pagemap: Read compressed page images

Existing pagemap readers derive each pages-image offset from a fixed
PAGE_SIZE payload per present page. Packed compressed blocks make that
calculation invalid and must be reconstructed before reaching restore
destinations.

Track block and intra-region cursors while advancing, skipping, and
reading pagemap entries. Validate block counts, decoded sizes, totals,
and aligned-payload offsets before I/O. Read per-page and region images
locally, preserve ordinary parent entries, and support sequential
per-page image streams.

Carry block metadata with queued iovecs, coalesce only compatible
regions, read each packed batch once, and rebuild its destinations block
by block. Allow exactly IOV_MAX destinations and roll back iovec changes
if appending metadata fails. Decompress complete regions directly into
a contiguous destination and use scratch space for partial or split
regions. Copy the metadata into restorer memory for the later PIE decode
path.

Keep compressed images on buffered I/O, avoid hole punching packed
payloads, handle short reads, and keep compressed local reads out of PIE
until that path can decode them.

Assisted-by: Codex:GPT-5
Signed-off-by: Radostin Stoyanov <rstoyanov@fedoraproject.org>

criu/include/pagemap.h

@@ -98,6 +98,23 @@ struct page_read {
 
 	/* Current offset in pages file */
 	off_t pi_off;
+	/* Alignment bytes a sequential image-streamer reader must discard. */
+	size_t stream_padding;
+
+	/*
+	 * Index into pe->compressed_size[] for the current pagemap
+	 * entry. Tracks which compressed block (page in per-page mode,
+	 * region in region mode) we are on when reading or skipping
+	 * compressed pages. Reset to 0 on advance().
+	 */
+	size_t compressed_size_index;
+
+	/*
+	 * In region mode: pages already consumed (read or skipped) from
+	 * the current block. Always 0 in per-page mode. Reset to 0 on
+	 * advance() and whenever the reader crosses a block boundary.
+	 */
+	unsigned int region_block_offset;
 
 	/* Record consequent neighbour iov-ecs to punch together */
 	struct iovec bunch;
@@ -201,4 +218,12 @@ static inline bool pagemap_payload_aligned(PagemapEntry *pe)
 	return !!(pe->flags & PE_PAYLOAD_ALIGNED);
 }
 
+/* Keep the mask as wide as off_t so offsets above 4 GiB remain intact. */
+static inline off_t pagemap_page_align_offset(off_t offset)
+{
+	off_t mask = ~((off_t)PAGE_SIZE - 1);
+
+	return (offset + (off_t)PAGE_SIZE - 1) & mask;
+}
+
 #endif /* __CR_PAGE_READ_H__ */

criu/include/restorer.h

@@ -138,6 +138,20 @@ typedef long (*thread_restore_fcall_t)(struct thread_restore_args *args);
 struct restore_vma_io {
 	int nr_iovs;
 	loff_t off;
+	uint32_t *compressed_size;
+	uint64_t total_compressed_size;
+	int n_compressed_size;
+	/*
+	 * Region compression metadata. region_pages == 0 means per-page
+	 * compression and block_pages is unused. When region_pages > 0,
+	 * compressed_size[] holds n_compressed_size (== n_blocks) entries
+	 * and block_pages is a parallel uint16_t-per-block array giving
+	 * each block's page count (the last block of any pagemap entry
+	 * spanning this iov may be shorter than region_pages).
+	 */
+	uint32_t region_pages;
+	int n_pages;
+	uint16_t *block_pages;
 	struct iovec iovs[0];
 };