kernel: backport ntfs3 driver from 5.19

Signed-off-by: Tianling Shen <cnsztl@immortalwrt.org>

package/kernel/linux/modules/fs.mk

@@ -480,6 +480,26 @@ endef
 $(eval $(call KernelPackage,fs-ntfs))
 
 
+define KernelPackage/fs-ntfs3
+  SUBMENU:=$(FS_MENU)
+  TITLE:=NTFS3 Read-Write file system support
+  DEPENDS:=@LINUX_4_19 +kmod-nls-base
+  KCONFIG:= \
+	CONFIG_NTFS3_FS \
+	CONFIG_NTFS3_64BIT_CLUSTER=y \
+	CONFIG_NTFS3_LZX_XPRESS=y \
+	CONFIG_NTFS3_FS_POSIX_ACL=y
+  FILES:=$(LINUX_DIR)/fs/ntfs3/ntfs3.ko
+  AUTOLOAD:=$(call AutoLoad,30,ntfs3)
+endef
+
+define KernelPackage/fs-ntfs3/description
+ Kernel module for NTFS3 filesystem support
+endef
+
+$(eval $(call KernelPackage,fs-ntfs3))
+
+
 define KernelPackage/fs-reiserfs
   SUBMENU:=$(FS_MENU)
   TITLE:=ReiserFS filesystem support

target/linux/generic/backport-4.19/510-v5.15-fs-ntfs3-Add-NTFS3-in-fs-Kconfig-and-fs-Makefile.patch

@@ -0,0 +1,40 @@
+From e64a5b846cd19f86eae8d8f2708f2e483dc37feb Mon Sep 17 00:00:00 2003
+From: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
+Date: Thu, 28 Jan 2021 12:04:54 +0300
+Subject: [PATCH 09/11] fs/ntfs3: Add NTFS3 in fs/Kconfig and fs/Makefile
+
+This adds NTFS3 in fs/Kconfig and fs/Makefile
+
+Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
+---
+ fs/Kconfig  | 1 +
+ fs/Makefile | 1 +
+ 2 files changed, 2 insertions(+)
+
+--- a/fs/Kconfig
++++ b/fs/Kconfig
+@@ -143,9 +143,10 @@
+ menu "DOS/FAT/NT Filesystems"
+ 
+ source "fs/fat/Kconfig"
+ source "fs/ntfs/Kconfig"
++source "fs/ntfs3/Kconfig"
+ 
+ endmenu
+ endif # BLOCK
+ 
+ menu "Pseudo filesystems"
+--- a/fs/Makefile
++++ b/fs/Makefile
+@@ -98,10 +98,11 @@
+ obj-$(CONFIG_UNICODE)		+= unicode/
+ obj-$(CONFIG_SYSV_FS)		+= sysv/
+ obj-$(CONFIG_CIFS)		+= cifs/
+ obj-$(CONFIG_HPFS_FS)		+= hpfs/
+ obj-$(CONFIG_NTFS_FS)		+= ntfs/
++obj-$(CONFIG_NTFS3_FS)		+= ntfs3/
+ obj-$(CONFIG_UFS_FS)		+= ufs/
+ obj-$(CONFIG_EFS_FS)		+= efs/
+ obj-$(CONFIG_JFFS2_FS)		+= jffs2/
+ obj-$(CONFIG_UBIFS_FS)		+= ubifs/
+ obj-$(CONFIG_AFFS_FS)		+= affs/

target/linux/generic/files/fs/ntfs3/Kconfig

@@ -0,0 +1,46 @@
+# SPDX-License-Identifier: GPL-2.0-only
+config NTFS3_FS
+	tristate "NTFS Read-Write file system support"
+	select NLS
+	help
+	  Windows OS native file system (NTFS) support up to NTFS version 3.1.
+
+	  Y or M enables the NTFS3 driver with full features enabled (read,
+	  write, journal replaying, sparse/compressed files support).
+	  File system type to use on mount is "ntfs3". Module name (M option)
+	  is also "ntfs3".
+
+	  Documentation: <file:Documentation/filesystems/ntfs3.rst>
+
+config NTFS3_64BIT_CLUSTER
+	bool "64 bits per NTFS clusters"
+	depends on NTFS3_FS && 64BIT
+	help
+	  Windows implementation of ntfs.sys uses 32 bits per clusters.
+	  If activated 64 bits per clusters you will be able to use 4k cluster
+	  for 16T+ volumes. Windows will not be able to mount such volumes.
+
+	  It is recommended to say N here.
+
+config NTFS3_LZX_XPRESS
+	bool "activate support of external compressions lzx/xpress"
+	depends on NTFS3_FS
+	help
+	  In Windows 10 one can use command "compact" to compress any files.
+	  4 possible variants of compression are: xpress4k, xpress8k, xpress16k and lzx.
+	  If activated you will be able to read such files correctly.
+
+	  It is recommended to say Y here.
+
+config NTFS3_FS_POSIX_ACL
+	bool "NTFS POSIX Access Control Lists"
+	depends on NTFS3_FS
+	select FS_POSIX_ACL
+	help
+	  POSIX Access Control Lists (ACLs) support additional access rights
+	  for users and groups beyond the standard owner/group/world scheme,
+	  and this option selects support for ACLs specifically for ntfs
+	  filesystems.
+	  NOTE: this is linux only feature. Windows will ignore these ACLs.
+
+	  If you don't know what Access Control Lists are, say N.

target/linux/generic/files/fs/ntfs3/Makefile

@@ -0,0 +1,43 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# Makefile for the ntfs3 filesystem support.
+#
+
+# to check robot warnings
+ccflags-y += -Wunused-but-set-variable -Wold-style-declaration -Wint-to-pointer-cast
+
+obj-$(CONFIG_NTFS3_FS) += ntfs3.o
+
+ntfs3-y :=	attrib.o \
+		attrlist.o \
+		bitfunc.o \
+		bitmap.o \
+		dir.o \
+		fsntfs.o \
+		frecord.o \
+		file.o \
+		fslog.o \
+		inode.o \
+		index.o \
+		lznt.o \
+		namei.o \
+		record.o \
+		run.o \
+		super.o \
+		upcase.o \
+		xattr.o
+
+ntfs3-$(CONFIG_NTFS3_LZX_XPRESS) += $(addprefix lib/,\
+		decompress_common.o \
+		lzx_decompress.o \
+		xpress_decompress.o \
+		)
+
+ccflags-$(CONFIG_NTFS3_LZX_XPRESS) += -DCONFIG_NTFS3_LZX_XPRESS
+ccflags-$(CONFIG_NTFS3_FS_POSIX_ACL) += -DCONFIG_NTFS3_FS_POSIX_ACL
+
+all:
+	make -C /lib/modules/$(KVERSION)/build M=$(PWD) modules
+
+clean:
+	make -C /lib/modules/$(KVERSION)/build M=$(PWD) clean

target/linux/generic/files/fs/ntfs3/attrlist.c

@@ -0,0 +1,456 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+/* Returns true if le is valid */
+static inline bool al_is_valid_le(const struct ntfs_inode *ni,
+				  struct ATTR_LIST_ENTRY *le)
+{
+	if (!le || !ni->attr_list.le || !ni->attr_list.size)
+		return false;
+
+	return PtrOffset(ni->attr_list.le, le) + le16_to_cpu(le->size) <=
+	       ni->attr_list.size;
+}
+
+void al_destroy(struct ntfs_inode *ni)
+{
+	run_close(&ni->attr_list.run);
+	ntfs_free(ni->attr_list.le);
+	ni->attr_list.le = NULL;
+	ni->attr_list.size = 0;
+	ni->attr_list.dirty = false;
+}
+
+/*
+ * ntfs_load_attr_list
+ *
+ * This method makes sure that the ATTRIB list, if present,
+ * has been properly set up.
+ */
+int ntfs_load_attr_list(struct ntfs_inode *ni, struct ATTRIB *attr)
+{
+	int err;
+	size_t lsize;
+	void *le = NULL;
+
+	if (ni->attr_list.size)
+		return 0;
+
+	if (!attr->non_res) {
+		lsize = le32_to_cpu(attr->res.data_size);
+		le = ntfs_malloc(al_aligned(lsize));
+		if (!le) {
+			err = -ENOMEM;
+			goto out;
+		}
+		memcpy(le, resident_data(attr), lsize);
+	} else if (attr->nres.svcn) {
+		err = -EINVAL;
+		goto out;
+	} else {
+		u16 run_off = le16_to_cpu(attr->nres.run_off);
+
+		lsize = le64_to_cpu(attr->nres.data_size);
+
+		run_init(&ni->attr_list.run);
+
+		err = run_unpack_ex(&ni->attr_list.run, ni->mi.sbi, ni->mi.rno,
+				    0, le64_to_cpu(attr->nres.evcn), 0,
+				    Add2Ptr(attr, run_off),
+				    le32_to_cpu(attr->size) - run_off);
+		if (err < 0)
+			goto out;
+
+		le = ntfs_malloc(al_aligned(lsize));
+		if (!le) {
+			err = -ENOMEM;
+			goto out;
+		}
+
+		err = ntfs_read_run_nb(ni->mi.sbi, &ni->attr_list.run, 0, le,
+				       lsize, NULL);
+		if (err)
+			goto out;
+	}
+
+	ni->attr_list.size = lsize;
+	ni->attr_list.le = le;
+
+	return 0;
+
+out:
+	ni->attr_list.le = le;
+	al_destroy(ni);
+
+	return err;
+}
+
+/*
+ * al_enumerate
+ *
+ * Returns the next list 'le'
+ * if 'le' is NULL then returns the first 'le'
+ */
+struct ATTR_LIST_ENTRY *al_enumerate(struct ntfs_inode *ni,
+				     struct ATTR_LIST_ENTRY *le)
+{
+	size_t off;
+	u16 sz;
+
+	if (!le) {
+		le = ni->attr_list.le;
+	} else {
+		sz = le16_to_cpu(le->size);
+		if (sz < sizeof(struct ATTR_LIST_ENTRY)) {
+			/* Impossible 'cause we should not return such 'le' */
+			return NULL;
+		}
+		le = Add2Ptr(le, sz);
+	}
+
+	/* Check boundary */
+	off = PtrOffset(ni->attr_list.le, le);
+	if (off + sizeof(struct ATTR_LIST_ENTRY) > ni->attr_list.size) {
+		// The regular end of list
+		return NULL;
+	}
+
+	sz = le16_to_cpu(le->size);
+
+	/* Check 'le' for errors */
+	if (sz < sizeof(struct ATTR_LIST_ENTRY) ||
+	    off + sz > ni->attr_list.size ||
+	    sz < le->name_off + le->name_len * sizeof(short)) {
+		return NULL;
+	}
+
+	return le;
+}
+
+/*
+ * al_find_le
+ *
+ * finds the first 'le' in the list which matches type, name and vcn
+ * Returns NULL if not found
+ */
+struct ATTR_LIST_ENTRY *al_find_le(struct ntfs_inode *ni,
+				   struct ATTR_LIST_ENTRY *le,
+				   const struct ATTRIB *attr)
+{
+	CLST svcn = attr_svcn(attr);
+
+	return al_find_ex(ni, le, attr->type, attr_name(attr), attr->name_len,
+			  &svcn);
+}
+
+/*
+ * al_find_ex
+ *
+ * finds the first 'le' in the list which matches type, name and vcn
+ * Returns NULL if not found
+ */
+struct ATTR_LIST_ENTRY *al_find_ex(struct ntfs_inode *ni,
+				   struct ATTR_LIST_ENTRY *le,
+				   enum ATTR_TYPE type, const __le16 *name,
+				   u8 name_len, const CLST *vcn)
+{
+	struct ATTR_LIST_ENTRY *ret = NULL;
+	u32 type_in = le32_to_cpu(type);
+
+	while ((le = al_enumerate(ni, le))) {
+		u64 le_vcn;
+		int diff = le32_to_cpu(le->type) - type_in;
+
+		/* List entries are sorted by type, name and vcn */
+		if (diff < 0)
+			continue;
+
+		if (diff > 0)
+			return ret;
+
+		if (le->name_len != name_len)
+			continue;
+
+		le_vcn = le64_to_cpu(le->vcn);
+		if (!le_vcn) {
+			/*
+			 * compare entry names only for entry with vcn == 0
+			 */
+			diff = ntfs_cmp_names(le_name(le), name_len, name,
+					      name_len, ni->mi.sbi->upcase,
+					      true);
+			if (diff < 0)
+				continue;
+
+			if (diff > 0)
+				return ret;
+		}
+
+		if (!vcn)
+			return le;
+
+		if (*vcn == le_vcn)
+			return le;
+
+		if (*vcn < le_vcn)
+			return ret;
+
+		ret = le;
+	}
+
+	return ret;
+}
+
+/*
+ * al_find_le_to_insert
+ *
+ * finds the first list entry which matches type, name and vcn
+ */
+static struct ATTR_LIST_ENTRY *al_find_le_to_insert(struct ntfs_inode *ni,
+						    enum ATTR_TYPE type,
+						    const __le16 *name,
+						    u8 name_len, CLST vcn)
+{
+	struct ATTR_LIST_ENTRY *le = NULL, *prev;
+	u32 type_in = le32_to_cpu(type);
+
+	/* List entries are sorted by type, name, vcn */
+	while ((le = al_enumerate(ni, prev = le))) {
+		int diff = le32_to_cpu(le->type) - type_in;
+
+		if (diff < 0)
+			continue;
+
+		if (diff > 0)
+			return le;
+
+		if (!le->vcn) {
+			/*
+			 * compare entry names only for entry with vcn == 0
+			 */
+			diff = ntfs_cmp_names(le_name(le), le->name_len, name,
+					      name_len, ni->mi.sbi->upcase,
+					      true);
+			if (diff < 0)
+				continue;
+
+			if (diff > 0)
+				return le;
+		}
+
+		if (le64_to_cpu(le->vcn) >= vcn)
+			return le;
+	}
+
+	return prev ? Add2Ptr(prev, le16_to_cpu(prev->size)) : ni->attr_list.le;
+}
+
+/*
+ * al_add_le
+ *
+ * adds an "attribute list entry" to the list.
+ */
+int al_add_le(struct ntfs_inode *ni, enum ATTR_TYPE type, const __le16 *name,
+	      u8 name_len, CLST svcn, __le16 id, const struct MFT_REF *ref,
+	      struct ATTR_LIST_ENTRY **new_le)
+{
+	int err;
+	struct ATTRIB *attr;
+	struct ATTR_LIST_ENTRY *le;
+	size_t off;
+	u16 sz;
+	size_t asize, new_asize;
+	u64 new_size;
+	typeof(ni->attr_list) *al = &ni->attr_list;
+
+	/*
+	 * Compute the size of the new 'le'
+	 */
+	sz = le_size(name_len);
+	new_size = al->size + sz;
+	asize = al_aligned(al->size);
+	new_asize = al_aligned(new_size);
+
+	/* Scan forward to the point at which the new 'le' should be inserted. */
+	le = al_find_le_to_insert(ni, type, name, name_len, svcn);
+	off = PtrOffset(al->le, le);
+
+	if (new_size > asize) {
+		void *ptr = ntfs_malloc(new_asize);
+
+		if (!ptr)
+			return -ENOMEM;
+
+		memcpy(ptr, al->le, off);
+		memcpy(Add2Ptr(ptr, off + sz), le, al->size - off);
+		le = Add2Ptr(ptr, off);
+		ntfs_free(al->le);
+		al->le = ptr;
+	} else {
+		memmove(Add2Ptr(le, sz), le, al->size - off);
+	}
+
+	al->size = new_size;
+
+	le->type = type;
+	le->size = cpu_to_le16(sz);
+	le->name_len = name_len;
+	le->name_off = offsetof(struct ATTR_LIST_ENTRY, name);
+	le->vcn = cpu_to_le64(svcn);
+	le->ref = *ref;
+	le->id = id;
+	memcpy(le->name, name, sizeof(short) * name_len);
+
+	al->dirty = true;
+
+	err = attr_set_size(ni, ATTR_LIST, NULL, 0, &al->run, new_size,
+			    &new_size, true, &attr);
+	if (err)
+		return err;
+
+	if (attr && attr->non_res) {
+		err = ntfs_sb_write_run(ni->mi.sbi, &al->run, 0, al->le,
+					al->size);
+		if (err)
+			return err;
+	}
+
+	al->dirty = false;
+	*new_le = le;
+
+	return 0;
+}
+
+/*
+ * al_remove_le
+ *
+ * removes 'le' from attribute list
+ */
+bool al_remove_le(struct ntfs_inode *ni, struct ATTR_LIST_ENTRY *le)
+{
+	u16 size;
+	size_t off;
+	typeof(ni->attr_list) *al = &ni->attr_list;
+
+	if (!al_is_valid_le(ni, le))
+		return false;
+
+	/* Save on stack the size of 'le' */
+	size = le16_to_cpu(le->size);
+	off = PtrOffset(al->le, le);
+
+	memmove(le, Add2Ptr(le, size), al->size - (off + size));
+
+	al->size -= size;
+	al->dirty = true;
+
+	return true;
+}
+
+/*
+ * al_delete_le
+ *
+ * deletes from the list the first 'le' which matches its parameters.
+ */
+bool al_delete_le(struct ntfs_inode *ni, enum ATTR_TYPE type, CLST vcn,
+		  const __le16 *name, size_t name_len,
+		  const struct MFT_REF *ref)
+{
+	u16 size;
+	struct ATTR_LIST_ENTRY *le;
+	size_t off;
+	typeof(ni->attr_list) *al = &ni->attr_list;
+
+	/* Scan forward to the first 'le' that matches the input */
+	le = al_find_ex(ni, NULL, type, name, name_len, &vcn);
+	if (!le)
+		return false;
+
+	off = PtrOffset(al->le, le);
+
+next:
+	if (off >= al->size)
+		return false;
+	if (le->type != type)
+		return false;
+	if (le->name_len != name_len)
+		return false;
+	if (name_len && ntfs_cmp_names(le_name(le), name_len, name, name_len,
+				       ni->mi.sbi->upcase, true))
+		return false;
+	if (le64_to_cpu(le->vcn) != vcn)
+		return false;
+
+	/*
+	 * The caller specified a segment reference, so we have to
+	 * scan through the matching entries until we find that segment
+	 * reference or we run of matching entries.
+	 */
+	if (ref && memcmp(ref, &le->ref, sizeof(*ref))) {
+		off += le16_to_cpu(le->size);
+		le = Add2Ptr(al->le, off);
+		goto next;
+	}
+
+	/* Save on stack the size of 'le' */
+	size = le16_to_cpu(le->size);
+	/* Delete 'le'. */
+	memmove(le, Add2Ptr(le, size), al->size - (off + size));
+
+	al->size -= size;
+	al->dirty = true;
+
+	return true;
+}
+
+/*
+ * al_update
+ */
+int al_update(struct ntfs_inode *ni)
+{
+	int err;
+	struct ATTRIB *attr;
+	typeof(ni->attr_list) *al = &ni->attr_list;
+
+	if (!al->dirty || !al->size)
+		return 0;
+
+	/*
+	 * attribute list increased on demand in al_add_le
+	 * attribute list decreased here
+	 */
+	err = attr_set_size(ni, ATTR_LIST, NULL, 0, &al->run, al->size, NULL,
+			    false, &attr);
+	if (err)
+		goto out;
+
+	if (!attr->non_res) {
+		memcpy(resident_data(attr), al->le, al->size);
+	} else {
+		err = ntfs_sb_write_run(ni->mi.sbi, &al->run, 0, al->le,
+					al->size);
+		if (err)
+			goto out;
+
+		attr->nres.valid_size = attr->nres.data_size;
+	}
+
+	ni->mi.dirty = true;
+	al->dirty = false;
+
+out:
+	return err;
+}

target/linux/generic/files/fs/ntfs3/bitfunc.c

@@ -0,0 +1,135 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
+ *
+ */
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+#define BITS_IN_SIZE_T (sizeof(size_t) * 8)
+
+/*
+ * fill_mask[i] - first i bits are '1' , i = 0,1,2,3,4,5,6,7,8
+ * fill_mask[i] = 0xFF >> (8-i)
+ */
+static const u8 fill_mask[] = { 0x00, 0x01, 0x03, 0x07, 0x0F,
+				0x1F, 0x3F, 0x7F, 0xFF };
+
+/*
+ * zero_mask[i] - first i bits are '0' , i = 0,1,2,3,4,5,6,7,8
+ * zero_mask[i] = 0xFF << i
+ */
+static const u8 zero_mask[] = { 0xFF, 0xFE, 0xFC, 0xF8, 0xF0,
+				0xE0, 0xC0, 0x80, 0x00 };
+
+/*
+ * are_bits_clear
+ *
+ * Returns true if all bits [bit, bit+nbits) are zeros "0"
+ */
+bool are_bits_clear(const ulong *lmap, size_t bit, size_t nbits)
+{
+	size_t pos = bit & 7;
+	const u8 *map = (u8 *)lmap + (bit >> 3);
+
+	if (pos) {
+		if (8 - pos >= nbits)
+			return !nbits || !(*map & fill_mask[pos + nbits] &
+					   zero_mask[pos]);
+
+		if (*map++ & zero_mask[pos])
+			return false;
+		nbits -= 8 - pos;
+	}
+
+	pos = ((size_t)map) & (sizeof(size_t) - 1);
+	if (pos) {
+		pos = sizeof(size_t) - pos;
+		if (nbits >= pos * 8) {
+			for (nbits -= pos * 8; pos; pos--, map++) {
+				if (*map)
+					return false;
+			}
+		}
+	}
+
+	for (pos = nbits / BITS_IN_SIZE_T; pos; pos--, map += sizeof(size_t)) {
+		if (*((size_t *)map))
+			return false;
+	}
+
+	for (pos = (nbits % BITS_IN_SIZE_T) >> 3; pos; pos--, map++) {
+		if (*map)
+			return false;
+	}
+
+	pos = nbits & 7;
+	if (pos && (*map & fill_mask[pos]))
+		return false;
+
+	// All bits are zero
+	return true;
+}
+
+/*
+ * are_bits_set
+ *
+ * Returns true if all bits [bit, bit+nbits) are ones "1"
+ */
+bool are_bits_set(const ulong *lmap, size_t bit, size_t nbits)
+{
+	u8 mask;
+	size_t pos = bit & 7;
+	const u8 *map = (u8 *)lmap + (bit >> 3);
+
+	if (pos) {
+		if (8 - pos >= nbits) {
+			mask = fill_mask[pos + nbits] & zero_mask[pos];
+			return !nbits || (*map & mask) == mask;
+		}
+
+		mask = zero_mask[pos];
+		if ((*map++ & mask) != mask)
+			return false;
+		nbits -= 8 - pos;
+	}
+
+	pos = ((size_t)map) & (sizeof(size_t) - 1);
+	if (pos) {
+		pos = sizeof(size_t) - pos;
+		if (nbits >= pos * 8) {
+			for (nbits -= pos * 8; pos; pos--, map++) {
+				if (*map != 0xFF)
+					return false;
+			}
+		}
+	}
+
+	for (pos = nbits / BITS_IN_SIZE_T; pos; pos--, map += sizeof(size_t)) {
+		if (*((size_t *)map) != MINUS_ONE_T)
+			return false;
+	}
+
+	for (pos = (nbits % BITS_IN_SIZE_T) >> 3; pos; pos--, map++) {
+		if (*map != 0xFF)
+			return false;
+	}
+
+	pos = nbits & 7;
+	if (pos) {
+		u8 mask = fill_mask[pos];
+
+		if ((*map & mask) != mask)
+			return false;
+	}
+
+	// All bits are ones
+	return true;
+}

target/linux/generic/files/fs/ntfs3/compat.h

@@ -0,0 +1,39 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __bitmap_set
+#define __bitmap_set(a, b, c)	bitmap_set(a, b, c)
+#endif
+
+#ifndef __bitmap_clear
+#define __bitmap_clear(a, b, c)	bitmap_clear(a, b, c)
+#endif
+
+/*
+ * Copy from include/linux/compiler_attributes.h
+ */
+#ifndef __has_attribute
+#define __has_attribute(x) 0
+#endif
+
+#ifndef fallthrough
+#if __has_attribute(__fallthrough__)
+#define fallthrough __attribute__((__fallthrough__))
+#else
+#define fallthrough do {} while (0)  /* fallthrough */
+#endif
+#endif
+
+/*
+ * Copy from include/linux/build_bug.h
+ */
+#ifndef static_assert
+#define static_assert(expr, ...) __static_assert(expr, ##__VA_ARGS__, #expr)
+#define __static_assert(expr, msg, ...) _Static_assert(expr, msg)
+#endif
+
+/*
+ * Copy from include/linux/overflow.h
+ */
+#ifndef struct_size
+#define struct_size(p, member, n) (sizeof(*(p)) + n * sizeof(*(p)->member))
+#endif

target/linux/generic/files/fs/ntfs3/debug.h

@@ -0,0 +1,64 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ *
+ * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
+ *
+ * useful functions for debuging
+ */
+
+// clang-format off
+#ifndef Add2Ptr
+#define Add2Ptr(P, I)		((void *)((u8 *)(P) + (I)))
+#define PtrOffset(B, O)		((size_t)((size_t)(O) - (size_t)(B)))
+#endif
+
+#define QuadAlign(n)		(((n) + 7u) & (~7u))
+#define IsQuadAligned(n)	(!((size_t)(n)&7u))
+#define Quad2Align(n)		(((n) + 15u) & (~15u))
+#define IsQuad2Aligned(n)	(!((size_t)(n)&15u))
+#define Quad4Align(n)		(((n) + 31u) & (~31u))
+#define IsSizeTAligned(n)	(!((size_t)(n) & (sizeof(size_t) - 1)))
+#define DwordAlign(n)		(((n) + 3u) & (~3u))
+#define IsDwordAligned(n)	(!((size_t)(n)&3u))
+#define WordAlign(n)		(((n) + 1u) & (~1u))
+#define IsWordAligned(n)	(!((size_t)(n)&1u))
+
+#ifdef CONFIG_PRINTK
+__printf(2, 3)
+void ntfs_printk(const struct super_block *sb, const char *fmt, ...);
+__printf(2, 3)
+void ntfs_inode_printk(struct inode *inode, const char *fmt, ...);
+#else
+static inline __printf(2, 3)
+void ntfs_printk(const struct super_block *sb, const char *fmt, ...)
+{
+}
+
+static inline __printf(2, 3)
+void ntfs_inode_printk(struct inode *inode, const char *fmt, ...)
+{
+}
+#endif
+
+/*
+ * Logging macros ( thanks Joe Perches <joe@perches.com> for implementation )
+ */
+
+#define ntfs_err(sb, fmt, ...)  ntfs_printk(sb, KERN_ERR fmt, ##__VA_ARGS__)
+#define ntfs_warn(sb, fmt, ...) ntfs_printk(sb, KERN_WARNING fmt, ##__VA_ARGS__)
+#define ntfs_info(sb, fmt, ...) ntfs_printk(sb, KERN_INFO fmt, ##__VA_ARGS__)
+#define ntfs_notice(sb, fmt, ...)                                              \
+	ntfs_printk(sb, KERN_NOTICE fmt, ##__VA_ARGS__)
+
+#define ntfs_inode_err(inode, fmt, ...)                                        \
+	ntfs_inode_printk(inode, KERN_ERR fmt, ##__VA_ARGS__)
+#define ntfs_inode_warn(inode, fmt, ...)                                       \
+	ntfs_inode_printk(inode, KERN_WARNING fmt, ##__VA_ARGS__)
+
+#define ntfs_malloc(s)		kmalloc(s, GFP_NOFS)
+#define ntfs_zalloc(s)		kzalloc(s, GFP_NOFS)
+#define ntfs_vmalloc(s)		kvmalloc(s, GFP_KERNEL)
+#define ntfs_free(p)		kfree(p)
+#define ntfs_vfree(p)		kvfree(p)
+#define ntfs_memdup(src, len)	kmemdup(src, len, GFP_NOFS)
+// clang-format on

target/linux/generic/files/fs/ntfs3/dir.c

@@ -0,0 +1,598 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
+ *
+ *  directory handling functions for ntfs-based filesystems
+ *
+ */
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+#include <linux/version.h>
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 16, 0)
+#include <linux/iversion.h>
+#endif
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+/*
+ * Convert little endian utf16 to nls string
+ */
+int ntfs_utf16_to_nls(struct ntfs_sb_info *sbi, const struct le_str *uni,
+		      u8 *buf, int buf_len)
+{
+	int ret, uni_len, warn;
+	const __le16 *ip;
+	u8 *op;
+	struct nls_table *nls = sbi->options.nls;
+
+	static_assert(sizeof(wchar_t) == sizeof(__le16));
+
+	if (!nls) {
+		/* utf16 -> utf8 */
+		ret = utf16s_to_utf8s((wchar_t *)uni->name, uni->len,
+				      UTF16_LITTLE_ENDIAN, buf, buf_len);
+		buf[ret] = '\0';
+		return ret;
+	}
+
+	ip = uni->name;
+	op = buf;
+	uni_len = uni->len;
+	warn = 0;
+
+	while (uni_len--) {
+		u16 ec;
+		int charlen;
+		char dump[5];
+
+		if (buf_len < NLS_MAX_CHARSET_SIZE) {
+			ntfs_warn(sbi->sb,
+				  "filename was truncated while converting.");
+			break;
+		}
+
+		ec = le16_to_cpu(*ip++);
+		charlen = nls->uni2char(ec, op, buf_len);
+
+		if (charlen > 0) {
+			op += charlen;
+			buf_len -= charlen;
+			continue;
+		}
+
+		*op++ = '_';
+		buf_len -= 1;
+		if (warn)
+			continue;
+
+		warn = 1;
+		hex_byte_pack(&dump[0], ec >> 8);
+		hex_byte_pack(&dump[2], ec);
+		dump[4] = 0;
+
+		ntfs_err(sbi->sb, "failed to convert \"%s\" to %s", dump,
+			 nls->charset);
+	}
+
+	*op = '\0';
+	return op - buf;
+}
+
+// clang-format off
+#define PLANE_SIZE	0x00010000
+
+#define SURROGATE_PAIR	0x0000d800
+#define SURROGATE_LOW	0x00000400
+#define SURROGATE_BITS	0x000003ff
+// clang-format on
+
+/*
+ * modified version of put_utf16 from fs/nls/nls_base.c
+ * is sparse warnings free
+ */
+static inline void put_utf16(wchar_t *s, unsigned int c,
+			     enum utf16_endian endian)
+{
+	static_assert(sizeof(wchar_t) == sizeof(__le16));
+	static_assert(sizeof(wchar_t) == sizeof(__be16));
+
+	switch (endian) {
+	default:
+		*s = (wchar_t)c;
+		break;
+	case UTF16_LITTLE_ENDIAN:
+		*(__le16 *)s = __cpu_to_le16(c);
+		break;
+	case UTF16_BIG_ENDIAN:
+		*(__be16 *)s = __cpu_to_be16(c);
+		break;
+	}
+}
+
+/*
+ * modified version of 'utf8s_to_utf16s' allows to
+ * detect -ENAMETOOLONG without writing out of expected maximum
+ */
+static int _utf8s_to_utf16s(const u8 *s, int inlen, enum utf16_endian endian,
+			    wchar_t *pwcs, int maxout)
+{
+	u16 *op;
+	int size;
+	unicode_t u;
+
+	op = pwcs;
+	while (inlen > 0 && *s) {
+		if (*s & 0x80) {
+			size = utf8_to_utf32(s, inlen, &u);
+			if (size < 0)
+				return -EINVAL;
+			s += size;
+			inlen -= size;
+
+			if (u >= PLANE_SIZE) {
+				if (maxout < 2)
+					return -ENAMETOOLONG;
+
+				u -= PLANE_SIZE;
+				put_utf16(op++,
+					  SURROGATE_PAIR |
+						  ((u >> 10) & SURROGATE_BITS),
+					  endian);
+				put_utf16(op++,
+					  SURROGATE_PAIR | SURROGATE_LOW |
+						  (u & SURROGATE_BITS),
+					  endian);
+				maxout -= 2;
+			} else {
+				if (maxout < 1)
+					return -ENAMETOOLONG;
+
+				put_utf16(op++, u, endian);
+				maxout--;
+			}
+		} else {
+			if (maxout < 1)
+				return -ENAMETOOLONG;
+
+			put_utf16(op++, *s++, endian);
+			inlen--;
+			maxout--;
+		}
+	}
+	return op - pwcs;
+}
+
+/*
+ * Convert input string to utf16
+ *
+ * name, name_len - input name
+ * uni, max_ulen - destination memory
+ * endian - endian of target utf16 string
+ *
+ * This function is called:
+ * - to create ntfs name
+ * - to create symlink
+ *
+ * returns utf16 string length or error (if negative)
+ */
+int ntfs_nls_to_utf16(struct ntfs_sb_info *sbi, const u8 *name, u32 name_len,
+		      struct cpu_str *uni, u32 max_ulen,
+		      enum utf16_endian endian)
+{
+	int ret, slen;
+	const u8 *end;
+	struct nls_table *nls = sbi->options.nls;
+	u16 *uname = uni->name;
+
+	static_assert(sizeof(wchar_t) == sizeof(u16));
+
+	if (!nls) {
+		/* utf8 -> utf16 */
+		ret = _utf8s_to_utf16s(name, name_len, endian, uname, max_ulen);
+		uni->len = ret;
+		return ret;
+	}
+
+	for (ret = 0, end = name + name_len; name < end; ret++, name += slen) {
+		if (ret >= max_ulen)
+			return -ENAMETOOLONG;
+
+		slen = nls->char2uni(name, end - name, uname + ret);
+		if (!slen)
+			return -EINVAL;
+		if (slen < 0)
+			return slen;
+	}
+
+#ifdef __BIG_ENDIAN
+	if (endian == UTF16_LITTLE_ENDIAN) {
+		int i = ret;
+
+		while (i--) {
+			__cpu_to_le16s(uname);
+			uname++;
+		}
+	}
+#else
+	if (endian == UTF16_BIG_ENDIAN) {
+		int i = ret;
+
+		while (i--) {
+			__cpu_to_be16s(uname);
+			uname++;
+		}
+	}
+#endif
+
+	uni->len = ret;
+	return ret;
+}
+
+/* helper function */
+struct inode *dir_search_u(struct inode *dir, const struct cpu_str *uni,
+			   struct ntfs_fnd *fnd)
+{
+	int err = 0;
+	struct super_block *sb = dir->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct ntfs_inode *ni = ntfs_i(dir);
+	struct NTFS_DE *e;
+	int diff;
+	struct inode *inode = NULL;
+	struct ntfs_fnd *fnd_a = NULL;
+
+	if (!fnd) {
+		fnd_a = fnd_get();
+		if (!fnd_a) {
+			err = -ENOMEM;
+			goto out;
+		}
+		fnd = fnd_a;
+	}
+
+	err = indx_find(&ni->dir, ni, NULL, uni, 0, sbi, &diff, &e, fnd);
+
+	if (err)
+		goto out;
+
+	if (diff) {
+		err = -ENOENT;
+		goto out;
+	}
+
+	inode = ntfs_iget5(sb, &e->ref, uni);
+	if (!IS_ERR(inode) && is_bad_inode(inode)) {
+		iput(inode);
+		err = -EINVAL;
+	}
+out:
+	fnd_put(fnd_a);
+
+	return err == -ENOENT ? NULL : err ? ERR_PTR(err) : inode;
+}
+
+static inline int ntfs_filldir(struct ntfs_sb_info *sbi, struct ntfs_inode *ni,
+			       const struct NTFS_DE *e, u8 *name,
+			       struct dir_context *ctx)
+{
+	const struct ATTR_FILE_NAME *fname;
+	unsigned long ino;
+	int name_len;
+	u32 dt_type;
+
+	fname = Add2Ptr(e, sizeof(struct NTFS_DE));
+
+	if (fname->type == FILE_NAME_DOS)
+		return 0;
+
+	if (!mi_is_ref(&ni->mi, &fname->home))
+		return 0;
+
+	ino = ino_get(&e->ref);
+
+	if (ino == MFT_REC_ROOT)
+		return 0;
+
+	/* Skip meta files ( unless option to show metafiles is set ) */
+	if (!sbi->options.showmeta && ntfs_is_meta_file(sbi, ino))
+		return 0;
+
+	if (sbi->options.nohidden && (fname->dup.fa & FILE_ATTRIBUTE_HIDDEN))
+		return 0;
+
+	name_len = ntfs_utf16_to_nls(sbi, (struct le_str *)&fname->name_len,
+				     name, PATH_MAX);
+	if (name_len <= 0) {
+		ntfs_warn(sbi->sb, "failed to convert name for inode %lx.",
+			  ino);
+		return 0;
+	}
+
+	dt_type = (fname->dup.fa & FILE_ATTRIBUTE_DIRECTORY) ? DT_DIR : DT_REG;
+
+	return !dir_emit(ctx, (s8 *)name, name_len, ino, dt_type);
+}
+
+/*
+ * ntfs_read_hdr
+ *
+ * helper function 'ntfs_readdir'
+ */
+static int ntfs_read_hdr(struct ntfs_sb_info *sbi, struct ntfs_inode *ni,
+			 const struct INDEX_HDR *hdr, u64 vbo, u64 pos,
+			 u8 *name, struct dir_context *ctx)
+{
+	int err;
+	const struct NTFS_DE *e;
+	u32 e_size;
+	u32 end = le32_to_cpu(hdr->used);
+	u32 off = le32_to_cpu(hdr->de_off);
+
+	for (;; off += e_size) {
+		if (off + sizeof(struct NTFS_DE) > end)
+			return -1;
+
+		e = Add2Ptr(hdr, off);
+		e_size = le16_to_cpu(e->size);
+		if (e_size < sizeof(struct NTFS_DE) || off + e_size > end)
+			return -1;
+
+		if (de_is_last(e))
+			return 0;
+
+		/* Skip already enumerated*/
+		if (vbo + off < pos)
+			continue;
+
+		if (le16_to_cpu(e->key_size) < SIZEOF_ATTRIBUTE_FILENAME)
+			return -1;
+
+		ctx->pos = vbo + off;
+
+		/* Submit the name to the filldir callback. */
+		err = ntfs_filldir(sbi, ni, e, name, ctx);
+		if (err)
+			return err;
+	}
+}
+
+/*
+ * file_operations::iterate_shared
+ *
+ * Use non sorted enumeration.
+ * We have an example of broken volume where sorted enumeration
+ * counts each name twice
+ */
+static int ntfs_readdir(struct file *file, struct dir_context *ctx)
+{
+	const struct INDEX_ROOT *root;
+	u64 vbo;
+	size_t bit;
+	loff_t eod;
+	int err = 0;
+	struct inode *dir = file_inode(file);
+	struct ntfs_inode *ni = ntfs_i(dir);
+	struct super_block *sb = dir->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	loff_t i_size = i_size_read(dir);
+	u32 pos = ctx->pos;
+	u8 *name = NULL;
+	struct indx_node *node = NULL;
+	u8 index_bits = ni->dir.index_bits;
+
+	/* name is a buffer of PATH_MAX length */
+	static_assert(NTFS_NAME_LEN * 4 < PATH_MAX);
+
+	eod = i_size + sbi->record_size;
+
+	if (pos >= eod)
+		return 0;
+
+	if (!dir_emit_dots(file, ctx))
+		return 0;
+
+	/* allocate PATH_MAX bytes */
+	name = __getname();
+	if (!name)
+		return -ENOMEM;
+
+	if (!ni->mi_loaded && ni->attr_list.size) {
+		/*
+		 * directory inode is locked for read
+		 * load all subrecords to avoid 'write' access to 'ni' during
+		 * directory reading
+		 */
+		ni_lock(ni);
+		if (!ni->mi_loaded && ni->attr_list.size) {
+			err = ni_load_all_mi(ni);
+			if (!err)
+				ni->mi_loaded = true;
+		}
+		ni_unlock(ni);
+		if (err)
+			goto out;
+	}
+
+	root = indx_get_root(&ni->dir, ni, NULL, NULL);
+	if (!root) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (pos >= sbi->record_size) {
+		bit = (pos - sbi->record_size) >> index_bits;
+	} else {
+		err = ntfs_read_hdr(sbi, ni, &root->ihdr, 0, pos, name, ctx);
+		if (err)
+			goto out;
+		bit = 0;
+	}
+
+	if (!i_size) {
+		ctx->pos = eod;
+		goto out;
+	}
+
+	for (;;) {
+		vbo = (u64)bit << index_bits;
+		if (vbo >= i_size) {
+			ctx->pos = eod;
+			goto out;
+		}
+
+		err = indx_used_bit(&ni->dir, ni, &bit);
+		if (err)
+			goto out;
+
+		if (bit == MINUS_ONE_T) {
+			ctx->pos = eod;
+			goto out;
+		}
+
+		vbo = (u64)bit << index_bits;
+		if (vbo >= i_size) {
+			ntfs_inode_err(dir, "Looks like your dir is corrupt");
+			err = -EINVAL;
+			goto out;
+		}
+
+		err = indx_read(&ni->dir, ni, bit << ni->dir.idx2vbn_bits,
+				&node);
+		if (err)
+			goto out;
+
+		err = ntfs_read_hdr(sbi, ni, &node->index->ihdr,
+				    vbo + sbi->record_size, pos, name, ctx);
+		if (err)
+			goto out;
+
+		bit += 1;
+	}
+
+out:
+
+	__putname(name);
+	put_indx_node(node);
+
+	if (err == -ENOENT) {
+		err = 0;
+		ctx->pos = pos;
+	}
+
+	return err;
+}
+
+static int ntfs_dir_count(struct inode *dir, bool *is_empty, size_t *dirs,
+			  size_t *files)
+{
+	int err = 0;
+	struct ntfs_inode *ni = ntfs_i(dir);
+	struct NTFS_DE *e = NULL;
+	struct INDEX_ROOT *root;
+	struct INDEX_HDR *hdr;
+	const struct ATTR_FILE_NAME *fname;
+	u32 e_size, off, end;
+	u64 vbo = 0;
+	size_t drs = 0, fles = 0, bit = 0;
+	loff_t i_size = ni->vfs_inode.i_size;
+	struct indx_node *node = NULL;
+	u8 index_bits = ni->dir.index_bits;
+
+	if (is_empty)
+		*is_empty = true;
+
+	root = indx_get_root(&ni->dir, ni, NULL, NULL);
+	if (!root)
+		return -EINVAL;
+
+	hdr = &root->ihdr;
+
+	for (;;) {
+		end = le32_to_cpu(hdr->used);
+		off = le32_to_cpu(hdr->de_off);
+
+		for (; off + sizeof(struct NTFS_DE) <= end; off += e_size) {
+			e = Add2Ptr(hdr, off);
+			e_size = le16_to_cpu(e->size);
+			if (e_size < sizeof(struct NTFS_DE) ||
+			    off + e_size > end)
+				break;
+
+			if (de_is_last(e))
+				break;
+
+			fname = de_get_fname(e);
+			if (!fname)
+				continue;
+
+			if (fname->type == FILE_NAME_DOS)
+				continue;
+
+			if (is_empty) {
+				*is_empty = false;
+				if (!dirs && !files)
+					goto out;
+			}
+
+			if (fname->dup.fa & FILE_ATTRIBUTE_DIRECTORY)
+				drs += 1;
+			else
+				fles += 1;
+		}
+
+		if (vbo >= i_size)
+			goto out;
+
+		err = indx_used_bit(&ni->dir, ni, &bit);
+		if (err)
+			goto out;
+
+		if (bit == MINUS_ONE_T)
+			goto out;
+
+		vbo = (u64)bit << index_bits;
+		if (vbo >= i_size)
+			goto out;
+
+		err = indx_read(&ni->dir, ni, bit << ni->dir.idx2vbn_bits,
+				&node);
+		if (err)
+			goto out;
+
+		hdr = &node->index->ihdr;
+		bit += 1;
+		vbo = (u64)bit << ni->dir.idx2vbn_bits;
+	}
+
+out:
+	put_indx_node(node);
+	if (dirs)
+		*dirs = drs;
+	if (files)
+		*files = fles;
+
+	return err;
+}
+
+bool dir_is_empty(struct inode *dir)
+{
+	bool is_empty = false;
+
+	ntfs_dir_count(dir, &is_empty, NULL, NULL);
+
+	return is_empty;
+}
+
+const struct file_operations ntfs_dir_operations = {
+	.llseek = generic_file_llseek,
+	.read = generic_read_dir,
+	.iterate_shared = ntfs_readdir,
+	.fsync = generic_file_fsync,
+	.open = ntfs_file_open,
+};

target/linux/generic/files/fs/ntfs3/lib/decompress_common.c

@@ -0,0 +1,332 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * decompress_common.c - Code shared by the XPRESS and LZX decompressors
+ *
+ * Copyright (C) 2015 Eric Biggers
+ *
+ * This program is free software: you can redistribute it and/or modify it under
+ * the terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 2 of the License, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "decompress_common.h"
+
+/*
+ * make_huffman_decode_table() -
+ *
+ * Build a decoding table for a canonical prefix code, or "Huffman code".
+ *
+ * This is an internal function, not part of the library API!
+ *
+ * This takes as input the length of the codeword for each symbol in the
+ * alphabet and produces as output a table that can be used for fast
+ * decoding of prefix-encoded symbols using read_huffsym().
+ *
+ * Strictly speaking, a canonical prefix code might not be a Huffman
+ * code.  But this algorithm will work either way; and in fact, since
+ * Huffman codes are defined in terms of symbol frequencies, there is no
+ * way for the decompressor to know whether the code is a true Huffman
+ * code or not until all symbols have been decoded.
+ *
+ * Because the prefix code is assumed to be "canonical", it can be
+ * reconstructed directly from the codeword lengths.  A prefix code is
+ * canonical if and only if a longer codeword never lexicographically
+ * precedes a shorter codeword, and the lexicographic ordering of
+ * codewords of the same length is the same as the lexicographic ordering
+ * of the corresponding symbols.  Consequently, we can sort the symbols
+ * primarily by codeword length and secondarily by symbol value, then
+ * reconstruct the prefix code by generating codewords lexicographically
+ * in that order.
+ *
+ * This function does not, however, generate the prefix code explicitly.
+ * Instead, it directly builds a table for decoding symbols using the
+ * code.  The basic idea is this: given the next 'max_codeword_len' bits
+ * in the input, we can look up the decoded symbol by indexing a table
+ * containing 2**max_codeword_len entries.  A codeword with length
+ * 'max_codeword_len' will have exactly one entry in this table, whereas
+ * a codeword shorter than 'max_codeword_len' will have multiple entries
+ * in this table.  Precisely, a codeword of length n will be represented
+ * by 2**(max_codeword_len - n) entries in this table.  The 0-based index
+ * of each such entry will contain the corresponding codeword as a prefix
+ * when zero-padded on the left to 'max_codeword_len' binary digits.
+ *
+ * That's the basic idea, but we implement two optimizations regarding
+ * the format of the decode table itself:
+ *
+ * - For many compression formats, the maximum codeword length is too
+ *   long for it to be efficient to build the full decoding table
+ *   whenever a new prefix code is used.  Instead, we can build the table
+ *   using only 2**table_bits entries, where 'table_bits' is some number
+ *   less than or equal to 'max_codeword_len'.  Then, only codewords of
+ *   length 'table_bits' and shorter can be directly looked up.  For
+ *   longer codewords, the direct lookup instead produces the root of a
+ *   binary tree.  Using this tree, the decoder can do traditional
+ *   bit-by-bit decoding of the remainder of the codeword.  Child nodes
+ *   are allocated in extra entries at the end of the table; leaf nodes
+ *   contain symbols.  Note that the long-codeword case is, in general,
+ *   not performance critical, since in Huffman codes the most frequently
+ *   used symbols are assigned the shortest codeword lengths.
+ *
+ * - When we decode a symbol using a direct lookup of the table, we still
+ *   need to know its length so that the bitstream can be advanced by the
+ *   appropriate number of bits.  The simple solution is to simply retain
+ *   the 'lens' array and use the decoded symbol as an index into it.
+ *   However, this requires two separate array accesses in the fast path.
+ *   The optimization is to store the length directly in the decode
+ *   table.  We use the bottom 11 bits for the symbol and the top 5 bits
+ *   for the length.  In addition, to combine this optimization with the
+ *   previous one, we introduce a special case where the top 2 bits of
+ *   the length are both set if the entry is actually the root of a
+ *   binary tree.
+ *
+ * @decode_table:
+ *	The array in which to create the decoding table.  This must have
+ *	a length of at least ((2**table_bits) + 2 * num_syms) entries.
+ *
+ * @num_syms:
+ *	The number of symbols in the alphabet; also, the length of the
+ *	'lens' array.  Must be less than or equal to 2048.
+ *
+ * @table_bits:
+ *	The order of the decode table size, as explained above.  Must be
+ *	less than or equal to 13.
+ *
+ * @lens:
+ *	An array of length @num_syms, indexable by symbol, that gives the
+ *	length of the codeword, in bits, for that symbol.  The length can
+ *	be 0, which means that the symbol does not have a codeword
+ *	assigned.
+ *
+ * @max_codeword_len:
+ *	The longest codeword length allowed in the compression format.
+ *	All entries in 'lens' must be less than or equal to this value.
+ *	This must be less than or equal to 23.
+ *
+ * @working_space
+ *	A temporary array of length '2 * (max_codeword_len + 1) +
+ *	num_syms'.
+ *
+ * Returns 0 on success, or -1 if the lengths do not form a valid prefix
+ * code.
+ */
+int make_huffman_decode_table(u16 decode_table[], const u32 num_syms,
+			      const u32 table_bits, const u8 lens[],
+			      const u32 max_codeword_len,
+			      u16 working_space[])
+{
+	const u32 table_num_entries = 1 << table_bits;
+	u16 * const len_counts = &working_space[0];
+	u16 * const offsets = &working_space[1 * (max_codeword_len + 1)];
+	u16 * const sorted_syms = &working_space[2 * (max_codeword_len + 1)];
+	int left;
+	void *decode_table_ptr;
+	u32 sym_idx;
+	u32 codeword_len;
+	u32 stores_per_loop;
+	u32 decode_table_pos;
+	u32 len;
+	u32 sym;
+
+	/* Count how many symbols have each possible codeword length.
+	 * Note that a length of 0 indicates the corresponding symbol is not
+	 * used in the code and therefore does not have a codeword.
+	 */
+	for (len = 0; len <= max_codeword_len; len++)
+		len_counts[len] = 0;
+	for (sym = 0; sym < num_syms; sym++)
+		len_counts[lens[sym]]++;
+
+	/* We can assume all lengths are <= max_codeword_len, but we
+	 * cannot assume they form a valid prefix code.  A codeword of
+	 * length n should require a proportion of the codespace equaling
+	 * (1/2)^n.  The code is valid if and only if the codespace is
+	 * exactly filled by the lengths, by this measure.
+	 */
+	left = 1;
+	for (len = 1; len <= max_codeword_len; len++) {
+		left <<= 1;
+		left -= len_counts[len];
+		if (left < 0) {
+			/* The lengths overflow the codespace; that is, the code
+			 * is over-subscribed.
+			 */
+			return -1;
+		}
+	}
+
+	if (left) {
+		/* The lengths do not fill the codespace; that is, they form an
+		 * incomplete set.
+		 */
+		if (left == (1 << max_codeword_len)) {
+			/* The code is completely empty.  This is arguably
+			 * invalid, but in fact it is valid in LZX and XPRESS,
+			 * so we must allow it.  By definition, no symbols can
+			 * be decoded with an empty code.  Consequently, we
+			 * technically don't even need to fill in the decode
+			 * table.  However, to avoid accessing uninitialized
+			 * memory if the algorithm nevertheless attempts to
+			 * decode symbols using such a code, we zero out the
+			 * decode table.
+			 */
+			memset(decode_table, 0,
+			       table_num_entries * sizeof(decode_table[0]));
+			return 0;
+		}
+		return -1;
+	}
+
+	/* Sort the symbols primarily by length and secondarily by symbol order.
+	 */
+
+	/* Initialize 'offsets' so that offsets[len] for 1 <= len <=
+	 * max_codeword_len is the number of codewords shorter than 'len' bits.
+	 */
+	offsets[1] = 0;
+	for (len = 1; len < max_codeword_len; len++)
+		offsets[len + 1] = offsets[len] + len_counts[len];
+
+	/* Use the 'offsets' array to sort the symbols.  Note that we do not
+	 * include symbols that are not used in the code.  Consequently, fewer
+	 * than 'num_syms' entries in 'sorted_syms' may be filled.
+	 */
+	for (sym = 0; sym < num_syms; sym++)
+		if (lens[sym])
+			sorted_syms[offsets[lens[sym]]++] = sym;
+
+	/* Fill entries for codewords with length <= table_bits
+	 * --- that is, those short enough for a direct mapping.
+	 *
+	 * The table will start with entries for the shortest codeword(s), which
+	 * have the most entries.  From there, the number of entries per
+	 * codeword will decrease.
+	 */
+	decode_table_ptr = decode_table;
+	sym_idx = 0;
+	codeword_len = 1;
+	stores_per_loop = (1 << (table_bits - codeword_len));
+	for (; stores_per_loop != 0; codeword_len++, stores_per_loop >>= 1) {
+		u32 end_sym_idx = sym_idx + len_counts[codeword_len];
+
+		for (; sym_idx < end_sym_idx; sym_idx++) {
+			u16 entry;
+			u16 *p;
+			u32 n;
+
+			entry = ((u32)codeword_len << 11) | sorted_syms[sym_idx];
+			p = (u16 *)decode_table_ptr;
+			n = stores_per_loop;
+
+			do {
+				*p++ = entry;
+			} while (--n);
+
+			decode_table_ptr = p;
+		}
+	}
+
+	/* If we've filled in the entire table, we are done.  Otherwise,
+	 * there are codewords longer than table_bits for which we must
+	 * generate binary trees.
+	 */
+	decode_table_pos = (u16 *)decode_table_ptr - decode_table;
+	if (decode_table_pos != table_num_entries) {
+		u32 j;
+		u32 next_free_tree_slot;
+		u32 cur_codeword;
+
+		/* First, zero out the remaining entries.  This is
+		 * necessary so that these entries appear as
+		 * "unallocated" in the next part.  Each of these entries
+		 * will eventually be filled with the representation of
+		 * the root node of a binary tree.
+		 */
+		j = decode_table_pos;
+		do {
+			decode_table[j] = 0;
+		} while (++j != table_num_entries);
+
+		/* We allocate child nodes starting at the end of the
+		 * direct lookup table.  Note that there should be
+		 * 2*num_syms extra entries for this purpose, although
+		 * fewer than this may actually be needed.
+		 */
+		next_free_tree_slot = table_num_entries;
+
+		/* Iterate through each codeword with length greater than
+		 * 'table_bits', primarily in order of codeword length
+		 * and secondarily in order of symbol.
+		 */
+		for (cur_codeword = decode_table_pos << 1;
+		     codeword_len <= max_codeword_len;
+		     codeword_len++, cur_codeword <<= 1) {
+			u32 end_sym_idx = sym_idx + len_counts[codeword_len];
+
+			for (; sym_idx < end_sym_idx; sym_idx++, cur_codeword++) {
+				/* 'sorted_sym' is the symbol represented by the
+				 * codeword.
+				 */
+				u32 sorted_sym = sorted_syms[sym_idx];
+				u32 extra_bits = codeword_len - table_bits;
+				u32 node_idx = cur_codeword >> extra_bits;
+
+				/* Go through each bit of the current codeword
+				 * beyond the prefix of length @table_bits and
+				 * walk the appropriate binary tree, allocating
+				 * any slots that have not yet been allocated.
+				 *
+				 * Note that the 'pointer' entry to the binary
+				 * tree, which is stored in the direct lookup
+				 * portion of the table, is represented
+				 * identically to other internal (non-leaf)
+				 * nodes of the binary tree; it can be thought
+				 * of as simply the root of the tree.  The
+				 * representation of these internal nodes is
+				 * simply the index of the left child combined
+				 * with the special bits 0xC000 to distingush
+				 * the entry from direct mapping and leaf node
+				 * entries.
+				 */
+				do {
+					/* At least one bit remains in the
+					 * codeword, but the current node is an
+					 * unallocated leaf.  Change it to an
+					 * internal node.
+					 */
+					if (decode_table[node_idx] == 0) {
+						decode_table[node_idx] =
+							next_free_tree_slot | 0xC000;
+						decode_table[next_free_tree_slot++] = 0;
+						decode_table[next_free_tree_slot++] = 0;
+					}
+
+					/* Go to the left child if the next bit
+					 * in the codeword is 0; otherwise go to
+					 * the right child.
+					 */
+					node_idx = decode_table[node_idx] & 0x3FFF;
+					--extra_bits;
+					node_idx += (cur_codeword >> extra_bits) & 1;
+				} while (extra_bits != 0);
+
+				/* We've traversed the tree using the entire
+				 * codeword, and we're now at the entry where
+				 * the actual symbol will be stored.  This is
+				 * distinguished from internal nodes by not
+				 * having its high two bits set.
+				 */
+				decode_table[node_idx] = sorted_sym;
+			}
+		}
+	}
+	return 0;
+}

target/linux/generic/files/fs/ntfs3/lib/decompress_common.h

@@ -0,0 +1,352 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+
+/*
+ * decompress_common.h - Code shared by the XPRESS and LZX decompressors
+ *
+ * Copyright (C) 2015 Eric Biggers
+ *
+ * This program is free software: you can redistribute it and/or modify it under
+ * the terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 2 of the License, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/string.h>
+#include <linux/compiler.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <asm/unaligned.h>
+
+
+/* "Force inline" macro (not required, but helpful for performance)  */
+#define forceinline __always_inline
+
+/* Enable whole-word match copying on selected architectures  */
+#if defined(__i386__) || defined(__x86_64__) || defined(__ARM_FEATURE_UNALIGNED)
+#  define FAST_UNALIGNED_ACCESS
+#endif
+
+/* Size of a machine word  */
+#define WORDBYTES (sizeof(size_t))
+
+static forceinline void
+copy_unaligned_word(const void *src, void *dst)
+{
+	put_unaligned(get_unaligned((const size_t *)src), (size_t *)dst);
+}
+
+
+/* Generate a "word" with platform-dependent size whose bytes all contain the
+ * value 'b'.
+ */
+static forceinline size_t repeat_byte(u8 b)
+{
+	size_t v;
+
+	v = b;
+	v |= v << 8;
+	v |= v << 16;
+	v |= v << ((WORDBYTES == 8) ? 32 : 0);
+	return v;
+}
+
+/* Structure that encapsulates a block of in-memory data being interpreted as a
+ * stream of bits, optionally with interwoven literal bytes.  Bits are assumed
+ * to be stored in little endian 16-bit coding units, with the bits ordered high
+ * to low.
+ */
+struct input_bitstream {
+
+	/* Bits that have been read from the input buffer.  The bits are
+	 * left-justified; the next bit is always bit 31.
+	 */
+	u32 bitbuf;
+
+	/* Number of bits currently held in @bitbuf.  */
+	u32 bitsleft;
+
+	/* Pointer to the next byte to be retrieved from the input buffer.  */
+	const u8 *next;
+
+	/* Pointer to just past the end of the input buffer.  */
+	const u8 *end;
+};
+
+/* Initialize a bitstream to read from the specified input buffer.  */
+static forceinline void init_input_bitstream(struct input_bitstream *is,
+					     const void *buffer, u32 size)
+{
+	is->bitbuf = 0;
+	is->bitsleft = 0;
+	is->next = buffer;
+	is->end = is->next + size;
+}
+
+/* Ensure the bit buffer variable for the bitstream contains at least @num_bits
+ * bits.  Following this, bitstream_peek_bits() and/or bitstream_remove_bits()
+ * may be called on the bitstream to peek or remove up to @num_bits bits.  Note
+ * that @num_bits must be <= 16.
+ */
+static forceinline void bitstream_ensure_bits(struct input_bitstream *is,
+					      u32 num_bits)
+{
+	if (is->bitsleft < num_bits) {
+		if (is->end - is->next >= 2) {
+			is->bitbuf |= (u32)get_unaligned_le16(is->next)
+					<< (16 - is->bitsleft);
+			is->next += 2;
+		}
+		is->bitsleft += 16;
+	}
+}
+
+/* Return the next @num_bits bits from the bitstream, without removing them.
+ * There must be at least @num_bits remaining in the buffer variable, from a
+ * previous call to bitstream_ensure_bits().
+ */
+static forceinline u32
+bitstream_peek_bits(const struct input_bitstream *is, const u32 num_bits)
+{
+	return (is->bitbuf >> 1) >> (sizeof(is->bitbuf) * 8 - num_bits - 1);
+}
+
+/* Remove @num_bits from the bitstream.  There must be at least @num_bits
+ * remaining in the buffer variable, from a previous call to
+ * bitstream_ensure_bits().
+ */
+static forceinline void
+bitstream_remove_bits(struct input_bitstream *is, u32 num_bits)
+{
+	is->bitbuf <<= num_bits;
+	is->bitsleft -= num_bits;
+}
+
+/* Remove and return @num_bits bits from the bitstream.  There must be at least
+ * @num_bits remaining in the buffer variable, from a previous call to
+ * bitstream_ensure_bits().
+ */
+static forceinline u32
+bitstream_pop_bits(struct input_bitstream *is, u32 num_bits)
+{
+	u32 bits = bitstream_peek_bits(is, num_bits);
+
+	bitstream_remove_bits(is, num_bits);
+	return bits;
+}
+
+/* Read and return the next @num_bits bits from the bitstream.  */
+static forceinline u32
+bitstream_read_bits(struct input_bitstream *is, u32 num_bits)
+{
+	bitstream_ensure_bits(is, num_bits);
+	return bitstream_pop_bits(is, num_bits);
+}
+
+/* Read and return the next literal byte embedded in the bitstream.  */
+static forceinline u8
+bitstream_read_byte(struct input_bitstream *is)
+{
+	if (unlikely(is->end == is->next))
+		return 0;
+	return *is->next++;
+}
+
+/* Read and return the next 16-bit integer embedded in the bitstream.  */
+static forceinline u16
+bitstream_read_u16(struct input_bitstream *is)
+{
+	u16 v;
+
+	if (unlikely(is->end - is->next < 2))
+		return 0;
+	v = get_unaligned_le16(is->next);
+	is->next += 2;
+	return v;
+}
+
+/* Read and return the next 32-bit integer embedded in the bitstream.  */
+static forceinline u32
+bitstream_read_u32(struct input_bitstream *is)
+{
+	u32 v;
+
+	if (unlikely(is->end - is->next < 4))
+		return 0;
+	v = get_unaligned_le32(is->next);
+	is->next += 4;
+	return v;
+}
+
+/* Read into @dst_buffer an array of literal bytes embedded in the bitstream.
+ * Return either a pointer to the byte past the last written, or NULL if the
+ * read overflows the input buffer.
+ */
+static forceinline void *bitstream_read_bytes(struct input_bitstream *is,
+					      void *dst_buffer, size_t count)
+{
+	if ((size_t)(is->end - is->next) < count)
+		return NULL;
+	memcpy(dst_buffer, is->next, count);
+	is->next += count;
+	return (u8 *)dst_buffer + count;
+}
+
+/* Align the input bitstream on a coding-unit boundary.  */
+static forceinline void bitstream_align(struct input_bitstream *is)
+{
+	is->bitsleft = 0;
+	is->bitbuf = 0;
+}
+
+extern int make_huffman_decode_table(u16 decode_table[], const u32 num_syms,
+				     const u32 num_bits, const u8 lens[],
+				     const u32 max_codeword_len,
+				     u16 working_space[]);
+
+
+/* Reads and returns the next Huffman-encoded symbol from a bitstream.  If the
+ * input data is exhausted, the Huffman symbol is decoded as if the missing bits
+ * are all zeroes.
+ */
+static forceinline u32 read_huffsym(struct input_bitstream *istream,
+					 const u16 decode_table[],
+					 u32 table_bits,
+					 u32 max_codeword_len)
+{
+	u32 entry;
+	u32 key_bits;
+
+	bitstream_ensure_bits(istream, max_codeword_len);
+
+	/* Index the decode table by the next table_bits bits of the input.  */
+	key_bits = bitstream_peek_bits(istream, table_bits);
+	entry = decode_table[key_bits];
+	if (entry < 0xC000) {
+		/* Fast case: The decode table directly provided the
+		 * symbol and codeword length.  The low 11 bits are the
+		 * symbol, and the high 5 bits are the codeword length.
+		 */
+		bitstream_remove_bits(istream, entry >> 11);
+		return entry & 0x7FF;
+	}
+	/* Slow case: The codeword for the symbol is longer than
+	 * table_bits, so the symbol does not have an entry
+	 * directly in the first (1 << table_bits) entries of the
+	 * decode table.  Traverse the appropriate binary tree
+	 * bit-by-bit to decode the symbol.
+	 */
+	bitstream_remove_bits(istream, table_bits);
+	do {
+		key_bits = (entry & 0x3FFF) + bitstream_pop_bits(istream, 1);
+	} while ((entry = decode_table[key_bits]) >= 0xC000);
+	return entry;
+}
+
+/*
+ * Copy an LZ77 match at (dst - offset) to dst.
+ *
+ * The length and offset must be already validated --- that is, (dst - offset)
+ * can't underrun the output buffer, and (dst + length) can't overrun the output
+ * buffer.  Also, the length cannot be 0.
+ *
+ * @bufend points to the byte past the end of the output buffer.  This function
+ * won't write any data beyond this position.
+ *
+ * Returns dst + length.
+ */
+static forceinline u8 *lz_copy(u8 *dst, u32 length, u32 offset, const u8 *bufend,
+			       u32 min_length)
+{
+	const u8 *src = dst - offset;
+
+	/*
+	 * Try to copy one machine word at a time.  On i386 and x86_64 this is
+	 * faster than copying one byte at a time, unless the data is
+	 * near-random and all the matches have very short lengths.  Note that
+	 * since this requires unaligned memory accesses, it won't necessarily
+	 * be faster on every architecture.
+	 *
+	 * Also note that we might copy more than the length of the match.  For
+	 * example, if a word is 8 bytes and the match is of length 5, then
+	 * we'll simply copy 8 bytes.  This is okay as long as we don't write
+	 * beyond the end of the output buffer, hence the check for (bufend -
+	 * end >= WORDBYTES - 1).
+	 */
+#ifdef FAST_UNALIGNED_ACCESS
+	u8 * const end = dst + length;
+
+	if (bufend - end >= (ptrdiff_t)(WORDBYTES - 1)) {
+
+		if (offset >= WORDBYTES) {
+			/* The source and destination words don't overlap.  */
+
+			/* To improve branch prediction, one iteration of this
+			 * loop is unrolled.  Most matches are short and will
+			 * fail the first check.  But if that check passes, then
+			 * it becomes increasing likely that the match is long
+			 * and we'll need to continue copying.
+			 */
+
+			copy_unaligned_word(src, dst);
+			src += WORDBYTES;
+			dst += WORDBYTES;
+
+			if (dst < end) {
+				do {
+					copy_unaligned_word(src, dst);
+					src += WORDBYTES;
+					dst += WORDBYTES;
+				} while (dst < end);
+			}
+			return end;
+		} else if (offset == 1) {
+
+			/* Offset 1 matches are equivalent to run-length
+			 * encoding of the previous byte.  This case is common
+			 * if the data contains many repeated bytes.
+			 */
+			size_t v = repeat_byte(*(dst - 1));
+
+			do {
+				put_unaligned(v, (size_t *)dst);
+				src += WORDBYTES;
+				dst += WORDBYTES;
+			} while (dst < end);
+			return end;
+		}
+		/*
+		 * We don't bother with special cases for other 'offset <
+		 * WORDBYTES', which are usually rarer than 'offset == 1'.  Extra
+		 * checks will just slow things down.  Actually, it's possible
+		 * to handle all the 'offset < WORDBYTES' cases using the same
+		 * code, but it still becomes more complicated doesn't seem any
+		 * faster overall; it definitely slows down the more common
+		 * 'offset == 1' case.
+		 */
+	}
+#endif /* FAST_UNALIGNED_ACCESS */
+
+	/* Fall back to a bytewise copy.  */
+
+	if (min_length >= 2) {
+		*dst++ = *src++;
+		length--;
+	}
+	if (min_length >= 3) {
+		*dst++ = *src++;
+		length--;
+	}
+	do {
+		*dst++ = *src++;
+	} while (--length);
+
+	return dst;
+}

target/linux/generic/files/fs/ntfs3/lib/lib.h

@@ -0,0 +1,37 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Adapted for linux kernel by Alexander Mamaev:
+ * - remove implementations of get_unaligned_
+ * - assume GCC is always defined
+ * - ISO C90
+ * - linux kernel code style
+ */
+
+#ifndef __has_attribute
+#define __has_attribute(x) 0
+#endif
+
+#ifndef fallthrough
+#if __has_attribute(__fallthrough__)
+#define fallthrough __attribute__((__fallthrough__))
+#else
+#define fallthrough do {} while (0)  /* fallthrough */
+#endif
+#endif
+
+/* globals from xpress_decompress.c */
+struct xpress_decompressor *xpress_allocate_decompressor(void);
+void xpress_free_decompressor(struct xpress_decompressor *d);
+int xpress_decompress(struct xpress_decompressor *__restrict d,
+		      const void *__restrict compressed_data,
+		      size_t compressed_size,
+		      void *__restrict uncompressed_data,
+		      size_t uncompressed_size);
+
+/* globals from lzx_decompress.c */
+struct lzx_decompressor *lzx_allocate_decompressor(void);
+void lzx_free_decompressor(struct lzx_decompressor *d);
+int lzx_decompress(struct lzx_decompressor *__restrict d,
+		   const void *__restrict compressed_data,
+		   size_t compressed_size, void *__restrict uncompressed_data,
+		   size_t uncompressed_size);

target/linux/generic/files/fs/ntfs3/lib/lzx_decompress.c

@@ -0,0 +1,683 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * lzx_decompress.c - A decompressor for the LZX compression format, which can
+ * be used in "System Compressed" files.  This is based on the code from wimlib.
+ * This code only supports a window size (dictionary size) of 32768 bytes, since
+ * this is the only size used in System Compression.
+ *
+ * Copyright (C) 2015 Eric Biggers
+ *
+ * This program is free software: you can redistribute it and/or modify it under
+ * the terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 2 of the License, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "decompress_common.h"
+#include "lib.h"
+
+/* Number of literal byte values  */
+#define LZX_NUM_CHARS			256
+
+/* The smallest and largest allowed match lengths  */
+#define LZX_MIN_MATCH_LEN		2
+#define LZX_MAX_MATCH_LEN		257
+
+/* Number of distinct match lengths that can be represented  */
+#define LZX_NUM_LENS			(LZX_MAX_MATCH_LEN - LZX_MIN_MATCH_LEN + 1)
+
+/* Number of match lengths for which no length symbol is required  */
+#define LZX_NUM_PRIMARY_LENS		7
+#define LZX_NUM_LEN_HEADERS		(LZX_NUM_PRIMARY_LENS + 1)
+
+/* Valid values of the 3-bit block type field  */
+#define LZX_BLOCKTYPE_VERBATIM		1
+#define LZX_BLOCKTYPE_ALIGNED		2
+#define LZX_BLOCKTYPE_UNCOMPRESSED	3
+
+/* Number of offset slots for a window size of 32768  */
+#define LZX_NUM_OFFSET_SLOTS		30
+
+/* Number of symbols in the main code for a window size of 32768  */
+#define LZX_MAINCODE_NUM_SYMBOLS	\
+	(LZX_NUM_CHARS + (LZX_NUM_OFFSET_SLOTS * LZX_NUM_LEN_HEADERS))
+
+/* Number of symbols in the length code  */
+#define LZX_LENCODE_NUM_SYMBOLS		(LZX_NUM_LENS - LZX_NUM_PRIMARY_LENS)
+
+/* Number of symbols in the precode  */
+#define LZX_PRECODE_NUM_SYMBOLS		20
+
+/* Number of bits in which each precode codeword length is represented  */
+#define LZX_PRECODE_ELEMENT_SIZE	4
+
+/* Number of low-order bits of each match offset that are entropy-encoded in
+ * aligned offset blocks
+ */
+#define LZX_NUM_ALIGNED_OFFSET_BITS	3
+
+/* Number of symbols in the aligned offset code  */
+#define LZX_ALIGNEDCODE_NUM_SYMBOLS	(1 << LZX_NUM_ALIGNED_OFFSET_BITS)
+
+/* Mask for the match offset bits that are entropy-encoded in aligned offset
+ * blocks
+ */
+#define LZX_ALIGNED_OFFSET_BITMASK	((1 << LZX_NUM_ALIGNED_OFFSET_BITS) - 1)
+
+/* Number of bits in which each aligned offset codeword length is represented  */
+#define LZX_ALIGNEDCODE_ELEMENT_SIZE	3
+
+/* Maximum lengths (in bits) of the codewords in each Huffman code  */
+#define LZX_MAX_MAIN_CODEWORD_LEN	16
+#define LZX_MAX_LEN_CODEWORD_LEN	16
+#define LZX_MAX_PRE_CODEWORD_LEN	((1 << LZX_PRECODE_ELEMENT_SIZE) - 1)
+#define LZX_MAX_ALIGNED_CODEWORD_LEN	((1 << LZX_ALIGNEDCODE_ELEMENT_SIZE) - 1)
+
+/* The default "filesize" value used in pre/post-processing.  In the LZX format
+ * used in cabinet files this value must be given to the decompressor, whereas
+ * in the LZX format used in WIM files and system-compressed files this value is
+ * fixed at 12000000.
+ */
+#define LZX_DEFAULT_FILESIZE		12000000
+
+/* Assumed block size when the encoded block size begins with a 0 bit.  */
+#define LZX_DEFAULT_BLOCK_SIZE		32768
+
+/* Number of offsets in the recent (or "repeat") offsets queue.  */
+#define LZX_NUM_RECENT_OFFSETS		3
+
+/* These values are chosen for fast decompression.  */
+#define LZX_MAINCODE_TABLEBITS		11
+#define LZX_LENCODE_TABLEBITS		10
+#define LZX_PRECODE_TABLEBITS		6
+#define LZX_ALIGNEDCODE_TABLEBITS	7
+
+#define LZX_READ_LENS_MAX_OVERRUN	50
+
+/* Mapping: offset slot => first match offset that uses that offset slot.
+ */
+static const u32 lzx_offset_slot_base[LZX_NUM_OFFSET_SLOTS + 1] = {
+	0,	1,	2,	3,	4,	/* 0  --- 4  */
+	6,	8,	12,	16,	24,	/* 5  --- 9  */
+	32,	48,	64,	96,	128,	/* 10 --- 14 */
+	192,	256,	384,	512,	768,	/* 15 --- 19 */
+	1024,	1536,	2048,	3072,	4096,   /* 20 --- 24 */
+	6144,	8192,	12288,	16384,	24576,	/* 25 --- 29 */
+	32768,					/* extra     */
+};
+
+/* Mapping: offset slot => how many extra bits must be read and added to the
+ * corresponding offset slot base to decode the match offset.
+ */
+static const u8 lzx_extra_offset_bits[LZX_NUM_OFFSET_SLOTS] = {
+	0,	0,	0,	0,	1,
+	1,	2,	2,	3,	3,
+	4,	4,	5,	5,	6,
+	6,	7,	7,	8,	8,
+	9,	9,	10,	10,	11,
+	11,	12,	12,	13,	13,
+};
+
+/* Reusable heap-allocated memory for LZX decompression  */
+struct lzx_decompressor {
+
+	/* Huffman decoding tables, and arrays that map symbols to codeword
+	 * lengths
+	 */
+
+	u16 maincode_decode_table[(1 << LZX_MAINCODE_TABLEBITS) +
+					(LZX_MAINCODE_NUM_SYMBOLS * 2)];
+	u8 maincode_lens[LZX_MAINCODE_NUM_SYMBOLS + LZX_READ_LENS_MAX_OVERRUN];
+
+
+	u16 lencode_decode_table[(1 << LZX_LENCODE_TABLEBITS) +
+					(LZX_LENCODE_NUM_SYMBOLS * 2)];
+	u8 lencode_lens[LZX_LENCODE_NUM_SYMBOLS + LZX_READ_LENS_MAX_OVERRUN];
+
+
+	u16 alignedcode_decode_table[(1 << LZX_ALIGNEDCODE_TABLEBITS) +
+					(LZX_ALIGNEDCODE_NUM_SYMBOLS * 2)];
+	u8 alignedcode_lens[LZX_ALIGNEDCODE_NUM_SYMBOLS];
+
+	u16 precode_decode_table[(1 << LZX_PRECODE_TABLEBITS) +
+				 (LZX_PRECODE_NUM_SYMBOLS * 2)];
+	u8 precode_lens[LZX_PRECODE_NUM_SYMBOLS];
+
+	/* Temporary space for make_huffman_decode_table()  */
+	u16 working_space[2 * (1 + LZX_MAX_MAIN_CODEWORD_LEN) +
+			  LZX_MAINCODE_NUM_SYMBOLS];
+};
+
+static void undo_e8_translation(void *target, s32 input_pos)
+{
+	s32 abs_offset, rel_offset;
+
+	abs_offset = get_unaligned_le32(target);
+	if (abs_offset >= 0) {
+		if (abs_offset < LZX_DEFAULT_FILESIZE) {
+			/* "good translation" */
+			rel_offset = abs_offset - input_pos;
+			put_unaligned_le32(rel_offset, target);
+		}
+	} else {
+		if (abs_offset >= -input_pos) {
+			/* "compensating translation" */
+			rel_offset = abs_offset + LZX_DEFAULT_FILESIZE;
+			put_unaligned_le32(rel_offset, target);
+		}
+	}
+}
+
+/*
+ * Undo the 'E8' preprocessing used in LZX.  Before compression, the
+ * uncompressed data was preprocessed by changing the targets of suspected x86
+ * CALL instructions from relative offsets to absolute offsets.  After
+ * match/literal decoding, the decompressor must undo the translation.
+ */
+static void lzx_postprocess(u8 *data, u32 size)
+{
+	/*
+	 * A worthwhile optimization is to push the end-of-buffer check into the
+	 * relatively rare E8 case.  This is possible if we replace the last six
+	 * bytes of data with E8 bytes; then we are guaranteed to hit an E8 byte
+	 * before reaching end-of-buffer.  In addition, this scheme guarantees
+	 * that no translation can begin following an E8 byte in the last 10
+	 * bytes because a 4-byte offset containing E8 as its high byte is a
+	 * large negative number that is not valid for translation.  That is
+	 * exactly what we need.
+	 */
+	u8 *tail;
+	u8 saved_bytes[6];
+	u8 *p;
+
+	if (size <= 10)
+		return;
+
+	tail = &data[size - 6];
+	memcpy(saved_bytes, tail, 6);
+	memset(tail, 0xE8, 6);
+	p = data;
+	for (;;) {
+		while (*p != 0xE8)
+			p++;
+		if (p >= tail)
+			break;
+		undo_e8_translation(p + 1, p - data);
+		p += 5;
+	}
+	memcpy(tail, saved_bytes, 6);
+}
+
+/* Read a Huffman-encoded symbol using the precode.  */
+static forceinline u32 read_presym(const struct lzx_decompressor *d,
+					struct input_bitstream *is)
+{
+	return read_huffsym(is, d->precode_decode_table,
+			    LZX_PRECODE_TABLEBITS, LZX_MAX_PRE_CODEWORD_LEN);
+}
+
+/* Read a Huffman-encoded symbol using the main code.  */
+static forceinline u32 read_mainsym(const struct lzx_decompressor *d,
+					 struct input_bitstream *is)
+{
+	return read_huffsym(is, d->maincode_decode_table,
+			    LZX_MAINCODE_TABLEBITS, LZX_MAX_MAIN_CODEWORD_LEN);
+}
+
+/* Read a Huffman-encoded symbol using the length code.  */
+static forceinline u32 read_lensym(const struct lzx_decompressor *d,
+					struct input_bitstream *is)
+{
+	return read_huffsym(is, d->lencode_decode_table,
+			    LZX_LENCODE_TABLEBITS, LZX_MAX_LEN_CODEWORD_LEN);
+}
+
+/* Read a Huffman-encoded symbol using the aligned offset code.  */
+static forceinline u32 read_alignedsym(const struct lzx_decompressor *d,
+					    struct input_bitstream *is)
+{
+	return read_huffsym(is, d->alignedcode_decode_table,
+			    LZX_ALIGNEDCODE_TABLEBITS,
+			    LZX_MAX_ALIGNED_CODEWORD_LEN);
+}
+
+/*
+ * Read the precode from the compressed input bitstream, then use it to decode
+ * @num_lens codeword length values.
+ *
+ * @is:		The input bitstream.
+ *
+ * @lens:	An array that contains the length values from the previous time
+ *		the codeword lengths for this Huffman code were read, or all 0's
+ *		if this is the first time.  This array must have at least
+ *		(@num_lens + LZX_READ_LENS_MAX_OVERRUN) entries.
+ *
+ * @num_lens:	Number of length values to decode.
+ *
+ * Returns 0 on success, or -1 if the data was invalid.
+ */
+static int lzx_read_codeword_lens(struct lzx_decompressor *d,
+				  struct input_bitstream *is,
+				  u8 *lens, u32 num_lens)
+{
+	u8 *len_ptr = lens;
+	u8 *lens_end = lens + num_lens;
+	int i;
+
+	/* Read the lengths of the precode codewords.  These are given
+	 * explicitly.
+	 */
+	for (i = 0; i < LZX_PRECODE_NUM_SYMBOLS; i++) {
+		d->precode_lens[i] =
+			bitstream_read_bits(is, LZX_PRECODE_ELEMENT_SIZE);
+	}
+
+	/* Make the decoding table for the precode.  */
+	if (make_huffman_decode_table(d->precode_decode_table,
+				      LZX_PRECODE_NUM_SYMBOLS,
+				      LZX_PRECODE_TABLEBITS,
+				      d->precode_lens,
+				      LZX_MAX_PRE_CODEWORD_LEN,
+				      d->working_space))
+		return -1;
+
+	/* Decode the codeword lengths.  */
+	do {
+		u32 presym;
+		u8 len;
+
+		/* Read the next precode symbol.  */
+		presym = read_presym(d, is);
+		if (presym < 17) {
+			/* Difference from old length  */
+			len = *len_ptr - presym;
+			if ((s8)len < 0)
+				len += 17;
+			*len_ptr++ = len;
+		} else {
+			/* Special RLE values  */
+
+			u32 run_len;
+
+			if (presym == 17) {
+				/* Run of 0's  */
+				run_len = 4 + bitstream_read_bits(is, 4);
+				len = 0;
+			} else if (presym == 18) {
+				/* Longer run of 0's  */
+				run_len = 20 + bitstream_read_bits(is, 5);
+				len = 0;
+			} else {
+				/* Run of identical lengths  */
+				run_len = 4 + bitstream_read_bits(is, 1);
+				presym = read_presym(d, is);
+				if (presym > 17)
+					return -1;
+				len = *len_ptr - presym;
+				if ((s8)len < 0)
+					len += 17;
+			}
+
+			do {
+				*len_ptr++ = len;
+			} while (--run_len);
+			/* Worst case overrun is when presym == 18,
+			 * run_len == 20 + 31, and only 1 length was remaining.
+			 * So LZX_READ_LENS_MAX_OVERRUN == 50.
+			 *
+			 * Overrun while reading the first half of maincode_lens
+			 * can corrupt the previous values in the second half.
+			 * This doesn't really matter because the resulting
+			 * lengths will still be in range, and data that
+			 * generates overruns is invalid anyway.
+			 */
+		}
+	} while (len_ptr < lens_end);
+
+	return 0;
+}
+
+/*
+ * Read the header of an LZX block and save the block type and (uncompressed)
+ * size in *block_type_ret and *block_size_ret, respectively.
+ *
+ * If the block is compressed, also update the Huffman decode @tables with the
+ * new Huffman codes.  If the block is uncompressed, also update the match
+ * offset @queue with the new match offsets.
+ *
+ * Return 0 on success, or -1 if the data was invalid.
+ */
+static int lzx_read_block_header(struct lzx_decompressor *d,
+				 struct input_bitstream *is,
+				 int *block_type_ret,
+				 u32 *block_size_ret,
+				 u32 recent_offsets[])
+{
+	int block_type;
+	u32 block_size;
+	int i;
+
+	bitstream_ensure_bits(is, 4);
+
+	/* The first three bits tell us what kind of block it is, and should be
+	 * one of the LZX_BLOCKTYPE_* values.
+	 */
+	block_type = bitstream_pop_bits(is, 3);
+
+	/* Read the block size.  */
+	if (bitstream_pop_bits(is, 1)) {
+		block_size = LZX_DEFAULT_BLOCK_SIZE;
+	} else {
+		block_size = 0;
+		block_size |= bitstream_read_bits(is, 8);
+		block_size <<= 8;
+		block_size |= bitstream_read_bits(is, 8);
+	}
+
+	switch (block_type) {
+
+	case LZX_BLOCKTYPE_ALIGNED:
+
+		/* Read the aligned offset code and prepare its decode table.
+		 */
+
+		for (i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) {
+			d->alignedcode_lens[i] =
+				bitstream_read_bits(is,
+						    LZX_ALIGNEDCODE_ELEMENT_SIZE);
+		}
+
+		if (make_huffman_decode_table(d->alignedcode_decode_table,
+					      LZX_ALIGNEDCODE_NUM_SYMBOLS,
+					      LZX_ALIGNEDCODE_TABLEBITS,
+					      d->alignedcode_lens,
+					      LZX_MAX_ALIGNED_CODEWORD_LEN,
+					      d->working_space))
+			return -1;
+
+		/* Fall though, since the rest of the header for aligned offset
+		 * blocks is the same as that for verbatim blocks.
+		 */
+		fallthrough;
+
+	case LZX_BLOCKTYPE_VERBATIM:
+
+		/* Read the main code and prepare its decode table.
+		 *
+		 * Note that the codeword lengths in the main code are encoded
+		 * in two parts: one part for literal symbols, and one part for
+		 * match symbols.
+		 */
+
+		if (lzx_read_codeword_lens(d, is, d->maincode_lens,
+					   LZX_NUM_CHARS))
+			return -1;
+
+		if (lzx_read_codeword_lens(d, is,
+					   d->maincode_lens + LZX_NUM_CHARS,
+					   LZX_MAINCODE_NUM_SYMBOLS - LZX_NUM_CHARS))
+			return -1;
+
+		if (make_huffman_decode_table(d->maincode_decode_table,
+					      LZX_MAINCODE_NUM_SYMBOLS,
+					      LZX_MAINCODE_TABLEBITS,
+					      d->maincode_lens,
+					      LZX_MAX_MAIN_CODEWORD_LEN,
+					      d->working_space))
+			return -1;
+
+		/* Read the length code and prepare its decode table.  */
+
+		if (lzx_read_codeword_lens(d, is, d->lencode_lens,
+					   LZX_LENCODE_NUM_SYMBOLS))
+			return -1;
+
+		if (make_huffman_decode_table(d->lencode_decode_table,
+					      LZX_LENCODE_NUM_SYMBOLS,
+					      LZX_LENCODE_TABLEBITS,
+					      d->lencode_lens,
+					      LZX_MAX_LEN_CODEWORD_LEN,
+					      d->working_space))
+			return -1;
+
+		break;
+
+	case LZX_BLOCKTYPE_UNCOMPRESSED:
+
+		/* Before reading the three recent offsets from the uncompressed
+		 * block header, the stream must be aligned on a 16-bit
+		 * boundary.  But if the stream is *already* aligned, then the
+		 * next 16 bits must be discarded.
+		 */
+		bitstream_ensure_bits(is, 1);
+		bitstream_align(is);
+
+		recent_offsets[0] = bitstream_read_u32(is);
+		recent_offsets[1] = bitstream_read_u32(is);
+		recent_offsets[2] = bitstream_read_u32(is);
+
+		/* Offsets of 0 are invalid.  */
+		if (recent_offsets[0] == 0 || recent_offsets[1] == 0 ||
+		    recent_offsets[2] == 0)
+			return -1;
+		break;
+
+	default:
+		/* Unrecognized block type.  */
+		return -1;
+	}
+
+	*block_type_ret = block_type;
+	*block_size_ret = block_size;
+	return 0;
+}
+
+/* Decompress a block of LZX-compressed data.  */
+static int lzx_decompress_block(const struct lzx_decompressor *d,
+				struct input_bitstream *is,
+				int block_type, u32 block_size,
+				u8 * const out_begin, u8 *out_next,
+				u32 recent_offsets[])
+{
+	u8 * const block_end = out_next + block_size;
+	u32 ones_if_aligned = 0U - (block_type == LZX_BLOCKTYPE_ALIGNED);
+
+	do {
+		u32 mainsym;
+		u32 match_len;
+		u32 match_offset;
+		u32 offset_slot;
+		u32 num_extra_bits;
+
+		mainsym = read_mainsym(d, is);
+		if (mainsym < LZX_NUM_CHARS) {
+			/* Literal  */
+			*out_next++ = mainsym;
+			continue;
+		}
+
+		/* Match  */
+
+		/* Decode the length header and offset slot.  */
+		mainsym -= LZX_NUM_CHARS;
+		match_len = mainsym % LZX_NUM_LEN_HEADERS;
+		offset_slot = mainsym / LZX_NUM_LEN_HEADERS;
+
+		/* If needed, read a length symbol to decode the full length. */
+		if (match_len == LZX_NUM_PRIMARY_LENS)
+			match_len += read_lensym(d, is);
+		match_len += LZX_MIN_MATCH_LEN;
+
+		if (offset_slot < LZX_NUM_RECENT_OFFSETS) {
+			/* Repeat offset  */
+
+			/* Note: This isn't a real LRU queue, since using the R2
+			 * offset doesn't bump the R1 offset down to R2.  This
+			 * quirk allows all 3 recent offsets to be handled by
+			 * the same code.  (For R0, the swap is a no-op.)
+			 */
+			match_offset = recent_offsets[offset_slot];
+			recent_offsets[offset_slot] = recent_offsets[0];
+			recent_offsets[0] = match_offset;
+		} else {
+			/* Explicit offset  */
+
+			/* Look up the number of extra bits that need to be read
+			 * to decode offsets with this offset slot.
+			 */
+			num_extra_bits = lzx_extra_offset_bits[offset_slot];
+
+			/* Start with the offset slot base value.  */
+			match_offset = lzx_offset_slot_base[offset_slot];
+
+			/* In aligned offset blocks, the low-order 3 bits of
+			 * each offset are encoded using the aligned offset
+			 * code.  Otherwise, all the extra bits are literal.
+			 */
+
+			if ((num_extra_bits & ones_if_aligned) >= LZX_NUM_ALIGNED_OFFSET_BITS) {
+				match_offset +=
+					bitstream_read_bits(is, num_extra_bits -
+								LZX_NUM_ALIGNED_OFFSET_BITS)
+							<< LZX_NUM_ALIGNED_OFFSET_BITS;
+				match_offset += read_alignedsym(d, is);
+			} else {
+				match_offset += bitstream_read_bits(is, num_extra_bits);
+			}
+
+			/* Adjust the offset.  */
+			match_offset -= (LZX_NUM_RECENT_OFFSETS - 1);
+
+			/* Update the recent offsets.  */
+			recent_offsets[2] = recent_offsets[1];
+			recent_offsets[1] = recent_offsets[0];
+			recent_offsets[0] = match_offset;
+		}
+
+		/* Validate the match, then copy it to the current position.  */
+
+		if (match_len > (size_t)(block_end - out_next))
+			return -1;
+
+		if (match_offset > (size_t)(out_next - out_begin))
+			return -1;
+
+		out_next = lz_copy(out_next, match_len, match_offset,
+				   block_end, LZX_MIN_MATCH_LEN);
+
+	} while (out_next != block_end);
+
+	return 0;
+}
+
+/*
+ * lzx_allocate_decompressor - Allocate an LZX decompressor
+ *
+ * Return the pointer to the decompressor on success, or return NULL and set
+ * errno on failure.
+ */
+struct lzx_decompressor *lzx_allocate_decompressor(void)
+{
+	return kmalloc(sizeof(struct lzx_decompressor), GFP_NOFS);
+}
+
+/*
+ * lzx_decompress - Decompress a buffer of LZX-compressed data
+ *
+ * @decompressor:      A decompressor allocated with lzx_allocate_decompressor()
+ * @compressed_data:	The buffer of data to decompress
+ * @compressed_size:	Number of bytes of compressed data
+ * @uncompressed_data:	The buffer in which to store the decompressed data
+ * @uncompressed_size:	The number of bytes the data decompresses into
+ *
+ * Return 0 on success, or return -1 and set errno on failure.
+ */
+int lzx_decompress(struct lzx_decompressor *decompressor,
+		   const void *compressed_data, size_t compressed_size,
+		   void *uncompressed_data, size_t uncompressed_size)
+{
+	struct lzx_decompressor *d = decompressor;
+	u8 * const out_begin = uncompressed_data;
+	u8 *out_next = out_begin;
+	u8 * const out_end = out_begin + uncompressed_size;
+	struct input_bitstream is;
+	u32 recent_offsets[LZX_NUM_RECENT_OFFSETS] = {1, 1, 1};
+	int e8_status = 0;
+
+	init_input_bitstream(&is, compressed_data, compressed_size);
+
+	/* Codeword lengths begin as all 0's for delta encoding purposes.  */
+	memset(d->maincode_lens, 0, LZX_MAINCODE_NUM_SYMBOLS);
+	memset(d->lencode_lens, 0, LZX_LENCODE_NUM_SYMBOLS);
+
+	/* Decompress blocks until we have all the uncompressed data.  */
+
+	while (out_next != out_end) {
+		int block_type;
+		u32 block_size;
+
+		if (lzx_read_block_header(d, &is, &block_type, &block_size,
+					  recent_offsets))
+			goto invalid;
+
+		if (block_size < 1 || block_size > (size_t)(out_end - out_next))
+			goto invalid;
+
+		if (block_type != LZX_BLOCKTYPE_UNCOMPRESSED) {
+
+			/* Compressed block  */
+
+			if (lzx_decompress_block(d,
+						 &is,
+						 block_type,
+						 block_size,
+						 out_begin,
+						 out_next,
+						 recent_offsets))
+				goto invalid;
+
+			e8_status |= d->maincode_lens[0xe8];
+			out_next += block_size;
+		} else {
+			/* Uncompressed block  */
+
+			out_next = bitstream_read_bytes(&is, out_next,
+							block_size);
+			if (!out_next)
+				goto invalid;
+
+			if (block_size & 1)
+				bitstream_read_byte(&is);
+
+			e8_status = 1;
+		}
+	}
+
+	/* Postprocess the data unless it cannot possibly contain 0xe8 bytes. */
+	if (e8_status)
+		lzx_postprocess(uncompressed_data, uncompressed_size);
+
+	return 0;
+
+invalid:
+	return -1;
+}
+
+/*
+ * lzx_free_decompressor - Free an LZX decompressor
+ *
+ * @decompressor:       A decompressor that was allocated with
+ *			lzx_allocate_decompressor(), or NULL.
+ */
+void lzx_free_decompressor(struct lzx_decompressor *decompressor)
+{
+	kfree(decompressor);
+}

target/linux/generic/files/fs/ntfs3/lib/xpress_decompress.c

@@ -0,0 +1,155 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * xpress_decompress.c - A decompressor for the XPRESS compression format
+ * (Huffman variant), which can be used in "System Compressed" files.  This is
+ * based on the code from wimlib.
+ *
+ * Copyright (C) 2015 Eric Biggers
+ *
+ * This program is free software: you can redistribute it and/or modify it under
+ * the terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 2 of the License, or (at your option) any later
+ * version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "decompress_common.h"
+#include "lib.h"
+
+#define XPRESS_NUM_SYMBOLS	512
+#define XPRESS_MAX_CODEWORD_LEN	15
+#define XPRESS_MIN_MATCH_LEN	3
+
+/* This value is chosen for fast decompression.  */
+#define XPRESS_TABLEBITS 12
+
+/* Reusable heap-allocated memory for XPRESS decompression  */
+struct xpress_decompressor {
+
+	/* The Huffman decoding table  */
+	u16 decode_table[(1 << XPRESS_TABLEBITS) + 2 * XPRESS_NUM_SYMBOLS];
+
+	/* An array that maps symbols to codeword lengths  */
+	u8 lens[XPRESS_NUM_SYMBOLS];
+
+	/* Temporary space for make_huffman_decode_table()  */
+	u16 working_space[2 * (1 + XPRESS_MAX_CODEWORD_LEN) +
+			  XPRESS_NUM_SYMBOLS];
+};
+
+/*
+ * xpress_allocate_decompressor - Allocate an XPRESS decompressor
+ *
+ * Return the pointer to the decompressor on success, or return NULL and set
+ * errno on failure.
+ */
+struct xpress_decompressor *xpress_allocate_decompressor(void)
+{
+	return kmalloc(sizeof(struct xpress_decompressor), GFP_NOFS);
+}
+
+/*
+ * xpress_decompress - Decompress a buffer of XPRESS-compressed data
+ *
+ * @decompressor:       A decompressor that was allocated with
+ *			xpress_allocate_decompressor()
+ * @compressed_data:	The buffer of data to decompress
+ * @compressed_size:	Number of bytes of compressed data
+ * @uncompressed_data:	The buffer in which to store the decompressed data
+ * @uncompressed_size:	The number of bytes the data decompresses into
+ *
+ * Return 0 on success, or return -1 and set errno on failure.
+ */
+int xpress_decompress(struct xpress_decompressor *decompressor,
+		      const void *compressed_data, size_t compressed_size,
+		      void *uncompressed_data, size_t uncompressed_size)
+{
+	struct xpress_decompressor *d = decompressor;
+	const u8 * const in_begin = compressed_data;
+	u8 * const out_begin = uncompressed_data;
+	u8 *out_next = out_begin;
+	u8 * const out_end = out_begin + uncompressed_size;
+	struct input_bitstream is;
+	u32 i;
+
+	/* Read the Huffman codeword lengths.  */
+	if (compressed_size < XPRESS_NUM_SYMBOLS / 2)
+		goto invalid;
+	for (i = 0; i < XPRESS_NUM_SYMBOLS / 2; i++) {
+		d->lens[i*2 + 0] = in_begin[i] & 0xF;
+		d->lens[i*2 + 1] = in_begin[i] >> 4;
+	}
+
+	/* Build a decoding table for the Huffman code.  */
+	if (make_huffman_decode_table(d->decode_table, XPRESS_NUM_SYMBOLS,
+				      XPRESS_TABLEBITS, d->lens,
+				      XPRESS_MAX_CODEWORD_LEN,
+				      d->working_space))
+		goto invalid;
+
+	/* Decode the matches and literals.  */
+
+	init_input_bitstream(&is, in_begin + XPRESS_NUM_SYMBOLS / 2,
+			     compressed_size - XPRESS_NUM_SYMBOLS / 2);
+
+	while (out_next != out_end) {
+		u32 sym;
+		u32 log2_offset;
+		u32 length;
+		u32 offset;
+
+		sym = read_huffsym(&is, d->decode_table,
+				   XPRESS_TABLEBITS, XPRESS_MAX_CODEWORD_LEN);
+		if (sym < 256) {
+			/* Literal  */
+			*out_next++ = sym;
+		} else {
+			/* Match  */
+			length = sym & 0xf;
+			log2_offset = (sym >> 4) & 0xf;
+
+			bitstream_ensure_bits(&is, 16);
+
+			offset = ((u32)1 << log2_offset) |
+				 bitstream_pop_bits(&is, log2_offset);
+
+			if (length == 0xf) {
+				length += bitstream_read_byte(&is);
+				if (length == 0xf + 0xff)
+					length = bitstream_read_u16(&is);
+			}
+			length += XPRESS_MIN_MATCH_LEN;
+
+			if (offset > (size_t)(out_next - out_begin))
+				goto invalid;
+
+			if (length > (size_t)(out_end - out_next))
+				goto invalid;
+
+			out_next = lz_copy(out_next, length, offset, out_end,
+					   XPRESS_MIN_MATCH_LEN);
+		}
+	}
+	return 0;
+
+invalid:
+	return -1;
+}
+
+/*
+ * xpress_free_decompressor - Free an XPRESS decompressor
+ *
+ * @decompressor:       A decompressor that was allocated with
+ *			xpress_allocate_decompressor(), or NULL.
+ */
+void xpress_free_decompressor(struct xpress_decompressor *decompressor)
+{
+	kfree(decompressor);
+}

target/linux/generic/files/fs/ntfs3/lznt.c

@@ -0,0 +1,452 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
+ *
+ */
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+// clang-format off
+/* src buffer is zero */
+#define LZNT_ERROR_ALL_ZEROS	1
+#define LZNT_CHUNK_SIZE		0x1000
+// clang-format on
+
+struct lznt_hash {
+	const u8 *p1;
+	const u8 *p2;
+};
+
+struct lznt {
+	const u8 *unc;
+	const u8 *unc_end;
+	const u8 *best_match;
+	size_t max_len;
+	bool std;
+
+	struct lznt_hash hash[LZNT_CHUNK_SIZE];
+};
+
+static inline size_t get_match_len(const u8 *ptr, const u8 *end, const u8 *prev,
+				   size_t max_len)
+{
+	size_t len = 0;
+
+	while (ptr + len < end && ptr[len] == prev[len] && ++len < max_len)
+		;
+	return len;
+}
+
+static size_t longest_match_std(const u8 *src, struct lznt *ctx)
+{
+	size_t hash_index;
+	size_t len1 = 0, len2 = 0;
+	const u8 **hash;
+
+	hash_index =
+		((40543U * ((((src[0] << 4) ^ src[1]) << 4) ^ src[2])) >> 4) &
+		(LZNT_CHUNK_SIZE - 1);
+
+	hash = &(ctx->hash[hash_index].p1);
+
+	if (hash[0] >= ctx->unc && hash[0] < src && hash[0][0] == src[0] &&
+	    hash[0][1] == src[1] && hash[0][2] == src[2]) {
+		len1 = 3;
+		if (ctx->max_len > 3)
+			len1 += get_match_len(src + 3, ctx->unc_end,
+					      hash[0] + 3, ctx->max_len - 3);
+	}
+
+	if (hash[1] >= ctx->unc && hash[1] < src && hash[1][0] == src[0] &&
+	    hash[1][1] == src[1] && hash[1][2] == src[2]) {
+		len2 = 3;
+		if (ctx->max_len > 3)
+			len2 += get_match_len(src + 3, ctx->unc_end,
+					      hash[1] + 3, ctx->max_len - 3);
+	}
+
+	/* Compare two matches and select the best one */
+	if (len1 < len2) {
+		ctx->best_match = hash[1];
+		len1 = len2;
+	} else {
+		ctx->best_match = hash[0];
+	}
+
+	hash[1] = hash[0];
+	hash[0] = src;
+	return len1;
+}
+
+static size_t longest_match_best(const u8 *src, struct lznt *ctx)
+{
+	size_t max_len;
+	const u8 *ptr;
+
+	if (ctx->unc >= src || !ctx->max_len)
+		return 0;
+
+	max_len = 0;
+	for (ptr = ctx->unc; ptr < src; ++ptr) {
+		size_t len =
+			get_match_len(src, ctx->unc_end, ptr, ctx->max_len);
+		if (len >= max_len) {
+			max_len = len;
+			ctx->best_match = ptr;
+		}
+	}
+
+	return max_len >= 3 ? max_len : 0;
+}
+
+static const size_t s_max_len[] = {
+	0x1002, 0x802, 0x402, 0x202, 0x102, 0x82, 0x42, 0x22, 0x12,
+};
+
+static const size_t s_max_off[] = {
+	0x10, 0x20, 0x40, 0x80, 0x100, 0x200, 0x400, 0x800, 0x1000,
+};
+
+static inline u16 make_pair(size_t offset, size_t len, size_t index)
+{
+	return ((offset - 1) << (12 - index)) |
+	       ((len - 3) & (((1 << (12 - index)) - 1)));
+}
+
+static inline size_t parse_pair(u16 pair, size_t *offset, size_t index)
+{
+	*offset = 1 + (pair >> (12 - index));
+	return 3 + (pair & ((1 << (12 - index)) - 1));
+}
+
+/*
+ * compress_chunk
+ *
+ * returns one of the three values:
+ * 0 - ok, 'cmpr' contains 'cmpr_chunk_size' bytes of compressed data
+ * 1 - input buffer is full zero
+ * -2 - the compressed buffer is too small to hold the compressed data
+ */
+static inline int compress_chunk(size_t (*match)(const u8 *, struct lznt *),
+				 const u8 *unc, const u8 *unc_end, u8 *cmpr,
+				 u8 *cmpr_end, size_t *cmpr_chunk_size,
+				 struct lznt *ctx)
+{
+	size_t cnt = 0;
+	size_t idx = 0;
+	const u8 *up = unc;
+	u8 *cp = cmpr + 3;
+	u8 *cp2 = cmpr + 2;
+	u8 not_zero = 0;
+	/* Control byte of 8-bit values: ( 0 - means byte as is, 1 - short pair ) */
+	u8 ohdr = 0;
+	u8 *last;
+	u16 t16;
+
+	if (unc + LZNT_CHUNK_SIZE < unc_end)
+		unc_end = unc + LZNT_CHUNK_SIZE;
+
+	last = min(cmpr + LZNT_CHUNK_SIZE + sizeof(short), cmpr_end);
+
+	ctx->unc = unc;
+	ctx->unc_end = unc_end;
+	ctx->max_len = s_max_len[0];
+
+	while (up < unc_end) {
+		size_t max_len;
+
+		while (unc + s_max_off[idx] < up)
+			ctx->max_len = s_max_len[++idx];
+
+		// Find match
+		max_len = up + 3 <= unc_end ? (*match)(up, ctx) : 0;
+
+		if (!max_len) {
+			if (cp >= last)
+				goto NotCompressed;
+			not_zero |= *cp++ = *up++;
+		} else if (cp + 1 >= last) {
+			goto NotCompressed;
+		} else {
+			t16 = make_pair(up - ctx->best_match, max_len, idx);
+			*cp++ = t16;
+			*cp++ = t16 >> 8;
+
+			ohdr |= 1 << cnt;
+			up += max_len;
+		}
+
+		cnt = (cnt + 1) & 7;
+		if (!cnt) {
+			*cp2 = ohdr;
+			ohdr = 0;
+			cp2 = cp;
+			cp += 1;
+		}
+	}
+
+	if (cp2 < last)
+		*cp2 = ohdr;
+	else
+		cp -= 1;
+
+	*cmpr_chunk_size = cp - cmpr;
+
+	t16 = (*cmpr_chunk_size - 3) | 0xB000;
+	cmpr[0] = t16;
+	cmpr[1] = t16 >> 8;
+
+	return not_zero ? 0 : LZNT_ERROR_ALL_ZEROS;
+
+NotCompressed:
+
+	if ((cmpr + LZNT_CHUNK_SIZE + sizeof(short)) > last)
+		return -2;
+
+	/*
+	 * Copy non cmpr data
+	 * 0x3FFF == ((LZNT_CHUNK_SIZE + 2 - 3) | 0x3000)
+	 */
+	cmpr[0] = 0xff;
+	cmpr[1] = 0x3f;
+
+	memcpy(cmpr + sizeof(short), unc, LZNT_CHUNK_SIZE);
+	*cmpr_chunk_size = LZNT_CHUNK_SIZE + sizeof(short);
+
+	return 0;
+}
+
+static inline ssize_t decompress_chunk(u8 *unc, u8 *unc_end, const u8 *cmpr,
+				       const u8 *cmpr_end)
+{
+	u8 *up = unc;
+	u8 ch = *cmpr++;
+	size_t bit = 0;
+	size_t index = 0;
+	u16 pair;
+	size_t offset, length;
+
+	/* Do decompression until pointers are inside range */
+	while (up < unc_end && cmpr < cmpr_end) {
+		/* Correct index */
+		while (unc + s_max_off[index] < up)
+			index += 1;
+
+		/* Check the current flag for zero */
+		if (!(ch & (1 << bit))) {
+			/* Just copy byte */
+			*up++ = *cmpr++;
+			goto next;
+		}
+
+		/* Check for boundary */
+		if (cmpr + 1 >= cmpr_end)
+			return -EINVAL;
+
+		/* Read a short from little endian stream */
+		pair = cmpr[1];
+		pair <<= 8;
+		pair |= cmpr[0];
+
+		cmpr += 2;
+
+		/* Translate packed information into offset and length */
+		length = parse_pair(pair, &offset, index);
+
+		/* Check offset for boundary */
+		if (unc + offset > up)
+			return -EINVAL;
+
+		/* Truncate the length if necessary */
+		if (up + length >= unc_end)
+			length = unc_end - up;
+
+		/* Now we copy bytes. This is the heart of LZ algorithm. */
+		for (; length > 0; length--, up++)
+			*up = *(up - offset);
+
+next:
+		/* Advance flag bit value */
+		bit = (bit + 1) & 7;
+
+		if (!bit) {
+			if (cmpr >= cmpr_end)
+				break;
+
+			ch = *cmpr++;
+		}
+	}
+
+	/* return the size of uncompressed data */
+	return up - unc;
+}
+
+/*
+ * 0 - standard compression
+ * !0 - best compression, requires a lot of cpu
+ */
+struct lznt *get_lznt_ctx(int level)
+{
+	struct lznt *r = ntfs_zalloc(level ? offsetof(struct lznt, hash)
+					   : sizeof(struct lznt));
+
+	if (r)
+		r->std = !level;
+	return r;
+}
+
+/*
+ * compress_lznt
+ *
+ * Compresses "unc" into "cmpr"
+ * +x - ok, 'cmpr' contains 'final_compressed_size' bytes of compressed data
+ * 0 - input buffer is full zero
+ */
+size_t compress_lznt(const void *unc, size_t unc_size, void *cmpr,
+		     size_t cmpr_size, struct lznt *ctx)
+{
+	int err;
+	size_t (*match)(const u8 *src, struct lznt *ctx);
+	u8 *p = cmpr;
+	u8 *end = p + cmpr_size;
+	const u8 *unc_chunk = unc;
+	const u8 *unc_end = unc_chunk + unc_size;
+	bool is_zero = true;
+
+	if (ctx->std) {
+		match = &longest_match_std;
+		memset(ctx->hash, 0, sizeof(ctx->hash));
+	} else {
+		match = &longest_match_best;
+	}
+
+	/* compression cycle */
+	for (; unc_chunk < unc_end; unc_chunk += LZNT_CHUNK_SIZE) {
+		cmpr_size = 0;
+		err = compress_chunk(match, unc_chunk, unc_end, p, end,
+				     &cmpr_size, ctx);
+		if (err < 0)
+			return unc_size;
+
+		if (is_zero && err != LZNT_ERROR_ALL_ZEROS)
+			is_zero = false;
+
+		p += cmpr_size;
+	}
+
+	if (p <= end - 2)
+		p[0] = p[1] = 0;
+
+	return is_zero ? 0 : PtrOffset(cmpr, p);
+}
+
+/*
+ * decompress_lznt
+ *
+ * decompresses "cmpr" into "unc"
+ */
+ssize_t decompress_lznt(const void *cmpr, size_t cmpr_size, void *unc,
+			size_t unc_size)
+{
+	const u8 *cmpr_chunk = cmpr;
+	const u8 *cmpr_end = cmpr_chunk + cmpr_size;
+	u8 *unc_chunk = unc;
+	u8 *unc_end = unc_chunk + unc_size;
+	u16 chunk_hdr;
+
+	if (cmpr_size < sizeof(short))
+		return -EINVAL;
+
+	/* read chunk header */
+	chunk_hdr = cmpr_chunk[1];
+	chunk_hdr <<= 8;
+	chunk_hdr |= cmpr_chunk[0];
+
+	/* loop through decompressing chunks */
+	for (;;) {
+		size_t chunk_size_saved;
+		size_t unc_use;
+		size_t cmpr_use = 3 + (chunk_hdr & (LZNT_CHUNK_SIZE - 1));
+
+		/* Check that the chunk actually fits the supplied buffer */
+		if (cmpr_chunk + cmpr_use > cmpr_end)
+			return -EINVAL;
+
+		/* First make sure the chunk contains compressed data */
+		if (chunk_hdr & 0x8000) {
+			/* Decompress a chunk and return if we get an error */
+			ssize_t err =
+				decompress_chunk(unc_chunk, unc_end,
+						 cmpr_chunk + sizeof(chunk_hdr),
+						 cmpr_chunk + cmpr_use);
+			if (err < 0)
+				return err;
+			unc_use = err;
+		} else {
+			/* This chunk does not contain compressed data */
+			unc_use = unc_chunk + LZNT_CHUNK_SIZE > unc_end
+					  ? unc_end - unc_chunk
+					  : LZNT_CHUNK_SIZE;
+
+			if (cmpr_chunk + sizeof(chunk_hdr) + unc_use >
+			    cmpr_end) {
+				return -EINVAL;
+			}
+
+			memcpy(unc_chunk, cmpr_chunk + sizeof(chunk_hdr),
+			       unc_use);
+		}
+
+		/* Advance pointers */
+		cmpr_chunk += cmpr_use;
+		unc_chunk += unc_use;
+
+		/* Check for the end of unc buffer */
+		if (unc_chunk >= unc_end)
+			break;
+
+		/* Proceed the next chunk */
+		if (cmpr_chunk > cmpr_end - 2)
+			break;
+
+		chunk_size_saved = LZNT_CHUNK_SIZE;
+
+		/* read chunk header */
+		chunk_hdr = cmpr_chunk[1];
+		chunk_hdr <<= 8;
+		chunk_hdr |= cmpr_chunk[0];
+
+		if (!chunk_hdr)
+			break;
+
+		/* Check the size of unc buffer */
+		if (unc_use < chunk_size_saved) {
+			size_t t1 = chunk_size_saved - unc_use;
+			u8 *t2 = unc_chunk + t1;
+
+			/* 'Zero' memory */
+			if (t2 >= unc_end)
+				break;
+
+			memset(unc_chunk, 0, t1);
+			unc_chunk = t2;
+		}
+	}
+
+	/* Check compression boundary */
+	if (cmpr_chunk > cmpr_end)
+		return -EINVAL;
+
+	/*
+	 * The unc size is just a difference between current
+	 * pointer and original one
+	 */
+	return PtrOffset(unc, unc_chunk);
+}

target/linux/generic/files/fs/ntfs3/namei.c

@@ -0,0 +1,626 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/namei.h>
+#include <linux/nls.h>
+#include <linux/version.h>
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 16, 0)
+#include <linux/iversion.h>
+#endif
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+/*
+ * fill_name_de
+ *
+ * formats NTFS_DE in 'buf'
+ */
+int fill_name_de(struct ntfs_sb_info *sbi, void *buf, const struct qstr *name,
+		 const struct cpu_str *uni)
+{
+	int err;
+	struct NTFS_DE *e = buf;
+	u16 data_size;
+	struct ATTR_FILE_NAME *fname = (struct ATTR_FILE_NAME *)(e + 1);
+
+#ifndef CONFIG_NTFS3_64BIT_CLUSTER
+	e->ref.high = fname->home.high = 0;
+#endif
+	if (uni) {
+#ifdef __BIG_ENDIAN
+		int ulen = uni->len;
+		__le16 *uname = fname->name;
+		const u16 *name_cpu = uni->name;
+
+		while (ulen--)
+			*uname++ = cpu_to_le16(*name_cpu++);
+#else
+		memcpy(fname->name, uni->name, uni->len * sizeof(u16));
+#endif
+		fname->name_len = uni->len;
+
+	} else {
+		/* Convert input string to unicode */
+		err = ntfs_nls_to_utf16(sbi, name->name, name->len,
+					(struct cpu_str *)&fname->name_len,
+					NTFS_NAME_LEN, UTF16_LITTLE_ENDIAN);
+		if (err < 0)
+			return err;
+	}
+
+	fname->type = FILE_NAME_POSIX;
+	data_size = fname_full_size(fname);
+
+	e->size = cpu_to_le16(QuadAlign(data_size) + sizeof(struct NTFS_DE));
+	e->key_size = cpu_to_le16(data_size);
+	e->flags = 0;
+	e->res = 0;
+
+	return 0;
+}
+
+/*
+ * ntfs_lookup
+ *
+ * inode_operations::lookup
+ */
+static struct dentry *ntfs_lookup(struct inode *dir, struct dentry *dentry,
+				  u32 flags)
+{
+	struct ntfs_inode *ni = ntfs_i(dir);
+	struct cpu_str *uni = __getname();
+	struct inode *inode;
+	int err;
+
+	if (!uni)
+		inode = ERR_PTR(-ENOMEM);
+	else {
+		err = ntfs_nls_to_utf16(ni->mi.sbi, dentry->d_name.name,
+					dentry->d_name.len, uni, NTFS_NAME_LEN,
+					UTF16_HOST_ENDIAN);
+		if (err < 0)
+			inode = ERR_PTR(err);
+		else {
+			ni_lock(ni);
+			inode = dir_search_u(dir, uni, NULL);
+			ni_unlock(ni);
+		}
+		__putname(uni);
+	}
+
+	return d_splice_alias(inode, dentry);
+}
+
+/*
+ * ntfs_create
+ *
+ * inode_operations::create
+ */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 12, 0)
+static int ntfs_create(struct user_namespace *mnt_userns, struct inode *dir,
+#else
+static int ntfs_create(struct inode *dir,
+#endif
+		       struct dentry *dentry, umode_t mode, bool excl)
+{
+	struct ntfs_inode *ni = ntfs_i(dir);
+	struct inode *inode;
+
+	ni_lock_dir(ni);
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 12, 0)
+	inode = ntfs_create_inode(mnt_userns, dir, dentry, NULL, S_IFREG | mode,
+#else
+	inode = ntfs_create_inode(dir, dentry, NULL, S_IFREG | mode,
+#endif
+				  0, NULL, 0, excl, NULL);
+
+	ni_unlock(ni);
+
+	return IS_ERR(inode) ? PTR_ERR(inode) : 0;
+}
+
+/*
+ * ntfs_link
+ *
+ * inode_operations::link
+ */
+static int ntfs_link(struct dentry *ode, struct inode *dir, struct dentry *de)
+{
+	int err;
+	struct inode *inode = d_inode(ode);
+	struct ntfs_inode *ni = ntfs_i(inode);
+
+	if (S_ISDIR(inode->i_mode))
+		return -EPERM;
+
+	if (inode->i_nlink >= NTFS_LINK_MAX)
+		return -EMLINK;
+
+	ni_lock_dir(ntfs_i(dir));
+	if (inode != dir)
+		ni_lock(ni);
+
+	dir->i_ctime = dir->i_mtime = inode->i_ctime = current_time(inode);
+	inc_nlink(inode);
+	ihold(inode);
+
+	err = ntfs_link_inode(inode, de);
+	if (!err) {
+		mark_inode_dirty(inode);
+		mark_inode_dirty(dir);
+		d_instantiate(de, inode);
+	} else {
+		drop_nlink(inode);
+		iput(inode);
+	}
+
+	if (inode != dir)
+		ni_unlock(ni);
+	ni_unlock(ntfs_i(dir));
+
+	return err;
+}
+
+/*
+ * ntfs_unlink
+ *
+ * inode_operations::unlink
+ */
+static int ntfs_unlink(struct inode *dir, struct dentry *dentry)
+{
+	struct ntfs_inode *ni = ntfs_i(dir);
+	int err;
+
+	ni_lock_dir(ni);
+
+	err = ntfs_unlink_inode(dir, dentry);
+
+	ni_unlock(ni);
+
+	return err;
+}
+
+/*
+ * ntfs_symlink
+ *
+ * inode_operations::symlink
+ */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 12, 0)
+static int ntfs_symlink(struct user_namespace *mnt_userns, struct inode *dir,
+#else
+static int ntfs_symlink(struct inode *dir,
+#endif
+			struct dentry *dentry, const char *symname)
+{
+	u32 size = strlen(symname);
+	struct inode *inode;
+	struct ntfs_inode *ni = ntfs_i(dir);
+
+	ni_lock_dir(ni);
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 12, 0)
+	inode = ntfs_create_inode(mnt_userns, dir, dentry, NULL, S_IFLNK | 0777,
+#else
+	inode = ntfs_create_inode(dir, dentry, NULL, S_IFLNK | 0777,
+#endif
+				  0, symname, size, 0, NULL);
+
+	ni_unlock(ni);
+
+	return IS_ERR(inode) ? PTR_ERR(inode) : 0;
+}
+
+/*
+ * ntfs_mkdir
+ *
+ * inode_operations::mkdir
+ */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 12, 0)
+static int ntfs_mkdir(struct user_namespace *mnt_userns, struct inode *dir,
+#else
+static int ntfs_mkdir(struct inode *dir,
+#endif
+		      struct dentry *dentry, umode_t mode)
+{
+	struct inode *inode;
+	struct ntfs_inode *ni = ntfs_i(dir);
+
+	ni_lock_dir(ni);
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 12, 0)
+	inode = ntfs_create_inode(mnt_userns, dir, dentry, NULL, S_IFDIR | mode,
+#else
+	inode = ntfs_create_inode(dir, dentry, NULL, S_IFDIR | mode,
+#endif
+				  0, NULL, -1, 0, NULL);
+
+	ni_unlock(ni);
+
+	return IS_ERR(inode) ? PTR_ERR(inode) : 0;
+}
+
+/*
+ * ntfs_rmdir
+ *
+ * inode_operations::rm_dir
+ */
+static int ntfs_rmdir(struct inode *dir, struct dentry *dentry)
+{
+	struct ntfs_inode *ni = ntfs_i(dir);
+	int err;
+
+	ni_lock_dir(ni);
+
+	err = ntfs_unlink_inode(dir, dentry);
+
+	ni_unlock(ni);
+
+	return err;
+}
+
+/*
+ * ntfs_rename
+ *
+ * inode_operations::rename
+ */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 12, 0)
+static int ntfs_rename(struct user_namespace *mnt_userns, struct inode *old_dir,
+#else
+static int ntfs_rename(struct inode *old_dir,
+#endif
+		       struct dentry *old_dentry, struct inode *new_dir,
+		       struct dentry *new_dentry, u32 flags)
+{
+	int err;
+	struct super_block *sb = old_dir->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct ntfs_inode *old_dir_ni = ntfs_i(old_dir);
+	struct ntfs_inode *new_dir_ni = ntfs_i(new_dir);
+	struct ntfs_inode *old_ni;
+	struct ATTR_FILE_NAME *old_name, *new_name, *fname;
+	u8 name_type;
+	bool is_same;
+	struct inode *old_inode, *new_inode;
+	struct NTFS_DE *old_de, *new_de;
+	struct ATTRIB *attr;
+	struct ATTR_LIST_ENTRY *le;
+	u16 new_de_key_size;
+
+	static_assert(SIZEOF_ATTRIBUTE_FILENAME_MAX + SIZEOF_RESIDENT < 1024);
+	static_assert(SIZEOF_ATTRIBUTE_FILENAME_MAX + sizeof(struct NTFS_DE) <
+		      1024);
+	static_assert(PATH_MAX >= 4 * 1024);
+
+	if (flags & ~RENAME_NOREPLACE)
+		return -EINVAL;
+
+	old_inode = d_inode(old_dentry);
+	new_inode = d_inode(new_dentry);
+
+	old_ni = ntfs_i(old_inode);
+
+	is_same = old_dentry->d_name.len == new_dentry->d_name.len &&
+		  !memcmp(old_dentry->d_name.name, new_dentry->d_name.name,
+			  old_dentry->d_name.len);
+
+	if (is_same && old_dir == new_dir) {
+		/* Nothing to do */
+		err = 0;
+		goto out;
+	}
+
+	if (ntfs_is_meta_file(sbi, old_inode->i_ino)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (new_inode) {
+		/*target name exists. unlink it*/
+		dget(new_dentry);
+		ni_lock_dir(new_dir_ni);
+		err = ntfs_unlink_inode(new_dir, new_dentry);
+		ni_unlock(new_dir_ni);
+		dput(new_dentry);
+		if (err)
+			goto out;
+	}
+
+	/* allocate PATH_MAX bytes */
+	old_de = __getname();
+	if (!old_de) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	err = fill_name_de(sbi, old_de, &old_dentry->d_name, NULL);
+	if (err < 0)
+		goto out1;
+
+	old_name = (struct ATTR_FILE_NAME *)(old_de + 1);
+
+	if (is_same) {
+		new_de = old_de;
+	} else {
+		new_de = Add2Ptr(old_de, 1024);
+		err = fill_name_de(sbi, new_de, &new_dentry->d_name, NULL);
+		if (err < 0)
+			goto out1;
+	}
+
+	ni_lock_dir(old_dir_ni);
+	ni_lock(old_ni);
+
+	mi_get_ref(&old_dir_ni->mi, &old_name->home);
+
+	/*get pointer to file_name in mft*/
+	fname = ni_fname_name(old_ni, (struct cpu_str *)&old_name->name_len,
+			      &old_name->home, &le);
+	if (!fname) {
+		err = -EINVAL;
+		goto out2;
+	}
+
+	/* Copy fname info from record into new fname */
+	new_name = (struct ATTR_FILE_NAME *)(new_de + 1);
+	memcpy(&new_name->dup, &fname->dup, sizeof(fname->dup));
+
+	name_type = paired_name(fname->type);
+
+	/* remove first name from directory */
+	err = indx_delete_entry(&old_dir_ni->dir, old_dir_ni, old_de + 1,
+				le16_to_cpu(old_de->key_size), sbi);
+	if (err)
+		goto out3;
+
+	/* remove first name from mft */
+	err = ni_remove_attr_le(old_ni, attr_from_name(fname), le);
+	if (err)
+		goto out4;
+
+	le16_add_cpu(&old_ni->mi.mrec->hard_links, -1);
+	old_ni->mi.dirty = true;
+
+	if (name_type != FILE_NAME_POSIX) {
+		/* get paired name */
+		fname = ni_fname_type(old_ni, name_type, &le);
+		if (fname) {
+			/* remove second name from directory */
+			err = indx_delete_entry(&old_dir_ni->dir, old_dir_ni,
+						fname, fname_full_size(fname),
+						sbi);
+			if (err)
+				goto out5;
+
+			/* remove second name from mft */
+			err = ni_remove_attr_le(old_ni, attr_from_name(fname),
+						le);
+			if (err)
+				goto out6;
+
+			le16_add_cpu(&old_ni->mi.mrec->hard_links, -1);
+			old_ni->mi.dirty = true;
+		}
+	}
+
+	/* Add new name */
+	mi_get_ref(&old_ni->mi, &new_de->ref);
+	mi_get_ref(&ntfs_i(new_dir)->mi, &new_name->home);
+
+	new_de_key_size = le16_to_cpu(new_de->key_size);
+
+	/* insert new name in mft */
+	err = ni_insert_resident(old_ni, new_de_key_size, ATTR_NAME, NULL, 0,
+				 &attr, NULL);
+	if (err)
+		goto out7;
+
+	attr->res.flags = RESIDENT_FLAG_INDEXED;
+
+	memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), new_name, new_de_key_size);
+
+	le16_add_cpu(&old_ni->mi.mrec->hard_links, 1);
+	old_ni->mi.dirty = true;
+
+	/* insert new name in directory */
+	err = indx_insert_entry(&new_dir_ni->dir, new_dir_ni, new_de, sbi,
+				NULL);
+	if (err)
+		goto out8;
+
+	if (IS_DIRSYNC(new_dir))
+		err = ntfs_sync_inode(old_inode);
+	else
+		mark_inode_dirty(old_inode);
+
+	old_dir->i_ctime = old_dir->i_mtime = current_time(old_dir);
+	if (IS_DIRSYNC(old_dir))
+		(void)ntfs_sync_inode(old_dir);
+	else
+		mark_inode_dirty(old_dir);
+
+	if (old_dir != new_dir) {
+		new_dir->i_mtime = new_dir->i_ctime = old_dir->i_ctime;
+		mark_inode_dirty(new_dir);
+	}
+
+	if (old_inode) {
+		old_inode->i_ctime = old_dir->i_ctime;
+		mark_inode_dirty(old_inode);
+	}
+
+	err = 0;
+	/* normal way */
+	goto out2;
+
+out8:
+	/* undo
+	 * ni_insert_resident(old_ni, new_de_key_size, ATTR_NAME, NULL, 0,
+	 *			 &attr, NULL);
+	 */
+	mi_remove_attr(&old_ni->mi, attr);
+out7:
+	/* undo
+	 * ni_remove_attr_le(old_ni, attr_from_name(fname), le);
+	 */
+out6:
+	/* undo
+	 * indx_delete_entry(&old_dir_ni->dir, old_dir_ni,
+	 *					fname, fname_full_size(fname),
+	 *					sbi);
+	 */
+out5:
+	/* undo
+	 * ni_remove_attr_le(old_ni, attr_from_name(fname), le);
+	 */
+out4:
+	/* undo:
+	 * indx_delete_entry(&old_dir_ni->dir, old_dir_ni, old_de + 1,
+	 *			old_de->key_size, NULL);
+	 */
+out3:
+out2:
+	ni_unlock(old_ni);
+	ni_unlock(old_dir_ni);
+out1:
+	__putname(old_de);
+out:
+	return err;
+}
+
+/*
+ * ntfs_atomic_open
+ *
+ * inode_operations::atomic_open
+ */
+static int ntfs_atomic_open(struct inode *dir, struct dentry *dentry,
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 19, 0)
+			    struct file *file, u32 flags, umode_t mode)
+#else
+			    struct file *file, u32 flags, umode_t mode, int *opened)
+#endif
+{
+	int err;
+	bool excl = !!(flags & O_EXCL);
+	struct inode *inode;
+	struct ntfs_fnd *fnd = NULL;
+	struct ntfs_inode *ni = ntfs_i(dir);
+	struct dentry *d = NULL;
+	struct cpu_str *uni = __getname();
+
+	if (!uni)
+		return -ENOMEM;
+
+	err = ntfs_nls_to_utf16(ni->mi.sbi, dentry->d_name.name,
+				dentry->d_name.len, uni, NTFS_NAME_LEN,
+				UTF16_HOST_ENDIAN);
+	if (err < 0)
+		goto out;
+
+	ni_lock_dir(ni);
+
+	if (d_in_lookup(dentry)) {
+		fnd = fnd_get();
+		if (!fnd) {
+			err = -ENOMEM;
+			goto out1;
+		}
+
+		d = d_splice_alias(dir_search_u(dir, uni, fnd), dentry);
+		if (IS_ERR(d)) {
+			err = PTR_ERR(d);
+			d = NULL;
+			goto out2;
+		}
+
+		if (d)
+			dentry = d;
+	}
+
+	if (!(flags & O_CREAT) || d_really_is_positive(dentry)) {
+		err = finish_no_open(file, d);
+		goto out2;
+	}
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 19, 0)
+	file->f_mode |= FMODE_CREATED;
+#else
+	*opened |= FILE_CREATED;
+#endif
+
+	/*fnd contains tree's path to insert to*/
+	/* TODO: init_user_ns? */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 12, 0)
+	inode = ntfs_create_inode(&init_user_ns, dir, dentry, uni, mode, 0,
+#else
+	inode = ntfs_create_inode(dir, dentry, uni, mode, 0,
+#endif
+				  NULL, 0, excl, fnd);
+	err = IS_ERR(inode) ? PTR_ERR(inode)
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 19, 0)
+			    : finish_open(file, dentry, ntfs_file_open);
+#else
+			    : finish_open(file, dentry, ntfs_file_open, opened);
+#endif
+	dput(d);
+
+out2:
+	fnd_put(fnd);
+out1:
+	ni_unlock(ni);
+out:
+	__putname(uni);
+
+	return err;
+}
+
+struct dentry *ntfs3_get_parent(struct dentry *child)
+{
+	struct inode *inode = d_inode(child);
+	struct ntfs_inode *ni = ntfs_i(inode);
+
+	struct ATTR_LIST_ENTRY *le = NULL;
+	struct ATTRIB *attr = NULL;
+	struct ATTR_FILE_NAME *fname;
+
+	while ((attr = ni_find_attr(ni, attr, &le, ATTR_NAME, NULL, 0, NULL,
+				    NULL))) {
+		fname = resident_data_ex(attr, SIZEOF_ATTRIBUTE_FILENAME);
+		if (!fname)
+			continue;
+
+		return d_obtain_alias(
+			ntfs_iget5(inode->i_sb, &fname->home, NULL));
+	}
+
+	return ERR_PTR(-ENOENT);
+}
+
+const struct inode_operations ntfs_dir_inode_operations = {
+	.lookup = ntfs_lookup,
+	.create = ntfs_create,
+	.link = ntfs_link,
+	.unlink = ntfs_unlink,
+	.symlink = ntfs_symlink,
+	.mkdir = ntfs_mkdir,
+	.rmdir = ntfs_rmdir,
+	.rename = ntfs_rename,
+	.permission = ntfs_permission,
+	.get_acl = ntfs_get_acl,
+	.set_acl = ntfs_set_acl,
+	.setattr = ntfs3_setattr,
+	.getattr = ntfs_getattr,
+	.listxattr = ntfs_listxattr,
+	.atomic_open = ntfs_atomic_open,
+	.fiemap = ntfs_fiemap,
+};

target/linux/generic/files/fs/ntfs3/record.c

@@ -0,0 +1,609 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+static inline int compare_attr(const struct ATTRIB *left, enum ATTR_TYPE type,
+			       const __le16 *name, u8 name_len,
+			       const u16 *upcase)
+{
+	/* First, compare the type codes: */
+	int diff = le32_to_cpu(left->type) - le32_to_cpu(type);
+
+	if (diff)
+		return diff;
+
+	/*
+	 * They have the same type code, so we have to compare the names.
+	 */
+	return ntfs_cmp_names(attr_name(left), left->name_len, name, name_len,
+			      upcase, true);
+}
+
+/*
+ * mi_new_attt_id
+ *
+ * returns unused attribute id that is less than mrec->next_attr_id
+ */
+static __le16 mi_new_attt_id(struct mft_inode *mi)
+{
+	u16 free_id, max_id, t16;
+	struct MFT_REC *rec = mi->mrec;
+	struct ATTRIB *attr;
+	__le16 id;
+
+	id = rec->next_attr_id;
+	free_id = le16_to_cpu(id);
+	if (free_id < 0x7FFF) {
+		rec->next_attr_id = cpu_to_le16(free_id + 1);
+		return id;
+	}
+
+	/* One record can store up to 1024/24 ~= 42 attributes */
+	free_id = 0;
+	max_id = 0;
+
+	attr = NULL;
+
+	for (;;) {
+		attr = mi_enum_attr(mi, attr);
+		if (!attr) {
+			rec->next_attr_id = cpu_to_le16(max_id + 1);
+			mi->dirty = true;
+			return cpu_to_le16(free_id);
+		}
+
+		t16 = le16_to_cpu(attr->id);
+		if (t16 == free_id) {
+			free_id += 1;
+			attr = NULL;
+		} else if (max_id < t16)
+			max_id = t16;
+	}
+}
+
+int mi_get(struct ntfs_sb_info *sbi, CLST rno, struct mft_inode **mi)
+{
+	int err;
+	struct mft_inode *m = ntfs_zalloc(sizeof(struct mft_inode));
+
+	if (!m)
+		return -ENOMEM;
+
+	err = mi_init(m, sbi, rno);
+	if (err) {
+		ntfs_free(m);
+		return err;
+	}
+
+	err = mi_read(m, false);
+	if (err) {
+		mi_put(m);
+		return err;
+	}
+
+	*mi = m;
+	return 0;
+}
+
+void mi_put(struct mft_inode *mi)
+{
+	mi_clear(mi);
+	ntfs_free(mi);
+}
+
+int mi_init(struct mft_inode *mi, struct ntfs_sb_info *sbi, CLST rno)
+{
+	mi->sbi = sbi;
+	mi->rno = rno;
+	mi->mrec = ntfs_malloc(sbi->record_size);
+	if (!mi->mrec)
+		return -ENOMEM;
+
+	return 0;
+}
+
+/*
+ * mi_read
+ *
+ * reads MFT data
+ */
+int mi_read(struct mft_inode *mi, bool is_mft)
+{
+	int err;
+	struct MFT_REC *rec = mi->mrec;
+	struct ntfs_sb_info *sbi = mi->sbi;
+	u32 bpr = sbi->record_size;
+	u64 vbo = (u64)mi->rno << sbi->record_bits;
+	struct ntfs_inode *mft_ni = sbi->mft.ni;
+	struct runs_tree *run = mft_ni ? &mft_ni->file.run : NULL;
+	struct rw_semaphore *rw_lock = NULL;
+
+	if (is_mounted(sbi)) {
+		if (!is_mft) {
+			rw_lock = &mft_ni->file.run_lock;
+			down_read(rw_lock);
+		}
+	}
+
+	err = ntfs_read_bh(sbi, run, vbo, &rec->rhdr, bpr, &mi->nb);
+	if (rw_lock)
+		up_read(rw_lock);
+	if (!err)
+		goto ok;
+
+	if (err == -E_NTFS_FIXUP) {
+		mi->dirty = true;
+		goto ok;
+	}
+
+	if (err != -ENOENT)
+		goto out;
+
+	if (rw_lock) {
+		ni_lock(mft_ni);
+		down_write(rw_lock);
+	}
+	err = attr_load_runs_vcn(mft_ni, ATTR_DATA, NULL, 0, &mft_ni->file.run,
+				 vbo >> sbi->cluster_bits);
+	if (rw_lock) {
+		up_write(rw_lock);
+		ni_unlock(mft_ni);
+	}
+	if (err)
+		goto out;
+
+	if (rw_lock)
+		down_read(rw_lock);
+	err = ntfs_read_bh(sbi, run, vbo, &rec->rhdr, bpr, &mi->nb);
+	if (rw_lock)
+		up_read(rw_lock);
+
+	if (err == -E_NTFS_FIXUP) {
+		mi->dirty = true;
+		goto ok;
+	}
+	if (err)
+		goto out;
+
+ok:
+	/* check field 'total' only here */
+	if (le32_to_cpu(rec->total) != bpr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	return 0;
+
+out:
+	return err;
+}
+
+struct ATTRIB *mi_enum_attr(struct mft_inode *mi, struct ATTRIB *attr)
+{
+	const struct MFT_REC *rec = mi->mrec;
+	u32 used = le32_to_cpu(rec->used);
+	u32 t32, off, asize;
+	u16 t16;
+
+	if (!attr) {
+		u32 total = le32_to_cpu(rec->total);
+
+		off = le16_to_cpu(rec->attr_off);
+
+		if (used > total)
+			return NULL;
+
+		if (off >= used || off < MFTRECORD_FIXUP_OFFSET_1 ||
+		    !IsDwordAligned(off)) {
+			return NULL;
+		}
+
+		/* Skip non-resident records */
+		if (!is_rec_inuse(rec))
+			return NULL;
+
+		attr = Add2Ptr(rec, off);
+	} else {
+		/* Check if input attr inside record */
+		off = PtrOffset(rec, attr);
+		if (off >= used)
+			return NULL;
+
+		asize = le32_to_cpu(attr->size);
+		if (asize < SIZEOF_RESIDENT) {
+			/* Impossible 'cause we should not return such attribute */
+			return NULL;
+		}
+
+		attr = Add2Ptr(attr, asize);
+		off += asize;
+	}
+
+	asize = le32_to_cpu(attr->size);
+
+	/* Can we use the first field (attr->type) */
+	if (off + 8 > used) {
+		static_assert(QuadAlign(sizeof(enum ATTR_TYPE)) == 8);
+		return NULL;
+	}
+
+	if (attr->type == ATTR_END) {
+		/* end of enumeration */
+		return NULL;
+	}
+
+	/* 0x100 is last known attribute for now*/
+	t32 = le32_to_cpu(attr->type);
+	if ((t32 & 0xf) || (t32 > 0x100))
+		return NULL;
+
+	/* Check boundary */
+	if (off + asize > used)
+		return NULL;
+
+	/* Check size of attribute */
+	if (!attr->non_res) {
+		if (asize < SIZEOF_RESIDENT)
+			return NULL;
+
+		t16 = le16_to_cpu(attr->res.data_off);
+
+		if (t16 > asize)
+			return NULL;
+
+		t32 = le32_to_cpu(attr->res.data_size);
+		if (t16 + t32 > asize)
+			return NULL;
+
+		return attr;
+	}
+
+	/* Check some nonresident fields */
+	if (attr->name_len &&
+	    le16_to_cpu(attr->name_off) + sizeof(short) * attr->name_len >
+		    le16_to_cpu(attr->nres.run_off)) {
+		return NULL;
+	}
+
+	if (attr->nres.svcn || !is_attr_ext(attr)) {
+		if (asize + 8 < SIZEOF_NONRESIDENT)
+			return NULL;
+
+		if (attr->nres.c_unit)
+			return NULL;
+	} else if (asize + 8 < SIZEOF_NONRESIDENT_EX)
+		return NULL;
+
+	return attr;
+}
+
+/*
+ * mi_find_attr
+ *
+ * finds the attribute by type and name and id
+ */
+struct ATTRIB *mi_find_attr(struct mft_inode *mi, struct ATTRIB *attr,
+			    enum ATTR_TYPE type, const __le16 *name,
+			    size_t name_len, const __le16 *id)
+{
+	u32 type_in = le32_to_cpu(type);
+	u32 atype;
+
+next_attr:
+	attr = mi_enum_attr(mi, attr);
+	if (!attr)
+		return NULL;
+
+	atype = le32_to_cpu(attr->type);
+	if (atype > type_in)
+		return NULL;
+
+	if (atype < type_in)
+		goto next_attr;
+
+	if (attr->name_len != name_len)
+		goto next_attr;
+
+	if (name_len && memcmp(attr_name(attr), name, name_len * sizeof(short)))
+		goto next_attr;
+
+	if (id && *id != attr->id)
+		goto next_attr;
+
+	return attr;
+}
+
+int mi_write(struct mft_inode *mi, int wait)
+{
+	struct MFT_REC *rec;
+	int err;
+	struct ntfs_sb_info *sbi;
+
+	if (!mi->dirty)
+		return 0;
+
+	sbi = mi->sbi;
+	rec = mi->mrec;
+
+	err = ntfs_write_bh(sbi, &rec->rhdr, &mi->nb, wait);
+	if (err)
+		return err;
+
+	if (mi->rno < sbi->mft.recs_mirr)
+		sbi->flags |= NTFS_FLAGS_MFTMIRR;
+
+	mi->dirty = false;
+
+	return 0;
+}
+
+int mi_format_new(struct mft_inode *mi, struct ntfs_sb_info *sbi, CLST rno,
+		  __le16 flags, bool is_mft)
+{
+	int err;
+	u16 seq = 1;
+	struct MFT_REC *rec;
+	u64 vbo = (u64)rno << sbi->record_bits;
+
+	err = mi_init(mi, sbi, rno);
+	if (err)
+		return err;
+
+	rec = mi->mrec;
+
+	if (rno == MFT_REC_MFT) {
+		;
+	} else if (rno < MFT_REC_FREE) {
+		seq = rno;
+	} else if (rno >= sbi->mft.used) {
+		;
+	} else if (mi_read(mi, is_mft)) {
+		;
+	} else if (rec->rhdr.sign == NTFS_FILE_SIGNATURE) {
+		/* Record is reused. Update its sequence number */
+		seq = le16_to_cpu(rec->seq) + 1;
+		if (!seq)
+			seq = 1;
+	}
+
+	memcpy(rec, sbi->new_rec, sbi->record_size);
+
+	rec->seq = cpu_to_le16(seq);
+	rec->flags = RECORD_FLAG_IN_USE | flags;
+
+	mi->dirty = true;
+
+	if (!mi->nb.nbufs) {
+		struct ntfs_inode *ni = sbi->mft.ni;
+		bool lock = false;
+
+		if (is_mounted(sbi) && !is_mft) {
+			down_read(&ni->file.run_lock);
+			lock = true;
+		}
+
+		err = ntfs_get_bh(sbi, &ni->file.run, vbo, sbi->record_size,
+				  &mi->nb);
+		if (lock)
+			up_read(&ni->file.run_lock);
+	}
+
+	return err;
+}
+
+/*
+ * mi_mark_free
+ *
+ * marks record as unused and marks it as free in bitmap
+ */
+void mi_mark_free(struct mft_inode *mi)
+{
+	CLST rno = mi->rno;
+	struct ntfs_sb_info *sbi = mi->sbi;
+
+	if (rno >= MFT_REC_RESERVED && rno < MFT_REC_FREE) {
+		ntfs_clear_mft_tail(sbi, rno, rno + 1);
+		mi->dirty = false;
+		return;
+	}
+
+	if (mi->mrec) {
+		clear_rec_inuse(mi->mrec);
+		mi->dirty = true;
+		mi_write(mi, 0);
+	}
+	ntfs_mark_rec_free(sbi, rno);
+}
+
+/*
+ * mi_insert_attr
+ *
+ * reserves space for new attribute
+ * returns not full constructed attribute or NULL if not possible to create
+ */
+struct ATTRIB *mi_insert_attr(struct mft_inode *mi, enum ATTR_TYPE type,
+			      const __le16 *name, u8 name_len, u32 asize,
+			      u16 name_off)
+{
+	size_t tail;
+	struct ATTRIB *attr;
+	__le16 id;
+	struct MFT_REC *rec = mi->mrec;
+	struct ntfs_sb_info *sbi = mi->sbi;
+	u32 used = le32_to_cpu(rec->used);
+	const u16 *upcase = sbi->upcase;
+	int diff;
+
+	/* Can we insert mi attribute? */
+	if (used + asize > mi->sbi->record_size)
+		return NULL;
+
+	/*
+	 * Scan through the list of attributes to find the point
+	 * at which we should insert it.
+	 */
+	attr = NULL;
+	while ((attr = mi_enum_attr(mi, attr))) {
+		diff = compare_attr(attr, type, name, name_len, upcase);
+		if (diff > 0)
+			break;
+		if (diff < 0)
+			continue;
+
+		if (!is_attr_indexed(attr))
+			return NULL;
+		break;
+	}
+
+	if (!attr) {
+		tail = 8; /* not used, just to suppress warning */
+		attr = Add2Ptr(rec, used - 8);
+	} else {
+		tail = used - PtrOffset(rec, attr);
+	}
+
+	id = mi_new_attt_id(mi);
+
+	memmove(Add2Ptr(attr, asize), attr, tail);
+	memset(attr, 0, asize);
+
+	attr->type = type;
+	attr->size = cpu_to_le32(asize);
+	attr->name_len = name_len;
+	attr->name_off = cpu_to_le16(name_off);
+	attr->id = id;
+
+	memmove(Add2Ptr(attr, name_off), name, name_len * sizeof(short));
+	rec->used = cpu_to_le32(used + asize);
+
+	mi->dirty = true;
+
+	return attr;
+}
+
+/*
+ * mi_remove_attr
+ *
+ * removes the attribute from record
+ * NOTE: The source attr will point to next attribute
+ */
+bool mi_remove_attr(struct mft_inode *mi, struct ATTRIB *attr)
+{
+	struct MFT_REC *rec = mi->mrec;
+	u32 aoff = PtrOffset(rec, attr);
+	u32 used = le32_to_cpu(rec->used);
+	u32 asize = le32_to_cpu(attr->size);
+
+	if (aoff + asize > used)
+		return false;
+
+	used -= asize;
+	memmove(attr, Add2Ptr(attr, asize), used - aoff);
+	rec->used = cpu_to_le32(used);
+	mi->dirty = true;
+
+	return true;
+}
+
+/* bytes = "new attribute size" - "old attribute size" */
+bool mi_resize_attr(struct mft_inode *mi, struct ATTRIB *attr, int bytes)
+{
+	struct MFT_REC *rec = mi->mrec;
+	u32 aoff = PtrOffset(rec, attr);
+	u32 total, used = le32_to_cpu(rec->used);
+	u32 nsize, asize = le32_to_cpu(attr->size);
+	u32 rsize = le32_to_cpu(attr->res.data_size);
+	int tail = (int)(used - aoff - asize);
+	int dsize;
+	char *next;
+
+	if (tail < 0 || aoff >= used)
+		return false;
+
+	if (!bytes)
+		return true;
+
+	total = le32_to_cpu(rec->total);
+	next = Add2Ptr(attr, asize);
+
+	if (bytes > 0) {
+		dsize = QuadAlign(bytes);
+		if (used + dsize > total)
+			return false;
+		nsize = asize + dsize;
+		// move tail
+		memmove(next + dsize, next, tail);
+		memset(next, 0, dsize);
+		used += dsize;
+		rsize += dsize;
+	} else {
+		dsize = QuadAlign(-bytes);
+		if (dsize > asize)
+			return false;
+		nsize = asize - dsize;
+		memmove(next - dsize, next, tail);
+		used -= dsize;
+		rsize -= dsize;
+	}
+
+	rec->used = cpu_to_le32(used);
+	attr->size = cpu_to_le32(nsize);
+	if (!attr->non_res)
+		attr->res.data_size = cpu_to_le32(rsize);
+	mi->dirty = true;
+
+	return true;
+}
+
+int mi_pack_runs(struct mft_inode *mi, struct ATTRIB *attr,
+		 struct runs_tree *run, CLST len)
+{
+	int err = 0;
+	struct ntfs_sb_info *sbi = mi->sbi;
+	u32 new_run_size;
+	CLST plen;
+	struct MFT_REC *rec = mi->mrec;
+	CLST svcn = le64_to_cpu(attr->nres.svcn);
+	u32 used = le32_to_cpu(rec->used);
+	u32 aoff = PtrOffset(rec, attr);
+	u32 asize = le32_to_cpu(attr->size);
+	char *next = Add2Ptr(attr, asize);
+	u16 run_off = le16_to_cpu(attr->nres.run_off);
+	u32 run_size = asize - run_off;
+	u32 tail = used - aoff - asize;
+	u32 dsize = sbi->record_size - used;
+
+	/* Make a maximum gap in current record */
+	memmove(next + dsize, next, tail);
+
+	/* Pack as much as possible */
+	err = run_pack(run, svcn, len, Add2Ptr(attr, run_off), run_size + dsize,
+		       &plen);
+	if (err < 0) {
+		memmove(next, next + dsize, tail);
+		return err;
+	}
+
+	new_run_size = QuadAlign(err);
+
+	memmove(next + new_run_size - run_size, next + dsize, tail);
+
+	attr->size = cpu_to_le32(asize + new_run_size - run_size);
+	attr->nres.evcn = cpu_to_le64(svcn + plen - 1);
+	rec->used = cpu_to_le32(used + new_run_size - run_size);
+	mi->dirty = true;
+
+	return 0;
+}

target/linux/generic/files/fs/ntfs3/upcase.c

@@ -0,0 +1,105 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
+ *
+ */
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/module.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+static inline u16 upcase_unicode_char(const u16 *upcase, u16 chr)
+{
+	if (chr < 'a')
+		return chr;
+
+	if (chr <= 'z')
+		return chr - ('a' - 'A');
+
+	return upcase[chr];
+}
+
+/*
+ * Thanks Kari Argillander <kari.argillander@gmail.com> for idea and implementation 'bothcase'
+ *
+ * Straigth way to compare names:
+ * - case insensitive
+ * - if name equals and 'bothcases' then
+ * - case sensitive
+ * 'Straigth way' code scans input names twice in worst case
+ * Optimized code scans input names only once
+ */
+int ntfs_cmp_names(const __le16 *s1, size_t l1, const __le16 *s2, size_t l2,
+		   const u16 *upcase, bool bothcase)
+{
+	int diff1 = 0;
+	int diff2;
+	size_t len = min(l1, l2);
+
+	if (!bothcase && upcase)
+		goto case_insentive;
+
+	for (; len; s1++, s2++, len--) {
+		diff1 = le16_to_cpu(*s1) - le16_to_cpu(*s2);
+		if (diff1) {
+			if (bothcase && upcase)
+				goto case_insentive;
+
+			return diff1;
+		}
+	}
+	return l1 - l2;
+
+case_insentive:
+	for (; len; s1++, s2++, len--) {
+		diff2 = upcase_unicode_char(upcase, le16_to_cpu(*s1)) -
+			upcase_unicode_char(upcase, le16_to_cpu(*s2));
+		if (diff2)
+			return diff2;
+	}
+
+	diff2 = l1 - l2;
+	return diff2 ? diff2 : diff1;
+}
+
+int ntfs_cmp_names_cpu(const struct cpu_str *uni1, const struct le_str *uni2,
+		       const u16 *upcase, bool bothcase)
+{
+	const u16 *s1 = uni1->name;
+	const __le16 *s2 = uni2->name;
+	size_t l1 = uni1->len;
+	size_t l2 = uni2->len;
+	size_t len = min(l1, l2);
+	int diff1 = 0;
+	int diff2;
+
+	if (!bothcase && upcase)
+		goto case_insentive;
+
+	for (; len; s1++, s2++, len--) {
+		diff1 = *s1 - le16_to_cpu(*s2);
+		if (diff1) {
+			if (bothcase && upcase)
+				goto case_insentive;
+
+			return diff1;
+		}
+	}
+	return l1 - l2;
+
+case_insentive:
+	for (; len; s1++, s2++, len--) {
+		diff2 = upcase_unicode_char(upcase, *s1) -
+			upcase_unicode_char(upcase, le16_to_cpu(*s2));
+		if (diff2)
+			return diff2;
+	}
+
+	diff2 = l1 - l2;
+	return diff2 ? diff2 : diff1;
+}