buf: Fix possible infinite loop in EscapeString, VSnprintf
The current code will go into an infinite loop if the printf generated
string is >= 1000, AND exactly 1 character smaller than the amount of free
space in the buffer. When this happens, we are dropped into the loop body,
but nothing will actually change, because count == (buf->size - buf->use - 1),
and virBufferGrow returns unchanged if count < (buf->size - buf->use)
Fix this by removing the '- 1' bit from 'size'. The *nprintf functions handle
the NULL byte for us anyways, so we shouldn't need to manually accommodate
for it.
Here's a bug where we are actually hitting this issue:
https://bugzilla.redhat.com/show_bug.cgi?id=602772
v2: Eric's improvements: while -> if (), remove extra va_list variable,
make sure we report buffer error if snprintf fails
v3: Add tests/virbuftest which reproduces the infinite loop before this
patch, works correctly after
Apparently the xen block device statistics moved from
"/sys/devices/xen-backend/vbd-%d-%d/statistics/%s"
to
"/sys/bus/xen-backend/devices/vbd-%d-%d/statistics/%s"
* src/xen/block_stats.c: try the extra path in case of failure to
find the statistics in /sys
virsh: Option for overriding disk type in attach-disk
Unless --driver tap|file option was given to attach-disk, virsh would
generate <disk type='block'> XML which might be fine for Xen but not for
other hypervisors. This patch introduces a new option --sourcetype which
can be used to explicitly set the type of disk source. The option
accepts either "file" or "block" types.
Very occasionally during a parallel make, dispatch.c would
be compiled before the generated remote headers had been
fully written. This would cause it to compile an empty
union, and result in really wierd runtime bugs that are
near impossible to diagnose.
Ensure remote daemon unions are always non-zero length
If the remote daemon args/ret unions ever become zero length
(due to a build / Makefile bug) then bad stuff happens at
runtime. Add a compile time assertion to check for this kind
of problem
A QEMU guest can have upto VIR_DOMAIN_BOOT_LAST boot entries
defined. When building the QEMU arg, each entry takes a
single byte. This means the array must be declared to be
VIR_DOMAIN_BOOT_LAST+1 bytes in length to allow for the
trailing null
For static-only DHCP, i.e. with no <range> but at least one <host>
element within <dhcp> element, we have to add "--dhcp-range IP,static"
option to dnsmasq to actually enable the service. Without this option,
dnsmasq will not respond to DHCP requests.
Before this commit SessionIsActive was not used because ESX(i)
doesn't implement it. vCenter supports SessionIsActive, so use
it here, but keep the fall back mechanism for ESX(i) and GSX.
esx: Fall back to path as key when QueryVirtualDiskUuid isn't available
QueryVirtualDiskUuid is only available on an ESX(i) server. vCenter
returns an NotImplemented fault and a GSX server is missing the
VirtualDiskManager completely. Therefore only use QueryVirtualDiskUuid
with an ESX(i) server and fall back to path as storage volume key for
vCenter and GSX server.
Matthias Bolte [Sun, 29 Aug 2010 17:33:49 +0000 (19:33 +0200)]
esx: Use the VirtualDisk UUID as storage volume key
VirtualDisks are .vmdk file based. Other files in a datastore
like .iso or .flp files don't have a UUID attached, fall back
to the path as key for them.
This patch adds support for ethernet interface type to OpenVZ domains
as stated in this previous message: http://www.redhat.com/archives/libvir-
list/2010-July/msg00658.html
This simplifies code using esxUtil_ParseDatastorePath, because
directoryAndFileName is used more often than fileName. Also the
old approach expected the datastore path to reference an actual
file, but this isn't always correct, especially when listing
volumes. In that case esxUtil_ParseDatastorePath is used to parse
a path that references a directory. This fails for a vpx://
connection because the vCenter returns directory paths with a
trailing '/'. The new approach is robust against this and the
actual decision if the datastore path should reference a file or
a directory is up to the caller of esxUtil_ParseDatastorePath.
Eric Blake [Tue, 31 Aug 2010 22:27:24 +0000 (16:27 -0600)]
build: add some modules
snprintf is currently implicitly picked up by getaddrinfo, but we
might as well make it explicit so that mingw doesn't break if
getaddrinfo changes to drop the dependency.
func doesn't matter for gcc compilation, but may help other compilers
cope with our use of __func__.
* bootstrap.conf (gnulib_modules): Add snprintf and func.
Ryan Harper [Tue, 24 Aug 2010 16:09:00 +0000 (11:09 -0500)]
virsh: remove driver check from attach-disk command
Virsh shouldn't check for driver support but rather let the backend handled this.
After removing the check, I can successfully attach file-based images to a qemu
VM with attach-disk.
% virsh attach-disk vm2 /images/test02.img vdc --driver qemu --type disk --subdriver raw
Disk attached successfully
Soren Hansen [Wed, 25 Aug 2010 09:03:42 +0000 (11:03 +0200)]
Explicitly pass uml_dir argument to user-mode-linux
uml_dir overrides user-mode-linux's default of ~/.uml. This is needed
for a couple of different reasons:
libvirt expects this to default to virGetUserDirectory(geteuid()) +
'/.uml'. However, user-mode-linux actually uses the HOME environment
variable to determine where to look for the uml sockets, but if running
libvirtd under sudo (which I routinely do during development), $HOME is
pointing at my user's homedir, while my euid is 0, so libvirt looks in
/root.
Also (and this was my actual motivation for this patch), if HOME isn't
set at all, user-mode-linux utterly fails. Looking at the code, it seems
it's meant to emit a warning, but alas, it doesn't for some reason.
If running libvirtd from upstart, HOME is not set, so any system using
upstart will need this change.
Jim Fehlig [Tue, 24 Aug 2010 17:31:14 +0000 (11:31 -0600)]
Add blktap2 support to xend driver
Xen4.0 includes a new blktap2 implementation, which is specified
with 'tap2' prefix. AFAICT it's configuration syntax is identical
to blktap, with exception of 'tap2' vs 'tap' prefix. This patch
takes the simple approach of accepting and generating sexp
containing 'tap2' prefix.
Eduardo Otubo [Tue, 17 Aug 2010 19:04:49 +0000 (16:04 -0300)]
PHYP: Bad comparison when checking for existing domain name
When creating a new domain from XML, the check for an existing
domain name should compare the return of the function to a valid
LPAR ID (!= -1) and not to error (== -1).
The check was altered in 8c48743b9737ad4d246ab6043fd299316f9a2091
and got too strict, I've no clue how that snuck in. This check
makes every try to open a connection using the ESX driver fail
with an invalid argument error.
Revert the change to the check and add a comment to prevent future
mistakes with this check.
Jiri Denemark [Wed, 25 Aug 2010 12:07:26 +0000 (14:07 +0200)]
Move libvirt-guests init script and config to tools
Since libvirt-guests init script and its configuration do not require
libvirtd to be running/installed, it was a bad idea to put them into
daemon directory. libvirt.spec even includes these files in
libvirt-client subpackage, which may result in build failure for
client-only builds when the whole daemon directory is just skipped.
When finding a sparse NUMA topology, libnuma will return ENOENT
the first time it is invoked. On subsequent invocations it
will return success, but with an all-1's CPU mask. Check for
this, to avoid polluting the capabilities XML with 4096 bogus
CPUs
Add explicit warning messages when failing to serialize to XDR
When libvirtd fails to serialize a message to XDR the client
connection is terminated immediately. To enable this to be
diagnosed, log the message which caused the problem on the
server
Soren Hansen [Mon, 23 Aug 2010 11:25:50 +0000 (13:25 +0200)]
Allow chardev of type 'file' for UML domains.
Like the comment suggested, we just open the file and pass the file
descriptor to uml. The input "stream" is set to "null", since I couldn't
find any useful way to actually use a file for input for a chardev and
this also mimics what e.g. QEmu does internally.
Matthias Bolte [Sat, 21 Aug 2010 16:49:18 +0000 (18:49 +0200)]
esx: Improve object-by-type lookup performance
Instead of using one big traversal spec for lookup use a set of
more fine grained traversal specs that are selected based on the
actual needs of the lookup.
This gives up to 20% speedup for certain operations like domain
listing due to less HTTP(S) traffic.
Jiri Denemark [Mon, 23 Aug 2010 13:43:51 +0000 (15:43 +0200)]
xen tests: Fix PV-VFB tests with RHEL-5 API
RHEL-5 Xen doesn't support the old style vnc configuration. In sexpr, we
can't really check it with rhel5-api turned on. However, for XM
configuration files it's sufficient to use cfg version 1 instead of 2.
Jiri Denemark [Mon, 23 Aug 2010 12:58:08 +0000 (14:58 +0200)]
xen tests: Fix missing "type ioemu" with rhel5-api
The most common cause of errors with rhel5-api turn on was missing
"(type ioemu)" in sexpr or its equivalent in XM configuration file. This
happens because the presence of that part in sexpr (or cfg) depends on
xen version the host is running. Let's avoid it by explicitly specifying
interface model which ensures "type ioemu" will always be emitted.
This patch adds
<model type='e1000'/>
withing the interface element in all affected xml files. And
(model 'e1000')
to all corresponding sexpr files with similar fix to cfg files. Such
configuration works regardless on Xen version.
Eric Blake [Fri, 20 Aug 2010 23:35:54 +0000 (17:35 -0600)]
xenapi: support xenapi 5.6.0 headers
* src/xenapi/xenapi_driver.c (xenapiDomainGetInfo): Avoid using
XEN_VM_POWER_STATE_UNKNOWN, which disappeared in newer xenapi.
* src/xenapi/xenapi_utils.c (mapPowerState): Likewise.
Previously QEMU enabled KQEMU by default and had -no-kqemu.
0.11.x switched to requiring -enable-kqemu. 0.12.x dropped
kqemu entirely. This patch adds support for -enable-kqemu
so 0.11.x works. It replaces a huge set of if() with a
switch() to make the code a bit more readable.
* src/qemu/qemu_conf.c, src/qemu/qemu_conf.h: Support
-enable-kqemu
Matthias Bolte [Sun, 8 Aug 2010 18:45:12 +0000 (20:45 +0200)]
esx: Use MD5 sum of mount path as storage pool UUID
With the previous storage pool UUID source not all storage pools
had a proper UUID, especially GSX storage pools. The mount path
is unique per host and cannot change during the lifetime of the
datastore. Therefore, it's MD5 sum can be used as UUID.
Use gnulib's crypto/md5 module to generate the MD5 sum.
Jiri Denemark [Fri, 20 Aug 2010 11:57:35 +0000 (13:57 +0200)]
qemu: Remove code duplication
We already filled the PCI address structure when we checked whether it's
free or not, so let's just use the structure here instead of filling it
again.
Patrick Dignan [Tue, 17 Aug 2010 17:44:27 +0000 (12:44 -0500)]
storage: add support for Vendor and Model in XML
I wrote a patch to add support for listing the Vendor and Model of a
storage pool in the storage pool XML. This would allow vendor
extensions of specific devices. The patch includes a test for the new
attributes as well.
Eric Blake [Mon, 16 Aug 2010 21:21:38 +0000 (15:21 -0600)]
uml: fix logic bug in checking reply length
* src/uml/uml_driver.c (umlMonitorCommand): Validate that enough
bytes were read to dereference both res.length, and that many
bytes from res.data.
Reported by Soren Hansen.
Eric Blake [Wed, 18 Aug 2010 19:43:26 +0000 (13:43 -0600)]
build: fix compiler warning
node_device/node_device_driver.c: In function 'nodeDeviceVportCreateDelete':
node_device/node_device_driver.c:423: error: implicit declaration of function 'stat' [-Wimplicit-function-declaration]
Daniel Veillard [Tue, 17 Aug 2010 16:30:17 +0000 (18:30 +0200)]
xen: Fix scheduler setting problems
Doing `virsh schedinfo rhel5u3 --cap 65535' the hypervisor does the
call, but does not change the value nor raise an error. Best is just to
consider it's not in the allowed values. The problem is that the error
won't be output since the xend driver will then be called and raise an
error
error: this function is not supported by the hypervisor: unsupported
in xendConfigVersion < 4
which will override the useful information from
xenUnifiedDomainSetSchedulerParameters(). So best is to also invert the
order in which the xen sub-drivers are called.
* src/xen/xen_hypervisor.c: mark 65535 cap value as out of bound
* src/xen/xen_hypervisor.c: reverse the order of the calls to the xen
sub drivers to get the error message if needed
Dave Allan [Tue, 17 Aug 2010 15:21:16 +0000 (17:21 +0200)]
nodedev: Fix sysfs paths for vport operations
Some kernels, such as the one used in RHEL-5, have vport_create and
vport_delete operation files in /sys/class/scsi_host/hostN directory
instead of /sys/class/fc_host/hostN. Let's check both paths for
compatibility reasons.
This also removes unnecessary '/' characters from sysfs paths containing
LINUX_SYSFS_FC_HOST_PREFIX.
Daniel Berrange [Mon, 9 Aug 2010 12:16:04 +0000 (14:16 +0200)]
remote: Fix incorrect use of private data field
NodeDeviceCreateXML and NodeDeviceDestroy methods added for NPIV were
using the wrong privateData field for the remote driver. This doesn't
impact KVM, since the remote driver handles everything, thus
privateData == devMonPrivateData. It does impact Xen though, because
the remote driver only handles a subset of methods and thus
privateData != devMonPrivateData.
Chris Lalancette [Fri, 13 Aug 2010 13:23:11 +0000 (09:23 -0400)]
Fix up qemu domain save/managed save locking.
The current version of the qemu managed save implementation
is subject to a race where the domain shuts down between
the time that we start the command and the time that we
actually try to do the save. Close this race by making
qemuDomainSaveFlags() expect both the driver and the passed-in
vm object to be locked before executing.
Signed-off-by: Chris Lalancette <clalance@redhat.com>
Jiri Denemark [Fri, 13 Aug 2010 14:16:19 +0000 (16:16 +0200)]
qemu: Re-reserve all PCI addresses on libvirtd restart
When reconnecting to existing VMs, we re-reserved only those PCI
addresses which were explicitly mentioned in domain XML. Since some
addresses are always reserved (e.g., 0:0:0 and 0:0:1), we need to handle
those too.
Also all this should only be done if device flag is supported by qemu.
Stefan Berger [Mon, 16 Aug 2010 16:59:54 +0000 (12:59 -0400)]
nwfilter: extend nwfilter reload support
In this patch I am extending and fixing the nwfilter module's reload support to stop all ongoing threads (for learning IP addresses of interfaces) and rebuild the filtering rules of all interfaces of all VMs when libvirt is started. Now libvirtd rebuilds the filters upon the SIGHUP signal and libvirtd restart.
About the patch: The nwfilter functions require a virConnectPtr. Therefore I am opening a connection in qemudStartup, which later on needs to be closed outside where the driver lock is held since otherwise it ends up in a deadlock due to virConnectClose() trying to lock the driver as well.
I have tested this now for a while with several machines running and needing the IP address learner thread(s). The rebuilding of the firewall rules seems to work fine following libvirtd restart or a SIGHUP. Also the termination of libvirtd worked fine.
Matthias Bolte [Fri, 6 Aug 2010 00:02:03 +0000 (02:02 +0200)]
esx: Explicitly disable unused floppy devices
floppy0.present defaults to true. Therefore, it needs to be
explicitly set to false when the XML config doesn't specify the
corresponding floppy device.
Eduardo Otubo [Thu, 12 Aug 2010 14:33:12 +0000 (11:33 -0300)]
PHYP: Add rudimentary network driver
I changed virStorage[Open|Close] to virVIOSDriver[Open|Close] so
the network driver can use it - since the network driver deals
with Open/Close in the same way.
Soren Hansen [Thu, 12 Aug 2010 13:06:34 +0000 (15:06 +0200)]
Make umlConnectTapDevice ask brAddTap for a persistent tap device.
This patch does two things:
* It makes umlConnectTapDevice ask brAddTap for a persistent tap by
passing it a NULL tapfd argument.
* Stops umlConnectTapDevice from immediately dismantling the bridge
it just set up.
Soren Hansen [Wed, 11 Aug 2010 21:51:41 +0000 (23:51 +0200)]
Close fd's of persistent tap devices
When passing a NULL tapfd argument to brAddTap, we need to close the fd
of the tap device. If we don't, libvirt will keep the fd open
indefinitely and renders the the guest unable to configure its side of
the tap device.
Soren Hansen [Thu, 12 Aug 2010 13:42:34 +0000 (15:42 +0200)]
Make sure all command line arguments get passed to UML
If umlBuildCommandLineChr fails (e.g. due to an unsupported chardev
type), it returns NULL. umlBuildCommandLine does not check for this and
sets this as an argument on the comand line, effectively ending the
argument list. This patch checks for this case and sets the chardev to
"none".
Stefan Berger [Fri, 13 Aug 2010 20:41:39 +0000 (16:41 -0400)]
nwfilter: Discard class D,E IP addresses when sniffing pkts
When sniffing the network traffic, discard class D and E IP addresses when sniffing traffic. This was a reason why filters were not correctly rebuilt on VMs on the local 192.* network when libvirt was restarted and those VMs did not use a DHCP request to get its IP address.
Stefan Berger [Fri, 13 Aug 2010 19:47:10 +0000 (15:47 -0400)]
nwfilter: serialize execution of scripts with ebtables cmds
While testing the SIGHUP handling and reloading of the nwfilter driver, I found that when the filters are rebuilt and mutlipe threads handled the individual interfaces, concurrently running multiple external bash scripts causes strange failures even though the executed ebtables commands are working on different tables for different interfaces. I cannot say for sure where the concurrency problems are caused, but introducing this lock definitely helps.
Chris Lalancette [Thu, 12 Aug 2010 16:51:31 +0000 (12:51 -0400)]
Move the tunnelled migration unix socket to /var/lib/libvirt/qemu
Since the qemu process is running as qemu:qemu, it can't actually
look at the unix socket in /var/run/libvirt/qemu which is owned by
root and has permission 700. Move the unix socket to
/var/lib/libvirt/qemu, which is already owned by qemu:qemu.
Thanks to Justin Clift for test this out for me.
Signed-off-by: Chris Lalancette <clalance@redhat.com>