Modify qemuBlockJobSyncBeginDisk to operate on qemuBlockt sJobDataPtr and
rename it accordingly.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

We can properly track the job type when starting the job so that we
don't have to infer it later.

This patch also adds an enum of block job types specific to qemu
(qemuBlockjobType) which mirrors the public block job types
(virDomainBlockJobType) but allows for other types to be added later
which will not be public.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

If the job wasn't started, we don't need to end the synchronous job. Add
a note and drop the unnecessary calls.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

Rather than directly modifying fields in the qemuBlockJobDataPtr
structure add a bunch of fields which allow to do the transitions.

This will help later when adding more complexity to the job handling.

APIs introduced in this patch are:

qemuBlockJobDiskNew - prepare for starting a new blockjob on a disk
qemuBlockJobDiskGetJob - get the block job data structure for a disk

For individual job state manipulation the following APIs are added:
qemuBlockJobStarted - Sets the job as started with qemu. Until that
                      the job can be cancelled without asking qemu.

qemuBlockJobStartupFinalize - finalize job startup. If the job was
                              started in qemu already, just releases
                              reference to the job object. Otherwise
                              clears everything as if the job was never
                              started.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

Extract the disk mirroring startup code from the loop into a separate
function to allow cleaner cleanup paths.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

When cancelling job after a reconnect we can now use the disk block job
state rather than having to re-detect it in the migration code.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

Internally we do a 'block-copy' to accomodate non-shared storage
migration but the code did not fill in that the block job was active on
the disk when starting the copy job. Since we handle block jobs finishes
regardless of having it registered it's not a problem but soon will
become one.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

All the public APIs of the qemu_blockjob module operate on a 'disk'.
Since I'll be adding APIs which operate on a job later let's rename the
existing ones.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

Signed-off-by: NYuri Chornoivan <yurchor@ukr.net>
Reviewed-by: NJán Tomko <jtomko@redhat.com>
Signed-off-by: NJán Tomko <jtomko@redhat.com>

If migration is cancelled or confirm phase fails the domain
should be kept on the source even if VIR_MIGRATE_UNDEFINE_SOURCE
was requested.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJiri Denemark <jdenemar@redhat.com>

There are some checks done when parsing a migration cookie. For
instance, one of the checks ensures that the domain is not being
migrated onto the same host. If that is the case, then we are in
big trouble because the @vm is the same domain object used by
source and it has some jobs sets and everything so recovering
from failed cookie parsing would be needlessly hard.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJiri Denemark <jdenemar@redhat.com>

The function currently takes virDomainObjPtr because it's using
both: the domain definition and domain private data.
Unfortunately, this means that in prepare phase we can't parse
migration cookie before putting incoming domain def onto domain
objects list (addressed in the very next commit). Change the
arguments so that virDomainDef and private data are passed
instead of virDomainObjPtr.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJiri Denemark <jdenemar@redhat.com>

There are several functions called in the cleanup path. Some of
them do report error (e.g. qemuDomainRemoveInactiveJob()) which
may result in overwriting an error reported earlier with some
less useful message.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NJiri Denemark <jdenemar@redhat.com>

The only place where VIR_DOMAIN_EVENT_RESUMED should be generated is the
RESUME event handler to make sure we don't generate duplicate events or
state changes. In the worse case the duplicity can revert or cover
changes done by other event handlers.

For example, after QEMU sent RESUME, BLOCK_IO_ERROR, and STOP events
we could happily mark the domain as running and report
VIR_DOMAIN_EVENT_RESUMED to registered clients.

https://bugzilla.redhat.com/show_bug.cgi?id=1612943Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

When a domain is killed on the source host while it is being migrated
and libvirtd is waiting for the migration to finish (waiting for the
domain condition in qemuMigrationSrcWaitForCompletion), the run-time
state including priv->job.current may already be freed once
virDomainObjWait returns with -1. Thus the priv->job.current pointer
cached in jobInfo is no longer valid and setting jobInfo->status may
crash the daemon.

https://bugzilla.redhat.com/show_bug.cgi?id=1593137Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

Once we called qemuDomainObjEnterRemote to talk to the destination
daemon during a peer to peer migration, the vm lock is released and we
only hold an async job. If the source domain dies at this point the
monitor EOF callback is allowed to do its job and (among other things)
clear all private data irrelevant for stopped domain. Thus when we call
qemuDomainObjExitRemote, the domain may already be gone and we should
avoid touching runtime private data (such as current job info).

In other words after acquiring the lock in qemuDomainObjExitRemote, we
need to check the domain is still alive. Unless we're doing offline
migration.

https://bugzilla.redhat.com/show_bug.cgi?id=1589730Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

The variable is used to store the offline migration capability of the
destination daemon. Let's call it 'dstOffline' so that we can later use
'offline' to indicate whether we were asked to do offline migration.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

And replace all calls with virObjectEventStateQueue such that:

    qemuDomainEventQueue(driver, event);

becomes:

    virObjectEventStateQueue(driver->domainEventState, event);

And remove NULL checking from all callers.
Signed-off-by: NAnya Harter <aharter@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>
Signed-off-by: NJán Tomko <jtomko@redhat.com>

Replace instances where we previously called virGetLastError just to
either get the code or to check if an error exists with
virGetLastErrorCode to avoid a validity pre-check.
Signed-off-by: NRamy Elkest <ramyelkest@gmail.com>
Reviewed-by: NErik Skultety <eskultet@redhat.com>

Emphasize that it's for the 'drive' part of the disk.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

The alias of the secret for decrypting the TLS passphrase is useless
besides for TLS setup. Stop passing it around.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

This has been broken since commit v4.0.0-165-g93412bb8 which added
jobInfo->statsType enum to distinguish various statistics types. During
migration the type will always be QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION,
however the destination code consuming the statistics data from
migration cookie failed to properly set the type. So even though
everything was filled in, the type remained *_NONE and any attempt to
fetch the statistics data of a completed migration on the destination
host failed.

https://bugzilla.redhat.com/show_bug.cgi?id=1584071Signed-off-by: NJiri Denemark <jdenemar@redhat.com>

Implement the secure way to transport non-shared storage data across
migrations. The new approach uses blockdev-add to create the NBD client
so that the TLS secret object can be specified.

https://bugzilla.redhat.com/show_bug.cgi?id=1300772Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Separate the code relevant for this approach so that we can later add a
second implementation without making the function messy.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Drop the mention of 'drive mirror' from the function names and mention
NBD. This will help when adding the 'blockdev mirror' migration code
which will allow using TLS.

Additionally fix some of the function comments to make more sense
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

The initiation of a synchronous block job in the NBD storage migration
code was placed after entering the monitor thus after the lock on the VM
object was unlocked. Thankfully nothing bad could happen in this
situation since the migration job prevents any disk detaches or other
modifications of the domain object.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJohn Ferlan <jferlan@redhat.com>

Signed-off-by: NClementine Hayat <clem@lse.epita.fr>
Reviewed-by: NJán Tomko <jtomko@redhat.com>
Signed-off-by: NJán Tomko <jtomko@redhat.com>

The pointer to the qemu driver is already included in domain object's
private data, so does not need to be passed as yet another parameter
when the domain object is already passed.

Also removes parameter 'driver' from functions which had it just because of
qemuBlockJobUpdate.
Signed-off-by: NRoland Schulz <schullzroll@gmail.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>
Signed-off-by: NJán Tomko <jtomko@redhat.com>

When adding a new object to the domain object list, there should
have been 2 virObjectRef calls made one for each list into which
the object was placed to match the 2 virObjectUnref calls that
would occur during Remove as part of virHashRemoveEntry when
virObjectFreeHashData is called when the element is removed from
the hash table as set up in virDomainObjListNew.

Some drivers (libxl, lxc, qemu, and vz) handled this inconsistency
by calling virObjectRef upon successful return from virDomainObjListAdd
in order to use virDomainObjEndAPI when done with the returned @vm.
While others (bhyve, openvz, test, and vmware) handled this via only
calling virObjectUnlock upon successful return from virDomainObjListAdd.

This patch will "unify" the approach to use virDomainObjEndAPI
for any @vm successfully returned from virDomainObjListAdd.

Because list removal is so tightly coupled with list addition,
this patch fixes the list removal algorithm to return the object
as entered - "locked and reffed".  This way, the callers can then
decide how to uniformly handle add/remove success and failure.
This removes the onus on the caller to "specially handle" the
@vm during removal processing.

The Add/Remove logic allows for some logic simplification such
as in libxl where we can Remove the @vm directly rather than
needing to set a @remove_dom boolean and removing after the
libxlDomainObjEndJob completes as the @vm is locked/reffed.
Signed-off-by: NJohn Ferlan <jferlan@redhat.com>
Reviewed-by: NErik Skultety <eskultet@redhat.com>

Signed-off-by: NMartin Kletzander <mkletzan@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

Use the TLS env for migration when starting the NBD server if TLS is
enabled for migration.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NDaniel P. Berrangé <berrange@redhat.com>

To allow encryption of the non-shared storage migration NBD connection
we will need to instantiated the NBD server with the TLS env.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NDaniel P. Berrangé <berrange@redhat.com>

When a VM is destroyed while being migrated (waiting in
qemuMigrationSrcWaitForCompletion) the private object cleanup code frees
the 'current' job info. Since the migration code attempts to setup
various aspects of the current job even on failure this results into a
crash.

Job data is cleared in qemuDomainObjPrivateDataClear since commit
888aa4b6

Fix this by skipping all of the code which requires the qemu process to
be alive if the VM is not active any more.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJiri Denemark <jdenemar@redhat.com>

Since libvirt is currently not able to setup the NBD migration stream
secured by TLS we should not allow such migration since data would be
transferred unencrypted.

This will break compatibility of TLS migration if non-shared storage is
requested but the security implications are more severe.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>

Implied by QEMU >= 1.2.0.
Signed-off-by: NJán Tomko <jtomko@redhat.com>
Reviewed-by: NAndrea Bolognani <abologna@redhat.com>

Trying to delete the non-existent TLS objects results in ugly error
messages in the log, which could easily confuse users. Let's avoid this
confusion by not trying to delete the objects if we were not asked to
enable TLS migration and thus we didn't created the objects anyway.

This patch restores the behavior to the state before "qemu: Reset all
migration parameters".
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

We store the flags passed to the API which started the migration. Let's
use them instead of a separate bool to check if post-copy migration was
requested.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

When an async job is running, we sometimes need to know how it was
started to distinguish between several types of the job, e.g., post-copy
vs. normal migration. So far we added a specific bool item to
qemuDomainJobObj for such cases, which doesn't scale very well and
storing such bools in status XML would be painful so we didn't do it.

A better approach is to store the flags passed to the API which started
the async job, which can be easily stored in status XML.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

When an always-on migration capability is supposed to be enabled on both
sides of migration, each side can only enable the feature if it is
enabled by the other side.

Thus the source host sends a list of supported migration capabilities in
the migration cookie generated in the Begin phase. The destination host
consumes the list in the Prepare phase and decides what capabilities can
be enabled when starting a QEMU process for incoming migration. Once
done the destination sends the list of supported capabilities back to
the source where it is used during the Perform phase to determine what
capabilities can be automatically enabled.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

Some migration capabilities may be enabled automatically, but only if
both sides of migration support them. Thus we need to be able transfer
the list of supported migration capabilities in migration cookie.
Signed-off-by: NJiri Denemark <jdenemar@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>