We can exploit the fact that gcc warns about int-to-pointer conversion
in ternary cond?(void*):(int) in order to prevent future mistakes of
calling VIR_FREE on a scalar lvalue.  For example, between commits
158ba873 and 802e2df9, we would have had this warning:

cc1: warnings being treated as errors
remote.c: In function 'remoteDispatchListNetworks':
remote.c:3684:70: error: pointer/integer type mismatch in conditional expression

There are still a number of places that malloc into a const char*;
while it would probably be worth scrubbing them to use char*
instead, that is a separate patch, so we have to cast away const
in VIR_FREE for now.

* src/util/memory.h (VIR_FREE): Make gcc warn about integers.
Iteratively developed from a patch by Christophe Fergeau.

mingw lacks the counterpart to PTHREAD_MUTEX_INITIALIZER, so the
best we can do is portably expose once-only runtime initialization.

* src/util/threads.h (virOnceControlPtr): New opaque type.
(virOnceFunc): New callback type.
(virOnce): New prototype.
* src/util/threads-pthread.h (virOnceControl): Declare.
(VIR_ONCE_CONTROL_INITIALIZER): Define.
* src/util/threads-win32.h (virOnceControl)
(VIR_ONCE_CONTROL_INITIALIZER): Likewise.
* src/util/threads-pthread.c (virOnce): Implement in pthreads.
* src/util/threads-win32.c (virOnce): Implement in WIN32.
* src/libvirt_private.syms: Export it.

To be used to share a CURL handle between multiple threads in the
upcoming domain event support.

Commit 36b65213 renamed err to rerr, do the same in the generator.

This patch strips reusable part of qemuDomainUpdateDeviceFlags()
and consolidate it to qemuDomainModifyDeviceFlags().
No functional changes.

* src/qemu/qemu_driver.c
(qemuDomainChangeDiskMediaLive) : pulled out code for updating disks.
(qemuDomainUpdateDeviceLive) : core of UpdateDevice, extracted from
UpdateDeviceFlags()
(qemuDomainModifyDeviceFlags): add support for updating device in live domain.
(qemuDomainUpdateDeviceFlags): reworked as a wrapper function of
qemuDomainModifyDeviceFlags()
Signed-off-by: NKAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: NEric Blake <eblake@redhat.com>

clean up At(De)tachDeviceFlags() for consolidation.

qemuDomainAttachDeviceFlags()/qemuDomainDetachFlags()/
qemuDomainUpdateDeviceFlags() has similar logics and copied codes.

This patch series tries to unify them to use shared code when it can.
At first, clean up At(De)tachDeviceFlags() and devide it into functions.

By this, this patch pulls out shared components between functions.
Based on patch series by Eric Blake, I added some modification as
switch-case with QEMU_DEVICE_ATTACH, QEMU_DEVICE_DETACH, QEMU_DEVICE_UPDATE

* src/qemu/qemu_driver.c
(qemuDomainAt(De)tachDeviceFlags) : pulled out to qemuDomainModifyDeviceFlags()
(qemuDomainModifyDeviceFlags) : implements generic code for modifying domain.
(qemuDomainAt(De)tachDeviceFlagsLive) : code for at(de)taching devices to
domain in line. no changes in logic from old code.
(qemuDomainAt(De)tachDeviceDiskLive) : for at(de)taching Disks.
(qemuDomainAt(De)tachDeviceControllerLive) : for at(de)taching Controllers
Signed-off-by: NKAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: NEric Blake <eblake@redhat.com>

Centralize device modification in the more flexible APIs, to allow future
honoring of additional flags.  Explicitly reject the
VIR_DOMAIN_DEVICE_MODIFY_FORCE flag on attach/detach.

Based on Eric Blake<eblake@redhat.com>'s work.

* src/qemu/qemu_driver.c
(qemudDomainAttachDevice)(qemudDomainAttachDeviceFlags): Swap bodies,rename...
(qemudDomainDetachDevice, qemudDomainDetachDeviceFlags): Likewise.
Signed-off-by: NKAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: NEric Blake <eblake@redhat.com>

Up to now we missed parser for cpuinfo on x390(x) machines. Those machines
have only 1 thread, core, socket. What is missing is information about
CPU frequency.

The two ends of the pipe used for feeding QEMU tunnelled
migration data were interchanged, so QEMU got given the
"write" end instead of the "read" end.

The qemuMigrationPrepareTunnel method was also immediately
closing the "write" end of the pipe, so the stream failed
to actually write anything.

* src/qemu/qemu_migration.c: Swap tunnelled migration
  pipe FDs & don't close pipe given to stream

ARRAY_CARDINALITY is typed as size_t, not long; this matters on 32-bit
platforms:

hashtest.c: In function 'testHashRemoveForEach':
hashtest.c:114: error: format '%lu' expects type 'long unsigned int', but argument 4 has type 'unsigned int' [-Wformat]

* tests/hashtest.c (testHashRemoveForEach): Use correct format.

Remove the artificial minimum of 4096 KB for guest memory. It's drivers'
job to set the limit if needed.

Adding reboot <domain> function for pHyp driver.

* .mailmap: Add an alias for last commit.

If the migrateFrom is "stdio" not "stdin", qemuBuildCommandLine()
will convert it to exec:cat or fd:n.

Ebtables filtering doesn't work on macvtap device. Remove support for direct type of interface.

Gnulib already guarantees <stdbool.h>, so it is easier to just
use the standardized spellings.

* tools/virsh.c (vshCmdDef): Change callback to return real bool.
(__vshControl): Change several fields to bool.
(vshCommandOptBool): Change return type.
All callers updated.
* tools/Makefile.am (virsh-net-edit.c, virsh-pool-edit.c):
Likewise.

If vm is unlocked in virDomainObjUnref(), the return value is 0, not less
than 0.

Here is a new version of this patch:
https://www.redhat.com/archives/libvir-list/2011-April/msg00337.html

v2:
  - store the cputune info for the whole runtime of the domain
  - remove cputune info when domain is destroyed

The nodeGetInfo code had to be moved into a helper
function to reuse it without a virConnectPtr.

Inspired by Eric Blake

Rather than copying and pasting lots of code, factor it into a
single helper function.

This commit adds a warning if tighter integer parsing would fail
due to any stray bytes after the number, but should not change
any behavior other than the bug fix for phypNumDomainsGeneric
looking only at numeric lines.

* src/phyp/phyp_driver.c (phypExecInt): New function.
(phypGetVIOSPartitionID, phypNumDomainsGeneric, phypGetLparID)
(phypGetLparMem, phypGetLparCPUGeneric, phypGetRemoteSlot)
(phypGetVIOSNextSlotNumber, phypAttachDevice)
(phypGetStoragePoolSize, phypStoragePoolNumOfVolumes)
(phypNumOfStoragePools, phypInterfaceDestroy)
(phypInterfaceDefineXML, phypInterfaceLookupByName)
(phypInterfaceIsActive, phypNumOfInterfaces): Use it.
(phypNumDomainsGeneric): Correctly find numeric line.

* .gitignore: Add exemption for hashtest.

This last minute addition caused a build failure

cc1: warnings being treated as errors
qemu/qemu_process.c: In function 'qemuProcessHandleWatchdog':
qemu/qemu_process.c:436:34: error: ignoring return value of 'virDomainObjUnref', declared with attribute warn_unused_result [-Wunused-result]
make[3]: *** [libvirt_driver_qemu_la-qemu_process.lo] Error 1

Replace some occurrances of

  virDomainPtr domain;
  virNetworkPtr network;

With

  virDomainPtr dom;
  virNetworkPtr net;

* daemon/remote.c: Fix variable naming to follow standard

Replace cases of

     type = virConnectGetType(conn);
     if (type == NULL)
         goto cleanup;

With

     if (!(type = virConnectGetType(conn)))
         goto cleanup;

* daemon/remote.c: Write error checks in compat form

Replace all occurrances of

   if (....) {
      goto cleanup;
   }

With

   if (.....)
      goto cleanup;

to save one line of code

* daemon/remote.c: Remove curly braces on single line conditionals

The libvirt APIs reserve any negative value for indicating an
error. Thus checks

    if (virXXXX() == -1)

Should instead be

    if (virXXXX() < 0)

* daemon/remote.c: s/ == -1/ < 0/

The dispatcher functions have numerous places where they
return to the caller. This leads to duplicated cleanup
code, often resulting in memory leaks. It makes it harder
to ensure that errors are dispatched before freeing objects,
which may overwrite the original error.

The standard pattern is now

    remoteDispatchXXX(...) {
        int rv = -1;

        ....
        if (XXX < 0)
          goto cleanup;
        ...
        if (XXXX < 0)
          goto cleanup;
        ...

        rv = 0;
    cleanup:
        if (rv < 0)
           remoteDispatchError(rerr);
        ...free all other stuff..
        return rv;
    }

* daemon/remote.c: Centralize all cleanup paths
* daemon/stream.c: s/remoteDispatchConnError/remoteDispatchError/
* daemon/dispatch.c, daemon/dispatch.h: Replace
  remoteDispatchConnError with remoteDispatchError
  removing unused virConnectPtr

To install it, disable libvirtd sysv initscript:
    chkconfig libvirtd off
    service libvirtd stop

and enable libvirtd upstart job:
    cp  /usr/share/doc/libvirt-*/libvirtd.upstart \
        /etc/init/libvirtd.conf
    initctl reload-configuration
    initctl start libvirtd

Test:
    initctl status libvirtd
libvirtd start/running, process 3929
    killall -9 libvirtd
    initctl status libvirtd
libvirtd start/running, process 4047

I looked into the possibility to use the upstart script from Ubuntu or
at least getting inspiration from it but that's not possible. "expect
daemon" is a nice thing but it only works if the process is defined with
exec stanza instead of script ... no script. Unfortunately, with exec
stanza environment variables can only be set within upstart script
(i.e., configuration in /etc/sysconfig/libvirtd can't work). Hence, we
need to use script stanza, source sysconfig, and execute libvirtd
without --daemon. For similar reasons we can't use limit stanza and need
to handle DAEMON_COREFILE_LIMIT in job's script.

This patch does the following two things:
1. hold an extra reference while handling watchdog event
   If the domain is not persistent, and qemu quits unexpectedly before
   calling processWatchdogEvent(), vm will be freed and the function
   processWatchdogEvent() will be dangerous.

2. unlock qemu driver and vm before returning from processWatchdogEvent()
   When the function processWatchdogEvent() failed, we only free wdEvent,
   but forget to unlock qemu driver and vm, free dumpfile.

We do not lock qemu_driver when calling virThreadPoolNew(). If it failed,
we will unlock qemu_driver. It is dangerous.

We may use this pool during auto starting domains. So we must create it before
calling qemuAutostartDomains(). Otherwise, libvirtd will crash.

Also mark error messages in block_stats.c for translation, add the
new macro to the msg_gen functions in cfg.mk and add block_stats.c
to po/POTFILES.in

And from all related macros and functions.

This is a basic set of tests for testing removals of hash entries during
iteration.

commit d4601696 introduces two more generated files: esx_vi.generated.h
and esx_vi.generated.h. But we do not include them into dist file.
It will break building if using dist file to build.

* tests/virsh-optparse: New file.
* tests/Makefile.am (test_scripts): Use it.

Regression introduced in 0.8.5, commit c1564268.  The command
'virsh freecell 0' quit working when it changed from an optional
string to an optional integer.

This patch introduces a slight change that specifying an option
twice is now detected as an error.  It also changes things so
that a command that has more than 1 required option will not
complain about missing options if one but not all of the options
were given in long format, as in 'virsh vol-create --pool p file',
as well as making positional parsing work for all optional
options (each positional argument is associated with the earliest
option that has not yet been seen by name).

Optional boolean options can appear before required argument
options, because they don't affect positional argument parsing,
and obviously a required boolean option makes no sense.

Technically, this patch renders VSH_OT_STRING and VSH_OT_DATA
redundant; but cleaning that up can be a separate patch.

No command should ever need more than 32 options, right? :)

* tools/virsh.c (vshCmddefGetData, vshCmddefGetOption)
(vshCommandCheckOpts): Alter parameters to use bitmaps.
(vshCmddefOptParse): New function.
(vshCommandParse): Update for better handling of positional
arguments.
(vshCmddefHelp): Allow unit tests to validate options.

The current state of virsh parsing is that:

$ virsh vol-info /path/to/image
$ virsh vol-info --pool default /path/to/image
$ virsh vol-info --pool default --vol /path/to/image

all lookup the volume by path (technically, the last two also attempt
a name lookup within a pool, whereas the first skips that step, but
the end result is the same); meanwhile:

$ virsh vol-info default /path/to/image

complains about unexpected data.  Why?  Because the --pool option is
optional, so default was parsed as the --vol argument, and
/path/to/image.img doesn't match up with any remaining options that
require an argument.  For proof, note that:

$ virsh vol-info default --vol /path/to/image

complains about looking up 'default' - the parser mis-associated both
arguments with --vol.  Given the above, the only way to specify pool
is with an explicit "--pool" argument (you can't specify it
positionally).  However, named arguments can appear in any order, so:

$ virsh vol-info /path/to/image --pool default
$ virsh vol-info --vol /path/to/image --pool default

have also always worked.  Therefore, this patch has no functional
change on vol-info option parsing, but only on 'virsh help vol-info'
synopsis layout.  However, it also allows the next patch to 1) enforce
that required options are always first (without this patch, the next
patch would fail the testsuite), and 2) allow the user to omit the
"--pool" argument.  That is, the next patch makes it possible to do:

$ virsh vol-info /path/to/image default

which to date was not possible.

* tools/virsh.c (opts_vol_create_from, opts_vol_clone)
(opts_vol_upload, opts_vol_download, opts_vol_delete)
(opts_vol_wipe, opts_vol_info, opts_vol_dumpxml, opts_vol_key)
(opts_vol_path): List optional pool parameter after required
arguments.