]> wimlib.net Git - wimlib/blobdiff - src/ntfs-3g_capture.c
resource: pass blob and offset to consume_chunk
[wimlib] / src / ntfs-3g_capture.c
index 439d17a61aa98c7931b9b59646427f62aa72ac6d..ca8d1b2ffe4a335002569c0a6de780a52d13e9f2 100644 (file)
@@ -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 +31,8 @@
 #include <errno.h>
 
 #include <ntfs-3g/attrib.h>
+#include <ntfs-3g/compat.h> /* for ENODATA, if needed */
+#include <ntfs-3g/object_id.h>
 #include <ntfs-3g/reparse.h>
 #include <ntfs-3g/security.h>
 #include <ntfs-3g/volume.h>
 #include "wimlib/alloca.h"
 #include "wimlib/assert.h"
 #include "wimlib/blob_table.h"
-#include "wimlib/capture.h"
 #include "wimlib/dentry.h"
 #include "wimlib/encoding.h"
 #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/scan.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 {
        ntfs_volume *vol;
        size_t refcnt;
+       bool dedup_warned;
 };
 
 /* Description of where data is located in an NTFS volume  */
@@ -61,7 +74,7 @@ struct ntfs_location {
        u64 mft_no;
        ATTR_TYPES attr_type;
        u32 attr_name_nchars;
-       utf16lechar *attr_name;
+       ntfschar *attr_name;
        u64 sort_key;
 };
 
@@ -104,7 +117,7 @@ open_ntfs_attr(ntfs_inode *ni, const struct ntfs_location *loc)
 
 int
 read_ntfs_attribute_prefix(const struct blob_descriptor *blob, u64 size,
-                          const struct read_blob_callbacks *cbs)
+                          const struct consume_chunk_callback *cb)
 {
        const struct ntfs_location *loc = blob->ntfs_loc;
        ntfs_volume *vol = loc->volume->vol;
@@ -141,7 +154,7 @@ read_ntfs_attribute_prefix(const struct blob_descriptor *blob, u64 size,
                }
                pos += to_read;
                bytes_remaining -= to_read;
-               ret = call_consume_chunk(buf, to_read, cbs);
+               ret = consume_chunk(cb, buf, to_read);
                if (ret)
                        goto out_close_ntfs_attr;
        }
@@ -158,7 +171,8 @@ 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 +182,7 @@ 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;
@@ -218,6 +232,27 @@ read_reparse_header(ntfs_inode *ni, struct wim_inode *inode)
        return 0;
 }
 
+
+static void
+warn_special_reparse_points(const struct wim_inode *inode,
+                           const struct scan_params *params,
+                           struct ntfs_volume_wrapper *volume)
+{
+       if (inode->i_reparse_tag == WIM_IO_REPARSE_TAG_DEDUP &&
+           (params->add_flags & WIMLIB_ADD_FLAG_WINCONFIG) &&
+           !volume->dedup_warned)
+       {
+               WARNING(
+         "Filesystem includes files deduplicated with Windows'\n"
+"          Data Deduplication feature, which to properly restore\n"
+"          would require that the chunk store in \"System Volume Information\"\n"
+"          be included in the WIM image. By default \"System Volume Information\"\n"
+"          is excluded, so you may want to use a custom capture configuration\n"
+"          file which includes it.");
+               volume->dedup_warned = true;
+       }
+}
+
 static int
 attr_type_to_wimlib_stream_type(ATTR_TYPES type)
 {
@@ -256,12 +291,49 @@ 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,
@@ -270,8 +342,7 @@ scan_ntfs_attr(struct wim_inode *inode,
        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)) {
@@ -309,7 +380,7 @@ 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;
@@ -318,16 +389,19 @@ 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);
@@ -335,20 +409,12 @@ scan_ntfs_attr(struct wim_inode *inode,
                        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;
 }
 
@@ -357,7 +423,6 @@ 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)
@@ -378,7 +443,6 @@ scan_ntfs_attrs_with_type(struct wim_inode *inode,
                ret = scan_ntfs_attr(inode,
                                     ni,
                                     path,
-                                    path_len,
                                     unhashed_blobs,
                                     volume,
                                     type,
@@ -397,6 +461,22 @@ 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
@@ -554,7 +634,7 @@ struct readdir_ctx {
        size_t path_len;
        struct dos_name_map dos_name_map;
        struct ntfs_volume_wrapper *volume;
-       struct capture_params *params;
+       struct scan_params *params;
        int ret;
 };
 
@@ -565,7 +645,7 @@ ntfs_3g_build_dentry_tree_recursive(struct wim_dentry **root_p,
                                    size_t path_len,
                                    int name_type,
                                    struct ntfs_volume_wrapper *volume,
-                                   struct capture_params *params);
+                                   struct scan_params *params);
 
 static int
 filldir(void *_ctx, const ntfschar *name, const int name_nchars,
@@ -619,7 +699,7 @@ static int
 ntfs_3g_recurse_directory(ntfs_inode *ni, char *path, size_t path_len,
                          struct wim_dentry *parent,
                          struct ntfs_volume_wrapper *volume,
-                         struct capture_params *params)
+                         struct scan_params *params)
 {
        int ret;
        s64 pos = 0;
@@ -664,7 +744,7 @@ ntfs_3g_build_dentry_tree_recursive(struct wim_dentry **root_ret,
                                    size_t path_len,
                                    int name_type,
                                    struct ntfs_volume_wrapper *volume,
-                                   struct capture_params *params)
+                                   struct scan_params *params)
 {
        u32 attributes;
        int ret;
@@ -696,13 +776,13 @@ ntfs_3g_build_dentry_tree_recursive(struct wim_dentry **root_ret,
        if (unlikely(attributes & FILE_ATTRIBUTE_ENCRYPTED)) {
                if (params->add_flags & WIMLIB_ADD_FLAG_NO_UNSUPPORTED_EXCLUDE)
                {
-                       ERROR("Can't archive \"%s\" because NTFS-3g capture mode "
+                       ERROR("Can't archive \"%s\" because NTFS-3G capture mode "
                              "does not support encrypted files and directories", path);
                        ret = WIMLIB_ERR_UNSUPPORTED_FILE;
                        goto out;
                }
                params->progress.scan.cur_path = path;
-               ret = do_capture_progress(params, WIMLIB_SCAN_DENTRY_UNSUPPORTED, NULL);
+               ret = do_scan_progress(params, WIMLIB_SCAN_DENTRY_UNSUPPORTED, NULL);
                goto out;
        }
 
@@ -730,11 +810,20 @@ 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;
+
+               warn_special_reparse_points(inode, params, volume);
+       }
+
+       /* 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.
@@ -743,13 +832,12 @@ ntfs_3g_build_dentry_tree_recursive(struct wim_dentry **root_ret,
         * 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;
 
-       /* Reparse-point fixups are a no-op because in NTFS-3g capture mode we
+       /* Reparse-point fixups are a no-op because in NTFS-3G capture mode we
         * only allow capturing an entire volume. */
        if (params->add_flags & WIMLIB_ADD_FLAG_RPFIX &&
            inode_is_symlink(inode))
@@ -775,16 +863,16 @@ ntfs_3g_build_dentry_tree_recursive(struct wim_dentry **root_ret,
 out_progress:
        params->progress.scan.cur_path = path;
        if (root == NULL)
-               ret = do_capture_progress(params, WIMLIB_SCAN_DENTRY_EXCLUDED, NULL);
+               ret = do_scan_progress(params, WIMLIB_SCAN_DENTRY_EXCLUDED, NULL);
        else
-               ret = do_capture_progress(params, WIMLIB_SCAN_DENTRY_OK, inode);
+               ret = do_scan_progress(params, WIMLIB_SCAN_DENTRY_OK, inode);
 out:
        if (ni)
                ntfs_inode_close(ni);
        if (unlikely(ret)) {
                free_dentry_tree(root, params->blob_table);
                root = NULL;
-               ret = report_capture_error(params, ret, path);
+               ret = report_scan_error(params, ret, path);
        }
        *root_ret = root;
        return ret;
@@ -792,35 +880,18 @@ out:
 
 int
 ntfs_3g_build_dentry_tree(struct wim_dentry **root_ret,
-                         const char *device,
-                         struct capture_params *params)
+                         const char *device, struct scan_params *params)
 {
        struct ntfs_volume_wrapper *volume;
        ntfs_volume *vol;
        char *path;
        int ret;
 
-       volume = MALLOC(sizeof(struct ntfs_volume_wrapper));
+       volume = CALLOC(1, sizeof(struct ntfs_volume_wrapper));
        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);
@@ -831,7 +902,16 @@ ntfs_3g_build_dentry_tree(struct wim_dentry **root_ret,
        volume->vol = vol;
        volume->refcnt = 1;
 
-       ntfs_open_secure(vol);
+       /* Currently, libntfs-3g users that need to read security descriptors
+        * are required to call ntfs_open_secure() to open the volume's security
+        * descriptor index, "$Secure".  This is only required to work on NTFS
+        * v3.0+, as older versions did not have a security descriptor index. */
+       if (ntfs_open_secure(vol) && vol->major_ver >= 3) {
+               ERROR_WITH_ERRNO("Unable to open security descriptor index of "
+                                "NTFS volume \"%s\"", device);
+               ret = WIMLIB_ERR_NTFS_3G;
+               goto out_put_ntfs_volume;
+       }
 
        /* We don't want to capture the special NTFS files such as $Bitmap.  Not
         * to be confused with "hidden" or "system" files which are real files
@@ -843,7 +923,7 @@ ntfs_3g_build_dentry_tree(struct wim_dentry **root_ret,
        path = MALLOC(32768);
        if (!path) {
                ret = WIMLIB_ERR_NOMEM;
-               goto out_put_ntfs_volume;
+               goto out_close_secure;
        }
 
        path[0] = '/';
@@ -852,10 +932,17 @@ ntfs_3g_build_dentry_tree(struct wim_dentry **root_ret,
                                                  FILE_NAME_POSIX, volume,
                                                  params);
        FREE(path);
+out_close_secure:
+       /* Undo the effects of ntfs_open_secure().  This is not yet done
+        * automatically by ntfs_umount().  But NULL out the inode to
+        * potentially be robust against future versions doing so. */
+       if (vol->secure_ni) {
+               ntfs_index_ctx_put(vol->secure_xsii);
+               ntfs_index_ctx_put(vol->secure_xsdh);
+               ntfs_inode_close(vol->secure_ni);
+               vol->secure_ni = NULL;
+       }
 out_put_ntfs_volume:
-       ntfs_index_ctx_put(vol->secure_xsii);
-       ntfs_index_ctx_put(vol->secure_xsdh);
-       ntfs_inode_close(vol->secure_ni);
        put_ntfs_volume(volume);
        return ret;
 }