Similar to the race fixed by
be34c3c7ef, make sure
to wait around for KVM to release the resources from
a hot-detached PCI device before attempting to
rebind that device to the host driver.
Signed-off-by: Chris Lalancette <clalance@redhat.com>
The QEMU driver contained code to generate a -device string for piix4-ide, but
wasn't using it. This change removes this string generation. It also adds a
comment explaining why IDE and FDC controllers don't generate -device strings.
The change also generates an error if a sata controller is specified for a QEMU
domain, as this isn't supported.
* src/qemu/qemu_conf.c: Remove VIR_DOMAIN_CONTROLLER_TYPE_IDE handler in
qemuBuildControllerDevStr(). Ignore IDE and FDC controllers. Error if
SATA controller is discovered. Add comments.
On RHEL-5 the qemu-kvm binary is located in /usr/libexec.
To reduce confusion for people trying to run upstream libvirt
on RHEL-5 machines, make the qemu driver look in /usr/libexec
for the qemu-kvm binary.
To make this work, I modified virFindFileInPath to handle an
absolute path correctly. I also ran into an issue where
NULL was sometimes being passed for the file parameter
to virFindFileInPath; it didn't crash prior to this patch
since it was building paths like /usr/bin/(null). This
is non-standard behavior, though, so I added a NULL
check at the beginning.
Signed-off-by: Chris Lalancette <clalance@redhat.com>
* src/util/util.c (virGetUserEnt): In the unlikely event that
sysconf(_SC_GETPW_R_SIZE_MAX) fails, don't use -1 as the size in
the subsequent allocation.
xen-unstable c/s 20762 bumped XEN_SYSCTL_INTERFACE_VERSION to 7. The
interface change does not affect libvirt, other than xenHypervisorInit()
failing since version 7 is not tried.
The attached patch accommodates the upcoming Xen 4.0 release by checking
for XEN_SYSCTL_INTERFACE_VERSION 7. If found, it sets
XEN_DOMCTL_INTERFACE_VERSION to 6, which is also new to Xen 4.0.
it causes a NULL-dereference on some systems like Solaris 10.
* src/node_device/node_device_linux_sysfs.c. Include <stdlib.h>.
(get_sriov_function): Use canonicalize_file_name, not realpath.
* bootstrap (modules): Add canonicalize-lgpl.
This fixes a segfault when the event handler is called after shutdown
when the global driver state is NULL again.
Also fix a locking issue in an error path.
virFileMakePath is a recursive function that was creates a buffer
PATH_MAX bytes long for each recursion (one recursion for each element
in the path). This changes it to have no buffers on the stack, and to
allocate just one buffer total, no matter how many elements are in the
path. Because the modified algorithm requires a char* to be passed in
rather than const char *, it is now 2 functions - a toplevel API
function that remains identical in function, and a 2nd helper function
called for the recursions, which 1) doesn't allocate anything, and 2)
takes a char* arg, so it can modify the contents.
* src/util/util.c: rewrite virFileMakePath
This reverts commit cdc42d0a48.
As DanB pointed out, this patch is actually wrong. The real
bug that was causing me to see this problem is a bug
introduced in a RHEL-5 libvirt snapshot, and I'm going to
fix the real bug there.
Signed-off-by: Chris Lalancette <clalance@redhat.com>
If you shutdown libvirtd while a domain with PCI
devices is running, then try to restart libvirtd,
libvirtd will crash.
This happens because qemuUpdateActivePciHostdevs() is calling
pciDeviceListSteal() with a dev of 0x0 (NULL), and then trying
to dereference it. This patch fixes it up so that
qemuUpdateActivePciHostdevs() steals the devices after first
Get()'ting them, avoiding the crash.
Signed-off-by: Chris Lalancette <clalance@redhat.com>
* src/qemu/qemu_monitor_text.c (qemuMonitorTextAttachDrive): Most other
failures in this function would "goto cleanup", but one mistakenly
returned directly, skipping the cleanup and resulting in a leak.
In addition, iterating the "try_command" loop would clobber, and
thus leak, the "cmd" allocated on the first iteration,
so be careful to free it in addition to "reply" beforehand.
The KVM build of QEMU includs the thread ID of each vCPU in the
'query-cpus' output. This is required for pinning guests to
particular host CPUs
* src/qemu/qemu_monitor_json.c: Extract 'thread_id' from CPU info
* src/util/json.c, src/util/json.h: Declare returned strings
to be const
* src/qemu/qemu_monitor.c: Wire up JSON mode for qemuMonitorGetPtyPaths
* src/qemu/qemu_monitor_json.c, src/qemu/qemu_monitor_json.h: Fix
const correctness. Add missing error message in the function
qemuMonitorJSONGetAllPCIAddresses. Add implementation of the
qemuMonitorGetPtyPaths function calling 'query-chardev'.
Two files were using functions from <sys/stat.h> but not including
in. Most of the time they got this automatically via another header,
but certain build flag combinations can reveal the problem
* src/lxc/lxc_container.c, src/node_device/node_device_linux_sysfs.c:
Add <sys/stat.h>
The <console> tag is supposed to result in addition of a single
<serial> device for HVM guests. The 'targetType' attribute was
missing though causing the compatibility code to add a second
<console> device
* src/conf/domain_conf.c: Set targetType for serial device
When libvirtd shuts down, it places a <state/> tag in the XML
state file it writes out for guests with PCI passthrough
devices. For devices that are attached at bootup time, the
state tag is empty. However, at libvirtd startup time, it
ignores anything with a <state/> tag in the XML, effectively
hiding the guest.
This patch remove the check for VIR_DOMAIN_XML_INTERNAL_STATUS
when parsing the XML.
* src/conf/domain_conf.c: remove VIR_DOMAIN_XML_INTERNAL_STATUS
flag check in virDomainHostdevSubsysPciDefParseXML()
Certain hypervisors (like qemu/kvm) map the PCI bar(s) on
the host when doing device passthrough. This can lead to a race
condition where the hypervisor is still cleaning up the device while
libvirt is trying to re-attach it to the host device driver. To avoid
this situation, we look through /proc/iomem, and if the hypervisor is
still holding onto the bar (denoted by the string in the matcher variable),
then we can wait around a bit for that to clear up.
v2: Thanks to review by DV, make sure we wait the full timeout per-device
Signed-off-by: Chris Lalancette <clalance@redhat.com>
The patches to add ACS checking to PCI device passthrough
introduced a bug. With the current code, if you try to
passthrough a device on the root bus (i.e. bus 0), then
it denies the passthrough. This is because the code in
pciDeviceIsBehindSwitchLackingACS() to check for a parent
device doesn't take into account the possibility of the
root bus. If we are on the root bus, it means we
legitimately can't find a parent, and it also means that
we don't have to worry about whether ACS is enabled.
Therefore return 0 (indicating we don't lack ACS) from
pciDeviceIsBehindSwitchLackingACS().
Signed-off-by: Chris Lalancette <clalance@redhat.com>
Fix a small problem with the qemu memory stats parsing algorithm. If qemu
reports a stat that libvirt does not recognize, skip past it so parsing can
continue. This corrects a potential infinite loop in the parsing code that can
only be triggered if new statistics are added to qemu.
* src/qemu/qemu_monitor_text.c: qemuMonitorParseExtraBalloonInfo add a
skip for extra ','
The loop looking for the controller associated with a SCI drive had
an off by one, causing it to miss the last controller.
* src/qemu/qemu_driver.c: Fix off-by-1 in searching for SCSI
drive hotplug
The hotplug code in QEMU was leaking memory because although the
inner device object was being moved into the main virDomainDefPtr
config object, the outer container virDomainDeviceDefPtr was not.
* src/qemu/qemu_driver.c: Clarify code to show that the inner
device object is owned by the main domain config upon
successfull attach.
Add the ability to turn off dynamic management of file permissions
for libvirt guests.
* qemu/libvirtd_qemu.aug: Support 'dynamic_ownership' flag
* qemu/qemu.conf: Document 'dynamic_ownership' flag.
* qemu/qemu_conf.c: Load 'dynamic_ownership' flag
* qemu/test_libvirtd_qemu.aug: Test 'dynamic_ownership' flag
The hotplug code was not correctly invoking the security driver
in error paths. If a hotplug attempt failed, the device would
be left with VM permissions applied, rather than restored to the
original permissions. Also, a CDROM media that is ejected was
not restored to original permissions. Finally there was a bogus
call to set hostdev permissions in the hostdev unplug code
* qemu/qemu_driver.c: Fix security driver usage in hotplug/unplug
If there is a problem with VM startup, PCI devices may be left
assigned to pci-stub / pci-back. Adding a call to reattach
host devices in the cleanup path is required.
* qemu/qemu_driver.c: qemuDomainReAttachHostDevices() when
VM startup fails
Remove all the QEMU driver calls for setting file ownership and
process uid/gid. Instead wire in the QEMU DAC security driver,
stacking it ontop of the primary SELinux/AppArmour driver.
* qemu/qemu_driver.c: Switch over to new DAC security driver
This new security driver is responsible for managing UID/GID changes
to the QEMU process, and any files/disks/devices assigned to it.
* qemu/qemu_conf.h: Add flag for disabling automatic file permission
changes
* qemu/qemu_security_dac.h, qemu/qemu_security_dac.c: New DAC driver
for QEMU guests
* Makefile.am: Add new files
Pulling the disk labelling code out of the exec hook, and into
libvirtd will allow it to access shared state in the daemon. It
will also make debugging & error reporting easier / more reliable.
* qemu/qemu_driver.c: Move initial disk labelling calls up into
libvirtd. Add cleanup of disk labels upon failure
If a VM fails to start, we can't simply free the security label
strings, we must call the domainReleaseSecurityLabel() method
otherwise the reserved 'mcs' level will be leaked in SElinux
* src/qemu/qemu_driver.c: Invoke domainReleaseSecurityLabel()
when domain fails to start
The current security driver architecture has the following
split of logic
* domainGenSecurityLabel
Allocate the unique label for the domain about to be started
* domainGetSecurityLabel
Retrieve the current live security label for a process
* domainSetSecurityLabel
Apply the previously allocated label to the current process
Setup all disk image / device labelling
* domainRestoreSecurityLabel
Restore the original disk image / device labelling.
Release the unique label for the domain
The 'domainSetSecurityLabel' method is special because it runs
in the context of the child process between the fork + exec.
This is require in order to set the process label. It is not
required in order to label disks/devices though. Having the
disk labelling code run in the child process limits what it
can do.
In particularly libvirtd would like to remember the current
disk image label, and only change shared image labels for the
first VM to start. This requires use & update of global state
in the libvirtd daemon, and thus cannot run in the child
process context.
The solution is to split domainSetSecurityLabel into two parts,
one applies process label, and the other handles disk image
labelling. At the same time domainRestoreSecurityLabel is
similarly split, just so that it matches the style. Thus the
previous 4 methods are replaced by the following 6 new methods
* domainGenSecurityLabel
Allocate the unique label for the domain about to be started
No actual change here.
* domainReleaseSecurityLabel
Release the unique label for the domain
* domainGetSecurityProcessLabel
Retrieve the current live security label for a process
Merely renamed for clarity.
* domainSetSecurityProcessLabel
Apply the previously allocated label to the current process
* domainRestoreSecurityAllLabel
Restore the original disk image / device labelling.
* domainSetSecurityAllLabel
Setup all disk image / device labelling
The SELinux and AppArmour drivers are then updated to comply with
this new spec. Notice that the AppArmour driver was actually a
little different. It was creating its profile for the disk image
and device labels in the 'domainGenSecurityLabel' method, where as
the SELinux driver did it in 'domainSetSecurityLabel'. With the
new method split, we can have consistency, with both drivers doing
that in the domainSetSecurityAllLabel method.
NB, the AppArmour changes here haven't been compiled so may not
build.
The QEMU driver is doing 90% of the calls to check for static vs
dynamic labelling. Except it is forgetting todo so in many places,
in particular hotplug is mistakenly assigning disk labels. Move
all this logic into the security drivers themselves, so the HV
drivers don't have to think about it.
* src/security/security_driver.h: Add virDomainObjPtr parameter
to virSecurityDomainRestoreHostdevLabel and to
virSecurityDomainRestoreSavedStateLabel
* src/security/security_selinux.c, src/security/security_apparmor.c:
Add explicit checks for VIR_DOMAIN_SECLABEL_STATIC and skip all
chcon() code in those cases
* src/qemu/qemu_driver.c: Remove all checks for VIR_DOMAIN_SECLABEL_STATIC
or VIR_DOMAIN_SECLABEL_DYNAMIC. Add missing checks for possibly NULL
driver entry points.
Allows the initiator to use a variety of IQNs rather than just the
system IQN when creating iSCSI pools.
* docs/schemas/storagepool.rng: extends the syntax with <iqn name="..."/>
* src/conf/storage_conf.[ch]: read and stores the iqn name
* src/storage/storage_backend_iscsi.[ch]: implement the IQN selection
when detected
* src/lxc/lxc_container.c src/lxc/lxc_controller.c src/lxc/lxc_driver.c
src/network/bridge_driver.c src/qemu/qemu_driver.c
src/uml/uml_driver.c: virFileMakePath returns 0 for success, or the
value of errno on failure, so error checking should be to test
if non-zero, not if lower than 0
Previously the uid/gid/mode in the xml was ignored when creating new
storage pool directories. This commit attempts to honor the requested
permissions, and spits out an error if it can't.
Note that when creating the directory, the rest of the path leading up
to the final element is created using current uid/gid/mode, and the
final element gets the settings from xml. It is NOT an error for the
directory to already exist; in this case, the perms for the existing
directory are just set (if necessary).
* src/storage/storage_backend_fs.c: update the virStorageBackendFileSystemBuild
function to check the directory hierarchy separately then create the
leaf directory with the right attributes
In order to avoid problems trying to chown files that were created by
root on a root-squashing nfs server, fork a new process that setuid's
to the desired uid before creating the file. (It's only done this way
if the pool containing the new volume is of type 'netfs', otherwise
the old method of creating the file followed by chown() is used.)
This changes the semantics of the "create_func" slightly - previously
it was assumed that this function just created the file, then the
caller would chown it to the desired uid. Now, create_func does both
operations.
There are multiple functions that can take on the role of create_func:
createFileDir - previously called mkdir(), now calls virDirCreate().
virStorageBackendCreateRaw - previously called open(),
now calls virFileCreate().
virStorageBackendCreateQemuImg - use virRunWithHook() to setuid/gid.
virStorageBackendCreateQcowCreate - same.
virStorageBackendCreateBlockFrom - preserve old behavior (but attempt
chown when necessary even if not root)
* src/storage/storage_backend.[ch] src/storage/storage_backend_disk.c
src/storage/storage_backend_fs.c src/storage/storage_backend_logical.c
src/storage/storage_driver.c: change the create_func implementations,
also propagate the pool information to be able to detect NETFS ones.
These functions create a new file or directory with the given
uid/gid. If the flag VIR_FILE_CREATE_AS_UID is given, they do this by
forking a new process, calling setuid/setgid in the new process, and
then creating the file. This is better than simply calling open then
fchown, because in the latter case, a root-squashing nfs server would
create the new file as user nobody, then refuse to allow fchown.
If VIR_FILE_CREATE_AS_UID is not specified, the simpler tactic of
creating the file/dir, then chowning is is used. This gives better
results in cases where the parent directory isn't on a root-squashing
NFS server, but doesn't give permission for the specified uid/gid to
create files. (Note that if the fork/setuid method fails to create the
file due to access privileges, the parent process will make a second
attempt using this simpler method.)
If the bit VIR_FILE_CREATE_ALLOW_EXIST is set in the flags, an
existing file/directory will not cause an error; in this case, the
function will simply set the permissions of the file/directory to
those requested. If VIR_FILE_CREATE_ALLOW_EXIST is not specified, an
existing file/directory is considered (and reported as) an error.
Return from both of these functions is 0 on success, or the value of
errno if there was a failure.
* src/util/util.[ch]: add the 2 new util functions
The test expected all environment variables copied in qemudBuildCommandLine
to have known values. So all of them have to be either set to a known value
or be unset. SDL_VIDEODRIVER and QEMU_AUDIO_DRV are not handled at all but
should be handled. Unset both, otherwise the test will fail if they are set
in the testing environment.
* src/qemu/qemu_conf.c: add a comment about copied environment variables
and qemuxml2argvtest
* tests/qemuxml2argvtest.c: unset SDL_VIDEODRIVER and QEMU_AUDIO_DRV
* src/conf/domain_conf.c (virDomainChrDefFormat): Plug a leak on
an error path, and at the same time, eliminate the need for a
"cleanup:" block. Before, the "return -1" after the switch
would leak an "addr" string. Now, by reversing the port,addr-
getting blocks we can free "addr" immediately and skip the goto.
The 'int virInterfaceIsActive()' method was directly returning the
value of the 'int active:1' bitfield in virIntefaceDefPtr. A bitfield
with a signed integer, will hold the values 0 and -1, not 0 and +1
as might be expected. This meant that virInterfaceIsActive() was
always returning -1 when the interface was active, not +1 & thus all
callers thought an error had occurred. To protect against this kind
of mistake again, change all bitfields to be unsigned ints
* daemon/libvirtd.h, src/conf/domain_conf.h, src/conf/interface_conf.h,
src/conf/network_conf.h: Change bitfields to unsigned int.
Invoking the virConnectGetCapabilities() method causes the QEMU
driver to rebuild its internal capabilities object. Unfortunately
it was forgetting to register the custom domain status XML hooks
again.
To avoid this kind of error in the future, the code which builds
capabilities is refactored into one single method, which can be
called from all locations, ensuring reliable rebuilds.
* src/qemu/qemu_driver.c: Fix rebuilding of capabilities XML and
guarentee it is always consistent
* src/util/logging.c (virLogMessage): Include "ignore-value.h".
Use it to ignore the return value of safewrite.
Use STDERR_FILENO, rather than "2".
* bootstrap (modules): Add ignore-value.
* gnulib: Update to latest, for ignore-value that is now LGPLv2+.
This was accomplished in xml parsing by doing away with the
stripped-down virInterfaceBareDef object, and just always using
virInterfaceDef, but with restrictions in certain places (eg, the type
of subordinate interface allowed in parsing depends on the parent
interface).
xml formatting was similarly adjusted. In addition, the formatting
functions keep track of the level of interface nesting, and insert
extra leading spaces on each line accordingly (using %*s).
The only change in formatted xml from previous (aside frmo supporting
new combinations of interface types) is that the subordinate ethernet
interfaces take up 2 lines rather than one, eg:
<interface type='ethernet' name='eth0'>
</interface>
instead of:
<interface type='ethernet' name='eth0'/>
I noticed some debug messages are printed with an empty lines after
them. This patch removes these empty lines from all invocations of the
following macros:
VIR_DEBUG
VIR_DEBUG0
VIR_ERROR
VIR_ERROR0
VIR_INFO
VIR_WARN
VIR_WARN0
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
New pciDeviceIsAssignable() function for checking whether a given PCI
device can be assigned to a guest was added. Currently it only checks
for ACS being enabled on all PCIe switches between root and the PCI
device. In the future, it could be the right place to check whether a
device is unbound or bound to a stub driver.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Currently CPU topology may only be specified together with CPU model:
<cpu match='exact'>
<model>name</model>
<topology sockets='1' cores='2' threads='3'/>
</cpu>
This patch allows for CPU topology specification without the need for
also specifying CPU model:
<cpu>
<topology sockets='1' cores='2' threads='3'/>
</cpu>
'match' attribute and 'model' element are made optional with the
restriction that 'match' attribute has to be set when 'model' is
present.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
When comparing x86 CPUs, features with 'disabled' policy were mistakenly
required to be supported by the host CPU.
Likewise, features with 'force' policy which were supported by host CPU
would make CPUs incompatible if 'strict' match was used by guest CPU.
This patch fixes both issues.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
QEMU's command line equivalent for the following domain XML fragment
<vcpus>2</vcpus>
<cpu ...>
...
<topology sockets='1' cores='2', threads='1'/>
</cpu>
is
-smp 2,sockets=1,cores=2,threads=1
This syntax was introduced in QEMU-0.12.
Version 2 changes:
- -smp argument build split into a separate function
- always add ",sockets=S,cores=C,threads=T" to -smp if qemu supports it
- use qemuParseCommandLineKeywords for command line parsing
Version 3 changes:
- ADD_ARG_LIT => ADD_ARG and line reordering in qemudBuildCommandLine
- rebased
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Current version expects name=value,... list and when an incorrect string
such as "a,b,c=d" would be parsed as "a,b,c" keyword with "d" value
without reporting any error, which is probably not the expected
behavior.
This patch adds an extra argument called allowEmptyValue, which if
non-zero will permit keywords with no value; "a,b=c,,d=" will be parsed
as follows:
keyword value
"a" NULL
"b" "c"
"" NULL
"d" ""
In case allowEmptyValue is zero, the string is required to contain
name=value pairs only; retvalues is guaranteed to contain non-NULL
pointers. Now, "a,b,c=d" will result in an error.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Replace
-balloon virtio
With
-device virtio-balloon-pci,id=balloon0,bus=pci.0,addr=0x3
This allows it to get correct assigned PCI address as declared in
previous patch
* src/qemu/qemu_conf.c: Convert Virtio ballon to -device and
give it an explicit PCI address
* tests/qemuxml2argvdata/qemuxml2argv-*args: Add in virtio balloon
where appropriate
Instead of relying on QEMU to assign PCI addresses and then querying
them with 'info pci', manually assign all PCI addresses before starting
the guest. These addresses are not stable across reboots. That will
come in a later patch
NB, the PIIX3 (IDE, FDC, ISA-Bridge) will always have slot 1 and
VGA will always have slot 2. We declare the Virtio Balloon gets
slot 3, and then all remaining slots are for configured devices.
* src/qemu/qemu_conf.c: If -device is supported, then assign all PCI
addresses when building the command line
* src/qemu/qemu_driver.c: Don't query monitor for PCI addresses if
they have already been assigned
* tests/qemuxml2argvdata/qemuxml2argv-hostdev-pci-address-device.args,
tests/qemuxml2argvdata/qemuxml2argv-net-virtio-device.args,
tests/qemuxml2argvdata/qemuxml2argv-sound-device.args,
tests/qemuxml2argvdata/qemuxml2argv-watchdog-device.args: Update
to include PCI slot/bus information
QEMU always configures a VGA card. If no video card is included in
the libvirt XML, it is neccessary to explicitly turn off the default
using -vga none
* src/qemu/qemu_conf.c: Pass -vga none if no video card is configured
* tests/qemuargv2xmltest.c, tests/qemuxml2argvtest.c: Test for
handling -vga none.
* tests/qemuxml2argvdata/qemuxml2argv-nographics-vga.args,
tests/qemuxml2argvdata/qemuxml2argv-nographics-vga.xml: Test
data files
Not all QEMU builds default to SDL graphics for their display.
Newer QEMU now has an explicit -sdl flag, which we can use to
explicitly request SDL intead of relying on the default. This
protects libvirt against unexpected changes in graphics default
* src/qemu/qemu_conf.c, src/qemu/qemu_conf.h: Probe for -sdl
flag and use it if it is found
* tests/qemuhelptest.c: Add SDL flag to tests
The old syntax was
-chardev SOMECONFIG
-nic user,guestfwd=tcp:IP:PORT-chardev:CHARDEV
The new syntax is
-chardev SOMECONFIG
-netdev user,guestfwd=tcp:IP:PORT,chardev=ID,id=user-ID
The old syntax was
-usbdevice host:PRODUCT:VENDOR
Or
-usbdevice host:BUS.DEV
The new syntax is
-device usb-host,product=PRODUCT,vendor=VENDOR
Or
-device usb-host,hostbus=BUS,hostaddr=DEV
The previous syntax was severely limited in its options
-usbdevice disk:/home/berrange/output.img
The new syntax is the same as for other disk types
-drive file=/home/berrange/output.img,if=none,id=usb-1,index=1
-device usb-storage,drive=usb-1
Again, the index= arg is wrong here, and will be removed in a
later merge
The current syntax uses a pair of args
-net nic,macaddr=52:54:00:56:6c:55,vlan=3,model=pcnet,name=pcnet.0
-net user,vlan=3,name=user.0
The new syntax does not need the vlan craziness anymore, and
so has a simplified pair of args
-netdev user,id=user.0
-device pcnet,netdev=user.0,id=pcnet.0,mac=52:54:00:56:6c:55,addr=<PCI SLOT>
The current preferred syntax for disk drives uses
-drive file=/vms/plain.qcow,if=virtio,index=0,boot=on,format=qcow
The new syntax splits this up into a pair of linked args
-drive file=/vms/plain.qcow,if=none,id=drive-virtio-0,format=qcow2
-device virtio-blk-pci,drive=drive-virtio-0,id=virtio-0,addr=<PCI SLOT>
SCSI/IDE devices also get a bus property linking them to the
controller
-device scsi-disk,drive=drive-scsi0-0-0,id=scsi0-0-0,bus=scsi0.0,scsi-id=0
-device ide-drive,drive=drive-ide0-0-0,id=ide0-0-0,bus=ide0,unit=0
The current syntax for audio devices is a horrible multiplexed
arg
-soundhw sb16,pcspk,ac97
The new syntax is
-device sb16,id=sound0
or
-device AC97,id=sound1,addr=<PCI SLOT>
NB, pcspk still uses the old -soundhw syntax
The current character device syntax uses either
-serial tty,path=/dev/ttyS2
Or
-chardev tty,id=serial0,path=/dev/ttyS2 -serial chardev:serial0
With the new -device support, we now prefer
-chardev file,id=serial0,path=/tmp/serial.log -device isa-serial,chardev=serial0
This patch changes the existing -chardev syntax to use this new
scheme, and fallbacks to the old plain -serial syntax for old
QEMU.
The monitor device changes to
-chardev socket,id=monitor,path=/tmp/test-monitor,server,nowait -mon chardev=monitor
In addition, this patch adds --nodefaults, which kills off the
default serial, parallel, vga and nic devices. THis avoids the
need for us to explicitly turn each off
When starting a guest, give every device a unique alias. This will
be used for the 'id' parameter in -device args in later patches.
It can also be used to uniquely identify devices in the monitor
For old QEMU without -device, assign disk names based on QEMU's
historical naming scheme.
* src/qemu/qemu_conf.c: Assign unique device aliases
* src/qemu/qemu_driver.c: Remove obsolete qemudDiskDeviceName
and use the device alias in eject & blockstats commands
* src/storage/storage_backend_fs.c (virStorageBackendFileSystemRefresh):
Correct parentheses. The documented intent is to ignore non-regular
files, yet due to a parenthesization error all errors were handled
that way.
Probe for the new -device flag and if available set the -nodefaults
flag, instead of using -net none, -serial none or -parallel none.
Other device types will be converted to use -device in later patches.
The -nodefaults flag will help avoid unwelcome surprises from future
QEMU releases
* src/qemu/qemu_conf.c: Probe for -device. Add -nodefaults flag.
Remove -net none, -serial none or -parallel none
* src/qemu/qemu_conf.h: Define QEMU_CMD_FLAG_DEVICE
* tests/qemuhelpdata/qemu-0.12.1: New data file for 0.12.1 QEMU
* tests/qemuhelptest.c: Test feature extraction from 0.12.1 QEMU
Although the serial, parallel, chanel, input & fs devices do
not have PCI address info, they can all have device aliases.
Thus it neccessary to associate the virDomainDeviceInfo data
with them all.
* src/conf/domain_conf.c, src/conf/domain_conf.h: Add hooks for
parsing / formatting device info for serial, parallel, channel
input and fs devices.
* docs/schemas/domain.rng: Associate device info with character
devices, input & fs device
This patch introduces the support for giving all devices a short,
unique name, henceforth known as a 'device alias'. These aliases
are not set by the end user, instead being assigned by the hypervisor
if it decides it want to support this concept.
The QEMU driver sets them whenever using the -device arg syntax
and uses them for improved hotplug/hotunplug. it is the intent
that other APIs (block / interface stats & device hotplug) be
able to accept device alias names in the future.
The XML syntax is
<alias name="video0"/>
This may appear in any type of device that supports device info.
* src/conf/domain_conf.c, src/conf/domain_conf.h: Add a 'alias'
field to virDomainDeviceInfo struct & parse/format it in XML
* src/libvirt_private.syms: Export virDomainDefClearDeviceAliases
* src/qemu/qemu_conf.c: Replace use of "nic_name" field with the
standard device alias
* src/qemu/qemu_driver.c: Clear device aliases at shutdown
The PCI device addresses are only valid while the VM is running,
since they are auto-assigned by QEMU. After shutdown they must
all be cleared. Future QEMU driver enhancement will allow for
persistent PCI address assignment
* src/conf/domain_conf.h, src/conf/domain_conf.c, src/libvirt_private.syms
Add virDomainDefClearPCIAddresses() method for wiping out auto assigned
PCI addresses
* src/qemu/qemu_driver.c: Clear PCI addresses at VM shutdown
Existing applications using libvirt are not aware of the disk
controller concept. Thus, after parsing the <disk> definitions
in the XML, it is neccessary to create <controller> elements
to satisfy all requested disks, as per their defined drive
addresses
* src/conf/domain_conf.c, src/conf/domain_conf.h,
src/libvirt_private.syms: Add virDomainDefAddDiskControllers()
method for populating disk controllers, and call it after
parsing disk definitions.
* src/qemu/qemu_conf.c: Call virDomainDefAddDiskControllers()
when doing ARGV -> XML conversion
* tests/qemuxml2argvdata/qemuxml2argv*.xml: Add disk controller
data to all data files which don't have it already
It is perfectly acceptable to have multiple sound devices of
same type in guest configuration. If the underlying hypervisor
does not like this, it is its job to complain, not the XML
parser's
* src/conf/domain_conf.c: Remove hack which deleted duplicated
sound device models.
* tests/xml2sexprdata/xml2sexpr-fv-sound.xml: Remove duplicate
models
Hotunplug of devices requires that we know their PCI address. Even
hotplug of SCSI drives, required that we know the PCI address of
the SCSI controller to attach the drive to. We can find this out
by running 'info pci' and then correlating the vendor/product IDs
with the devices we booted with.
Although this approach is somewhat fragile, it is the only viable
option with QEMU < 0.12, since there is no way for libvirto set
explicit PCI addresses when creating devices in the first place.
For QEMU > 0.12, this code will not be used.
* src/qemu/qemu_driver.c: Assign all dynamic PCI addresses on
startup of QEMU VM, matching vendor/product IDs
* src/qemu/qemu_monitor.c, src/qemu/qemu_monitor.h,
src/qemu/qemu_monitor_json.c, src/qemu/qemu_monitor_json.h,
src/qemu/qemu_monitor_text.c, src/qemu/qemu_monitor_text.h: Add
API for fetching PCI device address mapping
The current SCSI hotplug support attaches a brand new SCSI controller
for every disk. This is broken because the semantics differ from those
used when starting the VM initially. In the latter case, each SCSI
controller is filled before a new one is added.
If the user specifies an high drive index (sdazz) then at initial
startup, many intermediate SCSI controllers may be added with no
drives.
This patch changes SCSI hotplug so that it exactly matches the
behaviour of initial startup. First the SCSI controller number is
determined for the drive to be hotplugged. If any controller upto
and including that controller number is not yet present, it is
attached. Then finally the drive is attached to the last controller.
NB, this breaks SCSI hotunplug, because there is no 'drive_del'
command in current QEMU. Previous SCSI hotunplug was broken in
any case because it was unplugging the entire controller, not
just the drive in question.
A future QEMU will allow proper SCSI hotunplug of a drive.
This patch is derived from work done by Wolfgang Mauerer on disk
controllers.
* src/qemu/qemu_driver.c: Fix SCSI hotplug to add a drive to
the correct controller, instead of just attaching a new
controller.
* src/qemu/qemu_monitor.c, src/qemu/qemu_monitor.h,
src/qemu/qemu_monitor_json.c, src/qemu/qemu_monitor_json.h,
src/qemu/qemu_monitor_text.c, src/qemu/qemu_monitor_text.h: Add
support for 'drive_add' command
This patch allows for explicit hotplug/unplug of SCSI controllers.
Ordinarily this is not required, since QEMU/libvirt will attach
a new SCSI controller whenever one is required. Allowing explicit
hotplug of controllers though, enables the caller to specify a
static PCI address, instead of auto-assigning the next available
PCI slot. Or it will when we have static PCI addressing.
This patch is derived from Wolfgang Mauerer's disk controller
patch series.
* src/qemu/qemu_driver.c: Support hotplug & unplug of SCSI
controllers
* src/qemu/qemu_monitor.c, src/qemu/qemu_monitor.h,
src/qemu/qemu_monitor_json.c, src/qemu/qemu_monitor_json.h,
src/qemu/qemu_monitor_text.c, src/qemu/qemu_monitor_text.h: Add
new API for attaching PCI SCSI controllers
The latter is not officially "wrong", but *is* terribly anachronistic.
I think automake documentation or comments call that syntax obsolescent.
* cfg.mk (_makefile_at_at_check_exceptions): Exempt @SCHEMADIR@
and @SYSCONFDIR@ uses -- there are no Makefile variables for those.
* docs/Makefile.am: Use $(INSTALL), not @INSTALL@.
* examples/dominfo/Makefile.am: Similar.
* examples/domsuspend/Makefile.am: Similar.
* proxy/Makefile.am: Similar.
* python/Makefile.am: Similar.
* python/tests/Makefile.am: Similar.
* src/Makefile.am: Similar.
* tests/Makefile.am: Similar.
Until recently, some gnulib-generated replacement headers
included *other* headers that were not strictly necessary,
thus masking the need in this file for an explicit <stdlib.h>.
* src/util/util.c: Include <stdlib.h> for declarations of e.g.,
strtol, random_r, getenv, etc.
Current implementation of x86Decode() used for CPUID -> model+features
translation does not always select the closest CPU model. When walking
through all models from cpu_map.xml the function considers a new
candidate as a better choice than a previously selected candidate only
if the new one is a superset of the old one. In case the new candidate
is closer to host CPU but lacks some feature comparing to the old
candidate, the function does not choose well.
This patch changes the algorithm so that the closest model is always
selected. That is, the model which requires the lowest number of
additional features to describe host CPU.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
qemudFindCharDevicePTYsMonitor reports an error if 'info chardev' didn't
provide information for a requested device, even if the log output parsing
had found the pty path for that device. This makes pty assignment fail for
older QEMU/KVM versions. For example KVM 72 on Debian doesn't support
'info chardev', so qemuMonitorTextGetPtyPaths cannot parse any useful
information and the hash for device-id-to-pty-path mapping stays empty.
Make qemudFindCharDevicePTYsMonitor report an error only if the log output
parsing and the 'info chardev' parsing failed to provide the pty path.
* src/conf/domain_conf.c: add defaults for the video device
* src/esx/esx_vmx.[ch]: add VNC support to the VMX handling
* tests/vmx2xmltest.c, tests/xml2vmxtest.c: add tests for the VNC support
The current code for using -drive simply sets the -drive 'index'
parameter. QEMU internally converts this to bus/unit depending
on the type of drive. This does not give us precise control over
the bus/unit assignment though. This change switches over to make
libvirt explicitly calculate the bus/unit number.
In addition bus/unit/index are actually irrelevant for VirtIO
disks, since each virtio disk is a separate PCI device. No disk
controller is involved.
Doing the conversion to bus/unit in libvirt allows us to correctly
attach SCSI controllers when required.
* src/qemu/qemu_conf.c: Specify bus/unit instead of index for
disks
* tests/qemuxml2argvdata/qemuxml2argv-disk*.args: Switch over from
using index=NNNN, to bus=NN, unit=NN for SCSI/IDE/Floppy disks
To enable it to be called from multiple locations, split out
the code for building the -drive arg string. This will be needed
by later patches which do drive hotplug, the conversion to use
-device, and the conversion to controller/bus/unit addressing
* src/qemu/qemu_conf.c, src/qemu/qemu_conf.h: Add qemuBuildDriveStr
for building -drive arg string
Convert the QEMU monitor APIs over to use virDomainDeviceAddress
structs for passing addresses in/out, instead of individual bits.
This makes the number of parameters smaller & easier to deal with.
No functional change
* src/qemu/qemu_driver.c, src/qemu/qemu_monitor.c,
src/qemu/qemu_monitor.h, src/qemu/qemu_monitor_text.c,
src/qemu/qemu_monitor_text.h: Change monitor hotplug APIs to
take an explicit address ptr for all host/guest addresses
This augments virDomainDevice with a <controller> element
that is used to represent disk controllers (e.g., scsi
controllers). The XML format is given by
<controller type="scsi" index="<num>">
<address type="pci" domain="0xNUM" bus="0xNUM" slot="0xNUM"/>
</controller>
where type denotes the disk interface (scsi, ide,...), index
is an integer that identifies the controller for association
with disks, and the <address> element specifies the controller
address on the PCI bus as described in previous commits
The address element can be omitted; in this case, an address
will be assigned automatically.
Most of the code in this patch is from Wolfgang Mauerer's
previous disk controller series
* docs/schemas/domain.rng: Define syntax for <controller>
XML element
* src/conf/domain_conf.c, src/conf/domain_conf.h: Define
virDomainControllerDef struct, and routines for parsing
and formatting XML
* src/libvirt_private.syms: Add virDomainControllerInsert
and virDomainControllerDefFree
When parsing the <disk> element specification, if no <address>
is provided for the disk, then automatically assign one based on
the <target dev='sdXX'/> device name. This provides for backwards
compatability with existing applications using libvirt, while also
allowing new apps to have complete fine grained control.
* src/conf/domain_conf.h, src/conf/domain_conf.c,
src/libvirt_private.syms: Add virDomainDiskDefAssignAddress()
for assigning a controller/bus/unit address based on disk target
* src/qemu/qemu_conf.c: Call virDomainDiskDefAssignAddress() after
generating XML from ARGV
* tests/qemuxml2argvdata/*.xml: Add in drive address information
to all XML files
Add the virDomainDeviceAddress information to the sound, video
and watchdog devices. This means all of them gain the new XML
element
<address .... />
This brings them upto par with disk/net/hostdev devices which
already have address info
* src/conf/domain_conf.h: Add virDomainDeviceAddress to sound,
video & watchdog device struts.
* src/conf/domain_conf.c: Hook up parsing/formatting for
virDomainDeviceAddress in sound, video & watchdog devices
* docs/schemas/domain.rng: Associate device address info
with sound, video & watchdog
Introduce a new structure
struct _virDomainDeviceDriveAddress {
unsigned int controller;
unsigned int bus;
unsigned int unit;
};
and plug that into virDomainDeviceAddress and generates XML that
looks like
<address type='drive' controller='1' bus='0' unit='5'/>
This syntax will be used by the QEMU driver to explicitly control
how drives are attached to the bus
* src/conf/domain_conf.h, src/conf/domain_conf.c: Parsing and
formatting of drive addresses
* docs/schemas/domain.rng: Define new address format for drives
All guest devices now use a common device address structure
summarized by:
enum virDomainDeviceAddressType {
VIR_DOMAIN_DEVICE_ADDRESS_TYPE_NONE,
VIR_DOMAIN_DEVICE_ADDRESS_TYPE_PCI,
};
struct _virDomainDevicePCIAddress {
unsigned int domain;
unsigned int bus;
unsigned int slot;
unsigned int function;
};
struct _virDomainDeviceInfo {
int type;
union {
virDomainDevicePCIAddress pci;
} addr;
};
This replaces the anonymous structs in Disk/Net/Hostdev data
structures. Where available, the address is *always* printed
in the XML file, instead of being hidden in the internal state
file.
<address type='pci' domain='0x0000' bus='0x1e' slot='0x07' function='0x0'/>
The structure definition is based on Wolfgang Mauerer's disk
controller patch series.
* docs/schemas/domain.rng: Define the <address> syntax and
associate it with disk/net/hostdev devices
* src/conf/domain_conf.h, src/conf/domain_conf.c,
src/libvirt_private.syms: APIs for parsing/formatting address
information. Also remove the QEMU specific 'pci_addr' attributes
* src/qemu/qemu_driver.c: Replace use of 'pci_addr' attrs with
new standardized format.
With the introduction virDispatchError, hook function errors are
never sent through the error callback, so users will never see
these messages.
Fix this by calling virDispatchError after hook failure.
Based off how QEMU does it, look through /sys/bus/usb/devices/* for
matching vendor:product info, and if found, use info from the surrounding
files to build the device's /dev/bus/usb path.
This fixes USB device assignment by vendor:product when running qemu
as non-root (well, it should, but for some reason I couldn't reproduce
the failure people are seeing in [1], but it appears to work properly)
[1] https://bugzilla.redhat.com/show_bug.cgi?id=542450
udev doesn't prefix USB product/vendor info with '0x', so the
strtol conversions were wrong for the product field (vendor already
set the correct base). Make the change for PCI product/vendor as
well to be safe.
This fixes USB device assignment via virt-manager.
This allows debug statements and raised errors in hook functions to
actually be logged somewhere (stderr). Users can enable debugging in the
daemon and now see more info in /var/log/libvirt/...
Upstream xen has changed parameters to the migration operation
several times over the past 18 months. Changeset 17553 removed
the resouce parameter, Changesets 17709, 17753, and 20326 added
ssl, node, and change_home_server parameters respectively.
Fortunately, testing has revealed that xend will fail the
operation if a parameter is missing but happily honor it if
unknown parameters are provided. Thus all currently supported
parameters can be provided, satisfying current xend but not
regressing older versions.
The virRaiseErrorFull() may invoke the error handler callback
functions an application has registered. This is not good
because the connection object may not be available at this
point, and the caller may be holding locks. This creates a
problem if the error handler calls back into libvirt.
The solutuon is to move invocation of the handler into the
final cleanup code in the public API entry points, where it
is guarenteed to have safe state.
* src/libvirt.c: Invoke virDispatchError() in all error paths
* src/util/virterror.c: Remove virSetConnError/virSetGlobalError,
replacing with virDispatchError(). Move invocation of the
error callbacks into virDispatchError() instead of the
virRaiseErrorFull function which is not in a safe context
