Created attachment 1797017 [details] libvirtd log Description of problem: As subject Version-Release number of selected component (if applicable): qemu-kvm-6.0.0-7.el9.x86_64 libvirt-7.4.0-1.el9.x86_64 How reproducible: 100% Steps to Reproduce: 1. Start a vm with local storage 2. Pre-create the disk image on target host manually # qemu-img create -f qcow2 /var/lib/avocado/data/avocado-vt/images/jeos-27-x86_64.qcow2 10G # qemu-img info /var/lib/avocado/data/avocado-vt/images/jeos-27-x86_64.qcow2 -U image: /var/lib/avocado/data/avocado-vt/images/jeos-27-x86_64.qcow2 file format: qcow2 virtual size: 10 GiB (10737418240 bytes) disk size: 196 KiB cluster_size: 65536 Format specific information: compat: 1.1 compression type: zlib lazy refcounts: false refcount bits: 16 corrupt: false extended l2: false 2. Migrate vm with --copy-storage-inc # virsh migrate avocado-vt-vm1 qemu+ssh://******/system --live --verbose --copy-storage-inc 3. Check the disk image on dest host, found its size is small # qemu-img info /var/lib/avocado/data/avocado-vt/images/jeos-27-x86_64.qcow2 -U image: /var/lib/avocado/data/avocado-vt/images/jeos-27-x86_64.qcow2 file format: qcow2 virtual size: 10 GiB (10737418240 bytes) disk size: 5.01 MiB cluster_size: 65536 Format specific information: compat: 1.1 compression type: zlib lazy refcounts: false refcount bits: 16 corrupt: false extended l2: false 4. Try to login to vm on dest host, failed 5. Check libvirtd.log, can't find blockdev-add Actual results: As description Expected results: Disk should be copied during migration when --copy-storage-inc is used Additional info: Can't reproduce with --copy-storage-all
Created attachment 1797029 [details] domain xml
Created attachment 1797066 [details] The qmp log of src and dest hosts Reproduced on libvirt-7.4.0-1.el9.x86_64 qemu-kvm-6.0.0-4.el9.x86_64. See the qmp log files generated by qemu-monitor.stp.
The problem happens because a wrong constant was used in the logic expression which is used to determine whether storage migration needs to take place. The original condition is: bool storageMigration = flags & (VIR_MIGRATE_NON_SHARED_DISK | QEMU_MONITOR_MIGRATE_NON_SHARED_INC); The correct one is: bool storageMigration = flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC); QEMU_MONITOR_MIGRATE_NON_SHARED_INC equals to 0x04 VIR_MIGRATE_NON_SHARED_INC equals to 0x80
Fixed upstream: commit b249fa78718cd6c21109b385b568ecd3d6a3a8dd Author: Peter Krempa <pkrempa> Date: Fri Jul 2 14:17:58 2021 +0200 NEWS: Mention implications of the bug in migration code Wrong flag use could have user-visible implications. Mention the fix. Signed-off-by: Peter Krempa <pkrempa> Reviewed-by: Ján Tomko <jtomko> commit f58349c9c6d26d98e7c8c195b1160d0c0cfff080 Author: Peter Krempa <pkrempa> Date: Fri Jul 2 14:17:57 2021 +0200 qemu: migration: Use correct flag constant for enabling storage migration The 'storageMigration' flag is supposed to be true if storage migration is requested, which is based on VIR_MIGRATE_NON_SHARED_DISK or VIR_MIGRATE_NON_SHARED_INC flags. The assignment to the variable used QEMU_MONITOR_MIGRATE_NON_SHARED_INC (0x04) instead of VIR_MIGRATE_NON_SHARED_INC (0x80), caused libvirtd to skip the actual copy of data. Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1978526 Fixes: da69f4b2084bff140238e450e264d6036ebef898 Signed-off-by: Peter Krempa <pkrempa> Reviewed-by: Ján Tomko <jtomko> v7.5.0-44-gb249fa7871
Test on libvirt v7.5.0-97-g133d05a15e and qemu-6.0.0-1.fc35.x86_64 as comment0. PASS
Verified on libvirt-7.6.0-2.el9.x86_64 qemu-kvm-6.0.0-12.el9.x86_64: 1. Prepare a nbd based on an image with OS 2. Create backing file of nbd on both hosts # qemu-img info /var/lib/libvirt/images/backing.qcow2 image: /var/lib/libvirt/images/backing.qcow2 file format: qcow2 virtual size: 10 GiB (10737418240 bytes) disk size: 26 MiB cluster_size: 65536 backing file: json:{"file":{"driver":"nbd","server":{"type":"inet","host":"10.0.150.247","port":"10809"}}} backing file format: qcow2 Format specific information: compat: 1.1 compression type: zlib lazy refcounts: false refcount bits: 16 corrupt: false extended l2: false 3. Refresh the storage pool of backing image 4. Start a VM with backing image. Finish migration with --copy-storage-inc # virsh migrate fedora34 qemu+ssh://root.150.247/system --live --verbose --copy-storage-inc Migration: [100 %] 5. Login the VM and write some data to the disk # dd if=/dev/zero of=file bs=10M count=10 10+0 records in 10+0 records out 104857600 bytes (105 MB, 100 MiB) copied, 2.77315 s, 37.8 MB/s No I/O error.
Covered by RHEL-120247