During migration, the lock process is paused in the perform phase
but not resumed if there is a subsequent failure, leaving the locked
resource unprotected.
The perform phase itself can fail, in which case the lock process
should be resumed before returning from perform. The finish phase
could also fail on the destination host, in which case the migration
is canceled in the confirm phase and the VM is resumed. The lock
process needs to be resumed there as well.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Commit 8708ca01c added virNetDevSwitchdevFeature() to check if a network
device has Switchdev capabilities. virNetDevSwitchdevFeature() attempts
to retrieve the PCI device associated with the network device, ignoring
non-PCI devices. It does so via the following call chain
virNetDevSwitchdevFeature()->virNetDevGetPCIDevice()->
virPCIGetDeviceAddressFromSysfsLink()
For non-PCI network devices (qeth, Xen vif, etc),
virPCIGetDeviceAddressFromSysfsLink() will report an error when
virPCIDeviceAddressParse() fails. virPCIDeviceAddressParse() also
logs an error. After commit 8708ca01c there are now two errors reported
for each non-PCI network device even though the errors are harmless.
To avoid the errors, introduce virNetDevIsPCIDevice() and use it in
virNetDevGetPCIDevice() before attempting to retrieve the associated
PCI device. virNetDevIsPCIDevice() uses the 'subsystem' property of the
device to determine if it is PCI. See the sysfs rules in kernel
documentation for more details
https://www.kernel.org/doc/html/latest/admin-guide/sysfs-rules.html
The list_guests function in libvirt-guests uses 'grep -v' to filter
Xen domain0 from a list of guests. If domain0 is the only item in
the list, 'grep -v' returns 1, causing the 'stop' operation to fail
when action is 'suspend'. Improve the filtering by using sed to remove
domain0 from the list of guests.
Noticed the following denial in audit.log when shutting down
an apparmor confined domain
type=AVC msg=audit(1512002299.742:131): apparmor="DENIED"
operation="open" profile="libvirt-66154842-e926-4f92-92f0-1c1bf61dd1ff"
name="/proc/1475/cmdline" pid=2958 comm="qemu-system-x86"
requested_mask="r" denied_mask="r" fsuid=469 ouid=0
Squelch the denial by allowing read access to /proc/<pid>/cmdline.
Add two new entries under new features for 3.10.0. One
advertising support for specifying distance between vNUMA cells
and another advertising Xen's support for vNUMA configuration.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
There was a recent report of the xen-xl converter not handling
config files missing an ending newline
https://www.redhat.com/archives/libvir-list/2017-October/msg01353.html
Commit 3cc2a9e0 fixed a similar problem when parsing content of a
file but missed parsing in-memory content. But AFAICT, the better
fix is to properly set the end of the content when initializing the
virConfParserCtxt in virConfParse().
This commit reverts the part of 3cc2a9e0 that appends a newline to
files missing it, and fixes setting the end of content when
initializing virConfParserCtxt. A test is also added to check
parsing in-memory content missing an ending newline.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Commit 6c43149c removed the minsize directive from the qemu logrotate
file but missed other hypervisors. Remove minsize from the libxl, lxc,
and uml logrotate files as well.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
On a cloud host it is possible to create 100's of unique instances
per day, each leaving behind a /var/log/libvirt/qemu/instance-name.log
file that is < 100k. With the current 'minsize 100k' directive, these
files are never rotated and hence never removed. Over months of time,
tens of thousands of these files can accumulate on the host.
Dropping 'minsize 100k' allows rotating small files, which will
increase the number of log files, but 'rotate 4' ensures they will
be removed after a month.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Commit b482925c added ptrace rule for the apparmor profiles,
but one was missed in the libvirtd profile for dnsmasq. It was
overlooked since the test machine did not have an active libvirt
network requiring dnsmasq that was also set to autostart. With
one active and set to autostart, the following denial is observed
in audit.log when restarting libvirtd
type=AVC msg=audit(1507320136.306:298): apparmor="DENIED" \
operation="ptrace" profile="/usr/sbin/libvirtd" pid=5472 \
comm="libvirtd" requested_mask="trace" denied_mask="trace" \
peer="/usr/sbin/dnsmasq"
With an active network, I suspect a libvirtd restart causes access
to /proc/<dnsmasq-pid>/*, hence the resulting denial. As a nasty
side affect of the denial, libvirtd thinks it needs to spawn a
dnsmasq process even though one is already running for the network.
E.g. after two libvirtd restarts
dnsmasq 1683 0.0 0.0 51188 2612 ? S 12:03 0:00 \
/usr/sbin/dnsmasq --conf-file=/var/lib/libvirt/dnsmasq/default.conf \
--leasefile-ro --dhcp-script=/usr/lib64/libvirt/libvirt_leaseshelper
root 1684 0.0 0.0 51160 576 ? S 12:03 0:00 \
/usr/sbin/dnsmasq --conf-file=/var/lib/libvirt/dnsmasq/default.conf \
--leasefile-ro --dhcp-script=/usr/lib64/libvirt/libvirt_leaseshelper
dnsmasq 4706 0.0 0.0 51188 2572 ? S 13:54 0:00 \
/usr/sbin/dnsmasq --conf-file=/var/lib/libvirt/dnsmasq/default.conf \
--leasefile-ro --dhcp-script=/usr/lib64/libvirt/libvirt_leaseshelper
root 4707 0.0 0.0 51160 572 ? S 13:54 0:00 \
/usr/sbin/dnsmasq --conf-file=/var/lib/libvirt/dnsmasq/default.conf \
--leasefile-ro --dhcp-script=/usr/lib64/libvirt/libvirt_leaseshelper
dnsmasq 4791 0.0 0.0 51188 2580 ? S 13:56 0:00 \
/usr/sbin/dnsmasq --conf-file=/var/lib/libvirt/dnsmasq/default.conf \
--leasefile-ro --dhcp-script=/usr/lib64/libvirt/libvirt_leaseshelper
root 4792 0.0 0.0 51160 572 ? S 13:56 0:00 \
/usr/sbin/dnsmasq --conf-file=/var/lib/libvirt/dnsmasq/default.conf \
--leasefile-ro --dhcp-script=/usr/lib64/libvirt/libvirt_leaseshelper
A simple fix is to add a ptrace rule for dnsmasq.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-By: Guido Günther <agx@sigxcpu.org>
When security drivers are active but confinement is not enabled,
there is no need to autogenerate <seclabel> elements when starting
a domain def that contains no <seclabel> elements. In fact,
autogenerating the elements can result in needless save/restore and
migration failures when the security driver is not active on the
restore/migration target.
This patch changes the virSecurityManagerGenLabel function in
src/security_manager.c to only autogenerate a <seclabel> element
if none is already defined for the domain *and* default
confinement is enabled. Otherwise the needless <seclabel>
autogeneration is skipped.
Resolves: https://bugzilla.opensuse.org/show_bug.cgi?id=1051017
The virDomainDef created by testBuildDomainDef in securityselinuxtest
adds a seclabel but does not increment nseclabels. Also, it should
populate seclabel->model with 'selinux'.
While at it, use the secdef itself to populate values instead of
the indirection through def->seclabels[0].
The libxl library allows a libxl_domain_config object to be serialized
from/to a JSON string. Use this to allow testing of the XML to
libxl_domain_config conversion process. Test XML is converted to
libxl_domain_config, which is then serialized to json. A json template
corresponding to the test XML is converted to a libxl_domain_config
object using libxl_domain_config_from_json(), and then serialized
back to json using libxl_domain_config_to_json(). The two json
docs are then compared.
Using libxl to convert the json template to a libxl_domain_config
object and then back to json provides a simple way to account for
any changes or additions to the json representation across Xen
releases.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
[update to v3.5.0-rc1, improve error reporting, use /bin/true emulator]
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
/domain/devices/disk/driver/@name is not a required or mandatory
attribute according to formatdomain, and indeed it was agreed on
IRC that the attribute is "optional for input, recommended (but
not required) for output". Currently the schema requires the
attribute, causing virt-xml-validate to fail on disk config where
the driver name is not explicitly specified. E.g.
# cat test.xml | grep -A 5 cdrom
<disk type='file' device='cdrom'>
<driver type='raw'/>
<target dev='hdb' bus='ide'/>
<readonly/>
<address type='drive' controller='0' bus='0' target='0' unit='1'/>
</disk>
# virt-xml-validate test.xml
Relax-NG validity error : Extra element devices in interleave
test.xml:21: element devices: Relax-NG validity error : Element domain failed to validate content
test.xml fails to validate
Relaxing the name attribute to be optional fixes the validation
# virt-xml-validate test.xml
test.xml validates
Attempting to start a domain with USB hostdevs but no USB controllers
fails with the rather cryptic error
libxl: error: libxl_qmp.c:287:qmp_handle_error_response: received an
error message from QMP server: Bus 'xenusb-0.0' not found
This can be fixed by creating default USB controllers. When no USB
controllers are defined, create the number of 8 port controllers
necessary to accommodate the number of defined USB devices.
Note that USB controllers are already created as needed in the
domainAttachDevice code path. E.g. a USB controller will be created,
if necessary, when attaching a USB device with
'virsh attach-device dom usbdev.xml'.
If a VNC listen address is not specified in domXML, libxl
will default to 127.0.0.1, but this is never reflected in the domXML.
In the case of spice, a missing listen address resulted in listening
on all interfaces, i.e. '0.0.0.0'. If not specified, set the listen
address in virDomainGraphicsDef struct to the libxl default when
creating the frame buffer device. Additionally, set default spice
listen address to 127.0.0.1.
Nested HVM support in the libxl driver is a news-worthy
improvement for libvirt 3.3.0.
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
libvirtd can spawn threads/tasks when creating new domains for
some hypervisors such as Xen's libxl driver, quickly reaching
the cgroups pids controller default TasksMax setting of 512. When
the limit is reached, attempting to create additional domains
results in an error from the cgroups pids controller, e.g.
kernel: [71282.213347] cgroup: fork rejected by pids controller in
/system.slice/libvirtd.service
Depending on domain type and configuration, anywhere from 4-7
threads/tasks may be created by libxl when starting a domain.
In order to support 4096 domains, similar to commit 27cd763500,
increase the TasksMax setting in libvirtd.service to
4096 * 8 = 32768 tasks.
Fix xlconfig channel tests when OOM testing is enabled.
TEST: xlconfigtest
32) Xen XL-2-XML Format channel-unix ... OK
Test OOM for nalloc=55 ................................................*** Error in `/home/jfehlig/virt/upstream/libvirt/tests/.libs/xlconfigtest': double free or corruption (fasttop): 0x0000000000679550 ***
...
(gdb) bt
#0 0x00007ffff36875af in raise () from /lib64/libc.so.6
#1 0x00007ffff36889aa in abort () from /lib64/libc.so.6
#2 0x00007ffff36c5150 in __libc_message () from /lib64/libc.so.6
#3 0x00007ffff36cb4f6 in malloc_printerr () from /lib64/libc.so.6
#4 0x00007ffff36cbcee in _int_free () from /lib64/libc.so.6
#5 0x00007ffff782babf in virFree (ptrptr=0x7fffffffdca8) at util/viralloc.c:582
#6 0x000000000042f2f3 in xenParseXLChannel (conf=0x677350, def=0x6815b0) at xenconfig/xen_xl.c:788
#7 0x000000000042f44e in xenParseXL (conf=0x677350, caps=0x6832b0, xmlopt=0x67f6e0) at xenconfig/xen_xl.c:828
#8 0x00000000004105a3 in testCompareFormatXML (
xlcfg=0x6811e0 "/home/jfehlig/virt/upstream/libvirt/tests/xlconfigdata/test-channel-unix.cfg",
xml=0x681110 "/home/jfehlig/virt/upstream/libvirt/tests/xlconfigdata/test-channel-unix.xml", replaceVars=false)
at xlconfigtest.c:152
When a channel is successfully parsed and its path and name fields
assigned from local variables, set the local variables to NULL to
prevent a double free on error.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Commit 4ab0c959 fixed a memory leak in libxlDriverGetDom0MaxmemConf
but introduced a potential double free of mem_tokens
*** Error in `/usr/sbin/libvirtd': double free or corruption (out):
0x00007fffc808cfd0 ***
Avoid double free by setting mem_tokens to NULL after calling
virStringListFree.
If the apparmor security driver is loaded/enabled and domain config
contains a <seclabel> element whose type attribute is not 'apparmor',
starting the domain fails when attempting to label resources such
as tap FDs.
Many of the apparmor driver entry points attempt to retrieve the
apparmor security label from the domain def, returning failure if
not found. Functions such as AppArmorSetFDLabel fail even though
domain config contains an explicit 'none' secuirty driver, e.g.
<seclabel type='none' model='none'/>
Change the entry points to succeed if the domain config <seclabel>
is not apparmor. This matches the behavior of the selinux driver.
When the libxl driver is initialized, it creates a virDomainDef
object for dom0 and adds it to the list of domains. Total memory
for dom0 was being set from the max_memkb field of libxl_dominfo
struct retrieved from libxl, but this field can be set to
LIBXL_MEMKB_DEFAULT (~0ULL) if dom0 maximum memory has not been
explicitly set by the user.
This patch adds some simple parsing of the Xen commandline,
looking for a dom0_mem parameter that also specifies a 'max' value.
If not specified, dom0 maximum memory is effectively all physical
host memory.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
The libxl driver reports different values of maximum memory depending
on state of a domain. If inactive, maximum memory value is reported
correctly. When active, maximum memory is derived from max_pages value
returned by the XEN_SYSCTL_getdomaininfolist sysctl operation. But
max_pages can be changed by toolstacks and does not necessarily
represent the maximum memory a domain can use during its active
lifetime.
A better location for determining a domain's maximum memory is the
/local/domain/<id>/memory/static-max node in xenstore. This value
is set from the libxl_domain_build_info.max_memkb field when creating
the domain. Currently it cannot be changed nor can its value be
exceeded by a balloon operation. From libvirt's perspective, always
reporting maximum memory with virDomainDefGetMemoryTotal() will produce
the same results as reading the static-max node in xenstore.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
When a user does not explicitly set a <driver> in the disk config,
libvirt defers selection of a default to libxl. This approach works
fine when starting a domain with such configuration or attaching a
disk to a running domain. But when detaching such a disk, libxl
will fail with "unrecognized disk backend type: 0". libxl makes no
attempt to recalculate a default backend (driver) on detach and
simply fails when uninitialized.
This patch updates the libvirt disk config with the backend selected
by libxl when starting a domain or attaching a disk to a running
domain. Another benefit of this approach is that the live XML is
also updated with the backend driver selected by libxl.
When starting a domian, a libxl_domain_config object is created from
virDomainDef. Any virDomainDiskDef devices with a format of
VIR_STORAGE_FILE_NONE are mapped to LIBXL_DISK_FORMAT_RAW in the
corresponding libxl_disk_device, but the virDomainDiskDef format is
never updated to reflect the change.
A better place to set a default format for disk devices is the
device post-parse callback, ensuring the virDomainDiskDef object
reflects the default format.
The typical pattern when calling libxl functions that populate a
structure is
libxl_foo foo;
libxl_foo_init(&foo);
libxl_get_foo(ctx, &foo);
...
libxl_foo_dispose(&foo);
Fix several instances of libxl_physinfo missing the init and
dispose calls.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
libxlGetAutoballoonConf is supposed to honor user-specified
autoballoon setting in libxl.conf. As written, the user-specified
setting could be overwritten by the subsequent logic to check
dom0_mem parameter. If user-specified setting is present and
correct, accept it. Only fallback to checking Xen dom0_mem
command line parameter if user-specfied setting is not present.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
xen.git commit 57f8b13c changed several of the libxl memory
get/set functions to take 64 bit parameters. The libvirt
libxl driver still uses uint32_t variables for these various
parameters, which is particularly problematic for the
libxl_set_memory_target() function.
When dom0 autoballooning is enabled, libvirt (like xl) determines
the memory needed to start a domain and the memory available. If
memory available is less than memory needed, dom0 is ballooned
down by passing a negative value to libxl_set_memory_target()
'target_memkb' parameter. Prior to xen.git commit 57f8b13c,
'target_memkb' was an int32_t. Subtracting a larger uint32 from
a smaller uint32 and assigning it to int32 resulted in a negative
number. After commit 57f8b13c, the same subtraction is widened
to a int64, resulting in a large positive number. The simple
fix taken by this patch is to assign the difference of the
uint32 values to a temporary int32 variable, which is then
passed to 'target_memkb' parameter of libxl_set_memory_target().
Note that it is undesirable to change libvirt to use 64 bit
variables since it requires setting LIBXL_API_VERSION to 0x040800.
Currently libvirt supports LIBXL_API_VERSION >= 0x040400,
essentially Xen >= 4.4.
The current logic around configuring timers in libxl based on
virDomainDef object is a bit brain dead. Unsupported timers are
silently ignored and tsc is only recognized if it is the first
timer specified.
Change the logic to reject unsupported timers and honor the tsc
timer regardless of its order when multiple timers are specified.
While local builds succeed fine, a build worker building in a
chroot environment is encountering the following error with
libvirt 3.0.0 release candidates
[ 162s] shunloadtest.o: In function `main':
[ 162s] /home/abuild/rpmbuild/BUILD/libvirt-3.0.0/tests/shunloadtest.c:110: undefined reference to `dlopen'
[ 162s] /home/abuild/rpmbuild/BUILD/libvirt-3.0.0/tests/shunloadtest.c:114: undefined reference to `dlsym'
[ 162s] /home/abuild/rpmbuild/BUILD/libvirt-3.0.0/tests/shunloadtest.c:133: undefined reference to `dlclose'
[ 162s] /home/abuild/rpmbuild/BUILD/libvirt-3.0.0/tests/shunloadtest.c:111: undefined reference to `dlerror'
[ 162s] /home/abuild/rpmbuild/BUILD/libvirt-3.0.0/tests/shunloadtest.c:115: undefined reference to `dlerror'
[ 162s] /home/abuild/rpmbuild/BUILD/libvirt-3.0.0/tests/shunloadtest.c:116: undefined reference to `dlclose'
Fix by appending DLOPEN_LIBS to shunloadtest_LDADD.
When LIBXL_HAVE_QED is defined, xlconfigtest fails
9) Xen XL-2-XML Format disk-qed ... command line: config parsing error
in disk specification: no vdev specified in
`target=/var/lib/libvirt/images/XenGuest2,format=qed,backendtype=qdisk,vdev=hda,access=rw'
FAILED
As per the xl-disk-configuration(5) man page, target= must come
last in the disk specification when specified by name:
When this parameter is specified by name, ie with the target=
syntax in the configuration file, it consumes the whole rest of the
DISKSPEC including trailing whitespaces. Therefore in that case
it must come last.
Change tests/xlconfigdata/test-disk-qed.cfg to adhere to this
restriction.
For HVM domains, pae is only set in libxl_domain_build_info when
explicitly specified in the hypervisor <features> config. This is
fine for i686 machines, but is incorrect behavior for x86_64 machines
where pae must always be enabled. See the following discussion for
additional details
https://www.redhat.com/archives/libvir-list/2017-January/msg00254.html
Although the migration port is immediately released in the
finish phase of migration, it was never set in the domain
private object when allocated in the prepare phase. So
libxlDomainMigrationFinish() always released a 0-initialized
migrationPort, leaking any allocated port. After enough
migrations to exhaust the migration port pool, migration would
fail with
error: internal error: Unable to find an unused port in range
'migration' (49152-49216)
Fix it by setting libxlDomainObjPrivate->migrationPort to the
port allocated in the prepare phase. While at it, also fix
leaking an allocated port if the prepare phase fails.
Commit 6c504d6a added a note to the virsh man page about the
deprecation of 'cap' and 'weight' settings for the credit
scheduler. To this day, the default scheduler in Xen is credit
and it supports setting 'cap' and 'weight'. Remove the deprecation
notice from the note on the Xen credit scheduler.
Reported-by: Volo M. <vm@vovs.net>
Due to a copy and paste error, the scheduler 'cap' parameter
was over-writing the 'weight' parameter when preparing the
return parameters in libxlDomainGetSchedulerParametersFlags.
As a result, the scheduler weight was never shown when getting
schedinfo and setting the weight failed as well
virsh schedinfo testvm
Scheduler : credit
cap : 0
virsh schedinfo testvm --cap 50 --weight 500
Scheduler : credit
error: invalid scheduler option: weight
The obvious fix is to assign the 'caps' parameter to the correct
item in the parameter list.
Reported-by: Volo M. <vm@vovs.net>
libxl events are delivered to libvirt via the libxlDomainEventHandler
callback registered with libxl. Documenation in
$xensrc/tools/libxl/libxl_event.h states that the callback "may occur
on any thread in which the application calls libxl". This can result
in deadlock since many of the libvirt callees of libxl hold a lock on
the virDomainObj they are working on. When the callback is invoked, it
attempts to find a virDomainObj corresponding to the domain ID provided
by libxl. Searching the domain obj list results in locking each obj
before checking if it is active, and its ID equals the requested ID.
Deadlock is possible when attempting to lock an obj that is already
locked further up the call stack. Indeed, Max Ustermann reported an
instance of this deadlock
https://www.redhat.com/archives/libvir-list/2015-November/msg00130.html
Guido Rossmueller also recently stumbled across it
https://www.redhat.com/archives/libvir-list/2016-September/msg00287.html
Fix the deadlock by moving the lookup of virDomainObj to the
libxlDomainShutdownThread. After this patch, libxl events are
enqueued on the libvirt side and processed by dedicated thread,
avoiding the described deadlock.
Reported-by: Max Ustermann <ustermann78@web.de>
Reported-by: Guido Rossmueller <Guido.Rossmueller@gdata.de>
By default, virt-manager (and likely other libvirt-based apps) sets
the VIR_MIGRATE_PERSIST_DEST flag when invoking the migrate API, which
fails in a Xen setup since the libxl driver does not support the flag.
Persisting a domain is a trivial task in the grand scheme of migration,
so be nice to libvirt apps and add support for VIR_MIGRATE_PERSIST_DEST
in the libxl driver.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
libvirt uses the new_id PCI sysfs interface to bind a PCI stub driver
to a PCI device. The new_id interface is known to be buggy and racey,
hence a more deterministic interface was introduced in the 3.12 kernel:
driver_override. For more details see
https://www.redhat.com/archives/libvir-list/2016-June/msg02124.html
For more details about the driver_override interface and examples of
its usage, see
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/pci/pci-driver.c?h=v3.12&id=782a985d7af26db39e86070d28f987cad21313c0
This patch adds support for the driver_override interface by
- adding new virPCIDevice{BindTo,UnbindFrom}StubWithOverride functions
that use the driver_override interface
- renames the existing virPCIDevice{BindTo,UnbindFrom}Stub functions
to virPCIDevice{BindTo,UnbindFrom}StubWithNewid to perserve existing
behavior on new_id interface
- changes virPCIDevice{BindTo,UnbindFrom}Stub function to call one of
the above depending on availability of driver_override
The patch includes a bit of duplicate code, but allows for easily
dropping the new_id code once support for older kernels is no
longer desired.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
The libxl driver has long supported migration V3 but has never
indicated so in the connectSupportsFeature API. As a result, apps
such as virt-manager that use the more generic virDomainMigrate API
fail with
libvirtError: this function is not supported by the connection driver:
virDomainMigrate
Add VIR_DRV_FEATURE_MIGRATION_V3 to the list of features marked as
supported in the connectSupportsFeature API.
When starting a guest and copying host vendor cpuid to the guest
cpu, libvirtd would crash if the host cpu contained a NULL vendor
field. Avoid the crash by checking for a valid vendor in the host
cpu before copying the cpuid to the guest cpu.
For completeness, here is a backtrace from the crash
(gdb) bt
f0 0x00007ffff739bf33 in x86DataCpuid (cpuid=0x8, cpuid=0x8,
data=data@entry=0x7fffb800ee78) at cpu/cpu_x86.c:287
f1 virCPUx86DataAddCPUID (data=data@entry=0x7fffb800ee78, cpuid=0x8)
at cpu/cpu_x86.c:355
f2 0x00007ffff739ef47 in x86Compute (host=<optimized out>, cpu=0x7fffb8000cc0,
guest=0x7fffecca7348, message=<optimized out>) at cpu/cpu_x86.c:1580
f3 0x00007fffd2b38e53 in qemuBuildCpuModelArgStr (migrating=false,
hasHwVirt=<synthetic pointer>, qemuCaps=0x7fffb8001040, buf=0x7fffecca7360,
def=0x7fffc400ce20, driver=0x1c) at qemu/qemu_command.c:6283
f4 qemuBuildCpuCommandLine (cmd=cmd@entry=0x7fffb8002f60,
driver=driver@entry=0x7fffc80882c0, def=def@entry=0x7fffc400ce20,
qemuCaps=qemuCaps@entry=0x7fffb8001040, migrating=<optimized out>)
at qemu/qemu_command.c:6445
(gdb) f2
(gdb) p *host_model
$23 = {name = 0x7fffb800ec50 "qemu64", vendor = 0x0, signature = 0, data = {
len = 2, data = 0x7fffb800e720}}
Long, long ago before libxl_get_required_shadow_memory() was
made publicly available, its code was copied to the libxl driver
for calculating shadow memory requirements of HVM domains.
Long ago, libxl_get_required_shadow_memory() was exported in
libxl_utils.h and included in xen-devel packages everywhere.
Remove the copied code, which has become stale, and let libxl
provode a proper shadow memory value.
At least with systemd v210, NOTIFY_SOCKET is abstact, e.g.
@/org/freedesktop/systemd1/notify. sendmsg() fails on such a socket
with "Connection refused". The unix(7) man page contains the following
details wrt abstract socket addresses
abstract: an abstract socket address is distinguished (from a
pathname socket) by the fact that sun_path[0] is a null byte
('\0'). The socket's address in this namespace is given by the
additional bytes in sun_path that are covered by the specified
length of the address structure. (Null bytes in the name have
no special significance.)
So we need to be more precise about the address length, setting it to
the sizeof sa_family_t + length of address copied to sun_path instead
of setting it to the sizeof the entire sockaddr_un struct.
Resolves: https://bugzilla.opensuse.org/show_bug.cgi?id=987668
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
When domXML contains only <console type='pty'> and no corresponding
<serial>, the console is "stolen" [1] and used as the first <serial>
device. When this "stolen" console is accessed from the libxl driver
(in libxlConsoleCallback and libxlDomainOpenConsole), check if the
targetType is VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_SERIAL, and use the
"stolen" device in def->serials[0] instead. Prior to this change,
creating a domain with input XML containing only a <console> device
and subsequently attempting to access its console with
'virsh console' would fail
error: internal error: character device <null> is not using a PTY
[1] See comments associated with virDomainDefAddConsoleCompat() in
$LIBVIRT-SRC/src/conf/domain_conf.c:
Commit 2a58ed0b added support for creating guests with USB
hostdevs. Commit fc21d10 later added support for hotplut of
USB hostdevs. Advertise support for USB hostdevs in the
domcapabilities.
In addition add the appropriate caps for USB support on
domaincapstest when libvirt is built on a Xen with
LIBXL_HAVE_PVUSB. Otherwise domaincapstest would fail i.e.
testing the wrong domain capabilities.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
When converting domXML to xen xl.cfg, backendtype should
not be emitted if <driver> is not specified. Moreover,
<driver name='file'/> should be converted to backendtype
qdisk, similar to handling of <driver> in libxlMakeDisk()
in libxl_conf.c.
Prior to this change, connectDomainXMLToNative would
produce incorrect xl.cfg when the input domXML contained
<driver name='file'/>
domXML:
<disk type="file" device="disk">
<driver name="file"/>
<source file="/image/file/path"/>
<target dev="xvda" bus="xen"/>
</disk>
virsh domxml-to-native xen-xl domXML
disk = [ "format=raw,vdev=xvda,access=rw,backendtype=target=/image/file/path" ]
xl create xl.cfg
config parsing error in disk specification: unknown value
for backendtype: near `target=/image/file/path' in
`format=raw,vdev=xvda,access=rw,backendtype=target=/image/file/path'
Populate libxl_domain_build_info struct with bios and firmware
info from virDomainLoaderDef.
Note: Currently libxl only allows specifying the type of BIOS.
For type LIBXL_BIOS_TYPE_OVMF, the firmware path is configured
when building Xen using '--with-system-ovmf='. If not specified,
LIBXL_FIRMWARE_DIR/ovmf.bin is used. In the future, Xen will
support a user-specified firmware path. See
http://lists.xenproject.org/archives/html/xen-devel/2016-03/msg01628.html
Once that work is merged into xen.git, the libvirt libxl driver
will be able to honor a user-specified path. In the meantime use
the implicit path, which is tolerable since it is advertised in
domcapabilities.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Add support to xenconfig for conversion of xl.cfg(5) bios config
to/from libvirt domXml <loader> config. SeaBIOS is the default
for HVM guests using upstream QEMU. ROMBIOS is the default when
using the old qemu-dm. This patch allows specifying OVMF as an
alternate firmware.
Example xl.cfg:
bios = "ovmf"
Example domXML:
<os>
...
<loader readonly='yes' type='pflash'>/usr/lib/xen/boot/ovmf.bin</loader>
</os>
Note that currently Xen does not support a separate nvram for
non-volatile variables.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Prefer firmwares specified via --with-loader-nvram configure
option. If none are specified, use the Xen-provided default
firmwares found in LIBXL_FIRMWARE_DIR.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
The virQEMUDriverConfig object contains lists of
loader:nvram pairs to advertise firmwares supported by
by the driver, and qemu_conf.c contains code to populate
the lists, all of which is useful for other drivers too.
To avoid code duplication, introduce a virFirmware object
to encapsulate firmware details and switch the qemu driver
to use it.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Xen only supports network-based disks with the qemu (aka qdisk) driver.
Set the driverName to 'qemu' in libxlDomainDeviceDefPostParse() if
not already set. When starting a domain with network-based disks,
ensure the driverName is 'qemu'.
Resolves: https://bugzilla.opensuse.org/show_bug.cgi?id=981094
Some of the test configuration files in tests/xlconfigdata
use the old qemu-dm as the emulator. Many of the configuration
features tested (spice, rbd, multi-usb) are not even usable with
the old qemu. Change these files to use the new qemu-xen (also
known as qemu upstream) emulator.
Note: This change fixes xlconfigtest failures when the old
qemu is actually installed on the system. During device post
parse, the libxl driver attempts to invoke the emulator to
determine if it is the old or new qemu so it can properly set
video RAM defaults. With the old qemu installed, the default
video RAM was set differently than the expected value.
Changing all the test data files to use qemu-xen ensures
predictable results wrt default video RAM size.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
When probing the <emulator> with '-help' to determine if
it is the old qemu, errors are reported if the emulator
doesn't exist
libvirt: error : internal error: Child process
(/usr/lib/xen/bin/qemu-dm -help) unexpected exit status 127:
libvirt: error : cannot execute binary /usr/lib/xen/bin/qemu-dm:
No such file or directory
Avoid the probe if the specified emulator doesn't exist,
squelching the error. There is no behavior change since
libxlDomainGetEmulatorType() would return
LIBXL_DEVICE_MODEL_VERSION_QEMU_XEN if the probe failed
via virCommandRun().
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Similar to "support Xen migration stream V2 in save/restore",
add support for indicating the migration stream version in
the migration code. To accomplish this, add a minimal migration
cookie in the libxl driver that is passed between source and
destination hosts. Initially, the cookie is only used in
the Begin and Prepare phases of migration to communicate the
version of the migration stream produced by the source.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Xen 4.6 introduced a new migration stream commonly referred to as
"migration V2". Xen 4.6 and newer always produce this new stream,
whereas Xen 4.5 and older always produce the legacy stream.
Support for migration stream V2 can be detected at build time with
LIBXL_HAVE_SRM_V2 from libxl.h. The legacy and V2 streams are not
compatible, but a V2 host can accept and convert a legacy stream.
Commit e7440656 changed the libxl driver to use the lowest libxl
API version possible (version 0x040200) to ensure the driver
builds against older Xen releases. The old 4.2 restore API does
not support specifying a stream version and assumes a legacy
stream, even if the incoming stream is migration V2. Thinking it
has been given a legacy stream, libxl will fail to convert an
incoming stream that is already V2, which causes the entire
restore operation to fail. Xen's libvirt-related OSSTest has been
failing since commit e7440656 landed in libvirt.git master. One
of the more recent failures can be seen here
http://lists.xenproject.org/archives/html/xen-devel/2016-05/msg00071.html
This patch changes the call to libxl_domain_create_restore() to
include the stream version if LIBXL_HAVE_SRM_V2 is defined. The
version field of the libxlSavefileHeader struct is also updated
to '2' when LIBXL_HAVE_SRM_V2 is defined, ensuring the stream
version in the header matches the actual stream version produced
by Xen. Along with bumping the libxl API requirement to 0x040400,
this patch fixes save/restore on a migration V2 Xen host.
Oddly, migration has never used the libxlSavefileHeader. It
handles passing configuration in the Begin and Prepare phases,
and then calls libxl directly to transfer domain state/memory
in the Perform phase. A subsequent patch will add stream
version handling in the Begin and Prepare phase handshaking,
which will fix the migration related OSSTest failures.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
In LIBXL_API_VERSION 0x040400, the libxl_domain_create_restore API
gained a parameter for specifying restore parameters. Switch to
using version 0x040400, which will be useful in a subsequent commit
to specify the Xen migration stream version when restoring.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
To ensure the libvirt libxl driver will build with future versions
of Xen where the libxl API may change in incompatible ways,
explicitly use LIBXL_API_VERSION 0x040200. The libxl driver
does use new libxl APIs that have been added since Xen 4.2, but
currently it does not make use of any changes made to existing
APIs such as libxl_domain_create_restore or libxl_set_vcpuaffinity.
The version can be bumped if/when the libxl driver consumes the
changed APIs.
Further details can be found in the following discussion thread
https://www.redhat.com/archives/libvir-list/2016-April/msg00178.html
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Chunyan sent a nice cleanup patch for libxlDomainDetachNetDevice
https://www.redhat.com/archives/libvir-list/2016-March/msg00926.html
which I incorrectly modified before pushing as commit b5534e53. My
modification caused network devices of type hostdev to no longer
be removed. This patch changes b5534e53 to resemble Chunyan's
original, correct patch.
Chunyan sent a correct patch to fix a resource leak on error in
libxlDomainAttachNetDevice
https://www.redhat.com/archives/libvir-list/2016-March/msg00924.html
I made what was thought to be an improvement and pushed the patch as
commit e6336442. As it turns out, my change broke adding net devices
that are actually hostdevs to the list of nets in virDomainDef. This
patch changes e6336442 to resemble Chunyan's original, correct
patch.
Remove disabling domain death events from libxlDomainStart error
path. The domain death event is already disabled in libxlDomainCleanup.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Commit e6336442 changed the 'out:' label to 'cleanup' in
libxlDomainAttachNetDevice(), but missed a comment referencing
the 'out:' label. Remove it from the comment since it is no
longer accurate anyhow.
Until now, the libxl driver ignored any <hap> setting in domain XML
and deferred to libxl, which enables hap if not specified. While
this is a good default, it prevents disabling hap if desired.
This change allows disabling hap with <hap state='off'/>. hap is
explicitly enabled with <hap/> or <hap state='on/>. Absense of <hap>
retains current behavior of deferring default state to libxl.
hap is enabled by default in xm and xl config and usually only
specified when it is desirable to disable hap (hap = 0). Change
the xm,xl <-> xml converter to behave similarly. I.e. only
produce 'hap = 0' when <hap state='off'/> and vice versa.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Most hypervisors use Hardware Assisted Paging by default and don't
require specifying the feature in domain conf. But some hypervisors
support disabling HAP on a per-domain basis. To enable HAP by default
yet provide a knob to disable it, extend the <hap> feature with a
'state=on|off' attribute, similar to <pvspinlock> and <vmport> features.
In the absence of <hap>, the hypervisor default (on) is used. <hap>
without the state attribute would be the same as <hap state='on'/> for
backwards compatibility. And of course <hap state='off'/> disables hap.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
None of the existing domXML configs under tests/* specify a
default cache mode since default generally means "use the
hypervisor default" and is left unset by the various hypervisor
drivers. Add a config to tests/domainschemadata that specifies
cache='default'.
The docs claims the cache attribute of the disk <driver>
element supports 'default' as one of its permissible values,
but such configuration fails virt-xml-validate. Add 'default'
as one of the cache attribute choices in domaincommon.rng.
libxlMakeNic opens a virConnect object and takes a reference on a
virNetwork object, but doesn't drop the references on all error
paths. Rework the function to follow the standard libvirt pattern
of using a local 'ret' variable to hold the function return value,
performing all cleanup and returning 'ret' at a 'cleanup' label.
xl/libxl already supports qemu's network-based block backends
such as nbd and rbd. libvirt has supported configuring such
<disk>s for long time too. This patch adds support for rbd
disks in the libxl driver by generating a rbd device URL from
the virDomainDiskDef object. The URL is passed to libxl via the
pdev_path field of libxl_device_disk struct. libxl then passes
the URL to qemu for cosumption by the rbd backend.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
The target= setting in xl disk configuration can be used to encode
meta info that is meaningful to a backend. Leverage this fact to
support qdisk network disk types such as rbd. E.g. <disk> config
such as
<disk type='network' device='disk'>
<driver name='qemu' type='raw'/>
<source protocol='rbd' name='pool/image'>
<host name='mon1.example.org' port='6321'/>
<host name='mon2.example.org' port='6322'/>
<host name='mon3.example.org' port='6322'/>
</source>
<target dev='hdb' bus='ide'/>
<address type='drive' controller='0' bus='0' target='0' unit='1'/>
</disk>
can be converted to the following xl config (and vice versa)
disk = [ "format=raw,vdev=hdb,access=rw,backendtype=qdisk,
target=rbd:pool/image:auth_supported=none:mon_host=mon1.example.org\\:6321\\;mon2.example.org\\:6322\\;mon3.example.org\\:6322"
]
Note that in xl disk config, a literal backslash in target= must
be escaped with a backslash. Conversion of <auth> config is not
handled in this patch, but can be done in a follow-up patch.
Also add a test for the conversions.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
The most formal form of xl disk configuration uses key=value
syntax to define each configuration item, e.g.
format=raw, vdev=xvda, access=rw, backendtype=phy, target=disksrc
Change the xl disk formatter to produce this syntax, which allows
target= to contain meta info needed to setup a network-based
disksrc (e.g. rbd, nbd, iscsi). For details on xl disk config
format, see $xen-src/docs/misc/xl-disk-configuration.txt
Update the disk config in the tests to use the formal syntax.
But add tests to ensure disks specified with the positional
parameter syntax are correctly converted to <disk> XML.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
The virErrorDomain enum has VIR_FROM_XEN, VIR_FROM_XEND,
VIR_FROM_XENSTORE, VIR_FROM_SEXPR, and VIR_FROM_XENXM. Use
these elements in the corresponding .c files. While at it,
remove the VIR_FROM_THIS define in src/xenconfig/xenxs_private.h.
Commmit fd2e3c4c used the domctl version 8 structure for version 9
in the xen_getdomaininfolist union, resulting in insufficient buffer
size (and subsequent memory corruption) for the GETDOMAININFOLIST
ioctl.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
The libxl_device_nic structure supports specifying an outgoing rate
limit based on a time interval and bytes allowed per interval. In xl
config a rate limit is specified as "<RATE>/s@<INTERVAL>". INTERVAL
is optional and defaults to 50ms.
libvirt expresses outgoing limits by average (required), peak, burst,
and floor attributes in units of KB/s. This patch supports the outgoing
bandwidth limit by converting the average KB/s to bytes per interval
based on the same default interval (50ms) used by xl.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Both xm and xl config have long supported specifying vif rate
limiting, e.g.
vif = [ 'mac=00:16:3E:74:3d:76,bridge=br0,rate=10MB/s' ]
Add support for mapping rate to and from <bandwidth> in the xenconfig
parser and formatter. rate is mapped to the required 'average' attribute
of the <outbound> element, e.g.
<interface type='bridge'>
...
<bandwidth>
<outbound average='10240'/>
</bandwidth>
</interface>
Also add a unit test to check the conversion logic.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
The xen sexpr config format has long supported specifying vif rate
limiting, e.g.
(device
(vif
(mac '00:16:3e:1b:b1:47')
(rate '10240KB/s')
...
)
)
Add support for mapping rate to and from <bandwidth> in the xenconfig
sexpr parser and formatter. rate is mapped to the required 'average'
attribute of the <outbound> element, e.g.
<interface type='bridge'>
...
<bandwidth>
<outbound average='10240'/>
</bandwidth>
</interface>
Also add unit tests to check the conversion logic.
This patch benefits both the old xen driver and the libxl driver.
Both drivers gain support for vif bandwidth when converting to/from
domXML and xen-sxpr. In addition, the old xen driver will now be
able to handle vif 'rate' setting when communicating with xend.
From: Ian Campbell <ian.campbell@citrix.com>
xend prior to 4.0 understands vcpus as maxvcpus and vcpu_avail
as a bit map of which cpus are online (default is all).
xend from 4.0 onwards understands maxvcpus as maxvcpus and
vcpus as the number which are online (from 0..N-1). The
upstream commit (68a94cf528e6 "xm: Add maxvcpus support")
claims that if maxvcpus is omitted then the old behaviour
(i.e. obeying vcpu_avail) is retained, but AFAICT it was not,
in this case vcpu==maxcpus==online cpus. This is good for us
because handling anything else would be fiddly.
This patch changes parsing of the virDomainDef maxvcpus and vcpus
entries to use the corresponding 'maxvcpus' and 'vcpus' settings
from xm and xl config. It also drops use of the old Xen 3.x
'vcpu_avail' setting.
The change also removes the maxvcpus limit of MAX_VIRT_VCPUS (since
maxvcpus is simply a count, not a bit mask), which is particularly
crucial on ARM where MAX_VIRT_CPUS == 1 (since all guests are
expected to support vcpu placement, and therefore only the boot
vcpu's info lives in the shared info page).
Existing tests adjusted accordingly, and new tests added for the
'maxvcpus' setting.
Remove use of xendConfigVersion in the s-expresion config formatter/parser
in src/xenconfig/. Adjust callers in the xen and libxl drivers accordingly.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Jim Fehlig