QAPI will require glib/python, but for now the guest agent is the only
user. For now, make these dependencies an explicit guest agent one, and
give users the option to disable it if need be.
Once QAPI is adopted in core QEMU code, we would basically revert this
patch.
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The system emulation code was not merged before the branch.
Let's leave that work for the next release.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
With vhost_net="" (most non-Linux hosts), configure prints an
error message:
test: 2551: =: unexpected operator
Fix this and similar code by adding the missing "".
Cc: Wolfgang Mauerer <wolfgang.mauerer@siemens.com>
Cc: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Reviewed-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Migrating after unplugging a virtio-balloon device resulted in an error
message on the destination:
Unknown savevm section or instance '0000:00:04.0/virtio-balloon' 0
load of migration failed
Fix this by unregistering the section on device unplug.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Add an exit handler that will free up RAM after a virtio-balloon device
is unplugged.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Negative balloon values don't make sense, reject them and throw a qerror
with QERR_INVALID_PARAMETER_VALUE.
Reported-by: Mike Cao <bcao@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Multiple balloon registrations are not allowed; check if the
registration with the qemu balloon api succeeded. If not, fail the
device init.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Multiple balloon devices don't make sense; disallow more than one
registration attempt to register handlers.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Passing on '0' as ballooning target to indicate retrieval of stats is
bad API. It also makes 'balloon 0' in the monitor cause a segfault.
Have two different functions handle the different functionality instead.
Detailed explanation from Markus's review:
1. do_info_balloon() is an info_async() method. It receives a callback
with argument, to be called exactly once (callback frees the
argument). It passes the callback via qemu_balloon_status() and
indirectly through qemu_balloon_event to virtio_balloon_to_target().
virtio_balloon_to_target() executes its balloon stats half. It
stores the callback in the device state.
If it can't send a stats request, it resets stats and calls the
callback right away.
Else, it sends a stats request. The device model runs the callback
when it receives the answer.
Works.
2. do_balloon() is a cmd_async() method. It receives a callback with
argument, to be called when the command completes. do_balloon()
calls it right before it succeeds. Odd, but should work.
Nevertheless, it passes the callback on via qemu_ballon() and
indirectly through qemu_balloon_event to virtio_balloon_to_target().
a. If the argument is non-zero, virtio_balloon_to_target() executes
its balloon half, which doesn't use the callback in any way.
Odd, but works.
b. If the argument is zero, virtio_balloon_to_target() executes its
balloon stats half, just like in 1. It either calls the callback
right away, or arranges for it to be called later.
Thus, the callback runs twice: use after free and double free.
Test case: start with -S -device virtio-balloon, execute "balloon 0" in
human monitor. Runs the callback first from virtio_balloon_to_target(),
then again from do_balloon().
Reported-by: Mike Cao <bcao@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Separate out the code to retrieve balloon info from the code that sets
balloon values.
This will be used to separate the two callbacks from balloon.c and help
cope with 'balloon 0' on the monitor. Currently, 'balloon 0' causes a
segfault in monitor_resume().
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Replace:
if (foo) {
...
} else {
return 0;
}
by
if (!foo) {
return 0;
}
...
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
balloon.h had function declarations for a couple of functions that are
local to balloon.c. Make them static.
Drop the 'qemu_' prefix for balloon.c-local variables, and make them
static.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
If we're already in a coroutine, there is no reason to use the synchronous
version of block layer functions when a coroutine one exists. This makes
bdrv_read/write/flush use bdrv_co_* when used inside a coroutine.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The QEMU keyboard and mouse reports themselves as full speed devices,
though they are actually low speed devices. Until this is fixed, claim that
we are supporting full speed devices.
Acked-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
In order to be able to transparently replace bdrv_read calls by bdrv_co_read,
reading beyond EOF must produce zeros instead of short reads for AIO, too.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Now that AsyncContexts don't exist any more, we can use one global bottom half
for restarting coroutines instead of allocating a new one every time (before
removing AsyncContexts, the problem with having a global BH was that it had to
belong to a single AsyncContexts and wouldn't be executed in a different one -
which leads to deadlocks)
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The purpose of AsyncContexts was to protect qcow and qcow2 against reentrancy
during an emulated bdrv_read/write (which includes a qemu_aio_wait() call and
can run AIO callbacks of different requests if it weren't for AsyncContexts).
Now both qcow and qcow2 are protected by CoMutexes and AsyncContexts can be
removed.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The old qcow format is another user of the AsyncContext infrastructure.
Converting it to coroutines (and therefore CoMutexes) allows to remove
AsyncContexts.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
In order to be able to call bdrv_co_readv/writev for drivers that don't
implement the functions natively, add an emulation that uses the AIO functions
to implement them.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Use the bdrv_co_readv/writev callbacks to implement bdrv_aio_readv/writev and
bdrv_read/write if a driver provides the coroutine version instead of the
synchronous or AIO version.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Add new block driver callbacks bdrv_co_readv/writev, which work on a
QEMUIOVector like bdrv_aio_*, but don't need a callback. The function may only
be called inside a coroutine, so a block driver implementing this interface can
yield instead of blocking during I/O.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Add a microbenchmark for coroutine create, enter, and return (aka
lifecycle). This is a useful benchmark because users are expected to
create many coroutines, one per I/O request for example, and we
therefore need to provide good performance in that scenario.
To run:
make test-coroutine
./test-coroutine --benchmark-lifecycle 20000000
This will do 20,000,000 coroutine create, enter, return iterations and
print the resulting time.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
To run automated tests for coroutines:
make test-coroutine
./test-coroutine
On success the program terminates with exit status 0. On failure an
error message is written to stderr and the program exits with exit
status 1.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Commit b14ef7c9ab
introduced cpu_unassigned_access() function. On Sparc,
the function does not restore AREG0 used for global CPUState
on function exit, causing bugs with non-faulting unassigned
memory accesses. Alpha, Microblaze and MIPS are not affected.
Fix by restoring AREG0 on exit. Remove excess saving by
do_unassigned_access() functions.
Also ignore unassigned accesses outside of CPU context.
Reported-by: Bob Breuer <breuerr@mc.net>
Tested-by: Bob Breuer <breuerr@mc.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
On platforms that don't support makecontext(3) use gthread based
coroutine implementation.
Darwin has makecontext(3) but getcontext(3) is stubbed out to return
ENOTSUP. Andreas Färber <andreas.faerber@web.de> debugged this and
contributed the ./configure test which solves the issue for Darwin/ppc64
(and ppc) v10.5.
[Original patch by Aneesh, made consistent with coroutine-ucontext.c and
switched to GStaticPrivate by Stefan. Tested on Linux and OpenBSD.]
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Asynchronous code is becoming very complex. At the same time
synchronous code is growing because it is convenient to write.
Sometimes duplicate code paths are even added, one synchronous and the
other asynchronous. This patch introduces coroutines which allow code
that looks synchronous but is asynchronous under the covers.
A coroutine has its own stack and is therefore able to preserve state
across blocking operations, which traditionally require callback
functions and manual marshalling of parameters.
Creating and starting a coroutine is easy:
coroutine = qemu_coroutine_create(my_coroutine);
qemu_coroutine_enter(coroutine, my_data);
The coroutine then executes until it returns or yields:
void coroutine_fn my_coroutine(void *opaque) {
MyData *my_data = opaque;
/* do some work */
qemu_coroutine_yield();
/* do some more work */
}
Yielding switches control back to the caller of qemu_coroutine_enter().
This is typically used to switch back to the main thread's event loop
after issuing an asynchronous I/O request. The request callback will
then invoke qemu_coroutine_enter() once more to switch back to the
coroutine.
Note that if coroutines are used only from threads which hold the global
mutex they will never execute concurrently. This makes programming with
coroutines easier than with threads. Race conditions cannot occur since
only one coroutine may be active at any time. Other coroutines can only
run across yield.
This coroutines implementation is based on the gtk-vnc implementation
written by Anthony Liguori <anthony@codemonkey.ws> but it has been
significantly rewritten by Kevin Wolf <kwolf@redhat.com> to use
setjmp()/longjmp() instead of the more expensive swapcontext() and by
Paolo Bonzini <pbonzini@redhat.com> for Windows Fibers support.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Instead of using its own definitions scsi-disk should
be using the device type of the parent device.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
SET_WINDOW command is vendor-specific only.
So we shouldn't try to emulate it.
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
VHD files technically can be up to 2Tb, but virtual pc is limited
to 127G. Currently qemu-img refused to create vpc files > 127G,
but it is failing to return error when converting from a non-vpc
VHD file which is >127G. It returns success, but creates a truncated
converted image. Also, qemu-img info claims the vpc file is 127G
(and clean).
This patch detects a too-large vpc file and returns -EFBIG. Without
this patch,
=============================================================
root@ip-10-38-123-242:~/qemu-fixed# qemu-img info /mnt/140g-dynamic.vhd
image: /mnt/140g-dynamic.vhd
file format: vpc
virtual size: 127G (136899993600 bytes)
disk size: 284K
root@ip-10-38-123-242:~/qemu-fixed# qemu-img convert -f vpc -O raw /mnt/140g-dynamic.vhd /mnt/y
root@ip-10-38-123-242:~/qemu-fixed# echo $?
0
root@ip-10-38-123-242:~/qemu-fixed# qemu-img info /mnt/y
image: /mnt/y
file format: raw
virtual size: 127G (136899993600 bytes)
disk size: 0
=============================================================
(The 140G image was truncated with no warning or error.)
With the patch, I get:
=============================================================
root@ip-10-38-123-242:~/qemu-fixed# ./qemu-img info /mnt/140g-dynamic.vhd
qemu-img: Could not open '/mnt/140g-dynamic.vhd': File too large
root@ip-10-38-123-242:~/qemu-fixed# ./qemu-img convert -f vpc -O raw /mnt/140g-dynamic.vhd /mnt/y
qemu-img: Could not open '/mnt/140g-dynamic.vhd': File too large
qemu-img: Could not open '/mnt/140g-dynamic.vhd'
=============================================================
See https://bugs.launchpad.net/qemu/+bug/814222 for details.
Signed-off-by: Serge Hallyn <serge.hallyn@canonical.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Commit aea2a33c made bdrv_eject() obey the locked flag. Correct for
medium eject (eject_flag set), incorrect for medium load (eject_flag
clear). See MMC-5 Table 341 "Actions for Lock/Unlock/Eject".
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Callees always return 0, except for FreeBSD's cdrom_eject(), which
returns -ENOTSUP when the device is in a terminally wedged state.
The only caller is bdrv_eject(), and it maps -ENOTSUP to 0 since
commit 4be9762a.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The only caller is bdrv_set_locked(), and it ignores the value.
Callees always return 0, except for FreeBSD's cdrom_set_locked(),
which returns -ENOTSUP when the device is in a terminally wedged
state.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
It's been disabled since the start (commit 19cb3738, Aug 2006), and
has been untouched except for spelling fixes and such. I don't feel
like dragging it along any further.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
BlockDriverState members change_cb and change_opaque are initially
null. The device model may set them, with bdrv_set_change_cb(). If
the device model gets detached (hot unplug), they're left dangling.
Only safe because device hot unplug automatically destroys the
BlockDriverState. But that's a questionable feature, best not to rely
on it.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Ejecting hard disk platters can only end in tears.
If you need to revoke access to an image, use drive_del, not eject -f.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Fixes emulation with io-thread.
Signed-off-by: Peter A. G. Crosthwaite <peter.crosthwaite@petalogix.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
Anthony Liguori