Commit Graph

9103 Commits

Author SHA1 Message Date
Michal Privoznik
06b9c5b923 virCommand: Properly handle POLLHUP
It is a good practise to set revents to zero before doing any poll().
Moreover, we should check if event we waited for really occurred or
if any of fds we were polling on didn't encountered hangup.
2012-01-04 10:40:23 +01:00
Yuri Chornoivan
524ba58bb9 Fix typos in messages.
https://bugzilla.redhat.com/show_bug.cgi?id=770954
2012-01-03 20:30:33 -07:00
Jiri Denemark
66ca7ce573 virCPUDefCopy forgot to copy NUMA topology
As a result of it, guest NUMA topology would be lost during migration.
2012-01-03 21:05:54 +01:00
Eric Blake
851fc8139f qemu: fix block stat naming
Typo has existed since API introduction in commit ee0d8c3.

* src/qemu/qemu_driver.c (qemuDomainBlockStatsFlags): Use correct
name.
2012-01-02 20:43:07 -07:00
Eric Blake
269ce467fc domiftune: clean up previous patches
Most severe here is a latent (but currently untriggered) memory leak
if any hypervisor ever adds a string interface property; the
remainder are mainly cosmetic.

* include/libvirt/libvirt.h.in (VIR_DOMAIN_BANDWIDTH_*): Move
macros closer to interface that uses them, and document type.
* src/libvirt.c (virDomainSetInterfaceParameters)
(virDomainGetInterfaceParameters): Formatting tweaks.
* daemon/remote.c (remoteDispatchDomainGetInterfaceParameters):
Avoid memory leak.
* src/libvirt_public.syms (LIBVIRT_0.9.9): Sort lines.
* src/libvirt_private.syms (domain_conf.h): Likewise.
* src/qemu/qemu_driver.c (qemuDomainSetInterfaceParameters): Fix
comments, break long lines.
2012-01-02 14:35:12 -07:00
Peter Krempa
d82c6bcf30 virsh: Fix checking for reconnect conditions
virshReportError() function frees the most recent error reported from
libvirt. Condition that checks if connection to the daemon was broken
during last command was then limited to check for SIGPIPE signal not
taking into account possible errors signalized without SIGPIPE.

This patch moves the check before the error is freed, to take into
account code that does not emit SIGPIPE while failing.

* tools/virsh.c: - move check for broken connection before error print.
2012-01-02 18:01:09 +01:00
Peter Krempa
f4384b8439 network_conf: Fix whitespace to pass syntax-check 2012-01-02 17:59:05 +01:00
Michal Novotny
973af2362c Implement DNS SRV record into the bridge driver
Hi,
this is the fifth version of my SRV record for DNSMasq patch rebased
for the current codebase to the bridge driver and libvirt XML file to
include support for the SRV records in the DNS. The syntax is based on
DNSMasq man page and tests for both xml2xml and xml2argv were added as
well. There are some things written a better way in comparison with
version 4, mainly there's no hack in tests/networkxml2argvtest.c and
also the xPath context is changed to use a simpler query using the
virXPathInt() function relative to the current node.

Also, the patch is also fixing the networkxml2argv test to pass both
checks, i.e. both unit tests and also syntax check.

Please review,
Michal

Signed-off-by: Michal Novotny <minovotn@redhat.com>
2012-01-02 23:05:55 +08:00
Alex Jia
baade4cd2b qemu: Fix bandwidth memory leak on failure
Detected by Coverity. Leaks introduced in commit e8d6b29.

Signed-off-by: Alex Jia <ajia@redhat.com>
2011-12-31 16:42:23 -07:00
Eric Blake
8267aea5a6 qemu: fix blkio memory leak on failure
Leak detected by Coverity, and introduced in commit 93ab585.
Reported by Alex Jia.

* src/qemu/qemu_driver.c (qemuDomainSetBlkioParameters): Free
devices array on error.
2011-12-31 16:32:35 -07:00
Eric Blake
bc1b27e07e docs: re-fix stray /
Commit 6cb4acc reintroduced the bug fixed in commit d145fe3.

* docs/formatdomain.html.in (elementsDisks): Fix again.
2011-12-31 06:42:01 -07:00
Daniel Veillard
c4ac050fcb Fix build on s390(x) and other stange arches
The blocks to extract node information on a per-arch
basis wasn't well balanced leading to a compilation
failure if not on one of the handled arches (PCs and PPCs)
2011-12-30 14:15:26 +08:00
Eric Blake
904e05a292 seclabel: honor device override in selinux
This wires up the XML changes in the previous patch to let SELinux
labeling honor user overrides, as well as affecting the live XML
configuration in one case where the user didn't specify anything
in the offline XML.

I noticed that the logs contained messages like this:

2011-12-05 23:32:40.382+0000: 26569: warning : SELinuxRestoreSecurityFileLabel:533 : cannot lookup default selinux label for /nfs/libvirt/images/dom.img

for all my domain images living on NFS.  But if we would just remember
that on domain creation that we were unable to set a SELinux label (due to
NFSv3 lacking labels, or NFSv4 not being configured to expose attributes),
then we could avoid wasting the time trying to clear the label on
domain shutdown.  This in turn is one less point of NFS failure,
especially since there have been documented cases of virDomainDestroy
hanging during an attempted operation on a failed NFS connection.

* src/security/security_selinux.c (SELinuxSetFilecon): Move guts...
(SELinuxSetFileconHelper): ...to new function.
(SELinuxSetFileconOptional): New function.
(SELinuxSetSecurityFileLabel): Honor override label, and remember
if labeling failed.
(SELinuxRestoreSecurityImageLabelInt): Skip relabeling based on
override.
2011-12-30 10:57:59 +08:00
Eric Blake
b43432931a seclabel: allow a seclabel override on a disk src
Implement the parsing and formatting of the XML addition of
the previous commit.  The new XML doesn't affect qemu command
line, so we can now test round-trip XML->memory->XML handling.

I chose to reuse the existing structure, even though per-device
override doesn't use all of those fields, rather than create a
new structure, in order to reuse more code.

* src/conf/domain_conf.h (_virDomainDiskDef): Add seclabel member.
* src/conf/domain_conf.c (virDomainDiskDefFree): Free it.
(virSecurityLabelDefFree): New function.
(virDomainDiskDefFormat): Print it.
(virSecurityLabelDefFormat): Reduce output if model not present.
(virDomainDiskDefParseXML): Alter signature, and parse seclabel.
(virSecurityLabelDefParseXML): Split...
(virSecurityLabelDefParseXMLHelper): ...into new helper.
(virDomainDeviceDefParse, virDomainDefParseXML): Update callers.
* tests/qemuxml2argvdata/qemuxml2argv-seclabel-dynamic-override.args:
New file.
* tests/qemuxml2xmltest.c (mymain): Enhance test.
* tests/qemuxml2argvtest.c (mymain): Likewise.
2011-12-30 10:57:59 +08:00
Eric Blake
6cb4acce8b seclabel: extend XML to allow per-disk label overrides
When doing security relabeling, there are cases where a per-file
override might be appropriate.  For example, with a static label
and relabeling, it might be appropriate to skip relabeling on a
particular disk, where the backing file lives on NFS that lacks
the ability to track labeling.  Or with dynamic labeling, it might
be appropriate to use a custom (non-dynamic) label for a disk
specifically intended to be shared across domains.

The new XML resembles the top-level <seclabel>, but with fewer
options (basically relabel='no', or <label>text</label>):

<domain ...>
  ...
  <devices>
    <disk type='file' device='disk'>
      <source file='/path/to/image1'>
        <seclabel relabel='no'/> <!-- override for just this disk -->
      </source>
      ...
    </disk>
    <disk type='file' device='disk'>
      <source file='/path/to/image1'>
        <seclabel relabel='yes'> <!-- override for just this disk -->
          <label>system_u:object_r:shared_content_t:s0</label>
        </seclabel>
      </source>
      ...
    </disk>
    ...
  </devices>
  <seclabel type='dynamic' model='selinux'>
    <baselabel>text</baselabel> <!-- used for all devices without override -->
  </seclabel>
</domain>

This patch only introduces the XML and documentation; future patches
will actually parse and make use of it.  The intent is that we can
further extend things as needed, adding a per-device <seclabel> in
more places (such as the source of a console device), and possibly
allowing a <baselabel> instead of <label> for labeling where we want
to reuse the cNNN,cNNN pair of a dynamically labeled domain but a
different base label.

First suggested by Daniel P. Berrange here:
https://www.redhat.com/archives/libvir-list/2011-December/msg00258.html

* docs/schemas/domaincommon.rng (devSeclabel): New define.
(disk): Use it.
* docs/formatdomain.html.in (elementsDisks, seclabel): Document
the new XML.
* tests/qemuxml2argvdata/qemuxml2argv-seclabel-dynamic-override.xml:
New test, to validate RNG.
2011-12-30 10:57:58 +08:00
Eric Blake
e83837945c seclabel: move seclabel stuff earlier
Pure code motion; no semantic change.

* src/conf/domain_conf.h (virDomainSeclabelType)
(virSecurityLabelDefPtr): Declare earlier.
* src/conf/domain_conf.c (virSecurityLabelDefClear)
(virSecurityLabelDefParseXML): Move earlier.
(virDomainDefParseXML): Move seclabel parsing earlier.
2011-12-30 10:38:37 +08:00
Eric Blake
336df7966b seclabel: refactor existing domain_conf usage
A future patch will parse and output <seclabel> in more than one
location in a <domain> xml; make it easier to reuse code.

* src/conf/domain_conf.c (virSecurityLabelDefFree): Rename...
(virSecurityLabelDefClear): ...and make static.
(virSecurityLabelDefParseXML): Alter signature.
(virDomainDefParseXML, virDomainDefFree): Adjust callers.
(virDomainDefFormatInternal): Split output...
(virSecurityLabelDefFormat): ...into new helper.
2011-12-30 10:38:37 +08:00
Eric Blake
116d6af979 schema: rewrite seclabel rng to match code
The RNG for <seclabel> was too strict - if it was present, then it
had to have sub-elements, even if those didn't make sense for the
given attributes.  Also, we didn't have any tests of <seclabel>
parsing or XML output.

In this patch, I added more parsing tests than output tests (since
the output populates and/or reorders fields not present in certain
inputs).  Making the RNG reliable is a precursor to using <seclabel>
variants in more places in the XML in later patches.

See also:
http://berrange.com/posts/2011/09/29/two-small-improvements-to-svirt-guest-configuration-flexibility-with-kvmlibvirt/

* docs/schemas/domaincommon.rng (seclabel): Tighten rules.
* tests/qemuxml2argvtest.c (mymain): New tests.
* tests/qemuxml2xmltest.c (mymain): Likewise.
* tests/qemuxml2argvdata/qemuxml2argv-seclabel-*.*: New files.
2011-12-30 10:38:37 +08:00
Daniel Veillard
cbc7025942 Require avahi as an rpm dependancy
As seen in https://bugzilla.redhat.com/show_bug.cgi?id=746111

If compiled with the daemon and avahi support, then it's better
to require avahi at the rpm level to avoid daemon startup errors.
2011-12-29 22:01:57 +08:00
Hu Tao
b2310b2913 domiftune: Enable the virDomain{S,G}etInterfaceParameters in virsh
Add a new command domiftune to get/set interface parameters.

* tools/virsh.c: implement the new command
* tools/virsh.pod: documentation of the new command
2011-12-29 18:29:25 +08:00
Hu Tao
e8d6b293d8 domiftune: Add virDomain{S,G}etInterfaceParameters support to qemu driver
* src/qemu/qemu_driver.c: implement the qemu driver support
2011-12-29 18:28:47 +08:00
Hu Tao
ee3de186b3 domiftune: Add a util function virDomainNetFind
Add a util function virDomainNetFind to find a domain's net def.
2011-12-29 18:27:35 +08:00
Hu Tao
e7dfe00d06 domiftune: Add support of new APIs to the remote driver
* daemon/remote.c: implement the server side support
* src/remote/remote_driver.c: implement the client side support
* src/remote/remote_protocol.x: definitions for the new entry points
* src/remote_protocol-structs: structure definitions
2011-12-29 18:25:26 +08:00
Hu Tao
51fded0be9 domiftune: virDomain{S,G}etInterfaceParameters: the main entry points
* src/libvirt.c: implement the main entry points
2011-12-29 18:25:12 +08:00
Hu Tao
85f3493f34 domiftune: Add API virDomain{S,G}etInterfaceParameters
The APIs are used to set/get domain's network interface's parameters.
Currently supported parameters are bandwidth settings.

* include/libvirt/libvirt.h.in: new API and parameters definition
* python/generator.py: skip the Python API generation
* src/driver.h: add new entry to the driver structure
* src/libvirt_public.syms: export symbols
2011-12-29 18:24:43 +08:00
Daniel Veillard
f0293edc3f remove a static limit on max domains in python bindings
* python/libvirt-override.c: remove the predefined array in the
  virConnectListDomainsID binding and call virConnectNumOfDomains
  to do a proper allocation
2011-12-29 16:20:00 +08:00
Alex Jia
ae3315aa4a python: Fix problems of virDomain{Set, Get}BlockIoTune bindings
The parameter 'params' is useless for virDomainGetBlockIoTune API,
and the return value type should be a virTypedParameterPtr but not
integer. And "PyArg_ParseTuple" in functions
libvirt_virDomain{Set,Get}BlockIoTune misses format unit for "format"
argument.

* libvirt-override-api.xml: Remove useless the parameter 'params'
from virDomainGetBlockIoTune API, and change return value type from
integer to virTypedParameterPtr.

* python/libvirt-override.c: Add the missed format units.

RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=770683

Signed-off-by: Alex Jia <ajia@redhat.com>
2011-12-29 14:07:47 +08:00
Eric Blake
1a3f6608aa qemu: fix inf-loop in blkio parameters
https://bugzilla.redhat.com/show_bug.cgi?id=770520

We had two nested loops both trying to use 'i' as the iteration
variable, which can result in an infinite loop when the inner
loop interferes with the outer loop.  Introduced in commit 93ab585.

* src/qemu/qemu_driver.c (qemuDomainSetBlkioParameters): Don't
reuse iteration variable across two loops.
2011-12-28 06:57:42 -07:00
Lai Jiangshan
96b3716c2a virsh: move version command to host group
Trivial patch, move version command to host commands group.

It has no any related with any domain.

It may connect to the daemon, so the flag is 0 but not VSH_CMD_FLAG_NOCONNECT.
2011-12-28 05:18:51 -07:00
Eric Blake
e957b67061 daemon: clean up daemonization
Valgrind detected a pipe fd leak before the parent exits on success,
introduced in commit 4296cea; by itself, the leak is not bad, since
we immediately called _exit(), but we might as well be clean to make
valgrind analysis easier.  Meanwhile, if the daemon grandchild detects
an error, the parent failed to flush the error message before exiting.
Also, we had the possibility of both parent and child returning to the
caller, such that the user could see duplicated reports of failure
from the two return paths.  And we might as well be robust to the
(unlikely) situation of being started with stdin closed.

* daemon/libvirtd.c (daemonForkIntoBackground): Use exit if an
error message was generated, avoid fd leaks for valgrind's sake,
avoid returning to caller in both parent and child, and don't
close a just-dup'd stdin.
Based on a report by Alex Jia.

* How to reproduce?
  % service libvirtd stop
  % valgrind -v --track-fds=yes /usr/sbin/libvirtd --daemon

* Actual valgrind result:

==16804== FILE DESCRIPTORS: 7 open at exit.
==16804== Open file descriptor 7:
==16804==    at 0x321FAD8B87: pipe (in /lib64/libc-2.12.so)
==16804==    by 0x41F34D: daemonForkIntoBackground (libvirtd.c:186)
==16804==    by 0x4207A0: main (libvirtd.c:1420)

Signed-off-by: Alex Jia <ajia@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
2011-12-28 05:14:38 -07:00
Satoru SATOH
1f24ddf6bf docs: Move 'echo' command description into the generic commands section
Virsh's echo command looks not having any relations with domains and its
description should go into the generic commands section instead of the
domain commands section (current).
2011-12-26 11:29:23 +08:00
Satoru SATOH
b2c545bf80 docs: Move 'send-key' command description into the domain commands section
Virsh's send-key command manipulates domains and its description should
go into the domain commands section instead of generic commands section
(current).
2011-12-26 11:21:24 +08:00
Michal Privoznik
8a34f822e6 qemu: Keep list of USB devices attached to domains
In order to avoid situation where a USB device is
in use by two domains, we must keep a list of already
attached devices like we do for PCI.
2011-12-24 18:12:04 +01:00
Eric Blake
d145fe3bb3 docs: remove stray /
Commit e5a84d74 added a new attribute in the wrong location;
commit c8b9fa74 fixed the missing / at the end but not the extra
/ in the middle.

* docs/formatdomain.html.in (elementsDisks): Fix another typo.
2011-12-23 12:04:19 -07:00
Eric Blake
c8b9fa7434 docs: fix missing / in xml examples
* docs/formatdomain.html.in: Fix typos in examples.
2011-12-22 13:03:50 -07:00
Eric Blake
dc099b8338 tests: fix schema checks sorting
Commit 6fdbce12 attempted to sort the list of tests, but failed
(without quotes, echo merges all the tests into a single line,
so there was nothing to sort).

* tests/schematestutils.sh: Fix thinko in previous patch.
2011-12-22 13:01:09 -07:00
Michal Privoznik
d8db0f9690 qemu: Support for overriding NOFILE limit
This patch adds max_files option to qemu.conf which can be used to
override system default limit on number of opened files that are
allowed for qemu user.
2011-12-22 17:49:04 +01:00
Michal Privoznik
74ad69b708 virsh: Use vshWatchJob in cmdManagedSave
This patch alters saving code, so we can report progress
and allow cancel via ^C.
2011-12-22 11:31:51 +01:00
Michal Privoznik
6e4c540d28 virsh: Use vshWatchJob in cmdSave
This patch alters saving code, so we can report progress
and allow cancel via ^C.
2011-12-22 11:31:24 +01:00
Michal Privoznik
5ccc7f6488 virsh: Use vshWatchJob in cmdDump
This patch alters dumping code, so we can report progress
and allow cancel via ^C.
2011-12-22 11:11:28 +01:00
Michal Privoznik
8e7829f754 virsh: Move job watch code to a separate function
called vshWatchJob. This can be later used in other
job oriented commands like dump, save, managedsave
to report progress and allow user to cancel via ^C.
2011-12-22 11:05:15 +01:00
Michal Privoznik
6da91758d7 qemuhelptest: Add new qemuCap flag
Latest patch a1a83c5874 introduces new qemu capability flag
QEMU_CAPS_FSDEV_READONLY. However, it was missing in qemuhelptest
making test for qemu-1.0 fail.
2011-12-22 11:00:05 +01:00
Osier Yang
a1a83c5874 qemu: Support readonly filesystem passthrough
Upstream QEMU starts to support it from commit 2c74c2cb.
2011-12-22 12:29:58 +08:00
Stefan Berger
1c8f0cbb83 nwfilter: Do not require DHCP requests to be broadcasted
Remove the requirement that DHCP messages have to be broadcasted.
DHCP requests are most often sent via broadcast but can be directed
towards a specific DHCP server. For example 'dhclient' takes '-s <server>'
as a command line parameter thus allowing DHCP requests to be sent to a
specific DHCP server.
2011-12-21 10:54:47 -05:00
Osier Yang
33eca17f6a qemu: Release the lock on domobj if fails on finding the disk path 2011-12-21 10:22:08 +08:00
Eric Blake
6fdbce1232 tests: run schema checks in sorted order
Having a test that depends on file system timestamps and/or inode
allocation order gives non-deterministic output.

* tests/schematestutils.sh: Run test in deterministic order.
2011-12-20 17:34:12 -07:00
Michael Ellerman
bbae92f8b8 tests: Add fake PPC64 emulator for QEMU testing
Create a fake PPC64 QEMU so that we can run PPC64 QEMU tests when we
don't have a real version of the emulator available.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
2011-12-20 16:14:09 -07:00
Michael Ellerman
d64955a91a qemu: Add spapr-vio address assignment
Add logic to assign addresses for devices with spapr-vio addresses.

We also do validation of addresses specified by the user, ie. ensuring
that there are not duplicate addresses on the bus.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
2011-12-20 16:09:21 -07:00
Bharata B Rao
4ba56a9410 Add New address type spapr-vio to domain.rng
Original patch by Bharata. Updated to use {1,16} in spaprvioReg based
on example from Eric Blake.

Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
Signed-off-by: Prerna Saxena <prerna@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
2011-12-20 16:01:25 -07:00
Michael Ellerman
7e4d896b5e Add address type for SPAPR VIO devices
For QEMU PPC64 we have a machine type ("pseries") which has a virtual
bus called "spapr-vio". We need to be able to create devices on this
bus, and as such need a way to specify the address for those devices.

This patch adds a new address type "spapr-vio", which achieves this.

The addressing is specified with a "reg" property in the address
definition. The reg is optional, if it is not specified QEMU will
auto-assign an address for the device.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
2011-12-20 15:39:16 -07:00