'const fooPtr' is the same as 'foo * const' (the pointer won't
change, but it's contents can). But in general, if an interface
is trying to be const-correct, it should be using 'const foo *'
(the pointer is to data that can't be changed).
Fix up offenders in src/qemu.
* src/qemu/qemu_bridge_filter.h (networkAllowMacOnPort)
(networkDisallowMacOnPort): Use intended type.
* src/qemu/qemu_bridge_filter.c (networkAllowMacOnPort)
(networkDisallowMacOnPort): Likewise.
* src/qemu/qemu_command.c (qemuBuildTPMBackendStr)
(qemuBuildTPMDevStr, qemuBuildCpuArgStr)
(qemuBuildObsoleteAccelArg, qemuBuildMachineArgStr)
(qemuBuildSmpArgStr, qemuBuildNumaArgStr): Likewise.
* src/qemu/qemu_conf.c (qemuSharedDeviceEntryCopy): Likewise.
* src/qemu/qemu_driver.c (qemuDomainSaveImageStartVM): Likewise.
* src/qemu/qemu_hostdev.c
(qemuDomainHostdevNetConfigVirtPortProfile): Likewise.
* src/qemu/qemu_monitor_json.c
(qemuMonitorJSONAttachCharDevCommand): Likewise.
Signed-off-by: Eric Blake <eblake@redhat.com>
virDomainChrGetDomainPtrs() required 4 levels of pointers (taking
a parameter that will be used as an output variable to return the
address of another variable that contains an array of pointers).
This is rather complex to reason about, especially when outside
of the domain_conf file, no other caller should be modifying
the resulting array of pointers directly. Changing the public
signature gives something is easier to reason with, and actually
make const-correct; which is important as it was the only function
that was blocking virDomainDeviceDefCopy from treating its source
as const.
* src/conf/domain_conf.h (virDomainChrGetDomainPtrs): Use simpler
types, and make const-correct for external users.
* src/conf/domain_conf.c (virDomainChrGetDomainPtrs): Split...
(virDomainChrGetDomainPtrsInternal): ...into an internal version
that lets us modify terms, vs. external form that is read-only.
(virDomainDeviceDefPostParseInternal, virDomainChrFind)
(virDomainChrInsert): Adjust callers.
* src/qemu/qemu_command.c (qemuGetNextChrDevIndex): Adjust caller.
(qemuDomainDeviceAliasIndex): Make const-correct.
Signed-off-by: Eric Blake <eblake@redhat.com>
Prefer using VFIO (if available) to the legacy KVM device passthrough.
With this patch a PCI passthrough device without the driver configured
will be started with VFIO if it's available on the host. If not legacy
KVM passthrough is checked and error is reported if it's not available.
This resolves:
https://bugzilla.redhat.com/show_bug.cgi?id=1012824https://bugzilla.redhat.com/show_bug.cgi?id=1012834
Note that a similar problem was reported in:
https://bugzilla.redhat.com/show_bug.cgi?id=827519
but the fix only worked for <interface type='hostdev'>, *not* for
<interface type='network'> where the network itself was a pool of
hostdevs.
The symptom in both cases was this error message:
internal error: Unable to determine device index for network device
In both cases the cause was lack of proper handling for netdevs
(<interface>) of type='hostdev' when scanning the netdev list looking
for alias names in qemuAssignDeviceNetAlias() - those that aren't
type='hostdev' have an alias of the form "net%d", while those that are
hostdev use "hostdev%d". This special handling was completely lacking
prior to the fix for Bug 827519 which was:
When searching for the highest alias index, libvirt looks at the alias
for each netdev and if it is type='hostdev' it ignores the entry. If
the type is not hostdev, then it expects the "net%d" form; if it
doesn't find that, it fails and logs the above error message.
That fix works except in the case of <interface type='network'> where
the network uses hostdev (i.e. the network is a pool of VFs to be
assigned to the guests via PCI passthrough). In this case, the check
for type='hostdev' would fail because it was done as:
def->net[i]->type == VIR_DOMAIN_NET_TYPE_HOSTDEV
(which compares what was written in the config) when it actually
should have been:
virDomainNetGetActualType(def->net[i]) == VIR_DOMAIN_NET_TYPE_HOSTDEV
(which compares the type of netdev that was actually allocated from
the network at runtime).
Of course the latter wouldn't be of any use if the netdevs of
type='network' hadn't already acquired their actual network connection
yet, but manual examination of the code showed that this is never the
case.
While looking through qemu_command.c, two other places were found to
directly compare the net[i]->type field rather than getting actualType:
* qemuAssignDeviceAliases() - in this case, the incorrect comparison
would cause us to create a "net%d" alias for a netdev with
type='network' but actualType='hostdev'. This alias would be
subsequently overwritten by the proper "hostdev%d" form, so
everything would operate properly, but a string would be
leaked. This patch also fixes this problem.
* qemuAssignDevicePCISlots() - would defer assigning a PCI address to
a netdev if it was type='hostdev', but not for type='network +
actualType='hostdev'. In this case, the actual device usually hasn't
been acquired yet anyway, and even in the case that it has, there is
no practical difference between assigning a PCI address while
traversing the netdev list or while traversing the hostdev
list. Because changing it would be an effective NOP (but potentially
cause some unexpected regression), this usage was left unchanged.
If qemuParseCommandLine finds an arg it does not understand
it adds it to the QEMU passthrough custom arg list. If the
qemuParseCommandLine method hits an error for any reason
though, it just does 'VIR_FREE(cmd)' on the custom arg list.
This means all actual args / env vars are leaked. Introduce
a qemuDomainCmdlineDefFree method to be used for cleanup.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
If the call to virDomainControllerInsert fails in
qemuParseCommandLine, the controller struct is leaked.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The 'qemuStringToArgvEnv' method splits up a string of command
line env/args to an 'arglist' array. It then copies env vars
to a 'progenv' array and args to a 'progargv' array. When
copyin the env vars, it NULL-ifies the element in 'arglist'
that is copied.
Upon OOM the 'virStringListFree' is called on progenv and
arglist. Unfortunately, because the elements in 'arglist'
related to env vars have been set to NULL, the call to
virStringListFree(arglist) doesn't free anything, even
though some non-NULL args vars still exist later in the
array.
To fix this leak, stop NULL-ifying the 'arglist' elements,
and change the cleanup code to only free elements in the
'arglist' array, not 'progenv'.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
In a number of places in qemuParseCommandLineDisk, an error
is reported, but no 'goto error' jump is used. This causes
failure to report OOM conditions to the caller.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
If OOM occurs in qemuParseCommandLineDisk some intermediate
variables will be leaked when parsing Sheepdog or RBD disks.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The qemuBuildCommandLine code for parsing sound cards will leak
an intermediate variable if an OOM occurs. Move the free'ing of
the variable earlier to avoid the leak.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
In qemuParseNBDString, if the virURIParse fails, the
error is not reported to the caller. Instead execution
falls through to the non-URI codepath causing memory
leaks later on.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
If qemuAddRBDHost fails due to parsing problems or OOM, then
qemuParseRBDString cleanup is skipped causing a memory leak.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
qemuDomainPCIAddressGetNextSlot has a loop for finding
compatible PCI buses. In the loop body it creates a
PCI address string, but never frees this. This causes
a leak if the loop executes more than one iteration,
or if a call in the loop body fails.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
This resolves one of the issues listed in:
https://bugzilla.redhat.com/show_bug.cgi?id=1003983
00:1E.0 is the location of this controller on at least some actual Q35
hardware, so we try to replicate the placement. The bridge should work
just as well in any other location though, so if 00:1E.0 isn't
available, just allow it to be auto-assigned anywhere appropriate.
This resolves one of the issues in:
https://bugzilla.redhat.com/show_bug.cgi?id=1003983
This device is identical to qemu's "intel-hda" device (known as "ich6"
in libvirt), but has a different PCI device ID (which matches the ID
of the hda audio built into the ich9 chipset, of course). It's not
supported in earlier versions of qemu, so it requires a capability
bit.
I'm not sure why this code was written to compare the strings that it
had just retrieved from an enum->string conversion, rather than just
look at the original enum values, but this yields the same results,
and is much more efficient (especially as you add more devices).
This is a prerequisite for patches to resolve:
https://bugzilla.redhat.com/show_bug.cgi?id=1003983
Part of the resolution to:
https://bugzilla.redhat.com/show_bug.cgi?id=1003983
Although most devices available in qemu area defined as PCI devices,
and strictly speaking should only be attached via a PCI slot, in
practice qemu allows them to be attached to a PCIe slot and sometimes
this makes sense.
For example, The UHCI and EHCI USB controllers are usually attached
directly to the PCIe "root complex" (i.e. PCIe slots) on real
hardware, so that should be possible for a Q35-based qemu virtual
machine as well.
We still want to prefer a standard PCI slot when auto-assigning
addresses, though, and in general to disallow attaching PCI devices
via PCIe slots.
This patch makes that possible by adding a new
QEMU_PCI_CONNECT_TYPE_EITHER_IF_CONFIG flag. Three things are done
with this flag:
1) It is set for the "pcie-root" controller
2) qemuCollectPCIAddress() now has a set of nested switches that set
this "EITHER" flag for devices that we want to allow connecting to
pcie-root when specifically requested in the config.
3) qemuDomainPCIAddressFlagsCompatible() adds this new flag to the
"flagsMatchMask" if the address being checked came from config rather
than being newly auto-allocated by libvirt (this knowledge is
conveniently already available in the "fromConfig" arg).
Now any device having the EITHER flag set can be connected to
pcie-root if explicitly requested, but auto-allocated addresses for
those devices will still be standard PCI slots instead.
This patch only loosens the restrictions on devices that have been
specifically requested, but the setup is such that it should be fairly
easy to add new devices.
Replace them with switch cases. This will make it more efficient when
we add exceptions for more controller types, and other device types.
This is a prerequisite for patches to resolve:
https://bugzilla.redhat.com/show_bug.cgi?id=1003983
The parsing of '-usb' did not check for failure of the
virDomainControllerInsert method. As a result on OOM, the
parser mistakenly attached USB disks to the IDE controller.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The code formatting NUMA args was ignoring the return value
of virBitmapFormat, so on OOM, it would silently drop the
NUMA cpumask arg.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
When building boot menu args, if OOM occurred the CLI args
would end up containing 'order=(null)' due to a missing
call to 'virBufferError'.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The qemuParseCommandLine method did not check the return value of
virStringSplit to see if OOM had occurred. This lead to dereference
of a NULL pointer on OOM.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Most callers of qemuParseKeywords were assigning its return
value to a 'size_t' variable. Then then also checked '< 0'
for error condition, but this will never be true with the
unsigned size_t variable. Rather than using 'ssize_t', change
qemuParseKeywords so that the element count is returned via
an output parameter, leaving the return value solely as an
error indicator.
This avoids a crash accessing beyond the end of an error
upon OOM.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
In
commit 41b5505679
Author: Eric Blake <eblake@redhat.com>
Date: Wed Aug 28 15:01:23 2013 -0600
qemu: simplify list cleanup
The qemuStringToArgvEnv method was changed to use virStringFreeList
to free the 'arglist' array. This method assumes the string list
array is NULL terminated, however, qemuStringToArgvEnv was not
ensuring this when populating 'arglist'. This caused an out of
bounds access by virStringFreeList when OOM occured in the initial
loop of qemuStringToArgvEnv
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
When parsing the RBD hosts, it increments the 'nhosts' counter
before increasing the 'hosts' array allocation. If an OOM then
occurs when increasing the array allocation, the cleanup block
will attempt to access beyond the end of the array. Switch
to using VIR_EXPAND_N instead of VIR_REALLOC_N to protect against
this mistake
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
If OOM occurs in qemuDomainCCWAddressSetCreate, it jumps to
a cleanup block and frees the partially initialized object.
It then mistakenly returns the address of the just free'd
pointer instead of NULL.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
This resolves https://bugzilla.redhat.com/show_bug.cgi?id=1008903
The Q35 machinetype has an implicit SATA controller at 00:1F.2 which
isn't given the "expected" id of ahci0 by qemu when it's created. The
original suggested solution to this problem was to not specify any
controller for the disks that use the default controller and just
specify "unit=n" instead; qemu should then use the first IDE or SATA
controller for the disk.
Unfortunately, this "solution" is ignorant of the fact that in the
case of SATA disks, the "unit" attribute in the disk XML is actually
*not* being used for the unit, but is instead used to specify the
"bus" number; each SATA controller has 6 buses, and each bus only
allows a single unit. This makes it nonsensical to specify unit='n'
where n is anything other than 0. It also means that the only way to
connect more than a single device to the implicit SATA controller is
to explicitly give the bus names, which happen to be "ide.$n", where
$n can be replaced by the disk's "unit" number.
qemu/KVM also supports a tftp URL while specifying the cdrom ISO image.
The xml should be as following:
<disk type='network' device='cdrom'>
<source protocol='tftp' name='/url/path'>
<host name='host.name' port='69'/>
</source>
</disk>
Signed-off-by: Aline Manera <alinefm@br.ibm.com>
The ftps protocol is another protocol supported by qemu/KVM while specifying
the cdrom ISO image.
The xml should be as following:
<disk type='network' device='cdrom'>
<source protocol='ftps' name='/url/path'>
<host name='host.name' port='990'/>
</source>
</disk>
Signed-off-by: Aline Manera <alinefm@br.ibm.com>
The https protocol is also accepted by qemu/KVM when specifying the cdrom ISO
image.
The xml should be as following:
<disk type='network' device='cdrom'>
<source protocol='https' name='/url/path'>
<host name='host.name' port='443'/>
</source>
</disk>
Signed-off-by: Aline Manera <alinefm@br.ibm.com>
Currently, only X86 provides users CPU features with CPUID instruction.
If users specify the features for non-x86, it should tell users to
remove them.
This patch is to report one error if features are specified by
users for non-x86 platform.
Signed-off-by: Li Zhang <zhlcindy@linux.vnet.ibm.com>
In Fedora 19, 'qemu-kvm' is a simple wrapper that calls
'qemu-system-x86_64 -machine accel=kvm'. Attempting
to use 'virsh qemu-attach $pid' to a machine started as:
qemu-kvm -cdrom /var/lib/libvirt/images/foo.img \
-monitor unix:/tmp/demo,server,nowait -name foo \
--uuid cece4f9f-dff0-575d-0e8e-01fe380f12ea
was failing with:
error: XML error: No PCI buses available
because we did not see 'kvm' in the executable name read from
/proc/$pid/cmdline, and tried to assign os.machine as
"accel=kvm" instead of "pc"; this in turn led to refusal to
recognize the pci bus.
Noticed while investigating https://bugzilla.redhat.com/995312
although there are still other issues to fix before that bug
will be completely solved.
I've concluded that the existing parser code for native-to-xml
is a horrendous hodge-podge of ad-hoc approaches; I basically
rewrote the -machine section to be a bit saner.
* src/qemu/qemu_command.c (qemuParseCommandLine): Don't assume
-machine argument is always appropriate for os.machine; set
virtType if accel is present.
Signed-off-by: Eric Blake <eblake@redhat.com>
'virsh domxml-from-native' and 'virsh qemu-attach' could misbehave
for an emulator installed in (a somewhat unlikely) location
such as /usr/local/qemu-1.6/qemu-system-x86_64 or (an even less
likely) /opt/notxen/qemu-system-x86_64. Limit the strstr seach
to just the basename of the file where we are assuming details
about the binary based on its name.
While testing, I accidentally triggered a core dump during strcmp
when I forgot to set os.type on one of my code paths; this patch
changes such a coding error to raise a nicer internal error instead.
* src/qemu/qemu_command.c (qemuParseCommandLine): Compute basename
earlier.
* src/conf/domain_conf.c (virDomainDefPostParseInternal): Avoid
NULL deref.
Signed-off-by: Eric Blake <eblake@redhat.com>
CPU features are not supported on non-x86 and hasFeatures will be NULL.
This patch is to remove CPU features functions calling to avoid errors.
Signed-off-by: Li Zhang <zhlcindy@linux.vnet.ibm.com>
The VIR_FREE() macro will cast away any const-ness. This masked a
number of places where we passed a 'const char *' string to
VIR_FREE. Fortunately in all of these cases, the variable was not
in fact const data, but a heap allocated string. Fix all the
variable declarations to reflect this.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
No need to open code now that we have a nice function.
Interestingly, our virStringFreeList function is typed correctly
(a malloc'd list of malloc'd strings is NOT const, whether at the
point where it is created, or at the point where it is cleand up),
so using it with a 'const char **' argument would require a cast
to keep the compiler. I chose instead to remove const from code
even where we don't modify the argument, just to avoid the need
to cast.
* src/qemu/qemu_command.h (qemuParseCommandLine): Drop declaration.
* src/qemu/qemu_command.c (qemuParseProcFileStrings)
(qemuStringToArgvEnv): Don't force malloc'd result to be const.
(qemuParseCommandLinePid, qemuParseCommandLineString): Simplify
cleanup.
(qemuParseCommandLine, qemuFindEnv): Drop const-correctness to
avoid the need to cast in callers.
Signed-off-by: Eric Blake <eblake@redhat.com>
Currently, kernel supports up to 8 queues for a multiqueue tap device.
However, if user tries to enter a huge number (e.g. one million) the tap
allocation fails, as expected. But what is not expected is the log full
of warnings:
warning : virFileClose:83 : Tried to close invalid fd 0
The problem is, upon error we iterate over an array of FDs (handlers to
queues) and VIR_FORCE_CLOSE() over each item. However, the array is
pre-filled with zeros. Hence, we repeatedly close stdin. Ouch.
But there's more. The queues allocation is done in virNetDevTapCreate()
which cleans up the FDs in case of error. Then, its caller, the
virNetDevTapCreateInBridgePort() iterates over the FD array and tries to
close them too. And so does qemuNetworkIfaceConnect() and
qemuBuildInterfaceCommandLine().
Starting with qemu 1.6, the qemu-system-arm vexpress-a9 model has a
hardcoded virtio-mmio transport which enables attaching all virtio
devices.
On the command line, we have to use virtio-XXX-device rather than
virtio-XXX-pci, thankfully s390 already set the precedent here so
it's fairly straight forward.
At the XML level, this adds a new device address type virtio-mmio.
The controller and addressing don't have any subelements at the
moment because we they aren't needed for this usecase, but could
be added later if needed.
Add a test case for an ARM guest with one of every virtio device
enabled.
Similar to the chardev bit, ARM boards depend on the old style '-net nic'
for actually instantiating net devices. But we can't block out
-netdev altogether since it's needed for upcoming virtio support.
And add tests for working ARM XML with console, disk, and networking.
This corresponds to '-sd' and '-drive if=sd' on the qemu command line.
Needed for many ARM boards which don't provide any other way to
pass in storage.
QEMU ARM boards don't give us any way to explicitly wire in
a -chardev, so use the old style -serial options.
Unfortunately this isn't as simple as just turning off the CHARDEV flag
for qemu-system-arm, as upcoming virtio support _will_ use device/chardev.
On my machine, a guest fails to boot if it has a sound card, but not
graphical device/display is configured, because pulseaudio fails to
initialize since it can't access $HOME.
A workaround is removing the audio device, however on ARM boards there
isn't any option to do that, so -nographic always fails.
Set QEMU_AUDIO_DRV=none if no <graphics> are configured. Unfortunately
this has massive test suite fallout.
Add a qemu.conf parameter nographics_allow_host_audio, that if enabled
will pass through QEMU_AUDIO_DRV from sysconfig (similar to
vnc_allow_host_audio)
Add an attribute named 'removable' to the 'target' element of disks,
which controls the removable flag. For instance, on a Linux guest it
controls the value of /sys/block/$dev/removable. This option is only
valid for USB disks (i.e. bus='usb'), and its default value is 'off',
which is the same behaviour as before.
To achieve this, 'removable=on' (or 'off') is appended to the '-device
usb-storage' parameter sent to qemu when adding a USB disk via
'-disk'. A capability flag QEMU_CAPS_USB_STORAGE_REMOVABLE was added
to keep track if this option is supported by the qemu version used.
Bug: https://bugzilla.redhat.com/show_bug.cgi?id=922495
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Allow use of the usb-storage device only if the new capability flag
QEMU_CAPS_DEVICE_USB_STORAGE is set, which it is for qemu(-kvm)
versions >= 0.12.1.2-rhel62-beta.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
vhost only works in KVM mode at the moment, and is infact compiled
out if the emulator is built for non-native architecture. While it
may work at some point in the future for plain qemu, for now it's
just noise on the command line (and which contributes to arm cli
breakage).
QEMU commit 3984890 introduced the "pci-hole64-size" property,
to i440FX-pcihost and q35-pcihost with a default setting of 2 GB.
Translate <pcihole64>x<pcihole64/> to:
-global q35-pcihost.pci-hole64-size=x for q35 machines and
-global i440FX-pcihost.pci-hole64-size=x for i440FX-based machines.
Error out on other machine types or if the size was specified
but the pcihost device lacks 'pci-hole64-size' property.
https://bugzilla.redhat.com/show_bug.cgi?id=990418
The ftp protocol is already recognized by qemu/KVM so add this support to
libvirt as well.
The xml should be as following:
<disk type='network' device='cdrom'>
<source protocol='ftp' name='/url/path'>
<host name='host.name' port='21'/>
</source>
</disk>
Signed-off-by: Aline Manera <alinefm@br.ibm.com>
QEMU/KVM already allows a HTTP URL for the cdrom ISO image so add this support
to libvirt as well.
The xml should be as following:
<disk type='network' device='cdrom'>
<source protocol='http' name='/url/path'>
<host name='host.name' port='80'/>
</source>
</disk>
Signed-off-by: Aline Manera <alinefm@br.ibm.com>
If user requested multiqueue networking, beside multiple /dev/tap and
/dev/vhost-net openings, we forgot to pass mq=on onto the -device
virtio-net-pci command line. This is advised at:
http://www.linux-kvm.org/page/Multiqueue#Enable_MQ_feature
Each of the modules handled reporting error messages from the secret fetching
slightly differently with respect to the error. Provide a similar message
for each error case and provide as much data as possible.
If there's no hard_limit set and domain uses VFIO we still must lock the
guest memory (prerequisite from qemu). Hence, we should compute the
amount to be locked from max_balloon.
This function is to guess the correct limit for maximal memory
usage by qemu for given domain. This can never be guessed
correctly, not to mention all the pains and sleepless nights this
code has caused. Once somebody discovers algorithm to solve the
Halting Problem, we can compute the limit algorithmically. But
till then, this code should never see the light of the release
again.
This patch addresses two concerns with the error reporting when an
incompatible PCI address is specified for a device:
1) It wasn't always apparent which device had the problem. With this
patch applied, any error about an incompatible address will always
contain the full address as given in the config, so it will be easier
to determine which device's config aused the problem.
2) In some cases when the problem came from bad config, the error
message was erroneously classified as VIR_ERR_INTERNAL_ERROR. With
this patch applied, the same error message will be changed to indicate
either "internal" or "xml" error depending on whether the address came
from the config, or was automatically generated by libvirt.
Note that in the case of "internal" (due to bad auto-generation)
errors, the PCI address won't be of much use in finding the location
in config to change (because it was automatically generated). Of
course that makes perfect sense, but still the address could provide a
clue about a bug in libvirt attempting to use a type of pci bus that
doesn't have its flags set correctly (or something similar). In other
words, it's not perfect, but it is definitely better.
q35 machines have an implicit ahci (sata) controller at 00:1F.2 which
has no "id" associated with it. For this reason, we can't refer to it
as "ahci0". Instead, we don't give an id on the commandline, which
qemu interprets as "use the first ahci controller". We then need to
specify the unit with "unit=%d" rather than adding it onto the bus
arg.
We had been setting the device alias in the devinceinfo for pci
controllers to "pci%u", but then hardcoding "pci.%u" when creating the
device address for other devices using that pci bus. This all worked
just fine until we encountered the built-in "pcie.0" bus (the PCIe
root complex) in Q35 machines.
In order to create the correct commandline for this one case, this
patch:
1) sets the alias for PCI controllers correctly, to "pci.%u" (or
"pcie.%u" for the pcie-root controller)
2) eliminates the hardcoded "pci.%u" for pci controllers when
generatuing device address strings, and instead uses the controller's
alias.
3) plumbs a pointer to the virDomainDef all the way down to
qemuBuildDeviceAddressStr. This was necessary in order to make the
aliase of the controller *used by a device* available (previously
qemuBuildDeviceAddressStr only had the deviceinfo of the device
itself, *not* of the controller it was connecting to). This made for a
larger than desired diff, but at least in the future we won't have to
do it again, since all the information we could possibly ever need for
future enhancements is in the virDomainDef. (right?)
This should be done for *all* controllers, but for now we just do it
in the case of PCI controllers, to reduce the likelyhood of
regression.
This patch adds in special handling for a few devices that need to be
treated differently for q35 domains:
usb - there is no implicit/default usb controller for the q35
machinetype. This is done because normally the default usb controller
is added to a domain by just adding "-usb" to the qemu commandline,
and it's assumed that this will add a single piix3 usb1 controller at
slot 1 function 2. That's not what happens when the machinetype is
q35, though. Instead, adding -usb to the commandline adds 3 usb
(version 2) controllers to the domain at slot 0x1D.{1,2,7}. Rather
than having
<controller type='usb' index='0'/>
translate into 3 separate devices on the PCI bus, it's cleaner to not
automatically add a default usb device; one can always be added
explicitly if desired. Or we may decide that on q35 machines, 3 usb
controllers will be automatically added when none is given. But for
this initial commit, at least we aren't locking ourselves into
something we later won't want.
video - qemu always initializes the primary video device immediately
after any integrated devices for the machinetype. Unless instructed
otherwise (by using "-device vga..." instead of "-vga" which libvirt
uses in many cases to work around deficiencies and bugs in various
qemu versions) qemu will always pick the first unused slot. In the
case of the "pc" machinetype and its derivatives, this is always slot
2, but on q35 machinetypes, the first free slot is slot 1 (since the
q35's integrated peripheral devices are placed in other slots,
e.g. slot 0x1f). In order to make the PCI address of the video device
predictable, that slot (1 or 2, depending on machinetype) is reserved
even when no video device has been specified.
sata - a q35 machine always has a sata controller implicitly added at
slot 0x1F, function 2. There is no way to avoid this controller, so we
always add it. Note that the xml2xml tests for the pcie-root and q35
cases were changed to use DO_TEST_DIFFERENT() so that we can check for
the sata controller being automatically added. This is especially
important because we can't check for it in the xml2argv output (it has
no effect on that output since it's an implicit device).
ide - q35 has no ide controllers.
isa and smbus controllers - these two are always present in a q35 (at
slot 0x1F functions 0 and 3) but we have no way of modelling them in
our config. We do need to reserve those functions so that the user
doesn't attempt to put anything else there though. (note that the "pc"
machine type also has an ISA controller, which we also ignore).
This PCI controller, named "dmi-to-pci-bridge" in the libvirt config,
and implemented with qemu's "i82801b11-bridge" device, connects to a
PCI Express slot (e.g. one of the slots provided by the pcie-root
controller, aka "pcie.0" on the qemu commandline), and provides 31
*non-hot-pluggable* PCI (*not* PCIe) slots, numbered 1-31.
Any time a machine is defined which has a pcie-root controller
(i.e. any q35-based machinetype), libvirt will automatically add a
dmi-to-pci-bridge controller if one doesn't exist, and also add a
pci-bridge controller. The reasoning here is that any useful domain
will have either an immediate (startup time) or eventual (subsequent
hot-plug) need for a standard PCI slot; since the pcie-root controller
only provides PCIe slots, we need to connect a dmi-to-pci-bridge
controller to it in order to get a non-hot-plug PCI slot that we can
then use to connect a pci-bridge - the slots provided by the
pci-bridge will be both standard PCI and hot-pluggable.
Since pci-bridge devices themselves can not be hot-plugged into a
running system (although you can hot-plug other devices into a
pci-bridge's slots), any new pci-bridge controller that is added can
(and will) be plugged into the dmi-to-pci-bridge as long as it has
empty slots available.
This patch is also changing the qemuxml2xml-pcie test from a "DO_TEST"
to a "DO_DIFFERENT_TEST". This is so that the "before" xml can omit
the automatically added dmi-to-pci-bridge and pci-bridge devices, and
the "after" xml can include it - this way we are testing if libvirt is
properly adding these devices.
This controller is implicit on q35 machinetypes. It provides 31 PCIe
(*not* PCI) slots as controller 0.
Currently there are no devices that can connect to pcie-root, and no
implicit pci controller on a q35 machine, so q35 is still
unusable. For a usable q35 system, we need to add a
"dmi-to-pci-bridge" pci controller, which can connect to pcie-root,
and provides standard pci slots that can be used to connect other
devices.
Previous refactoring of the guest PCI address reservation/allocation
code allowed for slot types other than basic PCI (e.g. PCI express,
non-hotpluggable slots, etc) but would not auto-allocate a slot for a
device that required any type other than a basic hot-pluggable
PCI slot.
This patch refactors the code to be aware of different slot types
during auto-allocation of addresses as well - as long as there is an
empty slot of the required type, it will be found and used.
The piece that *wasn't* added is that we don't auto-create a new PCI
bus when needed for anything except basic PCI devices. This is because
there are multiple different types of controllers that can provide,
for example, a PCI express slot (in addition to the pcie-root
controller, these can also be found on a "root-port" or on a
"downstream-switch-port"). Since we currently don't support any PCIe
devices (except pending support for dmi-to-pci-bridge), we can defer
any decision on what to do about this.
* The functions qemuDomainPCIAddressReserveAddr and
qemuDomainPCIAddressReserveSlot were very similar (and should have
been more similar) and were about to get more code added to them which
would create even more duplicated code, so this patch gives
qemuDomainPCIAddressReserveAddr a "reserveEntireSlot" arg, then
replaces the body of qemuDomainPCIAddressReserveSlot with a call to
qemuDomainPCIAddressReserveAddr.
You will notice that addrs->lastaddr was previously set in
qemuDomainPCIAddressReserveAddr (but *not* set in
qemuDomainPCIAddressReserveSlot). For consistency and cleanliness of
code, that bit was removed and put into the one caller of
qemuDomainPCIAddressReserveAddr (there is a similar place where the
caller of qemuDomainPCIAddressReserveSlot sets lastaddr). This does
guarantee identical functionality to pre-patch code, but in practice
isn't really critical, because lastaddr is just keeping track of where
to start when looking for a free slot - if it isn't updated, we will
just start looking on a slot that's already occupied, then skip up to
one that isn't.
* qemuCollectPCIAddress was essentially doing the same thing as
qemuDomainPCIAddressReserveAddr, but with some extra special case
checking at the beginning. The duplicate code has been replaced with
a call to qemuDomainPCIAddressReserveAddr. This required adding a
"fromConfig" boolean, which is only used to change the log error
code from VIR_ERR_INTERNAL_ERROR (when the address was
auto-generated by libvirt) to VIR_ERR_XML_ERROR (when the address is
coming from the config); without this differentiation, it would be
difficult to tell if an error was caused by something wrong in
libvirt's auto-allocate code or just bad config.
* the bit of code in qemuDomainPCIAddressValidate that checks the
connect type flags is going to be used in a couple more places where
we don't need to also check the slot limits (because we're generating
the slot number ourselves), so that has been pulled out into a
separate qemuDomainPCIAddressFlagsCompatible function.
* qemuDomainPCIAddressSetNextAddr
The name of this function was confusing because 1) other functions in
the file that end in "Addr" are only operating on a single function of
one PCI slot, not the entire slot, while functions that do something
with the entire slot end in "Slot", and 2) it didn't contain a verb
describing what it is doing (the "Set" refers to the set that contains
all PCI buses in the system, used to keep track of which slots in
which buses are already reserved for use).
It is now renamed to qemuDomainPCIAddressReserveNextSlot, which more
clearly describes what it is doing. Arguably, it could have been
changed to qemuDomainPCIAddressSetReserveNextSlot, but 1) the word
"set" is confusing in this context because it could be intended as a
verb or as a noun, and 2) most other functions that operate on a
single slot or address within this set are also named
qemuDomainPCIAddress... rather than qemuDomainPCIAddressSet... Only
the Create, Free, and Grow functions for an address set (which modify the
entire set, not just one element) use "Set" in their name.
* qemuPCIAddressAsString, qemuPCIAddressValidate
All the other functions in this set are named
qemuDomainPCIAddressxxxxx, so I renamed these to be consistent.
Since PCI bridges, PCIe bridges, PCIe switches, and PCIe root ports
all share the same namespace, they are all defined as controllers of
type='pci' in libvirt (but with a differing model attribute). Each of
these controllers has a certain connection type upstream, allows
certain connection types downstream, and each can either allow a
single downstream connection at slot 0, or connections from slot 1 -
31.
Right now, we only support the pci-root and pci-bridge devices, both
of which only allow PCI devices to connect, and both which have usable
slots 1 - 31. In preparation for adding other types of controllers
that have different capabilities, this patch 1) adds info to the
qemuDomainPCIAddressBus object to indicate the capabilities, 2) sets
those capabilities appropriately for pci-root and pci-bridge devices,
and 3) validates that the controller being connected to is the proper
type when allocating slots or validating that a user-selected slot is
appropriate for a device..
Having this infrastructure in place will make it much easier to add
support for the other PCI controller types.
While it would be possible to do all the necessary checking by just
storing the controller model in the qemyuDomainPCIAddressBus, it
greatly simplifies all the validation code to also keep a "flags",
"minSlot" and "maxSlot" for each - that way we can just check those
attributes rather than requiring a nearly identical switch statement
everywhere we need to validate compatibility.
You may notice many places where the flags are seemingly hard-coded to
QEMU_PCI_CONNECT_HOTPLUGGABLE | QEMU_PCI_CONNECT_TYPE_PCI
This is currently the correct value for all PCI devices, and in the
future will be the default, with small bits of code added to change to
the flags for the few devices which are the exceptions to this rule.
Finally, there are a few places with "FIXME" comments. Note that these
aren't indicating places that are broken according to the currently
supported devices, they are places that will need fixing when support
for new PCI controller models is added.
To assure that there was no regression in the auto-allocation of PCI
addresses or auto-creation of integrated pci-root, ide, and usb
controllers, a new test case (pci-bridge-many-disks) has been added to
both the qemuxml2argv and qemuxml2xml tests. This new test defines a
domain with several dozen virtio disks but no pci-root or
pci-bridges. The .args file of the new test case was created using
libvirt sources from before this patch, and the test still passes
after this patch has been applied.
Although these two enums are named ..._LAST, they really had the value
of ..._SIZE. This patch changes their values so that, e.g.,
QEMU_PCI_ADDRESS_SLOT_LAST really is the slot number of the last slot
on a PCI bus.
The implicit IDE, USB, and video controllers provided by the PIIX3
chipset in the pc-* machinetypes are not present on other
machinetypes, so we shouldn't be doing the special checking for
them. This patch places those validation checks into a separate
function that is only called for machine types that have a PIIX3 chip
(which happens to be the i440fx-based pc-* machine types).
One qemuxml2argv test data file had to be changed - the
pseries-usb-multi test had included a piix3-usb-uhci device, which was
being placed at a specific address, and also had slot 2 auto reserved
for a video device, but the pseries virtual machine doesn't actually
have a PIIX3 chip, so even if there was a piix3-usb-uhci driver for
it, the device wouldn't need to reside at slot 1 function 2. I just
changed the .argv file to have the generic slot info for the two
devices that results when the special PIIX3 code isn't executed.
qemuDomainPCIAddressBus was an array of QEMU_PCI_ADDRESS_SLOT_LAST
uint8_t's, which worked fine as long as every PCI bus was
identical. In the future, some PCI busses will allow connecting PCI
devices, and some will allow PCIe devices; also some will only allow
connection of a single device, while others will allow connecting 31
devices.
In order to keep track of that information for each bus, we need to
turn qemuDomainPCIAddressBus into a struct, for now with just one
member:
uint8_t slots[QEMU_PCI_ADDRESS_SLOT_LAST];
Additional members will come in later patches.
The item in qemuDomainPCIAddresSet that contains the array of
qemuDomainPCIAddressBus is now called "buses" to be more consistent
with the already existing "nbuses" (and with the new "slots" array).
The difference with already supported pool types (dir, fs, block)
is: there are two modes for iscsi pool (or network pools in future),
one can specify it either to use the volume target path (the path
showed up on host) with mode='host', or to use the remote URI qemu
supports (e.g. file=iscsi://example.org:6000/iqn.1992-01.com.example/1)
with mode='direct'.
For 'host' mode, it copies the volume target path into disk->src. For
'direct' mode, the corresponding info in the *one* pool source host def
is copied to disk->hosts[0].
The alias for hostdevs of type SCSI can be too long for QEMU if
larger LUNs are encountered. Here's a real life example:
<hostdev mode='subsystem' type='scsi' managed='no'>
<source>
<adapter name='scsi_host0'/>
<address bus='0' target='19' unit='1088634913'/>
</source>
<address type='drive' controller='0' bus='0' target='0' unit='0'/>
</hostdev>
this results in a too long drive id, resulting in QEMU yelling
Property 'scsi-generic.drive' can't find value 'drive-hostdev-scsi_host0-0-19-1088634913'
This commit changes the alias back to the default hostdev$(index)
scheme.
Signed-off-by: Viktor Mihajlovski <mihajlov@linux.vnet.ibm.com>
When user does not specify any model for scsi controller, or worse, no
controller at all, but libvirt automatically adds scsi controller with
no model, we are not searching for virtio-scsi and thus this can fail
for example on qemu which doesn't support lsi logic adapter.
This means that when qemu on x86 doesn't support lsi53c895a and the
user adds the following to an XML without any scsi controller:
<disk ...>
...
<target dev='sda'>
</disk>
libvirt fails like this:
# virsh define asdf.xml
error: Failed to define domain from asdf.xml
error: internal error Unable to determine model for scsi controller
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=974943
When virAsprintf was changed from a function to a macro
reporting OOM error in dc6f2da, it was documented as returning
0 on success. This is incorrect, it returns the number of bytes
written as asprintf does.
Some of the functions were converted to use virAsprintf's return
value directly, changing the return value on success from 0 to >= 0.
For most of these, this is not a problem, but the change in
virPCIDriverDir breaks PCI passthrough.
The return value check in virhashtest pre-dates virAsprintf OOM
conversion.
vmwareMakePath seems to be unused.
Merge the virCommandPreserveFD / virCommandTransferFD methods
into a single virCommandPasFD method, and use a new
VIR_COMMAND_PASS_FD_CLOSE_PARENT to indicate their difference
in behaviour
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The function being introduced is responsible for creating command
line argument for '-device' for given character device. Based on
the chardev type, it calls appropriate qemuBuild.*ChrDeviceStr(),
e.g. qemuBuildSerialChrDeviceStr() for serial chardev and so on.
The chardev alias assignment is going to be needed in a separate
places, so it should be moved into a separate function rather
than copying code randomly around.
Convert the type of loop iterators named 'i', 'j', k',
'ii', 'jj', 'kk', to be 'size_t' instead of 'int' or
'unsigned int', also santizing 'ii', 'jj', 'kk' to use
the normal 'i', 'j', 'k' naming
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Add new CPU features for HyperV:
vapic for virtual APIC support
spinlocks for setting spinlock support
<features>
<hyperv>
<vapic state='on'/>
<spinlocks state='on' retries='4096'/>
</hyperv>
</features>
https://bugzilla.redhat.com/show_bug.cgi?id=784836
https://bugzilla.redhat.com/show_bug.cgi?id=964177
Though both libvirt and QEMU's document say RTC_CHANGE returns
the offset from the host UTC, qemu actually returns the offset
from the specified date instead when specific date is provided
(-rtc base=$date).
It's not safe for qemu to fix it in code, it worked like that
for 3 years, changing it now may break other QEMU use cases.
What qemu tries to do is to fix the document:
http://lists.gnu.org/archive/html/qemu-devel/2013-05/msg04782.html
And in libvirt side, instead of replying on the value from qemu,
this converts the offset returned from qemu to the offset from
host UTC, by:
/*
* a: the offset from qemu RTC_CHANGE event
* b: The specified date (-rtc base=$date)
* c: the host date when libvirt gets the RTC_CHANGE event
* offset: What libvirt will report
*/
offset = a + (b - c);
The specified date (-rtc base=$date) is recorded in clock's def as
an internal only member (may be useful to exposed outside?).
Internal only XML tag "basetime" is introduced to not lose the
guest's basetime after libvirt restarting/reloading:
<clock offset='variable' adjustment='304' basis='utc' basetime='1370423588'/>
Currently, if there's an error opening /dev/vhost-net (e.g. because
it doesn't exist) but it's not required we proceed with vhostfd array
filled with -1 and vhostfdSize unchanged. Later, when constructing
the qemu command line only non-negative items within vhostfd array
are taken into account. This means, vhostfdSize may be greater than
the actual count of non-negative items in vhostfd array. This results
in improper command line arguments being generated, e.g.:
-netdev tap,fd=21,id=hostnet0,vhost=on,vhostfd=(null)
With previous patch, we accept negative value as length of string to
duplicate. So there is no need to pass strlen(src) in case we want to do
duplicate the whole string.
In order to learn libvirt multiqueue several things must be done:
1) The '/dev/net/tun' device needs to be opened multiple times with
IFF_MULTI_QUEUE flag passed to ioctl(fd, TUNSETIFF, &ifr);
2) Similarly, '/dev/vhost-net' must be opened as many times as in 1)
in order to keep 1:1 ratio recommended by qemu and kernel folks.
3) The command line construction code needs to switch from 'fd=X' to
'fds=X:Y:...:Z' and from 'vhostfd=X' to 'vhostfds=X:Y:...:Z'.
4) The monitor handling code needs to learn to pass multiple FDs.
