add them by default for qemu 2.6
(support is already present in qemu 2.5, but we don't want to break live migration for current running vm)
vpindex && runtime need host kernel 4.4
Theses 3 enlightements are needed by windows to use vmbus
http://searchwindowsserver.techtarget.com/definition/Microsoft-Virtual-Machine-Bus-VMBus
details :
- When Hyper-V "vpindex" is on, guest can use MSR HV_X64_MSR_VP_INDEX
to get virtual processor ID.
- Hyper-V "runtime" enlightement feature allows to use MSR
HV_X64_MSR_VP_RUNTIME to get the time the virtual processor consumes
running guest code, as well as the time the hypervisor spends running
code on behalf of that guest.
- Hyper-V "reset" allows guest to reset VM.
Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
We cannot guarantee when the SSH forward Tunnel really becomes
ready. The check with the mtunnel API call did not help for this
prolem as it only checked that the SSH connection itself works and
that the destination node has quorum but the forwarded tunnel itself
was not checked.
The Forward tunnel is a different channel in the SSH connection,
independent of the SSH `qm mtunnel` channel, so only if that works
it does not guarantees that our migration tunnel is up and ready.
When the node(s) where under load, or when we did parallel
migrations (migrateall), the migrate command was often started
before a tunnel was open and ready to receive data. This led to
a direct abortion of the migration and is the main cause in why
parallel migrations often leave two thirds or more VMs on the
source node.
The issue was tracked down to SSH after debugging the QEMU
process and enabling debug logging showed that the tunnel became
often to late available and ready, or not at all.
Fixing the TCP forward tunnel is quirky and not straight ahead, the
only way SSH gives as a possibility is to use -N (no command)
-f (background) and -o "ExitOnForwardFailure=yes", then it would
wait in the foreground until the tunnel is ready and only then
background itself. This is not quite the nicest way for our special
use case and our code base.
Waiting for the local port to become open and ready (through
/proc/net/tcp[6]] as a proof of concept is not enough, even if the
port is in the listening state and should theoretically accept
connections this still failed often as the tunnel was not yet fully
ready.
Further another problem would still be open if we tried to patch the
SSH Forward method we currently use - which we solve for free with
the approach of this patch - namely the problem that the method
to get an available port (next_migration_port) has a serious race
condition which could lead to multiple use of the same port on a
parallel migration (I observed this on my many test, seldom but if
it happens its really bad).
So lets now use UNIX sockets, which ssh supports since version 5.7.
The end points are UNIX socket bound to the VMID - thus no port so
no race and also no limitation of available ports (we reserved 50 for
migration).
The endpoints get created in /run/qemu-server/VMID.migrate and as
KVM/QEMU in current versions is able to use UNIX socket just as well
as TCP we have not to change much on the interaction with QEMU.
QEMU is started with the migrate_incoming url at the local
destination endpoint and creates the socket file, we then create
a listening socket on the source side and connect over SSH to the
destination.
Now the migration can be started by issuing the migrate qmp command
with an updated uri.
This breaks live migration from new to old, but *not* from old to
new, so there is a upgrade path.
If a live migration from new to old must be made (for whatever
reason), use the unsecure_migration setting (man datacenter.conf)
to allow this, although that should only be done in trusted network.
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
With systemd-run qemu's --daemonize forks often happen
before systemd finishes setting up the scopes, which means
the limits we apply often don't work.
We now use enter_systemd_scope() to create the scope before
running qemu directly without systemd-run.
Note that vm_start() runs in a forked-worker or qm cli
command, so entering the scope in such a process should not
affect the rest of the pve daemon.
if we got an option which was not valid, we still
wrote it to the config, and subsequently returned
it on every api call
instead, now we die instead of warn and do not accept
invalid options
Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
otherwise, long kvm commands lead to systemd unit files with
very long lines, with confuses the systemd unit file parser.
apparently systemd has a length limit for unit file lines and
(line-)breaks the description string at that point. since
the rest of the description is probably not a valid key/value
pair, this leads to warnings. the default semantics of systemd-run
is to use the executed command as description unless a description
is specified explicitly.
note that this behaviour of systemd could allow an attacker
with access to the VM configuration to craft a kvm commandline
that starts or stops arbitrary systemd units.
previously, we did not check the file parameter of a disk,
allowing passthrough of a block device (by design)
with the change to the json parser for the disks, the format
became 'pve-volume-id' which is only valid for our volume ids
(and later we also allowed the value 'none')
this patch alternatively checks if the parameter is a path
or 'cdrom'
Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
Otherwise some move operations will fail to delete the old
disk (eg. when moving from ceph to local storage).
Note that in order for the deactivation to succeed we need
to make sure qemu has closed its file descriptors, so we
need to wait for the job to disappear the same way we do in
$cancel_job().
Factored the waiting out into $finish_job().
Additionally since the cpu and host node list isn't
restricted to a single range one can now provide multipel
ranges separated by semicolons. (eg. cpus=0-3;5;7)
The urlencoded format currently cannot check the real
decoded length, so we limit to an upper bound and document
the real limits. Ideally we'd introduce a decodedLength
schema parameter at some point...
/cirrur/cirrus/
/devive/device/
/Numa/NUMA/
and a few grammar fixes, rewrites of sentences
Also if already touching those lines lets break them up from one
liners to a column limit of ~80.
When using OVS tap_plug() resets rate limiting so we need
to pass it along to reapply it.
The rate on its own can still be hot-plugged with the
regular tap_rate_limit() call.
Drop snapshot_create, snapshot_delete and snapshot_rollback
in favour of PVE::AbstractConfig. Qemu-specific parts are
implemented in __snapshot_XX methods in PVE::QemuConfig.
has_feature is made an implementation of the abstract
has_feature, and thus moves to PVE::QemuConfig.
Note: a new hook method needed to be introduced to be called
before creating a volume snapshot, after creating a volume
snapshot, and after unfreezing the guestfs after creating a
volume snapshot. The base method in PVE::AbstractConfig is a
noop, the implemention in PVE::QemuConfig runs the necessary
Qemu monitor commands.
Drop load_config, write_config, lock_config[_xx],
check_lock, check_protection, is_template and config_file
in favour of implementions in PVE::AbstractConfig.
Implement guest_type, __config_max_unused_disks,
config_file_lock and cfs_config_path from
PVE::AbstractConfig in PVE::QemuConfig.
Previously, foreach_drive iterated over all configuration
keys (in a random order) and checked whether the current key
is a valid drive name. Instead, we now iterate over a list
of valid drive names (with deterministic order) and check
whether a drive with such a name exists in the
configuration.
Also rename the two involved methods from valid_drive_name
to is_valid_drive_name (for the check) and from disknames
to valid_drive_names (for the list of valid keys), for
consistency. These two were only used in the qemu-server
code base.
We hold a lock from snapshot_prepare until snapshot_commit,
so there is no need to copy back the snapshot config to the
actual config. This allows to drop a workaround for not
copying the 'machine' type config option.
We don't have any storage types other than LVM which react
to scsi inquiry, and we don't want to treat LVM as a scsi
device, so now we only query devices added as actual /dev
path. This was originally intended to be a pass-through
feature anyway, so this makes sense.
As there the signleton function "kvm_user_version" may not have been
called and with the machine alias q35 the regex from the
qemu_machine_feature_enabled method does not match and thus we
need a valid kvm version here
instead of hardcoding the storagetypes for writing zeros on a
backup restore, we use volume_has_feature with 'sparseinit'
for determining if we can omit writing zeros
Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
Currently migration is broken, because qemu_machine_pxe return nothing if no pxe rom exist.
That mean that we don't pass -machine flag to migration, and migration is broken between qemu 2.4->2.5
Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
qemu 2.5 support a new hyper-v feature: hv_vendor_id
This allow nvidia drivers to install on windows with hyper-v feature on.
Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
currently we leave orphaned vmstate files when we restore a
backup over a vm, which has snapshots with saved ram state.
this patch deletes those files on a restore.
Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
Since write_config was always called with skiplock=1 except
once, it makes sense to drop this parameter like in
PVE::LXC::write_config . If needed in the future, the
caller can use check_lock before write_config anyway.
The method update_config wrapped update_config_nolock
using lock_config, but to prevent update races the whole
"read config", "do something", "write config" flow was
always protected by lock_config anyway, and update_config
was never called.
Thus, we can safely drop update_config and rename
update_config_nolock to write_config like in PVE::LXC .
since we want the usb3 option to be really boolean and not only
'usb3=yes', we have to change the usb json format a little
to not break existing configs for 'usbX: spice', we set the 'host'
option as non-optional and default_key and allow 'spice' as its
content (this also makes the option less ambiguous)
another side effect is that previously accepted multiple 'host='
entries are now forbidden
Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
adding a flag for usb devices (usb3), if this is set to yes,
add a xhci controller and attach the specified devices to it
Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
The API passes $skiplock to vm_destroy() which performed a
check conditionally depending on the $skiplock parameter and
then simply calls destroy_vm() inside lock_config() which
did yet another check_lock() without any way to avoid that.
Added the $skiplock parameter to destroy_vm() and removed
the conditional check in vm_destroy() as both happened after
locking the config.
This add support for net trunks vlan filtering
for ovs and linux vlan-aware bridge
Can be mixed with current "tag" option
examples:
----------
allow only 802.1Q packets with vlanid 2,3,4 :
netx: .....,trunks=2,3,4
allow only 802.1Q packets with vlanid 2,3,4 and tag non-802.1Q packets to vlanid 5 :
netx: tag=5,trunks=2,3,4
tag non-802.1Q packets to vlanid 5
netx: tag=5
Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
x-vga vfio-pci flag is to enable seabios quirks only.
This patch keep using x-vga=on from proxmox config, to disable hyperv,kvm=off,vga=none by default
but don't pass x-vga to vfio-pci when ovmf is enabled.
Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
On some storages BLKZEROOUT commands do not work properly
and return without error while having no effect whatsoever.
This can produce various filesystem errors and thus needs
to be made optional.
A drive can now have 'detect_zeroes=off' to disable this
behavior. By default the behavior is the same as before:
always-on (and set to 'unmap' if discard is enabled).
get_used_paths returned a hash of used paths for all the
volumes in a VM's config, which is not enough to figure out
whether there are snapshots, as snapshots often have
different paths. Eg. on ZFS it is not enough to check for
/dev/zvol/tank/vm-123-disk-1 because the snapshot's path is
/dev/zvol/tank/vm-123-disk-1@snap1 and thus we allowed
deleting the drive. Then when trying to delete the snapshot
later you get:
zfs error: cannot open 'tank/vm-751-disk-1': dataset does not exist
and it refuses to delete the snapshot.
Since its only use was to check whether or not a drive is
still in use it is now renamed to is_volume_in_use and
beside checking paths now also checks volume-ids as those
should stay the same.
use it for nic hotplug, because pve-bridge script will
not work after a live migration, because of the PVE_MIGRATED_FROM env var.
Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
Users have reported resume bug when HA is used.
They seem to have a little race (bench show >0s < 1s) between the vm conf file move on source node and replication to,
and resume on target node.
I don't known why this is only with HA, maybe this occur will standard migration too.
Anyway, we don't need to read the vm config file to resume the vm on target host,
as we are sure that the vm is migrated, and config file move action is correct in the cluster.
Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
used to prevent an unintended virtual machine remove operation
v3 changes:
- changed man page message
- removed protection parameter (where not needed)
The -force flag didn't have any effect since the pending
changes didn't carry over the the flag.
Now forced deletes have an exclamation mark prepended to the
option name.
qemu 2.4 feature
changelog: rebase on last git
Note that currently linux guest don't support unplug of dimm when it'sused by kernel memory.
They are some tunning to do with memory zone movable.
http://events.linuxfoundation.org/sites/events/files/lcjp13_chen.pdf
Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
Changed from old, now missing, subroutine parse_startup() to new
pve_parse_startup_order() in qemu-server and pve-manager
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
Currently enforce with cpumodel=host on amd cpus don't work,
because amd cpus have unsupported flags in qemu.
This is a protection, and this is good.
but cpumodel host should be never use by users for production (only for testing).
For production and stability, users need to choose a true cpu model which filter
the supported cpuflags by qemu.
So I think we can remove the enforce for host model as for testing it's ok.
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 0]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 1]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 2]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 3]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 4]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 5]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 6]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 7]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 8]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 9]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 12]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 13]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 14]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 15]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 16]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 17]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 23]
warning: host doesn't support requested feature: CPUID.80000001H:EDX [bit 24]
Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
rdtscp is not supported by qemu and with enforce it's not starting
warning: host doesn't support requested feature: CPUID.80000001H:EDX.rdtscp [bit 27]
from to qemu wiki
http://wiki.qemu.org/Features/CPUModels#Disabling_features_that_were_always_disabled_on_KVM
"Fact: currently libvirt runs CPU models having rdtscp without the "enforce" flag, and rdtscp is silently disabled
Consequence: libvirt SHOULD use something like "-cpu Opteron_G5,-rdtscp",
especially when it starts using (or emulating) enforce mode
This will require a solution on libvirt side. QEMU will just provide the mechanisms to report CPU model information
and check what the host and QEMU supports, but the decision to disable rdtscp to be able
to run Opteron_G[2345] needs to be taken by libvirt."
Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
Only non-cdrom drives default to cache=none, so the check
for whether to default to aio=native needs to take the same
condition into account.
I combined them close together to make their relation more
visible.
drive-mirror is doing lseek on source image before starting, and this can take a lot of time for big nfs volume
during this time, qmp socket is hanging
http://lists.nongnu.org/archive/html/qemu-devel/2015-05/msg01838.html
so we need to setup a big timeout
qemu devs are currently working to fix this
Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
Since qemu 2.2, a new "ready" flag has been added to blockjob
http://git.qemu.org/?p=qemu.git;a=commit;h=ef6dbf1e46ebd1d41ab669df5bba0bbdec6bd374
to known if we can complete it.
we can't use len==offset to known if all block are mirrored, because behaviour will change soon in qemu 2.3
http://git.qemu.org/?p=qemu.git;a=commit;h=b21c76529d55bf7bb02ac736b312f5f8bf033ea2
"block/mirror: Improve progress report
Instead of taking the total length of the block device as the block
job's length, use the number of dirty sectors. The progress is now the
number of sectors mirrored to the target block device. Note that this
may result in the job's length increasing during operation, which is
however in fact desirable.
"
Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
It is better to check if a VM is running in QemuServer then in Storage.
for the Storage there is no difference if it is running or not.
Signed-off-by: Wolfgang Link <w.link@proxmox.com>
Currently qemu auto fallback to aio=threads if cache=none|directsync
It's better to handle that correctly
see:
https://bugzilla.redhat.com/show_bug.cgi?id=1086704http://wiki.qemu.org/ChangeLog/2.3
Future incompatible changes:
Block device parameter aio=native has no effect without cache.direct=on. It will be made an error.
Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
