device: 0.969
graphic: 0.949
virtual: 0.918
performance: 0.879
files: 0.869
architecture: 0.829
hypervisor: 0.819
debug: 0.816
PID: 0.792
KVM: 0.756
ppc: 0.736
x86: 0.732
kernel: 0.725
i386: 0.725
boot: 0.690
semantic: 0.669
risc-v: 0.659
network: 0.658
arm: 0.657
peripherals: 0.656
vnc: 0.651
permissions: 0.648
VMM: 0.645
mistranslation: 0.645
register: 0.636
socket: 0.617
TCG: 0.611
user-level: 0.589
assembly: 0.505

qcow2: disk size exceeds virtual size
Description of problem:
Disk size of qcow2 image file exceeds its virtual size after repeatedly writing, and deleting data in qemu vm.
Steps to reproduce:
1. qemu-img create -f qcow2 tmp.qcow2 32M
2. attach tmp.qcow2 as a device to qemu vm
3. mount the device in qemu vm, and repeatedly writing, and deleting data
Additional information:
xml for attaching tmp.qcow2
```xml
   <disk device="disk" type="file">
      <target bus="virtio" dev="vdb"/>
      <source file="/path/to/tmp.qcow2"/>
      <driver type="qcow2" name="qemu" cache="none" discard="unmap"/>
   </disk>
```
in fact, set discard="unmap" or not seems has `little impact` on the final result.
reproducible shell script.
```sh
#! /bin/sh

for i in {1..1000}; do
    for j in {1..27}; do
        dd if=/dev/zero of=/mnt/test-$j bs=1M count=1 &
    done
    sync
    sleep 10
    rm -f /mnt/test-*   
    fstrim /mnt
done
```
MOUNT the device and run this script, problem happens about 30 minutes.

final result looks like:
```sh
# qemu-img info tmp.qcow2 --force
image: tmp.qcow2
file format: qcow2
virtual size: 32 MiB (33554432 bytes)
disk size: 33 MiB
cluster_size: 65536
Format specific information:
    compat: 1.1
    compression type: zlib
    lazy refcounts: false
    refcount bits: 16
    corrupt: false
    extended l2: false
Child node '/file':
    filename: tmp.qcow2
    protocol type: file
    file length: 32.3 MiB (33882112 bytes)
    disk size: 33 MiB
    Format specific information:
        extent size hint: 1048576
```