block processor: Re-implement exact fragment matching

In the hash-table equals callback, if the hash and size match, do an
exact, byte-for-byte comparison of the fragment in question. The
fragment can either be in a fragment block that is in-flight (for which
we have the in-flight list), in the current, unfinished fragment block,
or it can be on disk.

In the later case, the fragment block is resolved through the fragment
table and read back from disk into a scratch buffer and decompressed.
After that, the fragment is checked for byte-for-byte equality with
the one we resolved through the hash table.

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

1 files changed, 106 insertions, 4 deletions

diff --git a/lib/sqfs/block_processor/block_processor.c b/lib/sqfs/block_processor/block_processor.c
index 19a538a..8303dce 100644
--- a/lib/sqfs/block_processor/block_processor.c
+++ b/lib/sqfs/block_processor/block_processor.c
@@ -44,11 +44,109 @@ static int process_block(void *userptr, void *workitem)
 	return 0;
 }
 
+static int load_frag_block(sqfs_block_processor_t *proc, sqfs_u32 index)
+{
+	sqfs_fragment_t info;
+	size_t size;
+	int ret;
+
+	if (proc->cached_frag_blk == NULL) {
+		size = sizeof(*proc->cached_frag_blk);
+
+		proc->cached_frag_blk = alloc_flex(size, 1,
+						   proc->max_block_size);
+
+		if (proc->cached_frag_blk == NULL)
+			return SQFS_ERROR_ALLOC;
+	} else {
+		if (proc->cached_frag_blk->index == index)
+			return 0;
+	}
+
+	ret = sqfs_frag_table_lookup(proc->frag_tbl, index, &info);
+	if (ret != 0)
+		return ret;
+
+	size = SQFS_ON_DISK_BLOCK_SIZE(info.size);
+	if (size >= proc->max_block_size)
+		return SQFS_ERROR_CORRUPTED;
+
+	if (SQFS_IS_BLOCK_COMPRESSED(info.size)) {
+		ret = proc->file->read_at(proc->file, info.start_offset,
+					  proc->scratch, size);
+		if (ret != 0)
+			return ret;
+
+		ret = proc->uncmp->do_block(proc->uncmp, proc->scratch, size,
+					    proc->cached_frag_blk->data,
+					    proc->max_block_size);
+		if (ret <= 0)
+			return ret ? ret : SQFS_ERROR_OVERFLOW;
+
+		size = ret;
+	} else {
+		ret = proc->file->read_at(proc->file, info.start_offset,
+					  proc->cached_frag_blk->data, size);
+		if (ret != 0)
+			return ret;
+	}
+
+	proc->cached_frag_blk->size = size;
+	proc->cached_frag_blk->index = index;
+	return 0;
+}
+
 static bool chunk_info_equals(void *user, const void *k, const void *c)
 {
 	const chunk_info_t *key = k, *cmp = c;
-	(void)user;
-	return key->size == cmp->size && key->hash == cmp->hash;
+	sqfs_block_processor_t *proc = user;
+	sqfs_block_t *it;
+	int ret;
+
+	if (key->size != cmp->size || key->hash != cmp->hash)
+		return false;
+
+	if (proc->uncmp == NULL || proc->file == NULL)
+		return true;
+
+	if (proc->current_frag == NULL || proc->frag_tbl == NULL)
+		return true;
+
+	if (proc->fblk_lookup_error != 0)
+		return false;
+
+	for (it = proc->fblk_in_flight; it != NULL; it = it->next) {
+		if (it->index == cmp->index)
+			break;
+	}
+
+	if (it == NULL && proc->frag_block != NULL) {
+		if (proc->frag_block->index == cmp->index)
+			it = proc->frag_block;
+	}
+
+	if (it == NULL) {
+		ret = load_frag_block(proc, cmp->index);
+		if (ret != 0) {
+			proc->fblk_lookup_error = ret;
+			return false;
+		}
+
+		it = proc->cached_frag_blk;
+	}
+
+	if (cmp->offset >= it->size || (it->size - cmp->offset) < cmp->size) {
+		proc->fblk_lookup_error = SQFS_ERROR_CORRUPTED;
+		return false;
+	}
+
+	if (cmp->size != proc->current_frag->size) {
+		proc->fblk_lookup_error = SQFS_ERROR_CORRUPTED;
+		return false;
+	}
+
+	return memcmp(it->data + cmp->offset,
+		      proc->current_frag->data, cmp->size) == 0;
 }
 
 static void ht_delete_function(struct hash_entry *entry)
@@ -71,6 +169,7 @@ static void block_processor_destroy(sqfs_object_t *base)
 
 	free(proc->frag_block);
 	free(proc->blk_current);
+	free(proc->cached_frag_blk);
 
 	free_block_list(proc->free_list);
 	free_block_list(proc->io_queue);
@@ -155,14 +254,17 @@ const sqfs_block_processor_stats_t
 int sqfs_block_processor_create_ex(const sqfs_block_processor_desc_t *desc,
 				   sqfs_block_processor_t **out)
 {
+	size_t i, count, scratch_size = 0;
 	sqfs_block_processor_t *proc;
-	size_t i, count;
 	int ret;
 
 	if (desc->size != sizeof(sqfs_block_processor_desc_t))
 		return SQFS_ERROR_ARG_INVALID;
 
-	proc = calloc(1, sizeof(*proc));
+	if (desc->file != NULL && desc->uncmp != NULL)
+		scratch_size = desc->max_block_size;
+
+	proc = alloc_flex(sizeof(*proc), 1, scratch_size);
 	if (proc == NULL)
 		return SQFS_ERROR_ALLOC;