QEMU migration pull adds fast snapshot load and RDMA hardening

QEMU's migration maintainers have staged a pull that adds faster snapshot restore, closes CVE-2026-6426, and hardens RDMA and vmstate paths against bad wire input.

The main user-facing change is Fast Snapshot Load, from Aadeshveer Singh. On restore from a mapped-ram snapshot, QEMU can now fault and prefetch guest pages from disk in parallel with startup instead of blocking until the full image is resident, so large snapshots become usable sooner. The work includes capability conflict handling with postcopy and new tests and docs.

Security work is the other focus. Peter Xu added destination-side sanity checks in RDMA migration after reports from Tristan: registration and compress control messages must now match expected buffer sizes, chunk counts, and address ranges inside the target RAM block. Previously, malformed values could slip through and produce undersized or out-of-range pointers. Fabiano Rosas, who sent the pull to the QEMU development list, separately hardened generic vmstate load against multiply overflow and failed huge allocations when sizes and element counts come from the migration stream, and simplified the surrounding macros and flag checks.

CVE-2026-6426 is fixed in the same series. The pull also stops a page-request tracking leak that could hang postcopy when vhost-user shared pages were involved.