This change adds backend qemu command line support for new libvirt
global feature 'acpi-bridge-hotplug'. This option can be used as
following:
<feature>
<pci>
<acpi-bridge-hotplug state='off|on'/>
</pci>
</feature>
The '<pci>' sub-element under '<feature>' is also newly introduced.
'acpi-bridge-hotplug' turns on the following command line option to
qemu for x86 guests:
(pc): -global PIIX4_PM.acpi-pci-hotplug-with-bridge-support=<off|on>
(q35): -global ICH9-LPC.acpi-pci-hotplug-with-bridge-support=<off|on>
This change also adds the required qemuxml2argv unit tests in order to
test correct qemu arguments. Unit tests have also been added to test
qemu capability validation checks as well as checks for using this
option with the right architecture.
Signed-off-by: Ani Sinha <ani@anisinha.ca>
Reviewed-by: Laine Stump <laine@redhat.com>
This change introduces a new libvirt sub-element <pci> under
<features> that can be used to configure all pci related features.
Currently the only sub-sub element supported by this sub-element is
'acpi-bridge-hotplug' as shown below:
<features>
<pci>
<acpi-bridge-hotplug state='on|off'/>
</pci>
</features>
The above option is only available for the QEMU driver, for x86 guests
only. It is a global option, affecting all PCI bridge controllers on
the guest.
The 'acpi-bridge-hotplug' option enables or disables ACPI hotplug
support for cold-plugged pci bridges. Examples of bridges include the
PCI-PCI bridge (pci-bridge controller) for pc (i440fx) machinetypes,
or PCIe-PCI bridges and pcie-root-port controllers for q35
machinetypes.
For pc machinetypes in x86, this option has been available in QEMU
since version 2.1. Please see the following changes in qemu repo:
9e047b982452c6 ("piix4: add acpi pci hotplug support")
133a2da488062e ("pc: acpi: generate AML only for PCI0 devices if PCI
bridge hotplug is disabled")
For q35 machinetypes, this was introduced in QEMU 6.1 with the
following changes in qemu repo:
(a) c0e427d6eb5fef ("hw/acpi/ich9: Enable ACPI PCI hot-plug")
(b) 17858a16950860 ("hw/acpi/ich9: Set ACPI PCI hot-plug as default on
Q35")
The reasons for enabling ACPI based hotplug for PCIe (q35) based
machines (as opposed to native hotplug) are outlined in (b). There are
use cases where users would still want to use native
hotplug. Therefore, this config option enables users to choose either
ACPI based hotplug or native hotplug for bridges (for example for pcie
root port controller in q35 machines).
Qemu capability validation checks have also been added along with
related unit tests to exercise the new conf option.
Signed-off-by: Ani Sinha <ani@anisinha.ca>
Reviewed-by: Laine Stump <laine@redhat.com>
qemu added support for i440fx specific global boolean flag
PIIX4_PM.acpi-pci-hotplug-with-bridge-support
around version 2.1. This flag is enabled by default. When disabled, it
turns off acpi pci hotplug for cold plugged pci bridges in i440fx
machine types.
Very recently, in qemu version 6.1, the same global option was also
added for q35 machine types as well.
ICH9-LPC.acpi-pci-hotplug-with-bridge-support
This option turns on or off acpi based hotplug for cold plugged pcie
bridges like pcie root ports. This flag is also enabled by
default. Please refer to the following qemu changes:
c0e427d6eb5fef ("hw/acpi/ich9: Enable ACPI PCI hot-plug")
17858a16950860 ("hw/acpi/ich9: Set ACPI PCI hot-plug as default on Q35")
This patch adds the corresponding qemu capabilities in libvirt. For
i440fx, the capability is detected as
QEMU_CAPS_PIIX_ACPI_HOTPLUG_BRIDGE. For q35, the capability is
detected as QEMU_CAPS_ICH9_ACPI_HOTPLUG_BRIDGE.
Please note that the test specific qemu capabilities .replies files
has already been updated as a part of regular refreshing them when a
new qemu version is released. Hence, no updates to those files are
required.
Signed-off-by: Ani Sinha <ani@anisinha.ca>
Reviewed-by: Laine Stump <laine@redhat.com>
When a server decides to close a client, the
virNetServerClientCloseLocked() is called. In here various
cleanup steps are taken, but the most important part (from this
commit's POV at least) is the way that the socket is closed.
Firstly, removal of the socket associated with the client from
the event loop is signalized and then the socket is unrefed. The
socket is not closed just yet though, because the event loop
holds a reference to it. This reference will be freed as soon as
the event loop wakes up and starts issuing callbacks (in this
case virNetSocketEventFree()).
So far, this is how things usually work. But if the daemon
reaches the number of opened files limit, things start to work
differently.
If the RLIMIT_NOFILE limit is reached and there's a client that
wants to connect then the event loop wakes up, sees POLLIN on the
socket and calls virNetServerServiceAccept() which in turn calls
virNetSocketAccept(). But because of the limit, accept() fails
with EMFILE leaving the POLLIN event unhandled. The dispatch then
continues to next FDs with events on them. BUT, it will NOT call
the socket removal callback (virNetSocketEventFree()) because it
has low priority (G_PRIORITY_DEFAULT_IDLE). Per glib's
documentation:
* Each event source is assigned a priority. The default priority,
* %G_PRIORITY_DEFAULT, is 0. Values less than 0 denote higher priorities.
* Values greater than 0 denote lower priorities. Events from high priority
* sources are always processed before events from lower priority sources.
and per g_idle_add() documentation:
* Adds a function to be called whenever there are no higher priority
* events pending to the default main loop. The function is given the
* default idle priority, %G_PRIORITY_DEFAULT_IDLE.
Now, because we did not accept() the client we are constantly
seeing POLLIN on the main socket and thus the removal of the
client socket won't ever happen.
The fix is to set at least the same priority as other sources,
but since we want to just close an FD, let's give it the highest
priority and call it before handling other events.
This issue can be easily reproduced, for instance:
# ulimit -S -n 40 (tweak this number if needed)
# ./src/libvirtd
from another terminal:
# for ((i=0; i<100; i++)); do virsh list & done; virsh list
The last `virsh list` must not get stuck.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2007168
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
With cgroup v1 I'm seeing LXC container startup failures:
$ sudo virt-install --connect lxc:/// --name test-container --memory 128
--boot init=/bin/sh
Starting install...
ERROR error from service:
GDBus.Error:org.freedesktop.machine1.NoMachineForPID: PID 2145047 does
not belong to any known machine
libvirt 7.0.0 works but 7.1.0+ does not. The root error seems to predate
that, showing up in syslog, but commit 9c1693eff made it fatal:
commit 9c1693eff4
Author: Pavel Hrdina <phrdina@redhat.com>
Date: Fri Feb 5 16:17:35 2021 +0100
vircgroup: use DBus call to systemd for some APIs
The error comes from virSystemdGetMachineByPID. The PID that shows up in
the above error message does not match the leader PID as reported by
machinectl.
This change fixes the error. Things seem to continue to work with
cgroupsv2 after this change.
https://gitlab.com/libvirt/libvirt/-/issues/182
Tested-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Cole Robinson <crobinso@redhat.com>
Other devices (includes 9p-based fsdev) call this wrapper
before formatting the device.
Add it here too.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
This reverts commit 801e6da29c
They are not needed anymore.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Reconstruct the socket path from priv->libDir in every user.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Intended as a replacement for qemuVirtioFSCreateSocketFilename,
to be used outside of qemu_virtiofs.c
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
The commit adding the vhost-user-fs device forgot to format
the device's alias on the command line.
Thankfully it was not needed yet because virtiofs migration
is not yet supported, but it will be needed in the future
to allow hot(un)plug.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
In case when libvirt runs inside a restricted container it may
not have enough permissions to modify unpriv_sgio. However, it
may have been set beforehand by sysadmin or an orchestration
tool. Therefore, let's check whether the currently set value is
the one we want and if it is refrain from writing to the file.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2010306
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
The capability name piix4-acpi-root-hotplug-en is not conventional and
appreared to be confusing to some. "en" suffix is also incorrect as the
capability in qemu is used to both enable and disable hotplug on the pci root
bus on the i440fx. Hence, rename it to piix4.acpi-root-pci-hotplug so that it
is clearer, less confusing and more accurate.
Signed-off-by: Ani Sinha <ani@anisinha.ca>
Reviewed-by: Laine Stump <laine@redhat.com>
Also introduces a G_DEFINE_AUTOPTR_CLEANUP_FUNC for virCHMonitor.
Signed-off-by: William Douglas <william.douglas@intel.com>
Reviewed-by: Laine Stump <laine@redhat.com>
It is already being unrefed in virCHMonitorDispose().
Signed-off-by: William Douglas <william.douglas@intel.com>
Reviewed-by: Laine Stump <laine@redhat.com>
In virCHMontiorNew the monitor object was referenced an additional
time incorrectly preventing it from being disposed of, and wasn't
always closed properly on failure.
Signed-off-by: William Douglas <william.douglas@intel.com>
Reviewed-by: Laine Stump <laine@redhat.com>
In virCHMonitorBuildKernelRelatedJson there are two cases of json
value objects being lost after the pointer being redefined. This
change removes the needless redefinition.
Signed-off-by: William Douglas <william.douglas@intel.com>
Reviewed-by: Laine Stump <laine@redhat.com>
This change adds qemu backend command line support for enabling or disabling
hotplug on the pci-root controller using the 'target' sub-element of the
pci-root controller as shown below:
<controller type='pci' model='pci-root'>
<target hotplug='off'/>
</controller>
'<target hotplug='off/on'/>' is only valid for pc (i440fx-based x86)
machinetypes and turns on the following command line option that is passed
to qemu for x86 guests:
-global PIIX4_PM.acpi-root-pci-hotplug=<off/on>
Before introduction of this attribute, hotplug was always enabled for
pci-root of an i440fx-based machinetype, and since its introduction
the default setting has always been "on" for those machinetypes.
This change also adds the required qemuxml2argv unit tests in order to test
correct qemu arguments. Unit tests have also been added to test qemu capability
validation checks.
Signed-off-by: Ani Sinha <ani@anisinha.ca>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>
This change introduces libvirt xml support to enable/disable hotplug on the
pci-root controller. It adds a 'target' subelement for the pci-root controller
with a 'hotplug' property. This property can be used to enable or disable
hotplug for the pci-root controller. For example, in order to disable hotplug
on the pci-root controller, one has to use set '<target hotplug='off'>' as
shown below:
<controller type='pci' model='pci-root'>
<target hotplug='off'/>
</controller>
'<target hotplug='on'>' option would enable hotplug for pci-root controller.
This is also the default value. This option is only available for pc machine
types and is applicable for qemu/kvm accelerator only.This feature was
introduced from qemu version 5.2 with the following change in qemu repository:
3d7e78aa7777f ("Introduce a new flag for i440fx to disable PCI hotplug on the root bus")
The above qemu commit describes some reasons why users might to disable hotplug
on PCI root buses.
Related unit tests to exercise the new conf option has also been added.
Signed-off-by: Ani Sinha <ani@anisinha.ca>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>
The following change in qemu added support for a global boolean flag specific
to i440fx machines that would turn off or on acpi based hotplug for pci root
bus:
3d7e78aa7777f ("Introduce a new flag for i440fx to disable PCI hotplug on the root bus")
The option is passed as "-global PIIX4_PM.acpi-root-pci-hotplug=on" etc in qemu
commandline. It is enabled by default. This patch adds the corresponding qemu
capabilities in libvirt as QEMU_CAPS_PIIX_ACPI_ROOT_PCI_HOTPLUG.
Please note that the test specific qemu capabilities .replies files has already
been updated as a part of regular refreshing them when a new qemu version is
released. Hence, no updates to those files are required.
Signed-off-by: Ani Sinha <ani@anisinha.ca>
Reviewed-by: Laine Stump <laine@redhat.com>
The qemuDomainSetMemoryFlags() allows for memballoon
(<currentMemory/>) changes for both active and inactive guests.
And just before doing any change, we have to make sure that the
new size is not greater than the total memory (<memory/>).
However, the total memory includes not only the regular guest
memory, but also sum of maximum sizes of all virtio-mems (in fact
all memory devices for that matter). But virtio-mem devices are
modified differently (via virDomainUpdateDevice()) and thus the
upper limit for new balloon size has to be lowered.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reporting how much memory is exposed to the guest happens under
<currentMemory/> which is taken from def->mem.cur_balloon. The
reported amount should account for both balloon size and the sum
of @currentsize of all virtio-mems. For instance, if domain has
4GiB via balloon and additional 2GiB via virtio-mem, then the
domain XML should report 6GiB. The same applies for domain
statistics.
The way to achieve this is to account for either balloon or
virtio-mem when the size of the other is changed, e.g. on balloon
change we have to add all @currentsize (for non virtio-mem these
will be zero, so the check for memory model is needless, but
makes it more obvious what's happening), and vice versa.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
If the QEMU driver restarts it loses the track of the current size
of virtio-mem (because it's runtime type of information and thus
not stored in XML) and therefore, we have to refresh it when
reconnecting to the domain monitor.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
As advertised in previous commit, this event is delivered to us
when virtio-mem module changes the allocation inside the guest.
It comes with one attribute - size - which holds the new size of
the virtio-mem (well, allocated size), in bytes.
Mind you, this is not necessarily the same number as 'requested
size'. It almost certainly will be when sizing the memory up, but
it might not be when sizing the memory down - the guest kernel
might be unable to free some blocks.
This current size is reported in the domain XML as an output
element only.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
This function will be needed in the next commit where we will
want to find virtio-mem given its alias by QEMU on the monitor.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
The virtio-mem has another property that isn't exposed yet:
current size exposed to the guest. Please note, that this is
different to <requested/> because esp. on sizing the memory
down guest may refuse to release some blocks. Therefore, let's
have another size to report in the XML. But because of its
nature, the <current/> won't be parsed and is report only (for
live XMLs).
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Updating offline XML of <memory/> devices might come handy when
dealing with virtio-mem devices. But it's implemented to just
replace one virDomainMemoryDef with another so it can be used to
change almost anything.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
As advertised in one of previous commits, we want to be able to
change 'requested-size' attribute of virtio-mem on the fly. This
commit does exactly that. Changing anything else is checked for
and forbidden.
Once guest has changed the allocation, QEMU emits an event which
we will use to track the allocation. In the next commit.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Nothing special is happening here. All important changes were
done when for 'virtio-pmem' (adjusting the code to put virtio
memory on PCI bus, generating alias using
qemuDomainDeviceAliasIndex(). The only bit that might look
suspicious is no prealloc for virtio-mem. But if you think about
it, the whole purpose of this device is to change amount of
memory exposed to guest on the fly. There is no point in locking
the whole backend in memory.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
The virtio-mem is paravirtualized mechanism of adding/removing
memory to/from a VM. A virtio-mem-pci device is split into blocks
of equal size which are then exposed (all or only a requested
portion of them) to the guest kernel to use as regular memory.
Therefore, the device has two important attributes:
1) block-size, which defines the size of a block
2) requested-size, which defines how much memory (in bytes)
is the device requested to expose to the guest.
The 'block-size' is configured on command line and immutable
throughout device's lifetime. The 'requested-size' can be set on
the command line too, but also is adjustable via monitor. In
fact, that is how management software places its requests to
change the memory allocation. If it wants to give more memory to
the guest it changes 'requested-size' to a bigger value, and if it
wants to shrink guest memory it changes the 'requested-size' to a
smaller value. Note, value of zero means that guest should
release all memory offered by the device. Of course, guest has to
cooperate. Therefore, there is a third attribute 'size' which is
read only and reflects how much memory the guest still has. This
can be different to 'requested-size', obviously. Because of name
clash, I've named it 'current' and it is dealt with in future
commits (it is a runtime information anyway).
In the backend, memory for virtio-mem is backed by usual objects:
memory-backend-{ram,file,memfd} and their size puts the cap on
the amount of memory that a virtio-mem device can offer to a
guest. But we are already able to express this info using <size/>
under <target/>.
Therefore, we need only two more elements to cover 'block-size'
and 'requested-size' attributes. This is the XML I've came up
with:
<memory model='virtio-mem'>
<source>
<nodemask>1-3</nodemask>
<pagesize unit='KiB'>2048</pagesize>
</source>
<target>
<size unit='KiB'>2097152</size>
<node>0</node>
<block unit='KiB'>2048</block>
<requested unit='KiB'>1048576</requested>
</target>
<address type='pci' domain='0x0000' bus='0x00' slot='0x04' function='0x0'/>
</memory>
I hope by now it is obvious that:
1) 'requested-size' must be an integer multiple of
'block-size', and
2) virtio-mem-pci device goes onto PCI bus and thus needs PCI
address.
Then there is a limitation that the minimal 'block-size' is
transparent huge page size (I'll leave this without explanation).
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
This capability tracks whether memory-backend-* supports .reserve
attribute which is going to be important for backends associated
with virtio-mem devices.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
This commit introduces a new capability that reflects virtio-mem-pci
device support in QEMU:
QEMU_CAPS_DEVICE_VIRTIO_MEM_PCI, /* -device virtio-mem-pci */
The virtio-mem-pci device was introduced in QEMU 5.1.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
New virHostMemGetTHPSize() is introduced which allows caller to
obtain THP PMD (Page Middle Directory) size, which is equal to
the minimal size that THP can use, taken from kernel doc
(Documentation/admin-guide/mm/transhuge.rst):
Some userspace (such as a test program, or an optimized memory allocation
library) may want to know the size (in bytes) of a transparent hugepage::
cat /sys/kernel/mm/transparent_hugepage/hpage_pmd_size
Since this size depends on the host architecture and the kernel
it won't change whilst libvirtd is running. Therefore, we can use
virOnce() and cache the value. Of course, we can be running under
kernel that has THP disabled or has no notion of THP at all. In
that case a negative value is returned to signal error.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
There are two variables that are used only in a single
loop. Move their definitions into their respective blocks.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
When parsing CPU topology, which is described in <topology/>
attributes we can use virXMLPropUInt() instead of virXPathUInt()
as the former results in shorter code.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
There is no need to use virXPathULong() and a temporary UL
variable if we can use virXPathUInt() directly.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Added by QEMU commit:
b96feb2cb9 "9pfs: local: Add support for custom fmode/dmode in 9ps
mapped security modes"
in 2.10.0
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Now that it's no longer used, remove probing for it
and mark it as deprecated.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Even though we only allow this option on x86,
all QEMUs report the command line option.
Added in QEMU v1.1:
6a48ffaaa7 "kvm: Activate in-kernel irqchip support"
Remove the pointless capability.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Assume the presence of the 'sandbox' option is enough,
no need to look at the parameters.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
