Laine Stump [Mon, 19 Mar 2012 16:49:17 +0000 (12:49 -0400)]
conf: forbid use of multicast mac addresses
A few times libvirt users manually setting mac addresses have
complained of a networking failure that ends up being due to a multicast
mac address being used for a guest interface. This patch prevents that
by logging an error and failing if a multicast mac address is
encountered in each of the three following cases:
1) domain xml <interface> mac address.
2) network xml bridge mac address.
3) network xml dhcp/host mac address.
There are several other places where a mac address can be input that
aren't controlled in this manner because failure to do so has no
consequences (e.g., if the address will be used to search through
existing interfaces for a match).
The RNG has been updated to add multiMacAddr and uniMacAddr along with
the existing macAddr, and macAddr was switched to uniMacAddr where
appropriate.
Laine Stump [Mon, 19 Mar 2012 03:57:50 +0000 (23:57 -0400)]
conf: return immediately on error in dhcp host element
If an error was encountered parsing a dhcp host entry mac address or
name, parsing would continue and log a less descriptive error that
might make it more difficult to notice the true nature of the problem.
This patch returns immediately on logging the first error.
If a guest's tap device is created using the same MAC address the
guest uses for its own network card (which connects to the tap
device), the Linux kernel will log the following message and traffic
will not pass:
kernel: vnet9: received packet with own address as source address
This patch disallows MAC addresses with a first byte of 0xFE, but only in
the case that the MAC address is used for a guest interface that's
connected by way of a standard tap device. (In other words, the
validation is done at runtime at the same place the MAC address is
modified for the tap device, rather than when mac address is parsed,
the idea being that it is then we know for sure the address will be
problematic.)
Using inheritance, this patch cleans up the cpu_map.xml file and also
sorts all CPU features according to the feature and registry
values. Model features are sorted the same way as foeatures in the
specification.
Also few models that are related were organized together and parts of
the XML are marked with comments
Eric Blake [Fri, 16 Mar 2012 20:43:58 +0000 (14:43 -0600)]
snapshot: make quiesce a bit safer
If a guest is paused, we were silently ignoring the quiesce flag,
which results in unclean snapshots, contrary to the intent of the
flag. Since we can't quiesce without guest agent support, we should
instead fail if the guest is not running.
Meanwhile, if we attempt a quiesce command, but the guest agent
doesn't respond, and we time out, we may have left the command
pending on the guest's queue, and when the guest resumes parsing
commands, it will freeze even though our command is no longer
around to issue a thaw. To be safe, we must _always_ pair every
quiesce call with a counterpart thaw, even if the quiesce call
failed due to a timeout, so that if a guest wakes up and starts
processing a command backlog, it will not get stuck in a frozen
state.
* src/qemu/qemu_driver.c (qemuDomainSnapshotCreateDiskActive):
Always issue thaw after a quiesce, even if quiesce failed.
(qemuDomainSnapshotFSThaw): Add a parameter.
This patch fixes a NULL pointer check that was causing SegFault on
some specific configurations. It also reverts commit 59d0c9801c1ab
that was checking for this value in one place.
A common coding pattern for changing blkio parameters is
1. virDomainGetBlkioParameters
2. change one or more params
3. virDomainSetBlkioParameters
For this to work, it must be possible to roundtrip through
the methods without error. Unfortunately virDomainGetBlkioParameters
will return "" for the deviceWeight parameter for guests by default,
which virDomainSetBlkioParameters will then reject as invalid.
This fixes the handling of "" to be a no-op, and also improves the
error message to tell you what was invalid
== 8 bytes in 1 blocks are definitely lost in loss record 1 of 28
== at 0x4A04A28: calloc (vg_replace_malloc.c:467)
== by 0x3EB7115FB8: xdr_reference (in /lib64/libc-2.12.so)
== by 0x3EB7115F10: xdr_pointer (in /lib64/libc-2.12.so)
== by 0x4D1EA84: xdr_remote_string (remote_protocol.c:40)
== by 0x4D1EAD8: xdr_remote_domain_migrate_prepare3_ret (remote_protocol.c:4772)
== by 0x4D2FFD2: virNetMessageDecodePayload (virnetmessage.c:382)
== by 0x4D2789C: virNetClientProgramCall (virnetclientprogram.c:382)
== by 0x4D0707D: callWithFD (remote_driver.c:4549)
== by 0x4D070FB: call (remote_driver.c:4570)
== by 0x4D12AEE: remoteDomainMigratePrepare3 (remote_driver.c:4138)
== by 0x4CF7BE9: virDomainMigrateVersion3 (libvirt.c:4815)
== by 0x4CF9432: virDomainMigrate2 (libvirt.c:5454)
==
== LEAK SUMMARY:
== definitely lost: 8 bytes in 1 blocks
== indirectly lost: 0 bytes in 0 blocks
== possibly lost: 0 bytes in 0 blocks
== still reachable: 126,995 bytes in 1,343 blocks
== suppressed: 0 bytes in 0 blocks
This patch also fixes the leaks in remoteDomainMigratePrepare and
remoteDomainMigratePrepare2.
Add support for forcing a private network namespace for LXC guests
If no <interface> elements are included in an LXC guest XML
description, then the LXC guest will just see the host's
network interfaces. It is desirable to be able to hide the
host interfaces, without having to define any guest interfaces.
This patch introduces a new feature flag <privnet/> to allow
forcing of a private network namespace for LXC. In the future
I also anticipate that we will add <privuser/> to force a
private user ID namespace.
* src/conf/domain_conf.c, src/conf/domain_conf.h: Add support
for <privnet/> feature. Auto-set <privnet> if any <interface>
devices are defined
* src/lxc/lxc_container.c: Honour request for private network
namespace
Peter Krempa [Thu, 15 Mar 2012 16:10:22 +0000 (17:10 +0100)]
lib: Don't access configuration if none is present
Commit e457d5ef2015e6106094b85f8bbd1582002edc4d adds ability to pass the
default URI using the client configuration file. If the file is not
present, it still accesses the NULL config object causing a segfault.
Laine Stump [Wed, 14 Mar 2012 05:41:35 +0000 (01:41 -0400)]
Emit graphics events when a SPICE client connects/disconnects
Wire up the domain graphics event notifications for SPICE. Adapted
from a RHEL-only patch written by Dan Berrange that used custom
__com.redhat_SPICE events - equivalent events are now available in
upstream QEMU (including a SPICE_CONNECTED event, which was missing in
the __COM.redhat_SPICE version).
* src/qemu/qemu_monitor_json.c: Wire up SPICE graphics events
Osier Yang [Thu, 8 Mar 2012 13:36:26 +0000 (21:36 +0800)]
qemu: Support numad
numad is an user-level daemon that monitors NUMA topology and
processes resource consumption to facilitate good NUMA resource
alignment of applications/virtual machines to improve performance
and minimize cost of remote memory latencies. It provides a
pre-placement advisory interface, so significant processes can
be pre-bound to nodes with sufficient available resources.
More details: http://fedoraproject.org/wiki/Features/numad
"numad -w ncpus:memory_amount" is the advisory interface numad
provides currently.
This patch add the support by introducing a new XML attribute
for <vcpu>. e.g.
The returned advisory nodeset from numad will be printed
in domain's dumped XML. e.g.
<vcpu placement="auto" cpuset="1-10^6">4</vcpu>
If placement is "auto", the number of vcpus and the current
memory amount specified in domain XML will be used for numad
command line (numad uses MB for memory amount):
numad -w $num_of_vcpus:$current_memory_amount / 1024
The advisory nodeset returned from numad will be used to set
domain process CPU affinity then. (e.g. qemuProcessInitCpuAffinity).
If the user specifies both CPU affinity policy (e.g.
(<vcpu cpuset="1-10,^7,^8">4</vcpu>) and placement == "auto"
the specified CPU affinity will be overridden.
Michal Privoznik [Wed, 14 Mar 2012 10:08:52 +0000 (11:08 +0100)]
qemu: Reverse condition in qemuDomainCheckDiskPresence
With current code, we pass true iff domain is cold booting. However,
if disk is inaccessible and startupPolicy for that disk is set to
'requisite' we have to fail iff cold booting.
AMD Bulldozer (or Opteron_G4 as called in QEMU) was added to the list
of cpu models, flags were taken from upstream qemu cpu specifications
and should be sorted by bit values (or first occurence in the feature
specification part of cpu_map.xml).
Michal Privoznik [Mon, 12 Mar 2012 15:50:39 +0000 (16:50 +0100)]
graphics: Cleanup port policy
Even though we say in documentation setting (tls-)port to -1 is legacy
compat style for enabling autoport, we're roughly doing this for VNC.
However, in case of SPICE auto enable autoport iff both port & tlsPort
are equal -1 as documentation says autoport plays with both.
Guannan Ren [Mon, 12 Mar 2012 15:50:02 +0000 (23:50 +0800)]
qemu: fix segfault when detaching non-existent network device
In qemuDomainDetachNetDevice, detach was being used before it had been
validated. If no matching device was found, this resulted in a
dereference of a NULL pointer.
Jiri Denemark [Fri, 9 Mar 2012 15:42:46 +0000 (16:42 +0100)]
qemu: Fix (managed)save and snapshots with host mode CPU
When host-model and host-passthrouh CPU modes were introduced, qemu
driver was properly modify to update guest CPU definition during
migration so that we use the right CPU at the destination. However,
similar treatment is needed for (managed)save and snapshots since they
need to save the exact CPU so that a domain can be properly restored.
To avoid repetition of such situation, all places that need live XML
share the code which generates it.
As a side effect, this patch fixes error reporting from
qemuDomainSnapshotWriteMetadata().
Eric Blake [Fri, 9 Mar 2012 00:35:59 +0000 (17:35 -0700)]
cpustats: collect VM user and sys times
As documented in linux.git/Documentation/cgroups/cpuacct.txt,
cpuacct.stat returns user and system time in ticks (the same
unit used in times(2)). It would be a bit nicer if it were like
getrusage(2) and reported timeval contents, or like cpuacct.usage
and in nanoseconds, but we can't be picky.
* src/util/cgroup.h (virCgroupGetCpuacctStat): New function.
* src/util/cgroup.c (virCgroupGetCpuacctStat): Implement it.
(virCgroupGetValueStr): Allow for multi-line files.
* src/libvirt_private.syms (cgroup.h): Export it.
Eric Blake [Fri, 9 Mar 2012 19:13:30 +0000 (16:13 -0300)]
qemu: support disk filenames with comma
If there is a disk file with a comma in the name, QEmu expects a double
comma instead of a single one (e.g., the file "virtual,disk.img" needs
to be specified as "virtual,,disk.img" in QEmu's command line). This
patch fixes libvirt to work with that feature. Fix RHBZ #801036.
Based on an initial patch by CrÃstian Viana.
* src/util/buf.h (virBufferEscape): Alter signature.
* src/util/buf.c (virBufferEscape): Add parameter.
(virBufferEscapeSexpr): Fix caller.
* src/qemu/qemu_command.c (qemuBuildRBDString): Likewise. Also
escape commas in file names.
(qemuBuildDriveStr): Escape commas in file names.
* docs/schemas/basictypes.rng (absFilePath): Relax RNG to allow
commas in input file names.
* tests/qemuxml2argvdata/*-disk-drive-network-sheepdog.*: Update
test.
Removed more AMD-specific features from cpu64-rhel* models
We found few more AMD-specific features in cpu64-rhel* models that
made it impossible to start qemu guest on Intel host (with this
setting) even though qemu itself starts correctly with them.
This impacts one test, thus the fix in tests/cputestdata/.
Laine Stump [Thu, 11 Aug 2011 03:34:07 +0000 (23:34 -0400)]
conf: eliminate redundant VIR_ALLOC of 1st element of network DNS hosts.
virNetworkDNSHostsDefParseXML was calling VIR_ALLOC(def->hosts) if
def->hosts was NULL. This is a waste of time, though, since
VIR_REALLOC_N is called a few lines further down, prior to any use of
def->hosts. (initializing def->nhosts to 0 is also redundant, because
the newly allocated memory will always be cleared to all 0's anyway).
qemuxml2argvtest: Pass some additional flags to graphics-spice-agentmouse
One of the recent commits introduced support for
spice agent-mouse. However, test for this feature
require some tweaking: pass QEMU_CAPS_CHARDEV_SPICEVMC |
QEMU_CAPS_NODEFCONFIG and add "-vga cirrus".
qemuBuildCommandLine: Don't add tlsPort if none set
If user hasn't supplied any tlsPort we default to setting it
to zero in our internal structure. However, when building command
line we test it against -1 which is obviously wrong.
Osier Yang [Thu, 8 Mar 2012 11:38:57 +0000 (19:38 +0800)]
virsh: Use option alias for outmoded "--persistent"
Since VIR_DOMAIN_AFFECT_{LIVE,CONFIG,CURRENT} was created,
all new virsh commands use "--config" to represents the
persistent changing. This patch add "--config" option
for the old commands which still use "--persistent",
and "--persistent" is now alias of "--config".
tools/virsh.c: (use "--config", and "--persistent" is
alias of "--config" now).
cmdDomIfSetLink, cmdDomIfGetLink, cmdAttachDevice,
cmdDetachDevice, cmdUpdateDevice, cmdAttachInterface,
cmdDetachInterface, cmdAttachDisk, cmdDetachDisk
toos/virsh.pod: Update docs of the changed commands, and
add some missed docs for "--config" (detach-interface,
detach-disk, and detach-device).
Laine Stump [Thu, 8 Mar 2012 20:44:38 +0000 (15:44 -0500)]
remove daemon/probes.h from .gitignore
The file daemon/probes.h used to be generated as part of a build, but
is no longer used. However, a stale copy of it lying around could
cause a build to fail. Removing it from .gitignore will make it more
likely someone will notice that they have it lying around.
Laine Stump [Thu, 8 Mar 2012 20:41:53 +0000 (15:41 -0500)]
util: consolidate duplicated error messages in pci.c
This is nearly identical to an earlier patch for virnetlink.c.
There are special stub versions of all public functions in this file
that are compiled when the platform isn't linux. Each of these
functions had an almost identical message, differing only in the
function name included in the message. Since log messages already
contain the function name, we can just define a const char* with the
common part of the string, and use that same string for all the log
messages.
If nothing else, this at least makes for less strings that need
translating...
Laine Stump [Thu, 8 Mar 2012 06:46:36 +0000 (01:46 -0500)]
qemu: eliminate memory leak in qemuDomainUpdateDeviceConfig
This function was freeing a virDomainNetDef with
VIR_FREE(). virDomainNetDef is a complex structure with many pointers
to other dynamically allocated data; to properly free it
virDomainNetDefFree() must be called instead, otherwise several
strings (and potentially other things) will be leaked.
Laine Stump [Wed, 7 Mar 2012 21:05:34 +0000 (16:05 -0500)]
qemu: support persistent hotplug of <hostdev> devices
For some reason, although live hotplug of <hostdev> devices is
supported, persistent hotplug is not. This patch adds the proper
VIR_DOMAIN_DEVICE_HOSTDEV cases to the switches in
qemuDomainAttachDeviceConfig and qemuDomainDetachDeviceConfig.
Laine Stump [Wed, 7 Mar 2012 17:44:56 +0000 (12:44 -0500)]
util: standardize return from functions calling virNetlinkCommand
There are several functions that call virNetlinkCommand, and they all
follow a common pattern, with three exit labels: err_exit (or
cleanup), malformed_resp, and buffer_too_small. All three of these
labels do their own cleanup and have their own return. However, the
malformed_resp label usually frees the same items as the
cleanup/err_exit label, and the buffer_too_small label just doesn't
free recvbuf (because it's known to always be NULL at the time we goto
buffer_too_small.
In order to simplify and standardize the code, I've made the following
changes to all of these functions:
1) err_exit is replaced with the more libvirt-ish "cleanup", which
makes sense because in all cases this code is also executed in the
case of success, so labelling it err_exit may be confusing.
2) rc is initialized to -1, and set to 0 just before the cleanup
label. Any code that currently sets rc = -1 is made to instead goto
cleanup.
3) malformed_resp and buffer_too_small just log their error and goto
cleanup. This gives us a single return path, and a single place to
free up resources.
4) In one instance, rather then logging an error immediately, a char*
msg was pointed to an error string, then goto cleanup (and cleanup
would log an error if msg != NULL). It takes no more lines of code
to just log the message as we encounter it.
Laine Stump [Tue, 6 Mar 2012 23:06:14 +0000 (18:06 -0500)]
util: eliminate device object leaks related to virDomain*Remove*()
There are several functions in domain_conf.c that remove a device
object from the domain's list of that object type, but don't free the
object or return it to the caller to free. In many cases this isn't a
problem because the caller already had a pointer to the object and
frees it afterward, but in several cases the removed object was just
left floating around with no references to it.
In particular, the function qemuDomainDetachDeviceConfig() calls
functions to locate and remove net (virDomainNetRemoveByMac), disk
(virDomainDiskRemoveByName()), and lease (virDomainLeaseRemove())
devices, but neither it nor its caller qemuDomainModifyDeviceConfig()
ever obtain a pointer to the device being removed, much less free it.
This patch modifies the following "remove" functions to return a
pointer to the device object being removed from the domain device
arrays, to give the caller the option of freeing the device object
using that pointer if needed. In places where the object was
previously leaked, it is now freed:
libxlDomainDetachConfig - leaked a virDomainDiskDef
qemuDomainDetachDeviceConfig - could leak a virDomainDiskDef,
a virDomainNetDef, or a
virDomainLeaseDef
qemuDomainDetachLease - leaked a virDomainLeaseDef
Laine Stump [Wed, 7 Mar 2012 01:43:22 +0000 (20:43 -0500)]
qemu: don't 'remove' hostdev objects from domain if operation fails
There were certain paths through the hostdev detach code that could
lead to the lower level function failing (and not removing the object
from the domain's hostdevs list), but the higher level function
free'ing the hostdev object anyway. This would leave a stale
hostdevdef pointer in the list, which would surely cause a problem
eventually.
This patch relocates virDomainHostdevRemove from the lower level
functions qemuDomainDetachThisHostDevice and
qemuDomainDetachHostPciDevice, to their caller
qemuDomainDetachThisHostDevice, placing it just before the call to
virDomainHostdevDefFree. This makes it easy to verify that either both
operations are done, or neither.
NB: The "dangling pointer" part of this problem was introduced in
commit 13d5a6, so it is not present in libvirt versions prior to
0.9.9. Earlier versions would return failure in certain cases even
though the the device object was removed/deleted, but the removal and
deletion operations would always both happen or neither.
Laine Stump [Tue, 6 Mar 2012 17:21:21 +0000 (12:21 -0500)]
util: consolidate duplicated error messages in virnetlink.c
There are special stub versions of all public functions in this file
that are compiled when either libnl isn't available or the platform
isn't linux. Each of these functions had two almost identical message,
differing only in the function name included in the message. Since log
messages already contain the function name, we can just define a const
char* with the common part of the string, and use that same string for
all the log messages.
Also, rather than doing #if defined ... #else ... #endif *inside the
error log macro invocation*, this patch does #if defined ... just
once, using it to decide which single string to define. This turns the
error log in each function from 6 lines, to 1 line.
Ansis Atteka [Wed, 7 Mar 2012 07:15:36 +0000 (23:15 -0800)]
Attach vm-id to Open vSwitch interfaces.
This patch will allow OpenFlow controllers to identify which interface
belongs to a particular VM by using the Domain UUID.
ovs-vsctl get Interface vnet0 external_ids
{attached-mac="52:54:00:8C:55:2C", iface-id="83ce45d6-3639-096e-ab3c-21f66a05f7fa", iface-status=active, vm-id="142a90a7-0acc-ab92-511c-586f12da8851"}
V2 changes:
Replaced vm-uuid with vm-id. There was a discussion in Open vSwitch
mailinglist that we should stick with the same DB key postfixes for the
sake of consistency (e.g iface-id, vm-id ...).
Laine Stump [Thu, 8 Mar 2012 18:31:06 +0000 (13:31 -0500)]
util: whitespace change to virNetDevOpenvswitchAddPort
The indentation on the final lines of the function was off by four
spaces, making me wonder for a second if there was something
missing. (There wasn't.)
Eric Blake [Tue, 6 Mar 2012 20:49:53 +0000 (13:49 -0700)]
rpc: generalize solution for VPATH builds
Commit 5d4b0c4c80 tried to fix certain classes of VPATH builds,
but was too limited. In particular, Guannan Ren reported:
> For example: The libvirt source code resides in /home/testuser,
> I make dist in /tmp/buildvpath, the XDR routine .c file will
> include full path of the header file like:
>
> #include "/home/testuser/src/rpc/virnetprotocol.h"
> #include "internal.h"
> #include <arpa/inet.h>
>
> If we distribute the tarball to another machine to compile,
> it will report error as follows:
>
> rpc/virnetprotocol.c:7:59: fatal error:
> /home/testuser/src/rpc/virnetprotocol.h: No such file or directory
util: Don't overflow on errno in virFileAccessibleAs
If we need to virFork() to check assess() under different
UID+GID we need to translate returned status via WEXITSTATUS().
Otherwise, we may return values greater than 255 which is
obviously wrong.
Peter Krempa [Wed, 7 Mar 2012 14:04:33 +0000 (15:04 +0100)]
sanlock: Use STREQ_NULLABLE instead of STREQ on strings that may be null
The function sanlock_inquire can return NULL in the state string if the
message consists only of a header. The return value is arbitrary and
sent by the server. We should proceed carefully while touching such
pointers.
Some members are generated during XML parse (e.g. MAC address of
an interface); However, with current implementation, if we
are plugging a device both to persistent and live config,
we parse given XML twice: first time for live, second for config.
This is wrong then as the second time we are not guaranteed
to generate same values as we did for the first time.
To prevent that we need to create a copy of DeviceDefPtr;
This is done through format/parse process instead of writing
functions for deep copy as it is easier to maintain:
adding new field to any virDomain*DefPtr doesn't require change
of copying function.
Currently, startupPolicy='requisite' was determining cold boot
by migrateFrom != NULL. That means, if domain was started up
with migrateFrom set we didn't require disk source path and allowed
it to be dropped. However, on snapshot-revert domain wasn't migrated
but according to documentation, requisite should drop disk source
as well.
Eric Blake [Tue, 6 Mar 2012 00:06:57 +0000 (17:06 -0700)]
virsh: improve memory unit parsing
The last vestige of the inaccurate 'kilobytes' when we meant 1024 is
now gone. And virsh is now useful for setting memory in units other
than KiB.
* tools/virsh.c (cmdSetmem, cmdSetmaxmem): Use new helper routine,
allow passing bogus arguments on to hypervisor to test driver
sanity checking, and fix leak on parse error.
(vshMemtuneGetSize): New helper.
(cmdMemtune): Use it.
* tools/virsh.pod (setmem, setmaxmem, memtune): Document this.
to get both interfaces to resize to 10MiB. The remaining wart
is that vol-resize defaults to bytes, but blockresize defaults
to KiB, but we can't break existing scripts; oh well, it's no
worse than the same wart of the underlying virDomainBlockResize.
The API for virStorageVolResize states that capacity must always
be positive, and that the presence of shrink and delta flags is
what implies a negative change.
* tools/virsh.c (vshCommandOptScaledInt): New function.
(cmdVolResize): Don't pass negative size.
(cmdVolSize): Rename...
(vshVolSize): ...and use new helper routine.
(cmdBlockResize): Use new helper routine, and support new bytes
flag.
* tools/virsh.pod (NOTES): Document suffixes.
(blockresize, vol-create-as, vol-resize): Point to notes.
Eric Blake [Fri, 2 Mar 2012 19:01:06 +0000 (12:01 -0700)]
virsh: add command aliases, and rename nodedev-detach
Just because our public API has a typo doesn't mean that virsh
has to keep the typo.
* tools/virsh.c (VSH_CMD_FLAG_ALIAS): New flag.
(nodedevCmds): Use it.
(cmdHelp): Omit alias commands.
(cmdNodeDeviceDettach): Rename...
(cmdNodeDeviceDetach): ...to this.
* tools/virsh.pod (nodedev-detach): Document it.
Eric Blake [Fri, 2 Mar 2012 18:01:15 +0000 (11:01 -0700)]
virsh: add option aliases
In the past, we have created some virsh options with less-than-stellar
names. For back-compat reasons, those names must continue to parse,
but we don't want to document them in help output. This introduces
a new option type, an alias, which points to a canonical option name
later in the option list.
I'm actually quite impressed that our code has already been factored
to do all option parsing through common entry points, such that I
got this added in relatively few lines of code!
* tools/virsh.c (VSH_OT_ALIAS): New option type.
(opts_echo): Hook up an alias, for easy testing.
(vshCmddefOptParse, vshCmddefHelp, vshCmddefGetOption): Allow for
aliases.
* tools/virsh.pod (NOTES): Document promise of back-compat.
* tests/virshtest.c (mymain): Test new feature.
Eric Blake [Fri, 2 Mar 2012 20:27:39 +0000 (13:27 -0700)]
xml: use better types for memory values
Using 'unsigned long' for memory values is risky on 32-bit platforms,
as a PAE guest can have more than 4GiB memory. Our API is
(unfortunately) locked at 'unsigned long' and a scale of 1024, but
the rest of our system should consistently use 64-bit values,
especially since the previous patch centralized overflow checking.
Eric Blake [Sat, 3 Mar 2012 00:47:16 +0000 (17:47 -0700)]
xml: use long long internally, to centralize overflow checks
On 64-bit platforms, unsigned long and unsigned long long are
identical, so we don't have to worry about overflow checks.
On 32-bit platforms, anywhere we narrow unsigned long long back
to unsigned long, we have to worry about overflow; it's easier
to do this in one place by having most of the code use the same
or wider types, and only doing the narrowing at the last minute.
Therefore, the memory set commands remain unsigned long, and
the memory get command now centralizes the overflow check into
libvirt.c, so that drivers don't have to repeat the work.
This also fixes a bug where xen returned the wrong value on
failure (most APIs return -1 on failure, but getMaxMemory
must return 0 on failure).
Eric Blake [Thu, 23 Feb 2012 04:18:39 +0000 (21:18 -0700)]
xml: drop unenforced minimum memory limit from RNG
The test domain allows <memory>0</memory>, but the RNG was stating
that memory had to be at least 4096000 bytes. Hypervisors should
enforce their own limits, rather than complicating the RNG.
Meanwhile, some copy and paste had introduced some fishy constructs
in various unit tests.
* docs/schemas/domaincommon.rng (memoryKB, memoryKBElement): Drop
limit that isn't enforced in code.
* src/conf/domain_conf.c (virDomainDefParseXML): Require current
<= maximum.
* tests/qemuxml2argvdata/*.xml: Fix offenders.
Eric Blake [Mon, 5 Mar 2012 21:06:33 +0000 (14:06 -0700)]
storage: support more scaling suffixes
Disk manufacturers are fond of quoting sizes in powers of 10,
rather than powers of 2 (after all, 2.1 GB sounds larger than
2.0 GiB, even though the exact opposite is true). So, we might
as well follow coreutils' lead in supporting three types of
suffix: single letter ${u} (which we already had) and ${u}iB
for the power of 2, and ${u}B for power of 10.
Additionally, it is impossible to create a file with more than
2**63 bytes, since off_t is signed (if you have enough storage
to even create one 8EiB file, I'm jealous). This now reports
failure up front rather than down the road when the kernel
finally refuses an impossible size.
* docs/schemas/basictypes.rng (unit): Add suffixes.
* src/conf/storage_conf.c (virStorageSize): Use new function.
* docs/formatstorage.html.in: Document it.
* tests/storagevolxml2xmlin/vol-file-backing.xml: Test it.
* tests/storagevolxml2xmlin/vol-file.xml: Likewise.
Eric Blake [Thu, 23 Feb 2012 00:48:38 +0000 (17:48 -0700)]
xml: output memory unit for clarity
Make it obvious to 'dumpxml' readers what unit we are using,
since our default of KiB for memory (1024) differs from qemu's
default of MiB; and differs from our use of bytes for storage.
Note that with this patch, the RNG for <memory> still forbids
validation of anything except unit='KiB', since the code silently
ignores the attribute; a later patch will expand <memory> to allow
scaled input in the code and update the RNG to match.
* docs/schemas/basictypes.rng (unit): Add 'bytes'.
(scaledInteger): New define.
* docs/schemas/storagevol.rng (sizing): Use it.
* docs/schemas/storagepool.rng (sizing): Likewise.
* docs/schemas/domaincommon.rng (memoryKBElement): New define; use
for memory elements.
* src/conf/storage_conf.c (virStoragePoolDefFormat)
(virStorageVolDefFormat): Likewise.
* src/conf/domain_conf.h (_virDomainDef): Document unit used
internally.
* src/conf/storage_conf.h (_virStoragePoolDef, _virStorageVolDef):
Likewise.
* tests/*data/*.xml: Update all tests.
* tests/*out/*.xml: Likewise.
* tests/define-dev-segfault: Likewise.
* tests/openvzutilstest.c (testReadNetworkConf): Likewise.
* tests/qemuargv2xmltest.c (blankProblemElements): Likewise.
Eric Blake [Wed, 29 Feb 2012 04:16:28 +0000 (21:16 -0700)]
xml: share 'unit' in RNG
The code supported unit='E' for "exabyte", but the RNG did not;
conversely, the RNG supported "z" and "y" but the code did not
(I'm jealous if you have that much storage, particularly since
it won't fit in 64-bit off_t). Also, the code supported
<allocation unit='...'>, but not the RNG.
In an effort to make 'unit' more worthwhile in future patches,
it's easier to share it between files.
In making this factorization, note that absFilePath is more
permissive than 'path', so storage pools and storage volumes
will now validate with a wider set of file names than before.
I don't think this should be a problem in practice.
* docs/schemas/storagepool.rng: Include basic types, rather than
repeating things here.
* docs/schemas/storagevol.rng: Likewise.
* docs/schemas/basictypes.rng: Add 'unsignedLong', 'unit', and fix
to match storage code.
Eric Blake [Mon, 5 Mar 2012 16:28:59 +0000 (09:28 -0700)]
util: new function for scaling numbers
Scaling an integer based on a suffix is something we plan on reusing
in several contexts: XML parsing, virsh CLI parsing, and possibly
elsewhere. Make it easy to reuse, as well as adding in support for
powers of 1000.
* src/util/util.h (virScaleInteger): New function.
* src/util/util.c (virScaleInteger): Implement it.
* src/libvirt_private.syms (util.h): Export it.
Eric Blake [Fri, 2 Mar 2012 23:58:05 +0000 (16:58 -0700)]
api: add overflow error
Overflow can be user-induced, so it deserves more than being called
an internal error. Note that in general, 32-bit platforms have
far more places to trigger this error (anywhere the public API
used 'unsigned long' but the other side of the connection is a
64-bit server); but some are possible on 64-bit platforms (where
the public API computes the product of two numbers).
* include/libvirt/virterror.h (VIR_ERR_OVERFLOW): New error.
* src/util/virterror.c (virErrorMsg): Translate it.
* src/libvirt.c (virDomainSetVcpusFlags, virDomainGetVcpuPinInfo)
(virDomainGetVcpus, virDomainGetCPUStats): Use it.
* daemon/remote.c (HYPER_TO_TYPE): Likewise.
* src/qemu/qemu_driver.c (qemuDomainBlockResize): Likewise.
Eric Blake [Fri, 2 Mar 2012 15:23:07 +0000 (08:23 -0700)]
docs: use correct terminology for 1024 bytes
Yes, I like kilobytes better than kibibytes (when I say kilobytes,
I generally mean 1024). But since the term is ambiguous, it can't
hurt to say what we mean, by using both the correct name and
calling out the numeric equivalent.
Laine Stump [Wed, 7 Mar 2012 18:16:53 +0000 (13:16 -0500)]
util: fix build mingw (and all non-linux) build failure
ATTRIBUTE_UNUSED was accidentally forgotten on one arg of a stub
function for functionality that's not present on non-linux
platforms. This causes a non-linux build with
--enable-compile-warnings=error to fail.
Eric Blake [Wed, 7 Mar 2012 04:36:53 +0000 (21:36 -0700)]
rpc: allow truncated return for virDomainGetCPUStats
The RPC code assumed that the array returned by the driver would be
fully populated; that is, ncpus on entry resulted in ncpus * return
value on exit. However, while we don't support holes in the middle
of ncpus, we do want to permit the case of ncpus on entry being
longer than the array returned by the driver (that is, it should be
safe for the caller to pass ncpus=128 on entry, and the driver will
stop populating the array when it hits max_id).
Additionally, a successful return implies that the caller will then
use virTypedParamArrayClear on the entire array; for this to not
free uninitialized memory, the driver must ensure that all skipped
entries are explicitly zeroed (the RPC driver did this, but not
the qemu driver).
There are now three cases:
server 0.9.10 and client 0.9.10 or newer: No impact - there were no
hypervisor drivers that supported cpu stats
server 0.9.11 or newer and client 0.9.10: if the client calls with
ncpus beyond the max, then the rpc call will fail on the client side
and disconnect the client, but the server is no worse for the wear
server 0.9.11 or newer and client 0.9.11: the server can return a
truncated array and the client will do just fine
I reproduced the problem by using a host with 2 CPUs, and doing:
virsh cpu-stats $dom --start 1 --count 2
* daemon/remote.c (remoteDispatchDomainGetCPUStats): Allow driver
to omit tail of array.
* src/remote/remote_driver.c (remoteDomainGetCPUStats):
Accommodate driver that omits tail of array.
* src/libvirt.c (virDomainGetCPUStats): Document this.
* src/qemu/qemu_driver.c (qemuDomainGetPercpuStats): Clear all
unpopulated entries.
qemu driver for virDomainGetCPUstats using cpuacct cgroup.
* For now, only "cpu_time" is supported.
* cpuacct cgroup is used for providing percpu cputime information.
* src/qemu/qemu.conf - take care of cpuacct cgroup.
* src/qemu/qemu_conf.c - take care of cpuacct cgroup.
* src/qemu/qemu_driver.c - added an interface
* src/util/cgroup.c/h - added interface for getting percpu cputime
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Roopa Prabhu [Tue, 6 Mar 2012 01:12:44 +0000 (17:12 -0800)]
qemu: install port profile and mac address on netdev hostdevs
These changes are applied only if the hostdev has a parent net device
(i.e. if it was defined as "<interface type='hostdev'>" rather than
just "<hostdev>"). If the parent netdevice has virtual port
information, the original virtualport associate functions are called
(these set and restore both mac and port profile on an
interface). Otherwise, only mac address is set on the device.
Note that This is only supported for SR-IOV Virtual Functions (not for
standard PCI or USB netdevs), and virtualport association is only
supported for 802.1Qbh. For all other types of cards and types of
virtualport, a "Config Unsupported" error is returned and the
operation fails.
Roopa Prabhu [Tue, 6 Mar 2012 01:12:39 +0000 (17:12 -0800)]
util: Changes to support portprofiles for hostdevs
This patch includes the following changes to virnetdevmacvlan.c and
virnetdevvportprofile.c:
- removes some netlink functions which are now available in
virnetdev.c
- Adds a vf argument to all port profile functions.
For 802.1Qbh devices, the port profile calls can use a vf argument if
passed by the caller. If the vf argument is -1 it will try to derive the vf
if the device passed is a virtual function.
For 802.1Qbg devices, This patch introduces a null check for the device
argument because during port profile assignment on a hostdev, this argument
can be null.
Roopa Prabhu [Tue, 6 Mar 2012 01:12:34 +0000 (17:12 -0800)]
util: support functions for mac/portprofile associations on hostdev
This patch adds the following:
- functions to set and get vf configs
- Functions to replace and store vf configs (Only mac address is handled today.
But the functions can be easily extended for vlans and other vf configs)
- function to dump link dev info (This is moved from virnetdevvportprofile.c)
Roopa Prabhu [Tue, 6 Mar 2012 01:12:23 +0000 (17:12 -0800)]
util: two new pci util functions
pciDeviceGetVirtualFunctionInfo returns pf netdevice name and virtual
function index for a given vf. This is just a wrapper around existing functions
to return vf's pf and vf_index with one api call
pciConfigAddressToSysfsfile returns the sysfile pci device link
from a 'struct pci_config_address'
Laine Stump [Mon, 27 Feb 2012 19:20:17 +0000 (14:20 -0500)]
qemu: support type=hostdev network device live hotplug attach/detach
qemuDomainAttachNetDevice
- re-ordered some things at start of function because
networkAllocateActualDevice should always be run and a slot
in def->nets always allocated, but host_net_add isn't needed
if the actual type is hostdev.
- if actual type is hostdev, defer to
qemuDomainAttachHostDevice (which will reach up to the NetDef
for things like MAC address when necessary). After return
from qemuDomainAttachHostDevice, slip directly to cleanup,
since the rest of the function is specific to emulated net
devices.
- put assignment of new NetDef into expanded def->nets down
below cleanup: (but only on success) since it is also needed
for emulated and hostdev net devices.
qemuDomainDetachHostDevice
- after locating the exact device to detach, check if it's a
network device and, if so, use toplevel
qemuDomainDetachNetDevice instead so that the def->nets list
is properly updated, and 'actual device' properly returned to
network pool if appropriate. Otherwise, for normal hostdevs,
call the lower level qemuDomainDetachThisDevice.
qemuDomainDetachNetDevice
- This is where it gets a bit tricky. After locating the device
on the def->nets list, if the network device type == hostdev,
call the *lower level* qemuDomainDetachThisDevice (which will
reach back up to the parent net device for MAC address /
virtualport when appropriate, then clear the device out of
def->hostdevs) before skipping past all the emulated
net-device-specific code to cleanup:, where the network
device is removed from def->nets, and the network device
object is freed.
In short, any time a hostdev-type network device is detached, we must
go through the toplevel virDomaineDetachNetDevice function first and
last, to make sure 1) the def->nnets list is properly managed, and 2)
any device allocated with networkAllocateActualDevice is properly
freed. At the same time, in the middle we need to go through the
lower-level vidDomainDetach*This*HostDevice to be sure that 1) the
def->hostdevs list is properly managed, 2) the PCI device is properly
detached from the guest and reattached to the host (if appropriate),
and 3) any higher level teardown is called at the appropriate time, by
reaching back up to the NetDef config (part (3) will be covered in a
separate patch).
Laine Stump [Thu, 23 Feb 2012 15:45:35 +0000 (10:45 -0500)]
qemu: support type='hostdev' network devices at domain start
This patch makes sure that each network device ("interface") of
type='hostdev' appears on both the hostdevs list and the nets list of
the virDomainDef, and it modifies the qemu driver startup code so that
these devices will be presented to qemu on the commandline as hostdevs
rather than as network devices.
It does not add support for hotplug of these type of devices, or code
to honor the <mac address> or <virtualport> given in the config (both
of those will be done in separate patches).
Once each device is placed on both lists, much of what this patch does
is modify places in the code that traverse all the device lists so
that these hybrid devices are only acted on once - either along with
the other hostdevs, or along with the other network interfaces. (In
many cases, only one of the lists is traversed / a specific operation
is performed on only one type of device. In those instances, the code
can remain unchanged.)
There is one special case - when building the commandline, interfaces
are allowed to proceed all the way through
networkAllocateActualDevice() before deciding to skip the rest of
netdev-specific processing - this is so that (once we have support for
networks with pools of hostdev devices) we can get the actual device
allocated, then rely on the loop processing all hostdevs to generate
the correct commandline.
(NB: <interface type='hostdev'> is only supported for PCI network
devices that are SR-IOV Virtual Functions (VF). Standard PCI[e] and
USB devices, and even the Physical Functions (PF) of SR-IOV devices
can only be assigned to a guest using the more basic <hostdev> device
entry. This limitation is mostly due to the fact that non-SR-IOV
ethernet devices tend to lose mac address configuration whenever the
card is reset, which happens when a card is assigned to a guest;
SR-IOV VFs fortunately don't suffer the same problem.)
This is the new interface type that sets up an SR-IOV PCI network
device to be assigned to the guest with PCI passthrough after
initializing some network device-specific things from the config
(e.g. MAC address, virtualport profile parameters). Here is an example
of the syntax:
This would assign the PCI card from bus 0 slot 4 function 3 on the
host, to bus 0 slot 7 function 0 on the guest, but would first set the
MAC address of the card to 00:11:22:33:44:55.
NB: The parser and formatter don't care if the PCI card being
specified is a standard single function network adapter, or a virtual
function (VF) of an SR-IOV capable network adapter, but the upcoming
code that implements the back end of this config will work *only* with
SR-IOV VFs. This is because modifying the mac address of a standard
network adapter prior to assigning it to a guest is pointless - part
of the device reset that occurs during that process will reset the MAC
address to the value programmed into the card's firmware.
Although it's not supported by any of libvirt's hypervisor drivers,
usb network hostdevs are also supported in the parser and formatter
for completeness and consistency. <source> syntax is identical to that
for plain <hostdev> devices, except that the <address> element should
have "type='usb'" added if bus/device are specified:
Again, the upcoming patch to fill in the backend of this functionality
will log an error and fail with "Unsupported Config" if you actually
try to assign a USB network adapter to a guest using <interface
type='hostdev'> - just use a standard <hostdev> entry in that case
(and also for single-port PCI adapters).
Laine Stump [Mon, 27 Feb 2012 11:53:19 +0000 (06:53 -0500)]
qemu: refactor hotplug detach of hostdevs
This refactoring is necessary to support hotplug detach of
type=hostdev network devices, but needs to be in a separate patch to
make potential debugging of regressions more practical.
Rather than the lowest level functions searching for a matching
device, the search is now done in the toplevel function, and an
intermediate-level function (qemuDomainDetachThisHostDevice()), which
expects that the device's entry is already found, is called (this
intermediate function will be called by qemuDomainDetachNetDevice() in
order to support detach of type=hostdev net devices)
This patch should result in 0 differences in functionality.
Laine Stump [Mon, 27 Feb 2012 06:46:47 +0000 (01:46 -0500)]
conf: hostdev utility functions
Three new functions useful in other files:
virDomainHostdevInsert:
Add a new hostdev at the end of the array. This would more sensibly be
called virDomainHostdevAppend, but the existing functions for other
types of devices are called Insert.
virDomainHostdevRemove:
Eliminates one entry from the hostdevs array, but doesn't free it;
patterned after the code at the end of the two
qemuDomainDetachHostXXXDevice functions (and also other pre-existing
virDomainXXXRemove functions for other device types).
virDomainHostdevFind:
This function is patterned from the search loops at the top of
qemuDomainDetachHostPciDevice and qemuDomainDetachHostUsbDevice, and
will be used to re-factor those (and other detach-related) functions.
Laine Stump [Mon, 27 Feb 2012 06:42:22 +0000 (01:42 -0500)]
conf: put subsys part of virDomainHostdevDef into its own struct
To shorten some new code that accesses the many fields within the
subsys struct of a hostdev, create a separate toplevel, typedefed
virDomainHostdevSubsys struct so that we can define temporary pointers
to the subsys part.
Laine Stump [Fri, 3 Feb 2012 16:13:08 +0000 (11:13 -0500)]
conf: give each hostdevdef a parent pointer
The parent can be any type of device. It defaults to type=none, and a
NULL pointer. The intent is that if a hostdevdef is contained in the
def for a higher level device (e.g. virDomainNetDef), hostdev->parent
will point to the higher level device, and type will be set to that
type of device. This way, during attach and detach of the device,
parent can be checked, and appropriate callouts made to do higher
level device initialization (e.g. setting MAC address).
Also, although these hostdevs with parents will be added to a domain's
hostdevs list, they will be treated slightly differently when
traversing the list, e.g. virDomainHostdefDefFree for a hostdev that
has a parent doesn't need to be called (and will be a NOP); it will
simply be removed from the list (since the parent device object is in
its own type-specific list, and will be freed from there).
Laine Stump [Sun, 19 Feb 2012 21:04:44 +0000 (16:04 -0500)]
conf: HostdevDef parse/format helper functions
In an upcoming patch, virDomainNetDef will acquire a
virDomainHostdevDef, and the <interface> XML will take on some of the
elements of a <hostdev>. To avoid duplicating the code for parsing and
formatting the <source> element (which will be nearly identical in
these two cases), this patch factors those parts out of the
HostdevDef's parse and format functions, and puts them into separate
helper functions that are now called by the HostdevDef
parser/formatter, and will soon be called by the NetDef
parser/formatter.
One change in behavior - previously virDomainHostdevDefParseXML() had
diverged from current common coding practice by logging an error and
failing if it found any subelements of <hostdev> other than those it
understood (standard libvirt practice is to ignore/discard unknown
elements and attributes during parse). The new helper function ignores
unknown elements, and thus so does the new
virDomainHostdevDefParseXML.
Laine Stump [Tue, 31 Jan 2012 09:16:54 +0000 (04:16 -0500)]
conf: make hostdev info a separate object
In order to allow for a virDomainHostdevDef that uses the
virDomainDeviceInfo of a "higher level" device (such as a
virDomainNetDef), this patch changes the virDomainDeviceInfo in the
HostdevDef into a virDomainDeviceInfoPtr. Rather than adding checks
all over the code to check for a null info, we just guarantee that it
is always valid. The new function virDomainHostdevDefAlloc() allocates
a virDomainDeviceInfo and plugs it in, and virDomainHostdevDefFree()
makes sure it is freed.
There were 4 places allocating virDomainHostdevDefs, all of them
parsers of one sort or another, and those have all had their
VIR_ALLOC(hostdev) changed to virDomainHostdevDefAlloc(). Other than
that, and the new functions, all the rest of the changes are just
mechanical removals of "&" or changing "." to "->".
Laine Stump [Wed, 22 Feb 2012 21:06:10 +0000 (16:06 -0500)]
conf: add device pointer to args of virDomainDeviceInfoIterate callback
There will be cases where the iterator callback will need to know the
type of the device whose info is being operated on, and possibly even
need to use some of the device's config. This patch adds a
virDomainDeviceDefPtr to the args of every callback, and fills it in
appropriately as the devices are iterated through.
Laine Stump [Thu, 23 Feb 2012 17:59:21 +0000 (12:59 -0500)]
qemu: rename virDomainDeviceInfoPtr variables to avoid confusion
The virDomainDeviceInfoPtrs in qemuCollectPCIAddress and
qemuComparePCIDevice are named "dev" and "dev1", but those functions
will be changed (in order to match a change in the args sent to
virDomainDeviceInfoIterate() callback args) to contain a
virDomainDeviceDefPtr device.
This patch renames "dev" to "info" (and "dev[n]" to "info[n]") to
avoid later confusion.
projects / libvirt.git / log