HS_PUBLIC_API
hs_error_t hs_open_stream(const hs_database_t *db, UNUSED unsigned flags,
hs_stream_t **stream) {
if (unlikely(!stream)) {
return HS_INVALID;
}
*stream = NULL;
hs_error_t err = validDatabase(db);
if (unlikely(err != HS_SUCCESS)) {
return err;
}
const struct RoseEngine *rose = hs_get_bytecode(db);
if (unlikely(!ISALIGNED_16(rose))) {
return HS_INVALID;
}
if (unlikely(rose->mode != HS_MODE_STREAM)) {
return HS_DB_MODE_ERROR;
}
size_t stateSize = rose->stateOffsets.end;
struct hs_stream *s = hs_stream_alloc(sizeof(struct hs_stream) + stateSize);
if (unlikely(!s)) {
return HS_NOMEM;
}
init_stream(s, rose);
*stream = s;
return HS_SUCCESS;
}
HS_PUBLIC_API
hs_error_t hs_stream_size(const hs_database_t *db, size_t *stream_size) {
if (!stream_size) {
return HS_INVALID;
}
hs_error_t ret = validDatabase(db);
if (ret != HS_SUCCESS) {
return ret;
}
const struct RoseEngine *rose = hs_get_bytecode(db);
if (!ISALIGNED_16(rose)) {
return HS_INVALID;
}
if (rose->mode != HS_MODE_STREAM) {
return HS_DB_MODE_ERROR;
}
u32 base_stream_size = rose->stateOffsets.end;
// stream state plus the hs_stream struct itself
*stream_size = base_stream_size + sizeof(struct hs_stream);
return HS_SUCCESS;
}
HS_PUBLIC_API
hs_error_t hs_scan_vector(const hs_database_t *db, const char * const * data,
const unsigned int *length, unsigned int count,
UNUSED unsigned int flags, hs_scratch_t *scratch,
match_event_handler onEvent, void *context) {
if (unlikely(!scratch || !data || !length)) {
return HS_INVALID;
}
hs_error_t err = validDatabase(db);
if (unlikely(err != HS_SUCCESS)) {
return err;
}
const struct RoseEngine *rose = hs_get_bytecode(db);
if (unlikely(!ISALIGNED_16(rose))) {
return HS_INVALID;
}
if (unlikely(rose->mode != HS_MODE_VECTORED)) {
return HS_DB_MODE_ERROR;
}
if (unlikely(!validScratch(rose, scratch))) {
return HS_INVALID;
}
hs_stream_t *id = (hs_stream_t *)(scratch->bstate);
init_stream(id, rose); /* open stream */
for (u32 i = 0; i < count; i++) {
DEBUG_PRINTF("block %u/%u offset=%llu len=%u\n", i, count, id->offset,
length[i]);
#ifdef DEBUG
dumpData(data[i], length[i]);
#endif
hs_error_t ret
= hs_scan_stream_internal(id, data[i], length[i], 0, scratch,
onEvent, context);
if (ret != HS_SUCCESS) {
return ret;
}
}
/* close stream */
if (onEvent) {
report_eod_matches(id, scratch, onEvent, context);
if (told_to_stop_matching(scratch)) {
return HS_SCAN_TERMINATED;
}
}
return HS_SUCCESS;
}
HS_PUBLIC_API
hs_error_t hs_alloc_scratch(const hs_database_t *db, hs_scratch_t **scratch) {
if (!db || !scratch) {
return HS_INVALID;
}
/* We need to do some real sanity checks on the database as some users mmap
* in old deserialised databases, so this is the first real opportunity we
* have to make sure it is sane.
*/
hs_error_t rv = dbIsValid(db);
if (rv != HS_SUCCESS) {
return rv;
}
/* We can also sanity-check the scratch parameter: if it points to an
* existing scratch area, that scratch should have valid magic bits. */
if (*scratch != NULL) {
/* has to be aligned before we can do anything with it */
if (!ISALIGNED_CL(*scratch)) {
return HS_INVALID;
}
if ((*scratch)->magic != SCRATCH_MAGIC) {
return HS_INVALID;
}
}
const struct RoseEngine *rose = hs_get_bytecode(db);
int resize = 0;
hs_scratch_t *proto;
hs_scratch_t *proto_tmp = hs_scratch_alloc(sizeof(struct hs_scratch) + 256);
hs_error_t proto_ret = hs_check_alloc(proto_tmp);
if (proto_ret != HS_SUCCESS) {
hs_scratch_free(proto_tmp);
hs_scratch_free(*scratch);
*scratch = NULL;
return proto_ret;
}
proto = ROUNDUP_PTR(proto_tmp, 64);
if (*scratch) {
*proto = **scratch;
} else {
memset(proto, 0, sizeof(*proto));
resize = 1;
}
proto->scratch_alloc = (char *)proto_tmp;
u32 max_anchored_match = rose->anchoredDistance;
if (max_anchored_match > rose->maxSafeAnchoredDROffset) {
u32 anchored_region_len = max_anchored_match
- rose->maxSafeAnchoredDROffset;
if (anchored_region_len > proto->anchored_region_len) {
resize = 1;
proto->anchored_region_len = anchored_region_len;
}
}
u32 anchored_region_width = rose->anchoredMatches;
if (anchored_region_width > proto->anchored_region_width) {
resize = 1;
proto->anchored_region_width = anchored_region_width;
}
if (rose->anchoredDistance > proto->anchored_literal_region_len) {
resize = 1;
proto->anchored_literal_region_len = rose->anchoredDistance;
}
if (rose->anchored_count > proto->anchored_literal_count) {
resize = 1;
proto->anchored_literal_count = rose->anchored_count;
}
if (rose->delay_count > proto->delay_count) {
resize = 1;
proto->delay_count = rose->delay_count;
}
if (rose->roleCount > proto->roleCount) {
resize = 1;
proto->roleCount = rose->roleCount;
}
if (rose->tStateSize > proto->tStateSize) {
resize = 1;
proto->tStateSize = rose->tStateSize;
}
const struct sidecar *side = getSLiteralMatcher(rose);
if (side && sidecarScratchSize(side) > proto->sideScratchSize) {
resize = 1;
proto->sideScratchSize = sidecarScratchSize(side);
}
u32 som_store_count = rose->somLocationCount;
if (som_store_count > proto->som_store_count) {
resize = 1;
proto->som_store_count = som_store_count;
}
u32 queueCount = rose->queueCount;
if (queueCount > proto->queueCount) {
resize = 1;
proto->queueCount = queueCount;
}
u32 bStateSize = 0;
if (rose->mode == HS_MODE_BLOCK) {
bStateSize = rose->stateOffsets.end;
} else if (rose->mode == HS_MODE_VECTORED) {
/* vectoring database require a full stream state (inc header) */
bStateSize = sizeof(struct hs_stream) + rose->stateOffsets.end;
}
if (bStateSize > proto->bStateSize) {
resize = 1;
proto->bStateSize = bStateSize;
}
u32 fullStateSize = rose->scratchStateSize;
if (fullStateSize > proto->fullStateSize) {
resize = 1;
proto->fullStateSize = fullStateSize;
}
if (rose->dkeyCount > proto->deduper.log_size) {
resize = 1;
proto->deduper.log_size = rose->dkeyCount;
}
if (resize) {
if (*scratch) {
hs_scratch_free((*scratch)->scratch_alloc);
}
hs_error_t alloc_ret = alloc_scratch(proto, scratch);
hs_scratch_free(proto_tmp); /* kill off temp used for sizing */
if (alloc_ret != HS_SUCCESS) {
*scratch = NULL;
return alloc_ret;
}
} else {
hs_scratch_free(proto_tmp); /* kill off temp used for sizing */
}
return HS_SUCCESS;
}
HS_PUBLIC_API
hs_error_t hs_scan(const hs_database_t *db, const char *data, unsigned length,
unsigned flags, hs_scratch_t *scratch,
match_event_handler onEvent, void *userCtx) {
if (unlikely(!scratch || !data)) {
return HS_INVALID;
}
hs_error_t err = validDatabase(db);
if (unlikely(err != HS_SUCCESS)) {
return err;
}
const struct RoseEngine *rose = hs_get_bytecode(db);
if (unlikely(!ISALIGNED_16(rose))) {
return HS_INVALID;
}
if (unlikely(rose->mode != HS_MODE_BLOCK)) {
return HS_DB_MODE_ERROR;
}
if (unlikely(!validScratch(rose, scratch))) {
return HS_INVALID;
}
if (rose->minWidth > length) {
DEBUG_PRINTF("minwidth=%u > length=%u\n", rose->minWidth, length);
return HS_SUCCESS;
}
prefetch_data(data, length);
/* populate core info in scratch */
populateCoreInfo(scratch, rose, scratch->bstate, onEvent, userCtx, data,
length, NULL, 0, 0, flags);
clearEvec(scratch->core_info.exhaustionVector, rose);
if (!length) {
if (rose->boundary.reportZeroEodOffset) {
processReportList(rose, rose->boundary.reportZeroEodOffset, 0,
scratch);
}
goto set_retval;
}
if (rose->boundary.reportZeroOffset) {
processReportList(rose, rose->boundary.reportZeroOffset, 0, scratch);
}
if (rose->minWidthExcludingBoundaries > length) {
DEBUG_PRINTF("minWidthExcludingBoundaries=%u > length=%u\n",
rose->minWidthExcludingBoundaries, length);
goto done_scan;
}
// Similarly, we may have a maximum width (for engines constructed entirely
// of bi-anchored patterns).
if (rose->maxBiAnchoredWidth != ROSE_BOUND_INF
&& length > rose->maxBiAnchoredWidth) {
DEBUG_PRINTF("block len=%u longer than maxBAWidth=%u\n", length,
rose->maxBiAnchoredWidth);
goto done_scan;
}
// Is this a small write case?
if (rose->smallWriteOffset) {
const struct SmallWriteEngine *smwr = getSmallWrite(rose);
assert(smwr);
// Apply the small write engine if and only if the block (buffer) is
// small enough. Otherwise, we allow rose &co to deal with it.
if (length < smwr->largestBuffer) {
DEBUG_PRINTF("Attempting small write of block %u bytes long.\n",
length);
runSmallWriteEngine(smwr, scratch);
goto done_scan;
}
}
switch (rose->runtimeImpl) {
default:
assert(0);
case ROSE_RUNTIME_FULL_ROSE:
rawBlockExec(rose, scratch);
break;
case ROSE_RUNTIME_PURE_LITERAL:
pureLiteralBlockExec(rose, scratch);
break;
case ROSE_RUNTIME_SINGLE_OUTFIX:
soleOutfixBlockExec(rose, scratch);
break;
}
done_scan:
if (told_to_stop_matching(scratch)) {
return HS_SCAN_TERMINATED;
}
if (rose->hasSom) {
int halt = flushStoredSomMatches(scratch, ~0ULL);
if (halt) {
return HS_SCAN_TERMINATED;
}
}
if (rose->boundary.reportEodOffset) {
processReportList(rose, rose->boundary.reportEodOffset, length, scratch);
}
set_retval:
DEBUG_PRINTF("done. told_to_stop_matching=%d\n",
told_to_stop_matching(scratch));
return told_to_stop_matching(scratch) ? HS_SCAN_TERMINATED : HS_SUCCESS;
}
