X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fadd_image.c;h=89721052d3a6156effd77fa1f07d873409312e3c;hp=7a8609fb7084c21656cb5ac2e54a76f356cae3e3;hb=cbeb81145e2532bfcf3f5e58c6be000615f54953;hpb=61edb3642435e6c494d7c7f3038334dd196f58fd diff --git a/src/add_image.c b/src/add_image.c index 7a8609fb..89721052 100644 --- a/src/add_image.c +++ b/src/add_image.c @@ -23,1142 +23,432 @@ #include "config.h" -#if defined(__CYGWIN__) || defined(__WIN32__) -# include -# include -# include -# ifdef ERROR -# undef ERROR -# endif -# include "security.h" +#ifdef __WIN32__ +# include "win32.h" #else -# include -# include -# include "timestamp.h" +# include +# include +# include +# include "timestamp.h" #endif #include "wimlib_internal.h" #include "dentry.h" #include "lookup_table.h" #include "xml.h" +#include "security.h" + #include #include -#include #include +#include #include + #include #ifdef HAVE_ALLOCA_H -#include +# include #endif -#define WIMLIB_ADD_IMAGE_FLAG_ROOT 0x80000000 -#define WIMLIB_ADD_IMAGE_FLAG_SOURCE 0x40000000 - /* * Adds the dentry tree and security data for a new image to the image metadata * array of the WIMStruct. */ -int add_new_dentry_tree(WIMStruct *w, struct wim_dentry *root_dentry, - struct wim_security_data *sd) +static int +add_new_dentry_tree(WIMStruct *w, struct wim_dentry *root_dentry, + struct wim_security_data *sd) { - struct wim_lookup_table_entry *metadata_lte; - struct wim_image_metadata *imd; struct wim_image_metadata *new_imd; - - wimlib_assert(root_dentry != NULL); - - DEBUG("Reallocating image metadata array for image_count = %u", - w->hdr.image_count + 1); - imd = CALLOC((w->hdr.image_count + 1), sizeof(struct wim_image_metadata)); - - if (!imd) { - ERROR("Failed to allocate memory for new image metadata array"); - goto err; - } - - memcpy(imd, w->image_metadata, - w->hdr.image_count * sizeof(struct wim_image_metadata)); + struct wim_lookup_table_entry *metadata_lte; + int ret; metadata_lte = new_lookup_table_entry(); if (!metadata_lte) - goto err_free_imd; + return WIMLIB_ERR_NOMEM; metadata_lte->resource_entry.flags = WIM_RESHDR_FLAG_METADATA; - random_hash(metadata_lte->hash); - lookup_table_insert(w->lookup_table, metadata_lte); + metadata_lte->unhashed = 1; - new_imd = &imd[w->hdr.image_count]; + new_imd = new_image_metadata(); + if (!new_imd) { + free_lookup_table_entry(metadata_lte); + return WIMLIB_ERR_NOMEM; + } new_imd->root_dentry = root_dentry; new_imd->metadata_lte = metadata_lte; new_imd->security_data = sd; new_imd->modified = 1; - FREE(w->image_metadata); - w->image_metadata = imd; - w->hdr.image_count++; - return 0; -err_free_imd: - FREE(imd); -err: - return WIMLIB_ERR_NOMEM; + ret = append_image_metadata(w, new_imd); + if (ret) + put_image_metadata(new_imd, NULL); + return ret; } -#if defined(__CYGWIN__) || defined(__WIN32__) +#ifndef __WIN32__ -static u64 FILETIME_to_u64(const FILETIME *ft) +static int +unix_capture_regular_file(const char *path, + u64 size, + struct wim_inode *inode, + struct wim_lookup_table *lookup_table) { - return ((u64)ft->dwHighDateTime << 32) | (u64)ft->dwLowDateTime; -} - + inode->i_attributes = FILE_ATTRIBUTE_NORMAL; -static int build_dentry_tree(struct wim_dentry **root_ret, - const char *root_disk_path, - struct wim_lookup_table *lookup_table, - struct wim_security_data *sd, - const struct capture_config *config, - int add_image_flags, - wimlib_progress_func_t progress_func, - void *extra_arg); + /* Empty files do not have to have a lookup table entry. */ + if (size != 0) { + struct wim_lookup_table_entry *lte; + char *file_on_disk; -static int win32_get_short_name(struct wim_dentry *dentry, - const wchar_t *path_utf16) -{ - WIN32_FIND_DATAW dat; - if (FindFirstFileW(path_utf16, &dat) && - dat.cAlternateFileName[0] != L'\0') - { - size_t short_name_len = wcslen(dat.cAlternateFileName) * 2; - size_t n = short_name_len + sizeof(wchar_t); - dentry->short_name = MALLOC(n); - if (!dentry->short_name) + file_on_disk = STRDUP(path); + if (!file_on_disk) + return WIMLIB_ERR_NOMEM; + lte = new_lookup_table_entry(); + if (!lte) { + FREE(file_on_disk); return WIMLIB_ERR_NOMEM; - memcpy(dentry->short_name, dat.cAlternateFileName, n); - dentry->short_name_len = short_name_len; - } - return 0; -} - -static int win32_get_security_descriptor(struct wim_dentry *dentry, - struct sd_set *sd_set, - const wchar_t *path_utf16) -{ - SECURITY_INFORMATION requestedInformation; - DWORD lenNeeded = 0; - BOOL status; - DWORD err; - - requestedInformation = DACL_SECURITY_INFORMATION | - SACL_SECURITY_INFORMATION | - OWNER_SECURITY_INFORMATION | - GROUP_SECURITY_INFORMATION; - /* Request length of security descriptor */ - status = GetFileSecurityW(path_utf16, requestedInformation, - NULL, 0, &lenNeeded); - err = GetLastError(); - if (!status && err == ERROR_INSUFFICIENT_BUFFER) { - DWORD len = lenNeeded; - char buf[len]; - if (GetFileSecurityW(path_utf16, requestedInformation, - buf, len, &lenNeeded)) - { - int security_id = sd_set_add_sd(sd_set, buf, len); - if (security_id < 0) - return WIMLIB_ERR_NOMEM; - else { - dentry->d_inode->i_security_id = security_id; - return 0; - } - } else { - err = GetLastError(); - } - } - ERROR("Win32 API: Failed to read security descriptor of \"%ls\"", - path_utf16); - win32_error(err); - return WIMLIB_ERR_READ; -} - -/* Reads the directory entries of directory using a Win32 API and recursively - * calls build_dentry_tree() on them. */ -static int win32_recurse_directory(struct wim_dentry *root, - const char *root_disk_path, - struct wim_lookup_table *lookup_table, - struct wim_security_data *sd, - const struct capture_config *config, - int add_image_flags, - wimlib_progress_func_t progress_func, - struct sd_set *sd_set, - const wchar_t *path_utf16, - size_t path_utf16_nchars) -{ - WIN32_FIND_DATAW dat; - HANDLE hFind; - DWORD err; - int ret; - - { - /* Begin reading the directory by calling FindFirstFileW. - * Unlink UNIX opendir(), FindFirstFileW has file globbing built - * into it. But this isn't what we actually want, so just add a - * dummy glob to get all entries. */ - wchar_t pattern_buf[path_utf16_nchars + 3]; - memcpy(pattern_buf, path_utf16, - path_utf16_nchars * sizeof(wchar_t)); - pattern_buf[path_utf16_nchars] = L'/'; - pattern_buf[path_utf16_nchars + 1] = L'*'; - pattern_buf[path_utf16_nchars + 2] = L'\0'; - hFind = FindFirstFileW(pattern_buf, &dat); - } - if (hFind == INVALID_HANDLE_VALUE) { - err = GetLastError(); - if (err == ERROR_FILE_NOT_FOUND) { - return 0; - } else { - ERROR("Win32 API: Failed to read directory \"%s\"", - root_disk_path); - win32_error(err); - return WIMLIB_ERR_READ; - } - } - ret = 0; - do { - /* Skip . and .. entries */ - if (!(dat.cFileName[0] == L'.' && - (dat.cFileName[1] == L'\0' || - (dat.cFileName[1] == L'.' && dat.cFileName[2] == L'\0')))) - { - struct wim_dentry *child; - - char *utf8_name; - size_t utf8_name_nbytes; - ret = utf16_to_utf8((const char*)dat.cFileName, - wcslen(dat.cFileName) * sizeof(wchar_t), - &utf8_name, - &utf8_name_nbytes); - if (ret) - goto out_find_close; - - char name[strlen(root_disk_path) + 1 + utf8_name_nbytes + 1]; - sprintf(name, "%s/%s", root_disk_path, utf8_name); - FREE(utf8_name); - ret = build_dentry_tree(&child, name, lookup_table, - sd, config, add_image_flags, - progress_func, sd_set); - if (ret) - goto out_find_close; - if (child) - dentry_add_child(root, child); } - } while (FindNextFileW(hFind, &dat)); - err = GetLastError(); - if (err != ERROR_NO_MORE_FILES) { - ERROR("Win32 API: Failed to read directory \"%s\"", root_disk_path); - win32_error(err); - if (ret == 0) - ret = WIMLIB_ERR_READ; + lte->file_on_disk = file_on_disk; + lte->resource_location = RESOURCE_IN_FILE_ON_DISK; + lte->resource_entry.original_size = size; + lookup_table_insert_unhashed(lookup_table, lte, inode, 0); + inode->i_lte = lte; } -out_find_close: - FindClose(hFind); - return ret; + return 0; } -/* Load a reparse point into a WIM inode. It is just stored in memory. - * - * @hFile: Open handle to a reparse point, with permission to read the reparse - * data. - * - * @inode: WIM inode for the reparse point. - * - * @lookup_table: Stream lookup table for the WIM; an entry will be added to it - * for the reparse point unless an entry already exists for - * the exact same data stream. - * - * @path: External path to the parse point (UTF-8). Used for error messages - * only. - * - * Returns 0 on success; nonzero on failure. */ -static int win32_capture_reparse_point(HANDLE hFile, - struct wim_inode *inode, - struct wim_lookup_table *lookup_table, - const char *path) +static int +unix_build_dentry_tree_recursive(struct wim_dentry **root_ret, + char *path, + size_t path_len, + struct add_image_params *params); + +static int +unix_capture_directory(struct wim_dentry *dir_dentry, + char *path, + size_t path_len, + struct add_image_params *params) { - /* "Reparse point data, including the tag and optional GUID, - * cannot exceed 16 kilobytes." - MSDN */ - char reparse_point_buf[16 * 1024]; - DWORD bytesReturned; - - if (!DeviceIoControl(hFile, FSCTL_GET_REPARSE_POINT, - NULL, 0, reparse_point_buf, - sizeof(reparse_point_buf), &bytesReturned, NULL)) - { - DWORD err = GetLastError(); - ERROR("Win32 API: Failed to get reparse data of \"%s\"", path); - win32_error(err); - return WIMLIB_ERR_READ; - } - if (bytesReturned < 8) { - ERROR("Reparse data on \"%s\" is invalid", path); - return WIMLIB_ERR_READ; - } - inode->i_reparse_tag = *(u32*)reparse_point_buf; - return inode_add_ads_with_data(inode, "", - (const u8*)reparse_point_buf + 8, - bytesReturned - 8, lookup_table); -} -/* Calculate the SHA1 message digest of a Win32 data stream, which may be either - * an unnamed or named data stream. - * - * @path: Path to the file, with the stream noted at the end for named - * streams. UTF-16LE encoding. - * - * @hash: On success, the SHA1 message digest of the stream is written to - * this location. - * - * Returns 0 on success; nonzero on failure. - */ -static int win32_sha1sum(const wchar_t *path, u8 hash[SHA1_HASH_SIZE]) -{ - HANDLE hFile; - SHA_CTX ctx; - u8 buf[32768]; - DWORD bytesRead; + DIR *dir; + struct dirent *entry; + struct wim_dentry *child; int ret; - hFile = win32_open_file_readonly(path); - if (hFile == INVALID_HANDLE_VALUE) + dir_dentry->d_inode->i_attributes = FILE_ATTRIBUTE_DIRECTORY; + dir = opendir(path); + if (!dir) { + ERROR_WITH_ERRNO("Failed to open the directory `%s'", + path); return WIMLIB_ERR_OPEN; + } - sha1_init(&ctx); + /* Recurse on directory contents */ + ret = 0; for (;;) { - if (!ReadFile(hFile, buf, sizeof(buf), &bytesRead, NULL)) { - ret = WIMLIB_ERR_READ; - goto out_close_handle; - } - if (bytesRead == 0) + errno = 0; + entry = readdir(dir); + if (!entry) { + if (errno) { + ret = WIMLIB_ERR_READ; + ERROR_WITH_ERRNO("Error reading the " + "directory `%s'", path); + } break; - sha1_update(&ctx, buf, bytesRead); - } - ret = 0; - sha1_final(hash, &ctx); -out_close_handle: - CloseHandle(hFile); - return ret; -} - -/* Scans an unnamed or named stream of a Win32 file (not a reparse point - * stream); calculates its SHA1 message digest and either creates a `struct - * wim_lookup_table_entry' in memory for it, or uses an existing 'struct - * wim_lookup_table_entry' for an identical stream. - * - * @path_utf16: Path to the file (UTF-16LE). - * - * @path_utf16_nchars: Number of 2-byte characters in @path_utf16. - * - * @inode: WIM inode to save the stream into. - * - * @lookup_table: Stream lookup table for the WIM. - * - * @dat: A `WIN32_FIND_STREAM_DATA' structure that specifies the - * stream name. - * - * Returns 0 on success; nonzero on failure. - */ -static int win32_capture_stream(const wchar_t *path_utf16, - size_t path_utf16_nchars, - struct wim_inode *inode, - struct wim_lookup_table *lookup_table, - WIN32_FIND_STREAM_DATA *dat) -{ - struct wim_ads_entry *ads_entry; - u8 hash[SHA1_HASH_SIZE]; - struct wim_lookup_table_entry *lte; - int ret; - wchar_t *p, *colon; - bool is_named_stream; - wchar_t *spath; - size_t spath_nchars; - DWORD err; - - /* The stream name should be returned as :NAME:TYPE */ - p = dat->cStreamName; - if (*p != L':') - goto out_invalid_stream_name; - p += 1; - colon = wcschr(p, L':'); - if (colon == NULL) - goto out_invalid_stream_name; - - if (wcscmp(colon + 1, L"$DATA")) { - /* Not a DATA stream */ - ret = 0; - goto out; - } - - is_named_stream = (p != colon); - if (is_named_stream) { - /* Allocate an ADS entry for the named stream. */ - char *utf8_stream_name; - size_t utf8_stream_name_len; - ret = utf16_to_utf8((const char *)p, - (colon - p) * sizeof(wchar_t), - &utf8_stream_name, - &utf8_stream_name_len); - if (ret) - goto out; - ads_entry = inode_add_ads(inode, utf8_stream_name); - FREE(utf8_stream_name); - if (!ads_entry) { - ret = WIMLIB_ERR_NOMEM; - goto out; } - } - /* Create a UTF-16 string @spath that gives the filename, then a colon, - * then the stream name. Or, if it's an unnamed stream, just the - * filename. It is MALLOC()'ed so that it can be saved in the - * wim_lookup_table_entry if needed. */ - *colon = '\0'; - spath_nchars = path_utf16_nchars; - if (is_named_stream) - spath_nchars += colon - p + 1; - - spath = MALLOC((spath_nchars + 1) * sizeof(wchar_t)); - memcpy(spath, path_utf16, path_utf16_nchars * sizeof(wchar_t)); - if (is_named_stream) { - spath[path_utf16_nchars] = L':'; - memcpy(&spath[path_utf16_nchars + 1], p, (colon - p) * sizeof(wchar_t)); - } - spath[spath_nchars] = L'\0'; + if (entry->d_name[0] == '.' && (entry->d_name[1] == '\0' + || (entry->d_name[1] == '.' && entry->d_name[2] == '\0'))) + continue; - ret = win32_sha1sum(spath, hash); - if (ret) { - err = GetLastError(); - ERROR("Win32 API: Failed to read \"%ls\" to calculate SHA1sum", - path_utf16); - win32_error(err); - goto out_free_spath; - } + size_t name_len = strlen(entry->d_name); - lte = __lookup_resource(lookup_table, hash); - if (lte) { - /* Use existing wim_lookup_table_entry that has the same SHA1 - * message digest */ - lte->refcnt++; - } else { - /* Make a new new wim_lookup_table_entry */ - lte = new_lookup_table_entry(); - if (!lte) { - ret = WIMLIB_ERR_NOMEM; - goto out_free_spath; - } - lte->file_on_disk = (char*)spath; - spath = NULL; - lte->resource_location = RESOURCE_WIN32; - lte->resource_entry.original_size = (uint64_t)dat->StreamSize.QuadPart; - lte->resource_entry.size = (uint64_t)dat->StreamSize.QuadPart; - copy_hash(lte->hash, hash); - lookup_table_insert(lookup_table, lte); + path[path_len] = '/'; + memcpy(&path[path_len + 1], entry->d_name, name_len + 1); + ret = unix_build_dentry_tree_recursive(&child, + path, + path_len + 1 + name_len, + params); + if (ret) + break; + if (child) + dentry_add_child(dir_dentry, child); } - if (is_named_stream) - ads_entry->lte = lte; - else - inode->i_lte = lte; -out_free_spath: - FREE(spath); -out: + closedir(dir); return ret; -out_invalid_stream_name: - ERROR("Invalid stream name: \"%ls:%ls\"", path_utf16, dat->cStreamName); - ret = WIMLIB_ERR_READ; - goto out; } -/* Scans a Win32 file for unnamed and named data streams (not reparse point - * streams). - * - * @path_utf16: Path to the file (UTF-16LE). - * - * @path_utf16_nchars: Number of 2-byte characters in @path_utf16. - * - * @inode: WIM inode to save the stream into. - * - * @lookup_table: Stream lookup table for the WIM. - * - * Returns 0 on success; nonzero on failure. - */ -static int win32_capture_streams(const wchar_t *path_utf16, - size_t path_utf16_nchars, - struct wim_inode *inode, - struct wim_lookup_table *lookup_table) +static int +unix_capture_symlink(struct wim_dentry **root_p, + const char *path, + struct wim_inode *inode, + struct add_image_params *params) { - WIN32_FIND_STREAM_DATA dat; + char deref_name_buf[4096]; + ssize_t deref_name_len; int ret; - HANDLE hFind; - DWORD err; - - hFind = FindFirstStreamW(path_utf16, FindStreamInfoStandard, &dat, 0); - if (hFind == INVALID_HANDLE_VALUE) { - err = GetLastError(); - - /* Seems legal for this to return ERROR_HANDLE_EOF on reparse - * points and directories */ - if ((inode->i_attributes & - (FILE_ATTRIBUTE_REPARSE_POINT | FILE_ATTRIBUTE_DIRECTORY)) - && err == ERROR_HANDLE_EOF) + + inode->i_attributes = FILE_ATTRIBUTE_REPARSE_POINT; + inode->i_reparse_tag = WIM_IO_REPARSE_TAG_SYMLINK; + + /* The idea here is to call readlink() to get the UNIX target of + * the symbolic link, then turn the target into a reparse point + * data buffer that contains a relative or absolute symbolic + * link (NOT a junction point or *full* path symbolic link with + * drive letter). + */ + deref_name_len = readlink(path, deref_name_buf, + sizeof(deref_name_buf) - 1); + if (deref_name_len >= 0) { + char *dest = deref_name_buf; + + dest[deref_name_len] = '\0'; + DEBUG("Read symlink `%s'", dest); + + if ((params->add_image_flags & WIMLIB_ADD_IMAGE_FLAG_RPFIX) && + dest[0] == '/') { - return 0; - } else { - ERROR("Win32 API: Failed to look up data streams of \"%ls\"", - path_utf16); - win32_error(err); - return WIMLIB_ERR_READ; + dest = fixup_symlink(dest, + params->capture_root_ino, + params->capture_root_dev); + if (!dest) { + WARNING("Ignoring out of tree absolute symlink " + "\"%s\" -> \"%s\"\n" + " (Use --norpfix to capture " + "absolute symlinks as-is)", + path, deref_name_buf); + free_dentry(*root_p); + *root_p = NULL; + return 0; + } + inode->i_not_rpfixed = 0; } + ret = inode_set_symlink(inode, dest, + params->lookup_table, NULL); + if (ret == 0) { + /* Unfortunately, Windows seems to have the concept of + * "file" symbolic links as being different from + * "directory" symbolic links... so + * FILE_ATTRIBUTE_DIRECTORY needs to be set on the + * symbolic link if the *target* of the symbolic link is + * a directory. */ + struct stat stbuf; + if (stat(path, &stbuf) == 0 && S_ISDIR(stbuf.st_mode)) + inode->i_attributes |= FILE_ATTRIBUTE_DIRECTORY; + } + } else { + ERROR_WITH_ERRNO("Failed to read target of " + "symbolic link `%s'", path); + ret = WIMLIB_ERR_READLINK; } - do { - ret = win32_capture_stream(path_utf16, - path_utf16_nchars, - inode, lookup_table, - &dat); - if (ret) - goto out_find_close; - } while (FindNextStreamW(hFind, &dat)); - err = GetLastError(); - if (err != ERROR_HANDLE_EOF) { - ERROR("Win32 API: Error reading data streams from \"%ls\"", path_utf16); - win32_error(err); - ret = WIMLIB_ERR_READ; - } -out_find_close: - FindClose(hFind); return ret; } -#endif -/* - * build_dentry_tree(): - * Recursively builds a tree of WIM dentries from an on-disk directory - * tree. - * - * @root_ret: Place to return a pointer to the root of the dentry tree. Only - * modified if successful. Set to NULL if the file or directory was - * excluded from capture. - * - * @root_disk_path: The path to the root of the directory tree on disk (UTF-8). - * - * @lookup_table: The lookup table for the WIM file. For each file added to the - * dentry tree being built, an entry is added to the lookup table, - * unless an identical stream is already in the lookup table. - * These lookup table entries that are added point to the path of - * the file on disk. - * - * @sd: Ignored. (Security data only captured in NTFS mode.) - * - * @capture_config: - * Configuration for files to be excluded from capture. - * - * @add_flags: Bitwise or of WIMLIB_ADD_IMAGE_FLAG_* - * - * @extra_arg: Ignored in UNIX builds; used to pass sd_set pointer in Windows - * builds. - * - * @return: 0 on success, nonzero on failure. It is a failure if any of - * the files cannot be `stat'ed, or if any of the needed - * directories cannot be opened or read. Failure to add the files - * to the WIM may still occur later when trying to actually read - * the on-disk files during a call to wimlib_write() or - * wimlib_overwrite(). - */ -static int build_dentry_tree(struct wim_dentry **root_ret, - const char *root_disk_path, - struct wim_lookup_table *lookup_table, - struct wim_security_data *sd, - const struct capture_config *config, - int add_image_flags, - wimlib_progress_func_t progress_func, - void *extra_arg) +static int +unix_build_dentry_tree_recursive(struct wim_dentry **root_ret, + char *path, + size_t path_len, + struct add_image_params *params) { struct wim_dentry *root = NULL; int ret = 0; struct wim_inode *inode; - if (exclude_path(root_disk_path, config, true)) { - if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_ROOT) { - ERROR("Cannot exclude the root directory from capture"); - ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG; - goto out; - } - if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE) - && progress_func) + if (exclude_path(path, path_len, params->config, true)) { + if ((params->add_image_flags & WIMLIB_ADD_IMAGE_FLAG_EXCLUDE_VERBOSE) + && params->progress_func) { union wimlib_progress_info info; - info.scan.cur_path = root_disk_path; + info.scan.cur_path = path; info.scan.excluded = true; - progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info); + params->progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info); } goto out; } - if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE) - && progress_func) + if ((params->add_image_flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE) + && params->progress_func) { union wimlib_progress_info info; - info.scan.cur_path = root_disk_path; + info.scan.cur_path = path; info.scan.excluded = false; - progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info); + params->progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info); } -#if !defined(__CYGWIN__) && !defined(__WIN32__) - /* UNIX version of capturing a directory tree */ - struct stat root_stbuf; + struct stat stbuf; int (*stat_fn)(const char *restrict, struct stat *restrict); - if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE) + if ((params->add_image_flags & WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE) || + (params->add_image_flags & WIMLIB_ADD_IMAGE_FLAG_ROOT)) stat_fn = stat; else stat_fn = lstat; - ret = (*stat_fn)(root_disk_path, &root_stbuf); + ret = (*stat_fn)(path, &stbuf); if (ret != 0) { - ERROR_WITH_ERRNO("Failed to stat `%s'", root_disk_path); + ERROR_WITH_ERRNO("Failed to stat `%s'", path); goto out; } - - if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_ROOT) && - !S_ISDIR(root_stbuf.st_mode)) - { - /* Do a dereference-stat in case the root is a symbolic link. - * This case is allowed, provided that the symbolic link points - * to a directory. */ - ret = stat(root_disk_path, &root_stbuf); - if (ret != 0) { - ERROR_WITH_ERRNO("Failed to stat `%s'", root_disk_path); - ret = WIMLIB_ERR_STAT; - goto out; - } - if (!S_ISDIR(root_stbuf.st_mode)) { - ERROR("`%s' is not a directory", root_disk_path); - ret = WIMLIB_ERR_NOTDIR; - goto out; - } - } - if (!S_ISREG(root_stbuf.st_mode) && !S_ISDIR(root_stbuf.st_mode) - && !S_ISLNK(root_stbuf.st_mode)) { + if (!S_ISREG(stbuf.st_mode) && !S_ISDIR(stbuf.st_mode) + && !S_ISLNK(stbuf.st_mode)) { ERROR("`%s' is not a regular file, directory, or symbolic link.", - root_disk_path); + path); ret = WIMLIB_ERR_SPECIAL_FILE; goto out; } - root = new_dentry_with_timeless_inode(path_basename(root_disk_path)); - if (!root) { - if (errno == EILSEQ) - ret = WIMLIB_ERR_INVALID_UTF8_STRING; - else if (errno == ENOMEM) - ret = WIMLIB_ERR_NOMEM; - else - ret = WIMLIB_ERR_ICONV_NOT_AVAILABLE; + ret = inode_table_new_dentry(params->inode_table, + path_basename_with_len(path, path_len), + stbuf.st_ino, + stbuf.st_dev, + &root); + if (ret) goto out; - } inode = root->d_inode; + if (inode->i_nlink > 1) /* Already captured this inode? */ + goto out; + #ifdef HAVE_STAT_NANOSECOND_PRECISION - inode->i_creation_time = timespec_to_wim_timestamp(&root_stbuf.st_mtim); - inode->i_last_write_time = timespec_to_wim_timestamp(&root_stbuf.st_mtim); - inode->i_last_access_time = timespec_to_wim_timestamp(&root_stbuf.st_atim); + inode->i_creation_time = timespec_to_wim_timestamp(stbuf.st_mtim); + inode->i_last_write_time = timespec_to_wim_timestamp(stbuf.st_mtim); + inode->i_last_access_time = timespec_to_wim_timestamp(stbuf.st_atim); #else - inode->i_creation_time = unix_timestamp_to_wim(root_stbuf.st_mtime); - inode->i_last_write_time = unix_timestamp_to_wim(root_stbuf.st_mtime); - inode->i_last_access_time = unix_timestamp_to_wim(root_stbuf.st_atime); + inode->i_creation_time = unix_timestamp_to_wim(stbuf.st_mtime); + inode->i_last_write_time = unix_timestamp_to_wim(stbuf.st_mtime); + inode->i_last_access_time = unix_timestamp_to_wim(stbuf.st_atime); #endif - if (sizeof(ino_t) >= 8) - inode->i_ino = (u64)root_stbuf.st_ino; - else - inode->i_ino = (u64)root_stbuf.st_ino | - ((u64)root_stbuf.st_dev << ((sizeof(ino_t) * 8) & 63)); inode->i_resolved = 1; - if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_UNIX_DATA) { - ret = inode_set_unix_data(inode, root_stbuf.st_uid, - root_stbuf.st_gid, - root_stbuf.st_mode, - lookup_table, + if (params->add_image_flags & WIMLIB_ADD_IMAGE_FLAG_UNIX_DATA) { + ret = inode_set_unix_data(inode, stbuf.st_uid, + stbuf.st_gid, + stbuf.st_mode, + params->lookup_table, UNIX_DATA_ALL | UNIX_DATA_CREATE); if (ret) goto out; } - add_image_flags &= ~(WIMLIB_ADD_IMAGE_FLAG_ROOT | WIMLIB_ADD_IMAGE_FLAG_SOURCE); - if (S_ISREG(root_stbuf.st_mode)) { /* Archiving a regular file */ - - struct wim_lookup_table_entry *lte; - u8 hash[SHA1_HASH_SIZE]; - - inode->i_attributes = FILE_ATTRIBUTE_NORMAL; - - /* Empty files do not have to have a lookup table entry. */ - if (root_stbuf.st_size == 0) - goto out; - - /* For each regular file, we must check to see if the file is in - * the lookup table already; if it is, we increment its refcnt; - * otherwise, we create a new lookup table entry and insert it. - * */ - - ret = sha1sum(root_disk_path, hash); - if (ret != 0) - goto out; - - lte = __lookup_resource(lookup_table, hash); - if (lte) { - lte->refcnt++; - DEBUG("Add lte reference %u for `%s'", lte->refcnt, - root_disk_path); - } else { - char *file_on_disk = STRDUP(root_disk_path); - if (!file_on_disk) { - ERROR("Failed to allocate memory for file path"); - ret = WIMLIB_ERR_NOMEM; - goto out; - } - lte = new_lookup_table_entry(); - if (!lte) { - FREE(file_on_disk); - ret = WIMLIB_ERR_NOMEM; - goto out; - } - lte->file_on_disk = file_on_disk; - lte->resource_location = RESOURCE_IN_FILE_ON_DISK; - lte->resource_entry.original_size = root_stbuf.st_size; - lte->resource_entry.size = root_stbuf.st_size; - copy_hash(lte->hash, hash); - lookup_table_insert(lookup_table, lte); - } - root->d_inode->i_lte = lte; - } else if (S_ISDIR(root_stbuf.st_mode)) { /* Archiving a directory */ - - inode->i_attributes = FILE_ATTRIBUTE_DIRECTORY; - - DIR *dir; - struct dirent entry, *result; - struct wim_dentry *child; - - dir = opendir(root_disk_path); - if (!dir) { - ERROR_WITH_ERRNO("Failed to open the directory `%s'", - root_disk_path); - ret = WIMLIB_ERR_OPEN; - goto out; - } - - /* Buffer for names of files in directory. */ - size_t len = strlen(root_disk_path); - char name[len + 1 + FILENAME_MAX + 1]; - memcpy(name, root_disk_path, len); - name[len] = '/'; - - /* Create a dentry for each entry in the directory on disk, and recurse - * to any subdirectories. */ - while (1) { - errno = 0; - ret = readdir_r(dir, &entry, &result); - if (ret != 0) { - ret = WIMLIB_ERR_READ; - ERROR_WITH_ERRNO("Error reading the " - "directory `%s'", - root_disk_path); - break; - } - if (result == NULL) - break; - if (result->d_name[0] == '.' && (result->d_name[1] == '\0' - || (result->d_name[1] == '.' && result->d_name[2] == '\0'))) - continue; - strcpy(name + len + 1, result->d_name); - ret = build_dentry_tree(&child, name, lookup_table, - NULL, config, add_image_flags, - progress_func, NULL); - if (ret != 0) - break; - if (child) - dentry_add_child(root, child); - } - closedir(dir); - } else { /* Archiving a symbolic link */ - inode->i_attributes = FILE_ATTRIBUTE_REPARSE_POINT; - inode->i_reparse_tag = WIM_IO_REPARSE_TAG_SYMLINK; - - /* The idea here is to call readlink() to get the UNIX target of - * the symbolic link, then turn the target into a reparse point - * data buffer that contains a relative or absolute symbolic - * link (NOT a junction point or *full* path symbolic link with - * drive letter). - */ - - char deref_name_buf[4096]; - ssize_t deref_name_len; - - deref_name_len = readlink(root_disk_path, deref_name_buf, - sizeof(deref_name_buf) - 1); - if (deref_name_len >= 0) { - deref_name_buf[deref_name_len] = '\0'; - DEBUG("Read symlink `%s'", deref_name_buf); - ret = inode_set_symlink(root->d_inode, deref_name_buf, - lookup_table, NULL); - if (ret == 0) { - /* - * Unfortunately, Windows seems to have the - * concept of "file" symbolic links as being - * different from "directory" symbolic links... - * so FILE_ATTRIBUTE_DIRECTORY needs to be set - * on the symbolic link if the *target* of the - * symbolic link is a directory. - */ - struct stat stbuf; - if (stat(root_disk_path, &stbuf) == 0 && - S_ISDIR(stbuf.st_mode)) - { - inode->i_attributes |= FILE_ATTRIBUTE_DIRECTORY; - } - } - } else { - ERROR_WITH_ERRNO("Failed to read target of " - "symbolic link `%s'", root_disk_path); - ret = WIMLIB_ERR_READLINK; - } - } -#else - /* Win32 version of capturing a directory tree */ - - wchar_t *path_utf16; - size_t path_utf16_nchars; - struct sd_set *sd_set; - DWORD err; - - if (extra_arg == NULL) { - sd_set = alloca(sizeof(struct sd_set)); - sd_set->rb_root.rb_node = NULL, - sd_set->sd = sd; - } else { - sd_set = extra_arg; - } - - ret = utf8_to_utf16(root_disk_path, strlen(root_disk_path), - (char**)&path_utf16, &path_utf16_nchars); - if (ret) - goto out; - path_utf16_nchars /= sizeof(wchar_t); - - HANDLE hFile = win32_open_file_readonly(path_utf16); - if (hFile == INVALID_HANDLE_VALUE) { - err = GetLastError(); - ERROR("Win32 API: Failed to open \"%s\"", root_disk_path); - win32_error(err); - ret = WIMLIB_ERR_OPEN; - goto out_free_path_utf16; - } - - BY_HANDLE_FILE_INFORMATION file_info; - if (!GetFileInformationByHandle(hFile, &file_info)) { - err = GetLastError(); - ERROR("Win32 API: Failed to get file information for \"%s\"", - root_disk_path); - win32_error(err); - ret = WIMLIB_ERR_STAT; - goto out_close_handle; - } - - /* Create a WIM dentry */ - root = new_dentry_with_timeless_inode(path_basename(root_disk_path)); - if (!root) { - if (errno == EILSEQ) - ret = WIMLIB_ERR_INVALID_UTF8_STRING; - else if (errno == ENOMEM) - ret = WIMLIB_ERR_NOMEM; - else - ret = WIMLIB_ERR_ICONV_NOT_AVAILABLE; - goto out_close_handle; - } - - /* Start preparing the associated WIM inode */ - inode = root->d_inode; - - inode->i_attributes = file_info.dwFileAttributes; - inode->i_creation_time = FILETIME_to_u64(&file_info.ftCreationTime); - inode->i_last_write_time = FILETIME_to_u64(&file_info.ftLastWriteTime); - inode->i_last_access_time = FILETIME_to_u64(&file_info.ftLastAccessTime); - inode->i_ino = ((u64)file_info.nFileIndexHigh << 32) | - (u64)file_info.nFileIndexLow; - - inode->i_resolved = 1; - add_image_flags &= ~(WIMLIB_ADD_IMAGE_FLAG_ROOT | WIMLIB_ADD_IMAGE_FLAG_SOURCE); - - /* Get DOS name and security descriptor (if any). */ - ret = win32_get_short_name(root, path_utf16); - if (ret) - goto out_close_handle; - ret = win32_get_security_descriptor(root, sd_set, path_utf16); - if (ret) - goto out_close_handle; - - if (inode_is_directory(inode)) { - /* Directory (not a reparse point) --- recurse to children */ - - /* But first... directories may have alternate data streams that - * need to be captured. */ - ret = win32_capture_streams(path_utf16, - path_utf16_nchars, - inode, - lookup_table); - if (ret) - goto out_close_handle; - ret = win32_recurse_directory(root, - root_disk_path, - lookup_table, - sd, - config, - add_image_flags, - progress_func, - sd_set, - path_utf16, - path_utf16_nchars); - } else if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) { - /* Reparse point: save the reparse tag and data */ - ret = win32_capture_reparse_point(hFile, - inode, - lookup_table, - root_disk_path); - } else { - /* Not a directory, not a reparse point; capture the default - * file contents and any alternate data streams. */ - ret = win32_capture_streams(path_utf16, - path_utf16_nchars, - inode, - lookup_table); - } -out_close_handle: - CloseHandle(hFile); -out_destroy_sd_set: - if (extra_arg == NULL) - destroy_sd_set(sd_set); -out_free_path_utf16: - FREE(path_utf16); -#endif - /* The below lines of code are common to both UNIX and Win32 builds. It - * simply returns the captured directory tree if the capture was - * successful, or frees it if the capture was unsuccessful. */ + params->add_image_flags &= + ~(WIMLIB_ADD_IMAGE_FLAG_ROOT | WIMLIB_ADD_IMAGE_FLAG_SOURCE); + if (S_ISREG(stbuf.st_mode)) + ret = unix_capture_regular_file(path, stbuf.st_size, + inode, params->lookup_table); + else if (S_ISDIR(stbuf.st_mode)) + ret = unix_capture_directory(root, path, path_len, params); + else + ret = unix_capture_symlink(&root, path, inode, params); out: if (ret == 0) *root_ret = root; else - free_dentry_tree(root, lookup_table); + free_dentry_tree(root, params->lookup_table); return ret; } -enum pattern_type { - NONE = 0, - EXCLUSION_LIST, - EXCLUSION_EXCEPTION, - COMPRESSION_EXCLUSION_LIST, - ALIGNMENT_LIST, -}; - -#define COMPAT_DEFAULT_CONFIG - -/* Default capture configuration file when none is specified. */ -static const char *default_config = -#ifdef COMPAT_DEFAULT_CONFIG /* XXX: This policy is being moved to library - users. The next ABI-incompatible library - version will default to the empty string here. */ -"[ExclusionList]\n" -"\\$ntfs.log\n" -"\\hiberfil.sys\n" -"\\pagefile.sys\n" -"\\System Volume Information\n" -"\\RECYCLER\n" -"\\Windows\\CSC\n" -"\n" -"[CompressionExclusionList]\n" -"*.mp3\n" -"*.zip\n" -"*.cab\n" -"\\WINDOWS\\inf\\*.pnf\n"; -#else -""; -#endif - -static void destroy_pattern_list(struct pattern_list *list) -{ - FREE(list->pats); -} - -static void destroy_capture_config(struct capture_config *config) -{ - destroy_pattern_list(&config->exclusion_list); - destroy_pattern_list(&config->exclusion_exception); - destroy_pattern_list(&config->compression_exclusion_list); - destroy_pattern_list(&config->alignment_list); - FREE(config->config_str); - FREE(config->prefix); - memset(config, 0, sizeof(*config)); -} - -static int pattern_list_add_pattern(struct pattern_list *list, - const char *pattern) -{ - const char **pats; - if (list->num_pats >= list->num_allocated_pats) { - pats = REALLOC(list->pats, - sizeof(list->pats[0]) * (list->num_allocated_pats + 8)); - if (!pats) - return WIMLIB_ERR_NOMEM; - list->num_allocated_pats += 8; - list->pats = pats; - } - list->pats[list->num_pats++] = pattern; - return 0; -} - -/* Parses the contents of the image capture configuration file and fills in a - * `struct capture_config'. */ -static int init_capture_config(struct capture_config *config, - const char *_config_str, size_t config_len) +/* + * unix_build_dentry_tree(): + * Builds a tree of WIM dentries from an on-disk directory tree (UNIX + * version; no NTFS-specific data is captured). + * + * @root_ret: Place to return a pointer to the root of the dentry tree. Only + * modified if successful. Set to NULL if the file or directory was + * excluded from capture. + * + * @root_disk_path: The path to the root of the directory tree on disk. + * + * @params: See doc for `struct add_image_params'. + * + * @return: 0 on success, nonzero on failure. It is a failure if any of + * the files cannot be `stat'ed, or if any of the needed + * directories cannot be opened or read. Failure to add the files + * to the WIM may still occur later when trying to actually read + * the on-disk files during a call to wimlib_write() or + * wimlib_overwrite(). + */ +static int +unix_build_dentry_tree(struct wim_dentry **root_ret, + const char *root_disk_path, + struct add_image_params *params) { - char *config_str; - char *p; - char *eol; - char *next_p; - size_t bytes_remaining; - enum pattern_type type = NONE; + char *path_buf; int ret; - unsigned long line_no = 0; - - DEBUG("config_len = %zu", config_len); - bytes_remaining = config_len; - memset(config, 0, sizeof(*config)); - config_str = MALLOC(config_len); - if (!config_str) { - ERROR("Could not duplicate capture config string"); - return WIMLIB_ERR_NOMEM; - } + size_t path_len; + size_t path_bufsz; - memcpy(config_str, _config_str, config_len); - next_p = config_str; - config->config_str = config_str; - while (bytes_remaining) { - line_no++; - p = next_p; - eol = memchr(p, '\n', bytes_remaining); - if (!eol) { - ERROR("Expected end-of-line in capture config file on " - "line %lu", line_no); - ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG; - goto out_destroy; + { + struct stat root_stbuf; + if (stat(root_disk_path, &root_stbuf)) { + ERROR_WITH_ERRNO("Failed to stat \"%s\"", root_disk_path); + return WIMLIB_ERR_STAT; } - next_p = eol + 1; - bytes_remaining -= (next_p - p); - if (eol == p) - continue; - - if (*(eol - 1) == '\r') - eol--; - *eol = '\0'; - - /* Translate backslash to forward slash */ - for (char *pp = p; pp != eol; pp++) - if (*pp == '\\') - *pp = '/'; - - /* Remove drive letter */ - if (eol - p > 2 && isalpha(*p) && *(p + 1) == ':') - p += 2; - - ret = 0; - if (strcmp(p, "[ExclusionList]") == 0) - type = EXCLUSION_LIST; - else if (strcmp(p, "[ExclusionException]") == 0) - type = EXCLUSION_EXCEPTION; - else if (strcmp(p, "[CompressionExclusionList]") == 0) - type = COMPRESSION_EXCLUSION_LIST; - else if (strcmp(p, "[AlignmentList]") == 0) - type = ALIGNMENT_LIST; - else if (p[0] == '[' && strrchr(p, ']')) { - ERROR("Unknown capture configuration section `%s'", p); - ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG; - } else switch (type) { - case EXCLUSION_LIST: - DEBUG("Adding pattern \"%s\" to exclusion list", p); - ret = pattern_list_add_pattern(&config->exclusion_list, p); - break; - case EXCLUSION_EXCEPTION: - DEBUG("Adding pattern \"%s\" to exclusion exception list", p); - ret = pattern_list_add_pattern(&config->exclusion_exception, p); - break; - case COMPRESSION_EXCLUSION_LIST: - DEBUG("Adding pattern \"%s\" to compression exclusion list", p); - ret = pattern_list_add_pattern(&config->compression_exclusion_list, p); - break; - case ALIGNMENT_LIST: - DEBUG("Adding pattern \"%s\" to alignment list", p); - ret = pattern_list_add_pattern(&config->alignment_list, p); - break; - default: - ERROR("Line %lu of capture configuration is not " - "in a block (such as [ExclusionList])", - line_no); - ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG; - break; + if ((params->add_image_flags & WIMLIB_ADD_IMAGE_FLAG_ROOT) && + !S_ISDIR(root_stbuf.st_mode)) + { + ERROR("Root of capture \"%s\" is not a directory", + root_disk_path); + return WIMLIB_ERR_NOTDIR; } - if (ret != 0) - goto out_destroy; + params->capture_root_ino = root_stbuf.st_ino; + params->capture_root_dev = root_stbuf.st_dev; } - return 0; -out_destroy: - destroy_capture_config(config); - return ret; -} -static int capture_config_set_prefix(struct capture_config *config, - const char *_prefix) -{ - char *prefix = STRDUP(_prefix); + path_bufsz = min(32790, PATH_MAX + 1); + path_len = strlen(root_disk_path); - if (!prefix) + if (path_len >= path_bufsz) + return WIMLIB_ERR_INVALID_PARAM; + + path_buf = MALLOC(path_bufsz); + if (!path_buf) return WIMLIB_ERR_NOMEM; - FREE(config->prefix); - config->prefix = prefix; - config->prefix_len = strlen(prefix); - return 0; + memcpy(path_buf, root_disk_path, path_len + 1); + + ret = unix_build_dentry_tree_recursive(root_ret, path_buf, + path_len, params); + FREE(path_buf); + return ret; } +#endif /* !__WIN32__ */ -static bool match_pattern(const char *path, const char *path_basename, - const struct pattern_list *list) +static bool +match_pattern(const tchar *path, + const tchar *path_basename, + const struct wimlib_pattern_list *list) { for (size_t i = 0; i < list->num_pats; i++) { - const char *pat = list->pats[i]; - const char *string; - if (pat[0] == '/') + + const tchar *pat = list->pats[i]; + const tchar *string; + + if (*pat == T('/')) { /* Absolute path from root of capture */ string = path; - else { - if (strchr(pat, '/')) + } else { + if (tstrchr(pat, T('/'))) /* Relative path from root of capture */ string = path + 1; else /* A file name pattern */ string = path_basename; } - if (fnmatch(pat, string, FNM_PATHNAME - #ifdef FNM_CASEFOLD - | FNM_CASEFOLD - #endif - ) == 0) + + /* Warning: on Windows native builds, fnmatch() calls the + * replacement function in win32.c. */ + if (fnmatch(pat, string, FNM_PATHNAME | FNM_NOESCAPE + #ifdef FNM_CASEFOLD + | FNM_CASEFOLD + #endif + ) == 0) { - DEBUG("`%s' matches the pattern \"%s\"", + DEBUG("\"%"TS"\" matches the pattern \"%"TS"\"", string, pat); return true; + } else { + DEBUG2("\"%"TS"\" does not match the pattern \"%"TS"\"", + string, pat); } } return false; @@ -1173,96 +463,93 @@ static bool match_pattern(const char *path, const char *path_basename, * file /mnt/windows7/hiberfil.sys if we are capturing the /mnt/windows7 * directory. */ -bool exclude_path(const char *path, const struct capture_config *config, - bool exclude_prefix) +bool +exclude_path(const tchar *path, size_t path_len, + const struct wimlib_capture_config *config, bool exclude_prefix) { - const char *basename = path_basename(path); + const tchar *basename = path_basename_with_len(path, path_len); if (exclude_prefix) { - wimlib_assert(strlen(path) >= config->prefix_len); - if (memcmp(config->prefix, path, config->prefix_len) == 0 - && path[config->prefix_len] == '/') - path += config->prefix_len; + wimlib_assert(path_len >= config->_prefix_num_tchars); + if (!tmemcmp(config->_prefix, path, config->_prefix_num_tchars) && + path[config->_prefix_num_tchars] == T('/')) + { + path += config->_prefix_num_tchars; + } } - return match_pattern(path, basename, &config->exclusion_list) && - !match_pattern(path, basename, &config->exclusion_exception); + return match_pattern(path, basename, &config->exclusion_pats) && + !match_pattern(path, basename, &config->exclusion_exception_pats); } /* Strip leading and trailing forward slashes from a string. Modifies it in * place and returns the stripped string. */ -static const char *canonicalize_target_path(char *target_path) +static const tchar * +canonicalize_target_path(tchar *target_path) { - char *p; + tchar *p; if (target_path == NULL) - target_path = ""; + return T(""); for (;;) { - if (*target_path == '\0') + if (*target_path == T('\0')) return target_path; - else if (*target_path == '/') + else if (*target_path == T('/')) target_path++; else break; } - p = target_path + strlen(target_path) - 1; - while (*p == '/') - *p-- = '\0'; + p = tstrchr(target_path, T('\0')) - 1; + while (*p == T('/')) + *p-- = T('\0'); return target_path; } -#if defined(__CYGWIN__) || defined(__WIN32__) -static void zap_backslashes(char *s) -{ - while (*s) { - if (*s == '\\') - *s = '/'; - s++; - } -} -#endif - -/* Strip leading and trailing slashes from the target paths */ -static void canonicalize_targets(struct wimlib_capture_source *sources, +/* Strip leading and trailing slashes from the target paths, and translate all + * backslashes in the source and target paths into forward slashes. */ +static void +canonicalize_sources_and_targets(struct wimlib_capture_source *sources, size_t num_sources) { while (num_sources--) { - DEBUG("Canonicalizing { source: \"%s\", target=\"%s\"}", + DEBUG("Canonicalizing { source: \"%"TS"\", target=\"%"TS"\"}", sources->fs_source_path, sources->wim_target_path); -#if defined(__CYGWIN__) || defined(__WIN32__) + /* The Windows API can handle forward slashes. Just get rid of * backslashes to avoid confusing other parts of the library * code. */ zap_backslashes(sources->fs_source_path); if (sources->wim_target_path) zap_backslashes(sources->wim_target_path); -#endif + sources->wim_target_path = - (char*)canonicalize_target_path(sources->wim_target_path); - DEBUG("Canonical target: \"%s\"", sources->wim_target_path); + (tchar*)canonicalize_target_path(sources->wim_target_path); + DEBUG("Canonical target: \"%"TS"\"", sources->wim_target_path); sources++; } } -static int capture_source_cmp(const void *p1, const void *p2) +static int +capture_source_cmp(const void *p1, const void *p2) { const struct wimlib_capture_source *s1 = p1, *s2 = p2; - return strcmp(s1->wim_target_path, s2->wim_target_path); + return tstrcmp(s1->wim_target_path, s2->wim_target_path); } /* Sorts the capture sources lexicographically by target path. This occurs * after leading and trailing forward slashes are stripped. * * One purpose of this is to make sure that target paths that are inside other - * target paths are extracted after the containing target paths. */ -static void sort_sources(struct wimlib_capture_source *sources, - size_t num_sources) + * target paths are added after the containing target paths. */ +static void +sort_sources(struct wimlib_capture_source *sources, size_t num_sources) { qsort(sources, num_sources, sizeof(sources[0]), capture_source_cmp); } -static int check_sorted_sources(struct wimlib_capture_source *sources, - size_t num_sources, int add_image_flags) +static int +check_sorted_sources(struct wimlib_capture_source *sources, size_t num_sources, + int add_image_flags) { if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_NTFS) { if (num_sources != 1) { @@ -1270,7 +557,7 @@ static int check_sorted_sources(struct wimlib_capture_source *sources, "(the NTFS volume) in NTFS mode!"); return WIMLIB_ERR_INVALID_PARAM; } - if (sources[0].wim_target_path[0] != '\0') { + if (sources[0].wim_target_path[0] != T('\0')) { ERROR("In NTFS capture mode the target path inside " "the image must be the root directory!"); return WIMLIB_ERR_INVALID_PARAM; @@ -1316,52 +603,63 @@ static int check_sorted_sources(struct wimlib_capture_source *sources, /* Creates a new directory to place in the WIM image. This is to create parent * directories that are not part of any target as needed. */ -static struct wim_dentry * -new_filler_directory(const char *name) +static int +new_filler_directory(const tchar *name, struct wim_dentry **dentry_ret) { + int ret; struct wim_dentry *dentry; - DEBUG("Creating filler directory \"%s\"", name); - dentry = new_dentry_with_inode(name); - if (dentry) { - /* Set the inode number to 0 for now. The final inode number - * will be assigned later by assign_inode_numbers(). */ - dentry->d_inode->i_ino = 0; + + DEBUG("Creating filler directory \"%"TS"\"", name); + ret = new_dentry_with_inode(name, &dentry); + if (ret == 0) { + /* Leave the inode number as 0; this is allowed for non + * hard-linked files. */ dentry->d_inode->i_resolved = 1; dentry->d_inode->i_attributes = FILE_ATTRIBUTE_DIRECTORY; + *dentry_ret = dentry; } - return dentry; + return ret; } -/* Transfers the children of @branch to @target. It is an error if @target is - * not a directory or if both @branch and @target contain a child dentry with - * the same name. */ -static int do_overlay(struct wim_dentry *target, struct wim_dentry *branch) +/* Overlays @branch onto @target, both of which must be directories. */ +static int +do_overlay(struct wim_dentry *target, struct wim_dentry *branch) { struct rb_root *rb_root; - if (!dentry_is_directory(target)) { - ERROR("Cannot overlay directory `%s' over non-directory", - branch->file_name_utf8); + DEBUG("Doing overlay \"%"WS"\" => \"%"WS"\"", + branch->file_name, target->file_name); + + if (!dentry_is_directory(branch) || !dentry_is_directory(target)) { + ERROR("Cannot overlay \"%"WS"\" onto existing dentry: " + "is not directory-on-directory!", branch->file_name); return WIMLIB_ERR_INVALID_OVERLAY; } rb_root = &branch->d_inode->i_children; while (rb_root->rb_node) { /* While @branch has children... */ struct wim_dentry *child = rbnode_dentry(rb_root->rb_node); + struct wim_dentry *existing; + /* Move @child to the directory @target */ unlink_dentry(child); - if (!dentry_add_child(target, child)) { - /* Revert the change to avoid leaking the directory tree - * rooted at @child */ - dentry_add_child(branch, child); - ERROR("Overlay error: file `%s' already exists " - "as a child of `%s'", - child->file_name_utf8, target->file_name_utf8); - return WIMLIB_ERR_INVALID_OVERLAY; + existing = dentry_add_child(target, child); + + /* File or directory with same name already exists */ + if (existing) { + int ret; + ret = do_overlay(existing, child); + if (ret) { + /* Overlay failed. Revert the change to avoid + * leaking the directory tree rooted at @child. + * */ + dentry_add_child(branch, child); + return ret; + } } } + free_dentry(branch); return 0; - } /* Attach or overlay a branch onto the WIM image. @@ -1375,14 +673,18 @@ static int do_overlay(struct wim_dentry *target, struct wim_dentry *branch) * Path in the WIM image to add the branch, with leading and trailing * slashes stripped. */ -static int attach_branch(struct wim_dentry **root_p, - struct wim_dentry *branch, - char *target_path) +static int +attach_branch(struct wim_dentry **root_p, struct wim_dentry *branch, + tchar *target_path) { - char *slash; + tchar *slash; struct wim_dentry *dentry, *parent, *target; + int ret; + + DEBUG("Attaching branch \"%"WS"\" => \"%"TS"\"", + branch->file_name, target_path); - if (*target_path == '\0') { + if (*target_path == T('\0')) { /* Target: root directory */ if (*root_p) { /* Overlay on existing root */ @@ -1397,21 +699,21 @@ static int attach_branch(struct wim_dentry **root_p, /* Adding a non-root branch. Create root if it hasn't been created * already. */ if (!*root_p) { - *root_p = new_filler_directory(""); - if (!*root_p) - return WIMLIB_ERR_NOMEM; + ret = new_filler_directory(T(""), root_p); + if (ret) + return ret; } /* Walk the path to the branch, creating filler directories as needed. * */ parent = *root_p; - while ((slash = strchr(target_path, '/'))) { - *slash = '\0'; + while ((slash = tstrchr(target_path, T('/')))) { + *slash = T('\0'); dentry = get_dentry_child_with_name(parent, target_path); if (!dentry) { - dentry = new_filler_directory(target_path); - if (!dentry) - return WIMLIB_ERR_NOMEM; + ret = new_filler_directory(target_path, &dentry); + if (ret) + return ret; dentry_add_child(parent, dentry); } parent = dentry; @@ -1421,12 +723,13 @@ static int attach_branch(struct wim_dentry **root_p, * trailing slashes were tripped. */ do { ++target_path; - } while (*target_path == '/'); + } while (*target_path == T('/')); } /* If the target path already existed, overlay the branch onto it. * Otherwise, set the branch as the target path. */ - target = get_dentry_child_with_name(parent, branch->file_name_utf8); + target = get_dentry_child_with_utf16le_name(parent, branch->file_name, + branch->file_name_nbytes); if (target) { return do_overlay(target, branch); } else { @@ -1435,27 +738,83 @@ static int attach_branch(struct wim_dentry **root_p, } } -WIMLIBAPI int wimlib_add_image_multisource(WIMStruct *w, - struct wimlib_capture_source *sources, - size_t num_sources, - const char *name, - const char *config_str, - size_t config_len, - int add_image_flags, - wimlib_progress_func_t progress_func) +static int +canonicalize_pat(tchar **pat_p) +{ + tchar *pat = *pat_p; + + /* Turn all backslashes in the pattern into forward slashes. */ + zap_backslashes(pat); + + if (*pat != T('/') && *pat != T('\0') && *(pat + 1) == T(':')) { + /* Pattern begins with drive letter */ + if (*(pat + 2) != T('/')) { + /* Something like c:file, which is actually a path + * relative to the current working directory on the c: + * drive. We require paths with drive letters to be + * absolute. */ + ERROR("Invalid path \"%"TS"\"; paths including drive letters " + "must be absolute!", pat); + ERROR("Maybe try \"%"TC":/%"TS"\"?", + *pat, pat + 2); + return WIMLIB_ERR_INVALID_CAPTURE_CONFIG; + } + + WARNING("Pattern \"%"TS"\" starts with a drive letter, which is " + "being removed.", pat); + /* Strip the drive letter */ + pat += 2; + *pat_p = pat; + } + return 0; +} + +static int +canonicalize_pat_list(struct wimlib_pattern_list *pat_list) { - int (*capture_tree)(struct wim_dentry **, const char *, - struct wim_lookup_table *, - struct wim_security_data *, - const struct capture_config *, - int, wimlib_progress_func_t, void *); + int ret = 0; + for (size_t i = 0; i < pat_list->num_pats; i++) { + ret = canonicalize_pat(&pat_list->pats[i]); + if (ret) + break; + } + return ret; +} + +static int +canonicalize_capture_config(struct wimlib_capture_config *config) +{ + int ret = canonicalize_pat_list(&config->exclusion_pats); + if (ret) + return ret; + return canonicalize_pat_list(&config->exclusion_exception_pats); +} + +WIMLIBAPI int +wimlib_add_image_multisource(WIMStruct *w, + struct wimlib_capture_source *sources, + size_t num_sources, + const tchar *name, + struct wimlib_capture_config *config, + int add_image_flags, + wimlib_progress_func_t progress_func) +{ + int (*capture_tree)(struct wim_dentry **, + const tchar *, + struct add_image_params *); void *extra_arg; struct wim_dentry *root_dentry; struct wim_dentry *branch; struct wim_security_data *sd; - struct capture_config config; struct wim_image_metadata *imd; + struct wim_inode_table inode_table; + struct list_head unhashed_streams; + struct add_image_params params; int ret; + struct sd_set sd_set; +#ifdef WITH_NTFS_3G + struct _ntfs_volume *ntfs_vol = NULL; +#endif if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_NTFS) { #ifdef WITH_NTFS_3G @@ -1469,18 +828,22 @@ WIMLIBAPI int wimlib_add_image_multisource(WIMStruct *w, return WIMLIB_ERR_INVALID_PARAM; } capture_tree = build_dentry_tree_ntfs; - extra_arg = &w->ntfs_vol; + extra_arg = &ntfs_vol; #else ERROR("wimlib was compiled without support for NTFS-3g, so\n" " cannot capture a WIM image directly from a NTFS volume!"); return WIMLIB_ERR_UNSUPPORTED; #endif } else { - capture_tree = build_dentry_tree; + #ifdef __WIN32__ + capture_tree = win32_build_dentry_tree; + #else + capture_tree = unix_build_dentry_tree; + #endif extra_arg = NULL; } -#if defined(__CYGWIN__) || defined(__WIN32__) +#ifdef __WIN32__ if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_UNIX_DATA) { ERROR("Capturing UNIX-specific data is not supported on Windows"); return WIMLIB_ERR_INVALID_PARAM; @@ -1491,6 +854,27 @@ WIMLIBAPI int wimlib_add_image_multisource(WIMStruct *w, } #endif + if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE) + add_image_flags |= WIMLIB_ADD_IMAGE_FLAG_EXCLUDE_VERBOSE; + + if ((add_image_flags & (WIMLIB_ADD_IMAGE_FLAG_RPFIX | + WIMLIB_ADD_IMAGE_FLAG_RPFIX)) == + (WIMLIB_ADD_IMAGE_FLAG_RPFIX | WIMLIB_ADD_IMAGE_FLAG_NORPFIX)) + { + ERROR("Cannot specify RPFIX and NORPFIX flags at the same time!"); + return WIMLIB_ERR_INVALID_PARAM; + } + + if ((add_image_flags & (WIMLIB_ADD_IMAGE_FLAG_RPFIX | + WIMLIB_ADD_IMAGE_FLAG_NORPFIX)) == 0) + { + /* Do reparse-point fixups by default if the header flag is set + * from previous images, or if this is the first image being + * added. */ + if ((w->hdr.flags & WIM_HDR_FLAG_RP_FIX) || w->hdr.image_count == 0) + add_image_flags |= WIMLIB_ADD_IMAGE_FLAG_RPFIX; + } + if (!name || !*name) { ERROR("Must specify a non-empty string for the image name"); return WIMLIB_ERR_INVALID_PARAM; @@ -1502,17 +886,22 @@ WIMLIBAPI int wimlib_add_image_multisource(WIMStruct *w, } if (wimlib_image_name_in_use(w, name)) { - ERROR("There is already an image named \"%s\" in `%s'", - name, w->filename); + ERROR("There is already an image named \"%"TS"\" in the WIM!", + name); return WIMLIB_ERR_IMAGE_NAME_COLLISION; } - if (!config_str) { - DEBUG("Using default capture configuration"); - config_str = default_config; - config_len = strlen(default_config); + if (!config) { + DEBUG("Capture config not provided; using empty config"); + config = alloca(sizeof(*config)); + memset(config, 0, sizeof(*config)); } - ret = init_capture_config(&config, config_str, config_len); + + ret = canonicalize_capture_config(config); + if (ret) + goto out; + + ret = init_inode_table(&inode_table, 9001); if (ret) goto out; @@ -1520,13 +909,16 @@ WIMLIBAPI int wimlib_add_image_multisource(WIMStruct *w, sd = CALLOC(1, sizeof(struct wim_security_data)); if (!sd) { ret = WIMLIB_ERR_NOMEM; - goto out_destroy_capture_config; + goto out_destroy_inode_table; } sd->total_length = 8; - sd->refcnt = 1; + + sd_set.sd = sd; + sd_set.rb_root.rb_node = NULL; + DEBUG("Using %zu capture sources", num_sources); - canonicalize_targets(sources, num_sources); + canonicalize_sources_and_targets(sources, num_sources); sort_sources(sources, num_sources); ret = check_sorted_sources(sources, num_sources, add_image_flags); if (ret) { @@ -1534,119 +926,129 @@ WIMLIBAPI int wimlib_add_image_multisource(WIMStruct *w, goto out_free_security_data; } - DEBUG("Building dentry tree."); - if (num_sources == 0) { - root_dentry = new_filler_directory(""); - if (!root_dentry) - goto out_free_security_data; - } else { - size_t i; + INIT_LIST_HEAD(&unhashed_streams); + w->lookup_table->unhashed_streams = &unhashed_streams; + root_dentry = NULL; + + params.lookup_table = w->lookup_table; + params.inode_table = &inode_table; + params.sd_set = &sd_set; + params.config = config; + params.add_image_flags = add_image_flags; + params.progress_func = progress_func; + params.extra_arg = extra_arg; + for (size_t i = 0; i < num_sources; i++) { + int flags; + union wimlib_progress_info progress; + + DEBUG("Building dentry tree for source %zu of %zu " + "(\"%"TS"\" => \"%"TS"\")", i + 1, num_sources, + sources[i].fs_source_path, + sources[i].wim_target_path); + if (progress_func) { + memset(&progress, 0, sizeof(progress)); + progress.scan.source = sources[i].fs_source_path; + progress.scan.wim_target_path = sources[i].wim_target_path; + progress_func(WIMLIB_PROGRESS_MSG_SCAN_BEGIN, &progress); + } + config->_prefix = sources[i].fs_source_path; + config->_prefix_num_tchars = tstrlen(sources[i].fs_source_path); + flags = add_image_flags | WIMLIB_ADD_IMAGE_FLAG_SOURCE; + if (!*sources[i].wim_target_path) + flags |= WIMLIB_ADD_IMAGE_FLAG_ROOT; + ret = (*capture_tree)(&branch, sources[i].fs_source_path, + ¶ms); + if (ret) { + ERROR("Failed to build dentry tree for `%"TS"'", + sources[i].fs_source_path); + goto out_free_dentry_tree; + } + if (branch) { + /* Use the target name, not the source name, for + * the root of each branch from a capture + * source. (This will also set the root dentry + * of the entire image to be unnamed.) */ + ret = set_dentry_name(branch, + path_basename(sources[i].wim_target_path)); + if (ret) + goto out_free_branch; - root_dentry = NULL; - i = 0; - do { - int flags; - union wimlib_progress_info progress; - - DEBUG("Building dentry tree for source %zu of %zu " - "(\"%s\" => \"%s\")", i + 1, num_sources, - sources[i].fs_source_path, - sources[i].wim_target_path); - if (progress_func) { - memset(&progress, 0, sizeof(progress)); - progress.scan.source = sources[i].fs_source_path; - progress.scan.wim_target_path = sources[i].wim_target_path; - progress_func(WIMLIB_PROGRESS_MSG_SCAN_BEGIN, &progress); - } - ret = capture_config_set_prefix(&config, - sources[i].fs_source_path); + ret = attach_branch(&root_dentry, branch, + sources[i].wim_target_path); if (ret) - goto out_free_dentry_tree; - flags = add_image_flags | WIMLIB_ADD_IMAGE_FLAG_SOURCE; - if (!*sources[i].wim_target_path) - flags |= WIMLIB_ADD_IMAGE_FLAG_ROOT; - ret = (*capture_tree)(&branch, sources[i].fs_source_path, - w->lookup_table, sd, - &config, - flags, - progress_func, extra_arg); - if (ret) { - ERROR("Failed to build dentry tree for `%s'", - sources[i].fs_source_path); - goto out_free_dentry_tree; - } - if (branch) { - /* Use the target name, not the source name, for - * the root of each branch from a capture - * source. (This will also set the root dentry - * of the entire image to be unnamed.) */ - ret = set_dentry_name(branch, - path_basename(sources[i].wim_target_path)); - if (ret) - goto out_free_branch; - - ret = attach_branch(&root_dentry, branch, - sources[i].wim_target_path); - if (ret) - goto out_free_branch; - } - if (progress_func) - progress_func(WIMLIB_PROGRESS_MSG_SCAN_END, &progress); - } while (++i != num_sources); + goto out_free_branch; + } + if (progress_func) + progress_func(WIMLIB_PROGRESS_MSG_SCAN_END, &progress); } - DEBUG("Calculating full paths of dentries."); - ret = for_dentry_in_tree(root_dentry, calculate_dentry_full_path, NULL); - if (ret != 0) - goto out_free_dentry_tree; + if (root_dentry == NULL) { + ret = new_filler_directory(T(""), &root_dentry); + if (ret) + goto out_free_dentry_tree; + } ret = add_new_dentry_tree(w, root_dentry, sd); - if (ret != 0) + + if (ret) { +#ifdef WITH_NTFS_3G + if (ntfs_vol) + do_ntfs_umount(ntfs_vol); +#endif goto out_free_dentry_tree; + } - imd = &w->image_metadata[w->hdr.image_count - 1]; + imd = w->image_metadata[w->hdr.image_count - 1]; + list_transfer(&unhashed_streams, &imd->unhashed_streams); - ret = dentry_tree_fix_inodes(root_dentry, &imd->inode_list); - if (ret != 0) - goto out_destroy_imd; +#ifdef WITH_NTFS_3G + imd->ntfs_vol = ntfs_vol; +#endif DEBUG("Assigning hard link group IDs"); - assign_inode_numbers(&imd->inode_list); + inode_table_prepare_inode_list(&inode_table, &imd->inode_list); ret = xml_add_image(w, name); - if (ret != 0) - goto out_destroy_imd; + if (ret) + goto out_put_imd; if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_BOOT) wimlib_set_boot_idx(w, w->hdr.image_count); + + if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_RPFIX) + w->hdr.flags |= WIM_HDR_FLAG_RP_FIX; + ret = 0; - goto out; -out_destroy_imd: - destroy_image_metadata(&w->image_metadata[w->hdr.image_count - 1], - w->lookup_table); - w->hdr.image_count--; - goto out; + goto out_destroy_inode_table; +out_put_imd: + put_image_metadata(w->image_metadata[--w->hdr.image_count], + w->lookup_table); + goto out_destroy_inode_table; out_free_branch: free_dentry_tree(branch, w->lookup_table); out_free_dentry_tree: free_dentry_tree(root_dentry, w->lookup_table); out_free_security_data: free_security_data(sd); -out_destroy_capture_config: - destroy_capture_config(&config); +out_destroy_inode_table: + destroy_inode_table(&inode_table); + destroy_sd_set(&sd_set); out: return ret; } -WIMLIBAPI int wimlib_add_image(WIMStruct *w, const char *source, - const char *name, const char *config_str, - size_t config_len, int add_image_flags, - wimlib_progress_func_t progress_func) +WIMLIBAPI int +wimlib_add_image(WIMStruct *w, + const tchar *source, + const tchar *name, + struct wimlib_capture_config *config, + int add_image_flags, + wimlib_progress_func_t progress_func) { if (!source || !*source) return WIMLIB_ERR_INVALID_PARAM; - char *fs_source_path = STRDUP(source); + tchar *fs_source_path = TSTRDUP(source); int ret; struct wimlib_capture_source capture_src = { .fs_source_path = fs_source_path, @@ -1654,8 +1056,8 @@ WIMLIBAPI int wimlib_add_image(WIMStruct *w, const char *source, .reserved = 0, }; ret = wimlib_add_image_multisource(w, &capture_src, 1, name, - config_str, config_len, - add_image_flags, progress_func); + config, add_image_flags, + progress_func); FREE(fs_source_path); return ret; }