qemu-system-x86_64: util/hbitmap.c:614: serialization_chunk: Assertion `(last >> hb->granularity) < hb->size' failed Description of problem: If a named dirty bitmap already exists on a disk and another disk is added via hotplug after the guest has booted, it will definitely cause the hot migration to fail. Steps to reproduce: 1. Create 2 images of type qcow2 ``` qemu-img create -f qcow2 vda.qcow2 50G qemu-img create -f qcow2 vdb.qcow2 2G # set to 2G ``` 2. Start the guest using the following libvirt xml ``` # virsh create i-btacsctt.xml i-btacsctt 973f7352-ad1d-31ea-9a9f-237f3e9a384f 2048 2 hvm /opt/qemu-5.1.0.9/usr/bin/qemu-system-x86_64 ``` 3. Create bitmap for vda ``` # The node name of vda is "libvirt-2-format" virsh qemu-monitor-command i-btacsctt --hmp "info block" libvirt-2-format: /tmp/echohu3/vda.qcow2 (qcow2) Attached to: /machine/peripheral/virtio-disk0/virtio-backend Cache mode: writethrough libvirt-1-format: /tmp/echohu3/vdb.qcow2 (qcow2) Attached to: /machine/peripheral/virtio-disk1/virtio-backend Cache mode: writeback, direct # Create bitmap virsh qemu-monitor-command i-btacsctt '{"execute":"block-dirty-bitmap-add","arguments":{"node":"libvirt-2-format","name":"bitmap0","persistent":true}}' ``` 4. Create vdc and run hotpluggin ``` qemu-img create -f qcow2 vdc.qcow2 50G cat disk.xml virsh attach-device i-btacsctt disk.xml ``` 5. Start live migrationg ``` # scp *.qcow2 172.31.68.42:/tmp/echohu3/ virsh qemu-monitor-command i-btacsctt --hmp "migrate_set_capability dirty-bitmaps on" virsh dumpxml --migratable i-btacsctt >/tmp/ivm-btacsctt.xml virsh migrate --live --abort-on-error --xml /tmp/ivm-btacsctt.xml i-btacsctt qemu+tcp://172.31.68.42/system error: internal error: qemu unexpectedly closed the monitor: qemu-system-x86_64: util/hbitmap.c:614: serialization_chunk: Assertion `(last >> hb->granularity) < hb->size' failed. ``` Additional information: Set breakpoints on the source side ``` gdb -p $pid -ex "break add_bitmaps_to_list" -ex "handle SIGUSR1 nostop" -ex "continue" (gdb) bt #0 add_bitmaps_to_list (bs=bs@entry=0x55c5bbaf85d0, bs_name=0x55c5bbafc674 "libvirt-2-format", alias_map=alias_map@entry=0x0, s=) at migration/block-dirty-bitmap.c:502 #1 0x000055c5ba3b2878 in init_dirty_bitmap_migration (s=0x55c5bb11a080 ) at migration/block-dirty-bitmap.c:660 #2 dirty_bitmap_save_setup (f=0x55c5bc981c40, opaque=0x55c5bb11a080 ) at migration/block-dirty-bitmap.c:1226 #3 0x000055c5ba3a3c4d in qemu_savevm_state_setup (f=0x55c5bc981c40) at migration/savevm.c:1176 #4 0x000055c5ba39e16b in migration_thread (opaque=opaque@entry=0x55c5bbaa2400) at migration/migration.c:3487 #5 0x000055c5ba530cf3 in qemu_thread_start (args=) at util/qemu-thread-posix.c:521 #6 0x00007f39846d9609 in start_thread (arg=) at pthread_create.c:477 #7 0x00007f3983d11293 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95 (gdb) p bs->node_name $4 = "libvirt-2-format", '\000' (gdb) p bitmap->name $5 = 0x55c5bbaf13d0 "bitmap0" ``` Set a breakpoint on the target side after hitting the breakpoint on the source side. ``` gdb -p $pid -ex "break serialization_chunk if ((start + count - 1) >> hb->granularity) >= hb->size" -ex "break dirty_bitmap_load_header" -ex "handle SIGUSR1 nostop" -ex "continue" (gdb) bt #0 dirty_bitmap_load_header (alias_map=0x0, s=0x557488aef0a8 , f=0x55748bcfd8f0) at migration/block-dirty-bitmap.c:1146 #1 dirty_bitmap_load (f=0x55748bcfd8f0, opaque=0x557488aef080 , version_id=) at migration/block-dirty-bitmap.c:1187 #2 0x0000557487d7759a in vmstate_load (se=0x55748adfb8b0, f=0x55748bcfd8f0) at migration/savevm.c:883 #3 vmstate_load (f=0x55748bcfd8f0, se=0x55748adfb8b0) at migration/savevm.c:879 #4 0x0000557487d79fdd in qemu_loadvm_section_part_end (mis=0x55748ad55be0, f=0x55748bcfd8f0) at migration/savevm.c:2365 #5 qemu_loadvm_state_main (f=f@entry=0x55748bcfd8f0, mis=mis@entry=0x55748ad55be0) at migration/savevm.c:2518 #6 0x0000557487d7b2ad in qemu_loadvm_state (f=0x55748bcfd8f0) at migration/savevm.c:2590 #7 0x0000557487d7078f in process_incoming_migration_co (opaque=) at migration/migration.c:480 #8 0x0000557487f15283 in coroutine_trampoline (i0=, i1=) at util/coroutine-ucontext.c:173 #9 0x00007f5360189660 in __start_context () at ../sysdeps/unix/sysv/linux/x86_64/__start_context.S:91 ``` in dirty_bitmap_load_header ``` s->bs = bdrv_lookup_bs(s->node_name, s->node_name, &local_err); // node_name is "libvirt-2-format" s->bitmap = bdrv_find_dirty_bitmap(s->bs, s->bitmap_name); // bitmap_name is "bitmap0" # Target side: “libvirt-2-format” is the node name of vdb. (gdb) p s->bs->node_name $10 = "libvirt-2-format", '\000' (gdb) p s->bs->filename $11 = "/tmp/echohu3/vdb.qcow2", '\000' ``` We can also see from the target /var/log/libvirt/qemu/i-btacsctt.log file that “libvirt-2-format” is the node name of the vdb,while the node name of vda is libvirt-3-format. ``` -blockdev '{"driver":"file","filename":"/tmp/echohu3/vda.qcow2","aio":"threads","node-name":"libvirt-3-storage","cache":{"direct":false,"no-flush":false},"auto-read-only":true,"discard":"unmap"}' \ -blockdev '{"node-name":"libvirt-3-format","read-only":false,"discard":"ignore","cache":{"direct":false,"no-flush":false},"driver":"qcow2","file":"libvirt-3-storage","backing":null}' \ -device virtio-blk-pci,bus=pci.0,addr=0x2,drive=libvirt-3-format,id=virtio-disk0,bootindex=1,write-cache=on \ -blockdev '{"driver":"file","filename":"/tmp/echohu3/vdb.qcow2","aio":"threads","node-name":"libvirt-2-storage","cache":{"direct":true,"no-flush":false},"auto-read-only":true,"discard":"unmap"}' \ -blockdev '{"node-name":"libvirt-2-format","read-only":false,"cache":{"direct":true,"no-flush":false},"driver":"qcow2","file":"libvirt-2-storage","backing":null}' \ -device virtio-blk-pci,bus=pci.0,addr=0x3,drive=libvirt-2-format,id=virtio-disk1,write-cache=on \ -blockdev '{"driver":"file","filename":"/tmp/echohu3/vdc.qcow2","aio":"threads","node-name":"libvirt-1-storage","cache":{"direct":true,"no-flush":false},"auto-read-only":true,"discard":"unmap"}' \ -blockdev '{"node-name":"libvirt-1-format","read-only":false,"discard":"ignore","cache":{"direct":true,"no-flush":false},"driver":"qcow2","file":"libvirt-1-storage","backing":null}' \ ``` From the source code, we know that HBitmap.size is from vdb size (2G), but bitmap is from vda (50G), so it triggers assert exception in serialization_chunk. ``` (gdb) bt #0 serialization_chunk (hb=hb@entry=0x55748ba28470, start=2147483648, count=536870912, first_el=first_el@entry=0x7f53503ffd20, el_count=el_count@entry=0x7f53503ffd18) at util/hbitmap.c:610 #1 0x0000557487f18654 in hbitmap_deserialize_zeroes (hb=0x55748ba28470, start=start@entry=2147483648, count=count@entry=536870912, finish=finish@entry=false) at util/hbitmap.c:701 #2 0x0000557487e7cfb0 in bdrv_dirty_bitmap_deserialize_zeroes (bitmap=, offset=offset@entry=2147483648, bytes=bytes@entry=536870912, finish=finish@entry=false) at block/dirty-bitmap.c:749 #3 0x0000557487d86b51 in dirty_bitmap_load_bits (s=0x557488aef0a8 , f=0x55748bcfd8f0) at migration/block-dirty-bitmap.c:992 #4 dirty_bitmap_load (f=0x55748bcfd8f0, opaque=0x557488aef080 , version_id=) at migration/block-dirty-bitmap.c:1198 #5 0x0000557487d7759a in vmstate_load (se=0x55748adfb8b0, f=0x55748bcfd8f0) at migration/savevm.c:883 #6 vmstate_load (f=0x55748bcfd8f0, se=0x55748adfb8b0) at migration/savevm.c:879 #7 0x0000557487d79fdd in qemu_loadvm_section_part_end (mis=0x55748ad55be0, f=0x55748bcfd8f0) at migration/savevm.c:2365 #8 qemu_loadvm_state_main (f=f@entry=0x55748bcfd8f0, mis=mis@entry=0x55748ad55be0) at migration/savevm.c:2518 #9 0x0000557487d7b2ad in qemu_loadvm_state (f=0x55748bcfd8f0) at migration/savevm.c:2590 #10 0x0000557487d7078f in process_incoming_migration_co (opaque=) at migration/migration.c:480 #11 0x0000557487f15283 in coroutine_trampoline (i0=, i1=) at util/coroutine-ucontext.c:173 #12 0x00007f5360189660 in __start_context () at ../sysdeps/unix/sysv/linux/x86_64/__start_context.S:91 #13 0x00007ffffb29c410 in () #14 0x0000000000000000 in () (gdb) p *hb $16 = {orig_size = 2147483648, size = 32768, count = 0, granularity = 16, meta = 0x0, levels = {0x55748ad55ad0, 0x55748acd8df0, 0x55748b0866a0, 0x55748acf8c10, 0x55748b1c4180, 0x55748b154f60, 0x55748adf2370}, sizes = {1, 1, 1, 1, 1, 8, 512}} ``` ``` (gdb) f 4 #4 dirty_bitmap_load (f=0x55748bcfd8f0, opaque=0x557488aef080 , version_id=) at migration/block-dirty-bitmap.c:1198 (gdb) p *s->bs $21 = {open_flags = 10274, read_only = false, encrypted = false, sg = false, probed = false, force_share = false, implicit = false, drv = 0x557488aa2ee0 , opaque = 0x55748acf8c90, aio_context = 0x55748acd1080, aio_notifiers = {lh_first = 0x0}, walking_aio_notifiers = false, filename = "/tmp/echohu3/vdb.qcow2", '\000' , backing_file = '\000' , auto_backing_file = '\000' , backing_format = '\000' , full_open_options = 0x55748b3c68e0, exact_filename = "/tmp/echohu3/vdb.qcow2", '\000' , backing = 0x0, file = 0x55748aa5de40, bl = {request_alignment = 1, max_pdiscard = 0, pdiscard_alignment = 65536, max_pwrite_zeroes = 0, pwrite_zeroes_alignment = 65536, opt_transfer = 0, max_transfer = 0, min_mem_alignment = 512, opt_mem_alignment = 4096, max_iov = 1024}, supported_write_flags = 0, supported_zero_flags = 260, supported_truncate_flags = 2, node_name = "libvirt-2-format", '\000' , node_list = {tqe_next = 0x55748adeb060, tqe_circ = {tql_next = 0x55748adeb060, tql_prev = 0x55748ad4d0e8}}, bs_list = {tqe_next = 0x55748adeb060, tqe_circ = {tql_next = 0x55748adeb060, tql_prev = 0x55748ad4d0f8}}, monitor_list = {tqe_next = 0x55748adeb060, tqe_circ = {tql_next = 0x55748adeb060, tql_prev = 0x55748ad4d108}}, refcnt = 2, op_blockers = {{lh_first = 0x0} }, inherits_from = 0x0, children = {lh_first = 0x55748aa5de40}, parents = {lh_first = 0x55748bbc0380}, options = 0x55748ad4d2d0, explicit_options = 0x55748ad525a0, detect_zeroes = BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF, backing_blocker = 0x0, total_sectors = 4194304, before_write_notifiers = {notifiers = {lh_first = 0x0}}, write_threshold_offset = 0, write_threshold_notifier = {notify = 0x0, node = { le_next = 0x0, le_prev = 0x0}}, dirty_bitmap_mutex = {lock = {__data = {__lock = 0, __count = 0, __owner = 0, __nusers = 0, __kind = 0, __spins = 0, __elision = 0, __list = {__prev = 0x0, __next = 0x0}}, __size = '\000' , __align = 0}, initialized = true}, dirty_bitmaps = {lh_first = 0x55748b4655f0}, wr_highest_offset = {value = 0}, copy_on_read = 0, in_flight = 0, serialising_in_flight = 0, io_plugged = 0, enable_write_cache = 0, quiesce_counter = 0, recursive_quiesce_counter = 0, write_gen = 0, reqs_lock = {locked = 0, ctx = 0x0, from_push = {slh_first = 0x0}, to_pop = {slh_first = 0x0}, handoff = 0, sequence = 0, holder = 0x0}, tracked_requests = {lh_first = 0x0}, flush_queue = {entries = {sqh_first = 0x0, sqh_last = 0x55748ad52570}}, active_flush_req = false, flushed_gen = 0, never_freeze = false} ``` When we merge into commit https://gitlab.com/qemu-project/qemu/-/commit/31e4c354b38cd42a051ad030eb7779d5e7ee32fe and then run `block-bitmap-mapping` before migration, the hot migration can be completed successfully. I would like to confirm with the community whether this solution is reasonable and if there are any other solutions to address this issue. ``` virsh qemu-monitor-command i-btacsctt '{"execute": "migrate-set-parameters", "arguments":{"block-bitmap-mapping":[{"node-name":"libvirt-2-format", "alias":"libvirt-3-format","bitmaps":[{"name":"bitmap0", "alias":"bitmap0"}]}]}}' ```