Bug 1657637

+++ This bug was initially created as a clone of Bug #1631615 +++

Description of problem:
 werror's auto mode can't work normally when start guest with -blockdev

Version-Release number of selected component (if applicable):
  kernel version: 3.10.0-945.el7.x86_64
  qemu-kvm-rhev version: qemu-kvm-rhev-2.12.0-17.el7.x86_64

How reproducible:
100%

Steps to Reproduce:
1. Install a guest with gluster backend.

2. Full write gluster backend to make it "No space left on device"

3. Start vm with -blockdev:
    /usr/libexec/qemu-kvm \
    -name 'avocado-vt-vm1'  \
    -sandbox off  \
    -machine pc  \
    -nodefaults \
    -device VGA,bus=pci.0,addr=0x2  \
    -chardev socket,id=qmp_id_qmpmonitor1,path=/var/tmp/monitor-qmpmonitor1-20180910-021412-u4bPHcZI,server,nowait \
    -mon chardev=qmp_id_qmpmonitor1,mode=control  \
    -chardev socket,id=qmp_id_catch_monitor,path=/var/tmp/monitor-catch_monitor-20180910-021412-u4bPHcZI,server,nowait \
    -mon chardev=qmp_id_catch_monitor,mode=control \
    -device pvpanic,ioport=0x505,id=idDQoT2q  \
    -chardev socket,id=serial_id_serial0,path=/var/tmp/serial-serial0-20180910-021412-u4bPHcZI,server,nowait \
    -device isa-serial,chardev=serial_id_serial0  \
    -chardev socket,id=seabioslog_id_20180910-021412-u4bPHcZI,path=/var/tmp/seabios-20180910-021412-u4bPHcZI,server,nowait \
    -device isa-debugcon,chardev=seabioslog_id_20180910-021412-u4bPHcZI,iobase=0x402 \
    -device nec-usb-xhci,id=usb1,bus=pci.0,addr=0x3 \
    -object secret,id=sec0,data=backing \
    -blockdev driver=qcow2,file.driver=gluster,node-name=drive_image1,cache.no-flush=on,cache.direct=off,encrypt.key-secret=sec0,encrypt.format=luks,file.server.0.type=inet,file.server.0.host=ibm-x3650m5-07.lab.eng.pek2.redhat.com,file.server.0.port=24007,file.volume=aliang,file.path=rhel76-64-virtio.qcow2.25 \
    -device virtio-blk-pci,id=image1,drive=drive_image1,bus=pci.0,addr=0x4,scsi=off,serial=0xdafafafafafaf12121212121212121212121212121212121212121212aaafffffffff,physical_block_size=4096,logical_block_size=512,disable-modern=on,disable-legacy=off \
    -device virtio-net-pci,mac=9a:5b:5c:5d:5e:5f,id=idqHhiSX,vectors=4,netdev=idLUiEfL,bus=pci.0,addr=0x5  \
    -netdev tap,id=idLUiEfL,vhost=on \
    -m 11264  \
    -blockdev driver=raw,file.driver=file,node-name=drive_cd1,cache.no-flush=on,cache.direct=off,file.filename=/home/kvm_autotest_root/iso/linux/RHEL7.6-Server-x86_64.iso,read-only=on \
    -device ide-cd,id=cd1,drive=drive_cd1 \
    -smp 2,maxcpus=2,cores=1,threads=1,sockets=2  \
    -cpu Penryn \
    -device usb-tablet,id=usb-tablet1,bus=usb1.0,port=1  \
    -vnc :0  \
    -rtc base=utc,clock=host,driftfix=slew  \
    -boot menu=on,strict=off,order=cdn  \
    -enable-kvm \
    -monitor stdio \
    -blockdev driver=raw,file.driver=iscsi,node-name=drive_data2,cache.no-flush=on,cache.direct=off,file.transport=tcp,file.portal=10.73.224.153,file.target=iqn.2018-09.com.example:t1,file.lun=1 \
    -device virtio-blk-pci,id=data2,drive=drive_data2,bus=pci.0,addr=0x07 \
    -blockdev driver=qcow2,file.driver=file,node-name=drive_data1,cache.no-flush=on,cache.direct=off,file.filename=/home/data.qcow2 \
    -device virtio-blk-pci,id=data1,drive=drive_data1,bus=pci.0 \
    -blockdev driver=raw,node-name=drive_data0,cache.no-flush=on,cache.direct=off,file.filename=/dev/disk/by-path/ip-10.73.224.153:3260-iscsi-iqn.2018-09.com.example:t1-lun-2,file.driver=host_device \
    -device virtio-blk-pci,bus=pci.0,id=data0,drive=drive_data0,scsi=on,disable-modern=on \
    -qmp tcp:0:3000,server,nowait \

4. Start some apps on guest, and check vm status by qmp monitor
   (qmp)#nc -U /var/tmp/monitor-qmpmonitor1-20180910-021412-u4bPHcZI
{"execute":"qmp_capabilities"}
{"timestamp": {"seconds": 1537499256, "microseconds": 309834}, "event": "BLOCK_IO_ERROR", "data": {"device": "", "__com.redhat_reason": "enospc", "node-name": "drive_image1", "reason": "No space left on device", "operation": "write", "action": "report"}}

5. Shutdown vm, then restart it with -drive:
    ...
    -drive if=none,format=qcow2,id=drive_image1,cache=none,encrypt.key-secret=sec0,encrypt.format=luks,file=gluster://ibm-x3650m5-07.lab.eng.pek2.redhat.com/aliang/rhel76-64-virtio.qcow2.25 \
    -device virtio-blk-pci,id=image1,drive=drive_image1,bus=pci.0,addr=0x4,scsi=off,serial=0xdafafafafafaf12121212121212121212121212121212121212121212aaafffffffff,physical_block_size=4096,logical_block_size=512,disable-modern=on,disable-legacy=off \
    ...

6. Start some apps on guest, and check vm status by qmp monitor:
   (qmp)#nc -U /var/tmp/monitor-qmpmonitor1-20180910-021412-u4bPHcZI
{"execute":"qmp_capabilities"}
{"timestamp": {"seconds": 1537498212, "microseconds": 218441}, "event": "BLOCK_IO_ERROR", "data": {"device": "drive_image1", "nospace": true, "__com.redhat_reason": "enospc", "node-name": "#block537", "reason": "No space left on device", "operation": "write", "action": "stop"}}
{"timestamp": {"seconds": 1537498212, "microseconds": 219006}, "event": "BLOCK_IO_ERROR", "data": {"device": "drive_image1", "nospace": false, "__com.redhat_reason": "eother", "node-name": "#block537", "reason": "No medium found", "operation": "write", "action": "report"}}
{"timestamp": {"seconds": 1537498212, "microseconds": 219960}, "event": "STOP"}

7. Shutdown vm, then start it with -blockdev+werror=enospc:
   ....
   -blockdev driver=qcow2,file.driver=gluster,node-name=drive_image1,cache.no-flush=on,cache.direct=off,encrypt.key-secret=sec0,encrypt.format=luks,file.server.0.type=inet,file.server.0.host=ibm-x3650m5-07.lab.eng.pek2.redhat.com,file.server.0.port=24007,file.volume=aliang,file.path=rhel76-64-virtio.qcow2.25 \
    -device virtio-blk-pci,id=image1,drive=drive_image1,werror=enospc,bus=pci.0,addr=0x4,scsi=off,serial=0xdafafafafafaf12121212121212121212121212121212121212121212aaafffffffff,physical_block_size=4096,logical_block_size=512,disable-modern=on,disable-legacy=off \
   ....

8. Start some apps on guest, and check vm status by qmp monitor:
    (qmp)#nc -U /var/tmp/monitor-qmpmonitor1-20180910-021412-u4bPHcZI
{"execute":"qmp_capabilities"}
{"timestamp": {"seconds": 1537498212, "microseconds": 218441}, "event": "BLOCK_IO_ERROR", "data": {"device": "drive_image1", "nospace": true, "__com.redhat_reason": "enospc", "node-name": "#block537", "reason": "No space left on device", "operation": "write", "action": "stop"}}
{"timestamp": {"seconds": 1537498212, "microseconds": 219960}, "event": "STOP"}
    
Actual results:
 Both -drive and -blockdev+werror=enospc work normally when "No space left on device" with qmp event: 
    {"timestamp": {"seconds": 1537498212, "microseconds": 218441}, "event": "BLOCK_IO_ERROR", "data": {"device": "drive_image1", "nospace": true, "__com.redhat_reason": "enospc", "node-name": "#block537", "reason": "No space left on device", "operation": "write", "action": "stop"}}

 -blockdev+werror=auto can't work normally when "No space left on device" with qmp event:
     {"timestamp": {"seconds": 1537501432, "microseconds": 641823}, "event": "BLOCK_IO_ERROR", "data": {"device": "", "__com.redhat_reason": "enospc", "node-name": "drive_image1", "reason": "No space left on device", "operation": "write", "action": "report"}}

Expected results:

 -blockdev+werror=auto can work normally.

Additional info:
  It's not a regression bug. Test on qemu-kvm-rhev-2.12.0-1.el7.x86_64 and qemu-kvm-rhev-2.10.0-21.el7_5.7.x86_64, both hit this issue.

--- Additional comment from Kevin Wolf on 2018-09-28 10:58:26 CEST ---

To be precise, the bug is with -device ...,drive=<node-name>. This can be reproduced with -drive as well if the node name is used to create the device.

The fix is simple, blk_new() should set the right defaults instead of relying on other code paths doing that.

--- Additional comment from Tingting Mao on 2018-09-28 13:23:23 CEST ---

The command could be like below for '-drive & -device' option(Thank Kevin and Miya).


/usr/libexec/qemu-kvm \

      ...
      -drive node-name=image_drive1,if=none,snapshot=off,aio=threads,cache=none,format=qcow2,file=base.qcow2 \
      -device virtio-blk-pci,drive=image_drive1,bus=pci.0,addr=05 \
      ...

--- Additional comment from aihua liang on 2018-09-30 08:39:11 CEST ---

Yes, tried with cmd: 
  ...
  -drive if=none,format=qcow2,node-name=drive_image1,cache=none,encrypt.key-secret=sec0,encrypt.format=luks,file=gluster://ibm-x3650m5-07.lab.eng.pek2.redhat.com/aliang/rhel76-64-virtio.qcow2.25 \
  -device virtio-blk-pci,id=image1,drive=drive_image1,bus=pci.0,addr=0x4,scsi=off,serial=0xdafafafafafaf12121212121212121212121212121212121212121212aaafffffffff,physical_block_size=4096,logical_block_size=512,disable-modern=on,disable-legacy=off
  ...

can also hit this issue.

--- Additional comment from Kevin Wolf on 2018-10-01 11:26:45 CEST ---

Peter, is this a problem for libvirt?

--- Additional comment from Peter Krempa on 2018-10-01 12:52:51 CEST ---

We specify the policies explicitly only if the user set a specific one, so without the user specifying anything the behaviour will change. I think that the libvirt API contract allows this as we specify that the "hypervisor default" will be used.

On the other hand it would be better if the behaviour did not change when switching to -blockdev.