https://bugzilla.redhat.com/show_bug.cgi?id=924153
Commit 904e05a2 (v0.9.9) added a per-<disk> seclabel element with
an attribute relabel='no' in order to try and minimize the
impact of shutdown delays when an NFS server disappears. The idea
was that if a disk is on NFS and can't be labeled in the first
place, there is no need to attempt the (no-op) relabel on domain
shutdown. Unfortunately, the way this was implemented was by
modifying the domain XML so that the optimization would survive
libvirtd restart, but in a way that is indistinguishable from an
explicit user setting. Furthermore, once the setting is turned
on, libvirt avoids attempts at labeling, even for operations like
snapshot or blockcopy where the chain is being extended or pivoted
onto non-NFS, where SELinux labeling is once again possible. As
a result, it was impossible to do a blockcopy to pivot from an
NFS image file onto a local file.
The solution is to separate the semantics of a chain that must
not be labeled (which the user can set even on persistent domains)
vs. the optimization of not attempting a relabel on cleanup (a
live-only annotation), and using only the user's explicit notation
rather than the optimization as the decision on whether to skip
a label attempt in the first place. When upgrading an older
libvirtd to a newer, an NFS volume will still attempt the relabel;
but as the avoidance of a relabel was only an optimization, this
shouldn't cause any problems.
In the ideal future, libvirt will eventually have XML describing
EVERY file in the backing chain, with each file having a separate
<seclabel> element. At that point, libvirt will be able to track
more closely which files need a relabel attempt at shutdown. But
until we reach that point, the single <seclabel> for the entire
<disk> chain is treated as a hint - when a chain has only one
file, then we know it is accurate; but if the chain has more than
one file, we have to attempt relabel in spite of the attribute,
in case part of the chain is local and SELinux mattered for that
portion of the chain.
* src/conf/domain_conf.h (_virSecurityDeviceLabelDef): Add new
member.
* src/conf/domain_conf.c (virSecurityDeviceLabelDefParseXML):
Parse it, for live images only.
(virSecurityDeviceLabelDefFormat): Output it.
(virDomainDiskDefParseXML, virDomainChrSourceDefParseXML)
(virDomainDiskSourceDefFormat, virDomainChrDefFormat)
(virDomainDiskDefFormat): Pass flags on through.
* src/security/security_selinux.c
(virSecuritySELinuxRestoreSecurityImageLabelInt): Honor labelskip
when possible.
(virSecuritySELinuxSetSecurityFileLabel): Set labelskip, not
norelabel, if labeling fails.
(virSecuritySELinuxSetFileconHelper): Fix indentation.
* docs/formatdomain.html.in (seclabel): Document new xml.
* docs/schemas/domaincommon.rng (devSeclabel): Allow it in RNG.
* tests/qemuxml2argvdata/qemuxml2argv-seclabel-*-labelskip.xml:
* tests/qemuxml2argvdata/qemuxml2argv-seclabel-*-labelskip.args:
* tests/qemuxml2xmloutdata/qemuxml2xmlout-seclabel-*-labelskip.xml:
New test files.
* tests/qemuxml2argvtest.c (mymain): Run the new tests.
* tests/qemuxml2xmltest.c (mymain): Likewise.
Signed-off-by: Eric Blake <eblake@redhat.com>
Use the new semantics of vshStringToArray to avoid leaking the array of
volumes to be deleted. The array would be leaked in case the first
volume was found in the domain definition. Also refactor the code a bit
to sanitize naming of variables hoding arrays and dimensions of the
arrays.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=996050
At a slightly larger memory expense allow stealing of items from the
string array returned from vshStringToArray and turn the result into a
string list compatible with virStringSplit. This will allow to use the
common dealloc function.
This patch also fixes a few forgotten checks of return from
vshStringToArray and one memory leak.
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.
Since 16bcb3 we have a regression. The hard_limit is set
unconditionally. By default the limit is zero. Hence, if user hasn't
configured any, we set the zero in cgroup subsystem making the kernel
kill the corresponding qemu process immediately. The proper fix is to
set hard_limit iff user has configured any.
We were failing to autoprobe which schema to use for several
top-level XML elements.
* tools/virt-xml-validate.in (TYPE): Recognize <domainsnapshot>,
<filter>, and <secret>.
Signed-off-by: Eric Blake <eblake@redhat.com>
All good tools should have --help and --version output :)
Furthermore, we want to ensure a failed exit if xmllint fails,
or even for 'virt-xml-validate > /dev/full'.
* tools/virt-xml-validate.in: Add option parsing. Output errors
to stderr. Update documentation to match.
* tools/Makefile.am (virt-xml-validate): Substitute version.
Signed-off-by: Eric Blake <eblake@redhat.com>
From: Dario Faggioli <dario.faggioli@citrix.com>
Starting from Xen 4.2, libxl has all the bits and pieces in place
for retrieving an adequate amount of information about the host
NUMA topology. It is therefore possible, after a bit of shuffling,
to arrange those information in the way libvirt wants to present
them to the outside world.
Therefore, with this patch, the <topology> section of the host
capabilities is properly populated, when running on Xen, so that
we can figure out whether or not we're running on a NUMA host,
and what its characteristics are.
[raistlin@Zhaman ~]$ sudo virsh --connect xen:/// capabilities
<capabilities>
<host>
<cpu>
....
<topology>
<cells num='2'>
<cell id='0'>
<memory unit='KiB'>6291456</memory>
<cpus num='8'>
<cpu id='0' socket_id='1' core_id='0' siblings='0-1'/>
<cpu id='1' socket_id='1' core_id='0' siblings='0-1'/>
<cpu id='2' socket_id='1' core_id='1' siblings='2-3'/>
<cpu id='3' socket_id='1' core_id='1' siblings='2-3'/>
<cpu id='4' socket_id='1' core_id='9' siblings='4-5'/>
<cpu id='5' socket_id='1' core_id='9' siblings='4-5'/>
<cpu id='6' socket_id='1' core_id='10' siblings='6-7'/>
<cpu id='7' socket_id='1' core_id='10' siblings='6-7'/>
</cpus>
</cell>
<cell id='1'>
<memory unit='KiB'>6881280</memory>
<cpus num='8'>
<cpu id='8' socket_id='0' core_id='0' siblings='8-9'/>
<cpu id='9' socket_id='0' core_id='0' siblings='8-9'/>
<cpu id='10' socket_id='0' core_id='1' siblings='10-11'/>
<cpu id='11' socket_id='0' core_id='1' siblings='10-11'/>
<cpu id='12' socket_id='0' core_id='9' siblings='12-13'/>
<cpu id='13' socket_id='0' core_id='9' siblings='12-13'/>
<cpu id='14' socket_id='0' core_id='10' siblings='14-15'/>
<cpu id='15' socket_id='0' core_id='10' siblings='14-15'/>
</cpus>
</cell>
</cells>
</topology>
</host>
....
When the daemon is compiled with firewalld support but the DBus message
bus isn't started in the system, the initialization of the nwfilter
driver fails even if there are fallback options.
On hosts that don't have the DBus service running or installed the new
systemd cgroups code failed with hard error instead of falling back to
"manual" cgroup creation.
Use the new helper to check for the system bus and use the fallback code
in case it isn't available.
Some systems may not use DBus in their system. Add a method to check if
the system bus is available that doesn't print error messages so that
code can later check for this condition and use an alternative approach.
Coverity reported a memleak in the test added in 7efd5fd1b0. In case
the code will be broken and the code will actually parse a faulty bitmap
the resulting pointer would be leaked. Free it although that shouldn't
ever happen.
Each new VM requires a new connection from libvirtd to virtlockd.
The default max clients limit in virtlockd of 20 is thus woefully
insufficient. virtlockd sockets are only accessible to matching
users, so there is no security need for such a tight limit. Make
it configurable and default to 1024.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
In one of my previous patches I am removing the hard_limit heuristic to
guess the correct value if none set. However, it turned out, this limit
is hard to guess even for users. We should advise them to not set the
limit as their domains may be OOM killed. Sigh.
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.
One has to refresh the pool to get the correct pool info after
adding/removing/resizing a volume, this updates the pool metadata
(allocation, available) after those operation are done.
The function that parses custom driver XML was getting pretty unruly,
split the object parsing into their own functions. Rename some variables
to be consistent across each function. This should be functionally
identical.
Currently the virConnectBaselineCPU API does not expose the CPU features
that are part of the CPU's model. This patch adds a new flag,
VIR_CONNECT_BASELINE_CPU_EXPAND_FEATURES, that causes the API to explicitly
list all features that are part of that model.
Signed-off-by: Don Dugger <donald.d.dugger@intel.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Rather than inlining gl_WARN_ADD loads of time, we can shave about
17k size off of the configure script by delaying it to a cleanup
shell loop.
* m4/virt-compile-warnings.m4 (LIBVIRT_COMPILE_WARNINGS): Track a
list of things to check, rather than inlining multiple checks.
Signed-off-by: Eric Blake <eblake@redhat.com>
Cleanup the libxl capabilities code to be a bit more extensible,
splitting out the creation of host and guest capabilities. This
should make it easier to implement additional capabilities in the
future, such as NUMA topology reporting.
Feedback from the polkit developers indicates that the
"_detail_XXXX" attributes are a private implementation
detail. Our examples should be recommending use of the
"action.lookup('XXX')" method instead.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Previous patch fixed an issue where, when parsing a bitmap from the
string, the bounds of the bitmap weren't checked. That flaw resulted into
crashes. This test tests that case to avoid it in the future.
The virBitmapParse function was calling virBitmapIsSet() function that
requires the caller to check the bounds of the bitmap without checking
them. This resulted into crashes when parsing a bitmap string that was
exceeding the bounds used as argument.
This patch refactors the function to use virBitmapSetBit without
checking if the bit is set (this function does the checks internally)
and then counts the bits in the bitmap afterwards (instead of keeping
track while parsing the string).
This patch also changes the "parse_error" label to a more common
"error".
The refactor should also get rid of the need to call sa_assert on the
returned variable as the callpath should allow coverity to infer the
possible return values.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=997367
Thanks to Alex Jia for tracking down the issue. This issue is introduced
by commit 0fc8909.
https://bugzilla.redhat.com/show_bug.cgi?id=997765
==1349431== 8 bytes in 1 blocks are definitely lost in loss record 11 of 760
==1349431== at 0x4C2A554: calloc (vg_replace_malloc.c:593)
==1349431== by 0x4E9AA3E: virAllocN (in /usr/lib64/libvirt.so.0.1001.1)
==1349431== by 0x4EF28C4: virXPathNodeSet (in /usr/lib64/libvirt.so.0.1001.1)
==1349431== by 0x130B83: cmdCPUBaseline (in /usr/bin/virsh)
==1349431== by 0x12C608: vshCommandRun (in /usr/bin/virsh)
==1349431== by 0x12889A: main (in /usr/bin/virsh)
Gnulib recently patched gcc warning detection so that it does
not treat things like -Wno-unused-command-line-argument as
supported in gcc (treating it as supported merely resulted in
extra verbosity when an actual compile error occurred). It
has also improved bootstrap to work with less hassle on OpenBSD.
* .gnulib: Update to latest, in part for bootstrap and warnings
improvements.
* bootstrap: Resync from gnulib.
Signed-off-by: Eric Blake <eblake@redhat.com>
I noticed that in a fresh checkout, autogen.sh generated the
following output, but continued on with execution:
autoreconf: running: automake --add-missing --copy --force-missing
gnulib/tests/Makefile.am:28: TESTS was already defined in condition TRUE, which includes condition WITH_EXPENSIVE_TESTS ...
gnulib/tests/gnulib.mk:28: ... `TESTS' previously defined here
gnulib/tests/Makefile.am:19: `gnulib/tests/gnulib.mk' included from here
and after the run, line 28 of gnulib.mk lists GNULIB_TESTS, not TESTS.
After more investigation, I found that it is because gnulib bootstrap
provides two hooks, one before automake, and the other after; we used
the one that ran after, and were then rerunning automake ourselves;
and the warning was from the first run. But a manual second run is
pointless if we use the right hook in the first place.
The wrong function name has been latent since commit 38c9440, and we
tried to work around it in commit 6cbab7c, but it took commit 70363ea
to finally change output enough for me to realize the root cause.
* bootstrap.conf (bootstrap_epilogue): Rename...
(bootstrap_post_import_hook): ...so that it gets run before automake.
Signed-off-by: Eric Blake <eblake@redhat.com>
There is a potential leak of a newly created libxlDomainObjPrivate
when subsequent allocation of the object's chrdev field fails.
Unref the object on such an error so that it is properly disposed.
This resolves the issue that prompted the filing of
https://bugzilla.redhat.com/show_bug.cgi?id=928638
(although the request there is for something much larger and more
general than this patch).
commit f3868259ca disabled the
forwarding to upstream DNS servers of unresolved DNS requests for
names that had no domain, but were just simple host names (no "."
character anywhere in the name). While this behavior is frowned upon
by DNS root servers (that's why it was changed in libvirt), it is
convenient in some cases, and since dnsmasq can be configured to allow
it, it must not be strictly forbidden.
This patch restores the old behavior, but since it is usually
undesirable, restoring it requires specification of a new option in
the network config. Adding the attribute "forwardPlainNames='yes'" to
the <dns> elemnt does the trick - when that attribute is added to a
network config, any simple hostnames that can't be resolved by the
network's dnsmasq instance will be forwarded to the DNS servers listed
in the host's /etc/resolv.conf for an attempt at resolution (just as
any FQDN would be forwarded).
When that attribute *isn't* specified, unresolved simple names will
*not* be forwarded to the upstream DNS server - this is the default
behavior.
The description for <permissions> was copied from the storage volume
section to the storage pool section, but the semantics are different:
1. Currently only the "dir", "fs" and "netfs" storage pools use it.
2. They use it only to build the final directory.
3. A default for the storage volumes can't be set.
Signed-off-by: Philipp Hahn <hahn@univention.de>
If booting a container with a root FS that isn't the host's
root, we must ensure that the /dev mount point exists.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The virtlockd/libvirtd daemons had listed '?' as the short option
for --help. getopt_long uses '?' for any unknown option. We want
to be able to distinguish unknown options (which use EXIT_FAILURE)
from correct usage of help (which should use EXIT_SUCCESS). Thus
we should use 'h' as a short option for --help. Also add this to
the man page docs
The virtlockd/libvirtd daemons did not list any short option
for the --version arg. Add -V as a valid short option, since
-v is already used for --verbose.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Address a number of code, style and docs issues identified
in review of virt-login-shell after it was merged.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
