liberasurecode/src/erasurecode_preprocessing.c
Jim Cheung 06e50ea86b Add Phazr.IO libphazr backend to liberasurecode
Currently, there are several implementations of erasure codes that are
available within OpenStack Swift.  Most, if not all, of which are based
on the Reed Solomon coding algorithm.

Phazr.IO’s Erasure Coding technology uses a patented algorithm which are
significantly more efficient and improves the speed of coding, decoding
and reconstruction.  In addition, Phazr.IO Erasure Code use a non-systematic
algorithm which provides data protection at rest and in transport without
the need to use encryption.

Please contact support@phazr.io for more info on our technology.

Change-Id: I4e40d02a8951e38409ad3c604c5dd6f050fa7ea0
2017-02-28 11:14:11 -08:00

385 lines
13 KiB
C

/*
* Copyright 2014 Tushar Gohad, Kevin M Greenan, Eric Lambert, Mark Storer
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* Redistributions of source code must retain the above copyright notice, this
* list of conditions and the following disclaimer.
*
* Redistributions in binary form must reproduce the above copyright notice, this
* list of conditions and the following disclaimer in the documentation and/or
* other materials provided with the distribution. THIS SOFTWARE IS PROVIDED BY
* THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED
* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO
* EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
* INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
* OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
* ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* liberasurecode proprocssing helpers implementation
*
* vi: set noai tw=79 ts=4 sw=4:
*/
#include "erasurecode_backend.h"
#include "erasurecode_helpers.h"
#include "erasurecode_helpers_ext.h"
#include "erasurecode_log.h"
#include "erasurecode_preprocessing.h"
#include "erasurecode_stdinc.h"
int prepare_fragments_for_encode(ec_backend_t instance,
int k, int m,
const char *orig_data, uint64_t orig_data_size, /* input */
char **encoded_data, char **encoded_parity, /* output */
int *blocksize)
{
int i, ret = 0;
int data_len; /* data len to write to fragment headers */
int aligned_data_len; /* EC algorithm compatible data length */
int buffer_size, payload_size = 0;
int metadata_size, data_offset = 0;
/* Calculate data sizes, aligned_data_len guaranteed to be divisible by k*/
data_len = orig_data_size;
aligned_data_len = get_aligned_data_size(instance, orig_data_size);
*blocksize = payload_size = (aligned_data_len / k);
metadata_size = instance->common.ops->get_backend_metadata_size(
instance->desc.backend_desc,
*blocksize);
data_offset = instance->common.ops->get_encode_offset(
instance->desc.backend_desc,
metadata_size);
buffer_size = payload_size + metadata_size;
for (i = 0; i < k; i++) {
int copy_size = data_len > payload_size ? payload_size : data_len;
char *fragment = (char *) alloc_fragment_buffer(buffer_size);
if (NULL == fragment) {
ret = -ENOMEM;
goto out_error;
}
/* Copy existing data into clean, zero'd out buffer */
encoded_data[i] = get_data_ptr_from_fragment(fragment);
if (data_len > 0) {
memcpy(encoded_data[i] + data_offset, orig_data, copy_size);
}
orig_data += copy_size;
data_len -= copy_size;
}
for (i = 0; i < m; i++) {
char *fragment = (char *) alloc_fragment_buffer(buffer_size);
if (NULL == fragment) {
ret = -ENOMEM;
goto out_error;
}
encoded_parity[i] = get_data_ptr_from_fragment(fragment);
}
out:
return ret;
out_error:
printf ("ERROR in encode\n");
if (encoded_data) {
for (i = 0; i < k; i++) {
if (encoded_data[i])
free_fragment_buffer(encoded_data[i]);
}
check_and_free_buffer(encoded_data);
}
if (encoded_parity) {
for (i = 0; i < m; i++) {
if (encoded_parity[i])
free_fragment_buffer(encoded_parity[i]);
}
check_and_free_buffer(encoded_parity);
}
goto out;
}
/*
* Note that the caller should always check realloc_bm during success or
* failure to free buffers allocated here. We could free up in this function,
* but it is internal to this library and only used in a few places. In any
* case, the caller has to free up in the success case, so it may as well do
* so in the failure case.
*/
int prepare_fragments_for_decode(
int k, int m,
char **data, char **parity,
int *missing_idxs,
int *orig_size, int *fragment_payload_size, int fragment_size,
uint64_t *realloc_bm)
{
int i; /* a counter */
unsigned long long missing_bm; /* bitmap form of missing indexes list */
int orig_data_size = -1;
int payload_size = -1;
missing_bm = convert_list_to_bitmap(missing_idxs);
/*
* Determine if each data fragment is:
* 1.) Alloc'd: if not, alloc new buffer (for missing fragments)
* 2.) Aligned to 16-byte boundaries: if not, alloc a new buffer
* memcpy the contents and free the old buffer
*/
for (i = 0; i < k; i++) {
/*
* Allocate or replace with aligned buffer if the buffer was not
* aligned.
* DO NOT FREE: the python GC should free the original when cleaning up
* 'data_list'
*/
if (NULL == data[i]) {
data[i] = alloc_fragment_buffer(fragment_size - sizeof(fragment_header_t));
if (NULL == data[i]) {
log_error("Could not allocate data buffer!");
return -ENOMEM;
}
*realloc_bm = *realloc_bm | (1 << i);
} else if (!is_addr_aligned((unsigned long)data[i], 16)) {
char *tmp_buf = alloc_fragment_buffer(fragment_size - sizeof(fragment_header_t));
if (NULL == tmp_buf) {
log_error("Could not allocate temp buffer!");
return -ENOMEM;
}
memcpy(tmp_buf, data[i], fragment_size);
data[i] = tmp_buf;
*realloc_bm = *realloc_bm | (1 << i);
}
/* Need to determine the size of the original data */
if (((missing_bm & (1 << i)) == 0) && orig_data_size < 0) {
orig_data_size = get_orig_data_size(data[i]);
if (orig_data_size < 0) {
log_error("Invalid orig_data_size in fragment header!");
return -EBADHEADER;
}
payload_size = get_fragment_payload_size(data[i]);
if (orig_data_size < 0) {
log_error("Invalid fragment_size in fragment header!");
return -EBADHEADER;
}
}
}
/* Perform the same allocation, alignment checks on the parity fragments */
for (i = 0; i < m; i++) {
/*
* Allocate or replace with aligned buffer, if the buffer was not aligned.
* DO NOT FREE: the python GC should free the original when cleaning up 'data_list'
*/
if (NULL == parity[i]) {
parity[i] = alloc_fragment_buffer(fragment_size-sizeof(fragment_header_t));
if (NULL == parity[i]) {
log_error("Could not allocate parity buffer!");
return -ENOMEM;
}
*realloc_bm = *realloc_bm | (1 << (k + i));
} else if (!is_addr_aligned((unsigned long)parity[i], 16)) {
char *tmp_buf = alloc_fragment_buffer(fragment_size-sizeof(fragment_header_t));
if (NULL == tmp_buf) {
log_error("Could not allocate temp buffer!");
return -ENOMEM;
}
memcpy(tmp_buf, parity[i], fragment_size);
parity[i] = tmp_buf;
*realloc_bm = *realloc_bm | (1 << (k + i));
}
/* Need to determine the size of the original data */
if (((missing_bm & (1 << (k + i))) == 0) && orig_data_size < 0) {
orig_data_size = get_orig_data_size(parity[i]);
if (orig_data_size < 0) {
log_error("Invalid orig_data_size in fragment header!");
return -EBADHEADER;
}
payload_size = get_fragment_payload_size(parity[i]);
if (orig_data_size < 0) {
log_error("Invalid fragment_size in fragment header!");
return -EBADHEADER;
}
}
}
*orig_size = orig_data_size;
*fragment_payload_size = payload_size;
return 0;
}
int get_fragment_partition(
int k, int m,
char **fragments, int num_fragments,
char **data, char **parity, int *missing)
{
int i = 0;
int num_missing = 0;
int index;
/*
* Set all data and parity entries to NULL
*/
for (i = 0; i < k; i++) {
data[i] = NULL;
}
for (i = 0; i < m; i++) {
parity[i] = NULL;
}
/*
* Fill in data and parity with available fragments
*/
for (i = 0; i < num_fragments; i++) {
index = get_fragment_idx(fragments[i]);
if (index < 0 || index > (k + m)) {
return -EBADHEADER;
}
if (index < k) {
data[index] = fragments[i];
} else {
parity[index - k] = fragments[i];
}
}
/*
* Fill in missing array with missing indexes
*/
for (i = 0; i < k; i++) {
if (NULL == data[i]) {
missing[num_missing] = i;
num_missing++;
}
}
for (i = 0; i < m; i++) {
if (NULL == parity[i]) {
missing[num_missing] = i + k;
num_missing++;
}
}
// TODO: In general, it is possible to reconstruct one or more fragments
// when more than m fragments are missing (e.g. flat XOR codes)
return (num_missing > m) ? -EINSUFFFRAGS : 0;
}
int fragments_to_string(int k, int m,
char **fragments, int num_fragments,
char **orig_payload, uint64_t *payload_len)
{
char *internal_payload = NULL;
char **data = NULL;
int orig_data_size = -1;
int i;
int index;
int data_size;
int num_data = 0;
int string_off = 0;
int ret = -1;
if (num_fragments < k) {
/*
* This is not necessarily an error condition, so *do not log here*
* We can maybe debug log, if necessary.
*/
goto out;
}
data = (char **) get_aligned_buffer16(sizeof(char *) * k);
if (NULL == data) {
log_error("Could not allocate buffer for data!!");
ret = -ENOMEM;
goto out;
}
for (i = 0; i < num_fragments; i++) {
index = get_fragment_idx(fragments[i]);
data_size = get_fragment_payload_size(fragments[i]);
if ((index < 0) || (data_size < 0)) {
log_error("Invalid fragment header information!");
ret = -EBADHEADER;
goto out;
}
/* Validate the original data size */
if (orig_data_size < 0) {
orig_data_size = get_orig_data_size(fragments[i]);
} else {
if (get_orig_data_size(fragments[i]) != orig_data_size) {
log_error("Inconsistent orig_data_size in fragment header!");
ret = -EBADHEADER;
goto out;
}
}
/* Skip parity fragments, put data fragments in index order */
if (index >= k) {
continue;
} else {
/* Make sure we account for duplicates */
if (NULL == data[index]) {
data[index] = fragments[i];
num_data++;
}
}
}
/* We do not have enough data fragments to do this! */
if (num_data != k) {
/*
* This is not necessarily an error condition, so *do not log here*
* We can maybe debug log, if necessary.
*/
goto out;
}
/* Create the string to return */
internal_payload = (char *) get_aligned_buffer16(orig_data_size);
if (NULL == internal_payload) {
log_error("Could not allocate buffer for decoded string!");
ret = -ENOMEM;
goto out;
}
/* Pass the original data length back */
*payload_len = orig_data_size;
/* Copy fragment data into cstring (fragments should be in index order) */
for (i = 0; i < num_data && orig_data_size > 0; i++) {
char* fragment_data = get_data_ptr_from_fragment(data[i]);
int fragment_size = get_fragment_payload_size(data[i]);
int payload_size = orig_data_size > fragment_size ? fragment_size : orig_data_size;
memcpy(internal_payload + string_off, fragment_data, payload_size);
orig_data_size -= payload_size;
string_off += payload_size;
}
/* Everything worked just fine */
ret = 0;
out:
if (NULL != data) {
free(data);
}
*orig_payload = internal_payload;
return ret;
}