This is the bug I'm facing. I deliberately configured a container
so that the source of a <filesystem/> to passthrough doesn't
exist. The start fails with:
lxcContainerPivotRoot:669 : Failed to create /non-existent/path/.oldroot: Permission denied
which is expected. But what is NOT expected is that CGroup
hierarchy is left behind. This is because the controller sets up
the CGroup hierarchy, user namespace, moves interfaces, etc. and
finally checks whether container setup (done in a separate
process) succeeded. Only after all this the error is propagated
to the LXC driver. The driver aborts the startup and tries to
perform the cleanup, but this is missing CGroups because those
weren't detected yet.
Ideally, whenever a function fails, it tries to unroll back so
that is has no artifacts left behind (look at all those frees/FD
closes/etc. at end of functions). But with CGroups it is
different - the controller process can't clean up after itself,
because it is still running inside that CGroup.
Therefore, what we have to do is to let the driver detect CGroups
as soon as they are created, and proceed with controller
execution only after that.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Currently, there is only a single pipe passed to lxc_controller
and it is used by lxc_controller to signal to the LXC driver that
the container is set up and ready to run. However, in the next
commit we will need to signal that the LXC driver has done its
part of startup process and thus the controller can proceed.
Unfortunately, virCommand handshake can't be used for this,
because it's already used to read controller's PID.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Future commits will want to reuse the handshakeFd and thus it
mustn't be closed in virLXCControllerDaemonHandshake(). Do the
closing explicitly afterwards.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
The lxc_controller has a structure that's keeping its internal
state, including so called handshakeFd which is the write end of
a pipe that's used to signal to the LXC driver that the container
is set up and ready to run. However, the struct member is not
initialized to -1, so if anything fails before it is set then the
virLXCControllerFree() function tries to close FD 0 (stdin).
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
The g_path_is_absolute() considers more situations
than just a simply "path[0] == '/'".
Related issue: https://gitlab.com/libvirt/libvirt/-/issues/12
Signed-off-by: Luke Yue <lukedyue@gmail.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The command line argument is called --hanshakefd (check out
lxc_controller.c:main()). But the command line builder puts only
--handshake. This works, because there is no other argument
sharing the prefix.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Historically, we declared pointer type to our types:
typedef struct _virXXX virXXX;
typedef virXXX *virXXXPtr;
But usefulness of such declaration is questionable, at best.
Unfortunately, we can't drop every such declaration - we have to
carry some over, because they are part of public API (e.g.
virDomainPtr). But for internal types - we can do drop them and
use what every other C project uses 'virXXX *'.
This change was generated by a very ugly shell script that
generated sed script which was then called over each file in the
repository. For the shell script refer to the cover letter:
https://listman.redhat.com/archives/libvir-list/2021-March/msg00537.html
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Recently, a few commits back I've switched bunch of code to
g_steal_pointer() using coccinelle. Problem was that the semantic
patch used was slightly off:
@@
expression a, b;
@@
+ b = g_steal_pointer(&a);
- b = a;
... when != a
- a = NULL;
Problem is that, "... when != a" is supposed to jump over those
lines, which don't contain expression a. My idea was to replace
the following pattern too:
ptrX = ptrY;
if (something(ptrZ) < 0) goto error;
ptrY = NULL;
But what I missed is that the following pattern is also matched
and replaced:
ptrX = ptrY;
if (something(ptrX) < 0) goto error;
ptrY = NULL;
This is not necessarily correct - as demonstrated by our hotplug
code. The real problem is ambiguous memory ownership transfer
(functions which add device to domain def take ownership only on
success), but to not tackle the real issue let's revert those
parts.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
The order in which virNetworkUpdate() accepts @section and
@command arguments is not the same as in which it passes them
onto networkUpdate() callback. Until recently, it did not really
matter, because calling the API on client side meant arguments
were encoded in reversed order (compared to the public API), but
then on the server it was fixed again - because the server
decoded RPC (still swapped), called public API (still swapped)
and in turn called the network driver callback (with reversing
the order - so magically fixing the order).
Long story short, if the public API is called even number of
times those swaps cancel each other out. The problem is when the
API is called an odd numbed of times - which happens with split
daemons and the right URI. There's one call in the client (e.g.
virsh net-update), the other in a hypervisor daemon (say
virtqemud) which ends up calling the API in the virnetworkd.
The fix is obvious - fix the order in which arguments are passed
to the callback.
But, to maintain compatibility with older, yet unfixed, daemons
new connection feature is introduced. The feature is detected
just before calling the callback and allows client to pass
arguments in correct order (talking to fixed daemon) or in
reversed order (talking to older daemon).
Unfortunately, older client talking to newer daemon can't be
fixed. Let's hope that it's less frequent scenario.
Fixes: 574b9bc66b
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1870552
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Generated by the following spatch:
@@
expression a, b;
@@
+ b = g_steal_pointer(&a);
- b = a;
... when != a
- a = NULL;
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
In short, virXXXPtr type is going away. With big bang. And to
help us rewrite the code with a sed script, it's better if each
variable is declared on its own line.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Generated using the following spatch:
@@
expression path;
@@
- virFileMakePath(path)
+ g_mkdir_with_parents(path, 0777)
However, 14 occurrences were not replaced, e.g. in
virHostdevManagerNew(). I don't really understand why.
Fixed by hand afterwards.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Now that we enforce the cpu.shares range kernel will no longer silently
change the value that libvirt configures so there is no need to read
the value back to get the actual configuration.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
'virLogGetFilters' doesn't return failure and 'virLogGetOutputs' reports
it's own errors.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>
In files: src/lxc/lxc_native: in lxcAddNetworkRouteDefinition(),
src/conf/networkcommon_conf: in virNetDevIPRouteCreate() and
virNetDevIPRouteParseXML()
Signed-off-by: Kristina Hanicova <khanicov@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>
In files: src/conf/domain_conf: in virDomainNetIPInfoParseXML(),
src/lxc/lxc_native: in lxcAddNetworkRouteDefinition(),
src/vz/vz_sdk: in prlsdkGetRoutes(), src/conf/networkcommon_conf:
in virNetDevIPRouteCreate()
Signed-off-by: Kristina Hanicova <khanicov@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>
Our implementation was heavily inspired by the glib version so it's a
drop-in replacement.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Some callers don't need to know the actual lenght of the list but only
care whether the required element is present or the list is non-empty.
Don't calculate the list length in those cases.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The glib variant doesn't accept NULL list, but there's just one caller
where it wasn't checked explicitly, thus there's no need for our own
wrapper.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The 'conflict' key in a virt_daemon_unit dictionary is not used when
generating systemd service and socket files. The comment associated
with the key claims the default is 'true', and a few build files
needlessly set it to 'true' when defining their virt_daemon_unit.
Remove the 'conflict' key and its use in the affect build files.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Now that this function can be called regardless of interface type (and
whether or not we have a conn for the network driver), let's actually
call it for all interface types. This will assure that we re-connect
any disconnected bridge devices for <interface type='bridge'> as
mentioned in https://bugzilla.redhat.com/show_bug.cgi?id=1730084#c26
(until now we've only been reconnecting bridge devices for <interface
type='network'>)
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The two scenarios were found by Coverity after a seemingly-unrelated
change to virLXCProcessSetupInterfaceTap() (in commit ecfc2d5f43), and
explained by John Ferlan here:
https://www.redhat.com/archives/libvir-list/2020-December/msg00810.html
To re-explain:
a) On entry to virLXCProcessSetupInterfaceTap() if net->ifname != NULL
then a copy of net->ifname is made into parentVeth, and a reference
to *that* pointer is sent down to virNetDevVethCreate().
b) If parentVeth (aka net->ifname) is a template name (e.g. "blah%d"),
then virNetDevVethCreate() calls virNetDevGenerateName(), and if
virNetDevGenerateName() successfully generates a usable name
(e.g. "blah27") then it will free the original template string
(which is pointed to by net->ifname and by parentVeth), then
replace the pointer in parentVeth with a pointer to the new
string. Note that net->ifname still points to the now-freed
template string.
c) returning back up to virLXCProcessSetupInterfaceTap(), we check if
net->ifname == NULL - it *isn't* (still contains stale pointer to
template string), so we don't replace it with the pointer to the new
string that is in parentVeth.
d) Result: the new string is leaked once we return from
virLXCProcessSetupInterfaceTap(), while there is a dangling pointer
to the old string in net->ifname.
There is also a leak if there is a failure somewhere between steps (b)
and (c) above - the failure cleanup in virNetDevVethCreate() will only
free the newly-generated parentVeth string if the original pointer was
NULL (narrator: "It wasn't."). But it's a new string allocated by
virNetDevGenerateName(), not the original string from net->ifname, so
it really does need to be freed.
The solution is to make a copy of the entire original string into a
g_autofree pointer, then iff everything is successful we g_free() the
original net->ifname and replace it by stealing the string returned by
virNetDevVethCreate().
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
In all cases *except* when parsing status XML as libvirt is being
restarted, the XML parser will delete any manually specified interface
name (aka "<target dev='blah'/>" aka net->ifname) that could have been
generated by virNetDevGenerateName(). This means that during the setup
when a domain is being started (e.g. during
virLXCProcessSetupInterfaceTap()) it is pointless to call
virNetDevReserveName() with any setting of net->ifname that has come
from the XML parser - it is guaranteed to not fit the pattern of any
auto-generated name, and so the call is just a NOP anyway.
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The only reason why virstoragefile.h needs to be included in virfile.h
is that virFileNBDDeviceAssociate() takes virStorageFileFormat argument.
The function doesn't need the enum value as it converts the value to
string and uses only that.
Change the argument to string which will allow us to remove that
include.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Our parser code relies on the fact that
VIR_LXC_DOMAIN_NAMESPACE_SOURCE_NONE has value of zero and thus
uses g_new0(). But strictly speaking, this is not mandated by
the enum typedef. Fix that.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>
While fixing our schema for <lxc:namespace/> I've looked into the
parser and realized it could use some treating.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>
As all other free functions, NULL should be accepted. Even though
there currently is no caller that would pass NULL, there will be
in future patches.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>
the lxc driver uses virNetDevGenerateName() for its veth device names
since patch 2dd0fb492, so it should be using virNetDevReserveName()
during daemon restart/reconnect to skip over the device names that are
in use.
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
In virLXCProcessSetupInterfaceTap, containerVeth needs to be freed on
failure.
Signed-off-by: Shi Lei <shi_lei@massclouds.com>
Reviewed-by: Laine Stump <laine@redhat.com>
Simplify virNetDevVethCreate by using common GenerateName/ReserveName
functions.
Signed-off-by: Shi Lei <shi_lei@massclouds.com>
Reviewed-by: Laine Stump <laine@redhat.com>
Commit 729a06c41 added code to the LXC driver (patterned after similar
code in the QEMU driver) that called
virNetDevMacVlanReserveName(net->ifname) for all type='direct'
interfaces during a libvirtd restart, to prevent other domains from
attempting to use a macvtap device name that was already in use by a
domain.
But, unlike a QEMU domain, when an LXC domain creates a macvtap
device, that device is almost immediately moved into the namespace of
the container (and it's then renamed, but that part isn't
important). Because of this, the LXC driver doesn't keep track (in
net->ifname) of the name used to create the device (as the QEMU driver
does).
The result of this is that if libvirtd is restarted while there is an
active LXC domain that has <interface type='direct'>, libvirtd will
segfault (since virNetDevMacVLanReserveName() doesn't check for a NULL
pointer).
The fix is to just not call that function in the case of the LXC
driver, since it is pointless anyway.
Fixes: 729a06c41a
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
The next objective is to move virDomainDeviceDefValidate() to
domain_validate.c. First let's move all the static helpers.
The net device validation functions are used across multiple
drivers, so let's move them separately first.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Attempting to create a lxc domain with <seclabel type='none'/> fails
virsh --connect lxc:/// create distro_nosec.xml
error: Failed to create domain from distro_nosec.xml
error: unsupported configuration: Security driver model '(null)' is not available
Commit 638ffa2228 adjusted the logic for setting a driver's default
security model.
The lxc driver does not set a default security driver model in the XML
parser config, causing seclabels of type='none' to have a null model.
The lxc driver's security manager is initialized in lxcStateInitialize()
by calling lxcSecurityInit(). Use the model of this manager as the
default in the XML parser config.
For the record, this is a regression caused by commit 638ffa2228, which
changed the logic for setting a driver's default security model. The
qemu driver was adjusted accordingly, but a similar change was missed
in the lxc driver.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
If starting an container fails, the virLXCProcessStop() is
called. But since vm->def->id is not set until libvirt_lxc is
spawned (the domain's ID is PID of that process),
virLXCProcessStop() returns early as virDomainObjIsActive()
returns false. But doing so leaves behind resources reserved for
the containers during the startup process. Most notably, hostdevs
are not re-attached to the host, the domain's transient XML is
not removed, etc.
To resolve this, virLXCProcessCleanup() is called in this case.
However, it is modified to accept @flags which allows caller to
run only specific cleanups (depending how far in container
creation the failure occurred). There is plenty of cleanups which
don't need this guard because either they detect a NULL pointer
or try to release an unique resource.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Glib provides g_auto(GStrv) which is in-place replacement of our
VIR_AUTOSTRINGLIST.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
This function always returns zero, so it might as well be void.
Signed-off-by: Matt Coleman <matt@datto.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The function only returns zero or aborts, so it might as well be void.
This has the added benefit of simplifying the code that calls it.
Signed-off-by: Matt Coleman <matt@datto.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Ján Tomko <jtomko@redhat.com>
This patch adds new schema and adds support for parsing and formatting
domain configurations that include vdpa devices.
vDPA network devices allow high-performance networking in a virtual
machine by providing a wire-speed data path. These devices require a
vendor-specific host driver but the data path follows the virtio
specification.
When a device on the host is bound to an appropriate vendor-specific
driver, it will create a chardev on the host at e.g. /dev/vhost-vdpa-0.
That chardev path can then be used to define a new interface with
type='vdpa'.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>
As preparation for g_autoptr() we need to change the function to take
only virCgroupPtr.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Jonathon Jongsma <jjongsma@redhat.com>
CVE-2020-25637
Add a requirement for domain:write if source is set to
VIR_DOMAIN_INTERFACE_ADDRESSES_SRC_AGENT.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reported-by: Ilja Van Sprundel <ivansprundel@ioactive.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Use https: links for websites that support them.
The URIs which are used as namespace identifiers
are left alone.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
Reviewed-by: Neal Gompa <ngompa13@gmail.com>
There have been some reports that, due to libvirt always trying to
assign the lowest numbered macvtap / tap device name possible, a new
guest would sometimes be started using the same tap device name as
previously used by another guest that is in the process of being
destroyed *as the new guest is starting.
In some cases this has led to, for example, the old guest's
qemuProcessStop() code deleting a port from an OVS switch that had
just been re-added by the new guest (because the port name is based on
only the device name using the port). Similar problems can happen (and
I believe have) with nwfilter rules and bandwidth rules (which are
both instantiated based on the name of the tap device).
A couple patches have been previously proposed to change the ordering
of startup and shutdown processing, or to put a mutex around
everything related to the tap/macvtap device name usage, but in the
end no matter what you do there will still be possible holes, because
the device could be deleted outside libvirt's control (for example,
regular tap devices are automatically deleted when the qemu process
terminates, and that isn't always initiated by libvirt but could
instead happen completely asynchronously - libvirt then has no control
over the ordering of shutdown operations, and no opportunity to
protect it with a mutex.)
But this only happens if a new device is created at the same time as
one is being deleted. We can effectively eliminate the chance of this
happening if we end the practice of always looking for the lowest
numbered available device name, and instead just keep an integer that
is incremented each time we need a new device name. At some point it
will need to wrap back around to 0 (in order to avoid the IFNAMSIZ 15
character limit if nothing else), and we can't guarantee that the new
name really will be the *least* recently used name, but "math"
suggests that it will be *much* less common that we'll try to re-use
the *most* recently used name.
This patch implements such a counter for macvtap/macvlan, replacing
the existing, and much more complicated, "ID reservation" system. The
counter is set according to whatever macvtap/macvlan devices are
already in use by guests when libvirtd is started, incremented each
time a new device name is needed, and wraps back to 0 when either
INT_MAX is reached, or when the resulting device name would be longer
than IFNAMSIZ-1 characters (which actually is what happens when the
template for the device name is "maccvtap%d"). The result is that no
macvtap name will be re-used until the host has created (and possibly
destroyed) 99,999,999 devices.
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Since the macro no longer includes the 'ignore_value'
statement, stop putting another empty statement after it.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Many of our functions start with a DEBUG statement.
Move the statements after declarations to appease
our coding style.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
