*/
/*
- * Copyright (C) 2012-2016 Eric Biggers
+ * Copyright (C) 2012-2017 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
* wimlib_extract_pathlist(). Internally, all end up calling
* do_wimlib_extract_paths() and extract_trees().
*
- * Although wimlib supports multiple extraction modes/backends (NTFS-3g, UNIX,
+ * Although wimlib supports multiple extraction modes/backends (NTFS-3G, UNIX,
* Win32), this file does not itself have code to extract files or directories
* to any specific target; instead, it handles generic functionality and relies
* on lower-level callback functions declared in `struct apply_operations' to do
#include <errno.h>
#include <fcntl.h>
+#include <stdlib.h>
#include <sys/stat.h>
#include <unistd.h>
#include "wimlib/endianness.h"
#include "wimlib/error.h"
#include "wimlib/metadata.h"
+#include "wimlib/object_id.h"
#include "wimlib/pathlist.h"
#include "wimlib/paths.h"
#include "wimlib/pattern.h"
#include "wimlib/unix_data.h"
#include "wimlib/wim.h"
#include "wimlib/win32.h" /* for realpath() equivalent */
+#include "wimlib/xattr.h"
#include "wimlib/xml.h"
#define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x80000000
return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
}
+/* Are all bytes in the specified buffer zero? */
+static bool
+is_all_zeroes(const u8 *p, const size_t size)
+{
+ const u8 * const end = p + size;
+
+ for (; (uintptr_t)p % WORDBYTES && p != end; p++)
+ if (*p)
+ return false;
+
+ for (; end - p >= WORDBYTES; p += WORDBYTES)
+ if (*(const machine_word_t *)p)
+ return false;
+
+ for (; p != end; p++)
+ if (*p)
+ return false;
+
+ return true;
+}
+
+/*
+ * Sparse regions should be detected at the granularity of the filesystem block
+ * size. For now just assume 4096 bytes, which is the default block size on
+ * NTFS and most Linux filesystems.
+ */
+#define SPARSE_UNIT 4096
+
+/*
+ * Detect whether the specified buffer begins with a region of all zero bytes.
+ * Return %true if a zero region was found or %false if a nonzero region was
+ * found, and sets *len_ret to the length of the region. This operates at a
+ * granularity of SPARSE_UNIT bytes, meaning that to extend a zero region, there
+ * must be SPARSE_UNIT zero bytes with no interruption, but to extend a nonzero
+ * region, just one nonzero byte in the next SPARSE_UNIT bytes is sufficient.
+ *
+ * Note: besides compression, the WIM format doesn't yet have a way to
+ * efficiently represent zero regions, so that's why we need to detect them
+ * ourselves. Things will still fall apart badly on extremely large sparse
+ * files, but this is a start...
+ */
+bool
+detect_sparse_region(const void *data, size_t size, size_t *len_ret)
+{
+ const void *p = data;
+ const void * const end = data + size;
+ size_t len = 0;
+ bool zeroes = false;
+
+ while (p != end) {
+ size_t n = min(end - p, SPARSE_UNIT);
+ bool z = is_all_zeroes(p, n);
+
+ if (len != 0 && z != zeroes)
+ break;
+ zeroes = z;
+ len += n;
+ p += n;
+ }
+
+ *len_ret = len;
+ return zeroes;
+}
+
#define PWM_FOUND_WIM_HDR (-1)
/* Read the header for a blob in a pipable WIM. If @pwm_hdr_ret is not NULL,
create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
{
tchar *name;
- int open_flags;
int raw_fd;
+#ifdef __WIN32__
retry:
- name = ttempnam(NULL, T("wimlib"));
+ name = _wtempnam(NULL, L"wimlib");
if (!name) {
ERROR_WITH_ERRNO("Failed to create temporary filename");
return WIMLIB_ERR_NOMEM;
}
-
- open_flags = O_WRONLY | O_CREAT | O_EXCL | O_BINARY;
-#ifdef __WIN32__
- open_flags |= _O_SHORT_LIVED;
-#endif
- raw_fd = topen(name, open_flags, 0600);
+ raw_fd = _wopen(name, O_WRONLY | O_CREAT | O_EXCL | O_BINARY |
+ _O_SHORT_LIVED, 0600);
+ if (raw_fd < 0 && errno == EEXIST) {
+ FREE(name);
+ goto retry;
+ }
+#else /* __WIN32__ */
+ const char *tmpdir = getenv("TMPDIR");
+ if (!tmpdir)
+ tmpdir = P_tmpdir;
+ name = MALLOC(strlen(tmpdir) + 1 + 6 + 6 + 1);
+ if (!name)
+ return WIMLIB_ERR_NOMEM;
+ sprintf(name, "%s/wimlibXXXXXX", tmpdir);
+ raw_fd = mkstemp(name);
+#endif /* !__WIN32__ */
if (raw_fd < 0) {
- if (errno == EEXIST) {
- FREE(name);
- goto retry;
- }
ERROR_WITH_ERRNO("Failed to create temporary file "
"\"%"TS"\"", name);
FREE(name);
}
static int
-begin_extract_blob_wrapper(struct blob_descriptor *blob, void *_ctx)
+begin_extract_blob(struct blob_descriptor *blob, void *_ctx)
{
struct apply_ctx *ctx = _ctx;
- ctx->cur_blob = blob;
- ctx->cur_blob_offset = 0;
-
if (unlikely(blob->out_refcnt > MAX_OPEN_FILES))
return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
}
static int
-extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx)
+extract_chunk(const struct blob_descriptor *blob, u64 offset,
+ const void *chunk, size_t size, void *_ctx)
{
struct apply_ctx *ctx = _ctx;
union wimlib_progress_info *progress = &ctx->progress;
+ bool last = (offset + size == blob->size);
int ret;
- ctx->cur_blob_offset += size;
-
if (likely(ctx->supported_features.hard_links)) {
progress->extract.completed_bytes +=
- (u64)size * ctx->cur_blob->out_refcnt;
- if (ctx->cur_blob_offset == ctx->cur_blob->size)
- progress->extract.completed_streams += ctx->cur_blob->out_refcnt;
+ (u64)size * blob->out_refcnt;
+ if (last)
+ progress->extract.completed_streams += blob->out_refcnt;
} else {
const struct blob_extraction_target *targets =
- blob_extraction_targets(ctx->cur_blob);
- for (u32 i = 0; i < ctx->cur_blob->out_refcnt; i++) {
+ blob_extraction_targets(blob);
+ for (u32 i = 0; i < blob->out_refcnt; i++) {
const struct wim_inode *inode = targets[i].inode;
const struct wim_dentry *dentry;
inode_for_each_extraction_alias(dentry, inode) {
progress->extract.completed_bytes += size;
- if (ctx->cur_blob_offset == ctx->cur_blob->size)
+ if (last)
progress->extract.completed_streams++;
}
}
return ret;
}
- return call_consume_chunk(chunk, size, ctx->saved_cbs);
+ return call_continue_blob(blob, offset, chunk, size, ctx->saved_cbs);
}
/* Copy the blob's data from the temporary file to each of its targets.
}
static int
-end_extract_blob_wrapper(struct blob_descriptor *blob, int status, void *_ctx)
+end_extract_blob(struct blob_descriptor *blob, int status, void *_ctx)
{
struct apply_ctx *ctx = _ctx;
extract_blob_list(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
{
struct read_blob_callbacks wrapper_cbs = {
- .begin_blob = begin_extract_blob_wrapper,
- .consume_chunk = extract_chunk_wrapper,
- .end_blob = end_extract_blob_wrapper,
+ .begin_blob = begin_extract_blob,
+ .continue_blob = extract_chunk,
+ .end_blob = end_extract_blob,
.ctx = ctx,
};
ctx->saved_cbs = cbs;
if (!ctx->supported_features.case_sensitive_filenames) {
struct wim_dentry *other;
- list_for_each_entry(other, &dentry->d_ci_conflict_list,
- d_ci_conflict_list)
- {
+ dentry_for_each_ci_match(other, dentry) {
if (will_extract_dentry(other)) {
if (ctx->extract_flags &
WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
features->security_descriptors++;
if (inode_has_unix_data(inode))
features->unix_data++;
+ if (inode_has_object_id(inode))
+ features->object_ids++;
+ if (inode_has_linux_xattrs(inode))
+ features->linux_xattrs++;
}
/* Tally features necessary to extract a dentry and the corresponding inode. */
WARNING("Ignoring Windows NT security descriptors of %lu files",
required_features->security_descriptors);
- /* UNIX data. */
- if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
- required_features->unix_data && !supported_features->unix_data)
+ /* Standard UNIX metadata */
+ if (required_features->unix_data &&
+ (!supported_features->unix_data ||
+ !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA)))
{
- ERROR("Extraction backend does not support UNIX data!");
- return WIMLIB_ERR_UNSUPPORTED;
+ if (extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
+ ERROR("Requested UNIX metadata extraction, but "
+ "extraction backend does not support it!");
+ return WIMLIB_ERR_UNSUPPORTED;
+ }
+ WARNING("Ignoring UNIX metadata (uid/gid/mode/rdev) of %lu files%"TS,
+ required_features->unix_data,
+ (supported_features->unix_data ?
+ T("\n (use --unix-data mode to extract these)") : T("")));
}
- if (required_features->unix_data &&
- !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
+ /* Linux-style extended attributes */
+ if (required_features->linux_xattrs &&
+ (!supported_features->linux_xattrs ||
+ !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA)))
{
- WARNING("Ignoring UNIX metadata of %lu files",
- required_features->unix_data);
+ WARNING("Ignoring Linux-style extended attributes of %lu files%"TS,
+ required_features->linux_xattrs,
+ (supported_features->linux_xattrs ?
+ T("\n (use --unix-data mode to extract these)") : T("")));
+ }
+
+ /* Object IDs. */
+ if (required_features->object_ids && !supported_features->object_ids) {
+ WARNING("Ignoring object IDs of %lu files",
+ required_features->object_ids);
}
/* DOS Names. */
#ifndef WITH_NTFS_3G
if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
- ERROR("wimlib was compiled without support for NTFS-3g, so\n"
+ ERROR("wimlib was compiled without support for NTFS-3G, so\n"
" it cannot apply a WIM image directly to an NTFS volume.");
return WIMLIB_ERR_UNSUPPORTED;
}
}
static const tchar * const filename_forbidden_chars =
-T(
#ifdef __WIN32__
-"<>:\"/\\|?*"
+T("<>:\"/\\|?*");
#else
-"/"
+T("/");
#endif
-);
/* This function checks if it is okay to use a WIM image's name as a directory
* name. */