sun4m: fix slavio timer RUN/STOP bit

The sun4m architecture has one 'system' timer and one timer per CPU.
The CPU timers can be configured in two modes:

  * 22 bits Counter/Timer. Periodic interrupts.
  * 54 bits User timer. For profiling. In this mode, the Run/Stop bit
    controls the timer.

The run/stop bit controls the timer only when it is in "User" mode, but
its state shall be persistent.

Signed-off-by: Olivier Danet <odanet@caramail.com>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>

sun4m: Set HostID in NVRAM

On SparcStations, the HostID field in the NVRAM is equal to the last
three bytes of the MAC address (which is also stored in the NVRAM).

This constant is used as an identification/serial number on Solaris.

Signed-off-by: Olivier Danet <odanet@caramail.com>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>

Merge remote-tracking branch 'remotes/juanquintela/tags/migration/20140225' into staging

migration/next for 20140225

# gpg: Signature made Tue 25 Feb 2014 14:04:31 GMT using RSA key ID 5872D723
# gpg: Can't check signature: public key not found

* remotes/juanquintela/tags/migration/20140225:
  rdma: rename 'x-rdma' => 'rdma'
  Fix two XBZRLE corruption issues
  Fix vmstate_info_int32_le comparison/assign
  qemu_file: use fwrite() correctly

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/stefanha/tags/net-pull-request' into staging

Net patches

# gpg: Signature made Tue 25 Feb 2014 13:32:33 GMT using RSA key ID 81AB73C8
# gpg: Good signature from "Stefan Hajnoczi <stefanha@redhat.com>"
# gpg:                 aka "Stefan Hajnoczi <stefanha@gmail.com>"
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 8695 A8BF D3F9 7CDA AC35  775A 9CA4 ABB3 81AB 73C8

* remotes/stefanha/tags/net-pull-request:
  virtio-net: use qemu_get_queue() where possible
  vhost_net: use offload API instead of bypassing it
  net: remove implicit peer from offload API
  net: Disable netmap backend when not supported
  net: add offloading support to netmap backend
  net: make tap offloading callbacks static
  net: virtio-net and vmxnet3 use offloading API
  net: TAP uses NetClientInfo offloading callbacks
  net: extend NetClientInfo for offloading
  net: change vnet-hdr TAP prototypes
  opencores_eth: flush queue whenever can_receive can go from false to true

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/kraxel/tags/pull-audio-3' into staging

hda-audio: qom cleanups

# gpg: Signature made Mon 24 Feb 2014 12:19:48 GMT using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>"
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>"
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>"

* remotes/kraxel/tags/pull-audio-3:
  hda-audio: qom cleanups

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

xilinx: Delete hw/include/xilinx.h

This is now obsolete - remove the header and all its inclusions.

Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>

xilinx: Inline usages of xilinx_axi*_init()

Inline the only usage of each of xilinx_axiethernet_init and
xilinx_axidma_init. Converts this init to at least a semi-recent QOM
styling.

Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>

xilinx: Inline usage of xilinx_ethlite_create()

Inline the only usage. Converts this init to at least a semi-recent QOM
styling.

Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>

xilinx: Inline usages of xilinx_timer_create()

Inline these usages. Converts these init to at least a semi-recent QOM
styling.

Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>

xilinx: Inline usages of xilinx_intc_create()

Inline these usages. Converts these init to at least a semi-recent QOM
styling.

Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>

microblaze/ml605: Define macros for irq/memory maps

Define (missing) macros for the interrupt and memory maps for the sake
of self documentation.

Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>

ppc/virtex_ml507: Define macros for irq/memory maps

Define macros for the interrupt and memory maps for the sake of self
documentation.

Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>

microblaze/s3adsp_1800: Define macros for irq map

Define macros for the interrupt map for the sake of self documentation.

Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>

modules: do not include gmodule-2.0 in static builds

gmodule-2.0's pkg-config files include -Wl,--export-dynamic, which breaks
static builds.  It is a glib bug, but we need to support --static builds for
the linux-user targets, and in the end all that is needed to fix this is:

* outlaw --enable-modules --static, which makes little sense anyway

* only include gmodule-2.0's cflags and ldflags if --enable-modules is
specified on the command line.

Reported-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1393346215-5636-1-git-send-email-pbonzini@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/spice/tags/pull-spice-3' into staging

qxl: add sanity check

# gpg: Signature made Mon 24 Feb 2014 12:01:27 GMT using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>"
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>"
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>"

* remotes/spice/tags/pull-spice-3:
  qxl: add sanity check

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

virtio-net: use qemu_get_queue() where possible

qemu_get_queue() is a shorthand for qemu_get_subqueue(n->nic, 0).  Use
the shorthand where possible.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

vhost_net: use offload API instead of bypassing it

There is no need to access backend->info->has_vnet_hdr() and friends
anymore.  Use the qemu_has_vnet_hdr() API instead.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

net: remove implicit peer from offload API

The virtio_net offload APIs are used on the NIC's peer (i.e. the tap
device).  The API was defined to implicitly use nc->peer, saving the
caller the trouble.

This wasn't ideal because:
1. There are callers who have the peer but not the NIC.  Currently they
   are forced to bypass the API and access peer->info->... directly.
2. The rest of the net.h API uses nc, not nc->peer, so it is
   inconsistent.

This patch pushes nc->peer back up to callers.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

rdma: rename 'x-rdma' => 'rdma'

As far as we can tell, all known bugs have been fixed:

1. Parallel migrations are working
2. IPv6 migration is working
3. virt-test is working

I'm not comfortable sending the revised libvirt patch
until this is accepted or review suggestions are addressed,
(including pin-all support. It does not make sense to
remove experimental for one thing and not the other. That's
too many trips through the libvirt community).

Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael R. Hines <mrhines@us.ibm.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

Fix two XBZRLE corruption issues

Push zero'd pages into the XBZRLE cache
    A page that was cached by XBZRLE, zero'd and then XBZRLE'd again
    was being compared against a stale cache value

Don't use 'qemu_put_buffer_async' to put pages from the XBZRLE cache
    Since the cache might change before the data hits the wire

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

Fix vmstate_info_int32_le comparison/assign

Fix comparison of vmstate_info_int32_le so that it succeeds if loaded
value is (l)ess than or (e)qual

When the comparison succeeds, assign the value loaded
  This is a change in behaviour but I think the original intent, since
  the idea is to check if the version/size of the thing you're loading is
  less than some limit, but you might well want to do something based on
  the actual version/size in the file

Fix up comment and name text

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

qemu_file: use fwrite() correctly

fwrite() returns the number of items written.  But when there is one
error, it can return a short write.

In the particular bug that I was tracking, I did a migration to a
read-only filesystem.  And it was able to finish the migration
correctly.  fwrite() never returned a negative error code, nor zero,
always 4096. (migration writes chunks of about 14000 bytes).  And it
was able to "complete" the migration with success (yes, reading the
file was a bit more difficult).

To add insult to injury, if your amount of memory was big enough (12GB
on my case), it overwrote some important structure, and from them,
malloc failed.  This check makes the problem go away.

Signed-off-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

Merge remote-tracking branch 'remotes/mdroth/qga-pull-2014-02-24' into staging

* remotes/mdroth/qga-pull-2014-02-24:
  qemu-ga: isa-serial support on Windows
  qga: Fix memory allocation pasto
  qga: Don't require 'time' argument in guest-set-time command
  qga: vss-win32: Fix interference with snapshot deletion by other VSS request
  qga: vss-win32: Fix interference with snapshot creation by other VSS requesters
  qga: vss-win32: Use NULL as an invalid pointer for OpenEvent and CreateEvent

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/xtensa/tags/20140224-xtensa' into staging

Xtensa fixes and improvements queue 2014-02-24:
- add support for ML605 and KC705 FPGA boards;
- flush opencores_eth queue when new RX descriptor is available;
- add basic checks to cache opcodes;
- make core configuration available to tests;
- implement HW config ID special registers.

# gpg: Signature made Mon 24 Feb 2014 00:52:42 GMT using RSA key ID F83FA044
# gpg: Good signature from "Max Filippov <max.filippov@cogentembedded.com>"
# gpg:                 aka "Max Filippov <jcmvbkbc@gmail.com>"

* remotes/xtensa/tags/20140224-xtensa:
  target-xtensa: provide HW confg ID registers
  target-xtensa: refactor standard core configuration
  target-xtensa: add basic tests for cache opcodes
  target-xtensa: allow using core configuration in tests
  target-xtensa: add overridable test_init macro
  target-xtensa: add basic checks to icache opcodes
  target-xtensa: add basic checks to dcache opcodes
  target-xtensa: add RRRI4 opcode format fields
  opencores_eth: flush queue whenever can_receive can go from false to true
  hw/xtensa: add support for ML605 and KC705 FPGA board

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

net: Disable netmap backend when not supported

This patch fixes configure so that the netmap backend is not compiled in if the
host doesn't support an API version >= 11. A version upper bound (15) has been
added so that the netmap API can be extended with some minor features without
requiring QEMU code modifications.

Moreover, some changes have been done to net/netmap.c in order to reflect the
current netmap API/ABI (11).

The NETMAP_WITH_LIBS macro makes possible to include some utilities (e.g.
netmap ring macros, D(), RD() and other high level functions) through the netmap
headers. In this way we get rid of the D and RD macro definitions in the QEMU
code, and we open the way for further code simplifications that will be
introduced by future patches.

Signed-off-by: Vincenzo Maffione <v.maffione@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

net: add offloading support to netmap backend

Whit this patch, the netmap backend supports TSO/UFO/CSUM
offloadings, and accepts the virtio-net header, similarly to what
happens with TAP. The offloading callbacks in the NetClientInfo
interface have been implemented.

Signed-off-by: Vincenzo Maffione <v.maffione@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

net: make tap offloading callbacks static

Since TAP offloadings are manipulated through a new API, it's
not necessary to export them in include/net/tap.h anymore.

Signed-off-by: Vincenzo Maffione <v.maffione@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

net: virtio-net and vmxnet3 use offloading API

With this patch, virtio-net and vmxnet3 frontends make
use of the qemu_peer_* API for backend offloadings manipulations,
instead of calling TAP-specific functions directly.
We also remove the existing checks which prevent those frontends
from using offloadings with backends different from TAP (e.g. netmap).

Signed-off-by: Vincenzo Maffione <v.maffione@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

net: TAP uses NetClientInfo offloading callbacks

The TAP NetClientInfo structure is inizialized with the TAP-specific
functions that manipulates offloading features.

Signed-off-by: Vincenzo Maffione <v.maffione@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

net: extend NetClientInfo for offloading

Some new callbacks have been added to generalize the operations done
by virtio-net and vmxnet3 frontends to manipulate TAP offloadings.

Signed-off-by: Vincenzo Maffione <v.maffione@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

net: change vnet-hdr TAP prototypes

The tap_has_vnet_hdr() and tap_has_vnet_hdr_len() functions used
to return int, even though they only return true/false values.
This patch changes the prototypes to return bool.

Signed-off-by: Vincenzo Maffione <v.maffione@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

opencores_eth: flush queue whenever can_receive can go from false to true

The following registers control whether MAC can receive frames:
- MODER.RXEN bit that enables/disables receiver;
- TX_BD_NUM register that specifies number of RX descriptors.
Notify QEMU networking core when the MAC is ready to receive frames.
Discard frame and raise BUSY interrupt when the frame arrives but the
current RX descriptor is not empty.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging

Block patches

# gpg: Signature made Fri 21 Feb 2014 21:42:24 GMT using RSA key ID C88F2FD6
# gpg: Good signature from "Kevin Wolf <kwolf@redhat.com>"

* remotes/kevin/tags/for-upstream: (54 commits)
  iotests: Mixed quorum child device specifications
  quorum: Simplify quorum_open()
  quorum: Add unit test.
  quorum: Add quorum_open() and quorum_close().
  quorum: Implement recursive .bdrv_recurse_is_first_non_filter in quorum.
  quorum: Add quorum_co_flush().
  quorum: Add quorum_invalidate_cache().
  quorum: Add quorum_getlength().
  quorum: Add quorum mechanism.
  quorum: Add quorum_aio_readv.
  blkverify: Extract qemu_iovec_clone() and qemu_iovec_compare() from blkverify.
  quorum: Add quorum_aio_writev and its dependencies.
  quorum: Create BDRVQuorumState and BlkDriver and do init.
  quorum: Create quorum.c, add QuorumChildRequest and QuorumAIOCB.
  check-qdict: Test termination of qdict_array_split()
  check-qdict: Adjust test for qdict_array_split()
  qdict: Extract non-QDicts in qdict_array_split()
  qemu-config: Sections must consist of keys
  qemu-iotests: Check qemu-img command line parsing
  qemu-img: Allow -o help with incomplete argument list
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/sstabellini/xen-140220' into staging

* remotes/sstabellini/xen-140220:
  xen_disk: fix io accounting
  Call pci_piix3_xen_ide_unplug from unplug_disks

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/bonzini/configure' into staging

* remotes/bonzini/configure:
  build: softmmu targets do not have a "main.o" file
  configure: Disable libtool if -fPIE does not work with it (bug #1257099)
  block: convert block drivers linked with libs to modules
  Makefile: introduce common-obj-m and block-obj-m for DSO
  Makefile: install modules with "make install"
  module: implement module loading
  rules.mak: introduce DSO rules
  darwin: do not use -mdynamic-no-pic
  block: use per-object cflags and libs
  rules.mak: allow per object cflags and libs
  rules.mak: fix $(obj) to a real relative path
  util: Split out exec_dir from os_find_datadir

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

configure: check that C++ compiler actually works

Check that the C++ compiler works with the C compiler; if it
does not, then don't pass CXX to the build process. This
fixes a regression where QEMU was no longer building if the
build environment didn't have a C++ compiler (introduced
in commit 3144f78b, which incorrectly assumed that rules.mak
would only see a non-empty $(CXX) if configure had actually
found a working C++ compiler).

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reported-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reported-by: Thomas Huth <thuth@linux.vnet.ibm.com>
Tested-by: Thomas Huth <thuth@linux.vnet.ibm.com>
Tested-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Message-id: 1392909016-14028-1-git-send-email-peter.maydell@linaro.org

qxl: add sanity check

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>

hda-audio: qom cleanups

Add HDA_AUDIO type and macro, drop DO_UPCAST().

Had to add a abstract hda audio class as parent
for all hda-* variants to make that fly.  Killed
some init code duplication while being at it.

Cc: Andreas Färber <afaerber@suse.de>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

target-xtensa: provide HW confg ID registers

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>

target-xtensa: refactor standard core configuration

Coalesce all standard configuration sections into single
DEFAULT_SECTIONS macro for all cores. This allows to add new features in
a single place: overlay_tool.h

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>

target-xtensa: add basic tests for cache opcodes

Test that non-locking prefetch operations don't cause exceptions on
missing TLB and that other 'hit' cache operations do.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>

target-xtensa: allow using core configuration in tests

Add path to the core configuration directory to test build command and
replace .include asm directive with #include to enable preprocessing.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>

target-xtensa: add overridable test_init macro

Some test suites, like MMU, need per-test initialization. Don't make them
redefine test macro, add test_init for that purpose.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>

target-xtensa: add basic checks to icache opcodes

Check privilege level for privileged instructions (IHU, III, IIU and IPFL
are privileged), memory accessibility for instructions that reference memory
(IH* and IPFL) and windowed register validity for all instruction cache
instructions.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>

target-xtensa: add basic checks to dcache opcodes

Check privilege level for privileged instructions (DHI, DHU, DII, DIU, DIWB,
DIWBI, DPFL are privileged), memory accessibility for instructions that
reference memory (all DH* and DPFL) and windowed register validity for all
data cache instructions.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>

target-xtensa: add RRRI4 opcode format fields

This encoding is used by cache instructions.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>

opencores_eth: flush queue whenever can_receive can go from false to true

The following registers control whether MAC can receive frames:
- MODER.RXEN bit that enables/disables receiver;
- TX_BD_NUM register that specifies number of RX descriptors.
Notify QEMU networking core when the MAC is ready to receive frames.
Discard frame and raise BUSY interrupt when the frame arrives but the
current RX descriptor is not empty.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>

hw/xtensa: add support for ML605 and KC705 FPGA board

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Reviewed-by: Andreas Färber <afaerber@suse.de>

qemu-ga: isa-serial support on Windows

Add support for isa-serial method for qemu-ga on Windows,
Added -p command line parameter for serial port name
specification, e.g. "-p COM15".

Signed-off-by: Miki Mishael <mmishael@redhat.com>
Signed-off-by: Dmitry Fleytman <dfleytma@redhat.com>
*added default isa-serial path to help output

Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>

qga: Fix memory allocation pasto

qmp_guest_file_seek() allocates memory for a GuestFileRead object
instead of the GuestFileSeek object it actually uses.  Harmless,
because the GuestFileRead is slightly larger.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>

qga: Don't require 'time' argument in guest-set-time command

As the description to the guest-set-time states, the command is
there to ease time synchronization after resume. If guest was
suspended for longer period of time, its system time can go off
so badly, that even NTP refuses to set it. That's why the command
was invented: to give users chance to set the time (not
necessarily 100% correct). However, there's is no real need for
us to require users to pass an arbitrary time. Especially if we
can read the correct value from RTC (boiling down to reading
host's time). Hence this commit enables logic:

guest-set-time() == guest-set-time($now_from_rtc)

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>

qga: vss-win32: Fix interference with snapshot deletion by other VSS request

When a VSS requester such as vshadow.exe or diskshadow.exe requests to
delete snapshots, qemu-ga VSS provider's DeleteSnapshots() is also called
and returns E_NOTIMPL, that makes the deletion fail.
To avoid this issue, return S_OK and set values that represent no snapshots
are deleted by qemu-ga VSS provider.

Signed-off-by: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Reviewed-by: Gal Hammer <ghammer@redhat.com>
Reviewed-by: Yan Vugenfirer <yvugenfi@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>

qga: vss-win32: Fix interference with snapshot creation by other VSS requesters

When a VSS requester such as vshadow.exe or diskshadow.exe requests to
create disk snapshots, Windows may choose qemu-ga VSS provider if it is
only provider registered on the system. However, because it provides only a
function to freeze the filesystem, the snapshotting fails.

This patch adds a check into CQGAVssProvider::IsVolumeSupported() to reject
the request from other VSS requesters, so that the other provider is chosen.

The check of requester is done by confirming event channels between
qemu-ga's requester and provider established. To ensure that the events are
initialized when CQGAVssProvider::IsVolumeSupported() is called, it moves
the initialization earlier.

Signed-off-by: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Reviewed-by: Gal Hammer <ghammer@redhat.com>
Reviewed-by: Yan Vugenfirer <yvugenfi@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>

qga: vss-win32: Use NULL as an invalid pointer for OpenEvent and CreateEvent

OpenEvent and CreateEvent WinAPI return NULL when failed to open/create
events handles, instead of INVALID_HANDLE_VALUE (although their return
types are HANDLE).
This replaces INVALID_HANDLE_VALUE related to event handles with NULL.

Signed-off-by: Tomoki Sekiyama <tomoki.sekiyama@hds.com>
Reviewed-by: Gal Hammer <ghammer@redhat.com>
Reviewed-by: Yan Vugenfirer <yvugenfi@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>

iotests: Mixed quorum child device specifications

Add a test case to test 081 for mixing full option dicts and reference
strings of specifying the quorum child block devices through QMP.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

quorum: Simplify quorum_open()

Although it may not look like it, this patch simplifies quorum_open().
qdict_array_split() is now able to return QLists with different objects
than only QDicts, therefore it will now do all the work and
quorum_open() does not have to handle reference strings by itself.

This allows mixing full option dicts and reference strings for
specifying the child block devices of quorum; furthermore, it improves
handling of malformed specifications.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

quorum: Add unit test.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

quorum: Add quorum_open() and quorum_close().

Example of command line:

-drive if=virtio,driver=quorum,\
children.0.file.filename=1.raw,\
children.0.node-name=1.raw,\
children.0.driver=raw,\
children.1.file.filename=2.raw,\
children.1.node-name=2.raw,\
children.1.driver=raw,\
children.2.file.filename=3.raw,\
children.2.node-name=3.raw,\
children.2.driver=raw,\
vote-threshold=2

blkverify=on with vote-threshold=2 and two files can be passed to
emulate blkverify.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

quorum: Implement recursive .bdrv_recurse_is_first_non_filter in quorum.

This is used to activate quorum snapshot.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

quorum: Add quorum_co_flush().

Makes a vote to select error if any.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

quorum: Add quorum_invalidate_cache().

We really want that live migration works with quorum so implement
quorum_invalidate_cache().

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

quorum: Add quorum_getlength().

Check that every bs file returns the same length.
Otherwise, return -EIO to disable the quorum and
avoid length discrepancy.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

quorum: Add quorum mechanism.

This patchset enables the core of the quorum mechanism.
The num_children reads are compared to get the majority version and if this
version exists more than threshold times the guest won't see the error at all.

If a block is corrupted or if an error occurs during an IO or if the quorum
cannot be established QMP events are used to report to the management.

Use gnutls's SHA-256 to compare versions.

--enable-quorum must be used to enable the feature.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

quorum: Add quorum_aio_readv.

Add code to do num_children reads in parallel and cleanup the structures
afterwards.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

blkverify: Extract qemu_iovec_clone() and qemu_iovec_compare() from blkverify.

qemu_iovec_compare() will be used to compare IOs vectors in quorum blkverify
mode. The patch extracts these functions in order to factorize the code.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

quorum: Add quorum_aio_writev and its dependencies.

Writes are mirrored num_children times on num_children devices.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

quorum: Create BDRVQuorumState and BlkDriver and do init.

Create the structure holding the quorum settings and write the minimal block
driver instanciation boilerplate.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

quorum: Create quorum.c, add QuorumChildRequest and QuorumAIOCB.

Quorum is a block filter mirroring writes to num_children children.
For reads quorum reads each children and does a vote.
If more than vote_threshold versions are identical the quorum is reached and
this winning version is returned to the guest. So quorum prevents bit corruption.
For high availability purpose minority errors are reported via QMP but the guest
does not see them.

This patch creates the driver C source file and introduces the structures that
will be used in asynchronous reads and writes.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

check-qdict: Test termination of qdict_array_split()

qdict_array_split() should terminate if it encounters both an entry with
a key of "%u" and entries with keys prefixed "%u." for the same index.

This patch adds a test for this case.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

check-qdict: Adjust test for qdict_array_split()

Test the new functionality of qdict_array_split(), that is, splitting
off single objects.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qdict: Extract non-QDicts in qdict_array_split()

Currently, qdict_array_split() only splits off entries with a key prefix
of "%u.", packing them into a new QDict. This patch makes it support
entries with the plain key "%u" as well, directly putting them into the
new QList without creating a QDict.

If there is both an entry with a key of "%u" and other entries with keys
prefixed "%u." (for the same index), the function simply terminates.

To do this, this patch also adds a static function which tests whether a
given QDict contains any keys with the given prefix. This is used to test
whether entries with a key prefixed "%u." do exist in the source QDict
without modifying it.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qemu-config: Sections must consist of keys

In config_parse_qdict_section(), the QList returned by
qdict_array_split() is assumed to only contain QDicts. Currently, this
is true but it may (and will) change in the future. Therefore, check
whether the assumption actually holds.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qemu-iotests: Check qemu-img command line parsing

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>

qemu-img: Allow -o help with incomplete argument list

This patch allows using 'qemu-img $subcmd -o help' for the create,
convert and amend subcommands, without specifying the previously
required filename arguments.

Note that it's still allowed and meaningful to specify a filename: An
invocation like 'qemu-img create -o help sheepdog:foo' will also display
options that are provided by the Sheepdog driver.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>

qemu-img amend: Support multiple -o options

Instead of ignoring all option values but the last one, multiple -o
options now have the same meaning as having a single option with all
settings in the order of their respective -o options.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>

qemu-img convert: Support multiple -o options

Instead of ignoring all option values but the last one, multiple -o
options now have the same meaning as having a single option with all
settings in the order of their respective -o options.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>

qemu-img create: Support multiple -o options

If you specified multiple -o options for qemu-img create, it would
silently ignore all but the last one. This patch fixes the problem.

Now multiple -o options has the same meaning as having a single option
with all settings in the order of their respective -o options.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>

qemu-option: has_help_option() and is_valid_option_list()

has_help_option() checks if any help option ('help' or '?') occurs
anywhere in an option string, so that things like 'cluster_size=4k,help'
are recognised.

is_valid_option_list() ensures that the option list doesn't have options
with leading commas or trailing unescaped commas.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>

vdi: say why an image is bad

Instead of just putting it in debugging output, we can now put the
value in an Error.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: do not abuse EMEDIUMTYPE

Returning "Wrong medium type" for an image that does not have a valid
header is a bit weird.  Improve the error by mentioning what format
was trying to open it.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

vmdk: correctly propagate errors

Now that we can return the "right" errors, use the Error** parameter
to pass them back instead of just printing them.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

vmdk: do not try opening a file as both image and descriptor

This prepares for propagating errors from vmdk_open_sparse and
vmdk_open_desc_file up to the caller of vmdk_open.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

vmdk: push vmdk_read_desc up to caller

Currently, we just try reading a VMDK file as both image and descriptor.
This makes it hard to choose which of the two attempts gave the best error.
We'll decide in advance if the file looks like an image or a descriptor,
and this patch is the first step to that end.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

vmdk: extract vmdk_read_desc

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

vvfat: correctly propagate errors

Before:
    $ ./qemu-io-old
    qemu-io-old> open -r -o driver=vvfat,fat-type=24,dir=i386-softmmu
    Valid FAT types are only 12, 16 and 32
    qemu-io-old: can't open device (null): Could not open image: Invalid argument

After:
    $ ./qemu-io
    qemu-io> open -r -o driver=vvfat,fat-type=24,dir=i386-softmmu
    qemu-io: can't open device (null): Valid FAT types are only 12, 16 and 32

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

vhdx: correctly propagate errors

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qed: correctly propagate errors

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qcow: correctly propagate errors

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

curl: correctly propagate errors

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

cow: correctly propagate errors

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

gluster: correctly propagate errors

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

gluster: default scheme to gluster:// and host to localhost.

Currently, "gluster:///volname/img" and (using file. options)
"file.driver=gluster,file.filename=foo" will segfault.  Also,
"//host/volname/img" will be rejected, but it is a valid URL
that should be accepted just fine with "file.driver=gluster".
Accept all of these, by inferring missing transport and host
as TCP and localhost respectively.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

iscsi: correctly propagate errors in iscsi_open

Before:
    $ ./qemu-io-old
    qemu-io-old> open -r -o file.driver=iscsi,file.filename=foo
    Failed to parse URL : foo
    qemu-io-old: can't open device (null): Could not open 'foo': Invalid argument

After:
    $ ./qemu-io
    qemu-io> open -r -o file.driver=iscsi,file.filename=foo
    qemu-io: can't open device (null): Failed to parse URL : foo

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

iscsi: fix indentation

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

nbd: move socket wrappers to qemu-nbd

qemu-nbd is one of the few valid users of qerror_report_err.  Move
the error-reporting socket wrappers there.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

nbd: inline tcp_socket_incoming_spec into sole caller

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

nbd: correctly propagate errors

Before:
    $ ./qemu-io-old
    qemu-io-old> open -r -o file.driver=nbd
    one of path and host must be specified.
    qemu-io-old: can't open device (null): Could not open image: Invalid argument
    $ ./qemu-io-old
    qemu-io-old> open -r -o file.driver=nbd,file.host=foo,file.path=bar
    path and host may not be used at the same time.
    qemu-io-old: can't open device (null): Could not open image: Invalid argument

After:
    $ ./qemu-io
    qemu-io> open -r -o file.driver=nbd
    qemu-io: can't open device (null): one of path and host must be specified.
    $ ./qemu-io
    qemu-io> open -r -o file.driver=nbd,file.host=foo,file.path=bar
    qemu-io: can't open device (null): path and host may not be used at the same time.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

nbd: produce a better error if neither host nor port is passed

Before:
    $ qemu-io-old
    qemu-io-old> open -r -o file.driver=nbd
    qemu-io-old: can't open device (null): Could not open image: Invalid argument
    $ ./qemu-io-old
    qemu-io-old> open -r -o file.driver=nbd,file.host=foo,file.path=bar
    path and host may not be used at the same time.
    qemu-io-old: can't open device (null): Could not open image: Invalid argument

After:
    $ ./qemu-io
    qemu-io> open -r -o file.driver=nbd
    one of path and host must be specified.
    qemu-io: can't open device (null): Could not open image: Invalid argument
    $ ./qemu-io
    qemu-io> open -r -o file.driver=nbd,file.host=foo,file.path=bar
    path and host may not be used at the same time.
    qemu-io: can't open device (null): Could not open image: Invalid argument

Next patch will fix the error propagation.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: Remove bdrv_open_image()'s force_raw option

This option is now unnecessary since specifying BDRV_O_PROTOCOL as flag
will do exactly the same.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: Reuse success path from bdrv_open()

The fail and success paths of bdrv_file_open() may be further shortened
by reusing code already existent in bdrv_open(). This includes
bdrv_file_open() not taking the reference to options which allows the
removal of QDECREF(options) in that function.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>