For a long time we've masked out VIR_DOMAIN_BLOCK_COPY_SHALLOW if
there's no backing chain for the copied disk to simplify the code.

One of the refactors of the block copy code caused that we no longer
update the 'flags' variable just the local copies. This was okay until
in ccd4228a we started storing the job flags in the block job data.

Given that we modify how we call qemu we also should modify @flags so
that the correct value is recorded in the block job data.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>

Move the check whether the job is already synchronised to the beginning
of the function so that we don't try to do some of the steps necessary
for pivoting prior to actually wanting to pivot.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NEric Blake <eblake@redhat.com>

'nfs' variable was set to -1 or -2 on agent failure. Cleanup then tried
to free 'nfs' elements of the array which resulted into a crash.

Make 'nfs' size_t and assign it only on successful agent call.

https://bugzilla.redhat.com/show_bug.cgi?id=1812965

Broken by commit 599ae372Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

The only caller doesn't check the value and also there are no real
errors to report anyways.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

The virQEMUCaps structure has usedQMP member which in the past
used to tell if qemu we are dealing with is capable of QMP. Well,
we don't support HMP anymore (minus a few HMP passthrough
commands, which are wrapped into QMP anyways) and the member is
not used really.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NDaniel Henrique Barboza <danielhb413@gmail.com>

needReply added in [1] looks redundant. Indeed it is set to false only
when mon->await_event is set too (the only exception qemuAgentFSTrim
which is mistaken).

However it fixes the issue when qemuAgentCommand exits on error path and
mon->await_event is not reset. Let's instead reset mon->await_event properly.

Also remove "Woken up by event" debug message as it can be misleading.
We can get it also if monitor is closed due to serial changed event
currently. Anyway both qemuAgentClose and qemuAgentNotifyEvent log
itself.

[1] qemu: make sure agent returns error when required data are missing
Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NMichal Privoznik <mprivozn@redhat.com>

Sync was introduced in [1] to check for ga presence. This
check is racy but in the era before serial events are available
there was not better solution I guess.

In case we have the events the sync function is different. It allows us
to flush stateless ga channel from remnants of previous communications.
But we need to do it only once. Until we get timeout on issued command
channel state is ok.

[1] qemu_agent: Issue guest-sync prior to every command
Signed-off-by: NNikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NMichal Privoznik <mprivozn@redhat.com>

If a disk has persistent reservations enabled, qemu-pr-helper
might open not only /dev/mapper/control but also individual
targets of the multipath device. We are already querying for them
in CGroups, but now we have to create them in the namespace too.
This was brought up in [1].

1: https://bugzilla.redhat.com/show_bug.cgi?id=1711045#c61Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Tested-by: NLin Ma <LMa@suse.com>
Reviewed-by: NJim Fehlig <jfehlig@suse.com>

This converts the QEMU agent APIs to use the per-VM
event loop, which involves switching from virEvent APIs
to GMainContext / GSource APIs.

A GSocket is used as a convenient way to create a GSource
for a socket, but is not yet used for actual I/O.
Reviewed-by: NMichal Privoznik <mprivozn@redhat.com>
Signed-off-by: NDaniel P. Berrangé <berrange@redhat.com>

We are dealing with the QEMU agent, not the monitor.
Reviewed-by: NMichal Privoznik <mprivozn@redhat.com>
Signed-off-by: NDaniel P. Berrangé <berrange@redhat.com>

This converts the QEMU monitor APIs to use the per-VM
event loop, which involves switching from virEvent APIs
to GMainContext / GSource APIs.

A GSocket is used as a convenient way to create a GSource
for a socket, but is not yet used for actual I/O.
Reviewed-by: NMichal Privoznik <mprivozn@redhat.com>
Signed-off-by: NDaniel P. Berrangé <berrange@redhat.com>

In common with regular QEMU guests, the QMP probing
will need an event loop for handling monitor I/O
operations.
Reviewed-by: NMichal Privoznik <mprivozn@redhat.com>
Signed-off-by: NDaniel P. Berrangé <berrange@redhat.com>

The event loop thread will be responsible for handling
any per-domain I/O operations, most notably the QEMU
monitor and agent sockets.

We start this event loop when launching QEMU, but stopping
the event loop is a little more complicated. The obvious
idea is to stop it in qemuProcessStop(), but if we do that
we risk loosing the final events from the QEMU monitor, as
they might not have been read by the event thread at the
time we tell the thread to stop.

The solution is to delay shutdown of the event thread until
we have seen EOF from the QEMU monitor, and thus we know
there are no further events to process.

Note that this assumes that we don't have events to process
from the QEMU agent.
Reviewed-by: NMichal Privoznik <mprivozn@redhat.com>
Signed-off-by: NDaniel P. Berrangé <berrange@redhat.com>

When preparing images for block jobs we modify their seclabels so
that QEMU can open them. However, as mentioned in the previous
commit, secdrivers base some it their decisions whether the image
they are working on is top of of the backing chain. Fortunately,
in places where we call secdrivers we know this and the
information can be passed to secdrivers.

The problem is the following: after the first blockcommit from
the base to one of the parents the XATTRs on the base image are
not cleared and therefore the second attempt to do another
blockcommit fails. This is caused by blockcommit code calling
qemuSecuritySetImageLabel() over the base image, possibly
multiple times (to ensure RW/RO access). A naive fix would be to
call the restore function. But this is not possible, because that
would deny QEMU the access to the base image.  Fortunately, we
can use the fact that seclabels are remembered only for the top
of the backing chain and not for the rest of the backing chain.
And thanks to the previous commit we can tell secdrivers which
images are top of the backing chain.

Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1803551Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NPeter Krempa <pkrempa@redhat.com>

Historically threads are given a name based on the C function,
and this name is just used inside libvirt. With OS level thread
naming this name is now visible to debuggers, but also has to
fit in 15 characters on Linux, so function names are too long
in some cases.
Reviewed-by: NMichal Privoznik <mprivozn@redhat.com>
Signed-off-by: NDaniel P. Berrangé <berrange@redhat.com>

The qemuMonitorOpenFD method has not been used since it
was first introduced.
Reviewed-by: NMichal Privoznik <mprivozn@redhat.com>
Signed-off-by: NDaniel P. Berrangé <berrange@redhat.com>

Libvirt has never configured the QEMU agent to support
running on a PTY implicitly. In theory an end user may
have written such an XML config, but this is reasonably
unlikely since when a bare <channel> is provided, libvirt
will auto-expand it to a UNIX socket backend.

With this change a user who has use the PTY backend will
have to switch to the UNIX backend if they wish to use
libvirt APIs for interacting with the agent. This will
not have guest ABI impact.
Reviewed-by: NMichal Privoznik <mprivozn@redhat.com>
Signed-off-by: NDaniel P. Berrangé <berrange@redhat.com>

qemuMonitorJSONMakeCommandInternal does the full command construction if
you pass in what would become the value of the 'arguments' key. Refactor
the open-coded implementation to use the helper and use modern cleanup
helpers at the same time.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NDaniel Henrique Barboza <danielhb413@gmail.com>

Make it obvious that the function always returns a valid pointer and fix
all callers.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NDaniel Henrique Barboza <danielhb413@gmail.com>

I've found that if my virtlogd is socket activated but the daemon
doesn't run yet, then the virt-qemu-run is killed right after it
tries to start the domain. The problem is that because the default
setting is to use virtlogd, the domain create code tries to
connect to virtlogd socket, which in turn tries to detect who is
connecting (virNetSocketGetUNIXIdentity()) and as a part of it,
it will try to open /proc/${PID_OF_SHIM}/stat which is denied by
SELinux:

  type=AVC msg=audit(1582903501.927:323): avc:  denied  { search } for  \
  pid=1210 comm="virtlogd" name="1843" dev="proc" ino=37224 \
  scontext=system_u:system_r:virtlogd_t:s0-s0:c0.c1023 \
  tcontext=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023 tclass=dir \
  permissive=0

Virtlogd reacts by closing the connection which the shim sees as
SIGPIPE. Since the default response to the signal is Term, we
don't even get to reporting any error nor to removing the
temporary directory.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NAndrea Bolognani <abologna@redhat.com>

When virt-qemu-run is ran without any root directory specified on
the command line, a temporary directory is made and used instead.
But since we are using g_dir_make_tmp() to create the directory
it is going to have 0700 mode. So even though we create the whole
directory structure under it and label everything, QEMU is very
likely to not have the access. This is because in this case there
is no qemu.conf and thus distro default UID:GID is used to run
QEMU (e.g. qemu:kvm on Fedora). Change the mode of the temporary
directory so that everybody has eXecute permission.
Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NAndrea Bolognani <abologna@redhat.com>

Libvirt tries to forbid migration onto the same host and it does
that by checking if local and remote hostnames are the same and
whether local and remote UUIDs are the same. Well, the latter
makes sense but the former doesn't really because libvirtd can be
running inside an UTS namespace and hostnames can appear the same
on both sides of migration. On the other hand, host UUIDs are
unique, so rely on them when trying to prevent migration onto the
same host.

Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1639596Signed-off-by: NMichal Privoznik <mprivozn@redhat.com>
Reviewed-by: NDaniel P. Berrangé <berrange@redhat.com>

Use the 'flat' flag for 'query-named-block-nodes' if qemu supports
QEMU_CAPS_QMP_QUERY_NAMED_BLOCK_NODES_FLAT in qemuBlockGetNamedNodeData.

We don't need the data so plumb in whether qemu supports the
'flat' output.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

Modern qemu allows to skip the nested redundant data in the output of
query-named-block-nodes. Plumb in the support for the argument that
enables it.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

Replace qemuMonitorBlockGetNamedNodeData by qemuBlockGetNamedNodeData.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

Use g_autoptr to get rid of the cleanup section.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

Detect the presence of the flag and make it available internally as
QEMU_CAPS_QMP_QUERY_NAMED_BLOCK_NODES_FLAT.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

The monitor password callback was removed long time ago but the callback
type and variable were left around. Finish the cleanup.
Signed-off-by: NPeter Krempa <pkrempa@redhat.com>
Reviewed-by: NJán Tomko <jtomko@redhat.com>

Format the 'vhost-user-fs' device on the QEMU command line.

This device provides shared file system access using the FUSE protocol
carried over virtio.
The actual file server is implemented in an external vhost-user-fs device
backend process.

https://bugzilla.redhat.com/show_bug.cgi?id=1694166Signed-off-by: NJán Tomko <jtomko@redhat.com>
Reviewed-by: NDaniel P. Berrangé <berrange@redhat.com>
Reviewed-by: NPeter Krempa <pkrempa@redhat.com>
Tested-by: NAndrea Bolognani <abologna@redhat.com>

Look into /usr/share/qemu/vhost-user to see whether we can find
a suitable virtiofsd binary, in case the user did not provide one
in the domain XML.
Signed-off-by: NJán Tomko <jtomko@redhat.com>
Reviewed-by: NDaniel P. Berrangé <berrange@redhat.com>
Reviewed-by: NPeter Krempa <pkrempa@redhat.com>
Tested-by: NAndrea Bolognani <abologna@redhat.com>

Wire up the code to put virtiofsd in the emulator cgroup on domain
startup.
Signed-off-by: NJán Tomko <jtomko@redhat.com>
Reviewed-by: NPeter Krempa <pkrempa@redhat.com>
Tested-by: NAndrea Bolognani <abologna@redhat.com>

Start virtiofsd for each <filesystem> device using it.

Pre-create the socket for communication with QEMU and pass it
to virtiofsd.

Note that virtiofsd needs to run as root.

https://bugzilla.redhat.com/show_bug.cgi?id=1694166

Introduced by QEMU commit a43efa34c7d7b628cbf1ec0fe60043e5c91043ea
Signed-off-by: NJán Tomko <jtomko@redhat.com>
Reviewed-by: NPeter Krempa <pkrempa@redhat.com>
Tested-by: NAndrea Bolognani <abologna@redhat.com>

This is not yet supported.
Signed-off-by: NJán Tomko <jtomko@redhat.com>
Reviewed-by: NDaniel P. Berrangé <berrange@redhat.com>
Reviewed-by: NPeter Krempa <pkrempa@redhat.com>
Tested-by: NAndrea Bolognani <abologna@redhat.com>

Reject unsupported configurations.
Signed-off-by: NJán Tomko <jtomko@redhat.com>
Reviewed-by: NPeter Krempa <pkrempa@redhat.com>
Tested-by: NAndrea Bolognani <abologna@redhat.com>
Reviewed-by: NMasayoshi Mizuma <m.mizuma@jp.fujitsu.com>

Add a 'virtiofsd_debug' option for tuning whether to run virtiofsd
in debug mode.
Signed-off-by: NJán Tomko <jtomko@redhat.com>
Reviewed-by: NDaniel P. Berrangé <berrange@redhat.com>
Reviewed-by: NPeter Krempa <pkrempa@redhat.com>
Tested-by: NAndrea Bolognani <abologna@redhat.com>

Introduce a new 'virtiofs' driver type for filesystem.

<filesystem type='mount' accessmode='passthrough'>
  <driver type='virtiofs'/>
  <source dir='/path'/>
  <target dir='mount_tag'>
  <address type='pci' domain='0x0000' bus='0x00' slot='0x02' function='0x0'/>
</filesystem>
Signed-off-by: NJán Tomko <jtomko@redhat.com>
Reviewed-by: NDaniel P. Berrangé <berrange@redhat.com>
Reviewed-by: NPeter Krempa <pkrempa@redhat.com>
Tested-by: NAndrea Bolognani <abologna@redhat.com>

Introduced by QEMU commit 98fc1ada4cf70af0f1df1a2d7183cf786fc7da05
    virtio: add vhost-user-fs base device

Released in QEMU v4.2.0.
Signed-off-by: NJán Tomko <jtomko@redhat.com>
Reviewed-by: NPeter Krempa <pkrempa@redhat.com>
Acked-by: NStefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: NDaniel P. Berrangé <berrange@redhat.com>
Tested-by: NAndrea Bolognani <abologna@redhat.com>

Signed-off-by: NJán Tomko <jtomko@redhat.com>
Reviewed-by: NPeter Krempa <pkrempa@redhat.com>
Tested-by: NAndrea Bolognani <abologna@redhat.com>

Pass logManager to qemuExtDevicesStart for future usage.
Signed-off-by: NJán Tomko <jtomko@redhat.com>
Reviewed-by: NDaniel P. Berrangé <berrange@redhat.com>
Tested-by: NAndrea Bolognani <abologna@redhat.com>

The default memlock limit is 64k which is not enough to start a single
VM. The requirements for one VM are 12k, 8k for eBPF map and 4k for eBPF
program, however, it fails to create eBPF map and program with 64k limit.
By testing I figured out that the minimal limit is 80k to start a single
VM with functional eBPF and if I add 12k I can start another one.

This leads into following calculation:

80k as memlock limit worked to start a VM with eBPF which means there
is 68k of lock memory that I was not able to figure out what was using
it.  So to get a number for 4096 VMs:

        68 + 12 * 4096 = 49220

If we round it up we will get 64M of memory lock limit to support 4096
VMs with default map size which can hold 64 entries for devices.

This should be good enough as a sane default and users can change it if
the need to.

Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1807090Signed-off-by: NPavel Hrdina <phrdina@redhat.com>
Reviewed-by: NMichal Privoznik <mprivozn@redhat.com>