Format all functions with two blank lines between, and return type
on separate line from function name.  Also break some lines longer
than 80 columns.  This makes the subsequent macro refactoring
less noisy.

* src/util/vircgroup.c: Match prevailing style.
Signed-off-by: NEric Blake <eblake@redhat.com>

Make the virCgroupNewMachine method try to use systemd-machined
first. If that fails, then fallback to using the traditional
cgroup setup code path.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

When systemd is involved in managing processes, it may start
killing off & tearing down croups associated with the process
while we're still doing virCgroupKillPainfully. We must
explicitly check for ENOENT and treat it as if we had finished
killing processes
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

Systemd uses a named cgroup mount for tracking processes. Add
it as another type of controller, albeit one which we have to
special case in a number of places. In particular we must
never create/delete directories there, nor add tasks. Essentially
the systemd mount is to be considered read-only for libvirt.

With this change both the virCgroupDetectPlacement and
virCgroupCopyPlacement methods must be invoked. The copy
placement method will copy setup for resource controllers
only. The detect placement method will probe for any
named controllers, or resource controllers not already
setup.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

The previous patch was incomplete.

  CC       libvirt_util_la-vircgroup.lo
../../src/util/vircgroup.c:70:12: error: 'virCgroupPartitionEscape' declared 'static' but never defined [-Werror=unused-function]
 static int virCgroupPartitionEscape(char **path);
            ^

* src/util/vircgroup.c (virCgroupPartitionEscape): Move forward
declaration inside conditional.
Signed-off-by: NEric Blake <eblake@redhat.com>

The virCgroupValidateMachineGroup method calls some functions
which are only conditionally compiled, thus it too must be
made conditional. This fixes the build on non-Linux hosts.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

If the app has provided a whitelist of controllers to be used,
we skip detecting its mount point. We still, however, fill in
the placement info which later confuses the machine name
validation code. Skip detecting placement if the controller
mount point is not set
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

When detecting cgroups we must honour any controllers
whitelist the driver may have.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

When a VM has an 'emulator' child cgroup present, we must
strip off that suffix when detecting the cgroup for a
machine

Rename the virCgroupIsValidMachineGroup method to
virCgroupValidateMachineGroup to make a bit clearer
that this isn't simply a boolean check, it will make
changes to the object.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

The virCgroupIsValidMachine does not need to be called from
outside the cgroups file now, so make it static.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

Instead of requiring drivers to use a combination of calls
to virCgroupNewDetect and virCgroupIsValidMachine, combine
the two into virCgroupNewDetectMachine
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

Add protection such that the virCgroupRemove and
virCgroupKill* do not do anything to the root cgroup.

Killing all PIDs in the root cgroup does not end well.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

Instead of requiring one API call to create a cgroup and
another to add a task to it, introduce a new API
virCgroupNewMachine which does both jobs at once. This
will facilitate the later code to talk to systemd to
achieve this job which is also atomic.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

virCgroupAvailable() implementation calls getmntent_r
without checking if HAVE_GETMNTENT_R is defined, so it fails
to build on platforms without getmntent_r support.

Make virCgroupAvailable() just return false without
HAVE_GETMNTENT_R.

The virCgroupNewDomainDriver and virCgroupNewDriver methods
are obsolete now that we can auto-detect existing cgroup
placement. Delete them to reduce code bloat.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

Add virCgroupIsValidMachine API to check whether an auto
detected cgroup is valid for a machine. This lets us
check if a VM has just been placed into some generic
shared cgroup, or worse, the root cgroup
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

Add a virCgroupNewDetect API which is used to initialize a
cgroup object with the placement of an arbitrary process.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

Convert the remaining methods in vircgroup.c to report errors
instead of returning errno values.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

Instead of returning errno values, change the virCgroupKill*
APIs to fully report errors.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

Instead of returning raw errno values, report full libvirt
errors in virCgroupNew* functions.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

Reuse the buffer for getline and track buffer allocation
separately from the string length to prevent unlikely
out-of-bounds memory access.

This fixes the following leak that happened when zero bytes were read:

==404== 120 bytes in 1 blocks are definitely lost in loss record 1,344 of 1,671
==404==    at 0x4C2C71B: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==404==    by 0x906F862: getdelim (iogetdelim.c:68)
==404==    by 0x52A48FB: virCgroupPartitionNeedsEscaping (vircgroup.c:1136)
==404==    by 0x52A0FB4: virCgroupPartitionEscape (vircgroup.c:1171)
==404==    by 0x52A0EA4: virCgroupNewDomainPartition (vircgroup.c:1450)

Convert the type of loop iterators named 'i', 'j', k',
'ii', 'jj', 'kk', to be 'size_t' instead of 'int' or
'unsigned int', also santizing 'ii', 'jj', 'kk' to use
the normal 'i', 'j', 'k' naming
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

Similarly to VIR_STRDUP, we want the OOM error to be reported in
VIR_ALLOC and friends.

The @newpath variable is allocated in virCgroupSetPartitionSuffix(). But
it's newer freed.

We only break out of the while loop if *content is an empty string.
However the buffer has been allocated to BUFSIZ + 1 (8193 in my case),
but it gets overwritten in the next for iteration.

Move VIR_FREE right before we overwrite it to avoid the leak.

==5777== 16,386 bytes in 2 blocks are definitely lost in loss record 1,022 of 1,027
==5777==    by 0x5296E28: virReallocN (viralloc.c:184)
==5777==    by 0x52B0C66: virFileReadLimFD (virfile.c:1137)
==5777==    by 0x52B0E1A: virFileReadAll (virfile.c:1199)
==5777==    by 0x529B092: virCgroupGetValueStr (vircgroup.c:534)
==5777==    by 0x529AF64: virCgroupMoveTask (vircgroup.c:1079)

Introduced by 83e4c775.

https://bugzilla.redhat.com/show_bug.cgi?id=978352

Don't check for '\n' at the end of file if zero bytes were read.

Found by valgrind:
==404== Invalid read of size 1
==404==    at 0x529B09F: virCgroupGetValueStr (vircgroup.c:540)
==404==    by 0x529AF64: virCgroupMoveTask (vircgroup.c:1079)
==404==    by 0x1EB475: qemuSetupCgroupForEmulator (qemu_cgroup.c:1061)
==404==    by 0x1D9489: qemuProcessStart (qemu_process.c:3801)
==404==    by 0x18557E: qemuDomainObjStart (qemu_driver.c:5787)
==404==    by 0x190FA4: qemuDomainCreateWithFlags (qemu_driver.c:5839)

Introduced by 0d0b4098.

https://bugzilla.redhat.com/show_bug.cgi?id=978356

Found with 'git grep "= 1"'.

Currently, the controllers argument to virCgroupDetect acts both as
a result filter and a required controller specification, which is
a bit overloaded. If both functionalities are needed, it would be
better to have them seperated into a filter and a requirement mask.
The only situation where it is used today is to ensure that only
CPU related controllers are used for the VCPU directories. But here
we clearly do not want to enforce the existence of cpu, cpuacct and
specifically not cpuset at the same time.
This commit changes the semantics of controllers to "filter only".
Should a required mask ever be needed, more work will have to be done.
Signed-off-by: NViktor Mihajlovski <mihajlov@linux.vnet.ibm.com>

This commit is separate due to unusual paradigm compared to the
most source files.

Within whole vircgroup.c we 'return -errno', e.g. 'return -ENOMEM'.
However, in this specific function virCgroupAddTaskStrController
we weren't returning -ENOMEM but -1 despite fact that later in
the function we are returning one of errno values indeed.

https://bugzilla.redhat.com/show_bug.cgi?id=965169 documents a
problem starting domains when cgroups are enabled; I was able
to reliably reproduce the race about 5% of the time when I added
hooks to domain startup by 3 seconds (as that seemed to be about
the length of time that qemu created and then closed a temporary
thread, probably related to aio handling of initially opening
a disk image).  The problem has existed since we introduced
virCgroupMoveTask in commit 91028296 (v0.10.0).

There are some inherent TOCTTOU races when moving tasks between
kernel cgroups, precisely because threads can be created or
completed in the window between when we read a thread id from the
source and when we write to the destination.  As the goal of
virCgroupMoveTask is merely to move ALL tasks into the new
cgroup, it is sufficient to iterate until no more threads are
being created in the old group, and ignoring any threads that
die before we can move them.

It would be nicer to start the threads in the right cgroup to
begin with, but by default, all child threads are created in
the same cgroup as their parent, and we don't want vcpu child
threads in the emulator cgroup, so I don't see any good way
of avoiding the move.  It would also be nice if the kernel were
to implement something like rename() as a way to atomically move
a group of threads from one cgroup to another, instead of forcing
a window where we have to read and parse the source, then format
and write back into the destination.

* src/util/vircgroup.c (virCgroupAddTaskStrController): Ignore
ESRCH, because a thread ended between read and write attempts.
(virCgroupMoveTask): Loop until all threads have moved.
Signed-off-by: NEric Blake <eblake@redhat.com>

Change bbe97ae9 caused the
QEMU driver to ignore ENOENT errors from cgroups, in order
to cope with missing /proc/cgroups. This is not good though
because many other things can cause ENOENT and should not
be ignored. The callers expect to see ENXIO when cgroups
are not present, so adjust the code to report that errno
when /proc/cgroups is missing
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

Found that I was unable to start existing domains after updating
to a kernel with no cgroups support

  # zgrep CGROUP /proc/config.gz
  # CONFIG_CGROUPS is not set
  # virsh start test
  error: Failed to start domain test
  error: Unable to initialize /machine cgroup: Cannot allocate memory

virCgroupPartitionNeedsEscaping() correctly returns errno (ENOENT) when
attempting to open /proc/cgroups on such a system, but it was being
dropped in virCgroupSetPartitionSuffix().

Change virCgroupSetPartitionSuffix() to propagate errors returned by
its callees.  Also check for ENOENT in qemuInitCgroup() when determining
if cgroups support is available.

Escaping a leading '.' with '_' in the cgroup names
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

virAsprintf(&foo, "%s", bar) is wasteful compared to
foo = strdup(bar) (or eventually, VIR_STRDUP(foo, bar),
but one thing at a time...).

Noticed while reviewing Laine's attempt to clean up broken
qemu:///session.

* cfg.mk (sc_prohibit_asprintf): Enhance rule.
* src/esx/esx_storage_backend_vmfs.c
(esxStorageBackendVMFSVolumeLookupByKey): Fix offender.
* src/network/bridge_driver.c (networkStateInitialize): Likewise.
* src/nwfilter/nwfilter_dhcpsnoop.c (virNWFilterSnoopDHCPOpen):
Likewise.
* src/storage/storage_backend_sheepdog.c
(virStorageBackendSheepdogRefreshVol): Likewise.
* src/util/vircgroup.c (virCgroupAddTaskStrController): Likewise.
* src/util/virdnsmasq.c (addnhostsAdd): Likewise.
* src/xen/block_stats.c (xenLinuxDomainDeviceID): Likewise.
* src/xen/xen_driver.c (xenUnifiedConnectOpen): Likewise.
* tools/virsh.c (vshGetTypedParamValue): Likewise.
Signed-off-by: NEric Blake <eblake@redhat.com>

If a user cgroup name begins with "cgroup.", "_" or with any of
the controllers from /proc/cgroups followed by a dot, then they
need to be prefixed with a single underscore. eg if there is
an object "cpu.service", then this would end up as "_cpu.service"
in the cgroup filesystem tree, however, "waldo.service" would
stay "waldo.service", at least as long as nobody comes up with
a cgroup controller called "waldo".

Since we require a '.XXXX' suffix on all partitions, there is
no scope for clashing with the kernel 'tasks' and 'release_agent'
files.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

If the partition named passed in the XML does not already have
a suffix, ensure it gets a '.partition' added to each component.
The exceptions are /machine, /user and /system which do not need
to have a suffix, since they are fixed partitions at the top
level.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

Recently we changed to create VM cgroups with the naming pattern
$VMNAME.$DRIVER.libvirt. Following discussions with the systemd
community it was decided that only having a single '.' in the
names is preferrable. So this changes the naming scheme to be
$VMNAME.libvirt-$DRIVER. eg for LXC 'mycontainer.libvirt-lxc' or
for KVM 'myvm.libvirt-qemu'.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>