sdk/emulator/qemu.git
11 years agoqcow2: handle_copied(): Get rid of nb_clusters parameter
Kevin Wolf [Tue, 26 Mar 2013 16:50:06 +0000 (17:50 +0100)]
qcow2: handle_copied(): Get rid of nb_clusters parameter

handle_copied() uses its bytes parameter now to determine how many
clusters it should try to find.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqcow2: Factor out handle_copied()
Kevin Wolf [Tue, 26 Mar 2013 16:50:05 +0000 (17:50 +0100)]
qcow2: Factor out handle_copied()

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqcow2: Clean up handle_alloc()
Kevin Wolf [Tue, 26 Mar 2013 16:50:04 +0000 (17:50 +0100)]
qcow2: Clean up handle_alloc()

Things can be simplified a bit now. No semantic changes.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqcow2: Finalise interface of handle_alloc()
Kevin Wolf [Tue, 26 Mar 2013 16:50:03 +0000 (17:50 +0100)]
qcow2: Finalise interface of handle_alloc()

The interface works completely on a byte granularity now and duplicated
parameters are removed.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqcow2: handle_alloc(): Get rid of keep_clusters parameter
Kevin Wolf [Tue, 26 Mar 2013 16:50:02 +0000 (17:50 +0100)]
qcow2: handle_alloc(): Get rid of keep_clusters parameter

handle_alloc() is now called with the offset at which the actual new
allocation starts instead of the offset at which the whole write request
starts, part of which may already be processed.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqcow2: handle_alloc(): Get rid of nb_clusters parameter
Kevin Wolf [Tue, 26 Mar 2013 16:50:01 +0000 (17:50 +0100)]
qcow2: handle_alloc(): Get rid of nb_clusters parameter

We already communicate the same information in *bytes.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqcow2: Factor out handle_alloc()
Kevin Wolf [Tue, 26 Mar 2013 16:50:00 +0000 (17:50 +0100)]
qcow2: Factor out handle_alloc()

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqcow2: Decouple cluster allocation from cluster reuse code
Kevin Wolf [Tue, 26 Mar 2013 16:49:59 +0000 (17:49 +0100)]
qcow2: Decouple cluster allocation from cluster reuse code

This moves some code that prepares the allocation of new clusters to
where the actual allocation happens. This is the minimum required to be
able to move it to a separate function in the next patch.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqcow2: Change handle_dependency to byte granularity
Kevin Wolf [Tue, 26 Mar 2013 16:49:58 +0000 (17:49 +0100)]
qcow2: Change handle_dependency to byte granularity

This is a more precise description of what really constitutes a
dependency. The behaviour doesn't change at this point because the COW
area of the old request is still aligned to cluster boundaries and
therefore an overlap is detected wheneven the requests touch any part of
the same cluster.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqcow2: Improve check for overlapping allocations
Kevin Wolf [Tue, 26 Mar 2013 16:49:57 +0000 (17:49 +0100)]
qcow2: Improve check for overlapping allocations

The old code detected an overlapping allocation even when the
allocations didn't actually overlap, but were only adjacent.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqcow2: Handle dependencies earlier
Kevin Wolf [Tue, 26 Mar 2013 16:49:56 +0000 (17:49 +0100)]
qcow2: Handle dependencies earlier

Handling overlapping allocations isn't just a detail of cluster
allocation. It is rather one of three ways to get the host cluster
offset for a write request:

1. If a request overlaps an in-flight allocations, the cluster offset
   can be taken from there (this is what handle_dependencies will evolve
   into) or the request must just wait until the allocation has
   completed. Accessing the L2 is not valid in this case, it has
   outdated information.

2. Outside overlapping areas, check the clusters that can be written to
   as they are, with no COW involved.

3. If a COW is required, allocate new clusters

Changing the code to reflect this doesn't change the behaviour because
overlaps cannot exist for clusters that are kept in step 2. It does
however make it easier for later patches to work on clusters that belong
to an allocation that is still in flight.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqcow2: Remove bogus unlock of s->lock
Kevin Wolf [Tue, 26 Mar 2013 16:49:55 +0000 (17:49 +0100)]
qcow2: Remove bogus unlock of s->lock

The unlock wakes up the next coroutine, but the currently running
coroutine will lock it again before it yields, so this doesn't make a
lot of sense.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqcow2: Fix "total clusters" number in bdrv_check
Kevin Wolf [Tue, 26 Mar 2013 16:49:54 +0000 (17:49 +0100)]
qcow2: Fix "total clusters" number in bdrv_check

This should be based on the virtual disk size, not on the size of the
image.

Interesting observation: With some VM state stored in the image file,
percentages higher than 100% are possible, even though snapshots
themselves are ignored. This is a qcow2 bug to be fixed another day: The
VM state should be discarded in the active L2 tables after completing
the snapshot creation.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqemu-iotests: More concurrent allocation scenarios
Kevin Wolf [Tue, 26 Mar 2013 16:49:53 +0000 (17:49 +0100)]
qemu-iotests: More concurrent allocation scenarios

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agovl.c: call bdrv_init_with_whitelist() before cmdline parsing
Peter Lieven [Thu, 21 Mar 2013 12:07:10 +0000 (13:07 +0100)]
vl.c: call bdrv_init_with_whitelist() before cmdline parsing

commit 4d454574 "qemu-option: move standard option definitions
out of qemu-config.c" broke support for commandline option
groups that where registered during bdrv_init(). In particular
support for -iscsi options was broken since that commit.

Fix by moving the bdrv_init_with_whitelist() before command
line argument parsing.

Signed-off-by: Peter Lieven <pl@kamp.de>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agospice-qemu-char: Drop hackish vmc_register on spice_chr_write
Hans de Goede [Tue, 26 Mar 2013 10:08:03 +0000 (11:08 +0100)]
spice-qemu-char: Drop hackish vmc_register on spice_chr_write

Now that the core takes care of fe_open tracking we no longer need this hack.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Message-id: 1364292483-16564-12-git-send-email-hdegoede@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agovirtio-serial: propagate guest_connected to the port on post_load
Alon Levy [Tue, 26 Mar 2013 10:08:02 +0000 (11:08 +0100)]
virtio-serial: propagate guest_connected to the port on post_load

When migrating a host with with a spice agent running the mouse becomes
non operational after the migration due to the agent state being
inconsistent between the guest and the client.

After migration the spicevmc backend on the destination has never been notified
of the (non 0) guest_connected state. Virtio-serial holds this state
information and migrates it, this patch properly propagates this information
to virtio-console and through that to interested chardev backends.

rhbz #725965

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Message-id: 1364292483-16564-11-git-send-email-hdegoede@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agovirtio-serial: Consolidate guest_open/guest_close into set_guest_connected
Hans de Goede [Tue, 26 Mar 2013 10:08:01 +0000 (11:08 +0100)]
virtio-serial: Consolidate guest_open/guest_close into set_guest_connected

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Message-id: 1364292483-16564-10-git-send-email-hdegoede@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agoqemu-char: add_handlers: Don't re-send the be_open event on unregister
Hans de Goede [Tue, 26 Mar 2013 10:08:00 +0000 (11:08 +0100)]
qemu-char: add_handlers: Don't re-send the be_open event on unregister

Resending the be_open event only is useful when a frontend is registering, not
when it is unregistering.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Message-id: 1364292483-16564-9-git-send-email-hdegoede@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agoqemu-char: Move incrementing of avail_connections to qdev-properties-system
Hans de Goede [Tue, 26 Mar 2013 10:07:59 +0000 (11:07 +0100)]
qemu-char: Move incrementing of avail_connections to qdev-properties-system

The decrement of avail_connections is done in qdev-properties-system move
the increment there too for proper balancing of the calls.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Message-id: 1364292483-16564-8-git-send-email-hdegoede@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agoqemu-char: Consolidate guest_close/guest_open into a set_fe_open callback
Hans de Goede [Tue, 26 Mar 2013 10:07:58 +0000 (11:07 +0100)]
qemu-char: Consolidate guest_close/guest_open into a set_fe_open callback

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Message-id: 1364292483-16564-7-git-send-email-hdegoede@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agoqemu-char: Cleanup: consolidate fe_open/fe_close into fe_set_open
Hans de Goede [Tue, 26 Mar 2013 10:07:57 +0000 (11:07 +0100)]
qemu-char: Cleanup: consolidate fe_open/fe_close into fe_set_open

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Message-id: 1364292483-16564-6-git-send-email-hdegoede@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agoqemu-char: Automatically do fe_open / fe_close on qemu_chr_add_handlers
Hans de Goede [Tue, 26 Mar 2013 10:07:56 +0000 (11:07 +0100)]
qemu-char: Automatically do fe_open / fe_close on qemu_chr_add_handlers

Most frontends can't really determine if the guest actually has the frontend
side open. So lets automatically generate fe_open / fe_close as soon as a
frontend becomes ready (as signalled by calling qemu_chr_add_handlers) /
becomes non ready (as signalled by setting all handlers to NULL).

And allow frontends which can actually determine if the guest is listening to
opt-out of this.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Message-id: 1364292483-16564-5-git-send-email-hdegoede@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agoqemu-char: Add fe_open tracking
Hans de Goede [Tue, 26 Mar 2013 10:07:55 +0000 (11:07 +0100)]
qemu-char: Add fe_open tracking

Add tracking of the fe_open state to struct CharDriverState.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Message-id: 1364292483-16564-4-git-send-email-hdegoede@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agoqemu-char: Rename qemu_chr_generic_open to qemu_chr_be_generic_open
Hans de Goede [Tue, 26 Mar 2013 10:07:54 +0000 (11:07 +0100)]
qemu-char: Rename qemu_chr_generic_open to qemu_chr_be_generic_open

To better reflect that it is for handling a backend being opened.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Message-id: 1364292483-16564-3-git-send-email-hdegoede@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agoqemu-char: Rename opened to be_open
Hans de Goede [Tue, 26 Mar 2013 10:07:53 +0000 (11:07 +0100)]
qemu-char: Rename opened to be_open

Rename the opened variable to be_open to reflect that it contains the
opened state of the backend.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Message-id: 1364292483-16564-2-git-send-email-hdegoede@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agocompiler: fix warning with GCC 4.8.0
Paolo Bonzini [Wed, 27 Mar 2013 13:34:32 +0000 (14:34 +0100)]
compiler: fix warning with GCC 4.8.0

GCC 4.8.0 introduces a new warning:

    block/qcow2-snapshot.c: In function 'qcow2_write_snapshots’:
    block/qcow2-snapshot.c:252:18: error: typedef 'qemu_build_bug_on__253'
              locally defined but not used [-Werror=unused-local-typedefs]
         QEMU_BUILD_BUG_ON(offsetof(QCowHeader, snapshots_offset) !=
                  ^
    cc1: all warnings being treated as errors

(Caret diagnostics aren't perfect yet with macros... :)) Work around it
with __attribute__((unused)).

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1364391272-1128-1-git-send-email-pbonzini@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agoMerge remote-tracking branch 'mst/tags/for_anthony' into staging
Anthony Liguori [Tue, 26 Mar 2013 21:16:43 +0000 (16:16 -0500)]
Merge remote-tracking branch 'mst/tags/for_anthony' into staging

virtio,pci,qom

Work by Alex to support VGA assignment,
pci and virtio fixes by Stefan, Jason and myself, and a
new qmp event for hotplug support by myself.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
# gpg: Signature made Tue 26 Mar 2013 02:02:24 PM CDT using RSA key ID D28D5469
# gpg: Can't check signature: public key not found

# By Alex Williamson (13) and others
# Via Michael S. Tsirkin
* mst/tags/for_anthony: (23 commits)
  pcie: Add endpoint capability initialization wrapper
  roms: switch oldnoconfig to olddefconfig
  pcie: Mangle types to match topology
  pci: Create and use API to determine root buses
  pci: Create pci_bus_is_express helper
  pci: Q35, Root Ports, and Switches create PCI Express buses
  pci: Allow PCI bus creation interfaces to specify the type of bus
  pci: Move PCI and PCIE type defines
  pci: Create and register a new PCI Express TypeInfo
  exec: assert that RAMBlock size is non-zero
  pci: refuse empty ROM files
  pci_bridge: Remove duplicate IRQ swizzle function
  pci_bridge: Use a default map_irq function
  pci: Fix INTx routing notifier recursion
  pci_bridge: drop formatting from source
  pci_bridge: factor out common code
  pci: Teach PCI Bridges about VGA routing
  pci: Add PCI VGA helpers
  virtio-pci: guest notifier mask without non-irqfd
  virtio-net: remove layout assumptions for mq ctrl
  ...

11 years agopcie: Add endpoint capability initialization wrapper
Alex Williamson [Tue, 19 Mar 2013 18:11:24 +0000 (12:11 -0600)]
pcie: Add endpoint capability initialization wrapper

Fix the awkward API of mangling the caller specified PCIe type and
just provide an interface to initialize an endpoint device.  This
will pick either a regular endpoint or integrated endpoint based on
the bus and return pcie_cap_init to doing exactly what is asked.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agoroms: switch oldnoconfig to olddefconfig
Michael S. Tsirkin [Thu, 21 Mar 2013 12:04:21 +0000 (14:04 +0200)]
roms: switch oldnoconfig to olddefconfig

When a new option is added that qemu does not know
about, the prudent thing is to use the default not
force it to "no".

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopcie: Mangle types to match topology
Alex Williamson [Thu, 14 Mar 2013 22:01:35 +0000 (16:01 -0600)]
pcie: Mangle types to match topology

Windows will fail to start drivers for devices with an Endpoint type
PCIe capability attached to a Root Complex (code 10 - Device cannot
start).  The proper type for such a device is Root Complex Integrated
Endpoint.  Devices don't care which they are, so do this conversion
automatically.

This allows the Windows driver to load for nec-usb-xhci when attached
to pcie.0 of a q35 machine.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci: Create and use API to determine root buses
Alex Williamson [Thu, 14 Mar 2013 22:01:29 +0000 (16:01 -0600)]
pci: Create and use API to determine root buses

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci: Create pci_bus_is_express helper
Alex Williamson [Thu, 14 Mar 2013 22:01:23 +0000 (16:01 -0600)]
pci: Create pci_bus_is_express helper

For testing the bus type.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci: Q35, Root Ports, and Switches create PCI Express buses
Alex Williamson [Thu, 14 Mar 2013 22:01:17 +0000 (16:01 -0600)]
pci: Q35, Root Ports, and Switches create PCI Express buses

Convert q35, ioh3420, xio3130_upstream, and xio3130_downstream to
use the new TYPE_PCIE_BUS.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci: Allow PCI bus creation interfaces to specify the type of bus
Alex Williamson [Thu, 14 Mar 2013 22:01:11 +0000 (16:01 -0600)]
pci: Allow PCI bus creation interfaces to specify the type of bus

No change to any types.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci: Move PCI and PCIE type defines
Alex Williamson [Thu, 14 Mar 2013 22:01:05 +0000 (16:01 -0600)]
pci: Move PCI and PCIE type defines

Move these so that we can reference them from a more common header
instead of including pci_bus.h everywhere.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci: Create and register a new PCI Express TypeInfo
Alex Williamson [Thu, 14 Mar 2013 22:00:59 +0000 (16:00 -0600)]
pci: Create and register a new PCI Express TypeInfo

This will allow us to differentiate Express and Legacy buses.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agoexec: assert that RAMBlock size is non-zero
Stefan Hajnoczi [Mon, 11 Mar 2013 09:20:21 +0000 (10:20 +0100)]
exec: assert that RAMBlock size is non-zero

find_ram_offset() does not handle size=0 gracefully.  It hands out the
same RAMBlock offset multiple times, leading to obscure failures later
on.

Add an assert to warn early if something is incorrectly allocating a
zero size RAMBlock.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci: refuse empty ROM files
Stefan Hajnoczi [Mon, 11 Mar 2013 09:20:20 +0000 (10:20 +0100)]
pci: refuse empty ROM files

A zero size ROM file is invalid and should produce a warning.
Attempting to use a zero size file ends up hitting an assertion
qemu_ram_set_idstr() because RAMBlocks with duplicate addresses are
allocated - due to zero size the allocator doesn't increment the next
available RAMBlock offset.

Also convert __FUNCTION__ to __func__ while we're touching this code.
There are no other __FUNCTION__ instances in pci.c anymore.

Reported-by: Milos Ivanovic <milosivanovic@orcon.net.nz>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci_bridge: Remove duplicate IRQ swizzle function
Alex Williamson [Thu, 7 Mar 2013 23:17:00 +0000 (16:17 -0700)]
pci_bridge: Remove duplicate IRQ swizzle function

pci_bridge_dev_map_irq_fn() is identical to pci_swizzle_map_irq_fn(),
which is now the default for all PCI bridges.  We can therefore remove
this function and the pci_bridge_map_irq() call that used it.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci_bridge: Use a default map_irq function
Alex Williamson [Thu, 7 Mar 2013 23:16:54 +0000 (16:16 -0700)]
pci_bridge: Use a default map_irq function

The PCI bridge spec defines a default swizzle for translating INTx
IRQs from secondary bus to primary.  Use this by default for any
bridge that doesn't set a function.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci: Fix INTx routing notifier recursion
Alex Williamson [Thu, 7 Mar 2013 18:29:19 +0000 (11:29 -0700)]
pci: Fix INTx routing notifier recursion

For some reason we recurse to fire the INTx routing notifier for each
child of a bus, for each possible device of a bus.  That means that if
we add a root port, the notifier gets called for that bridge 256
times.  If we add an upstream switch behind that root port, 256^2.  But
of course we need a downstream switch, 256^3.  This starts to be
noticeable.  Stop the insanity.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci_bridge: drop formatting from source
Michael S. Tsirkin [Mon, 4 Mar 2013 09:31:21 +0000 (11:31 +0200)]
pci_bridge: drop formatting from source

We use the same formatting for all files, it
doesn't make sense to have formatting directives only
in pci bridge header.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci_bridge: factor out common code
Michael S. Tsirkin [Mon, 4 Mar 2013 09:23:49 +0000 (11:23 +0200)]
pci_bridge: factor out common code

Reuse common code in pcie_port, override the hardwired-to-0
bits per PCI Express spec.
No functional change but makes the code easier to follow.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci: Teach PCI Bridges about VGA routing
Alex Williamson [Sun, 3 Mar 2013 17:21:32 +0000 (10:21 -0700)]
pci: Teach PCI Bridges about VGA routing

Each PCI Bridge has a set of implied VGA regions that are enabled when
the VGA bit is set in the bridge control register.  This allows VGA
devices behind bridges.  Unfortunately with VGA Enable, which we
formerly allowed but didn't back, comes along some required VGA
baggage.  VGA Palette Snooping is required, along with VGA 16-bit
decoding.  We don't yet have support for palette snooping.
We also don't have support for 10-bit VGA aliases, the default mode, but
we enable the register, even on root ports, to avoid confusing guests.
Fortunately there's likely nothing from this century that requires these
features, so the missing bits are noted with TODOs.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci: Add PCI VGA helpers
Alex Williamson [Sun, 3 Mar 2013 17:21:26 +0000 (10:21 -0700)]
pci: Add PCI VGA helpers

Allow devices to register VGA memory regions for handling PCI spec
defined VGA I/O port and MMIO areas.  PCI will attach these to the
bus address spaces and enable them according to the device command
register value.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agovirtio-pci: guest notifier mask without non-irqfd
Michael S. Tsirkin [Wed, 13 Mar 2013 19:37:08 +0000 (21:37 +0200)]
virtio-pci: guest notifier mask without non-irqfd

non-irqfd setups are currently broken with vhost:
we start up masked and nothing unmasks the interrupts.
Fix by using mask notifiers, same as the irqfd path.

Sharing irqchip/non irqchip code is always a good thing,
in this case it will help non irqchip benefit
from backend masking optimization.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agovirtio-net: remove layout assumptions for mq ctrl
Jason Wang [Wed, 6 Mar 2013 05:50:27 +0000 (13:50 +0800)]
virtio-net: remove layout assumptions for mq ctrl

Following commit 921ac5d0f3a0df869db5ce4edf752f51d8b1596a (virtio-net:
remove layout assumptions for ctrl vq), this patch makes multiqueue ctrl
handling not rely on the layout of descriptors.

Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agoqmp: add path to device_deleted event
Michael S. Tsirkin [Mon, 11 Mar 2013 13:11:04 +0000 (15:11 +0200)]
qmp: add path to device_deleted event

Add QOM path to device deleted event.  It now becomes useful to report
it for devices which don't have an ID assigned.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agoqom: call class destructor before unparent
Michael S. Tsirkin [Mon, 18 Mar 2013 19:01:37 +0000 (21:01 +0200)]
qom: call class destructor before unparent

It seems more logical to have destruction flow start with the subclass
and move up to the base class.  This ensures object has a valid
canonical path when destructor is called.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agoMerge remote-tracking branch 'quintela/migration.next' into staging
Anthony Liguori [Tue, 26 Mar 2013 18:38:00 +0000 (13:38 -0500)]
Merge remote-tracking branch 'quintela/migration.next' into staging

# By Peter Lieven (9) and others
# Via Juan Quintela
* quintela/migration.next: (22 commits)
  Use qemu_put_buffer_async for guest memory pages
  Add qemu_put_buffer_async
  Use writev ops if available
  Store the data to send also in iovec
  Update bytes_xfer in qemu_put_byte
  Add socket_writev_buffer function
  Add QemuFileWritevBuffer QemuFileOps
  migration: use XBZRLE only after bulk stage
  migration: do not search dirty pages in bulk stage
  migration: do not sent zero pages in bulk stage
  migration: add an indicator for bulk state of ram migration
  migration: search for zero instead of dup pages
  bitops: unroll while loop in find_next_bit()
  buffer_is_zero: use vector optimizations if possible
  cutils: add a function to find non-zero content in a buffer
  move vector definitions to qemu-common.h
  savevm: Fix bugs in the VMSTATE_VBUFFER_MULTIPLY definition
  savevm: Add VMSTATE_STRUCT_VARRAY_POINTER_UINT32
  savevm: Add VMSTATE_FLOAT64 helpers
  savevm: Add VMSTATE_UINTTL_EQUAL helper
  ...

11 years agoMerge remote-tracking branch 'cohuck/virtio-ccw-upstr' into staging
Anthony Liguori [Tue, 26 Mar 2013 18:34:38 +0000 (13:34 -0500)]
Merge remote-tracking branch 'cohuck/virtio-ccw-upstr' into staging

# By Cornelia Huck
# Via Cornelia Huck
* cohuck/virtio-ccw-upstr:
  virtio-ccw: Queue sanity check for notify hypercall.

11 years agovirtio-ccw: Queue sanity check for notify hypercall.
Cornelia Huck [Tue, 26 Mar 2013 16:32:44 +0000 (17:32 +0100)]
virtio-ccw: Queue sanity check for notify hypercall.

Verify that the virtio-ccw notify hypercall passed a reasonable
value for queue.

Cc: qemu-stable@nongnu.org
Reported-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
11 years agoqtest: use synchronous I/O for char device
Anthony Liguori [Tue, 26 Mar 2013 15:04:43 +0000 (10:04 -0500)]
qtest: use synchronous I/O for char device

Peter reported that rtc-test would periodically hang.  It turns out
this was due to an EAGAIN occurring on qemu_chr_fe_write.

Instead of heavily refactoring qtest, just use a synchronous version
of the write operation for qemu_chr_fe_write to address this problem.

Reported-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agochar: introduce a blocking version of qemu_chr_fe_write
Anthony Liguori [Tue, 26 Mar 2013 15:04:17 +0000 (10:04 -0500)]
char: introduce a blocking version of qemu_chr_fe_write

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agohw/qdev: Abort rather than ignoring errors adding device properties
Peter Maydell [Mon, 25 Mar 2013 13:15:14 +0000 (13:15 +0000)]
hw/qdev: Abort rather than ignoring errors adding device properties

Instead of ignoring any errors that occur when adding properties
to a new device in device_initfn(), check for them and abort if any
occur. The most likely cause is accidentally adding a duplicate
property, which is a programming error by the device author.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1364217314-7400-3-git-send-email-peter.maydell@linaro.org
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agoqom: Detect attempts to add a property that already exists
Peter Maydell [Mon, 25 Mar 2013 13:15:13 +0000 (13:15 +0000)]
qom: Detect attempts to add a property that already exists

Detect attempts to add a property to an object if one of
that name already exists, and report them as errors.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1364217314-7400-2-git-send-email-peter.maydell@linaro.org
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agohw/qdev-properties.c: Improve diagnostic for setting property after realize
Peter Maydell [Mon, 25 Mar 2013 13:40:44 +0000 (13:40 +0000)]
hw/qdev-properties.c: Improve diagnostic for setting property after realize

Now we have error_setg() we can improve the error message emitted if
you attempt to set a property of a device after the device is realized
(the previous message was "permission denied" which was not very
informative).

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1364218844-7509-1-git-send-email-peter.maydell@linaro.org
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agovirtio-scsi: cleanup: remove qdev field.
KONRAD Frederic [Thu, 21 Mar 2013 14:15:20 +0000 (15:15 +0100)]
virtio-scsi: cleanup: remove qdev field.

The qdev field is no longer needed. Just drop it.

Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Tested-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Message-id: 1363875320-7985-11-git-send-email-fred.konrad@greensocs.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agovirtio-scsi: cleanup: init and exit functions.
KONRAD Frederic [Thu, 21 Mar 2013 14:15:19 +0000 (15:15 +0100)]
virtio-scsi: cleanup: init and exit functions.

This remove old init and exit function as they are no longer needed.

Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Tested-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Message-id: 1363875320-7985-10-git-send-email-fred.konrad@greensocs.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agovirtio-scsi: cleanup: use QOM casts.
KONRAD Frederic [Thu, 21 Mar 2013 14:15:18 +0000 (15:15 +0100)]
virtio-scsi: cleanup: use QOM casts.

As the virtio-scsi-pci and virtio-scsi-s390 are switched to the new API,
we can use QOM casts.

Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Tested-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Message-id: 1363875320-7985-9-git-send-email-fred.konrad@greensocs.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agovirtio-scsi-ccw: switch to new API
KONRAD Frederic [Thu, 21 Mar 2013 14:15:17 +0000 (15:15 +0100)]
virtio-scsi-ccw: switch to new API

Here the virtio-scsi-ccw is modified for the new API. The device
virtio-scsi-ccw extends virtio-ccw-device as before. It creates and
connects a virtio-scsi during the init. The properties are not modified.

Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Tested-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Message-id: 1363875320-7985-8-git-send-email-fred.konrad@greensocs.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agovirtio-scsi-s390: switch to the new API.
KONRAD Frederic [Thu, 21 Mar 2013 14:15:16 +0000 (15:15 +0100)]
virtio-scsi-s390: switch to the new API.

Here the virtio-scsi-s390 is modified for the new API. The device
virtio-scsi-s390 extends virtio-s390-device as before. It creates and
connects a virtio-scsi during the init. The properties are not modified.

Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Tested-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Message-id: 1363875320-7985-7-git-send-email-fred.konrad@greensocs.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agovirtio-scsi-pci: switch to new API.
KONRAD Frederic [Thu, 21 Mar 2013 14:15:15 +0000 (15:15 +0100)]
virtio-scsi-pci: switch to new API.

Here the virtio-scsi-pci is modified for the new API. The device virtio-scsi-pci
extends virtio-pci. It creates and connects a virtio-scsi during the init.

Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Tested-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Message-id: 1363875320-7985-6-git-send-email-fred.konrad@greensocs.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agovirtio-scsi: add the virtio-scsi device.
KONRAD Frederic [Thu, 21 Mar 2013 14:15:14 +0000 (15:15 +0100)]
virtio-scsi: add the virtio-scsi device.

Create virtio-scsi which extends virtio-device, so it can be connected on
virtio-bus.

Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Tested-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Message-id: 1363875320-7985-5-git-send-email-fred.konrad@greensocs.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agovirtio-scsi: moving host_features from properties to transport properties.
KONRAD Frederic [Thu, 21 Mar 2013 14:15:13 +0000 (15:15 +0100)]
virtio-scsi: moving host_features from properties to transport properties.

host_features field is part of the transport device. So move all the
host_features related properties into transport device.

Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Tested-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Message-id: 1363875320-7985-4-git-send-email-fred.konrad@greensocs.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agovirtio-scsi: allocate cmd_vqs array separately.
KONRAD Frederic [Thu, 21 Mar 2013 14:15:12 +0000 (15:15 +0100)]
virtio-scsi: allocate cmd_vqs array separately.

Allocate/Free the cmd_vqs array separately to have a fixed size device.

Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Tested-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Message-id: 1363875320-7985-3-git-send-email-fred.konrad@greensocs.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agovirtio-scsi: don't use pointer for configuration.
KONRAD Frederic [Thu, 21 Mar 2013 14:15:11 +0000 (15:15 +0100)]
virtio-scsi: don't use pointer for configuration.

The configuration field must not be a pointer as it will be used for virtio-scsi
properties. So *conf is replaced by conf.

Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Tested-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Message-id: 1363875320-7985-2-git-send-email-fred.konrad@greensocs.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agoMerge remote-tracking branch 'luiz/queue/qmp' into staging
Anthony Liguori [Tue, 26 Mar 2013 14:25:45 +0000 (09:25 -0500)]
Merge remote-tracking branch 'luiz/queue/qmp' into staging

# By Corey Bryant (2) and others
# Via Luiz Capitulino
* luiz/queue/qmp:
  New QMP command query-cpu-max and HMP command cpu_max
  qmp: fix handling of boolean values in qmp-shell
  QMP: TPM QMP and man page documentation updates
  QMP: Remove duplicate TPM type from query-tpm

11 years agoUse qemu_put_buffer_async for guest memory pages
Orit Wasserman [Fri, 22 Mar 2013 14:48:03 +0000 (16:48 +0200)]
Use qemu_put_buffer_async for guest memory pages

This will remove an unneeded copy of guest memory pages.
For the page header and device state we still copy the data to the
static buffer the other option is to allocate the memory on demand
which is more expensive.

Signed-off-by: Orit Wasserman <owasserm@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agoAdd qemu_put_buffer_async
Orit Wasserman [Fri, 22 Mar 2013 14:48:02 +0000 (16:48 +0200)]
Add qemu_put_buffer_async

This allows us to add a buffer to the iovec to send without copying it
into the static buffer, the buffer will be sent later when qemu_fflush is called.

Signed-off-by: Orit Wasserman <owasserm@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agoUse writev ops if available
Orit Wasserman [Fri, 22 Mar 2013 14:48:01 +0000 (16:48 +0200)]
Use writev ops if available

Update qemu_fflush and stdio_close to use writev ops if they are available
Use the buffers stored in the iovec.

Signed-off-by: Orit Wasserman <owasserm@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agoStore the data to send also in iovec
Orit Wasserman [Fri, 22 Mar 2013 14:48:00 +0000 (16:48 +0200)]
Store the data to send also in iovec

All data is still copied into the static buffer.
Adjacent iovecs are coalesced so we send one big buffer
instead of many small buffers.

Signed-off-by: Orit Wasserman <owasserm@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agoUpdate bytes_xfer in qemu_put_byte
Orit Wasserman [Fri, 22 Mar 2013 14:47:59 +0000 (16:47 +0200)]
Update bytes_xfer in qemu_put_byte

Signed-off-by: Orit Wasserman <owasserm@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agoAdd socket_writev_buffer function
Orit Wasserman [Fri, 22 Mar 2013 14:47:58 +0000 (16:47 +0200)]
Add socket_writev_buffer function

Signed-off-by: Orit Wasserman <owasserm@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agoAdd QemuFileWritevBuffer QemuFileOps
Orit Wasserman [Fri, 22 Mar 2013 14:47:57 +0000 (16:47 +0200)]
Add QemuFileWritevBuffer QemuFileOps

This will allow us to write an iovec

Signed-off-by: Orit Wasserman <owasserm@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agomigration: use XBZRLE only after bulk stage
Peter Lieven [Tue, 26 Mar 2013 09:58:39 +0000 (10:58 +0100)]
migration: use XBZRLE only after bulk stage

at the beginning of migration all pages are marked dirty and
in the first round a bulk migration of all pages is performed.

currently all these pages are copied to the page cache regardless
of whether they are frequently updated or not. this doesn't make sense
since most of these pages are never transferred again.

this patch changes the XBZRLE transfer to only be used after
the bulk stage has been completed. that means a page is added
to the page cache the second time it is transferred and XBZRLE
can benefit from the third time of transfer.

since the page cache is likely smaller than the number of pages
it's also likely that in the second round the page is missing in the
cache due to collisions in the bulk phase.

on the other hand a lot of unnecessary mallocs, memdups and frees
are saved.

the following results have been taken earlier while executing
the test program from docs/xbzrle.txt. (+) with the patch and (-)
without. (thanks to Eric Blake for reformatting and comments)

+ total time: 22185 milliseconds
- total time: 22410 milliseconds

Shaved 0.3 seconds, better than 1%!

+ downtime: 29 milliseconds
- downtime: 21 milliseconds

Not sure why downtime seemed worse, but probably not the end of the world.

+ transferred ram: 706034 kbytes
- transferred ram: 721318 kbytes

Fewer bytes sent - good.

+ remaining ram: 0 kbytes
- remaining ram: 0 kbytes
+ total ram: 1057216 kbytes
- total ram: 1057216 kbytes
+ duplicate: 108556 pages
- duplicate: 105553 pages
+ normal: 175146 pages
- normal: 179589 pages
+ normal bytes: 700584 kbytes
- normal bytes: 718356 kbytes

Fewer normal bytes...

+ cache size: 67108864 bytes
- cache size: 67108864 bytes
+ xbzrle transferred: 3127 kbytes
- xbzrle transferred: 630 kbytes

...and more compressed pages sent - good.

+ xbzrle pages: 117811 pages
- xbzrle pages: 21527 pages
+ xbzrle cache miss: 18750
- xbzrle cache miss: 179589

And very good improvement on the cache miss rate.

+ xbzrle overflow : 0
- xbzrle overflow : 0

Signed-off-by: Peter Lieven <pl@kamp.de>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agomigration: do not search dirty pages in bulk stage
Peter Lieven [Tue, 26 Mar 2013 09:58:38 +0000 (10:58 +0100)]
migration: do not search dirty pages in bulk stage

avoid searching for dirty pages just increment the
page offset. all pages are dirty anyway.

Signed-off-by: Peter Lieven <pl@kamp.de>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agomigration: do not sent zero pages in bulk stage
Peter Lieven [Tue, 26 Mar 2013 09:58:37 +0000 (10:58 +0100)]
migration: do not sent zero pages in bulk stage

during bulk stage of ram migration if a page is a
zero page do not send it at all.
the memory at the destination reads as zero anyway.

even if there is an madvise with QEMU_MADV_DONTNEED
at the target upon receipt of a zero page I have observed
that the target starts swapping if the memory is overcommitted.
it seems that the pages are dropped asynchronously.

this patch also updates QMP to return the number of
skipped pages in MigrationStats.

Signed-off-by: Peter Lieven <pl@kamp.de>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agomigration: add an indicator for bulk state of ram migration
Peter Lieven [Tue, 26 Mar 2013 09:58:36 +0000 (10:58 +0100)]
migration: add an indicator for bulk state of ram migration

the first round of ram transfer is special since all pages
are dirty and thus all memory pages are transferred to
the target. this patch adds a boolean variable to track
this stage.

Signed-off-by: Peter Lieven <pl@kamp.de>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agomigration: search for zero instead of dup pages
Peter Lieven [Tue, 26 Mar 2013 09:58:35 +0000 (10:58 +0100)]
migration: search for zero instead of dup pages

virtually all dup pages are zero pages. remove
the special is_dup_page() function and use the
optimized buffer_find_nonzero_offset() function
instead.

here buffer_find_nonzero_offset() is used directly
to avoid the unnecssary additional checks in
buffer_is_zero().

raw performace gain checking 1 GByte zeroed memory
over is_dup_page() is approx. 10-12% with SSE2
and 8-10% with unsigned long arithmedtic.

Signed-off-by: Peter Lieven <pl@kamp.de>
Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agobitops: unroll while loop in find_next_bit()
Peter Lieven [Tue, 26 Mar 2013 09:58:34 +0000 (10:58 +0100)]
bitops: unroll while loop in find_next_bit()

this patch adopts the loop unrolling idea of bitmap_is_zero() to
speed up the skipping of large areas with zeros in find_next_bit().

this routine is extensively used to find dirty pages in
live migration.

testing only the find_next_bit performance on a zeroed bitfield
the loop onrolling decreased executing time by approx. 50% on x86_64.

Signed-off-by: Peter Lieven <pl@kamp.de>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agobuffer_is_zero: use vector optimizations if possible
Peter Lieven [Tue, 26 Mar 2013 09:58:33 +0000 (10:58 +0100)]
buffer_is_zero: use vector optimizations if possible

performance gain on SSE2 is approx. 20-25%. altivec
is not tested. performance for unsigned long arithmetic
is unchanged.

Signed-off-by: Peter Lieven <pl@kamp.de>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agocutils: add a function to find non-zero content in a buffer
Peter Lieven [Tue, 26 Mar 2013 09:58:32 +0000 (10:58 +0100)]
cutils: add a function to find non-zero content in a buffer

this adds buffer_find_nonzero_offset() which is a SSE2/Altivec
optimized function that searches for non-zero content in a
buffer.

the function starts full unrolling only after the first few chunks have
been checked one by one. analyzing real memory page data has revealed
that non-zero pages are non-zero within the first 256-512 bits in
most cases. as this function is also heavily used to check for zero memory
pages this tweak has been made to avoid the high setup costs of the fully
unrolled check for non-zero pages.

due to the optimizations used in the function there are restrictions
on buffer address and search length. the function
can_use_buffer_find_nonzero_content() can be used to check if
the function can be used safely.

Signed-off-by: Peter Lieven <pl@kamp.de>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agomove vector definitions to qemu-common.h
Peter Lieven [Tue, 26 Mar 2013 09:58:30 +0000 (10:58 +0100)]
move vector definitions to qemu-common.h

vector optimizations will now be used at various places
not just in is_dup_page() in arch_init.c

Signed-off-by: Peter Lieven <pl@kamp.de>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agosavevm: Fix bugs in the VMSTATE_VBUFFER_MULTIPLY definition
David Gibson [Tue, 12 Mar 2013 03:06:04 +0000 (14:06 +1100)]
savevm: Fix bugs in the VMSTATE_VBUFFER_MULTIPLY definition

The VMSTATE_BUFFER_MULTIPLY macro is misnamed - it actually specifies
a variably sized buffer with VMS_VBUFFER, so should be named
VMSTATE_VBUFFER_MULTIPLY.  This patch fixes this (the macro had no current
users under either name).

In addition, unlike the other VMSTATE_VBUFFER variants, this macro did not
specify VMS_POINTER.  This patch fixes this bug as well.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agosavevm: Add VMSTATE_STRUCT_VARRAY_POINTER_UINT32
David Gibson [Tue, 12 Mar 2013 03:06:03 +0000 (14:06 +1100)]
savevm: Add VMSTATE_STRUCT_VARRAY_POINTER_UINT32

Currently the savevm code contains a VMSTATE_STRUCT_VARRAY_POINTER_INT32
helper (a variably sized array with the number of elements in an int32_t),
but not VMSTATE_STRUCT_VARRAY_POINTER_UINT32 (... with the number of
elements in a uint32_t).  This patch (trivially) fixes the deficiency.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agosavevm: Add VMSTATE_FLOAT64 helpers
David Gibson [Tue, 12 Mar 2013 03:06:02 +0000 (14:06 +1100)]
savevm: Add VMSTATE_FLOAT64 helpers

The current savevm code includes VMSTATE helpers for a number of commonly
used data types, but not for the float64 type used by the internal floating
point emulation code.  This patch fixes the deficiency.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agosavevm: Add VMSTATE_UINTTL_EQUAL helper
David Gibson [Tue, 12 Mar 2013 03:06:01 +0000 (14:06 +1100)]
savevm: Add VMSTATE_UINTTL_EQUAL helper

This adds an _EQUAL VMSTATE helper for target_ulongs, defined in terms of
VMSTATE_UINT32_EQUAL or VMSTATE_UINT64_EQUAL as appropriate.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agosavevm: Add VMSTATE_UINT64_EQUAL helpers
David Gibson [Tue, 12 Mar 2013 03:06:00 +0000 (14:06 +1100)]
savevm: Add VMSTATE_UINT64_EQUAL helpers

The savevm code already includes a number of *_EQUAL helpers which act as
sanity checks verifying that the configuration of the saved state matches
that of the machine we're loading into to work.  Variants already exist
for 8 bit 16 bit and 32 bit integers, but not 64 bit integers.  This patch
fills that hole, adding a UINT64 version.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Juan Quintela <quintela@redhat.com>
11 years agomigration: Improve QMP documentation
Juan Quintela [Mon, 11 Feb 2013 14:11:10 +0000 (15:11 +0100)]
migration: Improve QMP documentation

Signed-off-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
11 years agoNew QMP command query-cpu-max and HMP command cpu_max
Michal Novotny [Mon, 25 Mar 2013 16:31:33 +0000 (17:31 +0100)]
New QMP command query-cpu-max and HMP command cpu_max

These commands return the maximum number of CPUs supported by the
currently running emulator instance, as defined in its QEMUMachine
struct.

Signed-off-by: Michal Novotny <minovotn@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
11 years agoqmp: fix handling of boolean values in qmp-shell
Igor Mammedov [Mon, 25 Mar 2013 14:48:46 +0000 (15:48 +0100)]
qmp: fix handling of boolean values in qmp-shell

qmp-shell converts only integer arguments and the rest
is assumed to be strings which are faithfully sent as
quoted strings by json. But QEMU refuses to accept qmp
command with boolean argument whose value is escaped
as string.

Fix it by special-casing true/false keywords and store
value as corresponding boolean.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
11 years agoQMP: TPM QMP and man page documentation updates
Corey Bryant [Wed, 20 Mar 2013 16:34:49 +0000 (12:34 -0400)]
QMP: TPM QMP and man page documentation updates

Signed-off-by: Corey Bryant <coreyb@linux.vnet.ibm.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
11 years agoQMP: Remove duplicate TPM type from query-tpm
Corey Bryant [Wed, 20 Mar 2013 16:34:48 +0000 (12:34 -0400)]
QMP: Remove duplicate TPM type from query-tpm

Signed-off-by: Corey Bryant <coreyb@linux.vnet.ibm.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
11 years agoMerge remote-tracking branch 'stefanha/net' into staging
Anthony Liguori [Mon, 25 Mar 2013 18:14:26 +0000 (13:14 -0500)]
Merge remote-tracking branch 'stefanha/net' into staging

# By Dmitry Fleytman (5) and others
# Via Stefan Hajnoczi
* stefanha/net:
  net: increase buffer size to accommodate Jumbo frame pkts
  VMXNET3 device implementation
  Packet abstraction for VMWARE network devices
  Common definitions for VMWARE devices
  net: iovec checksum calculator
  Checksum-related utility functions
  net: use socket_set_nodelay() for -netdev socket

11 years agoMerge remote-tracking branch 'stefanha/block' into staging
Anthony Liguori [Mon, 25 Mar 2013 18:14:20 +0000 (13:14 -0500)]
Merge remote-tracking branch 'stefanha/block' into staging

# By Liu Yuan (1) and Stefan Weil (1)
# Via Stefan Hajnoczi
* stefanha/block:
  block: Add options QDict to bdrv_file_open() prototypes (fix MinGW build)
  rbd: fix compile error

11 years agoMerge remote-tracking branch 'kraxel/ipxe.3' into staging
Anthony Liguori [Mon, 25 Mar 2013 18:13:53 +0000 (13:13 -0500)]
Merge remote-tracking branch 'kraxel/ipxe.3' into staging

# By Gerd Hoffmann
# Via Gerd Hoffmann
* kraxel/ipxe.3:
  ipxe: update binaries
  ipxe: disable two second timeout

11 years agoglib: add a compatibility interface for g_timeout_add_seconds
Anthony Liguori [Mon, 25 Mar 2013 15:23:56 +0000 (10:23 -0500)]
glib: add a compatibility interface for g_timeout_add_seconds

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
11 years agogtk: Release modifier when graphic console loses keyboard focus
Jan Kiszka [Sun, 24 Mar 2013 18:10:02 +0000 (19:10 +0100)]
gtk: Release modifier when graphic console loses keyboard focus

This solves, e.g., sticky ALT when selecting a GTK menu, switching to a
different window or selecting a different virtual console.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Message-id: 514F417A.6010908@web.de
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>