block: driver should override flags in bdrv_open()

The BDRV_O_PROTOCOL flag should have an impact only if no driver is
specified explicitly. Therefore, if bdrv_open() is called with an
explicit block driver argument (either through the options QDict or
through the drv parameter) and that block driver is a protocol block
driver, BDRV_O_PROTOCOL should be set; if it is a format block driver,
BDRV_O_PROTOCOL should be unset.

While there was code to unset the flag in case a format block driver
has been selected, it only followed the bdrv_fill_options() function
call whereas the flag in fact needs to be adjusted before it is used
there.

With that change, BDRV_O_PROTOCOL will always be set if the BDS should
be a protocol driver; if the driver has been specified explicitly, the
new code will set it; and bdrv_fill_options() will only "probe" a
protocol driver if BDRV_O_PROTOCOL is set. The probing after
bdrv_fill_options() cannot select a protocol driver.

Thus, bdrv_open_image() to open BDS.file is never called if a protocol
BDS is about to be created. With that change in turn it is impossible to
call bdrv_open_common() with a protocol drv and file != NULL, which
allows us to remove the bdrv_swap() call.

This change breaks a test case in qemu-iotest 051:
"-drive file=t.qcow2,file.driver=qcow2" now works because the explicitly
specified "qcow2" overrides the BDRV_O_PROTOCOL which is automatically
set for the "file" BDS (and the filename is just passed down).
Therefore, this patch removes that test case.

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

block: Change bitmap truncate conditional to assertion

This is an artifact of an older version that had both all-bitmap and
single-bitmap truncate functions, and some info got lost in the shuffle.

Bitmaps can only be frozen during a backup operation, and a backup
operation should prevent a resize operation, so just assert that this
cannot happen.

Suggested-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: record new size in bdrv_dirty_bitmap_truncate

ce1ffea8 neglected to update the BdrvDirtyBitmap structure
itself for internal consistency. It's currently not an issue,
but for migration and persistence series this will cause headaches.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

raw-posix: Fix .bdrv_co_get_block_status() for unaligned image size

Image files with an unaligned image size have a final hole that starts
at EOF, i.e. in the middle of a sector. Currently, *pnum == 0 is
returned when checking the status of this sector. In qemu-img, this
triggers an assertion failure.

In order to fix this, one type for the sector that contains EOF must be
found. Treating a hole as data is safe, so this patch rounds the
calculated number of data sectors up, so that a partial sector at EOF is
treated as a full data sector.

This fixes https://bugzilla.redhat.com/show_bug.cgi?id=1229394

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Tested-by: Cole Robinson <crobinso@redhat.com>

vmdk: Use vmdk_find_index_in_cluster everywhere

Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

vmdk: Fix index_in_cluster calculation in vmdk_co_get_block_status

It has the similar issue with b1649fae49a8. Since the calculation
is repeated for a few times already, introduce a function so it can be
reused.

Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qcow2: Add DEFAULT_L2_CACHE_CLUSTERS

If a relatively large cluster size is chosen, the default of 1 MB L2
cache is not really appropriate. In this case, unless overridden by the
user, the default cache size should not be determined by its size in
bytes but by the number of L2 tables (clusters) it is supposed to
contain.

Note that without this patch, MIN_L2_CACHE_SIZE will effectively take
over the same role. However, providing space for just two L2 tables is
not enough to be the default.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

iotests: qcow2 COW with minimal L2 cache size

This adds a test case to test 103 for performing a COW operation in a
qcow2 image using an L2 cache with minimal size (which should be at
least two clusters so the COW can access both source and destination
simultaneously).

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qcow2: Set MIN_L2_CACHE_SIZE to 2

The L2 cache must cover at least two L2 tables, because during COW two
L2 tables are accessed simultaneously.

Reported-by: Alexander Graf <agraf@suse.de>
Cc: qemu-stable <qemu-stable@nongnu.org>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Tested-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qemu-iotests: Fix 128 if sudo required

If passwordless "sudo" works, use it in the qemu-io cmd.

Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

iotests: remove assertIsNotNone call

RHEL6 doesn't have Python 2.7, so replace this call with
assertNotEqual(x, None) which will work just as well.

Reported-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

Merge remote-tracking branch 'remotes/lalrae/tags/mips-20150612' into staging

MIPS patches 2015-06-12

Changes:
* improve dp8393x network card and rc4030 chipset emulation
* support misaligned R6 and MSA memory accesses
* support MIPS eXtended and Large Physical Addressing
* add Config5.FRE bit and ERETNC instruction (Config5.LLB)
* support ememsize on MALTA

# gpg: Signature made Fri Jun 12 09:38:11 2015 BST using RSA key ID 0B29DA6B
# gpg: Good signature from "Leon Alrae <leon.alrae@imgtec.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: 8DD3 2F98 5495 9D66 35D4  4FC0 5211 8E3C 0B29 DA6B

* remotes/lalrae/tags/mips-20150612: (29 commits)
  target-mips: enable XPA and LPA features
  target-mips: remove misleading comments in translate_init.c
  target-mips: add MTHC0 and MFHC0 instructions
  target-mips: add CP0.PageGrain.ELPA support
  target-mips: support Page Frame Number Extension field
  target-mips: extend selected CP0 registers to 64-bits in MIPS32
  target-mips: correct MFC0 for CP0.EntryLo in MIPS64
  net/dp8393x: fix hardware reset
  net/dp8393x: correctly reset in_use field
  net/dp8393x: add load/save support
  net/dp8393x: add PROM to store MAC address
  net/dp8393x: QOM'ify
  net/dp8393x: use dp8393x_ prefix for all functions
  net/dp8393x: do not use old_mmio accesses
  net/dp8393x: always calculate proper checksums
  dma/rc4030: convert to QOM
  dma/rc4030: use trace events instead of custom logging
  dma/rc4030: document register at offset 0x210
  dma/rc4030: do not use old_mmio accesses
  dma/rc4030: use AddressSpace and address_space_rw in users
  ...

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

Merge remote-tracking branch 'remotes/juanquintela/tags/migration/20150612' into staging

migration/next for 20150612

# gpg: Signature made Fri Jun 12 05:56:21 2015 BST using RSA key ID 5872D723
# gpg: Good signature from "Juan Quintela <quintela@redhat.com>"
# gpg:                 aka "Juan Quintela <quintela@trasno.org>"

* remotes/juanquintela/tags/migration/20150612: (21 commits)
  Remove unneeded memset
  Rename RDMA structures to make destination clear
  Teach analyze-migration.py about section footers
  Add a protective section footer
  Disable section footers on older machine types
  Merge section header writing
  Move loadvm_handlers into MigrationIncomingState
  Move copy out of qemu_peek_buffer
  Create MigrationIncomingState
  qemu_ram_foreach_block: pass up error value, and down the ramblock name
  Split header writing out of qemu_savevm_state_begin
  Add qemu_get_counted_string to read a string prefixed by a count byte
  migration: Use normal VMStateDescriptions for Subsections
  migration: create savevm_state
  migration: Remove duplicated assignment of SETUP status
  rdma: Fix qemu crash when IPv6 address is used for migration
  arch_init: Clean up the duplicate variable 'len' defining in ram_load()
  migration: reduce include files
  migration: Add myself to the copyright list of both files
  migration: move savevm.c inside migration/
  ...

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

target-mips: enable XPA and LPA features

Enable XPA in MIPS32R5-generic and LPA in MIPS64R6-generic.

Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>

target-mips: remove misleading comments in translate_init.c

PABITS are not hardcoded to 36 bits and we do not model 59 PABITS (which is
the architectural limit) in QEMU.

Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>

target-mips: add MTHC0 and MFHC0 instructions

Implement MTHC0 and MFHC0 instructions. In MIPS32 they are used to access
upper word of extended to 64-bits CP0 registers.

In MIPS64, when CP0 destination register specified is the EntryLo0 or
EntryLo1, bits 1:0 of the GPR appear at bits 31:30 of EntryLo0 or
EntryLo1. This is to compensate for RI and XI, which were shifted to bits
63:62 by MTC0 to EntryLo0 or EntryLo1. Therefore creating separate
functions for EntryLo0 and EntryLo1.

Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>

target-mips: add CP0.PageGrain.ELPA support

CP0.PageGrain.ELPA enables support for large physical addresses. This field
is encoded as follows:
0: Large physical address support is disabled.
1: Large physical address support is enabled.

If this bit is a 1, the following changes occur to coprocessor 0 registers:
- The PFNX field of the EntryLo0 and EntryLo1 registers is writable and
  concatenated with the PFN field to form the full page frame number.
- Access to optional COP0 registers with PA extension, LLAddr, TagLo is
  defined.

P5600 can operate in 32-bit or 40-bit Physical Address Mode. Therefore if
XPA is disabled (CP0.PageGrain.ELPA = 0) then assume 32-bit Address Mode.
In MIPS64 assume 36 as default PABITS (when CP0.PageGrain.ELPA = 0).

env->PABITS value is constant and indicates maximum PABITS available on
a core, whereas env->PAMask is calculated from env->PABITS and is also
affected by CP0.PageGrain.ELPA.

Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>

target-mips: support Page Frame Number Extension field

Update tlb->PFN to contain PFN concatenated with PFNX. PFNX is 0 if large
physical address is not supported.

Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>

target-mips: extend selected CP0 registers to 64-bits in MIPS32

Extend EntryLo0, EntryLo1, LLAddr and TagLo from 32 to 64 bits in MIPS32.

Introduce gen_move_low32() function which moves low 32 bits from 64-bit
temp to GPR; it sign extends 32-bit value on MIPS64 and truncates on
MIPS32.

Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>

target-mips: correct MFC0 for CP0.EntryLo in MIPS64

CP0.EntryLo bits 31:30 have to be cleared.

Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>

Remove unneeded memset

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Michael R. Hines <mrhines@us.ibm.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

Rename RDMA structures to make destination clear

RDMA has two data types that are named confusingly;
   RDMALocalBlock (pointed to indirectly by local_ram_blocks)
   RDMARemoteBlock (pointed to by block in RDMAContext)

RDMALocalBlocks, as the name suggests is a data strucuture that
represents the RDMAable RAM Blocks on the current side of the migration
whichever that is.

RDMARemoteBlocks is always the shape of the RAMBlocks on the
destination, even on the destination.

Rename:
     RDMARemoteBlock -> RDMADestBlock
     context->'block' -> context->dest_blocks

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Michael R. Hines <mrhines@us.ibm.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

Teach analyze-migration.py about section footers

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

Add a protective section footer

Badly formatted migration streams can go undetected or produce
misleading errors due to a lock of checking at the end of sections.
In particular a section that adds an extra 0x00 at the end
causes what looks like a normal end of stream and thus doesn't produce
any errors, and something that ends in a 0x01..0x04 kind of look
like real section headers and then fail when the section parser tries
to figure out which section they are.  This is made worse by the
choice of 0x00..0x04 being small numbers that are particularly common
in normal section data.

This patch adds a section footer consisting of a marker (0x7e - ~)
followed by the section-id that was also sent in the header.  If
they mismatch then it throws an error explaining which section was
being loaded.

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

Disable section footers on older machine types

The next patch adds section footers; but we don't want to
break migration compatibility so disable them on older
machine types

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

Merge section header writing

The header writing for device sections is open coded in
a few places, merge it into one.

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

Move loadvm_handlers into MigrationIncomingState

In postcopy we need the loadvm_handlers to be used in a couple
of different instances of the loadvm loop/routine, and thus
it can't be local any more.

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

Move copy out of qemu_peek_buffer

qemu_peek_buffer currently copies the data it reads into a buffer,
however a future patch wants access to the buffer without the copy,
hence rework to remove the copy to the layer above.

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

Create MigrationIncomingState

There are currently lots of pieces of incoming migration state scattered
around, and postcopy is adding more, and it seems better to try and keep
it together.

allocate MIS in process_incoming_migration_co

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

qemu_ram_foreach_block: pass up error value, and down the ramblock name

check the return value of the function it calls and error if it's non-0
Fixup qemu_rdma_init_one_block that is the only current caller,
  and rdma_add_block the only function it calls using it.

Pass the name of the ramblock to the function; helps in debugging.

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Michael R. Hines <mrhines@us.ibm.com>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

Split header writing out of qemu_savevm_state_begin

Split qemu_savevm_state_begin to:
  qemu_savevm_state_header   That writes the initial file header.
  qemu_savevm_state_begin    That sets up devices and does the first
                             device pass.

Used later in postcopy.

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

Add qemu_get_counted_string to read a string prefixed by a count byte

and use it in loadvm_state and ram_load.

Where ever it's used, check the return and error if it failed.

Minor: ram_load was using a 257 byte array for its string, the
       maximum length is 255 bytes + 0 terminator, so fix to 256

Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

migration: Use normal VMStateDescriptions for Subsections

We create optional sections with this patch.  But we already have
optional subsections.  Instead of having two mechanism that do the
same, we can just generalize it.

For subsections we just change:

- Add a needed function to VMStateDescription
- Remove VMStateSubsection (after removal of the needed function
  it is just a VMStateDescription)
- Adjust the whole tree, moving the needed function to the corresponding
  VMStateDescription

Signed-off-by: Juan Quintela <quintela@redhat.com>

migration: create savevm_state

This way, we will put savevm global state here, instead of lots of variables.

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

migration: Remove duplicated assignment of SETUP status

We assign the MIGRATION_STATUS_SETUP status in two places.  Just in
succession.  Just remove the second one.

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

rdma: Fix qemu crash when IPv6 address is used for migration

Qemu crashes when IPv6 address is specified for migration and access
to any RDMA uverbs device available on the system is blocked using cgroups.
Fix the crash by checking the return value of ibv_open_device routine.

Signed-off-by: Meghana Cheripady <meghana.cheripady@avagotech.com>
Signed-off-by: Padmanabh Ratnakar <padmanabh.ratnakar@avagotech.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

arch_init: Clean up the duplicate variable 'len' defining in ram_load()

There are two places that define 'len' variable, It's OK for compiling,
but makes it difficult for reading.

Remove the local one which defined in the inside 'while' loop.

Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Juan Quintela <quintela@redhat.com>

migration: reduce include files

To make changes easier, with the copy, I maintained almost all include
files.  Now I remove the unnecessary ones on this patch.  This compiles
on linux x64 with all architectures configured, and cross-compiles for
windows 32 and 64 bits.

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

migration: Add myself to the copyright list of both files

If anyone feels like adding himself to the list, just sent me a patch.

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

migration: move savevm.c inside migration/

Now, everything is in place.

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

migration: move ram stuff to migration/ram

For historic reasons, ram migration have been on arch_init.c.  Just
split it into migration/ram.c, the same that happened with block.c.

There is only code movement, no changes altogether.

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

Merge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging

pc, acpi, virtio

Most notably this includes virtio 1 patches
Still not all devices converted, and not fully spec compliant,
so disabled by default.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
# gpg: Signature made Thu Jun 11 12:53:08 2015 BST using RSA key ID D28D5469
# gpg: Good signature from "Michael S. Tsirkin <mst@kernel.org>"
# gpg:                 aka "Michael S. Tsirkin <mst@redhat.com>"

* remotes/mst/tags/for_upstream: (42 commits)
  i386/acpi-build: fix PXB workarounds for unsupported BIOSes
  i386/acpi-build: more traditional _UID and _HID for PXB root buses
  vhost-scsi: move qdev properties into vhost-scsi.c
  virtio-9p-device: move qdev properties into virtio-9p-device.c
  virtio-serial-bus: move qdev properties into virtio-serial-bus.c
  virtio-rng: move qdev properties into virtio-rng.c
  virtio-scsi: move qdev properties into virtio-scsi.c
  virtio-net.h: Remove unsed DEFINE_VIRTIO_NET_PROPERTIES
  virtio-net: move qdev properties into virtio-net.c
  virtio-input: emulated devices [pci]
  virtio-input: core code & base class [pci]
  pci: add PCI_CLASS_INPUT_*
  virtio-pci: fill VirtIOPCIRegions early.
  virtio-pci: drop identical virtio_pci_cap
  virtio-pci: move cap type to VirtIOPCIRegion
  virtio-pci: move virtio_pci_add_mem_cap call to virtio_pci_modern_region_map
  virtio-pci: add virtio_pci_modern_region_map()
  virtio-pci: add virtio_pci_modern_regions_init()
  virtio-pci: add struct VirtIOPCIRegion for virtio-1 regions
  virtio-balloon: switch to virtio_add_feature
  ...

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

Merge remote-tracking branch 'remotes/kraxel/tags/pull-sdl-20150611-1' into staging

sdl2: fix crash in handle_windowevent() when restoring the screen size

# gpg: Signature made Thu Jun 11 08:57:38 2015 BST 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-sdl-20150611-1:
  sdl2: fix crash in handle_windowevent() when restoring the screen size

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

hw/vfio/platform: replace g_malloc0_n by g_new0

g_malloc0_n() is introduced since glib-2.24 while QEMU currently
requires glib-2.22. This may cause a link error on some distributions.

Signed-off-by: Eric Auger <eric.auger@linaro.org>
Reviewed-by: Gonglei <arei.gonglei@huawei.com>
Acked-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/spice/tags/pull-spice-20150611-1' into staging

spice: fix segfault in qemu_spice_create_update, ui_info tweaks.

# gpg: Signature made Thu Jun 11 08:48:49 2015 BST 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-20150611-1:
  spice: ui_info tweaks
  spice-display: fix segfault in qemu_spice_create_update

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

i386/acpi-build: fix PXB workarounds for unsupported BIOSes

The patch

  apci: fix PXB behaviour if used with unsupported BIOS

uses the following condition to see if a "PXB mem/IO chunk" has *not* been
configured by the BIOS:

  (!range_base || range_base > range_limit)

When this condition evaluates to true, said patch *omits* the
corresponding entry from the _CRS.

Later on the patch checks for the opposite condition (with the intent of
*adding* entries to the _CRS if the "PXB mem/IO chunks" *have* been
configured). Unfortunately, the condition was negated incorrectly: only
the first ! operator was removed, which led to the nonsensical expression

  (range_base || range_base > range_limit)

leading to bogus entries in the _CRS, and causing BSOD in Windows Server
2012 R2 when it runs on OVMF.

The correct negative of the condition seen at the top is

  (range_base && range_base <= range_limit)

Fix the expressions.

Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

i386/acpi-build: more traditional _UID and _HID for PXB root buses

The ACPI specification permits the _HID and _UID objects to evaluate to
strings. (See "6.1.5 _HID (Hardware ID)" and "6.1.12 _UID (Unique ID)" in
the ACPI v6.0 spec.)

With regard to related standards, the UEFI specification can also express
a device address composed from string _HID and _UID identifiers, inside
the Expanded ACPI Device Path Node. (See "9.3.3 ACPI Device Path", Table
49, in the UEFI v2.5 spec.)

However, numeric (integer) contents for both _HID and _UID are more
traditional. They are recommended by the UEFI spec for size reasons:

  [...] the ACPI Device Path node is smaller and should be used if
  possible to reduce the size of device paths that may potentially be
  stored in nonvolatile storage [...]

External tools support them better (for example the --acpi_hid and
--acpi_uid options of "efibootmgr" only take numeric identifiers).
Finally, numeric _HID and _UID contents are existing practice in the QEMU
source.

This patch was tested with a Fedora 20 LiveCD and a preexistent Windows
Server 2012 R2 guest. Using "acpidump" and "iasl" in the Fedora guest, we
get, in the SSDT:

> Scope (\_SB)
> {
>   Device (PC04)
>   {
>     Name (_UID, 0x04)  // _UID: Unique ID
>     Name (_HID, EisaId ("PNP0A03") /* PCI Bus */)  // _HID: Hardware ID

Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

Merge remote-tracking branch 'remotes/kraxel/tags/pull-gtk-20150611-1' into staging

gtk: don't exit early in case gtk init fails

# gpg: Signature made Thu Jun 11 10:38:29 2015 BST 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-gtk-20150611-1:
  gtk: don't exit early in case gtk init fails

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

gtk: don't exit early in case gtk init fails

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>

net/dp8393x: fix hardware reset

Documentation is not clear of what happens when doing a hardware reset,
but firmware expect all registers to be zero unless specified otherwise.

This fixes reboot on MIPS Magnum.

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

net/dp8393x: correctly reset in_use field

Don't write more than the field width, which is always 16 bit.
Fixes network in NetBSD 5.1/arc

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

net/dp8393x: add load/save support

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

net/dp8393x: add PROM to store MAC address

Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

net/dp8393x: QOM'ify

Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

net/dp8393x: use dp8393x_ prefix for all functions

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

net/dp8393x: do not use old_mmio accesses

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

net/dp8393x: always calculate proper checksums

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

dma/rc4030: convert to QOM

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

dma/rc4030: use trace events instead of custom logging

Remove also unneeded debug logs.

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

dma/rc4030: document register at offset 0x210

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

dma/rc4030: do not use old_mmio accesses

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

dma/rc4030: use AddressSpace and address_space_rw in users

Now that rc4030 internally uses an AddressSpace for DMA handling, make its root
memory region public. This is especially usefull for dp8393x netcard, which now
uses well known QEMU types and methods.

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

dma/rc4030: create custom DMA address space

Add a new memory region in system address space where DMA address space
definition (the 'translation table') belongs, so we can update on the fly
the DMA address space.

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

mips jazz: compile only in 64 bit

Remove now useless device models from other MIPS configurations

We're now compiling 12 files less than before.

Signed-off-by: Hervé Poussineau <hpoussin@reactos.org>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

target-mips: add ERETNC instruction and Config5.LLB bit

ERETNC is identical to ERET except that an ERETNC will not clear the LLbit
that is set by execution of an LL instruction, and thus when placed between
an LL and SC sequence, will never cause the SC to fail.

Presence of ERETNC is denoted by the Config5.LLB.

Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>

target-mips: Misaligned memory accesses for MSA

MIPS SIMD Architecture vector loads and stores require misalignment support.
MSA Memory access should work as an atomic operation. Therefore, it has to
check validity of all addresses for a vector store access if it is spanning
into two pages.

Separating helper functions for each data format as format is known in
translation.
To use mmu_idx from cpu_mmu_index() instead of calculating it from hflag.
Removing save_cpu_state() call in translation because it is able to use
cpu_restore_state() on fault as GETRA() is passed.

Signed-off-by: Yongbok Kim <yongbok.kim@imgtec.com>
Reviewed-by: Leon Alrae <leon.alrae@imgtec.com>
[leon.alrae@imgtec.com: remove unused do_* functions]
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

softmmu: Add probe_write()

Probe for whether the specified guest write access is permitted.
If it is not permitted then an exception will be taken in the same
way as if this were a real write access (and we will not return).
Otherwise the function will return, and there will be a valid
entry in the TLB for this access.

Signed-off-by: Yongbok Kim <yongbok.kim@imgtec.com>
Reviewed-by: Leon Alrae <leon.alrae@imgtec.com>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

target-mips: Misaligned memory accesses for R6

Release 6 requires misaligned memory access support for all ordinary memory
access instructions (for example, LW/SW, LWC1/SWC1).
However misaligned support is not provided for certain special memory accesses
such as atomics (for example, LL/SC).

Signed-off-by: Yongbok Kim <yongbok.kim@imgtec.com>
Reviewed-by: Leon Alrae <leon.alrae@imgtec.com>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

mips_malta: provide ememsize env variable to kernels

Commit 94c2b6aff43c (mips_malta: support up to 2GiB RAM) provided
support for using over 256MB of RAM with the MIPS Malta board, including
capping the memsize variable that QEMUs pseudo-bootloader provides to
the kernel at 256MB in order to match YAMON. It didn't however provide
the ememsize variable which kernels supporting memory outside of the
unmapped address spaces (ie. EVA or highmem) may use to determine the
true size of the RAM present in the system.

Set ememsize to the size of RAM so that such kernels may use all
available memory without the user having to manually specifying its size
& location.

Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Cc: Leon Alrae <leon.alrae@imgtec.com>
Reviewed-by: Leon Alrae <leon.alrae@imgtec.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

target-mips: add Config5.FRE support allowing Status.FR=0 emulation

This relatively small architectural feature adds the following:

FIR.FREP: Read-only. If FREP=1, then Config5.FRE and Config5.UFE are
          available.

Config5.FRE: When enabled all single-precision FP arithmetic instructions,
             LWC1/LWXC1/MTC1, SWC1/SWXC1/MFC1 cause a Reserved Instructions
             exception.

Config5.UFE: Allows user to write/read Config5.FRE using CTC1/CFC1
             instructions.

Enable the feature in MIPS64R6-generic CPU.

Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

target-mips: move group of functions above gen_load_fpr32()

Move the "Tests" group of functions so that gen_load_fpr32() and
gen_store_fpr32() can use generate_exception().

Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>

spice: ui_info tweaks

Use the new dpy_ui_info_supported function.
Clarifies the control flow.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

spice-display: fix segfault in qemu_spice_create_update

Although it is pretty unusual the stride for the guest image and the
mirror image maintained by spice-display can be different.  So use
separate variables for them.

https://bugzilla.redhat.com/show_bug.cgi?id=1163047

Cc: qemu-stable@nongnu.org
Reported-by: perrier vincent <clownix@clownix.net>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

Merge remote-tracking branch 'remotes/kraxel/tags/pull-vga-20150610-1' into staging

stdvga: factor out mmio subregion init
virtio-gpu: add virtio gpu core code, 2d mode

# gpg: Signature made Wed Jun 10 10:03:11 2015 BST 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-vga-20150610-1:
  virtio-gpu/2d: add virtio gpu core code
  virtio: update headers, add virtio-gpu (2d)
  stdvga: factor out mmio subregion init
  stdvga: pass VGACommonState instead of PCIVGAState
  stdvga: fix offset in pci_vga_ioport_read

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

vhost-scsi: move qdev properties into vhost-scsi.c

As only one place in vhost-scsi.c uses DEFINE_VHOST_SCSI_PROPERTIES,
there is no need to expose it. Inline it into vhost-scsi.c to avoid
wrongly use.

Signed-off-by: Shannon Zhao <zhaoshenglong@huawei.com>
Signed-off-by: Shannon Zhao <shannon.zhao@linaro.org>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-9p-device: move qdev properties into virtio-9p-device.c

As only one place in virtio-9p-device.c uses
DEFINE_VIRTIO_9P_PROPERTIES, there is no need to expose it. Inline it
into virtio-9p-device.c to avoid wrongly use.

Signed-off-by: Shannon Zhao <zhaoshenglong@huawei.com>
Signed-off-by: Shannon Zhao <shannon.zhao@linaro.org>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-serial-bus: move qdev properties into virtio-serial-bus.c

As only one place in virtio-serial-bus.c uses
DEFINE_VIRTIO_SERIAL_PROPERTIES, there is no need to expose it. Inline
it into virtio-serial-bus.c to avoid wrongly use.

Signed-off-by: Shannon Zhao <zhaoshenglong@huawei.com>
Signed-off-by: Shannon Zhao <shannon.zhao@linaro.org>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-rng: move qdev properties into virtio-rng.c

As only one place in virtio-rng.c uses DEFINE_VIRTIO_RNG_PROPERTIES,
there is no need to expose it. Inline it into virtio-rng.c to avoid
wrongly use.

Signed-off-by: Shannon Zhao <zhaoshenglong@huawei.com>
Signed-off-by: Shannon Zhao <shannon.zhao@linaro.org>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-scsi: move qdev properties into virtio-scsi.c

As only one place in virtio-scsi.c uses DEFINE_VIRTIO_SCSI_PROPERTIES
and DEFINE_VIRTIO_SCSI_FEATURES, there is no need to expose them. Inline
them into virtio-scsi.c to avoid wrongly use.

Signed-off-by: Shannon Zhao <zhaoshenglong@huawei.com>
Signed-off-by: Shannon Zhao <shannon.zhao@linaro.org>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-net.h: Remove unsed DEFINE_VIRTIO_NET_PROPERTIES

Remove unsed DEFINE_VIRTIO_NET_PROPERTIES in virtio-net.h and delete a
space typo.

Signed-off-by: Shannon Zhao <zhaoshenglong@huawei.com>
Signed-off-by: Shannon Zhao <shannon.zhao@linaro.org>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-net: move qdev properties into virtio-net.c

As only one place in virtio-net.c uses DEFINE_VIRTIO_NET_FEATURES,
there is no need to expose it. Inline it into virtio-net.c to avoid
wrongly use.

Signed-off-by: Shannon Zhao <zhaoshenglong@huawei.com>
Signed-off-by: Shannon Zhao <shannon.zhao@linaro.org>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-input: emulated devices [pci]

This patch adds virtio-pci support for the emulated virtio-input
devices.  Using them is as simple as adding "-device virtio-tablet-pci"
to your command line.  If you want add multiple devices but don't want
waste a pci slot for each you can compose a multifunction device this way:

qemu -device virtio-keyboard-pci,addr=0d.0,multifunction=on \
     -device virtio-tablet-pci,addr=0d.1,multifunction=on

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-input: core code & base class [pci]

This patch adds the virtio-pci support bits for virtio-input-device.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

pci: add PCI_CLASS_INPUT_*

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: fill VirtIOPCIRegions early.

Initialize the modern bar and the VirtIOPCIRegion fields early, in
realize.  Also add a size field to VirtIOPCIRegion and variables for
pci bars to VirtIOPCIProxy.

This allows virtio-pci subclasses to change things before the
device_plugged callback applies them.  virtio-vga will use that to
arrange regions in a way that virtio-vga is compatible to both stdvga
(in vga mode) and virtio-gpu-pci (in pci mode).

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: drop identical virtio_pci_cap

Now the three struct virtio_pci_caps are identical,
lets drop two of them ;)

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: move cap type to VirtIOPCIRegion

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: move virtio_pci_add_mem_cap call to virtio_pci_modern_region_map

Also fill offset and length automatically,
from VirtIOPCIRegion->offset and region size.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: add virtio_pci_modern_region_map()

Add function to map modern virtio regions.
Add offset to VirtIOPCIRegion.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: add virtio_pci_modern_regions_init()

Add init function for the modern pci regions,
move over the init code.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: add struct VirtIOPCIRegion for virtio-1 regions

For now just place the MemoryRegion there,
following patches will add more.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

virtio-balloon: switch to virtio_add_feature

This was missed during the conversion of feature bit manipulation.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>

virtio_balloon: header update

add modern header

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: correctly set host notifiers for modern bar

Currently, during host notifier set. We only add eventfd for legacy
bar, this is not correct since:

- Non-transitional device does not have legacy bar, so qemu will crash
  since proxy->bar was not initialized.
- Modern device uses modern bar and notify cap to notify the device,
  we should add eventfd for proxy->notify.

So this patch fixes the above two issues by adding eventfd based on
whether legacy or modern device were supported.

Signed-off-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: make modern bar 64bit + prefetchable

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: change & document virtio pci bar layout.

This patch adds variables for the pci bars (to get rid of the magic
numbers in the code) and moves the modern virtio bar to region 4 so
regions 2+3 are kept free.  virtio-vga wants use them.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: make QEMU_VIRTIO_PCI_QUEUE_MEM_MULT smaller

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: add flags to enable/disable legacy/modern

Add VIRTIO_PCI_FLAG_DISABLE_LEGACY and VIRTIO_PCI_FLAG_DISABLE_MODERN
for VirtIOPCIProxy->flags.  Also add properties for them.  They can be
used to disable modern (virtio 1.0) or legacy (virtio 0.9) modes.

By default only legacy is advertized, modern will be turned on by
default once all remaining spec compilance issues are addressed.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>

virtio-pci: switch to modern accessors for 1.0

virtio 1.0 config space is in LE format for all
devices, use modern wrappers when accessed through
the 1.0 BAR.

Reported-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>

virtio: add modern config accessors

virtio 1.0 defines config space as LE,
as opposed to pre-1.0 which was native endian.

Add API for transports to execute word/dword accesses in
little endian format - will be useful for mmio
and pci (byte access is also wrapped, for completeness).

For simplicity, we still keep config in host native
endian format, byteswap to LE on guest access.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>