sdk/emulator/qemu.git
11 years agoq35: use macro for MCFG property name
Michael S. Tsirkin [Mon, 2 Sep 2013 14:59:38 +0000 (17:59 +0300)]
q35: use macro for MCFG property name

Useful to make it accessible through QOM.

Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Tested-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Tested-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopcie_host: expose address format
Michael S. Tsirkin [Tue, 10 Sep 2013 07:15:00 +0000 (10:15 +0300)]
pcie_host: expose address format

Callers pass in the address so it's helpful for
them to be able to decode it.

Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Tested-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Tested-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopcie_host: expose UNMAPPED macro
Michael S. Tsirkin [Tue, 10 Sep 2013 05:43:48 +0000 (08:43 +0300)]
pcie_host: expose UNMAPPED macro

Make it possible to test unmapped status through QMP.

Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Tested-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Tested-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agoloader: support for unmapped ROM blobs
Michael S. Tsirkin [Sun, 18 Aug 2013 14:02:33 +0000 (17:02 +0300)]
loader: support for unmapped ROM blobs

Support ROM blobs not mapped into guest memory:
same as ROM files really but use caller's buffer.

Support invoking callback on access and
return memory pointer making it easier
for caller to update memory if necessary.

Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Tested-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Tested-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agofw_cfg: interface to trigger callback on read
Michael S. Tsirkin [Sun, 1 Sep 2013 14:56:20 +0000 (17:56 +0300)]
fw_cfg: interface to trigger callback on read

Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Tested-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Tested-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci: fix up w64 size calculation helper
Michael S. Tsirkin [Tue, 1 Oct 2013 12:39:13 +0000 (15:39 +0300)]
pci: fix up w64 size calculation helper

BAR base was calculated incorrectly.
Use existing pci_bar_address to get it right.

Tested-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agoqom: add pointer to int property helpers
Michael S. Tsirkin [Sun, 22 Sep 2013 07:10:17 +0000 (10:10 +0300)]
qom: add pointer to int property helpers

Make it easy to add read-only helpers for simple
integer properties in memory.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Tested-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Tested-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agoqom: cleanup struct Error references
Michael S. Tsirkin [Sun, 22 Sep 2013 07:08:14 +0000 (10:08 +0300)]
qom: cleanup struct Error references

now that a typedef for struct Error is available,
use it in qom/object.h to match coding style rules.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Tested-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Tested-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agocleanup object.h: include error.h directly
Igor Mammedov [Tue, 1 Oct 2013 15:44:32 +0000 (17:44 +0200)]
cleanup object.h: include error.h directly

qapi/error.h is simple enough to be included in qom/object.h
direcly and prepares qom/object.h to use Error typedef.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agohw/pci: removed irq field from PCIDevice
Marcel Apfelbaum [Mon, 7 Oct 2013 07:36:41 +0000 (10:36 +0300)]
hw/pci: removed irq field from PCIDevice

Instead of exposing the the irq field,
pci wrappers to qemu_set_irq or qemu_irq_*
can be used.

Signed-off-by: Marcel Apfelbaum <marcel.a@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agohw/pcie: AER and hot-plug events must use device's interrupt
Marcel Apfelbaum [Mon, 7 Oct 2013 07:36:40 +0000 (10:36 +0300)]
hw/pcie: AER and hot-plug events must use device's interrupt

The fields hpev_intx and aer_intx were removed because
both AER and hot-plug events must use device's interrupt.
Assert/deassert interrupts using pci irq wrappers instead.

Signed-off-by: Marcel Apfelbaum <marcel.a@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agohw: set interrupts using pci irq wrappers
Marcel Apfelbaum [Mon, 7 Oct 2013 07:36:39 +0000 (10:36 +0300)]
hw: set interrupts using pci irq wrappers

pci_set_irq and the other pci irq wrappers use
PCI_INTERRUPT_PIN config register to compute device
INTx pin to assert/deassert.

An irq is allocated using pci_allocate_irq wrapper
only if is needed by non pci devices.

Removed irq related fields from state if not used anymore.

Signed-off-by: Marcel Apfelbaum <marcel.a@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agohw/vfio: set interrupts using pci irq wrappers
Marcel Apfelbaum [Mon, 7 Oct 2013 07:36:38 +0000 (10:36 +0300)]
hw/vfio: set interrupts using pci irq wrappers

pci_set_irq and the other pci irq wrappers use
PCI_INTERRUPT_PIN config register to compute device
INTx pin to assert/deassert.

save INTX pin into the config register before calling
pci_set_irq

Signed-off-by: Marcel Apfelbaum <marcel.a@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agohw/vmxnet3: set interrupts using pci irq wrappers
Marcel Apfelbaum [Mon, 7 Oct 2013 07:36:37 +0000 (10:36 +0300)]
hw/vmxnet3: set interrupts using pci irq wrappers

pci_set_irq uses PCI_INTERRUPT_PIN config register
to compute device INTx pin to assert/deassert.

An assert is used to ensure that intx received
from the quest OS corresponds to PCI_INTERRUPT_PIN.

Signed-off-by: Marcel Apfelbaum <marcel.a@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agohw/pci-bridge: set PCI_INTERRUPT_PIN register before shpc init
Marcel Apfelbaum [Mon, 7 Oct 2013 07:36:36 +0000 (10:36 +0300)]
hw/pci-bridge: set PCI_INTERRUPT_PIN register before shpc init

The PCI_INTERRUPT_PIN will be used by shpc init, so
was moved before the call to shpc_init.

Signed-off-by: Marcel Apfelbaum <marcel.a@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agohw/pci: add pci wrappers for allocating and asserting irqs
Marcel Apfelbaum [Mon, 7 Oct 2013 07:36:35 +0000 (10:36 +0300)]
hw/pci: add pci wrappers for allocating and asserting irqs

Interrupt pin is selected and saved into PCI_INTERRUPT_PIN
register during device initialization. Devices should not call
directly qemu_set_irq and specify the INTx pin on each call.

Added pci_* wrappers to replace qemu_set_irq, qemu_irq_raise,
qemu_irq_lower and qemu_irq_pulse, setting the irq
based on PCI_INTERRUPT_PIN.

Added pci_allocate_irq wrapper to be used by devices that
still need PCIDevice infrastructure to assert irqs.

Renamed a static method which was named already pci_set_irq.

Signed-off-by: Marcel Apfelbaum <marcel.a@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agohw/core: Add interface to allocate and free a single IRQ
Marcel Apfelbaum [Mon, 7 Oct 2013 07:36:34 +0000 (10:36 +0300)]
hw/core: Add interface to allocate and free a single IRQ

qemu_allocate_irq returns a single qemu_irq.
The interface allows to specify an interrupt number.

qemu_free_irq frees it.

Signed-off-by: Marcel Apfelbaum <marcel.a@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agohw/pci: partially handle pci master abort
Marcel Apfelbaum [Mon, 16 Sep 2013 08:21:16 +0000 (11:21 +0300)]
hw/pci: partially handle pci master abort

A MemoryRegion with negative priority was created and
it spans over all the pci address space.
It "intercepts" the accesses to unassigned pci
address space and will follow the pci spec:
 1. returns -1 on read
 2. does nothing on write

Note: setting the RECEIVED MASTER ABORT bit in the STATUS register
      of the device that initiated the transaction will be
      implemented in another series

Signed-off-by: Marcel Apfelbaum <marcel.a@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agodocs/memory: Explictly state that MemoryRegion priority is signed
Marcel Apfelbaum [Mon, 16 Sep 2013 08:21:15 +0000 (11:21 +0300)]
docs/memory: Explictly state that MemoryRegion priority is signed

When memory regions overlap, priority can be used to specify
which of them takes priority. By making the priority values signed
rather than unsigned, we make it more convenient to implement
a situation where one "background" region should appear only
where no other region exists: rather than having to explicitly
specify a high priority for all the other regions, we can let them take
the default (zero) priority and specify a negative priority for the
background region.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Marcel Apfelbaum <marcel.a@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agomemory: Change MemoryRegion priorities from unsigned to signed
Marcel Apfelbaum [Mon, 16 Sep 2013 08:21:14 +0000 (11:21 +0300)]
memory: Change MemoryRegion priorities from unsigned to signed

When memory regions overlap, priority can be used to specify
which of them takes priority. By making the priority values signed
rather than unsigned, we make it more convenient to implement
a situation where one "background" region should appear only
where no other region exists: rather than having to explicitly
specify a high priority for all the other regions, we can let them take
the default (zero) priority and specify a negative priority for the
background region.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Marcel Apfelbaum <marcel.a@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agosmbios: Factor out smbios_maybe_add_str()
Markus Armbruster [Fri, 16 Aug 2013 13:18:32 +0000 (15:18 +0200)]
smbios: Factor out smbios_maybe_add_str()

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agosmbios: Make multiple -smbios type= accumulate sanely
Markus Armbruster [Fri, 16 Aug 2013 13:18:31 +0000 (15:18 +0200)]
smbios: Make multiple -smbios type= accumulate sanely

Currently, -smbios type=T,NAME=VAL,... adds one field (T,NAME) with
value VAL to fw_cfg for each unique NAME.  If NAME occurs multiple
times, the last one's VAL is used (before the QemuOpts conversion, the
first one was used).

Multiple -smbios can add multiple fields with the same (T, NAME).
SeaBIOS reads all of them from fw_cfg, but uses only the first field
(T, NAME).  The others are ignored.

"First one wins, subsequent ones get ignored silently" isn't nice.  We
commonly let the last option win.  Useful, because it lets you
-readconfig first, then selectively override with command line
options.

Clean up -smbios to work the common way.  Accumulate the settings,
with later ones overwriting earlier ones.  Put the result into fw_cfg
(no more useless duplicates).

Bonus cleanup: qemu_uuid_parse() no longer sets SMBIOS system uuid by
side effect.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agosmbios: Improve diagnostics for conflicting entries
Markus Armbruster [Fri, 16 Aug 2013 13:18:30 +0000 (15:18 +0200)]
smbios: Improve diagnostics for conflicting entries

We allow either tables or fields for the same type.  Makes sense,
because SeaBIOS uses fields only when no tables are present.

We do this by searching the SMBIOS blob for a previously added table
or field.  Error messages look like this:

    qemu-system-x86_64: -smbios type=1,serial=42: SMBIOS type 1 table already defined, cannot add field

User needs to know that "table" is defined by -smbios file=..., and
"field" by -smbios type=...

Instead of searching the blob, record additions of interest, and check
that.  Simpler, and makes better error messages possible:

    qemu-system-x86_64: -smbios file=smbios_type_1.bin: Can't mix file= and type= for same type
    qemu-system-x86_64: -smbios type=1,serial=42,serial=99: This is the conflicting setting

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agosmbios: Convert to QemuOpts
Markus Armbruster [Fri, 16 Aug 2013 13:18:29 +0000 (15:18 +0200)]
smbios: Convert to QemuOpts

So that it can be set in config file for -readconfig.

This tightens parsing of -smbios, and makes it more consistent with
other options: unknown parameters are rejected, numbers with trailing
junk are rejected, when a parameter is given multiple times, last
rather than first wins, ...

MST: drop one chunk to fix build errors

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agosmbios: Normalize smbios_entry_add()'s error handling to exit(1)
Markus Armbruster [Fri, 16 Aug 2013 13:18:28 +0000 (15:18 +0200)]
smbios: Normalize smbios_entry_add()'s error handling to exit(1)

It exits on all error conditions but one, where it returns -1.
Normalize, and return void.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agovirtio-net: fix up HMP NIC info string on reset
Michael S. Tsirkin [Tue, 17 Sep 2013 08:45:36 +0000 (11:45 +0300)]
virtio-net: fix up HMP NIC info string on reset

When mac is updated on reset, info string has stale data.
Fix it up.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci: remove explicit check to 64K ioport size
Hervé Poussineau [Fri, 13 Sep 2013 11:58:44 +0000 (13:58 +0200)]
pci: remove explicit check to 64K ioport size

This check is useless, as bigger addresses will be ignored when
added to 'io' MemoryRegion, which has a size of 64K.

However, some architectures don't use the 'io' MemoryRegion, like
the alpha and versatile platforms. They create a PCI I/O region
bigger than 64K, so let them handle PCI I/O BARs in the higher range.

MST: reinstated work-around for BAR sizing.

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopiix4: disable io on reset
Michael S. Tsirkin [Wed, 11 Sep 2013 10:33:31 +0000 (13:33 +0300)]
piix4: disable io on reset

io base register at 0x40 is cleared on reset,
but io is not disabled until some other event
happens to call pm_io_space_update.

Invoke pm_io_space_update directly to make this
consistent.

Cc: qemu-stable@nongnu.org
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopiix: use 64 bit window programmed by guest
Michael S. Tsirkin [Mon, 2 Sep 2013 09:58:57 +0000 (12:58 +0300)]
piix: use 64 bit window programmed by guest

Detect the 64 bit window programmed by firmware
and configure properties accordingly.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agoq35: use 64 bit window programmed by guest
Michael S. Tsirkin [Mon, 2 Sep 2013 09:57:36 +0000 (12:57 +0300)]
q35: use 64 bit window programmed by guest

Detect the 64 bit window programmed by firmware
and configure properties accordingly.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopci: add helper to retrieve the 64-bit range
Michael S. Tsirkin [Mon, 2 Sep 2013 08:37:02 +0000 (11:37 +0300)]
pci: add helper to retrieve the 64-bit range

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agorange: add min/max operations on ranges
Michael S. Tsirkin [Mon, 2 Sep 2013 08:04:39 +0000 (11:04 +0300)]
range: add min/max operations on ranges

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agorange: add Range to typedefs
Michael S. Tsirkin [Mon, 2 Sep 2013 08:41:37 +0000 (11:41 +0300)]
range: add Range to typedefs

will help simplify header dependencies.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agoq35: make pci window address/size match guest cfg
Michael S. Tsirkin [Sun, 1 Sep 2013 10:26:03 +0000 (13:26 +0300)]
q35: make pci window address/size match guest cfg

For Q35, MMCFG address and size are guest configurable.
Update w32 property to make it behave accordingly.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
11 years agoMerge remote-tracking branch 'pmaydell/tags/pull-target-arm-20130910' into staging
Anthony Liguori [Wed, 11 Sep 2013 19:46:52 +0000 (14:46 -0500)]
Merge remote-tracking branch 'pmaydell/tags/pull-target-arm-20130910' into staging

ARM queue:
 * aarch64 preparation patchset (excluding the defconfigs, so this
   doesn't actually enable the new targets yet)
 * minor bugfixes and cleanups
 * disable "-cpu any" in system emulation mode
 * fix ARMv7M stack alignment on reset

# gpg: Signature made Tue 10 Sep 2013 01:46:11 PM CDT using RSA key ID 14360CDE
# gpg: Can't check signature: public key not found

# By Alexander Graf (13) and others
# Via Peter Maydell
* pmaydell/tags/pull-target-arm-20130910: (28 commits)
  configure: Add handling code for AArch64 targets
  linux-user: Add AArch64 support
  linux-user: Allow targets to specify a minimum uname release
  linux-user: Add AArch64 termbits.h definitions
  linux-user: Implement cpu_set_tls() and cpu_clone_regs() for AArch64
  linux-user: Make sure NWFPE code is 32 bit ARM only
  linux-user: Add signal handling for AArch64
  linux-user: Fix up AArch64 syscall handlers
  linux-user: Add syscall number definitions for AArch64
  linux-user: Add cpu loop for AArch64
  linux-user: Don't treat AArch64 cpu names specially
  target-arm: Add AArch64 gdbstub support
  target-arm: Add AArch64 translation stub
  target-arm: Prepare translation for AArch64 code
  target-arm: Disable 32 bit CPUs in 64 bit linux-user builds
  target-arm: Add new AArch64CPUInfo base class and subclasses
  target-arm: Pass DisasContext* to gen_set_pc_im()
  target-arm: Fix target_ulong/uint32_t confusions
  target-arm: Export cpu_env
  target-arm: Extract the disas struct to a header file
  ...

Message-id: 1378839142-7726-1-git-send-email-peter.maydell@linaro.org
Signed-off-by: Anthony Liguori <anthony@codemonkey.ws>
11 years agoMerge remote-tracking branch 'luiz/queue/qmp' into staging
Anthony Liguori [Wed, 11 Sep 2013 19:46:44 +0000 (14:46 -0500)]
Merge remote-tracking branch 'luiz/queue/qmp' into staging

# By Cole Robinson
# Via Luiz Capitulino
* luiz/queue/qmp:
  qapi-types.py: Fix enum struct sizes on i686

Message-id: 1378822364-13887-1-git-send-email-lcapitulino@redhat.com
Signed-off-by: Anthony Liguori <anthony@codemonkey.ws>
11 years agoMerge remote-tracking branch 'spice/spice.v73' into staging
Anthony Liguori [Wed, 11 Sep 2013 19:46:26 +0000 (14:46 -0500)]
Merge remote-tracking branch 'spice/spice.v73' into staging

# By Gerd Hoffmann (2) and Christophe Fergeau (1)
# Via Gerd Hoffmann
* spice/spice.v73:
  qxl: fix local renderer
  qxl: trace io port name
  spice-core: Use g_strdup_printf instead of snprintf

Message-id: 1378807572-27902-1-git-send-email-kraxel@redhat.com
Signed-off-by: Anthony Liguori <anthony@codemonkey.ws>
11 years agoMerge remote-tracking branch 'kraxel/usb.89' into staging
Anthony Liguori [Wed, 11 Sep 2013 19:46:21 +0000 (14:46 -0500)]
Merge remote-tracking branch 'kraxel/usb.89' into staging

# By Gerd Hoffmann (2) and Miroslav Rezanina (2)
# Via Gerd Hoffmann
* kraxel/usb.89:
  ehci: save device pointer in EHCIState
  Remove dev-bluetooth.c dependency from vl.c
  Preparation for usb-bt-dongle conditional build
  usb: sanity check setup_index+setup_len in post_load

Message-id: 1378806073-25197-1-git-send-email-kraxel@redhat.com
Signed-off-by: Anthony Liguori <anthony@codemonkey.ws>
11 years agoMerge remote-tracking branch 'mdroth/qga-pull-2013-9-9' into staging
Anthony Liguori [Wed, 11 Sep 2013 19:46:08 +0000 (14:46 -0500)]
Merge remote-tracking branch 'mdroth/qga-pull-2013-9-9' into staging

# By Tomoki Sekiyama (10) and Paul Burton (1)
# Via Michael Roth
* mdroth/qga-pull-2013-9-9:
  QMP/qemu-ga-client: Make timeout longer for guest-fsfreeze-freeze command
  qemu-ga: Install Windows VSS provider on `qemu-ga -s install'
  qemu-ga: Call Windows VSS requester in fsfreeze command handler
  qemu-ga: Add Windows VSS provider and requester as DLL
  error: Add error_set_win32 and error_setg_win32
  qemu-ga: Add configure options to specify path to Windows/VSS SDK
  Add a script to extract VSS SDK headers on POSIX system
  checkpatch.pl: Check .cpp files
  Add c++ keywords to QAPI helper script
  configure: Support configuring C++ compiler
  mips_malta: support up to 2GiB RAM

Message-id: 1378755701-2051-1-git-send-email-mdroth@linux.vnet.ibm.com
Signed-off-by: Anthony Liguori <anthony@codemonkey.ws>
11 years agoMerge remote-tracking branch 'sstabellini/xen-2013-09-09' into staging
Anthony Liguori [Wed, 11 Sep 2013 19:45:57 +0000 (14:45 -0500)]
Merge remote-tracking branch 'sstabellini/xen-2013-09-09' into staging

# By Anthony PERARD
# Via Stefano Stabellini
* sstabellini/xen-2013-09-09:
  pc_q35: Initialize Xen.
  pc: Initializing ram_memory under Xen.

Message-id: alpine.DEB.2.02.1309091718030.6397@kaball.uk.xensource.com
Signed-off-by: Anthony Liguori <anthony@codemonkey.ws>
11 years agoMerge remote-tracking branch 'stefanha/block' into staging
Anthony Liguori [Wed, 11 Sep 2013 19:45:37 +0000 (14:45 -0500)]
Merge remote-tracking branch 'stefanha/block' into staging

# By Paolo Bonzini (21) and others
# Via Stefan Hajnoczi
* stefanha/block: (42 commits)
  qemu-iotests: Fixed test case 026
  qemu-iotests: Whitespace cleanup
  dataplane: Fix startup race.
  block: look for zero blocks in bs->file
  block: add default get_block_status implementation for protocols
  raw-posix: report unwritten extents as zero
  raw-posix: return get_block_status data and flags
  docs, qapi: document qemu-img map
  qemu-img: add a "map" subcommand
  block: return BDRV_BLOCK_ZERO past end of backing file
  block: use bdrv_has_zero_init to return BDRV_BLOCK_ZERO
  block: return get_block_status data and flags for formats
  block: define get_block_status return value
  block: introduce bdrv_get_block_status API
  block: make bdrv_has_zero_init return false for copy-on-write-images
  qemu-img: always probe the input image for allocated sectors
  block: expect errors from bdrv_co_is_allocated
  block: remove bdrv_is_allocated_above/bdrv_co_is_allocated_above distinction
  block: do not use ->total_sectors in bdrv_co_is_allocated
  block: make bdrv_co_is_allocated static
  ...

Message-id: 1378481953-23099-1-git-send-email-stefanha@redhat.com
Signed-off-by: Anthony Liguori <anthony@codemonkey.ws>
11 years agoMerge remote-tracking branch 'stefanha/net' into staging
Anthony Liguori [Wed, 11 Sep 2013 19:45:18 +0000 (14:45 -0500)]
Merge remote-tracking branch 'stefanha/net' into staging

# By Brad Smith (2) and others
# Via Stefan Hajnoczi
* stefanha/net:
  ne2000: mark I/O as LITTLE_ENDIAN
  vmxnet3: Eliminate __packed redefined warning
  e1000: add interrupt mitigation support
  net: Rename send_queue to incoming_queue
  tap: Use numbered tap/tun devices on all *BSD OS's

Message-id: 1378481624-20964-1-git-send-email-stefanha@redhat.com
Signed-off-by: Anthony Liguori <anthony@codemonkey.ws>
11 years agoconfigure: Add handling code for AArch64 targets
Alexander Graf [Tue, 3 Sep 2013 19:12:22 +0000 (20:12 +0100)]
configure: Add handling code for AArch64 targets

Add the necessary code to configure to handle AArch64 as a target
CPU (we already have some code for supporting it as host). Note
that this doesn't enable the AArch64 targets yet.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-23-git-send-email-peter.maydell@linaro.org
Message-id: 1368505980-17151-12-git-send-email-john.rigby@linaro.org
[PMM:
 * don't need to set TARGET_ABI_DIR to aarch64 as that is the default
 * don't build nwfpe -- this is 32 bit legacy only
 * rewrite commit message
 * add aarch64 to the list of "fdt required" targets
]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agolinux-user: Add AArch64 support
Alexander Graf [Tue, 3 Sep 2013 19:12:21 +0000 (20:12 +0100)]
linux-user: Add AArch64 support

This patch adds support for AArch64 in all the small corners of
linux-user (primarily in image loading and startup code).

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-22-git-send-email-peter.maydell@linaro.org
Message-id: 1368505980-17151-11-git-send-email-john.rigby@linaro.org
[PMM:
 * removed some unnecessary #defines from syscall.h
 * catch attempts to use a 32 bit only cpu with aarch64-linux-user
 * termios stuff moved into its own patch
 * we specify our minimum uname version here now
]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agolinux-user: Allow targets to specify a minimum uname release
Peter Maydell [Tue, 3 Sep 2013 19:12:20 +0000 (20:12 +0100)]
linux-user: Allow targets to specify a minimum uname release

For newer target architectures, glibc can be picky about the kernel
version: for example, it will not run on an aarch64 system unless
the kernel reports itself as at least 3.8.0. Accommodate this by
enhancing the existing support for faking the kernel version so
that each target can optionally specify a minimum version: if
the user doesn't force a specific fake version then we will override
with the minimum required version only if the real host kernel
version is insufficient.

Use this facility to let aarch64 report a minimum of 3.8.0.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-21-git-send-email-peter.maydell@linaro.org

11 years agolinux-user: Add AArch64 termbits.h definitions
Alexander Graf [Tue, 3 Sep 2013 19:12:19 +0000 (20:12 +0100)]
linux-user: Add AArch64 termbits.h definitions

Add the AArch64 termbits.h with all the target's termios related
constants and structures.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-20-git-send-email-peter.maydell@linaro.org
[PMM: split out from another patch]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agolinux-user: Implement cpu_set_tls() and cpu_clone_regs() for AArch64
Alexander Graf [Tue, 3 Sep 2013 19:12:18 +0000 (20:12 +0100)]
linux-user: Implement cpu_set_tls() and cpu_clone_regs() for AArch64

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-19-git-send-email-peter.maydell@linaro.org
[PMM: pulled out from another patch; don't use is_a64() here;
 moved to linux-user from target-arm]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agolinux-user: Make sure NWFPE code is 32 bit ARM only
Peter Maydell [Tue, 3 Sep 2013 19:12:17 +0000 (20:12 +0100)]
linux-user: Make sure NWFPE code is 32 bit ARM only

On ARM, linux-user emulation includes NWFPE support for emulating the
ancient FPA floating point coprocessor. This has long since been
superseded by VFP and is only required for legacy binaries. The
AArch64 linux-user target doesn't compile in NWFPE support, so make
sure the relevant code is protected by suitable ifdefs.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-18-git-send-email-peter.maydell@linaro.org

11 years agolinux-user: Add signal handling for AArch64
Andreas Schwab [Tue, 3 Sep 2013 19:12:16 +0000 (20:12 +0100)]
linux-user: Add signal handling for AArch64

This patch adds signal handling for AArch64. The code is based on the
respective source in the Linux kernel.

Signed-off-by: Andreas Schwab <schwab@suse.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-17-git-send-email-peter.maydell@linaro.org
Message-id: 1368505980-17151-10-git-send-email-john.rigby@linaro.org
[PMM: fixed style nits: tabs, long lines;
 pulled target_signal.h in from a later patch; it fits better here]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agolinux-user: Fix up AArch64 syscall handlers
Alexander Graf [Tue, 3 Sep 2013 19:12:15 +0000 (20:12 +0100)]
linux-user: Fix up AArch64 syscall handlers

Some syscall handlers have special code for ARM enabled that we don't
need on AArch64. Exclude AArch64 in those cases. In other places we
can share struct definitions with other targets or have to provide our
own.

With this patch applied, most syscall definitions in linux-user should
be sound for AArch64.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-16-git-send-email-peter.maydell@linaro.org
Message-id: 1368505980-17151-9-git-send-email-john.rigby@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agolinux-user: Add syscall number definitions for AArch64
Alexander Graf [Tue, 3 Sep 2013 19:12:14 +0000 (20:12 +0100)]
linux-user: Add syscall number definitions for AArch64

The AArch64 syscall definitions are all publicly available in the Linux
kernel. Let's add them to our linux-user emulation target, so that we
can easily handle AArch64 syscalls.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-15-git-send-email-peter.maydell@linaro.org
Message-id: 1368505980-17151-8-git-send-email-john.rigby@linaro.org
[PMM: changes relating to cpu_loop() removed as they are superseded
 by an earlier patch]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agolinux-user: Add cpu loop for AArch64
Peter Maydell [Tue, 3 Sep 2013 19:12:13 +0000 (20:12 +0100)]
linux-user: Add cpu loop for AArch64

Add the main linux-user cpu loop for AArch64. Since AArch64
has a different system call interface, doesn't need to worry
about FPA emulation and may in the future keep the prefetch/data
abort information in different system registers, it's simplest
just to use a completely separate loop from the 32 bit ARM
target, rather than peppering it with ifdefs.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-14-git-send-email-peter.maydell@linaro.org

11 years agolinux-user: Don't treat AArch64 cpu names specially
Alexander Graf [Tue, 3 Sep 2013 19:12:12 +0000 (20:12 +0100)]
linux-user: Don't treat AArch64 cpu names specially

32-bit ARM has a lot of different names for different types of CPUs it supports.
On AArch64, we don't have this, so we really don't want to execute the 32-bit
logic. Stub it out for AArch64 linux-user guests.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-13-git-send-email-peter.maydell@linaro.org
Message-id: 1368505980-17151-7-git-send-email-john.rigby@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agotarget-arm: Add AArch64 gdbstub support
Alexander Graf [Tue, 3 Sep 2013 19:12:11 +0000 (20:12 +0100)]
target-arm: Add AArch64 gdbstub support

We want to be able to debug AArch64 guests. So let's add the respective gdb
stub functions and xml descriptions that allow us to do so.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-12-git-send-email-peter.maydell@linaro.org
Message-id: 1368505980-17151-6-git-send-email-john.rigby@linaro.org
[PMM: dropped unused fp regs XML for now; moved 64 bit only functions
 to new gdbstub64.c; these are hooked up in AArch64CPU, not via
 ifdefs in ARMCPU]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agotarget-arm: Add AArch64 translation stub
Alexander Graf [Tue, 3 Sep 2013 19:12:10 +0000 (20:12 +0100)]
target-arm: Add AArch64 translation stub

We should translate AArch64 mode separately from AArch32 mode. In AArch64 mode,
registers look vastly different, instruction encoding is completely different,
basically the system turns into a different machine.

So let's do a simple if() in translate.c to decide whether we can handle the
current code in the legacy AArch32 code or in the new AArch64 code.

So far, the translation always complains about unallocated instructions. There
is no emulator functionality in this patch!

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-11-git-send-email-peter.maydell@linaro.org
Message-id: 1368505980-17151-5-git-send-email-john.rigby@linaro.org
[PMM:
 * provide no-op versions of a64 functions ifndef TARGET_AARCH64;
   this lets us avoid #ifdefs in translate.c
 * insert the missing call to disas_a64_insn()
 * stash the insn in the DisasContext rather than reloading it in
   real_unallocated_encoding()
]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agotarget-arm: Prepare translation for AArch64 code
Alexander Graf [Tue, 3 Sep 2013 19:12:09 +0000 (20:12 +0100)]
target-arm: Prepare translation for AArch64 code

This patch adds all the prerequisites for AArch64 support that didn't
fit into split up patches. It extends important bits in the core cpu
headers to also take AArch64 mode into account.

Add new ARM_TBFLAG_AARCH64_STATE translation buffer flag
indicate an ARMv8 cpu running in aarch64 mode vs aarch32 mode.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-10-git-send-email-peter.maydell@linaro.org
Message-id: 1368505980-17151-4-git-send-email-john.rigby@linaro.org
[PMM:
 * rearranged tbflags so AArch64? is bit 31 and if it is set then
  30..0 are freely available for whatever makes most sense for that mode
 * added version bump since we change VFP migration state
 * added a comment about how VFP/Neon register state works
 * physical address space is 48 bits, not 64
 * added ARM_FEATURE_AARCH64 flag to identify 64-bit capable CPUs
]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agotarget-arm: Disable 32 bit CPUs in 64 bit linux-user builds
Peter Maydell [Tue, 3 Sep 2013 19:12:08 +0000 (20:12 +0100)]
target-arm: Disable 32 bit CPUs in 64 bit linux-user builds

If we're building aarch64-linux-user then the 32 bit CPUs are
all unwanted, because they can't possibly execute the 64 bit
binaries we will be running; disable them.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-9-git-send-email-peter.maydell@linaro.org

11 years agotarget-arm: Add new AArch64CPUInfo base class and subclasses
Peter Maydell [Tue, 3 Sep 2013 19:12:07 +0000 (20:12 +0100)]
target-arm: Add new AArch64CPUInfo base class and subclasses

Create a new AArch64CPU class; all 64-bit capable ARM
CPUs are subclasses of this. (Currently we only support
one, the "any" CPU used by linux-user.)

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-8-git-send-email-peter.maydell@linaro.org

11 years agotarget-arm: Pass DisasContext* to gen_set_pc_im()
Peter Maydell [Tue, 3 Sep 2013 19:12:06 +0000 (20:12 +0100)]
target-arm: Pass DisasContext* to gen_set_pc_im()

We want gen_set_pc_im() to work for both AArch64 and AArch32, but
to do this we'll need the DisasContext* so we can tell which mode
we're in, so pass it in as a parameter.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-7-git-send-email-peter.maydell@linaro.org

11 years agotarget-arm: Fix target_ulong/uint32_t confusions
Alexander Graf [Tue, 3 Sep 2013 19:12:05 +0000 (20:12 +0100)]
target-arm: Fix target_ulong/uint32_t confusions

Correct a few places that were using uint32_t or a 32 bit
only format string to handle something that should be a target_ulong.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-6-git-send-email-peter.maydell@linaro.org
[PMM: split out to separate patch; added gen_goto_tb() and
gen_set_pc_im() dest params to list of things to change.]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agotarget-arm: Export cpu_env
Alexander Graf [Tue, 3 Sep 2013 19:12:04 +0000 (20:12 +0100)]
target-arm: Export cpu_env

The cpu_env tcg variable will be used by both the AArch32 and AArch64
handling code. Unstaticify it, so that both sides can make use of it.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-5-git-send-email-peter.maydell@linaro.org
Message-id: 1368505980-17151-3-git-send-email-john.rigby@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agotarget-arm: Extract the disas struct to a header file
Alexander Graf [Tue, 3 Sep 2013 19:12:03 +0000 (20:12 +0100)]
target-arm: Extract the disas struct to a header file

We will need to share the disassembly status struct between AArch32 and
AArch64 modes. So put it into a header file that both sides can use.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: John Rigby <john.rigby@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-4-git-send-email-peter.maydell@linaro.org
Message-id: 1368505980-17151-2-git-send-email-john.rigby@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agotarget-arm: Abstract out load/store from a vaddr in AArch32
Peter Maydell [Tue, 3 Sep 2013 19:12:02 +0000 (20:12 +0100)]
target-arm: Abstract out load/store from a vaddr in AArch32

AArch32 code (ie traditional 32 bit world) expects to be
able to pass a vaddr in a TCGv_i32. However when QEMU is
compiled with TARGET_LONG_BITS=32 the TCG load/store
functions take a TCGv_i64. Abstract out load/store with
a 32 bit vaddr so we have a place to put the zero extension
of the vaddr and the extension/truncation of the data value.

Apart from the function definitions most of this patch is
a simple s/tcg_gen_qemu_/gen_aa32_/.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378235544-22290-3-git-send-email-peter.maydell@linaro.org

11 years agoabitypes.h: Remove incorrect ARM ABI_LLONG_ALIGNMENT
Peter Maydell [Tue, 10 Sep 2013 18:09:33 +0000 (19:09 +0100)]
abitypes.h: Remove incorrect ARM ABI_LLONG_ALIGNMENT

The ARM EABI specifies that 64 bit integers should be
8 aligned; remove our incorrect setting of 4 alignment.
This has no actual effect since it only set the alignment
for the 'abi_ullong' and 'abi_llong' types, which are used
only inside code which is MIPS-specific, but it will
avoid problems later if we use the types elsewhere.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agopl110: Clarify comment about PL110 ID on VersatilePB
Peter Maydell [Tue, 10 Sep 2013 18:09:33 +0000 (19:09 +0100)]
pl110: Clarify comment about PL110 ID on VersatilePB

Clarify a comment about the ID register value presented by
the PL110 variant present on the VersatilePB board (based
on testing what the actual hardware does), to indicate that
this is not an error in our emulation, and to remove an #if-0.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agotarget-arm: Implement qmp query-cpu-definitions
Cole Robinson [Tue, 10 Sep 2013 18:09:33 +0000 (19:09 +0100)]
target-arm: Implement qmp query-cpu-definitions

Libvirt uses this to introspect available CPU models.

Signed-off-by: Cole Robinson <crobinso@redhat.com>
Reviewed-by: Andreas Färber <afaerber@suse.de>
Message-id: c0bdcd6c7ea6a085a6902ccaa73180fd771c8267.1378303555.git.crobinso@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agotarget-arm: fix ARMv7M stack alignment on reset
Sebastian Ottlik [Tue, 10 Sep 2013 18:09:32 +0000 (19:09 +0100)]
target-arm: fix ARMv7M stack alignment on reset

When the initial SP is loaded from the vector table on ARMv7M systems the two
least significant bits are ignored as the stack is always aligned at a four byte
boundary (see ARM DDI 0403C, B1.4.1 and B1.5.5). So far QEMU did not ignore
these bits leading to a stack alignment inconsitent with real hardware for
binaries that rely on this behaviour. This patch fixes this issue by masking the
two least significant bits when loading the SP.

Signed-off-by: Sebastian Ottlik <ottlik@fzi.de>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1378286595-27072-1-git-send-email-ottlik@fzi.de
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11 years agotarget-arm: Avoid "1 << 31" undefined behaviour
Peter Maydell [Tue, 10 Sep 2013 18:09:32 +0000 (19:09 +0100)]
target-arm: Avoid "1 << 31" undefined behaviour

Avoid the undefined behaviour of "1 << 31" by using 1U to make
the shift be of an unsigned value rather than shifting into the
sign bit of a signed integer. For consistency, we make all the
CPSR_* constants unsigned, though the only one which triggers
undefined behaviour is CPSR_N.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Message-id: 1378391908-22137-3-git-send-email-peter.maydell@linaro.org

11 years agotarget-arm: Use sextract32() in branch decode
Peter Maydell [Tue, 10 Sep 2013 18:09:32 +0000 (19:09 +0100)]
target-arm: Use sextract32() in branch decode

In the decode of ARM B and BL insns, swap the order of the
"append 2 implicit zeros to imm24" and the sign extend, and
use the new sextract32() utility function to do the latter.
This avoids a direct dependency on the undefined C behaviour
of shifting into the sign bit of an integer.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Message-id: 1378391908-22137-2-git-send-email-peter.maydell@linaro.org

11 years agotarget-arm: Make '-cpu any' available in linux-user mode only
Peter Maydell [Tue, 10 Sep 2013 18:09:32 +0000 (19:09 +0100)]
target-arm: Make '-cpu any' available in linux-user mode only

Make the 'any' CPU for target-arm available only in linux-user mode.
The ARM target provides a CPU named "any", which turns on support for
all user-level instruction set extensions we know about. This is
intended for linux-user emulation mode, where it is the default CPU type.
It makes no sense to try to use this for system emulation, since we don't
initialize it with any system-level information like feature register
values or implementation specific cp15 registers. (Unsurprisingly, some
boards won't boot at all, though you might get lucky in some cases where
the guest doesn't happen to prod things that aren't there.)

Prevent users from making this command line error by removing the
CPU definition from the softmmu build.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Andreas Färber <afaerber@suse.de>
Message-id: 1378213995-12945-1-git-send-email-peter.maydell@linaro.org

11 years agoqapi-types.py: Fix enum struct sizes on i686
Cole Robinson [Sat, 31 Aug 2013 22:36:17 +0000 (18:36 -0400)]
qapi-types.py: Fix enum struct sizes on i686

Unlike other list types, enum wasn't adding any padding, which caused
a mismatch between the generated struct size and GenericList struct
size. More details in a678e26cbe89f7a27cbce794c2c2784571ee9d21

This crashed qemu if calling qmp query-tpm-types for example, which
upsets libvirt capabilities probing. Reproducer on i686:

(sleep 5; printf '{"execute":"qmp_capabilities"}\n{"execute":"query-tpm-types"}\n') | ./i386-softmmu/qemu-system-i386 -S -nodefaults -nographic -M none -qmp stdio

https://bugs.launchpad.net/qemu/+bug/1219207

Cc: qemu-stable@nongnu.org
Signed-off-by: Cole Robinson <crobinso@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Tested-by: Richard W.M. Jones <rjones@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
11 years agoehci: save device pointer in EHCIState
Gerd Hoffmann [Mon, 9 Sep 2013 08:18:17 +0000 (10:18 +0200)]
ehci: save device pointer in EHCIState

We'll need a pointer to the actual pci/sysbus device,
stick a pointer to it into the EHCIState struct.

https://bugzilla.redhat.com/show_bug.cgi?id=1005495

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
11 years agoRemove dev-bluetooth.c dependency from vl.c
Miroslav Rezanina [Tue, 3 Sep 2013 09:23:09 +0000 (11:23 +0200)]
Remove dev-bluetooth.c dependency from vl.c

Use usb_legacy_register handling to create bt-dongle device and remove code
dependency from vl.c so CONFIG_USB_BLUETOOTH can be disabled.

Signed-off-by: Miroslav Rezanina <mrezanin@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
11 years agoPreparation for usb-bt-dongle conditional build
Miroslav Rezanina [Tue, 3 Sep 2013 09:23:08 +0000 (11:23 +0200)]
Preparation for usb-bt-dongle conditional build

To allow disable usb-bt-dongle device using CONFIG_BLUETOOTH option, some of
functions in vl.c file has to be made accessible in dev-bluetooth.c. This is
pure code moving.

Signed-off-by: Miroslav Rezanina <mrezanin@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
11 years agousb: sanity check setup_index+setup_len in post_load
Gerd Hoffmann [Wed, 28 Aug 2013 15:09:30 +0000 (17:09 +0200)]
usb: sanity check setup_index+setup_len in post_load

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
11 years agoqxl: fix local renderer
Gerd Hoffmann [Thu, 5 Sep 2013 19:57:19 +0000 (21:57 +0200)]
qxl: fix local renderer

The local spice renderer assumes the primary surface is located at the
start of the "ram" bar.  This used to be a requirement in qxl hardware
revision 1.  In revision 2+ this is relaxed.  Nevertheless guest drivers
continued to use the traditional location, for historical and backward
compatibility reasons.  The qxl kms driver doesn't though as it depends
on qxl revision 4+ anyway.

Result is that local rendering is hosed for recent linux guests, you'll
get pixel garbage with non-spice ui (gtk, sdl, vnc) and when doing
screendumps.  Fix that by doing a proper mapping of the guest-specified
memory location.

https://bugzilla.redhat.com/show_bug.cgi?id=948717

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
11 years agoqxl: trace io port name
Gerd Hoffmann [Thu, 5 Sep 2013 15:30:05 +0000 (17:30 +0200)]
qxl: trace io port name

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
11 years agospice-core: Use g_strdup_printf instead of snprintf
Christophe Fergeau [Mon, 2 Sep 2013 13:41:32 +0000 (15:41 +0200)]
spice-core: Use g_strdup_printf instead of snprintf

Several places in spice-core.c were using either g_malloc+snprintf
or snprintf+g_strdup to achieve the same result as g_strdup_printf.

Signed-off-by: Christophe Fergeau <cfergeau@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
11 years agoQMP/qemu-ga-client: Make timeout longer for guest-fsfreeze-freeze command
Tomoki Sekiyama [Wed, 7 Aug 2013 15:40:39 +0000 (11:40 -0400)]
QMP/qemu-ga-client: Make timeout longer for guest-fsfreeze-freeze command

guest-fsfreeze-freeze command can take longer than 3 seconds when heavy
disk I/O is running. To avoid unexpected timeout, this changes the timeout
to 60 seconds (timeout of pre-commit phase of VSS).

Signed-off-by: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
11 years agoqemu-ga: Install Windows VSS provider on `qemu-ga -s install'
Tomoki Sekiyama [Wed, 7 Aug 2013 15:40:32 +0000 (11:40 -0400)]
qemu-ga: Install Windows VSS provider on `qemu-ga -s install'

Register QGA VSS provider library into Windows when qemu-ga is installed as
Windows service ('-s install' option). It is deregistered when the service
is uninstalled ('-s uninstall' option).

Signed-off-by: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
11 years agoqemu-ga: Call Windows VSS requester in fsfreeze command handler
Tomoki Sekiyama [Wed, 7 Aug 2013 15:40:25 +0000 (11:40 -0400)]
qemu-ga: Call Windows VSS requester in fsfreeze command handler

Support guest-fsfreeze-freeze and guest-fsfreeze-thaw commands for Windows
guests. When fsfreeze command is issued, it calls the VSS requester to
freeze filesystems and applications. On thaw command, it again tells the VSS
requester to thaw them.

This also adds calling of initialize functions for the VSS requester.

Signed-off-by: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
11 years agoqemu-ga: Add Windows VSS provider and requester as DLL
Tomoki Sekiyama [Wed, 7 Aug 2013 15:40:18 +0000 (11:40 -0400)]
qemu-ga: Add Windows VSS provider and requester as DLL

Adds VSS provider and requester as a qga-vss.dll, which is loaded by
Windows VSS service as well as by qemu-ga.

"provider.cpp" implements a basic stub of a software VSS provider.
Currently, this module only relays a frozen event from VSS service to the
agent, and thaw event from the agent to VSS service, to block VSS process
to keep the system frozen while snapshots are taken at the host.

To register the provider to the guest system as COM+ application, the type
library (.tlb) for qga-vss.dll is required. To build it from COM IDL (.idl),
VisualC++, MIDL and stdole2.tlb in Windows SDK are required. This patch also
adds pre-compiled .tlb file in the repository in order to enable
cross-compile qemu-ga.exe for Windows with VSS support.

"requester.cpp" provides the VSS requester to kick the VSS snapshot process.
Qemu-ga.exe works without the DLL, although fsfreeze features are disabled.

These functions are only supported in Windows 2003 or later. In older
systems, fsfreeze features are disabled.

In several versions of Windows which don't support attribute
VSS_VOLSNAP_ATTR_NO_AUTORECOVERY, DoSnapshotSet fails with error
VSS_E_OBJECT_NOT_FOUND. In this patch, we just ignore this error.
To solve this fundamentally, we need a framework to handle mount writable
snapshot on guests, which is required by VSS auto-recovery feature
(cleanup phase after a snapshot is taken).

Signed-off-by: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
11 years agoerror: Add error_set_win32 and error_setg_win32
Tomoki Sekiyama [Wed, 7 Aug 2013 15:40:11 +0000 (11:40 -0400)]
error: Add error_set_win32 and error_setg_win32

These functions help maintaining homogeneous formatting of error messages
with Windows error code and description (generated by
g_win32_error_message()).

Signed-off-by: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
11 years agoqemu-ga: Add configure options to specify path to Windows/VSS SDK
Tomoki Sekiyama [Wed, 7 Aug 2013 15:40:03 +0000 (11:40 -0400)]
qemu-ga: Add configure options to specify path to Windows/VSS SDK

To enable VSS support in qemu-ga for Windows, header files included in
VSS SDK are required.
The VSS support is enabled by the configure option like below:
  ./configure --with-vss-sdk="/path/to/VSS SDK"

If the path is omitted, it tries to search the headers from default paths
and VSS support is enabled only if the SDK is found.
VSS support is disabled if --without-vss-sdk or --with-vss-sdk=no is
specified.

VSS SDK is available from:
  http://www.microsoft.com/en-us/download/details.aspx?id=23490

To cross-compile using mingw, you need to setup the SDK on Windows
environments to extract headers. You can also extract the SDK headers on
POSIX environments using scripts/extract-vss-headers and msitools.

In addition, --with-win-sdk="/path/to/Windows SDK" option is also added to
specify path to Windows SDK, which may be used for native-compile of .tlb
file of qemu-ga VSS provider. However, this is usually unnecessary because
pre-compiled .tlb file is included.

Signed-off-by: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
11 years agoAdd a script to extract VSS SDK headers on POSIX system
Tomoki Sekiyama [Wed, 7 Aug 2013 15:39:57 +0000 (11:39 -0400)]
Add a script to extract VSS SDK headers on POSIX system

VSS SDK(*) setup.exe is only runnable on Windows. This adds a script
to extract VSS SDK headers on POSIX-systems using msitools.

  * http://www.microsoft.com/en-us/download/details.aspx?id=23490

From: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
11 years agocheckpatch.pl: Check .cpp files
Tomoki Sekiyama [Wed, 7 Aug 2013 15:39:50 +0000 (11:39 -0400)]
checkpatch.pl: Check .cpp files

Enable checkpatch.pl to apply the same checks as C source files for
C++ files with .cpp extensions. It also adds some exceptions for C++
sources to suppress errors for:
  - <> used in C++ template arguments (e.g. template <class T>)
  - :: used to represent namespaces   (e.g. SomeClass::method())
  - : used in class declaration       (e.g. class T : public Super)
  - ~ used in destructor method name  (e.g. T::~T())
  - spacing around 'catch'            (e.g. catch (...))

Signed-off-by: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
11 years agoAdd c++ keywords to QAPI helper script
Tomoki Sekiyama [Wed, 7 Aug 2013 15:39:43 +0000 (11:39 -0400)]
Add c++ keywords to QAPI helper script

Add c++ keywords to avoid errors in compiling with c++ compiler.
This also renames class member of PciDeviceInfo to q_class.

Signed-off-by: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
11 years agoconfigure: Support configuring C++ compiler
Tomoki Sekiyama [Wed, 7 Aug 2013 15:39:36 +0000 (11:39 -0400)]
configure: Support configuring C++ compiler

Add configuration for C++ compiler in configure and Makefiles.
The C++ compiler is choosed as following:
 - ${CXX}, if it is specified.
 - ${cross_prefix}g++, if ${cross_prefix} is specified.
 - Otherwise, c++ is used.

Currently, usage of C++ language is only for access to Windows VSS
using COM+ services in qemu-guest-agent for Windows.

Signed-off-by: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Micael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
11 years agomips_malta: support up to 2GiB RAM
Paul Burton [Fri, 6 Sep 2013 12:57:44 +0000 (13:57 +0100)]
mips_malta: support up to 2GiB RAM

A Malta board can support up to 2GiB of RAM. Since the unmapped kseg0/1
regions are only 512MiB large & the latter 256MiB of those are taken up
by the IO region, access to RAM beyond 256MiB must be done through a
mapped region. In the case of a Linux guest this means we need to use
highmem.

The mainline Linux kernel does not support highmem for Malta at this
time, however this can be tested using the linux-mti-3.8 kernel branch
available from:

  git://git.linux-mips.org/pub/scm/linux-mti.git

You should be able to boot a Linux kernel built from the linux-mti-3.8
branch, with CONFIG_HIGHMEM enabled, using 2GiB RAM by passing "-m 2G"
to QEMU and appending the following kernel parameters:

  mem=256m@0x0 mem=256m@0x90000000 mem=1536m@0x20000000

Note that the upper half of the physical address space of a Malta
mirrors the lower half (hence the 2GiB limit) except that the IO region
(0x10000000-0x1fffffff in the lower half) is not mirrored in the upper
half. That is, physical addresses 0x90000000-0x9fffffff access RAM
rather than the IO region, resulting in a physical address space
resembling the following:

  0x00000000 -> 0x0fffffff  RAM
  0x10000000 -> 0x1fffffff  I/O
  0x20000000 -> 0x7fffffff  RAM
  0x80000000 -> 0x8fffffff  RAM (mirror of 0x00000000 -> 0x0fffffff)
  0x90000000 -> 0x9fffffff  RAM
  0xa0000000 -> 0xffffffff  RAM (mirror of 0x20000000 -> 0x7fffffff)

The second mem parameter provided to the kernel above accesses the
second 256MiB of RAM through the upper half of the physical address
space, making use of the aliasing described above in order to avoid
the IO region and use the whole 2GiB RAM.

The memory setup may be seen as 'backwards' in this commit since the
'real' memory is mapped in the upper half of the physical address space
and the lower half contains the aliases. On real hardware it would be
typical to see the upper half of the physical address space as the alias
since the bus addresses generated match the lower half of the physical
address space. However since the memory accessible in the upper half of
the physical address space is uninterrupted by the IO region it is
easiest to map the RAM as a whole there, and functionally it makes no
difference to the target code.

Due to the requirements of accessing the second 256MiB of RAM through
a mapping to the upper half of the physical address space it is usual
for the bootloader to indicate a maximum of 256MiB memory to a kernel.
This allows kernels which do not support such access to boot on systems
with more than 256MiB of RAM. It is also the behaviour assumed by Linux.
QEMUs small generated bootloader is modified to provide this behaviour.

Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Signed-off-by: Yongbok Kim <yongbok.kim@imgtec.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
11 years agopc_q35: Initialize Xen.
Anthony PERARD [Mon, 9 Sep 2013 16:15:53 +0000 (16:15 +0000)]
pc_q35: Initialize Xen.

Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
11 years agopc: Initializing ram_memory under Xen.
Anthony PERARD [Mon, 9 Sep 2013 16:15:52 +0000 (16:15 +0000)]
pc: Initializing ram_memory under Xen.

Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
CC: qemu-stable@nongnu.org
11 years agone2000: mark I/O as LITTLE_ENDIAN
Aurelien Jarno [Mon, 2 Sep 2013 11:10:34 +0000 (13:10 +0200)]
ne2000: mark I/O as LITTLE_ENDIAN

Now that the memory subsystem is propagating the endianness correctly,
the ne2000 device should have its I/O ports marked as LITTLE_ENDIAN, as
PCI devices are little endian.

This makes the ne2000 NIC to work again on PowerPC.

Cc: qemu-stable@nongnu.org
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agovmxnet3: Eliminate __packed redefined warning
Brad Smith [Fri, 23 Aug 2013 16:28:25 +0000 (12:28 -0400)]
vmxnet3: Eliminate __packed redefined warning

This eliminates a warning about __packed being redefined as exposed by the
vmxnet3 code. __packed is not used anywhere in the vmxnet3 code.

  CC    hw/net/vmxnet3.o
In file included from hw/net/vmxnet3.c:29:
hw/net/vmxnet3.h:37:1: warning: "__packed" redefined
In file included from /usr/include/stdlib.h:38,
                 from /buildbot-qemu/default_openbsd_current/build/include/qemu-common.h:26,
                 from /buildbot-qemu/default_openbsd_current/build/include/hw/hw.h:5,
                 from hw/net/vmxnet3.c:18:
/usr/include/sys/cdefs.h:209:1: warning: this is the location of the previous definition

Signed-off-by: Brad Smith <brad@comstyle.com>
Reviewed-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoe1000: add interrupt mitigation support
Vincenzo Maffione [Fri, 2 Aug 2013 16:30:52 +0000 (18:30 +0200)]
e1000: add interrupt mitigation support

This patch partially implements the e1000 interrupt mitigation mechanisms.
Using a single QEMUTimer, it emulates the ITR register (which is the newer
mitigation register, recommended by Intel) and approximately emulates
RADV and TADV registers. TIDV and RDTR register functionalities are not
emulated (RDTR is only used to validate RADV, according to the e1000 specs).

RADV, TADV, TIDV and RDTR registers make up the older e1000 mitigation
mechanism and would need a timer each to be completely emulated. However,
a single timer has been used in order to reach a good compromise between
emulation accuracy and simplicity/efficiency.

The implemented mechanism can be enabled/disabled specifying the command
line e1000-specific boolean parameter "mitigation", e.g.

    qemu-system-x86_64 -device e1000,mitigation=on,... ...

For more information, see the Software developer's manual at
http://download.intel.com/design/network/manuals/8254x_GBe_SDM.pdf.

Interrupt mitigation boosts performance when the guest suffers from
an high interrupt rate (i.e. receiving short UDP packets at high packet
rate). For some numerical results see the following link
http://info.iet.unipi.it/~luigi/papers/20130520-rizzo-vm.pdf

Signed-off-by: Vincenzo Maffione <v.maffione@gmail.com>
Reviewed-by: Andreas Färber <afaerber@suse.de> (for pc-* machines)
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agonet: Rename send_queue to incoming_queue
Jan Kiszka [Fri, 2 Aug 2013 19:47:08 +0000 (21:47 +0200)]
net: Rename send_queue to incoming_queue

Each networking client has a queue for packets that could not yet be
delivered to that client. Calling this queue "send_queue" is highly
confusing as it has nothing to to with packets send from this client but
to it. Avoid this confusing by renaming it to "incoming_queue".

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agotap: Use numbered tap/tun devices on all *BSD OS's
Brad Smith [Sun, 4 Aug 2013 02:20:41 +0000 (22:20 -0400)]
tap: Use numbered tap/tun devices on all *BSD OS's

The following patch simplifies the *BSD tap/tun code and makes use of numbered
tap/tun interfaces on all *BSD OS's. NetBSD has a patch in their pkgsrc tree
to make use of this feature and DragonFly also supports this as well.

Signed-off-by: Brad Smith <brad@comstyle.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqemu-iotests: Fixed test case 026
Kevin Wolf [Wed, 4 Sep 2013 11:16:05 +0000 (13:16 +0200)]
qemu-iotests: Fixed test case 026

The reference output for test case 026 hasn't been updated in a long
time and it's one of the "known failing" cases. This patch updates the
reference output so that unintentional changes can be reliably detected
again.

The problem with this test case is that it produces different output
depending on whether -nocache is used or not. The solution of this patch
is to actually have two different reference outputs. If nnn.out.nocache
exists, it is used as the reference output for -nocache; otherwise,
nnn.out stays valid for both cases.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoqemu-iotests: Whitespace cleanup
Kevin Wolf [Wed, 4 Sep 2013 11:16:04 +0000 (13:16 +0200)]
qemu-iotests: Whitespace cleanup

These scripts used to have a four characters indentation, with eight
consecutive spaces converted into a tab. Convert everything into spaces.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agodataplane: Fix startup race.
Cornelia Huck [Wed, 4 Sep 2013 12:16:15 +0000 (14:16 +0200)]
dataplane: Fix startup race.

Avoid trying to setup dataplane again if dataplane setup is already in
progress. This may happen if an eventfd is triggered during setup.

I saw this occasionally with an experimental s390 irqfd implementation:

virtio_blk_handle_output
-> virtio_blk_data_plane_start
-> virtio_ccw_set_host_notifier
...
-> virtio_queue_set_host_notifier_fd_handler
-> virtio_queue_host_notifier_read
-> virtio_queue_notify_vq
-> virtio_blk_handle_output
-> virtio_blk_data_plane_start
-> vring_setup
-> hostmem_init
-> memory_listener_register
-> BOOM

As virtio-ccw tries to follow what virtio-pci does, it might be triggerable
for other platforms as well.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11 years agoblock: look for zero blocks in bs->file
Paolo Bonzini [Wed, 4 Sep 2013 17:00:38 +0000 (19:00 +0200)]
block: look for zero blocks in bs->file

Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>