With the previous refactoring, this is a simple process, since the global
'CreateBlockFrom' in storage_backend does all the work.

We don't gain any space savings, so skip the detection to speed up
the cloning operation.

Add a 'CreateBlockFrom' in the global storage_backend, which sets up the
destination block device: CopyFromFD does the rest of the cloning.

The CreateRaw function has some 'file' only assumptions, so break the agnostic
cloning bits to a separate function.

Have storage building functions be definitions of
virStorageBackendBuildVolFrom: we will need to do this in the future anyways
if we ever support the flags attribute.

These will be used by other pool cloning implementations.

Break out separate functions for

- Determining the supported '*-img' tool,
- The tool's associated create function,
- Desired function for cloning (CreateXMLFrom).

This will be eventually used to unify cloning across all backends.

We need to unlock the first pool before looking up the second, since the
search locks every pool it checks.

There isn't any way to dictate allocation when creating disk volumes, so
capacity is the only relevant value.

Currently, if no format is specified for a new disk volume, we pass the
invalid value "none" as the FS type to 'parted mkpart'.

There doesn't seem to be a way to have parted not format the drive, so
just default to using 'ext2' in this case: this shouldn't cause any harm,
since we are creating a new partition in the first place.

Remove unneeded target path duplication, which could carelessly dereference
NULL. Make it clear where 'key' is actually filled in.

Typo was breaking 'parthelper -g', preventing disk pool definition.

The code in cfg.mk to detect when the git submodule was out of date
worked most of the time, but not when checked out in a certain way.
* cfg.mk: Extract submodule hash from command output and file,
and compare only that, since the format of the full line may vary.
Reported by Mike Burns, with some diagnosis by Daniel P Berrange.

* configure.in: Add --with-qemu-user and --with-qemu-group args
* libvirt.spec.in: use 'qemu' for user/group for Fedora >= 12
* qemud/libvirtd_qemu.arg, qemud/test_libvirtd_qemu.aug,
  src/qemu.conf: Add 'user' and 'group' args for configuration
* src/Makefile.am: Create %localstatedir/cache/libvirt/qemu
* src/qemu_conf.c, src/qemu_conf.h: Load user/group from config
* src/qemu_driver.c: Change user ID/group ID when launching QEMU
  guests. Change user/group ownership on disks/usb/pci devs.
  Put memory dumps in %localstatedir/cache/libvirt/qemu
* src/util.c, src/util.h: Add convenient APIs for converting
  username/groupname to user ID / group ID

* src/driver.h: add new driver functions virDrvNumOfDefinedInterfaces
  and virDrvListDefinedInterfaces
* src/libvirt.c: implements the entry points, calling new driver
  functions
* qemud/remote.c qemud/remote_dispatch_args.h qemud/remote_protocol.[chx]
  qemud/remote_dispatch_prototypes.h qemud/remote_dispatch_ret.h
  qemud/remote_dispatch_table.h src/remote_internal.c: implement the
  client/server side of the RPC

* include/libvirt/libvirt.h[.in]: adds signatures for the new exported
  functions virConnectNumOfDefinedInterfaces and
  virConnectListDefinedInterfaces
* src/libvirt_public.syms: export the new symbols

* libvirt.spec.in: we were still using deprecated configure switches

This renames a lot of the methods in the remote driver client
to more accurately reflect their responsibility of IO handling
vs message handling.

Remove redundant error reporting functions which obscured the
filename/line number reporting. Removed code which created a
virDomain/virNetwork object, since those are silently dropped
in error reporting functions now

* src/remote_internal.c: Remove error() and errorf() in favour of
 macros, and remove server_error in favour of direct call

Splits up the 'call' method moving generic IO code out into
separate method to allow it to be easily reused for sending
data streams

* src/remote_internal.c: Split 'call' into two methods, the first
  with same name serializes a set of method arguments into a
  message, the second 'remoteIO' takes a pre-serialized messages,
  sends it and awaits a reply

* src/remote_internal.c: Rename processCallRecvMsg to
  processCallDispatch, and move code specific to method replies
  into processCallDispatchReply, and rename processCallAsyncEvent
  to processCallDispatchMessage

The 'remote_message_header' struct has a mis-leadingly named
field 'direction'. It is really a reflection of the type of
message, and some types can be sent in either direction. Thus
the field is more accurately named 'type'. No function change.

* qemud/remote_protocol.x: Rename 'direction' to 'type' in
  'remote_message_header. Write better docs describing the
  message header field semantics & usage
* qemud/remote_protocol.c, qemud/remote_protocol.h: Regenerate
* qemud/remote.c, qemud/dispatch.c, src/remote_internal.c
  Update to reflect rename of 'direction' to 'type'

All incoming messages currently get routed to the generic method
remoteDispatchClientRequest() for processing. To allow incoming
data stream messages to bypass this and be routed to a specific
location, a concept of dispatch filters is introduced.

* qemud/qemud.h: Add a qemud_client_filter struct and a callback
  qemud_client_filter_func. Maintain a list of filters on every
  struct qemud_client
* qemud/qemud.c: Move remoteDecodeClientMessageHeader() out of
  qemudWorker() into qemudDispatchClientRead(). Check registered
  message filters in qemudDispatchClientRead() to decide where
  to send incoming messages for dispatch.

The remoteDispatchClientRequest() method is currently hardwired to
assume there is only one type of incoming message, a method call.
To allow for alternate types of incoming messags, the code that is
specific to method calls is being split into a separate method
remoteDispatchClientCall

* qemud/dispatch.c: Move method call specific code out into
  remoteDispatchClientCall. Add a helper remoteSerializeError
  for returning error messages to client

The current qemudRegisterClientEvent() code is used both for
registering the initial socket watch, and updating the already
registered watch. This causes unneccessary complexity in alot
of code which only cares about updating existing watches. The
updating of a watch cannot ever fail, nor is a reference to the
'qemud_server' object required.

This introduces a new qemudUpdateClientEvent() method for that
case, allowing the elimination of unneccessary error checking
and removal of the server back-reference in struct qemud_client.

* qemud/qemud.h: Remove 'server' field from struct qemud_client.
  Add qemudUpdateClientEvent() method. Remove 'update' param
  from qemudRegisterClientEvent method
* qemud/dispatch.c, qemud/qemud.c, qemud/remote.c: Update alot
  of code to use qemudUpdateClientEvent() instead of
  qemudRegisterClientEvent(). Move more logic from remoteRelayDomainEvent
  into remoteDispatchDomainEventSend.

This removes an assumption from qemudWorker() code that every
incoming message will generate a reply.

* qemud/dispatch.c: remoteDispatchClientRequest now has responsibility
  for queuing the reply message to the RPC call
* qemud/qemud.c: Do not queue the RPC call reply in qemudWorker(),
  allowing remoteDispatchClientRequest() to take care of it

The naming convention for structs used in the RPC layer is for
incoming requests to be called XXXX_args, and the associated
outgoing reply to be called XXXX_ret.  Asynchronously emitted
messages (eg events) are re-using the XXXX_ret naming scheme.
This patch changes that such that async messages are XXXX_msg,
and stops adding entries for them in the dispatch table, avoiding
the need for a dummy no-op implementation.

* qemud/remote.c: Remove dummy remoteDispatchDomainEvent, no
  longer required. Update to replace remote_domain_event_ret
  with xdr_remote_domain_event_msg
* qemud/remote_protocol.x: Rename remote_domain_event_ret to
  remote_domain_event_msg
* qemud/remote_generate_stubs.pl: Adding handling for new
  XXX_msg structs.
* src/remote_internal.c: Rename remote_domain_event_ret to
  remote_domain_event_msg
* qemud/remote_dispatch_prototypes.h, qemud/remote_dispatch_ret.h,
  qemud/remote_dispatch_table.h, qemud/remote_protocol.h,
  qemud/remote_protocol.c: auto-regenerate

Introduces an API for encoding the header field for outgoing messages
allowing some duplicated code to be eliminated

* qemud/dispatch.c, qemud/dispatch.h: add remoteEncodeClientMessageHeader
  for encoding message header. Update remoteDispatchClientRequest to
  use this method.
* qemud/remote.c: Update remoteDispatchDomainEventSend to use the
  generic remoteEncodeClientMessageHeader() for encoding event
  message hedaders. Push some logic from remoteRelayDomainEvent
  down into remoteDispatchDomainEventSend.

Separate the decoding of incoming request header out from the
dispatch code. This will allow later code to making dispatcher
routing decisions based on the header field data.

* qemud/dispatch.c, qemud/dispatch.h: Add remoteDecodeClientMessageHeader
  API for decoding the header of a client message. Update the
  remoteDispatchClientRequest method to assume a pre-decoded
  header.
* qemud/qemud.h: Include a 'remote_message_header' field in
  'struct qemud_client_message' for pre-decoded header data
* qemud/qemud.c: Decode the incoming client message header before
  invoking remoteDispatchClientRequest

* po/POTFILES.in: Add qemud/dispatch.c
* qemud/dispatch.c, qemud/dispatch.h: Generic code handling dispatch of
  RPC messages.
* qemud/Makefile.am: Add dispatch.c to build
* qemud/qemud.c: Include dispatch.h
* qemud/qemud.h: Remove remoteDispatchClientRequest, remoteRelayDomainEvent
  now in dispatch.h
* qemud/remote.c: Remove remoteDispatchClientRequest, remoteRelayDomainEvent
  now in dispatch.c, and dispatch_args, dispatch_ret, dispatch_fn & dispatch_data
  now in remote.h
* qemud/remote.h: Add typedefs for dispatch_args, dispatch_ret,
  dispatch_fn, dispath_data. Add remoteGetDispatchData() API

* src/qemu_driver.c (qemudDomainCoreDump): New
  (qemuDriver): Add core dump function. The behaviour is similar
  as the current Xen dump

* src/virsh.c: adds cd and pwd commands to virsh useful for save and
  restore commands
* docs/virsh.pod virsh.1: update the documentation
* AUTHORS: add Paolo Bonzini

This makes "make syntax-check" fail when a version-controlled
file contains a trailing blank line.
* cfg.mk (sc_prohibit_trailing_blank_lines): New rule.

by running this command:
git ls-files -z | xargs -0 perl -pi -0777 -e 's/\n\n+$/\n/'
This is in preparation for a more strict make syntax-check
rule that will detect trailing blank lines.

* src/qemu_driver.c: Relabel disk images *before* running QEMU
hotplug monitor commands

* src/qemu_driver.c: Try new monitor syntax for hotplug first. If
  that fails fallback to old KVM specific syntax

after a libvirtd daemon restart with active guests

* src/qemu_driver: Read and dicard pending monitor data
  before issuing new monitor commands.

* src/remote_internal.c: Disable libvirtd autostart if the
  LIBVIRT_AUTOSTART=0 env variable is set
* src/libvirt.c: Document environment variables can impact
  the virConnectOpen API