Currently only file backed memory backend can
be created with a "share" flag in order to allow
sharing guest RAM with other processes in the host.
Add the "share" flag also to RAM Memory Backend
in order to allow remapping parts of the guest RAM
to different host virtual addresses. This is needed
by the RDMA devices in order to remap non-contiguous
QEMU virtual addresses to a contiguous virtual address range.
Moved the "share" flag to the Host Memory base class,
modified phys_mem_alloc to include the new parameter
and a new interface memory_region_init_ram_shared_nomigrate.
There are no functional changes if the new flag is not used.
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Marcel Apfelbaum <marcel@redhat.com>
Now that qcow & qcow2 are wired up to get encryption keys
via the QCryptoSecret object, nothing is relying on the
interactive prompting for passwords. All the code related
to password prompting can thus be ripped out.
Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Message-id: 20170623162419.26068-17-berrange@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
The original ready < nhandles - 1 can be re-written as ready + 1 <
nhandles. The check was actually incorrect because
WAIT_OBJECT_0 was not subtracted from ready; it worked because
WAIT_OBJECT_0 is zero. After subtracting WAIT_OBJECT_0,
the result is the same condition that we are checking on the first
itteration of the for loop. This means we can remove the if statement
and let the for loop check the code.
Signed-off-by: Alistair Francis <alistair.francis@xilinx.com>
Message-Id: <a14083d681951f3999a0e9314605cb706381ae8d.1498756113.git.alistair.francis@xilinx.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Using "-mem-prealloc" option for a large guest leads to higher guest
start-up and migration time. This is because with "-mem-prealloc" option
qemu tries to map every guest page (create address translations), and
make sure the pages are available during runtime. virsh/libvirt by
default, seems to use "-mem-prealloc" option in case the guest is
configured to use huge pages. The patch tries to map all guest pages
simultaneously by spawning multiple threads. Currently limiting the
change to QEMU library functions on POSIX compliant host only, as we are
not sure if the problem exists on win32. Below are some stats with
"-mem-prealloc" option for guest configured to use huge pages.
------------------------------------------------------------------------
Idle Guest | Start-up time | Migration time
------------------------------------------------------------------------
Guest stats with 2M HugePage usage - single threaded (existing code)
------------------------------------------------------------------------
64 Core - 4TB | 54m11.796s | 75m43.843s
64 Core - 1TB | 8m56.576s | 14m29.049s
64 Core - 256GB | 2m11.245s | 3m26.598s
------------------------------------------------------------------------
Guest stats with 2M HugePage usage - map guest pages using 8 threads
------------------------------------------------------------------------
64 Core - 4TB | 5m1.027s | 34m10.565s
64 Core - 1TB | 1m10.366s | 8m28.188s
64 Core - 256GB | 0m19.040s | 2m10.148s
-----------------------------------------------------------------------
Guest stats with 2M HugePage usage - map guest pages using 16 threads
-----------------------------------------------------------------------
64 Core - 4TB | 1m58.970s | 31m43.400s
64 Core - 1TB | 0m39.885s | 7m55.289s
64 Core - 256GB | 0m11.960s | 2m0.135s
-----------------------------------------------------------------------
Changed in v2:
- modify number of memset threads spawned to min(smp_cpus, 16).
- removed 64GB memory restriction for spawning memset threads.
Changed in v3:
- limit number of threads spawned based on
min(sysconf(_SC_NPROCESSORS_ONLN), 16, smp_cpus)
- implement memset thread specific siglongjmp in SIGBUS signal_handler.
Changed in v4
- remove sigsetjmp/siglongjmp and SIGBUS unblock/block for main thread
as main thread no longer touches any pages.
- simplify code my returning memset_thread_failed status from
touch_all_pages.
Signed-off-by: Jitendra Kolhe <jitendra.kolhe@hpe.com>
Message-Id: <1487907103-32350-1-git-send-email-jitendra.kolhe@hpe.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
A fix has been committed in upstream glib commit
210a9796f78eb90f76f1bd6a304e9fea05e97617.
(See also related bug https://bugzilla.gnome.org/show_bug.cgi?id=764415)
It is desirable to use the glib version instead of qemu copy, since it
provides more debugging facilities (G_MAIN_POLL_DEBUG etc), and
hopefully has a better maintainance. Hopefully, we can drop the qemu
copy in a few years.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
This is a small helper that tries to fetch binary name for given
PID.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Message-Id: <4d75d475c1884f8e94ee8b1e57273ddf3ed68bf7.1474987617.git.mprivozn@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
When adding hostmem backend at runtime, QEMU might exit with error:
"os_mem_prealloc: Insufficient free host memory pages available to allocate guest RAM"
It happens due to os_mem_prealloc() not handling errors gracefully.
Fix it by passing errp argument so that os_mem_prealloc() could
report error to callers and undo performed allocation when
os_mem_prealloc() fails.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <1469008443-72059-1-git-send-email-imammedo@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Remove glib.h includes, as it is provided by osdep.h.
This commit was created with scripts/clean-includes.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Eric Blake <eblake@redhat.com>
Tested-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Move declarations out of qemu-common.h for functions declared in
utils/ files: e.g. include/qemu/path.h for utils/path.c.
Move inline functions out of qemu-common.h and into new files (e.g.
include/qemu/bcd.h)
Signed-off-by: Veronia Bahaa <veroniabahaa@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Commit 57cb38b included qapi/error.h into qemu/osdep.h to get the
Error typedef. Since then, we've moved to include qemu/osdep.h
everywhere. Its file comment explains: "To avoid getting into
possible circular include dependencies, this file should not include
any other QEMU headers, with the exceptions of config-host.h,
compiler.h, os-posix.h and os-win32.h, all of which are doing a
similar job to this file and are under similar constraints."
qapi/error.h doesn't do a similar job, and it doesn't adhere to
similar constraints: it includes qapi-types.h. That's in excess of
100KiB of crap most .c files don't actually need.
Add the typedef to qemu/typedefs.h, and include that instead of
qapi/error.h. Include qapi/error.h in .c files that need it and don't
get it now. Include qapi-types.h in qom/object.h for uint16List.
Update scripts/clean-includes accordingly. Update it further to match
reality: replace config.h by config-target.h, add sysemu/os-posix.h,
sysemu/os-win32.h. Update the list of includes in the qemu/osdep.h
comment quoted above similarly.
This reduces the number of objects depending on qapi/error.h from "all
of them" to less than a third. Unfortunately, the number depending on
qapi-types.h shrinks only a little. More work is needed for that one.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
[Fix compilation without the spice devel packages. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Now that QEMU wraps the Win32 sockets methods to automatically
set errno upon failure, there is no reason for callers to use
the socket_error() method. They can rely on accessing errno
even on Win32. Remove all use of socket_error() from general
code, leaving it as a static method in oslib-win32.c only.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The windows socket functions look identical to the normal POSIX
sockets functions, but instead of setting errno, the caller needs
to call WSAGetLastError(). QEMU has tried to deal with this
incompatibility by defining a socket_error() method that callers
must use that abstracts the difference between WSAGetLastError()
and errno.
This approach is somewhat error prone though - many callers of
the sockets functions are just using errno directly because it
is easy to forget the need use a QEMU specific wrapper. It is
not always immediately obvious that a particular function will
in fact call into Windows sockets functions, so the dev may not
even realize they need to use socket_error().
This introduces an alternative approach to portability inspired
by the way GNULIB fixes portability problems. We use a macro to
redefine the original socket function names to refer to a QEMU
wrapper function. The wrapper function calls the original Win32
sockets method and then sets errno from the WSAGetLastError()
value.
Thus all code can simply call the normal POSIX sockets APIs are
have standard errno reporting on error, even on Windows. This
makes the socket_error() method obsolete.
We also bring closesocket & ioctlsocket into this approach. Even
though they are non-standard Win32 names, we can't wrap the normal
close/ioctl methods since there's no reliable way to distinguish
between a file descriptor and HANDLE in Win32.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Historically QEMU has had a socket_error() macro that was
defined to map to WSASocketError(). The os-win32.h header
file would define errno constants that mapped to the
WSA error constants. This worked fine with Mingw32 since
its header files never defined any errno values, nor did
it even provide an errno.h. So callers of socket_error()
could match on traditional Exxxx constants and it would
all "just work".
With Mingw64 though, things work rather differently. First
there is an errno.h file which defines all the traditional
errno constants you'd expect from a UNIX platform. There
is then a winerror.h which defined the WSA error constants.
Crucially the WSAExxxx errno values in winerror.h do not
match the Exxxx errno values in error.h.
If QEMU had only imported winerror.h it would still work,
but the qemu/osdep.h file unconditionally imports errno.h.
So callers of socket_error() will get now WSAExxxx values
back and compare them to the Exxx constants. This will
always fail silently at runtime.
To solve this QEMU needs to stop assuming the WSAExxxx
constant values match the Exxx constant values. Thus the
socket_error() macro is turned into a small function that
re-maps WSAExxxx values into Exxx.
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Clean up includes so that osdep.h is included first and headers
which it implies are not included manually.
This commit was created with scripts/clean-includes.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1454089805-5470-6-git-send-email-peter.maydell@linaro.org
getpagesize on Linux returns an int. Fix QEMU's implementation for
Windows to return an int (instead of size_t), too.
This fixes a compiler warning which was introduced recently
(commit 093e3c42).
Signed-off-by: Stefan Weil <sw@weilnetz.de>
When using regular fork() the child process of course inherits
all the parents' signal handlers. If the child then proceeds
to close() any open file descriptors, it may break some of those
registered signal handlers. The child generally does not want to
ever run any of the signal handlers that the parent may have
installed in the short time before it exec's. The parent may also
have blocked various signals which the child process will want
enabled.
This introduces a wrapper qemu_fork() that takes care to sanitize
signal handling across fork. Before forking it blocks all signals
in the parent thread. After fork returns, the parent unblocks the
signals and carries on as usual. The child, however, resets all the
signal handlers back to their defaults before it unblocks signals.
The child process can now exec the binary in a "clean" signal
environment.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The oslib-win32 file currently provides a localtime_r and
gmtime_r replacement unconditionally. Some versions of
Mingw-w64 would provide crude macros for localtime_r/gmtime_r
which QEMU takes care to disable. Latest versions of Mingw-w64
now provide actual functions for localtime_r/gmtime_r, but
with a twist that you have to include unistd.h or pthread.h
before including time.h. By luck some files in QEMU have
such an include order, resulting in compile errors:
CC util/osdep.o
In file included from include/qemu-common.h:48:0,
from util/osdep.c:48:
include/sysemu/os-win32.h:77:12: error: redundant redeclaration of 'gmtime_r' [-Werror=redundant-decls]
struct tm *gmtime_r(const time_t *timep, struct tm *result);
^
In file included from include/qemu-common.h:35:0,
from util/osdep.c:48:
/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:272:107: note: previous definition of 'gmtime_r' was here
In file included from include/qemu-common.h:48:0,
from util/osdep.c:48:
include/sysemu/os-win32.h:79:12: error: redundant redeclaration of 'localtime_r' [-Werror=redundant-decls]
struct tm *localtime_r(const time_t *timep, struct tm *result);
^
In file included from include/qemu-common.h:35:0,
from util/osdep.c:48:
/usr/i686-w64-mingw32/sys-root/mingw/include/time.h:269:107: note: previous definition of 'localtime_r' was here
This change adds a configure test to see if localtime_r
exits, and only enables the QEMU impl if missing. We also
re-arrange qemu-common.h try attempt to guarantee that all
source files get unistd.h before time.h and thus see the
localtime_r/gmtime_r defs.
[sw: Use "official" spellings for Mingw-w64, MinGW in comments.]
[sw: Terminate sentences with a dot in comments.]
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: Stefan Weil <sw@weilnetz.de>
The qemu-img.c file has a read_password() method impl that is
used to prompt for passwords on the console, with impls for
POSIX and Windows. This will be needed by qemu-io.c too, so
move it into the QEMU osdep/oslib files where it can be shared
without code duplication
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
introduce memory_region_get_alignment() that returns
underlying memory block alignment or 0 if it's not
relevant/implemented for backend.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This function returns NULL instead of aborting when an allocation fails.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Benoit Canet <benoit@irqsave.net>
numa work by Hu Tao and others
memory hotplug by Igor
vhost-user by Nikolay, Antonios and others
guest virtio announcements by Jason
qtest fixes by Sergey
qdev hotplug fixes by Paolo
misc other fixes mostly by myself
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAABAgAGBQJTowW+AAoJECgfDbjSjVRpnMMH/jp3sKGzJumYLbi5ihjmYyND
jYd6ySXoVAjUTgaCvdje5srisOap8pbc783kQvQS4CeWsjZ5Vvh+PZjkBPIqF1pD
celxGQ43CY7QSUWq+02Dg9VIUwLwZqdKlxNsV01FligQn+ZBQ6sQ6ksWx7oGzqRt
5/HMZykbwUvSk/4xGUaMn2+/4uhQ0Wz5EsCkv9L/u8kS72k6ldc/tCGZMzBUNHTM
rW5FPYwMQP0MXgGTXnlLEQjJ7Lozc66IaMZoHw/a/aGSIxdag9Otj0ADuXq6yZaV
Xi4O/EOJWd1JpSG7w8LOyIZNakpHkU43fmJCLzBjDAupHeRp57TcW5ox4PJYAtg=
=Oxdt
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging
pc,pci,virtio,hotplug fixes, enhancements
numa work by Hu Tao and others
memory hotplug by Igor
vhost-user by Nikolay, Antonios and others
guest virtio announcements by Jason
qtest fixes by Sergey
qdev hotplug fixes by Paolo
misc other fixes mostly by myself
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
* remotes/mst/tags/for_upstream: (109 commits)
numa: use RAM_ADDR_FMT with ram_addr_t
qapi/string-output-visitor: fix bugs
tests: simplify code
qapi: fix input visitor bugs
acpi: rephrase comment
qmp: add ACPI_DEVICE_OST event handling
qmp: add query-acpi-ospm-status command
acpi: implement ospm_status() method for PIIX4/ICH9_LPC devices
acpi: introduce TYPE_ACPI_DEVICE_IF interface
qmp: add query-memory-devices command
numa: handle mmaped memory allocation failure correctly
pc: acpi: do not hardcode preprocessor
qmp: clean out whitespace
qdev: recursively unrealize devices when unrealizing bus
qdev: reorganize error reporting in bus_set_realized
qapi: fix build on glib < 2.28
qapi: make string output visitor parse int list
qapi: make string input visitor parse int list
tests: fix memory leak in test of string input visitor
hmp: add info memdev
...
Conflicts:
include/hw/i386/pc.h
[PMM: fixed minor conflict in pc.h]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
So that backends can use it.
Since we need the page size for efficiency, move code to compute it
out of translate-all.c and into util/oslib-win32.c.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Hu Tao <hutao@cn.fujitsu.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Commit 5a007547df tried to fix a
performance degradation caused by bad handling of small timeouts
in the original implementation of g_poll.
Since that commit, hard disk I/O no longer works.
Instead of rewriting the g_poll implementation, this patch simply copies
the original code (released under LGPL) from latest glib and only modifies
it where needed (see comments in the code). URL of the original code:
https://git.gnome.org/browse/glib/tree/glib/gpoll.c
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Message-id: 1401291744-14314-1-git-send-email-sw@weilnetz.de
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
g_poll has a problem on Windows when using
timeouts < 10ms, in glib/gpoll.c:
/* If not, and we have a significant timeout, poll again with
* timeout then. Note that this will return indication for only
* one event, or only for messages. We ignore timeouts less than
* ten milliseconds as they are mostly pointless on Windows, the
* MsgWaitForMultipleObjectsEx() call will timeout right away
* anyway.
*/
if (retval == 0 && (timeout == INFINITE || timeout >= 10))
retval = poll_rest (poll_msgs, handles, nhandles, fds, nfds, timeout);
so whenever g_poll is called with timeout < 10ms it does
a quick poll instead of wait, this causes significant performance
degradation of QEMU, thus we should use WaitForMultipleObjectsEx
directly
Signed-off-by: Stanislav Vorobiov <s.vorobiov@samsung.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
With this change, main() calls qemu_init_exec_dir and uses argv[0] to
init exec_dir. The saved value can be retrieved with
qemu_get_exec_dir later. It will be reused by module loading.
Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Using stdin with readline.c requires disabling echo and line buffering.
Add a portable wrapper to set the terminal attributes under Linux and
Windows.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
If a socket is closed it remains in TIME_WAIT state for some time. On operating
systems using BSD sockets the endpoint of the socket may not be reused while in
this state unless SO_REUSEADDR was set on the socket. On windows on the other
hand the default behaviour is to allow reuse (i.e. identical to SO_REUSEADDR on
other operating systems) and setting SO_REUSEADDR on a socket allows it to be
bound to a endpoint even if the endpoint is already used by another socket
independently of the other sockets state. This can even result in undefined
behaviour.
Many sockets used by QEMU should not block the use of their endpoint after being
closed while they are still in TIME_WAIT state. Currently QEMU sets SO_REUSEADDR
for such sockets, which can lead to problems on Windows. This patch introduces
the function socket_set_fast_reuse that should be used instead of setting
SO_REUSEADDR when fast socket reuse is desired and behaves correctly on all
operating systems.
As a failure of this function can only be caused by bad QEMU internal errors, an
assertion handles these situations. The return value is still passed on, to
minimize changes in client code and prevent unused variable warnings if NDEBUG
is defined.
Signed-off-by: Sebastian Ottlik <ottlik@fzi.de>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Stefan Weil <sw@weilnetz.de>
We abort() on memory allocation failure. abort() is appropriate for
programming errors. Maybe most memory allocation failures are
programming errors, maybe not. But guest memory allocation failure
isn't, and aborting when the user asks for more memory than we can
provide is not nice. exit(1) instead, and do it in just one place, so
the error message is consistent.
Tested-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Andreas Färber <afaerber@suse.de>
Acked-by: Laszlo Ersek <lersek@redhat.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Message-id: 1375276272-15988-8-git-send-email-armbru@redhat.com
Signed-off-by: Anthony Liguori <anthony@codemonkey.ws>
This function returns ${prefix}/var/RELATIVE_PATHNAME on POSIX-y systems,
and <CSIDL_COMMON_APPDATA>/RELATIVE_PATHNAME on Win32.
http://msdn.microsoft.com/en-us/library/bb762494.aspx
[...] This folder is used for application data that is not user
specific. For example, an application can store a spell-check
dictionary, a database of clip art, or a log file in the
CSIDL_COMMON_APPDATA folder. [...]
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
We switched from qemu_memalign to mmap() but then we don't modify
qemu_vfree() to do a munmap() over free(). Which we cannot do
because qemu_vfree() frees memory allocated by qemu_{mem,block}align.
Introduce a new function that does the munmap(), luckily the size is
available in the RAMBlock.
Reported-by: Amos Kong <akong@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Amos Kong <akong@redhat.com>
Message-id: 1368454796-14989-3-git-send-email-pbonzini@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This is preparatory to the introduction of a separate freeing API.
Reported-by: Amos Kong <akong@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Amos Kong <akong@redhat.com>
Message-id: 1368454796-14989-2-git-send-email-pbonzini@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The fcntl(fd, F_SETFL, O_NONBLOCK) flag is not specific to sockets.
Rename to qemu_set_nonblock() just like qemu_set_cloexec().
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
On POSIX, qemu_vfree() accepts NULL, because it's merely wrapper
around free(). As far as I can tell, the Windows implementation
doesn't. Breeds bugs that bite only under Windows.
Make the Windows implementation behave like the POSIX implementation.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
