X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fntfs-3g_capture.c;h=93d71e55bcefd951b84c5d57a271cd03285badff;hp=0180a48898c19d0fac7974bc148b7d01de7818e0;hb=51829aecdac415b417ab5b8ac897014bb780de10;hpb=80b48ed7aeb8e6c9460ab254920818f9e36d656e diff --git a/src/ntfs-3g_capture.c b/src/ntfs-3g_capture.c index 0180a488..93d71e55 100644 --- a/src/ntfs-3g_capture.c +++ b/src/ntfs-3g_capture.c @@ -6,7 +6,7 @@ */ /* - * Copyright (C) 2012, 2013, 2014, 2015 Eric Biggers + * Copyright (C) 2012-2016 Eric Biggers * * This file is free software; you can redistribute it and/or modify it under * the terms of the GNU Lesser General Public License as published by the Free @@ -31,6 +31,7 @@ #include #include +#include #include #include #include @@ -44,10 +45,20 @@ #include "wimlib/endianness.h" #include "wimlib/error.h" #include "wimlib/ntfs_3g.h" +#include "wimlib/object_id.h" #include "wimlib/paths.h" #include "wimlib/reparse.h" #include "wimlib/security.h" +/* NTFS-3g 2013 renamed MS_RDONLY to NTFS_MNT_RDONLY. We can't check for the + * existence of NTFS_MNT_RDONLY at compilation time because it's an enum. We + * also can't check for MS_RDONLY being missing because it's also a system + * constant. So check if the NTFS-3g specific MS_IGNORE_HIBERFILE is defined; + * if yes, then we need to use the old MS_RDONLY. */ +#ifdef MS_IGNORE_HIBERFILE +# define NTFS_MNT_RDONLY MS_RDONLY +#endif + /* A reference-counted NTFS volume than is automatically unmounted when the * reference count reaches 0 */ struct ntfs_volume_wrapper { @@ -61,7 +72,7 @@ struct ntfs_location { u64 mft_no; ATTR_TYPES attr_type; u32 attr_name_nchars; - utf16lechar *attr_name; + ntfschar *attr_name; u64 sort_key; }; @@ -158,7 +169,8 @@ void free_ntfs_location(struct ntfs_location *loc) { put_ntfs_volume(loc->volume); - FREE(loc->attr_name); + if (loc->attr_name != AT_UNNAMED) + FREE(loc->attr_name); FREE(loc); } @@ -168,7 +180,7 @@ clone_ntfs_location(const struct ntfs_location *loc) struct ntfs_location *new = memdup(loc, sizeof(*loc)); if (!new) goto err0; - if (loc->attr_name) { + if (loc->attr_name != AT_UNNAMED) { new->attr_name = utf16le_dup(loc->attr_name); if (!new->attr_name) goto err1; @@ -189,34 +201,33 @@ cmp_ntfs_locations(const struct ntfs_location *loc1, return cmp_u64(loc1->sort_key, loc2->sort_key); } +/* Read rptag and rpreserved from the NTFS inode and save them in the WIM inode. + */ static int -read_reparse_tag(ntfs_inode *ni, struct ntfs_location *loc, - u32 *reparse_tag_ret) +read_reparse_header(ntfs_inode *ni, struct wim_inode *inode) { - int ret; - le32 reparse_tag; + struct { + le32 rptag; + le16 rpdatalen; + le16 rpreserved; + } hdr; + s64 res; ntfs_attr *na; - na = open_ntfs_attr(ni, loc); - if (!na) { - ret = WIMLIB_ERR_NTFS_3G; - goto out; - } + na = ntfs_attr_open(ni, AT_REPARSE_POINT, AT_UNNAMED, 0); + if (!na) + return WIMLIB_ERR_NTFS_3G; + + res = ntfs_attr_pread(na, 0, sizeof(hdr), &hdr); - if (ntfs_attr_pread(na, 0, sizeof(reparse_tag), - &reparse_tag) != sizeof(reparse_tag)) - { - ERROR_WITH_ERRNO("Error reading reparse data"); - ret = WIMLIB_ERR_NTFS_3G; - goto out_close_ntfs_attr; - } - *reparse_tag_ret = le32_to_cpu(reparse_tag); - ret = 0; -out_close_ntfs_attr: ntfs_attr_close(na); -out: - return ret; + if (res != sizeof(hdr)) + return WIMLIB_ERR_NTFS_3G; + + inode->i_reparse_tag = le32_to_cpu(hdr.rptag); + inode->i_rp_reserved = le16_to_cpu(hdr.rpreserved); + return 0; } static int @@ -257,22 +268,58 @@ set_attr_sort_key(ntfs_inode *ni, struct ntfs_location *loc) return 0; } +/* + * Add a new stream to the specified inode, with duplicate checking. + * + * This works around a problem where NTFS-3g can list multiple unnamed data + * streams for a single file. In this case we can only keep one. We'll prefer + * one that is nonempty. + */ +static int +add_stream(struct wim_inode *inode, const char *path, int stream_type, + const utf16lechar *stream_name, struct blob_descriptor **blob_p, + struct list_head *unhashed_blobs) +{ + struct blob_descriptor *blob = *blob_p; + struct wim_inode_stream *strm; + + strm = inode_get_stream(inode, stream_type, stream_name); + if (unlikely(strm)) { + /* Stream already existed. */ + if (!blob) + return 0; + if (stream_blob_resolved(strm)) { + WARNING("\"%s\" has multiple nonempty streams " + "with the same type and name! Only the first " + "will be saved.", path); + return 0; + } + inode_replace_stream_blob(inode, strm, blob, NULL); + } else { + strm = inode_add_stream(inode, stream_type, stream_name, blob); + if (unlikely(!strm)) + return WIMLIB_ERR_NOMEM; + } + prepare_unhashed_blob(blob, inode, strm->stream_id, unhashed_blobs); + *blob_p = NULL; + return 0; + +} + /* Save information about an NTFS attribute (stream) to a WIM inode. */ static int scan_ntfs_attr(struct wim_inode *inode, ntfs_inode *ni, const char *path, - size_t path_len, struct list_head *unhashed_blobs, struct ntfs_volume_wrapper *volume, ATTR_TYPES type, const ATTR_RECORD *record) { - const u64 data_size = ntfs_get_attribute_value_length(record); + u64 data_size = ntfs_get_attribute_value_length(record); const u32 name_nchars = record->name_length; struct blob_descriptor *blob = NULL; - utf16lechar *stream_name = NULL; - struct wim_inode_stream *strm; + utf16lechar *stream_name = (utf16lechar *)NO_STREAM_NAME; int ret; if (unlikely(name_nchars)) { @@ -285,6 +332,23 @@ scan_ntfs_attr(struct wim_inode *inode, } } + if (unlikely(type == AT_REPARSE_POINT)) { + if (data_size < REPARSE_DATA_OFFSET) { + ERROR("Reparse point attribute of \"%s\" " + "is too short!", path); + ret = WIMLIB_ERR_INVALID_REPARSE_DATA; + goto out_cleanup; + } + data_size -= REPARSE_DATA_OFFSET; + + ret = read_reparse_header(ni, inode); + if (ret) { + ERROR_WITH_ERRNO("Error reading reparse point header " + "of \"%s\"", path); + goto out_cleanup; + } + } + /* If the stream is non-empty, set up a blob descriptor for it. */ if (data_size != 0) { blob = new_blob_descriptor(); @@ -293,7 +357,7 @@ scan_ntfs_attr(struct wim_inode *inode, goto out_cleanup; } - blob->ntfs_loc = CALLOC(1, sizeof(struct ntfs_location)); + blob->ntfs_loc = MALLOC(sizeof(struct ntfs_location)); if (unlikely(!blob->ntfs_loc)) { ret = WIMLIB_ERR_NOMEM; goto out_cleanup; @@ -302,52 +366,32 @@ scan_ntfs_attr(struct wim_inode *inode, blob->blob_location = BLOB_IN_NTFS_VOLUME; blob->size = data_size; blob->ntfs_loc->volume = get_ntfs_volume(volume); - blob->ntfs_loc->attr_type = type; blob->ntfs_loc->mft_no = ni->mft_no; + blob->ntfs_loc->attr_type = type; if (unlikely(name_nchars)) { + blob->ntfs_loc->attr_name_nchars = name_nchars; blob->ntfs_loc->attr_name = utf16le_dup(stream_name); if (!blob->ntfs_loc->attr_name) { ret = WIMLIB_ERR_NOMEM; goto out_cleanup; } - blob->ntfs_loc->attr_name_nchars = name_nchars; + } else { + blob->ntfs_loc->attr_name_nchars = 0; + blob->ntfs_loc->attr_name = AT_UNNAMED; } ret = set_attr_sort_key(ni, blob->ntfs_loc); if (ret) goto out_cleanup; - - if (unlikely(type == AT_REPARSE_POINT)) { - if (blob->size < REPARSE_DATA_OFFSET) { - ERROR("Reparse data of \"%s\" " - "is invalid (only %"PRIu64" bytes)!", - path, data_size); - ret = WIMLIB_ERR_INVALID_REPARSE_DATA; - goto out_cleanup; - } - blob->size -= REPARSE_DATA_OFFSET; - ret = read_reparse_tag(ni, blob->ntfs_loc, - &inode->i_reparse_tag); - if (ret) - goto out_cleanup; - } } - strm = inode_add_stream(inode, - attr_type_to_wimlib_stream_type(type), - stream_name ? stream_name : NO_STREAM_NAME, - blob); - if (unlikely(!strm)) { - ret = WIMLIB_ERR_NOMEM; - goto out_cleanup; - } - prepare_unhashed_blob(blob, inode, strm->stream_id, unhashed_blobs); - blob = NULL; - ret = 0; + ret = add_stream(inode, path, attr_type_to_wimlib_stream_type(type), + stream_name, &blob, unhashed_blobs); out_cleanup: free_blob_descriptor(blob); - FREE(stream_name); + if (stream_name != NO_STREAM_NAME) + FREE(stream_name); return ret; } @@ -356,7 +400,6 @@ static int scan_ntfs_attrs_with_type(struct wim_inode *inode, ntfs_inode *ni, const char *path, - size_t path_len, struct list_head *unhashed_blobs, struct ntfs_volume_wrapper *volume, ATTR_TYPES type) @@ -377,7 +420,6 @@ scan_ntfs_attrs_with_type(struct wim_inode *inode, ret = scan_ntfs_attr(inode, ni, path, - path_len, unhashed_blobs, volume, type, @@ -396,6 +438,22 @@ out_put_actx: return ret; } +static noinline_for_stack int +load_object_id(ntfs_inode *ni, struct wim_inode *inode) +{ + OBJECT_ID_ATTR attr; + int len; + + len = ntfs_get_ntfs_object_id(ni, (char *)&attr, sizeof(attr)); + if (likely(len == -ENODATA || len == 0)) + return 0; + if (len < 0) + return WIMLIB_ERR_NTFS_3G; + if (!inode_set_object_id(inode, &attr, len)) + return WIMLIB_ERR_NOMEM; + return 0; +} + /* Load the security descriptor of an NTFS inode into the corresponding WIM * inode and the WIM image's security descriptor set. */ static noinline_for_stack int @@ -590,17 +648,13 @@ filldir(void *_ctx, const ntfschar *name, const int name_nchars, goto out; } - /* Ignore . and .. entries */ - ret = 0; - if ((name_nchars == 1 && name[0] == cpu_to_le16('.')) || - (name_nchars == 2 && name[0] == cpu_to_le16('.') && - name[1] == cpu_to_le16('.'))) - goto out; - ret = utf16le_to_tstr(name, name_nbytes, &mbs_name, &mbs_name_nbytes); if (ret) goto out; + if (should_ignore_filename(mbs_name, mbs_name_nbytes)) + goto out_free_mbs_name; + path_len = ctx->path_len; if (path_len != 1) ctx->path[path_len++] = '/'; @@ -610,8 +664,8 @@ filldir(void *_ctx, const ntfschar *name, const int name_nchars, ret = ntfs_3g_build_dentry_tree_recursive(&child, mref, ctx->path, path_len, name_type, ctx->volume, ctx->params); - if (child) - dentry_add_child(ctx->parent, child); + attach_scanned_tree(ctx->parent, child, ctx->params->blob_table); +out_free_mbs_name: FREE(mbs_name); out: ctx->ret = ret; @@ -675,10 +729,10 @@ ntfs_3g_build_dentry_tree_recursive(struct wim_dentry **root_ret, struct wim_inode *inode = NULL; ntfs_inode *ni = NULL; - ret = try_exclude(path, path_len, params); - if (ret < 0) /* Excluded? */ + ret = try_exclude(path, params); + if (unlikely(ret < 0)) /* Excluded? */ goto out_progress; - if (ret > 0) /* Error? */ + if (unlikely(ret > 0)) /* Error? */ goto out; ni = ntfs_inode_open(volume->vol, mref); @@ -733,21 +787,27 @@ ntfs_3g_build_dentry_tree_recursive(struct wim_dentry **root_ret, if (attributes & FILE_ATTRIBUTE_REPARSE_POINT) { /* Scan the reparse point stream. */ - ret = scan_ntfs_attrs_with_type(inode, ni, path, path_len, + ret = scan_ntfs_attrs_with_type(inode, ni, path, params->unhashed_blobs, volume, AT_REPARSE_POINT); if (ret) goto out; } + /* Load the object ID. */ + ret = load_object_id(ni, inode); + if (ret) { + ERROR_WITH_ERRNO("Error reading object ID of \"%s\"", path); + goto out; + } + /* Scan the data streams. * * Note: directories should not have an unnamed data stream, but they * may have named data streams. Nondirectories (including reparse * points) can have an unnamed data stream as well as named data * streams. */ - ret = scan_ntfs_attrs_with_type(inode, ni, path, path_len, - params->unhashed_blobs, + ret = scan_ntfs_attrs_with_type(inode, ni, path, params->unhashed_blobs, volume, AT_DATA); if (ret) goto out; @@ -807,23 +867,7 @@ ntfs_3g_build_dentry_tree(struct wim_dentry **root_ret, if (!volume) return WIMLIB_ERR_NOMEM; - /* NTFS-3g 2013 renamed the "read-only" mount flag from MS_RDONLY to - * NTFS_MNT_RDONLY. - * - * Unfortunately we can't check for defined(NTFS_MNT_RDONLY) because - * NTFS_MNT_RDONLY is an enumerated constant. Also, the NTFS-3g headers - * don't seem to contain any explicit version information. So we have - * to rely on a test done at configure time to detect whether - * NTFS_MNT_RDONLY should be used. */ -#ifdef HAVE_NTFS_MNT_RDONLY - /* NTFS-3g 2013 */ vol = ntfs_mount(device, NTFS_MNT_RDONLY); -#elif defined(MS_RDONLY) - /* NTFS-3g 2011, 2012 */ - vol = ntfs_mount(device, MS_RDONLY); -#else - #error "Can't find NTFS_MNT_RDONLY or MS_RDONLY flags" -#endif if (!vol) { ERROR_WITH_ERRNO("Failed to mount NTFS volume \"%s\" read-only", device);