Cleanup: split libsquashfs xattr writer code

This commit moves the libsquashfs xattr related code into a sub
directory and splits the xattr writer code up into several files.
No actual code is changed.

Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>

6 files changed, 1092 insertions, 0 deletions

diff --git a/lib/sqfs/xattr/xattr.c b/lib/sqfs/xattr/xattr.c
new file mode 100644
index 0000000..29ecebf
--- /dev/null
+++ b/lib/sqfs/xattr/xattr.c
@@ -0,0 +1,49 @@
+/* SPDX-License-Identifier: LGPL-3.0-or-later */
+/*
+ * write_xattr.c
+ *
+ * Copyright (C) 2019 David Oberhollenzer <goliath@infraroot.at>
+ */
+#define SQFS_BUILDING_DLL
+#include "config.h"
+#include "sqfs/xattr.h"
+#include "sqfs/error.h"
+
+#include <string.h>
+
+static const struct {
+	const char *prefix;
+	SQFS_XATTR_TYPE type;
+} xattr_types[] = {
+	{ "user.", SQFS_XATTR_USER },
+	{ "trusted.", SQFS_XATTR_TRUSTED },
+	{ "security.", SQFS_XATTR_SECURITY },
+};
+
+int sqfs_get_xattr_prefix_id(const char *key)
+{
+	size_t i, len;
+
+	for (i = 0; i < sizeof(xattr_types) / sizeof(xattr_types[0]); ++i) {
+		len = strlen(xattr_types[i].prefix);
+
+		if (strncmp(key, xattr_types[i].prefix, len) == 0 &&
+		    strlen(key) > len) {
+			return xattr_types[i].type;
+		}
+	}
+
+	return SQFS_ERROR_UNSUPPORTED;
+}
+
+const char *sqfs_get_xattr_prefix(SQFS_XATTR_TYPE id)
+{
+	size_t i;
+
+	for (i = 0; i < sizeof(xattr_types) / sizeof(xattr_types[0]); ++i) {
+		if (xattr_types[i].type == id)
+			return xattr_types[i].prefix;
+	}
+
+	return NULL;
+}
diff --git a/lib/sqfs/xattr/xattr_reader.c b/lib/sqfs/xattr/xattr_reader.c
new file mode 100644
index 0000000..37f222a
--- /dev/null
+++ b/lib/sqfs/xattr/xattr_reader.c
@@ -0,0 +1,349 @@
+/* SPDX-License-Identifier: LGPL-3.0-or-later */
+/*
+ * xattr_reader.c
+ *
+ * Copyright (C) 2019 David Oberhollenzer <goliath@infraroot.at>
+ */
+#define SQFS_BUILDING_DLL
+#include "config.h"
+
+#include "sqfs/xattr_reader.h"
+#include "sqfs/meta_reader.h"
+#include "sqfs/super.h"
+#include "sqfs/xattr.h"
+#include "sqfs/error.h"
+#include "sqfs/block.h"
+#include "sqfs/io.h"
+#include "util.h"
+
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+
+struct sqfs_xattr_reader_t {
+	sqfs_object_t base;
+
+	sqfs_u64 xattr_start;
+	sqfs_u64 xattr_end;
+
+	size_t num_id_blocks;
+	size_t num_ids;
+
+	sqfs_u64 *id_block_starts;
+
+	sqfs_meta_reader_t *idrd;
+	sqfs_meta_reader_t *kvrd;
+};
+
+static sqfs_object_t *xattr_reader_copy(const sqfs_object_t *obj)
+{
+	const sqfs_xattr_reader_t *xr = (const sqfs_xattr_reader_t *)obj;
+	sqfs_xattr_reader_t *copy = malloc(sizeof(*copy));
+
+	if (copy == NULL)
+		return NULL;
+
+	memcpy(copy, xr, sizeof(*xr));
+
+	if (xr->kvrd != NULL) {
+		copy->kvrd = sqfs_copy(xr->kvrd);
+		if (copy->kvrd == NULL)
+			goto fail_kvrd;
+	}
+
+	if (xr->idrd != NULL) {
+		copy->idrd = sqfs_copy(xr->idrd);
+		if (copy->idrd == NULL)
+			goto fail_idrd;
+	}
+
+	if (xr->id_block_starts != NULL) {
+		copy->id_block_starts = alloc_array(sizeof(sqfs_u64),
+						    xr->num_id_blocks);
+		if (copy->id_block_starts == NULL)
+			goto fail_idblk;
+
+		memcpy(copy->id_block_starts, xr->id_block_starts,
+		       sizeof(sqfs_u64) * xr->num_id_blocks);
+	}
+
+	return (sqfs_object_t *)copy;
+fail_idblk:
+	if (copy->idrd != NULL)
+		sqfs_destroy(copy->idrd);
+fail_idrd:
+	if (copy->kvrd != NULL)
+		sqfs_destroy(copy->kvrd);
+fail_kvrd:
+	free(copy);
+	return NULL;
+}
+
+static void xattr_reader_destroy(sqfs_object_t *obj)
+{
+	sqfs_xattr_reader_t *xr = (sqfs_xattr_reader_t *)obj;
+
+	if (xr->kvrd != NULL)
+		sqfs_destroy(xr->kvrd);
+
+	if (xr->idrd != NULL)
+		sqfs_destroy(xr->idrd);
+
+	free(xr->id_block_starts);
+	free(xr);
+}
+
+int sqfs_xattr_reader_load(sqfs_xattr_reader_t *xr, const sqfs_super_t *super,
+			   sqfs_file_t *file, sqfs_compressor_t *cmp)
+{
+	sqfs_xattr_id_table_t idtbl;
+	size_t i;
+	int err;
+
+	/* sanity check */
+	if (super->flags & SQFS_FLAG_NO_XATTRS)
+		return 0;
+
+	if (super->xattr_id_table_start == 0xFFFFFFFFFFFFFFFF)
+		return 0;
+
+	if (super->xattr_id_table_start >= super->bytes_used)
+		return SQFS_ERROR_OUT_OF_BOUNDS;
+
+	/* cleanup pre-existing data */
+	if (xr->idrd != NULL) {
+		sqfs_destroy(xr->idrd);
+		xr->idrd = NULL;
+	}
+
+	if (xr->kvrd != NULL) {
+		sqfs_destroy(xr->kvrd);
+		xr->kvrd = NULL;
+	}
+
+	free(xr->id_block_starts);
+	xr->id_block_starts = NULL;
+
+	/* read the locations table */
+	err = file->read_at(file, super->xattr_id_table_start,
+			    &idtbl, sizeof(idtbl));
+	if (err)
+		return err;
+
+	xr->xattr_start = le64toh(idtbl.xattr_table_start);
+	xr->num_ids = le32toh(idtbl.xattr_ids);
+	xr->num_id_blocks =
+		(xr->num_ids * sizeof(sqfs_xattr_id_t)) / SQFS_META_BLOCK_SIZE;
+
+	if ((xr->num_ids * sizeof(sqfs_xattr_id_t)) % SQFS_META_BLOCK_SIZE)
+		xr->num_id_blocks += 1;
+
+	xr->id_block_starts = alloc_array(sizeof(sqfs_u64), xr->num_id_blocks);
+	if (xr->id_block_starts == NULL) {
+		if (errno == EOVERFLOW)
+			return SQFS_ERROR_OVERFLOW;
+		return SQFS_ERROR_ALLOC;
+	}
+
+	err = file->read_at(file, super->xattr_id_table_start + sizeof(idtbl),
+			    xr->id_block_starts,
+			    sizeof(sqfs_u64) * xr->num_id_blocks);
+	if (err)
+		goto fail_blocks;
+
+	for (i = 0; i < xr->num_id_blocks; ++i) {
+		xr->id_block_starts[i] = le64toh(xr->id_block_starts[i]);
+
+		if (xr->id_block_starts[i] > super->bytes_used) {
+			err = SQFS_ERROR_OUT_OF_BOUNDS;
+			goto fail_blocks;
+		}
+	}
+
+	/* create the meta data readers */
+	xr->idrd = sqfs_meta_reader_create(file, cmp, super->id_table_start,
+					   super->bytes_used);
+	if (xr->idrd == NULL)
+		goto fail_blocks;
+
+	xr->kvrd = sqfs_meta_reader_create(file, cmp, super->id_table_start,
+					   super->bytes_used);
+	if (xr->kvrd == NULL)
+		goto fail_idrd;
+
+	xr->xattr_end = super->bytes_used;
+	return 0;
+fail_idrd:
+	sqfs_destroy(xr->idrd);
+	xr->idrd = NULL;
+fail_blocks:
+	free(xr->id_block_starts);
+	xr->id_block_starts = NULL;
+	return err;
+}
+
+int sqfs_xattr_reader_read_key(sqfs_xattr_reader_t *xr,
+			       sqfs_xattr_entry_t **key_out)
+{
+	sqfs_xattr_entry_t key, *out;
+	const char *prefix;
+	size_t plen, total;
+	int ret;
+
+	ret = sqfs_meta_reader_read(xr->kvrd, &key, sizeof(key));
+	if (ret)
+		return ret;
+
+	key.type = le16toh(key.type);
+	key.size = le16toh(key.size);
+
+	prefix = sqfs_get_xattr_prefix(key.type & SQFS_XATTR_PREFIX_MASK);
+	if (prefix == NULL)
+		return SQFS_ERROR_UNSUPPORTED;
+
+	plen = strlen(prefix);
+
+	if (SZ_ADD_OV(plen, key.size, &total) || SZ_ADD_OV(total, 1, &total) ||
+	    SZ_ADD_OV(sizeof(*out), total, &total)) {
+		return SQFS_ERROR_OVERFLOW;
+	}
+
+	out = calloc(1, total);
+	if (out == NULL)
+		return SQFS_ERROR_ALLOC;
+
+	*out = key;
+	memcpy(out->key, prefix, plen);
+
+	ret = sqfs_meta_reader_read(xr->kvrd, out->key + plen, key.size);
+	if (ret) {
+		free(out);
+		return ret;
+	}
+
+	*key_out = out;
+	return 0;
+}
+
+int sqfs_xattr_reader_read_value(sqfs_xattr_reader_t *xr,
+				 const sqfs_xattr_entry_t *key,
+				 sqfs_xattr_value_t **val_out)
+{
+	size_t offset, new_offset, size;
+	sqfs_xattr_value_t value, *out;
+	sqfs_u64 ref, start, new_start;
+	int ret;
+
+	ret = sqfs_meta_reader_read(xr->kvrd, &value, sizeof(value));
+	if (ret)
+		return ret;
+
+	if (key->type & SQFS_XATTR_FLAG_OOL) {
+		ret = sqfs_meta_reader_read(xr->kvrd, &ref, sizeof(ref));
+		if (ret)
+			return ret;
+
+		sqfs_meta_reader_get_position(xr->kvrd, &start, &offset);
+
+		new_start = xr->xattr_start + (ref >> 16);
+		if (new_start >= xr->xattr_end)
+			return SQFS_ERROR_OUT_OF_BOUNDS;
+
+		new_offset = ref & 0xFFFF;
+		if (new_offset >= SQFS_META_BLOCK_SIZE)
+			return SQFS_ERROR_OUT_OF_BOUNDS;
+
+		ret = sqfs_meta_reader_seek(xr->kvrd, new_start, new_offset);
+		if (ret)
+			return ret;
+	}
+
+	value.size = le32toh(value.size);
+
+	if (SZ_ADD_OV(sizeof(*out), value.size, &size) ||
+	    SZ_ADD_OV(size, 1, &size)) {
+		return SQFS_ERROR_OVERFLOW;
+	}
+
+	out = calloc(1, size);
+	if (out == NULL)
+		return SQFS_ERROR_ALLOC;
+
+	*out = value;
+
+	ret = sqfs_meta_reader_read(xr->kvrd, out->value, value.size);
+	if (ret)
+		goto fail;
+
+	if (key->type & SQFS_XATTR_FLAG_OOL) {
+		ret = sqfs_meta_reader_seek(xr->kvrd, start, offset);
+		if (ret)
+			goto fail;
+	}
+
+	*val_out = out;
+	return 0;
+fail:
+	free(out);
+	return ret;
+}
+
+int sqfs_xattr_reader_seek_kv(sqfs_xattr_reader_t *xr,
+			      const sqfs_xattr_id_t *desc)
+{
+	sqfs_u32 offset = desc->xattr & 0xFFFF;
+	sqfs_u64 block = xr->xattr_start + (desc->xattr >> 16);
+
+	return sqfs_meta_reader_seek(xr->kvrd, block, offset);
+}
+
+int sqfs_xattr_reader_get_desc(sqfs_xattr_reader_t *xr, sqfs_u32 idx,
+			       sqfs_xattr_id_t *desc)
+{
+	size_t block, offset;
+	int ret;
+
+	memset(desc, 0, sizeof(*desc));
+
+	if (idx == 0xFFFFFFFF)
+		return 0;
+
+	if (xr->kvrd == NULL || xr->idrd == NULL)
+		return idx == 0 ? 0 : SQFS_ERROR_OUT_OF_BOUNDS;
+
+	if (idx >= xr->num_ids)
+		return SQFS_ERROR_OUT_OF_BOUNDS;
+
+	offset = (idx * sizeof(*desc)) % SQFS_META_BLOCK_SIZE;
+	block = (idx * sizeof(*desc)) / SQFS_META_BLOCK_SIZE;
+
+	ret = sqfs_meta_reader_seek(xr->idrd, xr->id_block_starts[block],
+				    offset);
+	if (ret)
+		return ret;
+
+	ret = sqfs_meta_reader_read(xr->idrd, desc, sizeof(*desc));
+	if (ret)
+		return ret;
+
+	desc->xattr = le64toh(desc->xattr);
+	desc->count = le32toh(desc->count);
+	desc->size = le32toh(desc->size);
+	return 0;
+}
+
+sqfs_xattr_reader_t *sqfs_xattr_reader_create(sqfs_u32 flags)
+{
+	sqfs_xattr_reader_t *xr;
+
+	if (flags != 0)
+		return NULL;
+
+	xr = calloc(1, sizeof(*xr));
+	if (xr == NULL)
+		return NULL;
+
+	((sqfs_object_t *)xr)->copy = xattr_reader_copy;
+	((sqfs_object_t *)xr)->destroy = xattr_reader_destroy;
+	return xr;
+}
diff --git a/lib/sqfs/xattr/xattr_writer.c b/lib/sqfs/xattr/xattr_writer.c
new file mode 100644
index 0000000..9de3823
--- /dev/null
+++ b/lib/sqfs/xattr/xattr_writer.c
@@ -0,0 +1,110 @@
+/* SPDX-License-Identifier: LGPL-3.0-or-later */
+/*
+ * xattr_writer.c
+ *
+ * Copyright (C) 2019 David Oberhollenzer <goliath@infraroot.at>
+ */
+#include "xattr_writer.h"
+
+static sqfs_object_t *xattr_writer_copy(const sqfs_object_t *obj)
+{
+	const sqfs_xattr_writer_t *xwr = (const sqfs_xattr_writer_t *)obj;
+	kv_block_desc_t *blk, *it, **next;
+	sqfs_xattr_writer_t *copy;
+
+	copy = calloc(1, sizeof(*copy));
+	if (copy == NULL)
+		return NULL;
+
+	memcpy(copy, xwr, sizeof(*xwr));
+
+	if (str_table_copy(&copy->keys, &xwr->keys))
+		goto fail_keys;
+
+	if (str_table_copy(&copy->values, &xwr->values))
+		goto fail_values;
+
+	copy->max_pairs = xwr->num_pairs;
+	copy->num_pairs = xwr->num_pairs;
+
+	copy->kv_pairs = malloc(sizeof(copy->kv_pairs[0]) * xwr->num_pairs);
+	if (copy->kv_pairs == NULL)
+		goto fail_pairs;
+
+	memcpy(copy->kv_pairs, xwr->kv_pairs,
+	       sizeof(copy->kv_pairs[0]) * xwr->num_pairs);
+
+	next = &(copy->kv_blocks);
+
+	for (it = xwr->kv_blocks; it != NULL; it = it->next) {
+		blk = malloc(sizeof(*blk));
+		if (blk == NULL)
+			goto fail_blk;
+
+		memcpy(blk, it, sizeof(*it));
+		blk->next = NULL;
+
+		*next = blk;
+		next = &(blk->next);
+	}
+
+	return (sqfs_object_t *)copy;
+fail_blk:
+	while (copy->kv_blocks != NULL) {
+		blk = copy->kv_blocks;
+		copy->kv_blocks = copy->kv_blocks->next;
+		free(blk);
+	}
+fail_pairs:
+	str_table_cleanup(&copy->values);
+fail_values:
+	str_table_cleanup(&copy->keys);
+fail_keys:
+	free(copy);
+	return NULL;
+}
+
+static void xattr_writer_destroy(sqfs_object_t *obj)
+{
+	sqfs_xattr_writer_t *xwr = (sqfs_xattr_writer_t *)obj;
+	kv_block_desc_t *blk;
+
+	while (xwr->kv_blocks != NULL) {
+		blk = xwr->kv_blocks;
+		xwr->kv_blocks = xwr->kv_blocks->next;
+		free(blk);
+	}
+
+	free(xwr->kv_pairs);
+	str_table_cleanup(&xwr->values);
+	str_table_cleanup(&xwr->keys);
+	free(xwr);
+}
+
+sqfs_xattr_writer_t *sqfs_xattr_writer_create(void)
+{
+	sqfs_xattr_writer_t *xwr = calloc(1, sizeof(*xwr));
+
+	if (str_table_init(&xwr->keys, XATTR_KEY_BUCKETS))
+		goto fail_keys;
+
+	if (str_table_init(&xwr->values, XATTR_VALUE_BUCKETS))
+		goto fail_values;
+
+	xwr->max_pairs = XATTR_INITIAL_PAIR_CAP;
+	xwr->kv_pairs = alloc_array(sizeof(xwr->kv_pairs[0]), xwr->max_pairs);
+
+	if (xwr->kv_pairs == NULL)
+		goto fail_pairs;
+
+	((sqfs_object_t *)xwr)->copy = xattr_writer_copy;
+	((sqfs_object_t *)xwr)->destroy = xattr_writer_destroy;
+	return xwr;
+fail_pairs:
+	str_table_cleanup(&xwr->values);
+fail_values:
+	str_table_cleanup(&xwr->keys);
+fail_keys:
+	free(xwr);
+	return NULL;
+}
diff --git a/lib/sqfs/xattr/xattr_writer.h b/lib/sqfs/xattr/xattr_writer.h
new file mode 100644
index 0000000..826b6f6
--- /dev/null
+++ b/lib/sqfs/xattr/xattr_writer.h
@@ -0,0 +1,63 @@
+/* SPDX-License-Identifier: LGPL-3.0-or-later */
+/*
+ * xattr_writer.h
+ *
+ * Copyright (C) 2019 David Oberhollenzer <goliath@infraroot.at>
+ */
+#ifndef XATTR_WRITER_H
+#define XATTR_WRITER_H
+
+#define SQFS_BUILDING_DLL
+#include "config.h"
+
+#include "sqfs/xattr_writer.h"
+#include "sqfs/meta_writer.h"
+#include "sqfs/super.h"
+#include "sqfs/xattr.h"
+#include "sqfs/error.h"
+#include "sqfs/block.h"
+#include "sqfs/io.h"
+
+#include "str_table.h"
+#include "util.h"
+
+#include <stdlib.h>
+#include <string.h>
+#include <assert.h>
+
+
+#define XATTR_KEY_BUCKETS 31
+#define XATTR_VALUE_BUCKETS 511
+#define XATTR_INITIAL_PAIR_CAP 128
+
+#define MK_PAIR(key, value) (((sqfs_u64)(key) << 32UL) | (sqfs_u64)(value))
+#define GET_KEY(pair) ((pair >> 32UL) & 0x0FFFFFFFFUL)
+#define GET_VALUE(pair) (pair & 0x0FFFFFFFFUL)
+
+
+typedef struct kv_block_desc_t {
+	struct kv_block_desc_t *next;
+	size_t start;
+	size_t count;
+
+	sqfs_u64 start_ref;
+	size_t size_bytes;
+} kv_block_desc_t;
+
+struct sqfs_xattr_writer_t {
+	sqfs_object_t base;
+
+	str_table_t keys;
+	str_table_t values;
+
+	sqfs_u64 *kv_pairs;
+	size_t max_pairs;
+	size_t num_pairs;
+
+	size_t kv_start;
+
+	kv_block_desc_t *kv_blocks;
+	size_t num_blocks;
+};
+
+#endif /* XATTR_WRITER_H */
diff --git a/lib/sqfs/xattr/xattr_writer_flush.c b/lib/sqfs/xattr/xattr_writer_flush.c
new file mode 100644
index 0000000..9b3de30
--- /dev/null
+++ b/lib/sqfs/xattr/xattr_writer_flush.c
@@ -0,0 +1,342 @@
+/* SPDX-License-Identifier: LGPL-3.0-or-later */
+/*
+ * xattr_writer_flush.c
+ *
+ * Copyright (C) 2019 David Oberhollenzer <goliath@infraroot.at>
+ */
+#include "xattr_writer.h"
+
+static const char *hexmap = "0123456789ABCDEF";
+
+static void *from_base32(const char *input, size_t *size_out)
+{
+	sqfs_u8 lo, hi, *out, *ptr;
+	size_t len;
+
+	len = strlen(input);
+	*size_out = len / 2;
+
+	out = malloc(*size_out);
+	if (out == NULL)
+		return NULL;
+
+	ptr = out;
+
+	while (*input != '\0') {
+		lo = strchr(hexmap, *(input++)) - hexmap;
+		hi = strchr(hexmap, *(input++)) - hexmap;
+
+		*(ptr++) = lo | (hi << 4);
+	}
+
+	return out;
+}
+
+static sqfs_s32 write_key(sqfs_meta_writer_t *mw, const char *key,
+			  bool value_is_ool)
+{
+	sqfs_xattr_entry_t kent;
+	int type, err;
+	size_t len;
+
+	type = sqfs_get_xattr_prefix_id(key);
+	assert(type >= 0);
+
+	key = strchr(key, '.');
+	assert(key != NULL);
+	++key;
+	len = strlen(key);
+
+	if (value_is_ool)
+		type |= SQFS_XATTR_FLAG_OOL;
+
+	memset(&kent, 0, sizeof(kent));
+	kent.type = htole16(type);
+	kent.size = htole16(len);
+
+	err = sqfs_meta_writer_append(mw, &kent, sizeof(kent));
+	if (err)
+		return err;
+
+	err = sqfs_meta_writer_append(mw, key, len);
+	if (err)
+		return err;
+
+	return sizeof(kent) + len;
+}
+
+static sqfs_s32 write_value(sqfs_meta_writer_t *mw, const char *value_str,
+			    sqfs_u64 *value_ref_out)
+{
+	sqfs_xattr_value_t vent;
+	sqfs_u32 offset;
+	sqfs_u64 block;
+	size_t size;
+	void *value;
+	int err;
+
+	value = from_base32(value_str, &size);
+	if (value == NULL)
+		return SQFS_ERROR_ALLOC;
+
+	memset(&vent, 0, sizeof(vent));
+	vent.size = htole32(size);
+
+	sqfs_meta_writer_get_position(mw, &block, &offset);
+	*value_ref_out = (block << 16) | (offset & 0xFFFF);
+
+	err = sqfs_meta_writer_append(mw, &vent, sizeof(vent));
+	if (err)
+		goto fail;
+
+	err = sqfs_meta_writer_append(mw, value, size);
+	if (err)
+		goto fail;
+
+	free(value);
+	return sizeof(vent) + size;
+fail:
+	free(value);
+	return err;
+}
+
+static sqfs_s32 write_value_ool(sqfs_meta_writer_t *mw, sqfs_u64 location)
+{
+	sqfs_xattr_value_t vent;
+	sqfs_u64 ref;
+	int err;
+
+	memset(&vent, 0, sizeof(vent));
+	vent.size = htole32(sizeof(location));
+	ref = htole64(location);
+
+	err = sqfs_meta_writer_append(mw, &vent, sizeof(vent));
+	if (err)
+		return err;
+
+	err = sqfs_meta_writer_append(mw, &ref, sizeof(ref));
+	if (err)
+		return err;
+
+	return sizeof(vent) + sizeof(ref);
+}
+
+static bool should_store_ool(const char *val_str, size_t refcount)
+{
+	if (refcount < 2)
+		return false;
+
+	/*
+	  Storing in line needs this many bytes: refcount * len
+
+	  Storing out-of-line needs this many: len + (refcount - 1) * 8
+
+	  Out-of-line prefereable iff refcount > 1 and:
+	      refcount * len > len + (refcount - 1) * 8
+	   => refcount * len - len > (refcount - 1) * 8
+	   => (refcount - 1) * len > (refcount - 1) * 8
+	   => len > 8
+	 */
+	return (strlen(val_str) / 2) > sizeof(sqfs_u64);
+}
+
+static int write_block_pairs(sqfs_xattr_writer_t *xwr, sqfs_meta_writer_t *mw,
+			     kv_block_desc_t *blk, sqfs_u64 *ool_locations)
+{
+	sqfs_u32 key_idx, val_idx;
+	const char *key_str, *value_str;
+	sqfs_s32 diff, total = 0;
+	size_t i, refcount;
+	sqfs_u64 ref;
+
+	for (i = 0; i < blk->count; ++i) {
+		key_idx = GET_KEY(xwr->kv_pairs[blk->start + i]);
+		val_idx = GET_VALUE(xwr->kv_pairs[blk->start + i]);
+
+		key_str = str_table_get_string(&xwr->keys, key_idx);
+		value_str = str_table_get_string(&xwr->values, val_idx);
+
+		if (ool_locations[val_idx] == 0xFFFFFFFFFFFFFFFFUL) {
+			diff = write_key(mw, key_str, false);
+			if (diff < 0)
+				return diff;
+			total += diff;
+
+			diff = write_value(mw, value_str, &ref);
+			if (diff < 0)
+				return diff;
+			total += diff;
+
+			refcount = str_table_get_ref_count(&xwr->values,
+							   val_idx);
+
+			if (should_store_ool(value_str, refcount))
+				ool_locations[val_idx] = ref;
+		} else {
+			diff = write_key(mw, key_str, true);
+			if (diff < 0)
+				return diff;
+			total += diff;
+
+			diff = write_value_ool(mw, ool_locations[val_idx]);
+			if (diff < 0)
+				return diff;
+			total += diff;
+		}
+	}
+
+	return total;
+}
+
+static int write_kv_pairs(sqfs_xattr_writer_t *xwr, sqfs_meta_writer_t *mw)
+{
+	sqfs_u64 block, *ool_locations;
+	kv_block_desc_t *blk;
+	sqfs_u32 offset;
+	sqfs_s32 size;
+	size_t i;
+
+	ool_locations = alloc_array(sizeof(ool_locations[0]),
+				    xwr->values.num_strings);
+	if (ool_locations == NULL)
+		return SQFS_ERROR_ALLOC;
+
+	for (i = 0; i < xwr->values.num_strings; ++i)
+		ool_locations[i] = 0xFFFFFFFFFFFFFFFFUL;
+
+	for (blk = xwr->kv_blocks; blk != NULL; blk = blk->next) {
+		sqfs_meta_writer_get_position(mw, &block, &offset);
+		blk->start_ref = (block << 16) | (offset & 0xFFFF);
+
+		size = write_block_pairs(xwr, mw, blk, ool_locations);
+		if (size < 0) {
+			free(ool_locations);
+			return size;
+		}
+
+		blk->size_bytes = size;
+	}
+
+	free(ool_locations);
+	return sqfs_meta_writer_flush(mw);
+}
+
+static int write_id_table(sqfs_xattr_writer_t *xwr, sqfs_meta_writer_t *mw,
+			  sqfs_u64 *locations)
+{
+	sqfs_xattr_id_t id_ent;
+	kv_block_desc_t *blk;
+	sqfs_u32 offset;
+	sqfs_u64 block;
+	size_t i = 0;
+	int err;
+
+	locations[i++] = 0;
+
+	for (blk = xwr->kv_blocks; blk != NULL; blk = blk->next) {
+		memset(&id_ent, 0, sizeof(id_ent));
+		id_ent.xattr = htole64(blk->start_ref);
+		id_ent.count = htole32(blk->count);
+		id_ent.size = htole32(blk->size_bytes);
+
+		err = sqfs_meta_writer_append(mw, &id_ent, sizeof(id_ent));
+		if (err)
+			return err;
+
+		sqfs_meta_writer_get_position(mw, &block, &offset);
+		if (block != locations[i - 1])
+			locations[i++] = block;
+	}
+
+	return sqfs_meta_writer_flush(mw);
+}
+
+static int write_location_table(sqfs_xattr_writer_t *xwr, sqfs_u64 kv_start,
+				sqfs_file_t *file, const sqfs_super_t *super,
+				sqfs_u64 *locations, size_t loc_count)
+{
+	sqfs_xattr_id_table_t idtbl;
+	int err;
+
+	memset(&idtbl, 0, sizeof(idtbl));
+	idtbl.xattr_table_start = htole64(kv_start);
+	idtbl.xattr_ids = htole32(xwr->num_blocks);
+
+	err = file->write_at(file, super->xattr_id_table_start,
+			     &idtbl, sizeof(idtbl));
+	if (err)
+		return err;
+
+	return file->write_at(file, super->xattr_id_table_start + sizeof(idtbl),
+			      locations, sizeof(locations[0]) * loc_count);
+}
+
+static int alloc_location_table(sqfs_xattr_writer_t *xwr, sqfs_u64 **tbl_out,
+				size_t *szout)
+{
+	sqfs_u64 *locations;
+	size_t size, count;
+
+	if (SZ_MUL_OV(xwr->num_blocks, sizeof(sqfs_xattr_id_t), &size))
+		return SQFS_ERROR_OVERFLOW;
+
+	count = size / SQFS_META_BLOCK_SIZE;
+	if (size % SQFS_META_BLOCK_SIZE)
+		++count;
+
+	locations = alloc_array(sizeof(sqfs_u64), count);
+	if (locations == NULL)
+		return SQFS_ERROR_ALLOC;
+
+	*tbl_out = locations;
+	*szout = count;
+	return 0;
+}
+
+int sqfs_xattr_writer_flush(sqfs_xattr_writer_t *xwr, sqfs_file_t *file,
+			    sqfs_super_t *super, sqfs_compressor_t *cmp)
+{
+	sqfs_u64 *locations = NULL, kv_start, id_start;
+	sqfs_meta_writer_t *mw;
+	size_t i, count;
+	int err;
+
+	if (xwr->num_pairs == 0 || xwr->num_blocks == 0) {
+		super->xattr_id_table_start = 0xFFFFFFFFFFFFFFFFUL;
+		super->flags |= SQFS_FLAG_NO_XATTRS;
+		return 0;
+	}
+
+	mw = sqfs_meta_writer_create(file, cmp, 0);
+	if (mw == NULL)
+		return SQFS_ERROR_ALLOC;
+
+	kv_start = file->get_size(file);
+	err = write_kv_pairs(xwr, mw);
+	if (err)
+		goto out;
+
+	sqfs_meta_writer_reset(mw);
+
+	id_start = file->get_size(file);
+	err = alloc_location_table(xwr, &locations, &count);
+	if (err)
+		goto out;
+
+	err = write_id_table(xwr, mw, locations);
+	if (err)
+		goto out;
+
+	super->xattr_id_table_start = file->get_size(file);
+	super->flags &= ~SQFS_FLAG_NO_XATTRS;
+
+	for (i = 0; i < count; ++i)
+		locations[i] = htole64(locations[i] + id_start);
+
+	err = write_location_table(xwr, kv_start, file, super,
+				   locations, count);
+out:
+	free(locations);
+	sqfs_destroy(mw);
+	return err;
+}
diff --git a/lib/sqfs/xattr/xattr_writer_record.c b/lib/sqfs/xattr/xattr_writer_record.c
new file mode 100644
index 0000000..5e97868
--- /dev/null
+++ b/lib/sqfs/xattr/xattr_writer_record.c
@@ -0,0 +1,179 @@
+/* SPDX-License-Identifier: LGPL-3.0-or-later */
+/*
+ * xattr_writer_record.c
+ *
+ * Copyright (C) 2019 David Oberhollenzer <goliath@infraroot.at>
+ */
+#include "xattr_writer.h"
+
+static const char *hexmap = "0123456789ABCDEF";
+
+static char *to_base32(const void *input, size_t size)
+{
+	const sqfs_u8 *in = input;
+	char *out, *ptr;
+	size_t i;
+
+	out = malloc(2 * size + 1);
+	if (out == NULL)
+		return NULL;
+
+	ptr = out;
+
+	for (i = 0; i < size; ++i) {
+		*(ptr++) = hexmap[ in[i]       & 0x0F];
+		*(ptr++) = hexmap[(in[i] >> 4) & 0x0F];
+	}
+
+	*ptr = '\0';
+	return out;
+}
+
+static int compare_u64(const void *a, const void *b)
+{
+	sqfs_u64 lhs = *((const sqfs_u64 *)a);
+	sqfs_u64 rhs = *((const sqfs_u64 *)b);
+
+	return (lhs < rhs ? -1 : (lhs > rhs ? 1 : 0));
+}
+
+int sqfs_xattr_writer_begin(sqfs_xattr_writer_t *xwr)
+{
+	xwr->kv_start = xwr->num_pairs;
+	return 0;
+}
+
+int sqfs_xattr_writer_add(sqfs_xattr_writer_t *xwr, const char *key,
+			  const void *value, size_t size)
+{
+	size_t i, key_index, old_value_index, value_index, new_count;
+	sqfs_u64 kv_pair, *new;
+	char *value_str;
+	int err;
+
+	if (sqfs_get_xattr_prefix_id(key) < 0)
+		return SQFS_ERROR_UNSUPPORTED;
+
+	/* resolve key and value into unique, incremental IDs */
+	err = str_table_get_index(&xwr->keys, key, &key_index);
+	if (err)
+		return err;
+
+	value_str = to_base32(value, size);
+	if (value_str == NULL)
+		return SQFS_ERROR_ALLOC;
+
+	err = str_table_get_index(&xwr->values, value_str, &value_index);
+	free(value_str);
+	if (err)
+		return err;
+
+	str_table_add_ref(&xwr->values, value_index);
+
+	if (sizeof(size_t) > sizeof(sqfs_u32)) {
+		if (key_index > 0x0FFFFFFFFUL || value_index > 0x0FFFFFFFFUL)
+			return SQFS_ERROR_OVERFLOW;
+	}
+
+	/* bail if already have the pair, overwrite if we have the key */
+	kv_pair = MK_PAIR(key_index, value_index);
+
+	for (i = xwr->kv_start; i < xwr->num_pairs; ++i) {
+		if (xwr->kv_pairs[i] == kv_pair)
+			return 0;
+
+		if (GET_KEY(xwr->kv_pairs[i]) == key_index) {
+			old_value_index = GET_VALUE(xwr->kv_pairs[i]);
+
+			str_table_del_ref(&xwr->values, old_value_index);
+
+			xwr->kv_pairs[i] = kv_pair;
+			return 0;
+		}
+	}
+
+	/* append it to the list */
+	if (xwr->max_pairs == xwr->num_pairs) {
+		new_count = xwr->max_pairs * 2;
+		new = realloc(xwr->kv_pairs,
+			      sizeof(xwr->kv_pairs[0]) * new_count);
+
+		if (new == NULL)
+			return SQFS_ERROR_ALLOC;
+
+		xwr->kv_pairs = new;
+		xwr->max_pairs = new_count;
+	}
+
+	xwr->kv_pairs[xwr->num_pairs++] = kv_pair;
+	return 0;
+}
+
+int sqfs_xattr_writer_end(sqfs_xattr_writer_t *xwr, sqfs_u32 *out)
+{
+	kv_block_desc_t *blk, *blk_prev;
+	size_t i, count, value_idx;
+	sqfs_u32 index;
+	int ret;
+
+	count = xwr->num_pairs - xwr->kv_start;
+	if (count == 0) {
+		*out = 0xFFFFFFFF;
+		return 0;
+	}
+
+	qsort(xwr->kv_pairs + xwr->kv_start, count,
+	      sizeof(xwr->kv_pairs[0]), compare_u64);
+
+	blk_prev = NULL;
+	blk = xwr->kv_blocks;
+	index = 0;
+
+	while (blk != NULL) {
+		if (blk->count == count) {
+			ret = memcmp(xwr->kv_pairs + blk->start,
+				     xwr->kv_pairs + xwr->kv_start,
+				     sizeof(xwr->kv_pairs[0]) * count);
+
+			if (ret == 0)
+				break;
+		}
+
+		if (index == 0xFFFFFFFF)
+			return SQFS_ERROR_OVERFLOW;
+
+		++index;
+		blk_prev = blk;
+		blk = blk->next;
+	}
+
+	if (blk != NULL) {
+		for (i = 0; i < count; ++i) {
+			value_idx = GET_VALUE(xwr->kv_pairs[xwr->kv_start + i]);
+			str_table_del_ref(&xwr->values, value_idx);
+
+			value_idx = GET_VALUE(xwr->kv_pairs[blk->start + i]);
+			str_table_add_ref(&xwr->values, value_idx);
+		}
+
+		xwr->num_pairs = xwr->kv_start;
+	} else {
+		blk = calloc(1, sizeof(*blk));
+		if (blk == NULL)
+			return SQFS_ERROR_ALLOC;
+
+		blk->start = xwr->kv_start;
+		blk->count = count;
+
+		if (blk_prev == NULL) {
+			xwr->kv_blocks = blk;
+		} else {
+			blk_prev->next = blk;
+		}
+
+		xwr->num_blocks += 1;
+	}
+
+	*out = index;
+	return 0;
+}