To every virDirOpen we must have VIR_DIR_CLOSE otherwise FD is
leaked.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
On systems with older glibc including fcntl.h for getting
FALLOC_FL_PUNCH_HOLE defined is not enough. We must also include
linux/falloc.h.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Some older systems (such as RHEL6) lack SEEK_HOLE and SEEK_DATA
which virFileInData relies on. Provide a stub for these systems.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Currently, we don't assign any meaning to that. Our current view
on virStream is that it's merely a pipe. And pipes don't support
seeking.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Similarly to previous commit, implement sparse streams feature
for vol-upload. This is, however, slightly different approach,
because we must implement a function that will tell us whether
we are in a data section or in a hole. But there's no magic
hidden in here.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Add a new --sparse switch that does nothing more than
enables the sparse streams feature for this command. Among with
the switch new helper function is introduced: virshStreamSkip().
This is the callback that is called whenever daemon sends us a
hole. In the callback we reflect the hole in underlying file by
seeking as many bytes as told.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
These flags to APIs will tell if caller wants to use sparse
stream for storage transfer. At the same time, it's safe to
enable them in storage driver frontend and rely on our backends
checking the flags. This way we can enable specific flags only on
some specific backends, e.g. enable
VIR_STORAGE_VOL_DOWNLOAD_SPARSE_STREAM for filesystem backend but
not iSCSI backend.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Now, not all APIs are going to support sparse streams. To some it
makes no sense at all, e.g. virDomainOpenConsole() or
virDomainOpenChannel(). To others, we will need a special flag to
indicate that client wants to enable sparse streams. Instead of
having to write RPC dispatchers by hand we can just annotate in
our .x files that a certain flag to certain RPC call enables this
feature. For instance:
/**
* @generate: both
* @readstream: 1
* @sparseflag: VIR_SPARSE_STREAM
* @acl: storage_vol:data_read
*/
REMOTE_PROC_DOMAIN_SOME_API = XXX,
Therefore, whenever client calls virDomainSomeAPI(..,
VIR_SPARSE_STREAM); daemon will mark that down and send stream
skips when possible.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Basically, what is needed here is to introduce new message type
for the messages passed between the event loop callbacks and the
worker thread that does all the I/O. The idea is that instead of
a queue of read buffers we will have a queue where "hole of size
X" messages appear. That way the event loop callbacks can just
check the head of the queue and see if the worker thread is in
data or a hole section and how long the section is.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Whenever client is able to receive some data from stream
daemonStreamHandleRead is called. But now the behaviour of this
function needs to be changed a bit. Previously it just read data
from underlying file (of chardev or whatever) and sent those
through the stream to client. This model will not work any longer
because it does not differentiate whether underlying file is in
data or hole section. Therefore, at the beginning of this
function add code that checks this situation and acts
accordingly.
So after the this, when wanting to send some data we always check
whether we are not in a hole and if so, skip it an inform client
about its size.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Whenever server sends a client stream packet (either regular with
actual data or stream skip one) it is queued on @st->rx. So the
list is a mixture of both types of stream packets. So now that we
have all the helpers needed we can wire their processing up. But
since virNetClientStreamRecvPacket doesn't support
VIR_STREAM_RECV_STOP_AT_HOLE flag yet, let's turn all received
skips into zeroes repeating requested times.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Now that we have RPC wrappers over VIR_NET_STREAM_HOLE we can
start wiring them up. This commit wires up situation when a
client wants to send a hole to daemon.
To keep stream offsets synchronous, upon successful call on the
daemon skip the same hole in local part of the stream.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This is a function that handles an incoming STREAM_HOLE packet.
Even though it is not wired up yet, it will be soon. At the
beginning do couple of checks whether server plays nicely and
sent us a STREAM_HOLE packed only after we've enabled sparse
streams. Then decodes the message payload to see how big the hole
is and stores it in passed @length argument.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Basically, whenever the new type of stream packet arrives to the
daemon call this function that decodes it and calls
virStreamSendHole(). Otherwise a regular data stream packet has
arrived and therefore continue its processing.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
While the previous commit implemented a helper for sending a
STREAM_HOLE packet for daemon, this is a client's counterpart.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This is just a helper function that takes in a length value,
encodes it into XDR and sends to client.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Ideally, this would be generated, but to achieve that
corresponding XDR definitions needed to go into a different .x
file. But they belong just to the one that they are right now.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This is a special type of stream packet, that is bidirectional
and contains information regarding how many bytes each side will
be skipping in the stream.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Add a new argument to daemonCreateClientStream in order to allow for
future expansion to mark that a specific stream can be used to skip
data, such as the case with sparsely populated files. The new flag will
be the eventual decision point between client/server to decide whether
both ends can support and want to use sparse streams.
A new bool 'allowSkip' is added to both _virNetClientStream and
daemonClientStream in order to perform the tracking.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Add a virStreamPtr pointer to the _virNetClientStream
in order to reverse track the parent stream.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This is just an internal API, that calls corresponding function
in stream driver. This function will set @data = 1 if the
underlying file is in data section, or @data = 0 if it is in a
hole. At any rate, @length is set to number of bytes remaining in
the section the file currently is.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This is just a wrapper over new function that have been just
introduced: virStreamSendHole() . It's very similar to
virStreamSendAll() except it handles sparse streams well.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This is just a wrapper over new functions that have been just
introduced: virStreamRecvFlags(), virStreamRecvHole(). It's very
similar to virStreamRecvAll() except it handles sparse streams
well.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Add a new flag to virStreamRecvFlags in order to handle being able to
stop reading from the stream so that the consumer can generate a "hole"
in stream target. Generation of a hole replaces the need to receive and
handle a sequence of zero bytes for sparse stream targets.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This function is basically a counterpart for virStreamSendHole().
If one side of a stream called virStreamSendHole() the other
should call virStreamRecvHole() to get the size of the hole.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This API is used to tell the other side of the stream to skip
some bytes in the stream. This can be used to create a sparse
file on the receiving side of a stream.
It takes @length argument, which says how big the hole is. This
skipping is done from the current point of stream. Since our
streams are not rewindable like regular files, we don't need
@whence argument like seek(2) has.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
There are three virStreamDriver's currently supported:
* virFDStream
* remote driver
* ESX driver
For now, backend virStreamRecvFlags support for only remote driver and
ESX driver is sufficient. Future patches will update virFDStream.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This patch is adding the virStreamRecvFlags as a variant to the
virStreamRecv function in order to allow for future expansion of
functionality for processing sparse streams using a @flags
argument.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This function takes a FD and determines whether the current
position is in data section or in a hole. In addition to that,
it also determines how much bytes are there remaining till the
current section ends.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
One big downside of using the pipe to transfer the data is that
we can really transfer just bare data. No metadata can be carried
through unless some formatted messages are introduced. That would
be quite painful to achieve so let's use a message queue. It's
fairly easy to exchange info between threads now that iohelper is
no longer used.
The reason why we cannot use the FD for plain files directly is
that despite us setting noblock flag on the FD, any
read()/write() blocks regardless (which is a show stopper since
those parts of the code are run from the event loop) and poll()
reports such FD as always readable/writable - even though the
subsequent operation might block.
The pipe is still not gone though. It is used to signal the event
loop that an event occurred (e.g. data is available for reading
in the queue, or vice versa).
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
When adding a nwfilter onto the list in
virNWFilterObjListAssignDef() this array is re-allocated to match
demand for new size. However, it is never freed leading to a
leak:
==26535== 136 bytes in 1 blocks are definitely lost in loss record 1,079 of 1,250
==26535== at 0x4C2E2BE: realloc (vg_replace_malloc.c:785)
==26535== by 0x54BA28E: virReallocN (viralloc.c:245)
==26535== by 0x54BA384: virExpandN (viralloc.c:294)
==26535== by 0x54BA657: virInsertElementsN (viralloc.c:436)
==26535== by 0x55DB011: virNWFilterObjListAssignDef (virnwfilterobj.c:362)
==26535== by 0x55DB530: virNWFilterObjListLoadConfig (virnwfilterobj.c:503)
==26535== by 0x55DB635: virNWFilterObjListLoadAllConfigs (virnwfilterobj.c:539)
==26535== by 0x2AC5A28B: nwfilterStateInitialize (nwfilter_driver.c:250)
==26535== by 0x5621C64: virStateInitialize (libvirt.c:770)
==26535== by 0x124379: daemonRunStateInit (libvirtd.c:881)
==26535== by 0x554AC78: virThreadHelper (virthread.c:206)
==26535== by 0x8F5F493: start_thread (in /lib64/libpthread-2.23.so)
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: John Ferlan <jferlan@redhat.com>
After bdcf6e481 there is a crasher in libvirt. The commit assumes
that priv->perf is always set. That is not true. For inactive
domains, the priv->perf is not allocated as it is set in
qemuProcessLaunch(). Now, usually we differentiate between
accesses to inactive and active definition and it works just
fine. Except for 'domstats'. There priv->perf is accessed without
prior check for domain inactivity. While we could check for that,
more robust solution is to make virPerfEventIsEnabled() accept
NULL.
How to reproduce:
1) ensure you have at least one inactive domain
2) virsh domstats
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Even though there are several checks before calling this function
and for some scenarios we don't call it at all (e.g. on disk hot
unplug), it may be possible to sneak in some weird files (e.g. if
domain would have RNG with /dev/shm/some_file as its backend). No
matter how improbable, we shouldn't unlink it as we would be
unlinking a file from the host which we haven't created in the
first place.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Cedric Bosdonnat <cbosdonnat@suse.com>
Just like in previous commit, this fixes the same issue for
hotplug.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Cedric Bosdonnat <cbosdonnat@suse.com>
While the code allows devices to already be there (by some
miracle), we shouldn't try to create devices that don't belong to
us. For instance, we shouldn't try to create /dev/shm/file
because /dev/shm is a mount point that is preserved. Therefore if
a file is created there from an outside (e.g. by mgmt application
or some other daemon running on the system like vhostmd), it
exists in the qemu namespace too as the mount point is the same.
It's only /dev and /dev only that is different. The same
reasoning applies to all other preserved mount points.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Cedric Bosdonnat <cbosdonnat@suse.com>
Currently, all we need to do in qemuDomainCreateDeviceRecursive() is to
take given @device, get all kinds of info on it (major & minor numbers,
owner, seclabels) and create its copy at a temporary location @path
(usually /var/run/libvirt/qemu/$domName.dev), if @device live under
/dev. This is, however, very loose condition, as it also means
/dev/shm/* is created too. Therefor, we will need to pass more arguments
into the function for better decision making (e.g. list of mount points
under /dev). Instead of adding more arguments to all the functions (not
easily reachable because some functions are callback with strictly
defined type), lets just turn this one 'const char *' into a 'struct *'.
New "arguments" can be then added at no cost.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Cedric Bosdonnat <cbosdonnat@suse.com>
When setting up mount namespace for a qemu domain the following
steps are executed:
1) get list of mountpoints under /dev/
2) move them to /var/run/libvirt/qemu/$domName.ext
3) start constructing new device tree under /var/run/libvirt/qemu/$domName.dev
4) move the mountpoint of the new device tree to /dev
5) restore original mountpoints from step 2)
Note the problem with this approach is that if some device in step
3) requires access to a mountpoint from step 2) it will fail as
the mountpoint is not there anymore. For instance consider the
following domain disk configuration:
<disk type='file' device='disk'>
<driver name='qemu' type='raw'/>
<source file='/dev/shm/vhostmd0'/>
<target dev='vdb' bus='virtio'/>
<address type='pci' domain='0x0000' bus='0x00' slot='0x0a' function='0x0'/>
</disk>
In this case operation fails as we are unable to create vhostmd0
in the new device tree because after step 2) there is no /dev/shm
anymore. Leave aside fact that we shouldn't try to create devices
living in other mountpoints. That's a separate bug that will be
addressed later.
Currently, the order described above is rearranged to:
1) get list of mountpoints under /dev/
2) start constructing new device tree under /var/run/libvirt/qemu/$domName.dev
3) move them to /var/run/libvirt/qemu/$domName.ext
4) move the mountpoint of the new device tree to /dev
5) restore original mountpoints from step 3)
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Cedric Bosdonnat <cbosdonnat@suse.com>
After 1eb6647979 nobody calls the iohelper with 6 arguments.
Everybody uses the other mode. Well, the only user of iohelper
after the previous commit is virFileWrapperFd really.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: John Ferlan <jferlan@redhat.com>
Currently we use iohelper for virFDStream implementation. This is
because UNIX I/O can lie sometimes: even though a FD for a
file/block device is set as unblocking, actual read()/write() can
block. To avoid this, a pipe is created and one end is kept for
read/write while the other is handed over to iohelper to
write/read the data for us. Thus it's iohelper which gets blocked
and not our event loop.
This approach has two problems:
1) we are spawning a new process.
2) any exchange of information between daemon and iohelper can be
done only through the pipe.
Therefore, iohelper is replaced with an implementation in thread
which is created just for the stream lifetime. The data are still
transferred through pipe (for now), but both problems described
above are solved.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: John Ferlan <jferlan@redhat.com>
While this is no functional change, it makes the code look a bit
nicer. Moreover, it prepares ground for future work.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: John Ferlan <jferlan@redhat.com>
There is really no reason why we should have to have 'struct'
everywhere.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: John Ferlan <jferlan@redhat.com>
It helps with debugging if we know what's the return value of
saferead().
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: John Ferlan <jferlan@redhat.com>
Because of copy-paste the temporary directory used for this test
is called "fakesysdir". That's probably misleading.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: John Ferlan <jferlan@redhat.com>
This function runs an iscsi command and parses its output.
However, due to the nature of things, virISCSIExtractSession()
callback can be called multiple times. In each run it would
allocate new memory and overwrite the variable where we keep
pointer to it and thus leaking old allocations.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Even though the virMacMap object is not necessarily created at
the same time as the network object, the former makes no sense
without the latter and thus should be unref'd in the network
object dispose function.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Imagine that this function is called twice over the same disk
source. While in the first run all allocated memory is freed, not
all pointers are set to NULL (e.g. def->srcpool). So when called
again, these poitners are freed again resulting in double free.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Currently, if we want to zero out disk source (e,g, due to
startupPolicy when starting up a domain) we use
virDomainDiskSetSource(disk, NULL). This works well for file
based storage (storage type file, dir, or block). But it doesn't
work at all for other types like volume and network.
So imagine that you have a domain that has a CDROM configured
which source is a volume from an inactive pool. Because it is
startupPolicy='optional', the CDROM is empty when the domain
starts. However, the source element is not cleared out in the
status XML and thus when the daemon restarts and tries to
reconnect to the domain it refreshes the disks (which fails - the
storage pool is still not running) and thus the domain is killed.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
The virMacMap module is there for dumping [domain, <list of is
MACs>] pairs into a file so that libvirt_guest NSS module can use
it. Whenever a interface is allocated from network (e.g. on
domain<F2> startup or NIC hotplug), network is notified and so is
virMacMap module subsequently. The module update functions
networkMacMgrAdd() and networkMacMgrDel() gracefully handle the
case when there's no module. The problem is, the module is
created if and only if network is freshly started, or if the
daemon restarts and network previously had the module.
This is not very user friendly - if users want to use the NSS
module they need to destroy their network and bring it up again
(and subsequently all the domains using it).
One disadvantage of this approach implemented here is that one
may get just partial results: any already running network does
not record mac maps, thus only newly plugged domains will be
stored in the module. The network restart scenario is not touched
by this of course. But one can argue that older libvirts had
never recorded the mac maps anyway.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
So far our code is full of the following pattern:
dom = virGetDomain(conn, name, uuid)
if (dom)
dom->id = 42;
There is no reasong why it couldn't be just:
dom = virGetDomain(conn, name, uuid, id);
After all, client domain representation consists of tuple (name,
uuid, id).
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
In 9e2465834 a check that denies internal snapshots when pflash
based loader is configured for the domain. However, if there's
none and an user tries to do an internal snapshot they will
witness daemon crash as in that case vm->def->os.loader is NULL
and we dereference it unconditionally.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
One of the problems with our virGetDomain function is that it
copies just domain name and domain UUID. Therefore it's very
easy to forget aboud domain ID. This can cause some bugs, like
virConnectGetAllDomainStats not reporting proper domain IDs.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1434882
Imagine the following scenario:
1) virsh net-start default
2) virsh start myFavouriteDomain
3) virsh net-destroy default
4) virsh destroy myFavouriteDomain
(assuming myFavouriteDomain has an interface from default
network)
Regardless of how unlikely this scenario looks like, we should
not crash. The problem is, on net-destroy in
networkShutdownNetworkVirtual() the virMacMap module is unrefed,
but the stale pointer is kept around. Thus when the domain
destroy procedure comes in, networkReleaseActualDevice() and
subsequently networkMacMgrDel() is called. This function sees the
stale pointer and starts calling the virMacMap module APIs which
work over freed memory.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
For some drivers the domain's machine type makes no sense. They
just don't use it. A great example is bhyve driver. Therefore it
makes very less sense to report machine in domain capabilities
XML.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
There were couple of reports on the list (e.g. [1]) that guests
with huge amounts of RAM are unable to start because libvirt
kills qemu in the initialization phase. The problem is that if
guest is configured to use hugepages kernel has to zero them all
out before handing over to qemu process. For instance, 402GiB
worth of 1GiB pages took around 105 seconds (~3.8GiB/s). Since we
do not want to make the timeout for connecting to monitor
configurable, we have to teach libvirt to count with this
fact. This commit implements "1s per each 1GiB of RAM" approach
as suggested here [2].
1: https://www.redhat.com/archives/libvir-list/2017-March/msg00373.html
2: https://www.redhat.com/archives/libvir-list/2017-March/msg00405.html
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
While connecting to qemu monitor, the first thing we do is wait
for it to show up. However, we are doing it with some timeout to
avoid indefinite waits (e.g. when qemu doesn't create the monitor
socket at all). After beaa447a29 we are using exponential back
off timeout meaning, after the first connection attempt we wait
1ms, then 2ms, then 4 and so on. This allows us to bring down
wait time for small domains where qemu initializes quickly.
However, on the other end of this scale are some domains with
huge amounts of guest memory. Now imagine that we've gotten up to
wait time of 15 seconds. The next one is going to be 30 seconds,
and the one after that whole minute. Well, okay - with current
code we are not going to wait longer than 30 seconds in total,
but this is going to change in the next commit.
The exponential back off is usable only for first few iterations.
Then it needs to be caped (one second was chosen as the limit)
and switch to constant wait time.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Some users might want to pass a blockdev or a chardev as a
backend for NVDIMM. In fact, this is expected to be the mostly
used configuration. Therefore libvirt should allow the device in
devices CGroup then.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Now that we have APIs for relabel memdevs on hotplug, fill in the
missing implementation in qemu hotplug code.
The qemuSecurity wrappers might look like overkill for now,
because qemu namespace code does not deal with the nvdimms yet.
Nor does our cgroup code. But hey, there's cgroup_device_acl
variable in qemu.conf. If users add their /dev/pmem* device in
there, the device is allowed in cgroups and created in the
namespace so they can successfully passthrough it to the domain.
It doesn't look like overkill after all, does it?
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
When domain is being started up, we ought to relabel the host
side of NVDIMM so qemu has access to it.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
When domain is being started up, we ought to relabel the host
side of NVDIMM so qemu has access to it.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
For NVDIMM devices it is optionally possible to specify the size
of internal storage for namespaces. Namespaces are a feature that
allows users to partition the NVDIMM for different uses.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Now that NVDIMM has found its way into libvirt, users might want
to fine tune some settings for each module separately. One such
setting is 'share=on|off' for the memory-backend-file object.
This setting - just like its name suggest already - enables
sharing the nvdimm module with other applications. Under the hood
it controls whether qemu mmaps() the file as MAP_PRIVATE or
MAP_SHARED.
Yet again, we have such config knob in domain XML, but it's just
an attribute to numa <cell/>. This does not give fine enough
tuning on per-memdevice basis so we need to have the attribute
for each device too.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
So, majority of the code is just ready as-is. Well, with one
slight change: differentiate between dimm and nvdimm in places
like device alias generation, generating the command line and so
on.
Speaking of the command line, we also need to append 'nvdimm=on'
to the '-machine' argument so that the nvdimm feature is
advertised in the ACPI tables properly.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
NVDIMM is new type of memory introduced into QEMU 2.6. The idea
is that we have a Non-Volatile memory module that keeps the data
persistent across domain reboots.
At the domain XML level, we already have some representation of
'dimm' modules. Long story short, NVDIMM will utilize the
existing <memory/> element that lives under <devices/> by adding
a new attribute 'nvdimm' to the existing @model and introduce a
new <path/> element for <source/> while reusing other fields. The
resulting XML would appear as:
<memory model='nvdimm'>
<source>
<path>/tmp/nvdimm</path>
</source>
<target>
<size unit='KiB'>523264</size>
<node>0</node>
</target>
<address type='dimm' slot='0'/>
</memory>
So far, this is just a XML parser/formatter extension. QEMU
driver implementation is in the next commit.
For more info on NVDIMM visit the following web page:
http://pmem.io/
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Frankly, this function is one big mess. A lot of arguments,
complicated behaviour. It's really surprising that arguments were
in random order (input and output arguments were mixed together),
the documentation was outdated, the description of return values
was bogus.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Even though this variable contains just values from an enum where
zero has the usual meaning, it's enum after all and we should
check it as such.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1431112
Yeah, that's right. A mount point doesn't have to be a directory.
It can be a file too. However, the code that tries to preserve
mount points under /dev for new namespace for qemu does not count
with that option.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1430634
If a qemu process has died, we get EOF on its monitor. At this
point, since qemu process was the only one running in the
namespace kernel has already cleaned the namespace up. Any
attempt of ours to enter it has to fail.
This really happened in the bug linked above. We've tried to
attach a disk to qemu and while we were in the monitor talking to
qemu it just died. Therefore our code tried to do some roll back
(e.g. deny the device in cgroups again, restore labels, etc.).
However, during the roll back (esp. when restoring labels) we
still thought that domain has a namespace. So we used secdriver's
transactions. This failed as there is no namespace to enter.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This is a very historic artefact. Back in the old days of
830ba76c3e when we had macros to add arguments onto qemu command
line (!) we thought it was a good idea to let qemu write out the
PID file. So we passed -pidfile $stateDir/$domName onto the
command line. Thus, in order for tests to work we needed stable
stateDir in the qemu driver. Unfortunately, after 16efa11aa6
where stateDir is mkdtemp()-d, this approach lead to a leak of
temp dir.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Some of our tests (e.g. qemuhotplugtest) call
virDomainSaveConfig(). Now the problem is, qemuTestDriverInit()
creates a fake qemu driver and fills it with some fake
configuration. At least so we hoped. The truth is, it calls
regular virQEMUDriverConfigNew() and then fix couple of paths.
Literally. Therefore our tests see regular stateDir and configDir
for the user that is running the tests. Directories, where live
domain XMLs are stored. Let's just hope our test suite hasn't
mangled any of them.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1379200
When we are restoring a domain from a saved image, or just
updating its XML in the saved image - we have to make sure that
the ABI guests sees will not change. We have a function for that
which reports errors. But for some reason if this function fails,
we call it again with slightly different argument. Therefore it
might happen that we overwrite the original error and leave user
with less helpful one.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Now that we have some qemuSecurity wrappers over
virSecurityManager APIs, lets make sure everybody sticks with
them. We have them for a reason and calling virSecurityManager
API directly instead of wrapper may lead into accidentally
labelling a file on the host instead of namespace.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This function is calling public APIs (virNodeDeviceLookupByName
etc.). That requires the driver lock to be unlocked and locked
again. If we, however, replace the public APIs calls with the
internal calls (that public APIs call anyway), we can drop the
lock/unlock exercise.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
After eca76884ea in case of error in qemuDomainSetPrivatePaths()
in pretended start we jump to stop. I've changed this during
review from 'cleanup' which turned out to be correct. Well, sort
of. We can't call qemuProcessStop() as it decrements
driver->nactive and we did not increment it. However, it calls
virDomainObjRemoveTransientDef() which is basically the only
function we need to call. So call that function and goto cleanup;
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
When a domain needs an access to some device (be it a disk, RNG,
chardev, whatever), we have to allow it in the devices CGroup (if
it is available), because by default we disallow all the devices.
But some of the functions that are responsible for setting up
devices CGroup are lacking check whether there is any CGroup
available. Thus users might be unable to hotplug some devices:
virsh # attach-device fedora rng.xml
error: Failed to attach device from rng.xml
error: internal error: Controller 'devices' is not mounted
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Our virSomeEnumTypeFromString() functions return either the value
of item from the enum or -1 on error. Usually however the value 0
means 'this value is not set in the domain XML, use some sensible
default'. Therefore, we don't accept corresponding string in
domain XML, for instance:
<memoryBacking>
<source mode="none"/>
<access mode="default"/>
<allocation mode="none"/>
</memoryBacking>
should be rejected as invalid XML.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
When enabling virgl, qemu opens /dev/dri/render*. So far, we are
not allowing that in devices CGroup nor creating the file in
domain's namespace and thus requiring users to set the paths in
qemu.conf. This, however, is suboptimal as it allows access to
ALL qemu processes even those which don't have virgl configured.
Now that we have a way to specify render node that qemu will use
we can be more cautious and enable just that.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
So far, qemuDomainGetHostdevPath has no knowledge of the reasong
it is called and thus reports /dev/vfio/vfio for every VFIO
backed device. This is suboptimal, as we want it to:
a) report /dev/vfio/vfio on every addition or domain startup
b) report /dev/vfio/vfio only on last VFIO device being unplugged
If a domain is being stopped then namespace and CGroup die with
it so no need to worry about that. I mean, even when a domain
that's exiting has more than one VFIO devices assigned to it,
this function does not clean /dev/vfio/vfio in CGroup nor in the
namespace. But that doesn't matter.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
So far, we are allowing /dev/vfio/vfio in the devices cgroup
unconditionally (and creating it in the namespace too). Even if
domain has no hostdev assignment configured. This is potential
security hole. Therefore, when starting the domain (or
hotplugging a hostdev) create & allow /dev/vfio/vfio too (if
needed).
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Since these two functions are nearly identical (with
qemuSetupHostdevCgroup actually calling virCgroupAllowDevicePath)
we can have one function call the other and thus de-duplicate
some code.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
There's no need for this function. Currently it is passed as a
callback to virSCSIVHostDeviceFileIterate(). However, SCSI host
devices have just one file path. Therefore we can mimic approach
used in qemuDomainGetHostdevPath() to get path and call
virCgroupAllowDevicePath() directly.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Michal Privoznik