Add drive_add/drive_del to qmp_commands.hx
PATCH 1/1 Add drive_add/drive_del to qmp_commands.hx
This patch enables drive_add/drive_del qmp commands, which are required by libvirt when connecting to qemu using QMP.
Signed-off-by Hu Tao <[email protected]>
--- qmp-commands.hx | 59 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 files changed, 59 insertions(+), 0 deletions(-)
qmp-commands.hx +59 -0
@@ -1774,3 +1774,62 @@ Example: EQMP + { + .name = "drive_add", + .args_type = "pci_addr:s,opts:s", + .params = "[[<domain>:]<bus>:]<slot>\n" + "[file=file][,if=type][,bus=n]\n" + "[,unit=m][,media=d][index=i]\n" + "[,cyls=c,heads=h,secs=s[,trans=t]]\n" + "[snapshot=on|off][,cache=on|off]", + + .help = "add drive to PCI storage controller", + .mhandler.cmd = drive_hot_add, + }, + +SQMP +drive_add +--------- + +Add drive to PCI storage controller. + +Arguments: + +- "pci_addr": pci address. +- "opts": options of the drive to be added. + +Example: + +-> { "execute": "drive_add", "arguments": { "pci_addr": "dummy", "opts": "file=/path/to/disk.img, if=none, id=drive-virtio-disk1, format=raw" } } +<- { "return": {} } + +EQMP + + { + .name = "drive_del", + .args_type = "id:s", + .params = "device", + .help = "remove host block device", + .user_print = monitor_user_noop, + .mhandler.cmd_new = do_drive_del, + }, + +SQMP +drive_del +--------- + +Remove host block device. The result is that guest generated IO is no longer +submitted against the host device underlying the disk. Once a drive has +been deleted, the QEMU Block layer returns -EIO which results in IO +errors in the guest for applications that are reading/writing to the device. + +Arguments: + +- "id": id of device to be removed. + +Examples: + +-> { "execute": "drive_del", "arguments": { "id": "drive-virtio-disk1"} } +<- { "return": {} } + +EQMP
Signature · 1 line
1.7.3.1
Quoted 34 lines · Hu Tao
This patch enables drive_add/drive_del qmp commands, which are required by libvirt when connecting to qemu using QMP.
Signed-off-by Hu Tao<[email protected]>--- qmp-commands.hx | 59 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 files changed, 59 insertions(+), 0 deletions(-)qmp-commands.hx +17
-0@@ -1774,3 +1774,62 @@ Example: EQMP + { + .name = "drive_add", + .args_type = "pci_addr:s,opts:s", + .params = "[[<domain>:]<bus>:]<slot>\n" + "[file=file][,if=type][,bus=n]\n" + "[,unit=m][,media=d][index=i]\n" + "[,cyls=c,heads=h,secs=s[,trans=t]]\n" + "[snapshot=on|off][,cache=on|off]", + + .help = "add drive to PCI storage controller", + .mhandler.cmd = drive_hot_add, + }, + +SQMP +drive_add +--------- +
qmp-commands.hx Example: · line 1803
Quoted 10 lines
qmp-commands.hx +10
-0+Add drive to PCI storage controller. + +Arguments: + +- "pci_addr": pci address. +- "opts": options of the drive to be added. + +Example: + +-> { "execute": "drive_add", "arguments": { "pci_addr": "dummy", "opts": "file=/path/to/disk.img, if=none, id=drive-virtio-disk1, format=raw" } }
QMP command should not take encoded arguments.
Regards,
Anthony Liguori
Quoted 33 lines
qmp-commands.hx +32
-0+<- { "return": {} } + +EQMP + + { + .name = "drive_del", + .args_type = "id:s", + .params = "device", + .help = "remove host block device", + .user_print = monitor_user_noop, + .mhandler.cmd_new = do_drive_del, + }, + +SQMP +drive_del +--------- + +Remove host block device. The result is that guest generated IO is no longer +submitted against the host device underlying the disk. Once a drive has +been deleted, the QEMU Block layer returns -EIO which results in IO +errors in the guest for applications that are reading/writing to the device. + +Arguments: + +- "id": id of device to be removed. + +Examples: + +-> { "execute": "drive_del", "arguments": { "id": "drive-virtio-disk1"} } +<- { "return": {} } + +EQMP
Quoted 2 lines · Hu Tao
This patch enables drive_add/drive_del qmp commands, which are required by libvirt when connecting to qemu using QMP.
These commands were intentionally left out of QMP because they are badly designed.
I understand the need for drive hot plug in QMP. I started to work on saner commands some time ago, but too many other things took priority. I still hope to finish the job.
Can you use human monitor passthrough in the meantime?