There is no reason to require a minimum size of 16 MB for the vram.
Lower the limit to 4096 (one page). Make it disapper completely would
break guests.
We used to assure the guest surfaces were saved before migration by
setting the whole vram dirty. This patch sets dirty only the areas
that are actually used in the vram.
Signed-off-by: Yonit Halperin <yhalperi@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Replace device_init() with generalized type_init().
While at it, unify naming convention: type_init([$prefix_]register_types)
Also, type_init() is a function, so add preceding blank line where
necessary and don't put a semicolon after the closing brace.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Cc: Anthony Liguori <anthony@codemonkey.ws>
Cc: malc <av1474@comtv.ru>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This was done in a mostly automated fashion. I did it in three steps and then
rebased it into a single step which avoids repeatedly touching every file in
the tree.
The first step was a sed-based addition of the parent type to the subclass
registration functions.
The second step was another sed-based removal of subclass registration functions
while also adding virtual functions from the base class into a class_init
function as appropriate.
Finally, a python script was used to convert the DeviceInfo structures and
qdev_register_subclass functions to TypeInfo structures, class_init functions,
and type_register_static calls.
We are almost fully converted to QOM after this commit.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
These are various small stylistic changes which help make things more
consistent such that the automated conversion script can be simpler.
It's not necessary to agree or disagree with these style changes because all
of this code is going to be rewritten by the patch monkey script anyway.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Instead of each target knowing or guessing the guest page size,
just pass the desired size of dirtied memory area.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Currently creating a memory region automatically registers it for
live migration. This differs from other state (which is enumerated
in a VMStateDescription structure) and ties the live migration code
into the memory core.
Decouple the two by introducing a separate API, vmstate_register_ram(),
for registering a RAM block for migration. Currently the same
implementation is reused, but later it can be moved into a separate list,
and registrations can be moved to VMStateDescription blocks.
Signed-off-by: Avi Kivity <avi@redhat.com>
Commit 0a039dc700 broke vga modes for
qxl-vga by loosing vga_ioport_read windup. qxl needs to hook into
vga port writes only and used to realize that by letting vga_init() do
the work for both reads and writes, then overwrite the write function.
That little detail was missed while doing the conversion ...
This patch fixes it. It also switch qxl vga ioport registration to
portio lists while being at it.
Cc: Hans de Goede <hdegoede@redhat.com>
Acked-by: Alon Levy <alevy@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
RHBZ 740547
If we migrate when the device is in vga state the guest
still believes the slots are created, and will cause operations
that reference the slots, causing a "panic: virtual address out of range"
on the first of them. Easy to see by migrating in vga mode with
a driver loaded, for instance windows cmd window in full screen mode,
and then exiting vga mode back to native mode will cause said panic.
Fixed by doing the slot recreation in post_load for vga mode as well.
Note that compat does not require any changes because it creates it's
only slot by a side effect of QXL_IO_SET_MODE.
Signed-off-by: Alon Levy <alevy@redhat.com>
update init_qxl_ram to reset update_surface to 0. This fixes one case
of breakage when installing an old driver in a vm that had a new driver
installed. The newer driver would know about surface creation and would
change update_surface to !=0, then a reset would happen, all surfaces
are destroyed, then the old driver is initialized and issues an
UPDATE_AREA, and spice server aborts on invalid surface.
RHBZ: 690427
Signed-off-by: Alon Levy <alevy@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
(1) If the guest cursor command is empty, don't reload it after migration.
(2) Cleaning the guest cursor when it is released by
the spice server. In addition, explicitly reset the
cursor in spice upon destroying the primary surface
(was done by spice-server implicitly). This will prevent
access to pci memory that was released.
RHBZ: 744518
Signed-off-by: Yonit Halperin <yhalperi@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
This was only a best-effort attempt, by far not guaranteed to have an
effect. Drop it so that also no direct pthread usage remain in the
device model.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Use QEMU thread API instead of pthread directly. We still need to get
rid of pthread_yield, though, to drop pthread.h inclusion.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
[jan: fix cut'n'paste errors]
[avi: adjust pci variants not to use isa functions]
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Today, when notifying a VM state change with vm_state_notify(),
we pass a VMSTOP macro as the 'reason' argument. This is not ideal
because the VMSTOP macros tell why qemu stopped and not exactly
what the current VM state is.
One example to demonstrate this problem is that vm_start() calls
vm_state_notify() with reason=0, which turns out to be VMSTOP_USER.
This commit fixes that by replacing the VMSTOP macros with a proper
state type called RunState.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
if qxl_send_events was called from spice server context, and then
migration had completed before a call to pipe_read, the target
guest qxl driver didn't get the interrupt. In addition,
qxl_send_events ignored further interrupts of the same kind, since
ram->int_pending was set. As a result, the guest driver was stacked
or very slow (when the waiting for the interrupt was with timeout).
Signed-off-by: Yonit Halperin <yhalperi@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Fix format string errors causing compile failure on 32 bit hosts
when spice is enabled.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Enabling the I/O thread by default seems like an important part of declaring
1.0. Besides allowing true SMP support with KVM, the I/O thread means that the
TCG VCPU doesn't have to multiplex itself with the I/O dispatch routines which
currently requires a (racey) signal based alarm system.
I know there have been concerns about performance. I think so far the ones that
have come up (virtio-net) are most likely due to secondary reasons like
decreased batching.
I think we ought to force enabling I/O thread early in 1.0 development and
commit to resolving any lingering issues.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Instead, use the bus accessors, or get the address space directly
from the board constructor.
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Break is only noticable with newer spice-server library (0.8.2 release
or 0.9.0 and newer on master branch).
ioport_write's val was changed from uint32_t to uint64_t, this
broke two printfs. Use PRId64 instead of %d.
Signed-off-by: Alon Levy <alevy@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
same as 8927cfbba2, but for qxl_check_state, that was
triggered by qxl_pre_load (which calls qxl_hard_reset, which calls qxl_soft_reset),
and caused the migration target to crash.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Convert all vga memory to the memory API. Note we need to fall back to
get_system_memory(), since the various buses don't pass the vga window
as a memory region.
We no longer need to sync the dirty bitmap of the cirrus mapped memory
banks, since the memory API takes care of that for us.
[jan: fix vga-pci logging]
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Inform guest drivers about the new features I/O commands we have
now (async commands, S3 support) if building with newer spice, i.e.
if SPICE_INTERFACE_QXL_MINOR >= 1.
sneaked in some 81+ column line spliting.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Alon Levy <alevy@redhat.com>
Add two new IOs.
QXL_IO_FLUSH_SURFACES - equivalent to update area for all surfaces, used
to reduce vmexits from NumSurfaces to 1 on guest S3, S4 and resolution change (windows
driver implementation is such that this is done on each of those occasions).
QXL_IO_FLUSH_RELEASE - used to ensure anything on last_release is put on the release ring
for the client to free.
Signed-off-by: Yonit Halperin <yhalperi@redhat.com>
Signed-off-by: Alon Levy <alevy@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Some of the QXL port i/o commands are waiting for the spice server to
complete certain actions. Add async versions for these commands, so we
don't block the vcpu while the spice server processses the command.
Instead the qxl device will raise an IRQ when done.
The async command processing relies on an added QXLInterface::async_complete
and added QXLWorker::*_async additions, in spice server qxl >= 3.1
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Alon Levy <alevy@redhat.com>
Since the driver is still in operation even after moving to UNDEFINED, i.e.
by destroying primary in any way.
Signed-off-by: Alon Levy <alevy@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Add qxl_guest_bug() function which is supposed to be called in case
sanity checks of guest requests fail. It raises an error IRQ and
logs a message in case guest debugging is enabled.
Make PANIC_ON() abort instead of exit. That macro should be used
for qemu bugs only, any guest-triggerable stuff should use the new
qxl_guest_bug() function instead.
Convert a few easy cases from PANIC_ON() to qxl_guest_bug() to
show intended usage.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Surface tracking needs proper locking since it is used from vcpu and spice
worker threads, add it. Also reset the surface counter when zapping all
surfaces.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Move the wrapper functions which are used by qxl only to qxl.c.
Rename them from qemu_spice_* to qxl_spice_*. Also pass in a
qxl state pointer instead of a SimpleSpiceDisplay pointer.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Factor out SimpleSpiceDisplay initialization into
qemu_spice_display_init_common() and call it from
both qxl.c (for vga mode) and spice-display.c
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Spice worker does no longer process commands when it is stopped.
Otherwise, it might crash during migration when attempting to process
commands while the guest is not completely loaded.
Cc: Alon Levy <alevy@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
report correct mode when in undefined mode.
introduces qxl_mode_to_string(), and uses it in other places too.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
