERROR_INVALID_PARAMETER, EINVAL }, + { ERROR_NO_PROC_SLOTS, EAGAIN }, + { ERROR_DRIVE_LOCKED, EACCES }, + { ERROR_BROKEN_PIPE, EPIPE }, + { ERROR_DISK_FULL, ENOSPC }, + { ERROR_INVALID_TARGET_HANDLE, EBADF }, + { ERROR_INVALID_HANDLE, EINVAL }, + { ERROR_WAIT_NO_CHILDREN, ECHILD }, + { ERROR_CHILD_NOT_COMPLETE, ECHILD }, + { ERROR_DIRECT_ACCESS_HANDLE, EBADF }, + { ERROR_NEGATIVE_SEEK, EINVAL }, + { ERROR_SEEK_ON_DEVICE, EACCES }, + { ERROR_DIR_NOT_EMPTY, ENOTEMPTY }, + { ERROR_NOT_LOCKED, EACCES }, + { ERROR_BAD_PATHNAME, ENOENT }, + { ERROR_MAX_THRDS_REACHED, EAGAIN }, + { ERROR_LOCK_FAILED, EACCES }, + { ERROR_ALREADY_EXISTS, EEXIST }, + { ERROR_FILENAME_EXCED_RANGE, ENOENT }, + { ERROR_NESTING_NOT_ALLOWED, EAGAIN }, + { ERROR_NOT_ENOUGH_QUOTA, ENOMEM } +}; + +void +__la_dosmaperr(unsigned long e) +{ + int i; + + if (e == 0) + { + errno = 0; + return; + } + + for (i = 0; i < (int)sizeof(doserrors); i++) + { + if (doserrors[i].winerr == e) + { + errno = doserrors[i].doserr; + return; + } + } + + /* fprintf(stderr, "unrecognized win32 error code: %lu", e); */ + errno = EINVAL; + return; +} + +#endif /* _WIN32 && !__CYGWIN__ */ diff --git a/libarchive/archive_windows.h b/libarchive/archive_windows.h new file mode 100644 index 000000000000..c6f5bc510513 --- /dev/null +++ b/libarchive/archive_windows.h @@ -0,0 +1,306 @@ +/*- + * Copyright (c) 2009-2011 Michihiro NAKAJIMA + * Copyright (c) 2003-2006 Tim Kientzle + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer + * in this position and unchanged. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + * $FreeBSD$ + */ + +#ifndef __LIBARCHIVE_BUILD +#error This header is only to be used internally to libarchive. +#endif + +/* + * TODO: A lot of stuff in here isn't actually used by libarchive and + * can be trimmed out. Note that this file is used by libarchive and + * libarchive_test but nowhere else. (But note that it gets compiled + * with many different Windows environments, including MinGW, Visual + * Studio, and Cygwin. Significant changes should be tested in all three.) + */ + +/* + * TODO: Don't use off_t in here. Use __int64 instead. Note that + * Visual Studio and the Windows SDK define off_t as 32 bits; Win32's + * more modern file handling APIs all use __int64 instead of off_t. + */ + +#ifndef LIBARCHIVE_ARCHIVE_WINDOWS_H_INCLUDED +#define LIBARCHIVE_ARCHIVE_WINDOWS_H_INCLUDED + +/* Start of configuration for native Win32 */ +#ifndef MINGW_HAS_SECURE_API +#define MINGW_HAS_SECURE_API 1 +#endif + +#include <errno.h> +#define set_errno(val) ((errno)=val) +#include <io.h> +#include <stdlib.h> //brings in NULL +#if defined(HAVE_STDINT_H) +#include <stdint.h> +#endif +#include <stdio.h> +#include <fcntl.h> +#include <sys/stat.h> +#include <process.h> +#include <direct.h> +#if defined(__MINGW32__) && defined(HAVE_UNISTD_H) +/* Prevent build error from a type mismatch of ftruncate(). + * This unistd.h defines it as ftruncate(int, off_t). */ +#include <unistd.h> +#endif +#define NOCRYPT +#include <windows.h> +//#define EFTYPE 7 + +#if defined(__BORLANDC__) +#pragma warn -8068 /* Constant out of range in comparison. */ +#pragma warn -8072 /* Suspicious pointer arithmetic. */ +#endif + +#ifndef NULL +#ifdef __cplusplus +#define NULL 0 +#else +#define NULL ((void *)0) +#endif +#endif + +/* Alias the Windows _function to the POSIX equivalent. */ +#define close _close +#define fcntl(fd, cmd, flg) /* No operation. */ +#ifndef fileno +#define fileno _fileno +#endif +#ifdef fstat +#undef fstat +#endif +#define fstat __la_fstat +#if !defined(__BORLANDC__) +#ifdef lseek +#undef lseek +#endif +#define lseek _lseeki64 +#else +#define lseek __la_lseek +#define __LA_LSEEK_NEEDED +#endif +#define lstat __la_stat +#define open __la_open +#define read __la_read +#if !defined(__BORLANDC__) +#define setmode _setmode +#endif +#ifdef stat +#undef stat +#endif +#define stat(path,stref) __la_stat(path,stref) +#if !defined(__BORLANDC__) +#define strdup _strdup +#endif +#define tzset _tzset +#if !defined(__BORLANDC__) +#define umask _umask +#endif +#define waitpid __la_waitpid +#define write __la_write + +#ifndef O_RDONLY +#define O_RDONLY _O_RDONLY +#define O_WRONLY _O_WRONLY +#define O_TRUNC _O_TRUNC +#define O_CREAT _O_CREAT +#define O_EXCL _O_EXCL +#define O_BINARY _O_BINARY +#endif + +#ifndef _S_IFIFO + #define _S_IFIFO 0010000 /* pipe */ +#endif +#ifndef _S_IFCHR + #define _S_IFCHR 0020000 /* character special */ +#endif +#ifndef _S_IFDIR + #define _S_IFDIR 0040000 /* directory */ +#endif +#ifndef _S_IFBLK + #define _S_IFBLK 0060000 /* block special */ +#endif +#ifndef _S_IFLNK + #define _S_IFLNK 0120000 /* symbolic link */ +#endif +#ifndef _S_IFSOCK + #define _S_IFSOCK 0140000 /* socket */ +#endif +#ifndef _S_IFREG + #define _S_IFREG 0100000 /* regular */ +#endif +#ifndef _S_IFMT + #define _S_IFMT 0170000 /* file type mask */ +#endif + +#ifndef S_IFIFO +#define S_IFIFO _S_IFIFO +#endif +//#define S_IFCHR _S_IFCHR +//#define S_IFDIR _S_IFDIR +#ifndef S_IFBLK +#define S_IFBLK _S_IFBLK +#endif +#ifndef S_IFLNK +#define S_IFLNK _S_IFLNK +#endif +#ifndef S_IFSOCK +#define S_IFSOCK _S_IFSOCK +#endif +//#define S_IFREG _S_IFREG +//#define S_IFMT _S_IFMT + +#ifndef S_ISBLK +#define S_ISBLK(m) (((m) & S_IFMT) == S_IFBLK) /* block special */ +#define S_ISFIFO(m) (((m) & S_IFMT) == S_IFIFO) /* fifo or socket */ +#define S_ISCHR(m) (((m) & S_IFMT) == S_IFCHR) /* char special */ +#define S_ISDIR(m) (((m) & S_IFMT) == S_IFDIR) /* directory */ +#define S_ISREG(m) (((m) & S_IFMT) == S_IFREG) /* regular file */ +#endif +#define S_ISLNK(m) (((m) & S_IFMT) == S_IFLNK) /* Symbolic link */ +#define S_ISSOCK(m) (((m) & S_IFMT) == S_IFSOCK) /* Socket */ + +#define _S_ISUID 0004000 /* set user id on execution */ +#define _S_ISGID 0002000 /* set group id on execution */ +#define _S_ISVTX 0001000 /* save swapped text even after use */ + +#define S_ISUID _S_ISUID +#define S_ISGID _S_ISGID +#define S_ISVTX _S_ISVTX + +#define _S_IRWXU (_S_IREAD | _S_IWRITE | _S_IEXEC) +#define _S_IXUSR _S_IEXEC /* read permission, user */ +#define _S_IWUSR _S_IWRITE /* write permission, user */ +#define _S_IRUSR _S_IREAD /* execute/search permission, user */ +#define _S_IRWXG (_S_IRWXU >> 3) +#define _S_IXGRP (_S_IXUSR >> 3) /* read permission, group */ +#define _S_IWGRP (_S_IWUSR >> 3) /* write permission, group */ +#define _S_IRGRP (_S_IRUSR >> 3) /* execute/search permission, group */ +#define _S_IRWXO (_S_IRWXG >> 3) +#define _S_IXOTH (_S_IXGRP >> 3) /* read permission, other */ +#define _S_IWOTH (_S_IWGRP >> 3) /* write permission, other */ +#define _S_IROTH (_S_IRGRP >> 3) /* execute/search permission, other */ + +#ifndef S_IRWXU +#define S_IRWXU _S_IRWXU +#define S_IXUSR _S_IXUSR +#define S_IWUSR _S_IWUSR +#define S_IRUSR _S_IRUSR +#endif +#define S_IRWXG _S_IRWXG +#define S_IXGRP _S_IXGRP +#define S_IWGRP _S_IWGRP +#define S_IRGRP _S_IRGRP +#define S_IRWXO _S_IRWXO +#define S_IXOTH _S_IXOTH +#define S_IWOTH _S_IWOTH +#define S_IROTH _S_IROTH + +#define F_DUPFD 0 /* Duplicate file descriptor. */ +#define F_GETFD 1 /* Get file descriptor flags. */ +#define F_SETFD 2 /* Set file descriptor flags. */ +#define F_GETFL 3 /* Get file status flags. */ +#define F_SETFL 4 /* Set file status flags. */ +#define F_GETOWN 5 /* Get owner (receiver of SIGIO). */ +#define F_SETOWN 6 /* Set owner (receiver of SIGIO). */ +#define F_GETLK 7 /* Get record locking info. */ +#define F_SETLK 8 /* Set record locking info (non-blocking). */ +#define F_SETLKW 9 /* Set record locking info (blocking). */ + +/* XXX missing */ +#define F_GETLK64 7 /* Get record locking info. */ +#define F_SETLK64 8 /* Set record locking info (non-blocking). */ +#define F_SETLKW64 9 /* Set record locking info (blocking). */ + +/* File descriptor flags used with F_GETFD and F_SETFD. */ +#define FD_CLOEXEC 1 /* Close on exec. */ + +//NOT SURE IF O_NONBLOCK is OK here but at least the 0x0004 flag is not used by anything else... +#define O_NONBLOCK 0x0004 /* Non-blocking I/O. */ +//#define O_NDELAY O_NONBLOCK + +/* Symbolic constants for the access() function */ +#if !defined(F_OK) + #define R_OK 4 /* Test for read permission */ + #define W_OK 2 /* Test for write permission */ + #define X_OK 1 /* Test for execute permission */ + #define F_OK 0 /* Test for existence of file */ +#endif + + +/* Replacement POSIX function */ +extern int __la_fstat(int fd, struct stat *st); +extern int __la_lstat(const char *path, struct stat *st); +#if defined(__LA_LSEEK_NEEDED) +extern __int64 __la_lseek(int fd, __int64 offset, int whence); +#endif +extern int __la_open(const char *path, int flags, ...); +extern ssize_t __la_read(int fd, void *buf, size_t nbytes); +extern int __la_stat(const char *path, struct stat *st); +extern pid_t __la_waitpid(HANDLE child, int *status, int option); +extern ssize_t __la_write(int fd, const void *buf, size_t nbytes); + +#define _stat64i32(path, st) __la_stat(path, st) +#define _stat64(path, st) __la_stat(path, st) +/* for status returned by la_waitpid */ +#define WIFEXITED(sts) ((sts & 0x100) == 0) +#define WEXITSTATUS(sts) (sts & 0x0FF) + +extern wchar_t *__la_win_permissive_name(const char *name); +extern wchar_t *__la_win_permissive_name_w(const wchar_t *wname); +extern void __la_dosmaperr(unsigned long e); +#define la_dosmaperr(e) __la_dosmaperr(e) +extern struct archive_entry *__la_win_entry_in_posix_pathseparator( + struct archive_entry *); + +#if defined(HAVE_WCRTOMB) && defined(__BORLANDC__) +typedef int mbstate_t; +size_t wcrtomb(char *, wchar_t, mbstate_t *); +#endif + +#if defined(_MSC_VER) && _MSC_VER < 1300 +WINBASEAPI BOOL WINAPI GetVolumePathNameW( + LPCWSTR lpszFileName, + LPWSTR lpszVolumePathName, + DWORD cchBufferLength + ); +# if _WIN32_WINNT < 0x0500 /* windows.h not providing 0x500 API */ +typedef struct _FILE_ALLOCATED_RANGE_BUFFER { + LARGE_INTEGER FileOffset; + LARGE_INTEGER Length; +} FILE_ALLOCATED_RANGE_BUFFER, *PFILE_ALLOCATED_RANGE_BUFFER; +# define FSCTL_SET_SPARSE \ + CTL_CODE(FILE_DEVICE_FILE_SYSTEM, 49, METHOD_BUFFERED, FILE_WRITE_DATA) +# define FSCTL_QUERY_ALLOCATED_RANGES \ + CTL_CODE(FILE_DEVICE_FILE_SYSTEM, 51, METHOD_NEITHER, FILE_READ_DATA) +# endif +#endif + +#endif /* LIBARCHIVE_ARCHIVE_WINDOWS_H_INCLUDED */ diff --git a/libarchive/archive_write_disk_windows.c b/libarchive/archive_write_disk_windows.c new file mode 100644 index 000000000000..0f0780a8e47e --- /dev/null +++ b/libarchive/archive_write_disk_windows.c @@ -0,0 +1,2502 @@ +/*- + * Copyright (c) 2003-2010 Tim Kientzle + * Copyright (c) 2011-2012 Michihiro NAKAJIMA + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer + * in this position and unchanged. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "archive_platform.h" +__FBSDID("$FreeBSD$"); + +#if defined(_WIN32) && !defined(__CYGWIN__) + +#ifdef HAVE_SYS_TYPES_H +#include <sys/types.h> +#endif +#ifdef HAVE_SYS_UTIME_H +#include <sys/utime.h> +#endif +#ifdef HAVE_ERRNO_H +#include <errno.h> +#endif +#ifdef HAVE_FCNTL_H +#include <fcntl.h> +#endif +#ifdef HAVE_LIMITS_H +#include <limits.h> +#endif +#ifdef HAVE_STDLIB_H +#include <stdlib.h> +#endif +#include <winioctl.h> + +/* TODO: Support Mac OS 'quarantine' feature. This is really just a + * standard tag to mark files that have been downloaded as "tainted". + * On Mac OS, we should mark the extracted files as tainted if the + * archive being read was tainted. Windows has a similar feature; we + * should investigate ways to support this generically. */ + +#include "archive.h" +#include "archive_acl_private.h" +#include "archive_string.h" +#include "archive_entry.h" +#include "archive_private.h" + +#ifndef O_BINARY +#define O_BINARY 0 +#endif +#ifndef IO_REPARSE_TAG_SYMLINK +/* Old SDKs do not provide IO_REPARSE_TAG_SYMLINK */ +#define IO_REPARSE_TAG_SYMLINK 0xA000000CL +#endif + +static BOOL SetFilePointerEx_perso(HANDLE hFile, + LARGE_INTEGER liDistanceToMove, + PLARGE_INTEGER lpNewFilePointer, + DWORD dwMoveMethod) +{ + LARGE_INTEGER li; + li.QuadPart = liDistanceToMove.QuadPart; + li.LowPart = SetFilePointer( + hFile, li.LowPart, &li.HighPart, dwMoveMethod); + if(lpNewFilePointer) { + lpNewFilePointer->QuadPart = li.QuadPart; + } + return li.LowPart != (DWORD)-1 || GetLastError() == NO_ERROR; +} + +struct fixup_entry { + struct fixup_entry *next; + struct archive_acl acl; + mode_t mode; + int64_t atime; + int64_t birthtime; + int64_t mtime; + int64_t ctime; + unsigned long atime_nanos; + unsigned long birthtime_nanos; + unsigned long mtime_nanos; + unsigned long ctime_nanos; + unsigned long fflags_set; + int fixup; /* bitmask of what needs fixing */ + wchar_t *name; +}; + +/* + * We use a bitmask to track which operations remain to be done for + * this file. In particular, this helps us avoid unnecessary + * operations when it's possible to take care of one step as a + * side-effect of another. For example, mkdir() can specify the mode + * for the newly-created object but symlink() cannot. This means we + * can skip chmod() if mkdir() succeeded, but we must explicitly + * chmod() if we're trying to create a directory that already exists + * (mkdir() failed) or if we're restoring a symlink. Similarly, we + * need to verify UID/GID before trying to restore SUID/SGID bits; + * that verification can occur explicitly through a stat() call or + * implicitly because of a successful chown() call. + */ +#define TODO_MODE_FORCE 0x40000000 +#define TODO_MODE_BASE 0x20000000 +#define TODO_SUID 0x10000000 +#define TODO_SUID_CHECK 0x08000000 +#define TODO_SGID 0x04000000 +#define TODO_SGID_CHECK 0x02000000 +#define TODO_MODE (TODO_MODE_BASE|TODO_SUID|TODO_SGID) +#define TODO_TIMES ARCHIVE_EXTRACT_TIME +#define TODO_OWNER ARCHIVE_EXTRACT_OWNER +#define TODO_FFLAGS ARCHIVE_EXTRACT_FFLAGS +#define TODO_ACLS ARCHIVE_EXTRACT_ACL +#define TODO_XATTR ARCHIVE_EXTRACT_XATTR +#define TODO_MAC_METADATA ARCHIVE_EXTRACT_MAC_METADATA + +struct archive_write_disk { + struct archive archive; + + mode_t user_umask; + struct fixup_entry *fixup_list; + struct fixup_entry *current_fixup; + int64_t user_uid; + int skip_file_set; + int64_t skip_file_dev; + int64_t skip_file_ino; + time_t start_time; + + int64_t (*lookup_gid)(void *private, const char *gname, int64_t gid); + void (*cleanup_gid)(void *private); + void *lookup_gid_data; + int64_t (*lookup_uid)(void *private, const char *uname, int64_t uid); + void (*cleanup_uid)(void *private); + void *lookup_uid_data; + + /* + * Full path of last file to satisfy symlink checks. + */ + struct archive_wstring path_safe; + + /* + * Cached stat data from disk for the current entry. + * If this is valid, pst points to st. Otherwise, + * pst is null. + */ + BY_HANDLE_FILE_INFORMATION st; + BY_HANDLE_FILE_INFORMATION *pst; + + /* Information about the object being restored right now. */ + struct archive_entry *entry; /* Entry being extracted. */ + wchar_t *name; /* Name of entry, possibly edited. */ + struct archive_wstring _name_data; /* backing store for 'name' */ + /* Tasks remaining for this object. */ + int todo; + /* Tasks deferred until end-of-archive. */ + int deferred; + /* Options requested by the client. */ + int flags; + /* Handle for the file we're restoring. */ + HANDLE fh; + /* Current offset for writing data to the file. */ + int64_t offset; + /* Last offset actually written to disk. */ + int64_t fd_offset; + /* Total bytes actually written to files. */ + int64_t total_bytes_written; + /* Maximum size of file, -1 if unknown. */ + int64_t filesize; + /* Dir we were in before this restore; only for deep paths. */ + int restore_pwd; + /* Mode we should use for this entry; affected by _PERM and umask. */ + mode_t mode; + /* UID/GID to use in restoring this entry. */ + int64_t uid; + int64_t gid; +}; + +/* + * Default mode for dirs created automatically (will be modified by umask). + * Note that POSIX specifies 0777 for implicity-created dirs, "modified + * by the process' file creation mask." + */ +#define DEFAULT_DIR_MODE 0777 +/* + * Dir modes are restored in two steps: During the extraction, the permissions + * in the archive are modified to match the following limits. During + * the post-extract fixup pass, the permissions from the archive are + * applied. + */ +#define MINIMUM_DIR_MODE 0700 +#define MAXIMUM_DIR_MODE 0775 + +static int check_symlinks(struct archive_write_disk *); +static int create_filesystem_object(struct archive_write_disk *); +static struct fixup_entry *current_fixup(struct archive_write_disk *, + const wchar_t *pathname); +static int cleanup_pathname(struct archive_write_disk *); +static int create_dir(struct archive_write_disk *, wchar_t *); +static int create_parent_dir(struct archive_write_disk *, wchar_t *); +static int la_chmod(const wchar_t *, mode_t); +static int older(BY_HANDLE_FILE_INFORMATION *, struct archive_entry *); +static int permissive_name_w(struct archive_write_disk *); +static int restore_entry(struct archive_write_disk *); +static int set_acls(struct archive_write_disk *, HANDLE h, + const wchar_t *, struct archive_acl *); +static int set_xattrs(struct archive_write_disk *); +static int set_fflags(struct archive_write_disk *); +static int set_ownership(struct archive_write_disk *); +static int set_mode(struct archive_write_disk *, int mode); +static int set_times(struct archive_write_disk *, HANDLE, int, + const wchar_t *, time_t, long, time_t, long, time_t, + long, time_t, long); +static int set_times_from_entry(struct archive_write_disk *); +static struct fixup_entry *sort_dir_list(struct fixup_entry *p); +static ssize_t write_data_block(struct archive_write_disk *, + const char *, size_t); + +static struct archive_vtable *archive_write_disk_vtable(void); + +static int _archive_write_disk_close(struct archive *); +static int _archive_write_disk_free(struct archive *); +static int _archive_write_disk_header(struct archive *, + struct archive_entry *); +static int64_t _archive_write_disk_filter_bytes(struct archive *, int); +static int _archive_write_disk_finish_entry(struct archive *); +static ssize_t _archive_write_disk_data(struct archive *, const void *, + size_t); +static ssize_t _archive_write_disk_data_block(struct archive *, const void *, + size_t, int64_t); + +#define bhfi_dev(bhfi) ((bhfi)->dwVolumeSerialNumber) +/* Treat FileIndex as i-node. We should remove a sequence number + * which is high-16-bits of nFileIndexHigh. */ +#define bhfi_ino(bhfi) \ + ((((int64_t)((bhfi)->nFileIndexHigh & 0x0000FFFFUL)) << 32) \ + + (bhfi)->nFileIndexLow) +#define bhfi_size(bhfi) \ + ((((int64_t)(bhfi)->nFileSizeHigh) << 32) + (bhfi)->nFileSizeLow) + +static int +file_information(struct archive_write_disk *a, wchar_t *path, + BY_HANDLE_FILE_INFORMATION *st, mode_t *mode, int sim_lstat) +{ + HANDLE h; + int r; + DWORD flag = FILE_FLAG_BACKUP_SEMANTICS; + WIN32_FIND_DATAW findData; + + if (sim_lstat || mode != NULL) { + h = FindFirstFileW(path, &findData); + if (h == INVALID_HANDLE_VALUE && + GetLastError() == ERROR_INVALID_NAME) { + wchar_t *full; + full = __la_win_permissive_name_w(path); + h = FindFirstFileW(full, &findData); + free(full); + } + if (h == INVALID_HANDLE_VALUE) { + la_dosmaperr(GetLastError()); + return (-1); + } + FindClose(h); + } + + /* Is symlink file ? */ + if (sim_lstat && + ((findData.dwFileAttributes + & FILE_ATTRIBUTE_REPARSE_POINT) && + (findData.dwReserved0 == IO_REPARSE_TAG_SYMLINK))) + flag |= FILE_FLAG_OPEN_REPARSE_POINT; + + h = CreateFileW(a->name, 0, 0, NULL, + OPEN_EXISTING, flag, NULL); + if (h == INVALID_HANDLE_VALUE && + GetLastError() == ERROR_INVALID_NAME) { + wchar_t *full; + full = __la_win_permissive_name_w(path); + h = CreateFileW(full, 0, 0, NULL, + OPEN_EXISTING, flag, NULL); + free(full); + } + if (h == INVALID_HANDLE_VALUE) { + la_dosmaperr(GetLastError()); + return (-1); + } + r = GetFileInformationByHandle(h, st); + CloseHandle(h); + if (r == 0) { + la_dosmaperr(GetLastError()); + return (-1); + } + + if (mode == NULL) + return (0); + + *mode = S_IRUSR | S_IRGRP | S_IROTH; + if ((st->dwFileAttributes & FILE_ATTRIBUTE_READONLY) == 0) + *mode |= S_IWUSR | S_IWGRP | S_IWOTH; + if ((st->dwFileAttributes & FILE_ATTRIBUTE_REPARSE_POINT) && + findData.dwReserved0 == IO_REPARSE_TAG_SYMLINK) + *mode |= S_IFLNK; + else if (st->dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) + *mode |= S_IFDIR | S_IXUSR | S_IXGRP | S_IXOTH; + else { + const wchar_t *p; + + *mode |= S_IFREG; + p = wcsrchr(path, L'.'); + if (p != NULL && wcslen(p) == 4) { + switch (p[1]) { + case L'B': case L'b': + if ((p[2] == L'A' || p[2] == L'a' ) && + (p[3] == L'T' || p[3] == L't' )) + *mode |= S_IXUSR | S_IXGRP | S_IXOTH; + break; + case L'C': case L'c': + if (((p[2] == L'M' || p[2] == L'm' ) && + (p[3] == L'D' || p[3] == L'd' )) || + ((p[2] == L'M' || p[2] == L'm' ) && + (p[3] == L'D' || p[3] == L'd' ))) + *mode |= S_IXUSR | S_IXGRP | S_IXOTH; + break; + case L'E': case L'e': + if ((p[2] == L'X' || p[2] == L'x' ) && + (p[3] == L'E' || p[3] == L'e' )) + *mode |= S_IXUSR | S_IXGRP | S_IXOTH; + break; + default: + break; + } + } + } + return (0); +} + +/* + * Note: The path, for example, "aa/a/../b../c" will be converted to "aa/c" + * by GetFullPathNameW() W32 API, which __la_win_permissive_name_w uses. + * It means we cannot handle multiple dirs in one archive_entry. + * So we have to make the full-pathname in another way, which does not + * break "../" path string. + */ +static int +permissive_name_w(struct archive_write_disk *a) +{ + wchar_t *wn, *wnp; + wchar_t *ws, *wsp; + DWORD l; + + wnp = a->name; + if (wnp[0] == L'\\' && wnp[1] == L'\\' && + wnp[2] == L'?' && wnp[3] == L'\\') + /* We have already a permissive name. */ + return (0); + + if (wnp[0] == L'\\' && wnp[1] == L'\\' && + wnp[2] == L'.' && wnp[3] == L'\\') { + /* This is a device name */ + if (((wnp[4] >= L'a' && wnp[4] <= L'z') || + (wnp[4] >= L'A' && wnp[4] <= L'Z')) && + wnp[5] == L':' && wnp[6] == L'\\') { + wnp[2] = L'?';/* Not device name. */ + return (0); + } + } + + /* + * A full-pathname starting with a drive name like "C:\abc". + */ + if (((wnp[0] >= L'a' && wnp[0] <= L'z') || + (wnp[0] >= L'A' && wnp[0] <= L'Z')) && + wnp[1] == L':' && wnp[2] == L'\\') { + wn = _wcsdup(wnp); + if (wn == NULL) + return (-1); + archive_wstring_ensure(&(a->_name_data), 4 + wcslen(wn) + 1); + a->name = a->_name_data.s; + /* Prepend "\\?\" */ + archive_wstrncpy(&(a->_name_data), L"\\\\?\\", 4); + archive_wstrcat(&(a->_name_data), wn); + free(wn); + return (0); + } + + /* + * A full-pathname pointig a network drive + * like "\\<server-name>\<share-name>\file". + */ + if (wnp[0] == L'\\' && wnp[1] == L'\\' && wnp[2] != L'\\') { + const wchar_t *p = &wnp[2]; + + /* Skip server-name letters. */ + while (*p != L'\\' && *p != L'\0') + ++p; + if (*p == L'\\') { + const wchar_t *rp = ++p; + /* Skip share-name letters. */ + while (*p != L'\\' && *p != L'\0') + ++p; + if (*p == L'\\' && p != rp) { + /* Now, match patterns such as + * "\\server-name\share-name\" */ + wn = _wcsdup(wnp); + if (wn == NULL) + return (-1); + archive_wstring_ensure(&(a->_name_data), + 8 + wcslen(wn) + 1); + a->name = a->_name_data.s; + /* Prepend "\\?\UNC\" */ + archive_wstrncpy(&(a->_name_data), + L"\\\\?\\UNC\\", 8); + archive_wstrcat(&(a->_name_data), wn+2); + free(wn); + return (0); + } + } + return (0); + } + + /* + * Get current working directory. + */ + l = GetCurrentDirectoryW(0, NULL); + if (l == 0) + return (-1); + ws = malloc(l * sizeof(wchar_t)); + l = GetCurrentDirectoryW(l, ws); + if (l == 0) { + free(ws); + return (-1); + } + wsp = ws; + + /* + * A full-pathname starting without a drive name like "\abc". + */ + if (wnp[0] == L'\\') { + wn = _wcsdup(wnp); + if (wn == NULL) + return (-1); + archive_wstring_ensure(&(a->_name_data), + 4 + 2 + wcslen(wn) + 1); + a->name = a->_name_data.s; + /* Prepend "\\?\" and drive name. */ + archive_wstrncpy(&(a->_name_data), L"\\\\?\\", 4); + archive_wstrncat(&(a->_name_data), wsp, 2); + archive_wstrcat(&(a->_name_data), wn); + free(wsp); + free(wn); + return (0); + } + + wn = _wcsdup(wnp); + if (wn == NULL) + return (-1); + archive_wstring_ensure(&(a->_name_data), 4 + l + 1 + wcslen(wn) + 1); + a->name = a->_name_data.s; + /* Prepend "\\?\" and drive name. */ + archive_wstrncpy(&(a->_name_data), L"\\\\?\\", 4); + archive_wstrncat(&(a->_name_data), wsp, l); + archive_wstrncat(&(a->_name_data), L"\\", 1); + archive_wstrcat(&(a->_name_data), wn); + a->name = a->_name_data.s; + free(wsp); + free(wn); + return (0); +} + +static int +la_chmod(const wchar_t *path, mode_t mode) +{ + DWORD attr; + BOOL r; + wchar_t *fullname; + int ret = 0; + + fullname = NULL; + attr = GetFileAttributesW(path); + if (attr == (DWORD)-1 && + GetLastError() == ERROR_INVALID_NAME) { + fullname = __la_win_permissive_name_w(path); + attr = GetFileAttributesW(fullname); + } + if (attr == (DWORD)-1) { + la_dosmaperr(GetLastError()); + ret = -1; + goto exit_chmode; + } + if (mode & _S_IWRITE) + attr &= ~FILE_ATTRIBUTE_READONLY; + else + attr |= FILE_ATTRIBUTE_READONLY; + if (fullname != NULL) + r = SetFileAttributesW(fullname, attr); + else + r = SetFileAttributesW(path, attr); + if (r == 0) { + la_dosmaperr(GetLastError()); + ret = -1; + } +exit_chmode: + free(fullname); + return (ret); +} + +static void * +la_GetFunctionKernel32(const char *name) +{ + static HINSTANCE lib; + static int set; + if (!set) { + set = 1; + lib = LoadLibrary("kernel32.dll"); + } + if (lib == NULL) { + fprintf(stderr, "Can't load kernel32.dll?!\n"); + exit(1); + } + return (void *)GetProcAddress(lib, name); +} + +static int +la_CreateHardLinkW(wchar_t *linkname, wchar_t *target) +{ + static BOOLEAN (WINAPI *f)(LPWSTR, LPWSTR, LPSECURITY_ATTRIBUTES); + static int set; + BOOL ret; + + if (!set) { + set = 1; + f = la_GetFunctionKernel32("CreateHardLinkW"); + } + if (!f) + return (0); + ret = (*f)(linkname, target, NULL); + if (!ret) { + /* Under windows 2000, it is necessary to remove + * the "\\?\" prefix. */ +#define IS_UNC(name) ((name[0] == L'U' || name[0] == L'u') && \ + (name[1] == L'N' || name[1] == L'n') && \ + (name[2] == L'C' || name[2] == L'c') && \ + name[3] == L'\\') + if (!wcsncmp(linkname,L"\\\\?\\", 4)) { + linkname += 4; + if (IS_UNC(linkname)) + linkname += 4; + } + if (!wcsncmp(target,L"\\\\?\\", 4)) { + target += 4; + if (IS_UNC(target)) + target += 4; + } +#undef IS_UNC + ret = (*f)(linkname, target, NULL); + } + return (ret); +} + +static int +la_ftruncate(HANDLE handle, int64_t length) +{ + LARGE_INTEGER distance; + + if (GetFileType(handle) != FILE_TYPE_DISK) { + errno = EBADF; + return (-1); + } + distance.QuadPart = length; + if (!SetFilePointerEx_perso(handle, distance, NULL, FILE_BEGIN)) { + la_dosmaperr(GetLastError()); + return (-1); + } + if (!SetEndOfFile(handle)) { + la_dosmaperr(GetLastError()); + return (-1); + } + return (0); +} + +static int +lazy_stat(struct archive_write_disk *a) +{ + if (a->pst != NULL) { + /* Already have stat() data available. */ + return (ARCHIVE_OK); + } + if (a->fh != INVALID_HANDLE_VALUE && + GetFileInformationByHandle(a->fh, &a->st) == 0) { + a->pst = &a->st; + return (ARCHIVE_OK); + } + + /* + * XXX At this point, symlinks should not be hit, otherwise + * XXX a race occurred. Do we want to check explicitly for that? + */ + if (file_information(a, a->name, &a->st, NULL, 1) == 0) { + a->pst = &a->st; + return (ARCHIVE_OK); + } + archive_set_error(&a->archive, errno, "Couldn't stat file"); + return (ARCHIVE_WARN); +} + +static struct archive_vtable * +archive_write_disk_vtable(void) +{ + static struct archive_vtable av; + static int inited = 0; + + if (!inited) { + av.archive_close = _archive_write_disk_close; + av.archive_filter_bytes = _archive_write_disk_filter_bytes; + av.archive_free = _archive_write_disk_free; + av.archive_write_header = _archive_write_disk_header; + av.archive_write_finish_entry + = _archive_write_disk_finish_entry; + av.archive_write_data = _archive_write_disk_data; + av.archive_write_data_block = _archive_write_disk_data_block; + inited = 1; + } + return (&av); +} + +static int64_t +_archive_write_disk_filter_bytes(struct archive *_a, int n) +{ + struct archive_write_disk *a = (struct archive_write_disk *)_a; + (void)n; /* UNUSED */ + if (n == -1 || n == 0) + return (a->total_bytes_written); + return (-1); +} + + +int +archive_write_disk_set_options(struct archive *_a, int flags) +{ + struct archive_write_disk *a = (struct archive_write_disk *)_a; + + a->flags = flags; + return (ARCHIVE_OK); +} + + +/* + * Extract this entry to disk. + * + * TODO: Validate hardlinks. According to the standards, we're + * supposed to check each extracted hardlink and squawk if it refers + * to a file that we didn't restore. I'm not entirely convinced this + * is a good idea, but more importantly: Is there any way to validate + * hardlinks without keeping a complete list of filenames from the + * entire archive?? Ugh. + * + */ +static int +_archive_write_disk_header(struct archive *_a, struct archive_entry *entry) +{ + struct archive_write_disk *a = (struct archive_write_disk *)_a; + struct fixup_entry *fe; + int ret, r; + + archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, + ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, + "archive_write_disk_header"); + archive_clear_error(&a->archive); + if (a->archive.state & ARCHIVE_STATE_DATA) { + r = _archive_write_disk_finish_entry(&a->archive); + if (r == ARCHIVE_FATAL) + return (r); + } + + /* Set up for this particular entry. */ + a->pst = NULL; + a->current_fixup = NULL; + a->deferred = 0; + if (a->entry) { + archive_entry_free(a->entry); + a->entry = NULL; + } + a->entry = archive_entry_clone(entry); + a->fh = INVALID_HANDLE_VALUE; + a->fd_offset = 0; + a->offset = 0; + a->restore_pwd = -1; + a->uid = a->user_uid; + a->mode = archive_entry_mode(a->entry); + if (archive_entry_size_is_set(a->entry)) + a->filesize = archive_entry_size(a->entry); + else + a->filesize = -1; + archive_wstrcpy(&(a->_name_data), archive_entry_pathname_w(a->entry)); + a->name = a->_name_data.s; + archive_clear_error(&a->archive); + + /* + * Clean up the requested path. This is necessary for correct + * dir restores; the dir restore logic otherwise gets messed + * up by nonsense like "dir/.". + */ + ret = cleanup_pathname(a); + if (ret != ARCHIVE_OK) + return (ret); + + /* + * Generate a full-pathname and use it from here. + */ + if (permissive_name_w(a) < 0) { + errno = EINVAL; + return (ARCHIVE_FAILED); + } + + /* + * Query the umask so we get predictable mode settings. + * This gets done on every call to _write_header in case the + * user edits their umask during the extraction for some + * reason. + */ + umask(a->user_umask = umask(0)); + + /* Figure out what we need to do for this entry. */ + a->todo = TODO_MODE_BASE; + if (a->flags & ARCHIVE_EXTRACT_PERM) { + a->todo |= TODO_MODE_FORCE; /* Be pushy about permissions. */ + /* + * SGID requires an extra "check" step because we + * cannot easily predict the GID that the system will + * assign. (Different systems assign GIDs to files + * based on a variety of criteria, including process + * credentials and the gid of the enclosing + * directory.) We can only restore the SGID bit if + * the file has the right GID, and we only know the + * GID if we either set it (see set_ownership) or if + * we've actually called stat() on the file after it + * was restored. Since there are several places at + * which we might verify the GID, we need a TODO bit + * to keep track. + */ + if (a->mode & S_ISGID) + a->todo |= TODO_SGID | TODO_SGID_CHECK; + /* + * Verifying the SUID is simpler, but can still be + * done in multiple ways, hence the separate "check" bit. + */ + if (a->mode & S_ISUID) + a->todo |= TODO_SUID | TODO_SUID_CHECK; + } else { + /* + * User didn't request full permissions, so don't + * restore SUID, SGID bits and obey umask. + */ + a->mode &= ~S_ISUID; + a->mode &= ~S_ISGID; + a->mode &= ~S_ISVTX; + a->mode &= ~a->user_umask; + } +#if 0 + if (a->flags & ARCHIVE_EXTRACT_OWNER) + a->todo |= TODO_OWNER; +#endif + if (a->flags & ARCHIVE_EXTRACT_TIME) + a->todo |= TODO_TIMES; + if (a->flags & ARCHIVE_EXTRACT_ACL) { + if (archive_entry_filetype(a->entry) == AE_IFDIR) + a->deferred |= TODO_ACLS; + else + a->todo |= TODO_ACLS; + } + if (a->flags & ARCHIVE_EXTRACT_XATTR) + a->todo |= TODO_XATTR; + if (a->flags & ARCHIVE_EXTRACT_FFLAGS) + a->todo |= TODO_FFLAGS; + if (a->flags & ARCHIVE_EXTRACT_SECURE_SYMLINKS) { + ret = check_symlinks(a); + if (ret != ARCHIVE_OK) + return (ret); + } + + ret = restore_entry(a); + + /* + * TODO: There are rumours that some extended attributes must + * be restored before file data is written. If this is true, + * then we either need to write all extended attributes both + * before and after restoring the data, or find some rule for + * determining which must go first and which last. Due to the + * many ways people are using xattrs, this may prove to be an + * intractable problem. + */ + + /* + * Fixup uses the unedited pathname from archive_entry_pathname(), + * because it is relative to the base dir and the edited path + * might be relative to some intermediate dir as a result of the + * deep restore logic. + */ + if (a->deferred & TODO_MODE) { + fe = current_fixup(a, archive_entry_pathname_w(entry)); + fe->fixup |= TODO_MODE_BASE; + fe->mode = a->mode; + } + + if ((a->deferred & TODO_TIMES) + && (archive_entry_mtime_is_set(entry) + || archive_entry_atime_is_set(entry))) { + fe = current_fixup(a, archive_entry_pathname_w(entry)); + fe->mode = a->mode; + fe->fixup |= TODO_TIMES; + if (archive_entry_atime_is_set(entry)) { + fe->atime = archive_entry_atime(entry); + fe->atime_nanos = archive_entry_atime_nsec(entry); + } else { + /* If atime is unset, use start time. */ + fe->atime = a->start_time; + fe->atime_nanos = 0; + } + if (archive_entry_mtime_is_set(entry)) { + fe->mtime = archive_entry_mtime(entry); + fe->mtime_nanos = archive_entry_mtime_nsec(entry); + } else { + /* If mtime is unset, use start time. */ + fe->mtime = a->start_time; + fe->mtime_nanos = 0; + } + if (archive_entry_birthtime_is_set(entry)) { + fe->birthtime = archive_entry_birthtime(entry); + fe->birthtime_nanos = archive_entry_birthtime_nsec(entry); + } else { + /* If birthtime is unset, use mtime. */ + fe->birthtime = fe->mtime; + fe->birthtime_nanos = fe->mtime_nanos; + } + } + + if (a->deferred & TODO_ACLS) { + fe = current_fixup(a, archive_entry_pathname_w(entry)); + archive_acl_copy(&fe->acl, archive_entry_acl(entry)); + } + + if (a->deferred & TODO_FFLAGS) { + fe = current_fixup(a, archive_entry_pathname_w(entry)); + fe->fixup |= TODO_FFLAGS; + /* TODO: Complete this.. defer fflags from below. */ + } + + /* + * On Windows, A creating sparse file requires a special mark. + */ + if (a->fh != INVALID_HANDLE_VALUE && + archive_entry_sparse_count(entry) > 0) { + int64_t base = 0, offset, length; + int i, cnt = archive_entry_sparse_reset(entry); + int sparse = 0; + + for (i = 0; i < cnt; i++) { + archive_entry_sparse_next(entry, &offset, &length); + if (offset - base >= 4096) { + sparse = 1;/* we have a hole. */ + break; + } + base = offset + length; + } + if (sparse) { + DWORD dmy; + /* Mark this file as sparse. */ + DeviceIoControl(a->fh, FSCTL_SET_SPARSE, + NULL, 0, NULL, 0, &dmy, NULL); + } + } + + /* We've created the object and are ready to pour data into it. */ + if (ret >= ARCHIVE_WARN) + a->archive.state = ARCHIVE_STATE_DATA; + /* + * If it's not open, tell our client not to try writing. + * In particular, dirs, links, etc, don't get written to. + */ + if (a->fh == INVALID_HANDLE_VALUE) { + archive_entry_set_size(entry, 0); + a->filesize = 0; + } + + return (ret); +} + +int +archive_write_disk_set_skip_file(struct archive *_a, int64_t d, int64_t i) +{ + struct archive_write_disk *a = (struct archive_write_disk *)_a; + archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_write_disk_set_skip_file"); + a->skip_file_set = 1; + a->skip_file_dev = d; + a->skip_file_ino = i; + return (ARCHIVE_OK); +} + +static ssize_t +write_data_block(struct archive_write_disk *a, const char *buff, size_t size) +{ + OVERLAPPED ol; + uint64_t start_size = size; + DWORD bytes_written = 0; + ssize_t block_size = 0, bytes_to_write; + + if (size == 0) + return (ARCHIVE_OK); + + if (a->filesize == 0 || a->fh == INVALID_HANDLE_VALUE) { + archive_set_error(&a->archive, 0, + "Attempt to write to an empty file"); + return (ARCHIVE_WARN); + } + + if (a->flags & ARCHIVE_EXTRACT_SPARSE) { + /* XXX TODO XXX Is there a more appropriate choice here ? */ + /* This needn't match the filesystem allocation size. */ + block_size = 16*1024; + } + + /* If this write would run beyond the file size, truncate it. */ + if (a->filesize >= 0 && (int64_t)(a->offset + size) > a->filesize) + start_size = size = (size_t)(a->filesize - a->offset); + + /* Write the data. */ + while (size > 0) { + if (block_size == 0) { + bytes_to_write = size; + } else { + /* We're sparsifying the file. */ + const char *p, *end; + int64_t block_end; + + /* Skip leading zero bytes. */ + for (p = buff, end = buff + size; p < end; ++p) { + if (*p != '\0') + break; + } + a->offset += p - buff; + size -= p - buff; + buff = p; + if (size == 0) + break; + + /* Calculate next block boundary after offset. */ + block_end + = (a->offset / block_size + 1) * block_size; + + /* If the adjusted write would cross block boundary, + * truncate it to the block boundary. */ + bytes_to_write = size; + if (a->offset + bytes_to_write > block_end) + bytes_to_write = (DWORD)(block_end - a->offset); + } + memset(&ol, 0, sizeof(ol)); + ol.Offset = (DWORD)(a->offset & 0xFFFFFFFF); + ol.OffsetHigh = (DWORD)(a->offset >> 32); + if (!WriteFile(a->fh, buff, (uint32_t)bytes_to_write, + &bytes_written, &ol)) { + DWORD lasterr; + + lasterr = GetLastError(); + if (lasterr == ERROR_ACCESS_DENIED) + errno = EBADF; + else + la_dosmaperr(lasterr); + archive_set_error(&a->archive, errno, "Write failed"); + return (ARCHIVE_WARN); + } + buff += bytes_written; + size -= bytes_written; + a->total_bytes_written += bytes_written; + a->offset += bytes_written; + a->fd_offset = a->offset; + } + return ((ssize_t)(start_size - size)); +} + +static ssize_t +_archive_write_disk_data_block(struct archive *_a, + const void *buff, size_t size, int64_t offset) +{ + struct archive_write_disk *a = (struct archive_write_disk *)_a; + ssize_t r; + + archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, + ARCHIVE_STATE_DATA, "archive_write_data_block"); + + a->offset = offset; + r = write_data_block(a, buff, size); + if (r < ARCHIVE_OK) + return (r); + if ((size_t)r < size) { + archive_set_error(&a->archive, 0, + "Write request too large"); + return (ARCHIVE_WARN); + } + return (ARCHIVE_OK); +} + +static ssize_t +_archive_write_disk_data(struct archive *_a, const void *buff, size_t size) +{ + struct archive_write_disk *a = (struct archive_write_disk *)_a; + + archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, + ARCHIVE_STATE_DATA, "archive_write_data"); + + return (write_data_block(a, buff, size)); +} + +static int +_archive_write_disk_finish_entry(struct archive *_a) +{ + struct archive_write_disk *a = (struct archive_write_disk *)_a; + int ret = ARCHIVE_OK; + + archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, + ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, + "archive_write_finish_entry"); + if (a->archive.state & ARCHIVE_STATE_HEADER) + return (ARCHIVE_OK); + archive_clear_error(&a->archive); + + /* Pad or truncate file to the right size. */ + if (a->fh == INVALID_HANDLE_VALUE) { + /* There's no file. */ + } else if (a->filesize < 0) { + /* File size is unknown, so we can't set the size. */ + } else if (a->fd_offset == a->filesize) { + /* Last write ended at exactly the filesize; we're done. */ + /* Hopefully, this is the common case. */ + } else { + if (la_ftruncate(a->fh, a->filesize) == -1) { + archive_set_error(&a->archive, errno, + "File size could not be restored"); + return (ARCHIVE_FAILED); + } + } + + /* Restore metadata. */ + + /* + * Look up the "real" UID only if we're going to need it. + * TODO: the TODO_SGID condition can be dropped here, can't it? + */ + if (a->todo & (TODO_OWNER | TODO_SUID | TODO_SGID)) { + a->uid = archive_write_disk_uid(&a->archive, + archive_entry_uname(a->entry), + archive_entry_uid(a->entry)); + } + /* Look up the "real" GID only if we're going to need it. */ + /* TODO: the TODO_SUID condition can be dropped here, can't it? */ + if (a->todo & (TODO_OWNER | TODO_SGID | TODO_SUID)) { + a->gid = archive_write_disk_gid(&a->archive, + archive_entry_gname(a->entry), + archive_entry_gid(a->entry)); + } + + /* + * Restore ownership before set_mode tries to restore suid/sgid + * bits. If we set the owner, we know what it is and can skip + * a stat() call to examine the ownership of the file on disk. + */ + if (a->todo & TODO_OWNER) + ret = set_ownership(a); + + /* + * set_mode must precede ACLs on systems such as Solaris and + * FreeBSD where setting the mode implicitly clears extended ACLs + */ + if (a->todo & TODO_MODE) { + int r2 = set_mode(a, a->mode); + if (r2 < ret) ret = r2; + } + + /* + * Security-related extended attributes (such as + * security.capability on Linux) have to be restored last, + * since they're implicitly removed by other file changes. + */ + if (a->todo & TODO_XATTR) { + int r2 = set_xattrs(a); + if (r2 < ret) ret = r2; + } + + /* + * Some flags prevent file modification; they must be restored after + * file contents are written. + */ + if (a->todo & TODO_FFLAGS) { + int r2 = set_fflags(a); + if (r2 < ret) ret = r2; + } + + /* + * Time must follow most other metadata; + * otherwise atime will get changed. + */ + if (a->todo & TODO_TIMES) { + int r2 = set_times_from_entry(a); + if (r2 < ret) ret = r2; + } + + /* + * ACLs must be restored after timestamps because there are + * ACLs that prevent attribute changes (including time). + */ + if (a->todo & TODO_ACLS) { + int r2 = set_acls(a, a->fh, + archive_entry_pathname_w(a->entry), + archive_entry_acl(a->entry)); + if (r2 < ret) ret = r2; + } + + /* If there's an fd, we can close it now. */ + if (a->fh != INVALID_HANDLE_VALUE) { + CloseHandle(a->fh); + a->fh = INVALID_HANDLE_VALUE; + } + /* If there's an entry, we can release it now. */ + if (a->entry) { + archive_entry_free(a->entry); + a->entry = NULL; + } + a->archive.state = ARCHIVE_STATE_HEADER; + return (ret); +} + +int +archive_write_disk_set_group_lookup(struct archive *_a, + void *private_data, + int64_t (*lookup_gid)(void *private, const char *gname, int64_t gid), + void (*cleanup_gid)(void *private)) +{ + struct archive_write_disk *a = (struct archive_write_disk *)_a; + archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_write_disk_set_group_lookup"); + + if (a->cleanup_gid != NULL && a->lookup_gid_data != NULL) + (a->cleanup_gid)(a->lookup_gid_data); + + a->lookup_gid = lookup_gid; + a->cleanup_gid = cleanup_gid; + a->lookup_gid_data = private_data; + return (ARCHIVE_OK); +} + +int +archive_write_disk_set_user_lookup(struct archive *_a, + void *private_data, + int64_t (*lookup_uid)(void *private, const char *uname, int64_t uid), + void (*cleanup_uid)(void *private)) +{ + struct archive_write_disk *a = (struct archive_write_disk *)_a; + archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_write_disk_set_user_lookup"); + + if (a->cleanup_uid != NULL && a->lookup_uid_data != NULL) + (a->cleanup_uid)(a->lookup_uid_data); + + a->lookup_uid = lookup_uid; + a->cleanup_uid = cleanup_uid; + a->lookup_uid_data = private_data; + return (ARCHIVE_OK); +} + +int64_t +archive_write_disk_gid(struct archive *_a, const char *name, int64_t id) +{ + struct archive_write_disk *a = (struct archive_write_disk *)_a; + archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_write_disk_gid"); + if (a->lookup_gid) + return (a->lookup_gid)(a->lookup_gid_data, name, id); + return (id); +} + +int64_t +archive_write_disk_uid(struct archive *_a, const char *name, int64_t id) +{ + struct archive_write_disk *a = (struct archive_write_disk *)_a; + archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, + ARCHIVE_STATE_ANY, "archive_write_disk_uid"); + if (a->lookup_uid) + return (a->lookup_uid)(a->lookup_uid_data, name, id); + return (id); +} + +/* + * Create a new archive_write_disk object and initialize it with global state. + */ +struct archive * +archive_write_disk_new(void) +{ + struct archive_write_disk *a; + + a = (struct archive_write_disk *)malloc(sizeof(*a)); + if (a == NULL) + return (NULL); + memset(a, 0, sizeof(*a)); + a->archive.magic = ARCHIVE_WRITE_DISK_MAGIC; + /* We're ready to write a header immediately. */ + a->archive.state = ARCHIVE_STATE_HEADER; + a->archive.vtable = archive_write_disk_vtable(); + a->start_time = time(NULL); + /* Query and restore the umask. */ + umask(a->user_umask = umask(0)); + if (archive_wstring_ensure(&a->path_safe, 512) == NULL) { + free(a); + return (NULL); + } + return (&a->archive); +} + +static int +disk_unlink(wchar_t *path) +{ + wchar_t *fullname; + int r; + + r = _wunlink(path); + if (r != 0 && GetLastError() == ERROR_INVALID_NAME) { + fullname = __la_win_permissive_name_w(path); + r = _wunlink(fullname); + free(fullname); + } + return (r); +} + +static int +disk_rmdir(wchar_t *path) +{ + wchar_t *fullname; + int r; + + r = _wrmdir(path); + if (r != 0 && GetLastError() == ERROR_INVALID_NAME) { + fullname = __la_win_permissive_name_w(path); + r = _wrmdir(fullname); + free(fullname); + } + return (r); +} + +/* + * The main restore function. + */ +static int +restore_entry(struct archive_write_disk *a) +{ + int ret = ARCHIVE_OK, en; + + if (a->flags & ARCHIVE_EXTRACT_UNLINK && !S_ISDIR(a->mode)) { + /* + * TODO: Fix this. Apparently, there are platforms + * that still allow root to hose the entire filesystem + * by unlinking a dir. The S_ISDIR() test above + * prevents us from using unlink() here if the new + * object is a dir, but that doesn't mean the old + * object isn't a dir. + */ + if (disk_unlink(a->name) == 0) { + /* We removed it, reset cached stat. */ + a->pst = NULL; + } else if (errno == ENOENT) { + /* File didn't exist, that's just as good. */ + } else if (disk_rmdir(a->name) == 0) { + /* It was a dir, but now it's gone. */ + a->pst = NULL; + } else { + /* We tried, but couldn't get rid of it. */ + archive_set_error(&a->archive, errno, + "Could not unlink"); + return(ARCHIVE_FAILED); + } + } + + /* Try creating it first; if this fails, we'll try to recover. */ + en = create_filesystem_object(a); + + if ((en == ENOTDIR || en == ENOENT) + && !(a->flags & ARCHIVE_EXTRACT_NO_AUTODIR)) { + wchar_t *full; + /* If the parent dir doesn't exist, try creating it. */ + create_parent_dir(a, a->name); + /* Now try to create the object again. */ + full = __la_win_permissive_name_w(a->name); + if (full == NULL) { + en = EINVAL; + } else { + /* Remove multiple directories such as "a/../b../c" */ + archive_wstrcpy(&(a->_name_data), full); + a->name = a->_name_data.s; + free(full); + en = create_filesystem_object(a); + } + } + + if ((en == EISDIR || en == EEXIST) + && (a->flags & ARCHIVE_EXTRACT_NO_OVERWRITE)) { + /* If we're not overwriting, we're done. */ + archive_entry_unset_size(a->entry); + return (ARCHIVE_OK); + } + + /* + * Some platforms return EISDIR if you call + * open(O_WRONLY | O_EXCL | O_CREAT) on a directory, some + * return EEXIST. POSIX is ambiguous, requiring EISDIR + * for open(O_WRONLY) on a dir and EEXIST for open(O_EXCL | O_CREAT) + * on an existing item. + */ + if (en == EISDIR) { + /* A dir is in the way of a non-dir, rmdir it. */ + if (disk_rmdir(a->name) != 0) { + archive_set_error(&a->archive, errno, + "Can't remove already-existing dir"); + return (ARCHIVE_FAILED); + } + a->pst = NULL; + /* Try again. */ + en = create_filesystem_object(a); + } else if (en == EEXIST) { + mode_t st_mode; + /* + * We know something is in the way, but we don't know what; + * we need to find out before we go any further. + */ + int r = 0; + /* + * The SECURE_SYMLINK logic has already removed a + * symlink to a dir if the client wants that. So + * follow the symlink if we're creating a dir. + */ + if (S_ISDIR(a->mode)) + r = file_information(a, a->name, &a->st, &st_mode, 0); + /* + * If it's not a dir (or it's a broken symlink), + * then don't follow it. + */ + if (r != 0 || !S_ISDIR(a->mode)) + r = file_information(a, a->name, &a->st, &st_mode, 1); + if (r != 0) { + archive_set_error(&a->archive, errno, + "Can't stat existing object"); + return (ARCHIVE_FAILED); + } + + /* + * NO_OVERWRITE_NEWER doesn't apply to directories. + */ + if ((a->flags & ARCHIVE_EXTRACT_NO_OVERWRITE_NEWER) + && !S_ISDIR(st_mode)) { + if (!older(&(a->st), a->entry)) { + archive_entry_unset_size(a->entry); + return (ARCHIVE_OK); + } + } + + /* If it's our archive, we're done. */ + if (a->skip_file_set && + bhfi_dev(&a->st) == a->skip_file_dev && + bhfi_ino(&a->st) == a->skip_file_ino) { + archive_set_error(&a->archive, 0, + "Refusing to overwrite archive"); + return (ARCHIVE_FAILED); + } + + if (!S_ISDIR(st_mode)) { + /* A non-dir is in the way, unlink it. */ + if (disk_unlink(a->name) != 0) { + archive_set_error(&a->archive, errno, + "Can't unlink already-existing object"); + return (ARCHIVE_FAILED); + } + a->pst = NULL; + /* Try again. */ + en = create_filesystem_object(a); + } else if (!S_ISDIR(a->mode)) { + /* A dir is in the way of a non-dir, rmdir it. */ + if (disk_rmdir(a->name) != 0) { + archive_set_error(&a->archive, errno, + "Can't remove already-existing dir"); + return (ARCHIVE_FAILED); + } + /* Try again. */ + en = create_filesystem_object(a); + } else { + /* + * There's a dir in the way of a dir. Don't + * waste time with rmdir()/mkdir(), just fix + * up the permissions on the existing dir. + * Note that we don't change perms on existing + * dirs unless _EXTRACT_PERM is specified. + */ + if ((a->mode != st_mode) + && (a->todo & TODO_MODE_FORCE)) + a->deferred |= (a->todo & TODO_MODE); + /* Ownership doesn't need deferred fixup. */ + en = 0; /* Forget the EEXIST. */ + } + } + + if (en) { + /* Everything failed; give up here. */ + archive_set_error(&a->archive, en, "Can't create '%ls'", + a->name); + return (ARCHIVE_FAILED); + } + + a->pst = NULL; /* Cached stat data no longer valid. */ + return (ret); +} + +/* + * Returns 0 if creation succeeds, or else returns errno value from + * the failed system call. Note: This function should only ever perform + * a single system call. + */ +static int +create_filesystem_object(struct archive_write_disk *a) +{ + /* Create the entry. */ + const wchar_t *linkname; + wchar_t *fullname; + mode_t final_mode, mode; + int r; + + /* We identify hard/symlinks according to the link names. */ + /* Since link(2) and symlink(2) don't handle modes, we're done here. */ + linkname = archive_entry_hardlink_w(a->entry); + if (linkname != NULL) { + wchar_t *linkfull, *namefull; + + linkfull = __la_win_permissive_name_w(linkname); + namefull = __la_win_permissive_name_w(a->name); + if (linkfull == NULL || namefull == NULL) { + errno = EINVAL; + r = -1; + } else { + r = la_CreateHardLinkW(namefull, linkfull); + if (r == 0) { + la_dosmaperr(GetLastError()); + r = errno; + } else + r = 0; + } + /* + * New cpio and pax formats allow hardlink entries + * to carry data, so we may have to open the file + * for hardlink entries. + * + * If the hardlink was successfully created and + * the archive doesn't have carry data for it, + * consider it to be non-authoritative for meta data. + * This is consistent with GNU tar and BSD pax. + * If the hardlink does carry data, let the last + * archive entry decide ownership. + */ + if (r == 0 && a->filesize <= 0) { + a->todo = 0; + a->deferred = 0; + } else if (r == 0 && a->filesize > 0) { + a->fh = CreateFileW(namefull, GENERIC_WRITE, 0, NULL, + TRUNCATE_EXISTING, FILE_ATTRIBUTE_NORMAL, NULL); + if (a->fh == INVALID_HANDLE_VALUE) { + la_dosmaperr(GetLastError()); + r = errno; + } + } + free(linkfull); + free(namefull); + return (r); + } + linkname = archive_entry_symlink_w(a->entry); + if (linkname != NULL) { +#if HAVE_SYMLINK + return symlink(linkname, a->name) ? errno : 0; +#else + return (EPERM); +#endif + } + + /* + * The remaining system calls all set permissions, so let's + * try to take advantage of that to avoid an extra chmod() + * call. (Recall that umask is set to zero right now!) + */ + + /* Mode we want for the final restored object (w/o file type bits). */ + final_mode = a->mode & 07777; + /* + * The mode that will actually be restored in this step. Note + * that SUID, SGID, etc, require additional work to ensure + * security, so we never restore them at this point. + */ + mode = final_mode & 0777 & ~a->user_umask; + + switch (a->mode & AE_IFMT) { + default: + /* POSIX requires that we fall through here. */ + /* FALLTHROUGH */ + case AE_IFREG: + fullname = a->name; + /* O_WRONLY | O_CREAT | O_EXCL */ + a->fh = CreateFileW(fullname, GENERIC_WRITE, 0, NULL, + CREATE_NEW, FILE_ATTRIBUTE_NORMAL, NULL); + if (a->fh == INVALID_HANDLE_VALUE && + GetLastError() == ERROR_INVALID_NAME && + fullname == a->name) { + fullname = __la_win_permissive_name_w(a->name); + a->fh = CreateFileW(fullname, GENERIC_WRITE, 0, NULL, + CREATE_NEW, FILE_ATTRIBUTE_NORMAL, NULL); + } + if (a->fh == INVALID_HANDLE_VALUE) { + if (GetLastError() == ERROR_ACCESS_DENIED) { + DWORD attr; + /* Simulate an errno of POSIX system. */ + attr = GetFileAttributesW(fullname); + if (attr == (DWORD)-1) + la_dosmaperr(GetLastError()); + else if (attr & FILE_ATTRIBUTE_DIRECTORY) + errno = EISDIR; + else + errno = EACCES; + } else + la_dosmaperr(GetLastError()); + r = 1; + } else + r = 0; + if (fullname != a->name) + free(fullname); + break; + case AE_IFCHR: + case AE_IFBLK: + /* TODO: Find a better way to warn about our inability + * to restore a block device node. */ + return (EINVAL); + case AE_IFDIR: + mode = (mode | MINIMUM_DIR_MODE) & MAXIMUM_DIR_MODE; + fullname = a->name; + r = CreateDirectoryW(fullname, NULL); + if (r == 0 && GetLastError() == ERROR_INVALID_NAME && + fullname == a->name) { + fullname = __la_win_permissive_name_w(a->name); + r = CreateDirectoryW(fullname, NULL); + } + if (r != 0) { + r = 0; + /* Defer setting dir times. */ + a->deferred |= (a->todo & TODO_TIMES); + a->todo &= ~TODO_TIMES; + /* Never use an immediate chmod(). */ + /* We can't avoid the chmod() entirely if EXTRACT_PERM + * because of SysV SGID inheritance. */ + if ((mode != final_mode) + || (a->flags & ARCHIVE_EXTRACT_PERM)) + a->deferred |= (a->todo & TODO_MODE); + a->todo &= ~TODO_MODE; + } else { + la_dosmaperr(GetLastError()); + r = -1; + } + if (fullname != a->name) + free(fullname); + break; + case AE_IFIFO: + /* TODO: Find a better way to warn about our inability + * to restore a fifo. */ + return (EINVAL); + } + + /* All the system calls above set errno on failure. */ + if (r) + return (errno); + + /* If we managed to set the final mode, we've avoided a chmod(). */ + if (mode == final_mode) + a->todo &= ~TODO_MODE; + return (0); +} + +/* + * Cleanup function for archive_extract. Mostly, this involves processing + * the fixup list, which is used to address a number of problems: + * * Dir permissions might prevent us from restoring a file in that + * dir, so we restore the dir with minimum 0700 permissions first, + * then correct the mode at the end. + * * Similarly, the act of restoring a file touches the directory + * and changes the timestamp on the dir, so we have to touch-up dir + * timestamps at the end as well. + * * Some file flags can interfere with the restore by, for example, + * preventing the creation of hardlinks to those files. + * * Mac OS extended metadata includes ACLs, so must be deferred on dirs. + * + * Note that tar/cpio do not require that archives be in a particular + * order; there is no way to know when the last file has been restored + * within a directory, so there's no way to optimize the memory usage + * here by fixing up the directory any earlier than the + * end-of-archive. + * + * XXX TODO: Directory ACLs should be restored here, for the same + * reason we set directory perms here. XXX + */ +static int +_archive_write_disk_close(struct archive *_a) +{ + struct archive_write_disk *a = (struct archive_write_disk *)_a; + struct fixup_entry *next, *p; + int ret; + + archive_check_magic(&a->archive, ARCHIVE_WRITE_DISK_MAGIC, + ARCHIVE_STATE_HEADER | ARCHIVE_STATE_DATA, + "archive_write_disk_close"); + ret = _archive_write_disk_finish_entry(&a->archive); + + /* Sort dir list so directories are fixed up in depth-first order. */ + p = sort_dir_list(a->fixup_list); + + while (p != NULL) { + a->pst = NULL; /* Mark stat cache as out-of-date. */ + if (p->fixup & TODO_TIMES) { + set_times(a, INVALID_HANDLE_VALUE, p->mode, p->name, + p->atime, p->atime_nanos, + p->birthtime, p->birthtime_nanos, + p->mtime, p->mtime_nanos, + p->ctime, p->ctime_nanos); + } + if (p->fixup & TODO_MODE_BASE) + la_chmod(p->name, p->mode); + if (p->fixup & TODO_ACLS) + set_acls(a, INVALID_HANDLE_VALUE, p->name, &p->acl); + next = p->next; + archive_acl_clear(&p->acl); + free(p->name); + free(p); + p = next; + } + a->fixup_list = NULL; + return (ret); +} + +static int +_archive_write_disk_free(struct archive *_a) +{ + struct archive_write_disk *a; + int ret; + if (_a == NULL) + return (ARCHIVE_OK); + archive_check_magic(_a, ARCHIVE_WRITE_DISK_MAGIC, + ARCHIVE_STATE_ANY | ARCHIVE_STATE_FATAL, "archive_write_disk_free"); + a = (struct archive_write_disk *)_a; + ret = _archive_write_disk_close(&a->archive); + archive_write_disk_set_group_lookup(&a->archive, NULL, NULL, NULL); + archive_write_disk_set_user_lookup(&a->archive, NULL, NULL, NULL); + if (a->entry) + archive_entry_free(a->entry); + archive_wstring_free(&a->_name_data); + archive_string_free(&a->archive.error_string); + archive_wstring_free(&a->path_safe); + a->archive.magic = 0; + __archive_clean(&a->archive); + free(a); + return (ret); +} + +/* + * Simple O(n log n) merge sort to order the fixup list. In + * particular, we want to restore dir timestamps depth-first. + */ +static struct fixup_entry * +sort_dir_list(struct fixup_entry *p) +{ + struct fixup_entry *a, *b, *t; + + if (p == NULL) + return (NULL); + /* A one-item list is already sorted. */ + if (p->next == NULL) + return (p); + + /* Step 1: split the list. */ + t = p; + a = p->next->next; + while (a != NULL) { + /* Step a twice, t once. */ + a = a->next; + if (a != NULL) + a = a->next; + t = t->next; + } + /* Now, t is at the mid-point, so break the list here. */ + b = t->next; + t->next = NULL; + a = p; + + /* Step 2: Recursively sort the two sub-lists. */ + a = sort_dir_list(a); + b = sort_dir_list(b); + + /* Step 3: Merge the returned lists. */ + /* Pick the first element for the merged list. */ + if (wcscmp(a->name, b->name) > 0) { + t = p = a; + a = a->next; + } else { + t = p = b; + b = b->next; + } + + /* Always put the later element on the list first. */ + while (a != NULL && b != NULL) { + if (wcscmp(a->name, b->name) > 0) { + t->next = a; + a = a->next; + } else { + t->next = b; + b = b->next; + } + t = t->next; + } + + /* Only one list is non-empty, so just splice it on. */ + if (a != NULL) + t->next = a; + if (b != NULL) + t->next = b; + + return (p); +} + +/* + * Returns a new, initialized fixup entry. + * + * TODO: Reduce the memory requirements for this list by using a tree + * structure rather than a simple list of names. + */ +static struct fixup_entry * +new_fixup(struct archive_write_disk *a, const wchar_t *pathname) +{ + struct fixup_entry *fe; + + fe = (struct fixup_entry *)calloc(1, sizeof(struct fixup_entry)); + if (fe == NULL) + return (NULL); + fe->next = a->fixup_list; + a->fixup_list = fe; + fe->fixup = 0; + fe->name = _wcsdup(pathname); + return (fe); +} + +/* + * Returns a fixup structure for the current entry. + */ +static struct fixup_entry * +current_fixup(struct archive_write_disk *a, const wchar_t *pathname) +{ + if (a->current_fixup == NULL) + a->current_fixup = new_fixup(a, pathname); + return (a->current_fixup); +} + +/* TODO: Make this work. */ +/* + * TODO: The deep-directory support bypasses this; disable deep directory + * support if we're doing symlink checks. + */ +/* + * TODO: Someday, integrate this with the deep dir support; they both + * scan the path and both can be optimized by comparing against other + * recent paths. + */ +/* TODO: Extend this to support symlinks on Windows Vista and later. */ +static int +check_symlinks(struct archive_write_disk *a) +{ + wchar_t *pn, *p; + wchar_t c; + int r; + BY_HANDLE_FILE_INFORMATION st; + mode_t st_mode; + + /* + * Guard against symlink tricks. Reject any archive entry whose + * destination would be altered by a symlink. + */ + /* Whatever we checked last time doesn't need to be re-checked. */ + pn = a->name; + p = a->path_safe.s; + while ((*pn != '\0') && (*p == *pn)) + ++p, ++pn; + c = pn[0]; + /* Keep going until we've checked the entire name. */ + while (pn[0] != '\0' && (pn[0] != '\\' || pn[1] != '\0')) { + /* Skip the next path element. */ + while (*pn != '\0' && *pn != '\\') + ++pn; + c = pn[0]; + pn[0] = '\0'; + /* Check that we haven't hit a symlink. */ + r = file_information(a, a->name, &st, &st_mode, 1); + if (r != 0) { + /* We've hit a dir that doesn't exist; stop now. */ + if (errno == ENOENT) + break; + } else if (S_ISLNK(st_mode)) { + if (c == '\0') { + /* + * Last element is symlink; remove it + * so we can overwrite it with the + * item being extracted. + */ + if (disk_unlink(a->name)) { + archive_set_error(&a->archive, errno, + "Could not remove symlink %ls", + a->name); + pn[0] = c; + return (ARCHIVE_FAILED); + } + a->pst = NULL; + /* + * Even if we did remove it, a warning + * is in order. The warning is silly, + * though, if we're just replacing one + * symlink with another symlink. + */ + if (!S_ISLNK(a->mode)) { + archive_set_error(&a->archive, 0, + "Removing symlink %ls", + a->name); + } + /* Symlink gone. No more problem! */ + pn[0] = c; + return (0); + } else if (a->flags & ARCHIVE_EXTRACT_UNLINK) { + /* User asked us to remove problems. */ + if (disk_unlink(a->name) != 0) { + archive_set_error(&a->archive, 0, + "Cannot remove intervening " + "symlink %ls", a->name); + pn[0] = c; + return (ARCHIVE_FAILED); + } + a->pst = NULL; + } else { + archive_set_error(&a->archive, 0, + "Cannot extract through symlink %ls", + a->name); + pn[0] = c; + return (ARCHIVE_FAILED); + } + } + } + pn[0] = c; + /* We've checked and/or cleaned the whole path, so remember it. */ + archive_wstrcpy(&a->path_safe, a->name); + return (ARCHIVE_OK); +} + +static int +guidword(wchar_t *p, int n) +{ + int i; + + for (i = 0; i < n; i++) { + if ((*p >= L'0' && *p <= L'9') || + (*p >= L'a' && *p <= L'f') || + (*p >= L'A' && *p <= L'F')) + p++; + else + return (-1); + } + return (0); +} + +/* + * Canonicalize the pathname. In particular, this strips duplicate + * '\' characters, '.' elements, and trailing '\'. It also raises an + * error for an empty path, a trailing '..' or (if _SECURE_NODOTDOT is + * set) any '..' in the path. + */ +static int +cleanup_pathname(struct archive_write_disk *a) +{ + wchar_t *dest, *src, *p, *top; + wchar_t separator = L'\0'; + + p = a->name; + if (*p == L'\0') { + archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, + "Invalid empty pathname"); + return (ARCHIVE_FAILED); + } + + /* Replace '/' by '\' */ + for (; *p != L'\0'; p++) { + if (*p == L'/') + *p = L'\\'; + } + p = a->name; + + /* Skip leading "\\.\" or "\\?\" or "\\?\UNC\" or + * "\\?\Volume{GUID}\" + * (absolute path prefixes used by Windows API) */ + if (p[0] == L'\\' && p[1] == L'\\' && + (p[2] == L'.' || p[2] == L'?') && p[3] == L'\\') + { + /* A path begin with "\\?\UNC\" */ + if (p[2] == L'?' && + (p[4] == L'U' || p[4] == L'u') && + (p[5] == L'N' || p[5] == L'n') && + (p[6] == L'C' || p[6] == L'c') && + p[7] == L'\\') + p += 8; + /* A path begin with "\\?\Volume{GUID}\" */ + else if (p[2] == L'?' && + (p[4] == L'V' || p[4] == L'v') && + (p[5] == L'O' || p[5] == L'o') && + (p[6] == L'L' || p[6] == L'l') && + (p[7] == L'U' || p[7] == L'u') && + (p[8] == L'M' || p[8] == L'm') && + (p[9] == L'E' || p[9] == L'e') && + p[10] == L'{') { + if (guidword(p+11, 8) == 0 && p[19] == L'-' && + guidword(p+20, 4) == 0 && p[24] == L'-' && + guidword(p+25, 4) == 0 && p[29] == L'-' && + guidword(p+30, 4) == 0 && p[34] == L'-' && + guidword(p+35, 12) == 0 && p[47] == L'}' && + p[48] == L'\\') + p += 49; + else + p += 4; + /* A path begin with "\\.\PhysicalDriveX" */ + } else if (p[2] == L'.' && + (p[4] == L'P' || p[4] == L'p') && + (p[5] == L'H' || p[5] == L'h') && + (p[6] == L'Y' || p[6] == L'y') && + (p[7] == L'S' || p[7] == L's') && + (p[8] == L'I' || p[8] == L'i') && + (p[9] == L'C' || p[9] == L'c') && + (p[9] == L'A' || p[9] == L'a') && + (p[9] == L'L' || p[9] == L'l') && + (p[9] == L'D' || p[9] == L'd') && + (p[9] == L'R' || p[9] == L'r') && + (p[9] == L'I' || p[9] == L'i') && + (p[9] == L'V' || p[9] == L'v') && + (p[9] == L'E' || p[9] == L'e') && + (p[10] >= L'0' && p[10] <= L'9') && + p[11] == L'\0') { + archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, + "Path is a physical drive name"); + return (ARCHIVE_FAILED); + } else + p += 4; + } + + /* Skip leading drive letter from archives created + * on Windows. */ + if (((p[0] >= L'a' && p[0] <= L'z') || + (p[0] >= L'A' && p[0] <= L'Z')) && + p[1] == L':') { + if (p[2] == L'\0') { + archive_set_error(&a->archive, ARCHIVE_ERRNO_MISC, + "Path is a drive name"); + return (ARCHIVE_FAILED); + } + if (p[2] == L'\\') + p += 2; + } + + top = dest = src = p; + /* Rewrite the path name if its character is a unusable. */ + for (; *p != L'\0'; p++) { + if (*p == L':' || *p == L'*' || *p == L'?' || *p == L'"' || + *p == L'<' || *p == L'>' || *p == L'|') + *p = L'_'; + } + /* Skip leading '\'. */ + if (*src == L'\\') + separator = *src++; + + /* Scan the pathname one element at a time. */ + for (;;) { + /* src points to first char after '\' */ + if (src[0] == L'\0') { + break; + } else if (src[0] == L'\\') { + /* Found '\\'('//'), ignore second one. */ + src++; + continue; + } else if (src[0] == L'.') { + if (src[1] == L'\0') { + /* Ignore trailing '.' */ + break; + } else if (src[1] == L'\\') { + /* Skip '.\'. */ + src += 2; + continue; + } else if (src[1] == L'.') { + if (src[2] == L'\\' || src[2] == L'\0') { + /* Conditionally warn about '..' */ + if (a->flags & + ARCHIVE_EXTRACT_SECURE_NODOTDOT) { + archive_set_error(&a->archive, + ARCHIVE_ERRNO_MISC, + "Path contains '..'"); + return (ARCHIVE_FAILED); + } + } + /* + * Note: Under no circumstances do we + * remove '..' elements. In + * particular, restoring + * '\foo\..\bar\' should create the + * 'foo' dir as a side-effect. + */ + } + } + + /* Copy current element, including leading '\'. */ + if (separator) + *dest++ = L'\\'; + while (*src != L'\0' && *src != L'\\') { + *dest++ = *src++; + } + + if (*src == L'\0') + break; + + /* Skip '\' separator. */ + separator = *src++; + } + /* + * We've just copied zero or more path elements, not including the + * final '\'. + */ + if (dest == top) { + /* + * Nothing got copied. The path must have been something + * like '.' or '\' or './' or '/././././/./'. + */ + if (separator) + *dest++ = L'\\'; + else + *dest++ = L'.'; + } + /* Terminate the result. */ + *dest = L'\0'; + return (ARCHIVE_OK); +} + +/* + * Create the parent directory of the specified path, assuming path + * is already in mutable storage. + */ +static int +create_parent_dir(struct archive_write_disk *a, wchar_t *path) +{ + wchar_t *slash; + int r; + + /* Remove tail element to obtain parent name. */ + slash = wcsrchr(path, L'\\'); + if (slash == NULL) + return (ARCHIVE_OK); + *slash = L'\0'; + r = create_dir(a, path); + *slash = L'\\'; + return (r); +} + +/* + * Create the specified dir, recursing to create parents as necessary. + * + * Returns ARCHIVE_OK if the path exists when we're done here. + * Otherwise, returns ARCHIVE_FAILED. + * Assumes path is in mutable storage; path is unchanged on exit. + */ +static int +create_dir(struct archive_write_disk *a, wchar_t *path) +{ + BY_HANDLE_FILE_INFORMATION st; + struct fixup_entry *le; + wchar_t *slash, *base, *full; + mode_t mode_final, mode, st_mode; + int r; + + /* Check for special names and just skip them. */ + slash = wcsrchr(path, L'\\'); + if (slash == NULL) + base = path; + else + base = slash + 1; + + if (base[0] == L'\0' || + (base[0] == L'.' && base[1] == L'\0') || + (base[0] == L'.' && base[1] == L'.' && base[2] == L'\0')) { + /* Don't bother trying to create null path, '.', or '..'. */ + if (slash != NULL) { + *slash = L'\0'; + r = create_dir(a, path); + *slash = L'\\'; + return (r); + } + return (ARCHIVE_OK); + } + + /* + * Yes, this should be stat() and not lstat(). Using lstat() + * here loses the ability to extract through symlinks. Also note + * that this should not use the a->st cache. + */ + if (file_information(a, path, &st, &st_mode, 0) == 0) { + if (S_ISDIR(st_mode)) + return (ARCHIVE_OK); + if ((a->flags & ARCHIVE_EXTRACT_NO_OVERWRITE)) { + archive_set_error(&a->archive, EEXIST, + "Can't create directory '%ls'", path); + return (ARCHIVE_FAILED); + } + if (disk_unlink(path) != 0) { + archive_set_error(&a->archive, errno, + "Can't create directory '%ls': " + "Conflicting file cannot be removed", + path); + return (ARCHIVE_FAILED); + } + } else if (errno != ENOENT && errno != ENOTDIR) { + /* Stat failed? */ + archive_set_error(&a->archive, errno, + "Can't test directory '%ls'", path); + return (ARCHIVE_FAILED); + } else if (slash != NULL) { + *slash = '\0'; + r = create_dir(a, path); + *slash = '\\'; + if (r != ARCHIVE_OK) + return (r); + } + + /* + * Mode we want for the final restored directory. Per POSIX, + * implicitly-created dirs must be created obeying the umask. + * There's no mention whether this is different for privileged + * restores (which the rest of this code handles by pretending + * umask=0). I've chosen here to always obey the user's umask for + * implicit dirs, even if _EXTRACT_PERM was specified. + */ + mode_final = DEFAULT_DIR_MODE & ~a->user_umask; + /* Mode we want on disk during the restore process. */ + mode = mode_final; + mode |= MINIMUM_DIR_MODE; + mode &= MAXIMUM_DIR_MODE; + /* + * Apply __la_win_permissive_name_w to path in order to + * remove '../' path string. + */ + full = __la_win_permissive_name_w(path); + if (full == NULL) + errno = EINVAL; + else if (CreateDirectoryW(full, NULL) != 0) { + if (mode != mode_final) { + le = new_fixup(a, path); + le->fixup |=TODO_MODE_BASE; + le->mode = mode_final; + } + free(full); + return (ARCHIVE_OK); + } else { + la_dosmaperr(GetLastError()); + } + free(full); + + /* + * Without the following check, a/b/../b/c/d fails at the + * second visit to 'b', so 'd' can't be created. Note that we + * don't add it to the fixup list here, as it's already been + * added. + */ + if (file_information(a, path, &st, &st_mode, 0) == 0 && + S_ISDIR(st_mode)) + return (ARCHIVE_OK); + + archive_set_error(&a->archive, errno, "Failed to create dir '%ls'", + path); + return (ARCHIVE_FAILED); +} + +/* + * Note: Although we can skip setting the user id if the desired user + * id matches the current user, we cannot skip setting the group, as + * many systems set the gid based on the containing directory. So + * we have to perform a chown syscall if we want to set the SGID + * bit. (The alternative is to stat() and then possibly chown(); it's + * more efficient to skip the stat() and just always chown().) Note + * that a successful chown() here clears the TODO_SGID_CHECK bit, which + * allows set_mode to skip the stat() check for the GID. + */ +static int +set_ownership(struct archive_write_disk *a) +{ +/* unfortunately, on win32 there is no 'root' user with uid 0, + so we just have to try the chown and see if it works */ + + /* If we know we can't change it, don't bother trying. */ + if (a->user_uid != 0 && a->user_uid != a->uid) { + archive_set_error(&a->archive, errno, + "Can't set UID=%jd", (intmax_t)a->uid); + return (ARCHIVE_WARN); + } + + archive_set_error(&a->archive, errno, + "Can't set user=%jd/group=%jd for %ls", + (intmax_t)a->uid, (intmax_t)a->gid, a->name); + return (ARCHIVE_WARN); +} + +static int +set_times(struct archive_write_disk *a, + HANDLE h, int mode, const wchar_t *name, + time_t atime, long atime_nanos, + time_t birthtime, long birthtime_nanos, + time_t mtime, long mtime_nanos, + time_t ctime_sec, long ctime_nanos) +{ +#define EPOC_TIME ARCHIVE_LITERAL_ULL(116444736000000000) +#define WINTIME(sec, nsec) ((Int32x32To64(sec, 10000000) + EPOC_TIME)\ + + (((nsec)/1000)*10)) + + HANDLE hw = 0; + ULARGE_INTEGER wintm; + FILETIME *pfbtime; + FILETIME fatime, fbtime, fmtime; + + (void)ctime_sec; /* UNUSED */ + (void)ctime_nanos; /* UNUSED */ + + if (h != INVALID_HANDLE_VALUE) { + hw = NULL; + } else { + wchar_t *ws; + + if (S_ISLNK(mode)) + return (ARCHIVE_OK); + ws = __la_win_permissive_name_w(name); + if (ws == NULL) + goto settimes_failed; + hw = CreateFileW(ws, FILE_WRITE_ATTRIBUTES, + 0, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); + free(ws); + if (hw == INVALID_HANDLE_VALUE) + goto settimes_failed; + h = hw; + } + + wintm.QuadPart = WINTIME(atime, atime_nanos); + fatime.dwLowDateTime = wintm.LowPart; + fatime.dwHighDateTime = wintm.HighPart; + wintm.QuadPart = WINTIME(mtime, mtime_nanos); + fmtime.dwLowDateTime = wintm.LowPart; + fmtime.dwHighDateTime = wintm.HighPart; + /* + * SetFileTime() supports birthtime. + */ + if (birthtime > 0 || birthtime_nanos > 0) { + wintm.QuadPart = WINTIME(birthtime, birthtime_nanos); + fbtime.dwLowDateTime = wintm.LowPart; + fbtime.dwHighDateTime = wintm.HighPart; + pfbtime = &fbtime; + } else + pfbtime = NULL; + if (SetFileTime(h, pfbtime, &fatime, &fmtime) == 0) + goto settimes_failed; + CloseHandle(hw); + return (ARCHIVE_OK); + +settimes_failed: + CloseHandle(hw); + archive_set_error(&a->archive, EINVAL, "Can't restore time"); + return (ARCHIVE_WARN); +} + +static int +set_times_from_entry(struct archive_write_disk *a) +{ + time_t atime, birthtime, mtime, ctime_sec; + long atime_nsec, birthtime_nsec, mtime_nsec, ctime_nsec; + + /* Suitable defaults. */ + atime = birthtime = mtime = ctime_sec = a->start_time; + atime_nsec = birthtime_nsec = mtime_nsec = ctime_nsec = 0; + + /* If no time was provided, we're done. */ + if (!archive_entry_atime_is_set(a->entry) + && !archive_entry_birthtime_is_set(a->entry) + && !archive_entry_mtime_is_set(a->entry)) + return (ARCHIVE_OK); + + if (archive_entry_atime_is_set(a->entry)) { + atime = archive_entry_atime(a->entry); + atime_nsec = archive_entry_atime_nsec(a->entry); + } + if (archive_entry_birthtime_is_set(a->entry)) { + birthtime = archive_entry_birthtime(a->entry); + birthtime_nsec = archive_entry_birthtime_nsec(a->entry); + } + if (archive_entry_mtime_is_set(a->entry)) { + mtime = archive_entry_mtime(a->entry); + mtime_nsec = archive_entry_mtime_nsec(a->entry); + } + if (archive_entry_ctime_is_set(a->entry)) { + ctime_sec = archive_entry_ctime(a->entry); + ctime_nsec = archive_entry_ctime_nsec(a->entry); + } + + return set_times(a, a->fh, a->mode, a->name, + atime, atime_nsec, + birthtime, birthtime_nsec, + mtime, mtime_nsec, + ctime_sec, ctime_nsec); +} + +static int +set_mode(struct archive_write_disk *a, int mode) +{ + int r = ARCHIVE_OK; + mode &= 07777; /* Strip off file type bits. */ + + if (a->todo & TODO_SGID_CHECK) { + /* + * If we don't know the GID is right, we must stat() + * to verify it. We can't just check the GID of this + * process, since systems sometimes set GID from + * the enclosing dir or based on ACLs. + */ + if ((r = lazy_stat(a)) != ARCHIVE_OK) + return (r); + if (0 != a->gid) { + mode &= ~ S_ISGID; + } + /* While we're here, double-check the UID. */ + if (0 != a->uid + && (a->todo & TODO_SUID)) { + mode &= ~ S_ISUID; + } + a->todo &= ~TODO_SGID_CHECK; + a->todo &= ~TODO_SUID_CHECK; + } else if (a->todo & TODO_SUID_CHECK) { + /* + * If we don't know the UID is right, we can just check + * the user, since all systems set the file UID from + * the process UID. + */ + if (a->user_uid != a->uid) { + mode &= ~ S_ISUID; + } + a->todo &= ~TODO_SUID_CHECK; + } + + if (S_ISLNK(a->mode)) { +#ifdef HAVE_LCHMOD + /* + * If this is a symlink, use lchmod(). If the + * platform doesn't support lchmod(), just skip it. A + * platform that doesn't provide a way to set + * permissions on symlinks probably ignores + * permissions on symlinks, so a failure here has no + * impact. + */ + if (lchmod(a->name, mode) != 0) { + archive_set_error(&a->archive, errno, + "Can't set permissions to 0%o", (int)mode); + r = ARCHIVE_WARN; + } +#endif + } else if (!S_ISDIR(a->mode)) { + /* + * If it's not a symlink and not a dir, then use + * fchmod() or chmod(), depending on whether we have + * an fd. Dirs get their perms set during the + * post-extract fixup, which is handled elsewhere. + */ +#ifdef HAVE_FCHMOD + if (a->fd >= 0) { + if (fchmod(a->fd, mode) != 0) { + archive_set_error(&a->archive, errno, + "Can't set permissions to 0%o", (int)mode); + r = ARCHIVE_WARN; + } + } else +#endif + /* If this platform lacks fchmod(), then + * we'll just use chmod(). */ + if (la_chmod(a->name, mode) != 0) { + archive_set_error(&a->archive, errno, + "Can't set permissions to 0%o", (int)mode); + r = ARCHIVE_WARN; + } + } + return (r); +} + +static int +set_fflags(struct archive_write_disk *a) +{ + (void)a; /* UNUSED */ + return (ARCHIVE_OK); +} + +/* Default empty function body to satisfy mainline code. */ +static int +set_acls(struct archive_write_disk *a, HANDLE h, const wchar_t *name, + struct archive_acl *acl) +{ + (void)a; /* UNUSED */ + (void)h; /* UNUSED */ + (void)name; /* UNUSED */ + (void)acl; /* UNUSED */ + return (ARCHIVE_OK); +} + +/* + * Restore extended attributes - stub implementation for unsupported systems + */ +static int +set_xattrs(struct archive_write_disk *a) +{ + static int warning_done = 0; + + /* If there aren't any extended attributes, then it's okay not + * to extract them, otherwise, issue a single warning. */ + if (archive_entry_xattr_count(a->entry) != 0 && !warning_done) { + warning_done = 1; + archive_set_error(&a->archive, ARCHIVE_ERRNO_FILE_FORMAT, + "Cannot restore extended attributes on this system"); + return (ARCHIVE_WARN); + } + /* Warning was already emitted; suppress further warnings. */ + return (ARCHIVE_OK); +} + +static void +fileTimeToUtc(const FILETIME *filetime, time_t *t, long *ns) +{ + ULARGE_INTEGER utc; + + utc.HighPart = filetime->dwHighDateTime; + utc.LowPart = filetime->dwLowDateTime; + if (utc.QuadPart >= EPOC_TIME) { + utc.QuadPart -= EPOC_TIME; + /* milli seconds base */ + *t = (time_t)(utc.QuadPart / 10000000); + /* nano seconds base */ + *ns = (long)(utc.QuadPart % 10000000) * 100; + } else { + *t = 0; + *ns = 0; + } +} +/* + * Test if file on disk is older than entry. + */ +static int +older(BY_HANDLE_FILE_INFORMATION *st, struct archive_entry *entry) +{ + time_t sec; + long nsec; + + fileTimeToUtc(&st->ftLastWriteTime, &sec, &nsec); + /* First, test the seconds and return if we have a definite answer. */ + /* Definitely older. */ + if (sec < archive_entry_mtime(entry)) + return (1); + /* Definitely younger. */ + if (sec > archive_entry_mtime(entry)) + return (0); + if (nsec < archive_entry_mtime_nsec(entry)) + return (1); + /* Same age or newer, so not older. */ + return (0); +} + +#endif /* _WIN32 && !__CYGWIN__ */ + diff --git a/libarchive/config_freebsd.h b/libarchive/config_freebsd.h new file mode 100644 index 000000000000..d61c4167b3b9 --- /dev/null +++ b/libarchive/config_freebsd.h @@ -0,0 +1,160 @@ +/*- + * Copyright (c) 2003-2007 Tim Kientzle + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + * $FreeBSD$ + */ + +/* FreeBSD 5.0 and later have ACL and extattr support. */ +#if __FreeBSD__ > 4 +#define HAVE_ACL_CREATE_ENTRY 1 +#define HAVE_ACL_GET_LINK_NP 1 +#define HAVE_ACL_GET_PERM_NP 1 +#define HAVE_ACL_INIT 1 +#define HAVE_ACL_SET_FD 1 +#define HAVE_ACL_SET_FD_NP 1 +#define HAVE_ACL_SET_FILE 1 +#define HAVE_ACL_USER 1 +#define HAVE_EXTATTR_GET_FILE 1 +#define HAVE_EXTATTR_LIST_FILE 1 +#define HAVE_EXTATTR_SET_FD 1 +#define HAVE_EXTATTR_SET_FILE 1 +#define HAVE_SYS_ACL_H 1 +#define HAVE_SYS_EXTATTR_H 1 +#endif + +#ifdef WITH_OPENSSL +#define HAVE_OPENSSL_MD5_H 1 +#define HAVE_OPENSSL_RIPEMD_H 1 +#define HAVE_OPENSSL_SHA_H 1 +#define HAVE_SHA384 1 +#define HAVE_SHA512 1 +#endif + +#define HAVE_BSDXML_H 1 +#define HAVE_BZLIB_H 1 +#define HAVE_CHFLAGS 1 +#define HAVE_CHOWN 1 +#define HAVE_DECL_INT64_MAX 1 +#define HAVE_DECL_INT64_MIN 1 +#define HAVE_DECL_SIZE_MAX 1 +#define HAVE_DECL_SSIZE_MAX 1 +#define HAVE_DECL_STRERROR_R 1 +#define HAVE_DECL_UINT32_MAX 1 +#define HAVE_DECL_UINT64_MAX 1 +#define HAVE_DIRENT_H 1 +#define HAVE_EFTYPE 1 +#define HAVE_EILSEQ 1 +#define HAVE_ERRNO_H 1 +#define HAVE_FCHDIR 1 +#define HAVE_FCHFLAGS 1 +#define HAVE_FCHMOD 1 +#define HAVE_FCHOWN 1 +#define HAVE_FCNTL 1 +#define HAVE_FCNTL_H 1 +#define HAVE_FSEEKO 1 +#define HAVE_FSTAT 1 +#define HAVE_FTRUNCATE 1 +#define HAVE_FUTIMES 1 +#define HAVE_GETEUID 1 +#define HAVE_GETGRGID_R 1 +#define HAVE_GETPID 1 +#define HAVE_GETPWUID_R 1 +#define HAVE_GRP_H 1 +#define HAVE_INTTYPES_H 1 +#define HAVE_LCHFLAGS 1 +#define HAVE_LCHMOD 1 +#define HAVE_LCHOWN 1 +#define HAVE_LIMITS_H 1 +#define HAVE_LINK 1 +#define HAVE_LSTAT 1 +#define HAVE_LUTIMES 1 +#define HAVE_MALLOC 1 +#define HAVE_MD5 1 +#define HAVE_MD5_H 1 +#define HAVE_MEMMOVE 1 +#define HAVE_MKDIR 1 +#define HAVE_MKFIFO 1 +#define HAVE_MKNOD 1 +#define HAVE_PIPE 1 +#define HAVE_POLL 1 +#define HAVE_POLL_H 1 +#define HAVE_PWD_H 1 +#define HAVE_READLINK 1 +#define HAVE_RMD160 1 +#define HAVE_SELECT 1 +#define HAVE_SETENV 1 +#define HAVE_SHA_H 1 +#define HAVE_SHA1 1 +#define HAVE_SHA256 1 +#define HAVE_SHA256_H 1 +#define HAVE_SIGNAL_H 1 +#define HAVE_STDINT_H 1 +#define HAVE_STDLIB_H 1 +#define HAVE_STRCHR 1 +#define HAVE_STRDUP 1 +#define HAVE_STRERROR 1 +#define HAVE_STRERROR_R 1 +#define HAVE_STRINGS_H 1 +#define HAVE_STRING_H 1 +#define HAVE_STRRCHR 1 +#define HAVE_STRUCT_STAT_ST_BLKSIZE 1 +#define HAVE_STRUCT_STAT_ST_BIRTHTIME 1 +#define HAVE_STRUCT_STAT_ST_BIRTHTIMESPEC_TV_NSEC 1 +#define HAVE_STRUCT_STAT_ST_FLAGS 1 +#define HAVE_STRUCT_STAT_ST_MTIMESPEC_TV_NSEC 1 +#define HAVE_STRUCT_TM_TM_GMTOFF 1 +#define HAVE_SYMLINK 1 +#define HAVE_SYS_CDEFS_H 1 +#define HAVE_SYS_IOCTL_H 1 +#define HAVE_SYS_MOUNT_H 1 +#define HAVE_SYS_PARAM_H 1 +#define HAVE_SYS_SELECT_H 1 +#define HAVE_SYS_STAT_H 1 +#define HAVE_SYS_TIME_H 1 +#define HAVE_SYS_TYPES_H 1 +#undef HAVE_SYS_UTIME_H +#define HAVE_SYS_UTSNAME_H 1 +#define HAVE_SYS_WAIT_H 1 +#define HAVE_TIMEGM 1 +#define HAVE_TZSET 1 +#define HAVE_UNISTD_H 1 +#define HAVE_UNSETENV 1 +#define HAVE_UTIME 1 +#define HAVE_UTIMES 1 +#define HAVE_UTIME_H 1 +#define HAVE_VFORK 1 +#define HAVE_WCHAR_H 1 +#define HAVE_WCSCPY 1 +#define HAVE_WCSLEN 1 +#define HAVE_WCTOMB 1 +#define HAVE_WMEMCMP 1 +#define HAVE_WMEMCPY 1 +#define HAVE_ZLIB_H 1 +#define TIME_WITH_SYS_TIME 1 + +/* FreeBSD 4 and earlier lack intmax_t/uintmax_t */ +#if __FreeBSD__ < 5 +#define intmax_t int64_t +#define uintmax_t uint64_t +#endif diff --git a/libarchive/filter_fork_windows.c b/libarchive/filter_fork_windows.c new file mode 100644 index 000000000000..fa59cc9e90ce --- /dev/null +++ b/libarchive/filter_fork_windows.c @@ -0,0 +1,190 @@ +/*- + * Copyright (c) 2009-2012 Michihiro NAKAJIMA + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + */ + +#include "archive_platform.h" + +#if defined(_WIN32) && !defined(__CYGWIN__) +#include "archive_cmdline_private.h" +#include "archive_string.h" + +#include "filter_fork.h" + +pid_t +__archive_create_child(const char *cmd, int *child_stdin, int *child_stdout) +{ + HANDLE childStdout[2], childStdin[2],childStderr; + SECURITY_ATTRIBUTES secAtts; + STARTUPINFO staInfo; + PROCESS_INFORMATION childInfo; + struct archive_string cmdline; + struct archive_string fullpath; + struct archive_cmdline *acmd; + char *arg0, *ext; + int i, l; + DWORD fl, fl_old; + + childStdout[0] = childStdout[1] = INVALID_HANDLE_VALUE; + childStdin[0] = childStdin[1] = INVALID_HANDLE_VALUE; + childStderr = INVALID_HANDLE_VALUE; + archive_string_init(&cmdline); + archive_string_init(&fullpath); + + acmd = __archive_cmdline_allocate(); + if (acmd == NULL) + goto fail; + if (__archive_cmdline_parse(acmd, cmd) != ARCHIVE_OK) + goto fail; + + /* + * Search the full path of 'path'. + * NOTE: This does not need if we give CreateProcessA 'path' as + * a part of the cmdline and give CreateProcessA NULL as first + * parameter, but I do not like that way. + */ + ext = strrchr(acmd->path, '.'); + if (ext == NULL || strlen(ext) > 4) + /* 'path' does not have a proper extension, so we have to + * give SearchPath() ".exe" as the extension. */ + ext = ".exe"; + else + ext = NULL;/* 'path' has an extension. */ + + fl = MAX_PATH; + do { + if (archive_string_ensure(&fullpath, fl) == NULL) + goto fail; + fl_old = fl; + fl = SearchPathA(NULL, acmd->path, ext, fl, fullpath.s, + &arg0); + } while (fl != 0 && fl > fl_old); + if (fl == 0) + goto fail; + + /* + * Make a command line. + */ + for (l = 0, i = 0; acmd->argv[i] != NULL; i++) { + if (i == 0) + continue; + l += (int)strlen(acmd->argv[i]) + 1; + } + if (archive_string_ensure(&cmdline, l + 1) == NULL) + goto fail; + for (i = 0; acmd->argv[i] != NULL; i++) { + if (i == 0) { + const char *p, *sp; + + if ((p = strchr(acmd->argv[i], '/')) != NULL || + (p = strchr(acmd->argv[i], '\\')) != NULL) + p++; + else + p = acmd->argv[i]; + if ((sp = strchr(p, ' ')) != NULL) + archive_strappend_char(&cmdline, '"'); + archive_strcat(&cmdline, p); + if (sp != NULL) + archive_strappend_char(&cmdline, '"'); + } else { + archive_strappend_char(&cmdline, ' '); + archive_strcat(&cmdline, acmd->argv[i]); + } + } + if (i <= 1) { + const char *sp; + + if ((sp = strchr(arg0, ' ')) != NULL) + archive_strappend_char(&cmdline, '"'); + archive_strcat(&cmdline, arg0); + if (sp != NULL) + archive_strappend_char(&cmdline, '"'); + } + + secAtts.nLength = sizeof(SECURITY_ATTRIBUTES); + secAtts.bInheritHandle = TRUE; + secAtts.lpSecurityDescriptor = NULL; + if (CreatePipe(&childStdout[0], &childStdout[1], &secAtts, 0) == 0) + goto fail; + if (!SetHandleInformation(childStdout[0], HANDLE_FLAG_INHERIT, 0)) + goto fail; + if (CreatePipe(&childStdin[0], &childStdin[1], &secAtts, 0) == 0) + goto fail; + if (!SetHandleInformation(childStdin[1], HANDLE_FLAG_INHERIT, 0)) + goto fail; + if (DuplicateHandle(GetCurrentProcess(), GetStdHandle(STD_ERROR_HANDLE), + GetCurrentProcess(), &childStderr, 0, TRUE, + DUPLICATE_SAME_ACCESS) == 0) + goto fail; + + memset(&staInfo, 0, sizeof(staInfo)); + staInfo.cb = sizeof(staInfo); + staInfo.hStdError = childStderr; + staInfo.hStdOutput = childStdout[1]; + staInfo.hStdInput = childStdin[0]; + staInfo.wShowWindow = SW_HIDE; + staInfo.dwFlags = STARTF_USESTDHANDLES | STARTF_USESHOWWINDOW; + if (CreateProcessA(fullpath.s, cmdline.s, NULL, NULL, TRUE, 0, + NULL, NULL, &staInfo, &childInfo) == 0) + goto fail; + WaitForInputIdle(childInfo.hProcess, INFINITE); + CloseHandle(childInfo.hProcess); + CloseHandle(childInfo.hThread); + + *child_stdout = _open_osfhandle((intptr_t)childStdout[0], _O_RDONLY); + *child_stdin = _open_osfhandle((intptr_t)childStdin[1], _O_WRONLY); + + CloseHandle(childStdout[1]); + CloseHandle(childStdin[0]); + + archive_string_free(&cmdline); + archive_string_free(&fullpath); + __archive_cmdline_free(acmd); + return (childInfo.dwProcessId); + +fail: + if (childStdout[0] != INVALID_HANDLE_VALUE) + CloseHandle(childStdout[0]); + if (childStdout[1] != INVALID_HANDLE_VALUE) + CloseHandle(childStdout[1]); + if (childStdin[0] != INVALID_HANDLE_VALUE) + CloseHandle(childStdin[0]); + if (childStdin[1] != INVALID_HANDLE_VALUE) + CloseHandle(childStdin[1]); + if (childStderr != INVALID_HANDLE_VALUE) + CloseHandle(childStderr); + archive_string_free(&cmdline); + archive_string_free(&fullpath); + __archive_cmdline_free(acmd); + return (-1); +} + +void +__archive_check_child(int in, int out) +{ + (void)in; /* UNUSED */ + (void)out; /* UNUSED */ + Sleep(100); +} + +#endif /* _WIN32 && !__CYGWIN__ */ diff --git a/libarchive/mtree.5 b/libarchive/mtree.5 new file mode 100644 index 000000000000..983fff723891 --- /dev/null +++ b/libarchive/mtree.5 @@ -0,0 +1,269 @@ +.\" Copyright (c) 1989, 1990, 1993 +.\" The Regents of the University of California. All rights reserved. +.\" +.\" Redistribution and use in source and binary forms, with or without +.\" modification, are permitted provided that the following conditions +.\" are met: +.\" 1. Redistributions of source code must retain the above copyright +.\" notice, this list of conditions and the following disclaimer. +.\" 2. Redistributions in binary form must reproduce the above copyright +.\" notice, this list of conditions and the following disclaimer in the +.\" documentation and/or other materials provided with the distribution. +.\" 4. Neither the name of the University nor the names of its contributors +.\" may be used to endorse or promote products derived from this software +.\" without specific prior written permission. +.\" +.\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND +.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +.\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE +.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS +.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) +.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT +.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY +.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF +.\" SUCH DAMAGE. +.\" +.\" From: @(#)mtree.8 8.2 (Berkeley) 12/11/93 +.\" $FreeBSD$ +.\" +.Dd May 6, 2008 +.Dt MTREE 5 +.Os +.Sh NAME +.Nm mtree +.Nd format of mtree dir hierarchy files +.Sh DESCRIPTION +The +.Nm +format is a textual format that describes a collection of filesystem objects. +Such files are typically used to create or verify directory hierarchies. +.Ss General Format +An +.Nm +file consists of a series of lines, each providing information +about a single filesystem object. +Leading whitespace is always ignored. +.Pp +When encoding file or pathnames, any backslash character or +character outside of the 95 printable ASCII characters must be +encoded as a a backslash followed by three +octal digits. +When reading mtree files, any appearance of a backslash +followed by three octal digits should be converted into the +corresponding character. +.Pp +Each line is interpreted independently as one of the following types: +.Bl -tag -width Cm +.It Signature +The first line of any mtree file must begin with +.Dq #mtree . +If a file contains any full path entries, the first line should +begin with +.Dq #mtree v2.0 , +otherwise, the first line should begin with +.Dq #mtree v1.0 . +.It Blank +Blank lines are ignored. +.It Comment +Lines beginning with +.Cm # +are ignored. +.It Special +Lines beginning with +.Cm / +are special commands that influence +the interpretation of later lines. +.It Relative +If the first whitespace-delimited word has no +.Cm / +characters, +it is the name of a file in the current directory. +Any relative entry that describes a directory changes the +current directory. +.It dot-dot +As a special case, a relative entry with the filename +.Pa .. +changes the current directory to the parent directory. +Options on dot-dot entries are always ignored. +.It Full +If the first whitespace-delimited word has a +.Cm / +character after +the first character, it is the pathname of a file relative to the +starting directory. +There can be multiple full entries describing the same file. +.El +.Pp +Some tools that process +.Nm +files may require that multiple lines describing the same file +occur consecutively. +It is not permitted for the same file to be mentioned using +both a relative and a full file specification. +.Ss Special commands +Two special commands are currently defined: +.Bl -tag -width Cm +.It Cm /set +This command defines default values for one or more keywords. +It is followed on the same line by one or more whitespace-separated +keyword definitions. +These definitions apply to all following files that do not specify +a value for that keyword. +.It Cm /unset +This command removes any default value set by a previous +.Cm /set +command. +It is followed on the same line by one or more keywords +separated by whitespace. +.El +.Ss Keywords +After the filename, a full or relative entry consists of zero +or more whitespace-separated keyword definitions. +Each such definition consists of a key from the following +list immediately followed by an '=' sign +and a value. +Software programs reading mtree files should warn about +unrecognized keywords. +.Pp +Currently supported keywords are as follows: +.Bl -tag -width Cm +.It Cm cksum +The checksum of the file using the default algorithm specified by +the +.Xr cksum 1 +utility. +.It Cm contents +The full pathname of a file that holds the contents of this file. +.It Cm flags +The file flags as a symbolic name. +See +.Xr chflags 1 +for information on these names. +If no flags are to be set the string +.Dq none +may be used to override the current default. +.It Cm gid +The file group as a numeric value. +.It Cm gname +The file group as a symbolic name. +.It Cm ignore +Ignore any file hierarchy below this file. +.It Cm link +The target of the symbolic link when type=link. +.It Cm md5 +The MD5 message digest of the file. +.It Cm md5digest +A synonym for +.Cm md5 . +.It Cm mode +The current file's permissions as a numeric (octal) or symbolic +value. +.It Cm nlink +The number of hard links the file is expected to have. +.It Cm nochange +Make sure this file or directory exists but otherwise ignore all attributes. +.It Cm ripemd160digest +The +.Tn RIPEMD160 +message digest of the file. +.It Cm rmd160 +A synonym for +.Cm ripemd160digest . +.It Cm rmd160digest +A synonym for +.Cm ripemd160digest . +.It Cm sha1 +The +.Tn FIPS +160-1 +.Pq Dq Tn SHA-1 +message digest of the file. +.It Cm sha1digest +A synonym for +.Cm sha1 . +.It Cm sha256 +The +.Tn FIPS +180-2 +.Pq Dq Tn SHA-256 +message digest of the file. +.It Cm sha256digest +A synonym for +.Cm sha256 . +.It Cm size +The size, in bytes, of the file. +.It Cm time +The last modification time of the file. +.It Cm type +The type of the file; may be set to any one of the following: +.Pp +.Bl -tag -width Cm -compact +.It Cm block +block special device +.It Cm char +character special device +.It Cm dir +directory +.It Cm fifo +fifo +.It Cm file +regular file +.It Cm link +symbolic link +.It Cm socket +socket +.El +.It Cm uid +The file owner as a numeric value. +.It Cm uname +The file owner as a symbolic name. +.El +.Pp +.Sh SEE ALSO +.Xr cksum 1 , +.Xr find 1 , +.Xr mtree 8 +.Sh BUGS +The +.Fx +implementation of mtree does not currently support +the +.Nm +2.0 +format. +The requirement for a +.Dq #mtree +signature line is new and not yet widely implemented. +.Sh HISTORY +The +.Nm +utility appeared in +.Bx 4.3 Reno . +The +.Tn MD5 +digest capability was added in +.Fx 2.1 , +in response to the widespread use of programs which can spoof +.Xr cksum 1 . +The +.Tn SHA-1 +and +.Tn RIPEMD160 +digests were added in +.Fx 4.0 , +as new attacks have demonstrated weaknesses in +.Tn MD5 . +The +.Tn SHA-256 +digest was added in +.Fx 6.0 . +Support for file flags was added in +.Fx 4.0 , +and mostly comes from +.Nx . +The +.Dq full +entry format was added by +.Nx . diff --git a/libarchive/test/.cvsignore b/libarchive/test/.cvsignore new file mode 100644 index 000000000000..b71f5a0dbd62 --- /dev/null +++ b/libarchive/test/.cvsignore @@ -0,0 +1,10 @@ +*.tar +*.tar.gz +*.tgz +*.zip +.depend +.deps +.dirstamp +archive.h +libarchive_test +list.h diff --git a/libarchive/test/CMakeLists.txt b/libarchive/test/CMakeLists.txt new file mode 100644 index 000000000000..d2eb2c2f39ef --- /dev/null +++ b/libarchive/test/CMakeLists.txt @@ -0,0 +1,256 @@ +############################################ +# +# How to build libarchive_test +# +############################################ +IF(ENABLE_TEST) + SET(libarchive_test_SOURCES + ../../test_utils/test_utils.c + main.c + read_open_memory.c + test.h + test_acl_freebsd_posix1e.c + test_acl_freebsd_nfs4.c + test_acl_nfs4.c + test_acl_pax.c + test_acl_posix1e.c + test_archive_api_feature.c + test_archive_clear_error.c + test_archive_cmdline.c + test_archive_crypto.c + test_archive_getdate.c + test_archive_match_owner.c + test_archive_match_path.c + test_archive_match_time.c + test_archive_pathmatch.c + test_archive_read_close_twice.c + test_archive_read_close_twice_open_fd.c + test_archive_read_close_twice_open_filename.c + test_archive_read_multiple_data_objects.c + test_archive_read_next_header_empty.c + test_archive_read_next_header_raw.c + test_archive_read_open2.c + test_archive_read_set_filter_option.c + test_archive_read_set_format_option.c + test_archive_read_set_option.c + test_archive_read_set_options.c + test_archive_read_support.c + test_archive_set_error.c + test_archive_string.c + test_archive_string_conversion.c + test_archive_write_add_filter_by_name.c + test_archive_write_set_filter_option.c + test_archive_write_set_format_by_name.c + test_archive_write_set_format_option.c + test_archive_write_set_option.c + test_archive_write_set_options.c + test_bad_fd.c + test_compat_bzip2.c + test_compat_cpio.c + test_compat_gtar.c + test_compat_gzip.c + test_compat_lzip.c + test_compat_lzma.c + test_compat_lzop.c + test_compat_mac.c + test_compat_pax_libarchive_2x.c + test_compat_solaris_tar_acl.c + test_compat_solaris_pax_sparse.c + test_compat_tar_hardlink.c + test_compat_uudecode.c + test_compat_xz.c + test_compat_zip.c + test_empty_write.c + test_entry.c + test_entry_strmode.c + test_extattr_freebsd.c + test_filter_count.c + test_fuzz.c + test_gnutar_filename_encoding.c + test_link_resolver.c + test_open_failure.c + test_open_fd.c + test_open_file.c + test_open_filename.c + test_pax_filename_encoding.c + test_read_data_large.c + test_read_disk.c + test_read_disk_directory_traversals.c + test_read_disk_entry_from_file.c + test_read_extract.c + test_read_file_nonexistent.c + test_read_filter_grzip.c + test_read_filter_lrzip.c + test_read_filter_lzop.c + test_read_filter_lzop_multiple_parts.c + test_read_filter_program.c + test_read_filter_program_signature.c + test_read_filter_uudecode.c + test_read_format_7zip.c + test_read_format_ar.c + test_read_format_cab.c + test_read_format_cab_filename.c + test_read_format_cpio_afio.c + test_read_format_cpio_bin.c + test_read_format_cpio_bin_Z.c + test_read_format_cpio_bin_be.c + test_read_format_cpio_bin_bz2.c + test_read_format_cpio_bin_gz.c + test_read_format_cpio_bin_lzip.c + test_read_format_cpio_bin_lzma.c + test_read_format_cpio_bin_xz.c + test_read_format_cpio_filename.c + test_read_format_cpio_odc.c + test_read_format_cpio_svr4_bzip2_rpm.c + test_read_format_cpio_svr4_gzip.c + test_read_format_cpio_svr4_gzip_rpm.c + test_read_format_cpio_svr4c_Z.c + test_read_format_empty.c + test_read_format_gtar_filename.c + test_read_format_gtar_gz.c + test_read_format_gtar_lzma.c + test_read_format_gtar_sparse.c + test_read_format_iso_Z.c + test_read_format_iso_multi_extent.c + test_read_format_iso_xorriso.c + test_read_format_isojoliet_bz2.c + test_read_format_isojoliet_long.c + test_read_format_isojoliet_rr.c + test_read_format_isojoliet_versioned.c + test_read_format_isorr_bz2.c + test_read_format_isorr_ce.c + test_read_format_isorr_new_bz2.c + test_read_format_isorr_rr_moved.c + test_read_format_isozisofs_bz2.c + test_read_format_lha.c + test_read_format_lha_filename.c + test_read_format_mtree.c + test_read_format_pax_bz2.c + test_read_format_rar.c + test_read_format_raw.c + test_read_format_tar.c + test_read_format_tar_empty_filename.c + test_read_format_tar_filename.c + test_read_format_tbz.c + test_read_format_tgz.c + test_read_format_tlz.c + test_read_format_txz.c + test_read_format_tz.c + test_read_format_ustar_filename.c + test_read_format_xar.c + test_read_format_zip.c + test_read_format_zip_comment_stored.c + test_read_format_zip_filename.c + test_read_format_zip_mac_metadata.c + test_read_format_zip_sfx.c + test_read_large.c + test_read_pax_truncated.c + test_read_position.c + test_read_set_format.c + test_read_truncated.c + test_read_truncated_filter.c + test_sparse_basic.c + test_tar_filenames.c + test_tar_large.c + test_ustar_filenames.c + test_ustar_filename_encoding.c + test_write_disk.c + test_write_disk_appledouble.c + test_write_disk_failures.c + test_write_disk_hardlink.c + test_write_disk_hfs_compression.c + test_write_disk_lookup.c + test_write_disk_mac_metadata.c + test_write_disk_no_hfs_compression.c + test_write_disk_perms.c + test_write_disk_secure.c + test_write_disk_sparse.c + test_write_disk_symlink.c + test_write_disk_times.c + test_write_filter_b64encode.c + test_write_filter_bzip2.c + test_write_filter_compress.c + test_write_filter_gzip.c + test_write_filter_gzip_timestamp.c + test_write_filter_lrzip.c + test_write_filter_lzip.c + test_write_filter_lzma.c + test_write_filter_lzop.c + test_write_filter_program.c + test_write_filter_uuencode.c + test_write_filter_xz.c + test_write_format_7zip.c + test_write_format_7zip_empty.c + test_write_format_7zip_large.c + test_write_format_ar.c + test_write_format_cpio.c + test_write_format_cpio_empty.c + test_write_format_cpio_newc.c + test_write_format_cpio_odc.c + test_write_format_gnutar.c + test_write_format_iso9660.c + test_write_format_iso9660_boot.c + test_write_format_iso9660_empty.c + test_write_format_iso9660_filename.c + test_write_format_iso9660_zisofs.c + test_write_format_mtree.c + test_write_format_mtree_absolute_path.c + test_write_format_mtree_classic.c + test_write_format_mtree_classic_indent.c + test_write_format_mtree_fflags.c + test_write_format_mtree_no_separator.c + test_write_format_mtree_quoted_filename.c + test_write_format_pax.c + test_write_format_shar_empty.c + test_write_format_tar.c + test_write_format_tar_empty.c + test_write_format_tar_sparse.c + test_write_format_tar_ustar.c + test_write_format_tar_v7tar.c + test_write_format_xar.c + test_write_format_xar_empty.c + test_write_format_zip.c + test_write_format_zip_empty.c + test_write_format_zip_no_compression.c + test_write_zip_set_compression_store.c + test_write_open_memory.c + test_zip_filename_encoding.c + ) + + # + # Register target + # + ADD_EXECUTABLE(libarchive_test ${libarchive_test_SOURCES}) + TARGET_LINK_LIBRARIES(libarchive_test archive_static ${ADDITIONAL_LIBS}) + SET_PROPERTY(TARGET libarchive_test PROPERTY COMPILE_DEFINITIONS + LIBARCHIVE_STATIC LIST_H) + + # + # Generate list.h by grepping DEFINE_TEST() lines out of the C sources. + # + GENERATE_LIST_H(${CMAKE_CURRENT_BINARY_DIR}/list.h + ${CMAKE_CURRENT_LIST_FILE} ${libarchive_test_SOURCES}) + SET_PROPERTY(DIRECTORY APPEND PROPERTY INCLUDE_DIRECTORIES + ${CMAKE_CURRENT_BINARY_DIR}) + + # list.h has a line DEFINE_TEST(testname) for every + # test. We can use that to define the tests for cmake by + # defining a DEFINE_TEST macro and reading list.h in. + MACRO (DEFINE_TEST _testname) + ADD_TEST( + NAME libarchive_${_testname} + COMMAND libarchive_test -vv + -r ${CMAKE_CURRENT_SOURCE_DIR} + ${_testname}) + ENDMACRO (DEFINE_TEST _testname) + + INCLUDE(${CMAKE_CURRENT_BINARY_DIR}/list.h) + INCLUDE_DIRECTORIES(${CMAKE_CURRENT_BINARY_DIR}) + INCLUDE_DIRECTORIES(${CMAKE_SOURCE_DIR}/test_utils) + + # Experimental new test handling + ADD_CUSTOM_TARGET(run_libarchive_test + COMMAND libarchive_test -r ${CMAKE_CURRENT_SOURCE_DIR}) + ADD_DEPENDENCIES(run_all_tests run_libarchive_test) +ENDIF(ENABLE_TEST) + |