We should bind pci device to original driver when pciBindDeviceToStub() failed.
If the pci device is not bound to any driver before calling pciBindDeviceToStub(),
we should only unbind it from pci-stub. If it is bound to pci-stub, we should not
unbind it from pci-stub.

This patch do the following things:
1. rename the function as 'Unbind' is better than 'UnBind'.
2. pciUnbindDeviceFromStub() will be used in the function pciBindDeviceToStub() in
   next patch. Float it up, instead of having to have a forward declaration

We should not mark pci devices as active when qemuPrepareHostdevPCIDevices()
failed.

This bug was introduce by commit 57162db8, and it will cause libvirtd crashed.

In file included from util/threads.c:31:
util/threads-pthread.c: In function 'virThreadSelfID':
util/threads-pthread.c:214: warning: cast from function call of type 'pthread_t' to non-matching type 'int' [-Wbad-function-cast]

* src/util/threads-pthread.c (virThreadSelfID) [!SYS_gettid]:
Add intermediate cast to silence gcc.

With gcc 4.3.4 I'm seeing the following warning failure

cc1: warnings being treated as errors
cc1: error: -funit-at-a-time is required for inlining of functions
that are only called once [-Wdisabled-optimization]

Add -funit-at-a-time to WARN_CFLAGS.

We're seeing bugs apparently resulting from thread unsafety of
libpciaccess, such as
https://bugs.launchpad.net/ubuntu/+source/libvirt/+bug/726099
To prevent those, as suggested by danpb on irc, move the
nodeDeviceLock(driverState) higher into the callers.  In
particular:

  udevDeviceMonitorStartup should hold the lock while calling
  udevEnumerateDevices(), and udevEventHandleCallback should hold it
  over its entire execution.

It's not clear to me whether it is ok to hold the
nodeDeviceLock while taking the virNodeDeviceObjLock(dev) on a
device.  If not, then the lock will need to be dropped around
the calling of udevSetupSystemDev(), and udevAddOneDevice()
may not actually be safe to call from higher layers with the
driverstate lock held.

libvirt 0.8.8 with this patch on it seems to work fine for me.
Assuming it looks ok and I haven't done anything obviously dumb,
I'll ask the bug submitters to try this patch.
Signed-off-by: NSerge Hallyn <serge.hallyn@ubuntu.com>

* tests/qemuxml2argvdata/qemuxml2argv-*.args: Reflect reserved VGA
port change.

This patch adds max_processes option to qemu.conf which can be used to
override system default limit on number of processes that are allowed to
be running for qemu user.

To address https://bugzilla.redhat.com/show_bug.cgi?id=692355

This fix is to reserve slot 0x02 for primary VGA even if there
is no "video" specified in domain XML to avoid the problem.

cc1: warnings being treated as errors
libxl/libxl_driver.c: In function 'libxlDomainSetVcpusFlags':
libxl/libxl_driver.c:1570:14: error: cast from function call of type 'double' to non-matching type 'unsigned int' [-Wbad-function-cast]
libxl/libxl_driver.c:1578:15: error: cast from function call of type 'double' to non-matching type 'unsigned int' [-Wbad-function-cast]

This was the only use of floor() and ceil(), and floating-point
is overkill for power-of-two manipulations.

* src/libxl/libxl_driver.c (libxlDomainSetVcpusFlags): Avoid -lm
for trivial computations.

The systemtap directory for tapsets is called

  /usr/share/systemtap/tapset

Not

 /usr/share/systemtap/tapsets

* daemon/Makefile.am,libvirt.spec.in: s/tapsets/tapset/

The GCC Win32 compiler will claim to support -fstack-protector,
but if it actually gets triggered by a suitable code pattern,
linking will fail. Other non-Linux OS likely suffer the same
way with gcc.

* m4/virt-compile-warnings.m4: Only use stack protector when
  the build target is Linux.

GCC is a little confused about the cast of beginthread/beginthreadex
from unsigned long -> void *. Go via an intermediate variable avoids
the bogus warning, and makes the code a little cleaner

* src/util/threads-win32.c: Avoid compiler warning in cast

The SCSI volumes get a better 'key' field based on the fully
qualified volume path. All SCSI volumes have a unique serial
available in hardware which can be obtained by sending a
suitable SCSI command. Call out to udev's 'scsi_id' command
to fetch this value

* src/storage/storage_backend_scsi.c: Improve volume key
  field value stability and uniqueness

When initializing qemu guest capabilities, we should ignore qemu
binaries that we are not able to extract version/help info from since
they will be unusable for creating domains anyway. Ignoring them is also
much better than letting initialization of qemu driver fail.

The macro is huge and gives us nothing but headache when maintaining it.

A couple of functions were declared using the old style foo()
for no-parameters, instead of foo(void)

* src/xen/xen_hypervisor.c, tests/testutils.c: Replace () with (void)
  in some function declarations
* m4/virt-compile-warnings.m4: Enable -Wold-style-definition

* src/internal.h: Define a ATTRIBUTE_NO_RETURN annotation
* src/lxc/lxc_container.c: Annotate lxcContainerDummyChild
  with ATTRIBUTE_NO_RETURN
* tests/eventtest.c: Mark async thread as ATTRIBUTE_NO_RETURN
* m4/virt-compile-warnings.m4: Enable -Wmissing-noreturn

Add a couple of missing ATTRIBUTE_FMT_PRINTF annotations

* tools/virsh.c, tests/testutils.c: Add printf format attribute
* m4/virt-compile-warnings.m4: Enable -Wmissing-format-attribute

Split the bit acinclude.m4 file into smaller pieces named
as m4/virt-XXXXX.m4

* .gitignore: Ignore gettext related files
* acinclude.m4: Delete
* m4/virt-compile-warnings.m4: Checks for GCC compiler flags
* m4/virt-pkgconfig-back-compat.m4: Backcompat check for
  pkgconfig program

Remove custom code for checking compiler warnings, using
gl_WARN_ADD instead. Don't list all flags ourselves, use
gnulib's gl_MANYWARN_ALL_GCC to get all possible GCC flags,
then turn off the ones we don't want yet.

* acinclude.m4: Rewrite to use gl_WARN_ADD and gl_MANYWARN_ALL_GCC
* bootstrap.conf: Add warnings & manywarnings
* configure.ac: Switch to gl_WARN_ADD
* m4/compiler-flags.m4: Obsoleted by gl_WARN_ADD

* src/openvz/openvz_driver.c: Initialize saveptr variable

Removes 4kb stack allocation in the XenD subdriver.

Replace openvz_readline with getline in several places to get rid of stack
allocated buffers to hold lines.

openvzReadConfigParam allocates memory for return values instead of
expecting a preexisting buffer.

Removes multiple 4kb stack allocations.

Removes TODO comments as suggested by Daniel P. Berrange.

Move the buffers to the heap allocated client/private data structs.

Removing a 4kb stack allocation.

Reduce stack buffer for virStrerror to the common 1kb instead of 4kb.

Allocate on the heap instead.