postcopy: helper for waking shared

Provide a helper to send a 'wake' request on a userfaultfd for
a shared process.
The address in the clients address space is specified together
with the RAMBlock it was resolved to.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

vhost+postcopy: Resolve client address

Resolve fault addresses read off the clients UFD into RAMBlock
and offset, and call back to the postcopy code to ask for the page.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Peter Xu <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

postcopy-ram: add a stub for postcopy_request_shared_page

This fixes the build on systems without userfaultfd.

Signed-off-by: Michael S. Tsirkin <[email protected]>

vhost+postcopy: Helper to send requests to source for shared pages

Provide a helper to be used by shared waker functions to request
shared pages from the source.
The last_rb pointer is moved into the incoming state since this
helper can update it as well as the main fault thread function.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

vhost+postcopy: Stash RAMBlock and offset

Stash the RAMBlock and offset for later use looking up
addresses.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

vhost+postcopy: Send address back to qemu

We need a better way, but at the moment we need the address of the
mappings sent back to qemu so it can interpret the messages on the
userfaultfd it reads.

This is done as a 3 stage set:
   QEMU -> client
      set_mem_table

   mmap stuff, get addresses

   client -> qemu
       here are the addresses

   qemu -> client
       OK - now you can use them

That ensures that qemu has registered the new addresses in it's
userfault code before the client starts accessing them.

Note: We don't ask for the default 'ack' reply since we've got our own.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

libvhost-user+postcopy: Register new regions with the ufd

When new regions are sent to the client using SET_MEM_TABLE, register
them with the userfaultfd.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

migration/ram: ramblock_recv_bitmap_test_byte_offset

Utility for testing the map when you already know the offset
in the RAMBlock.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Peter Xu <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

postcopy+vhost-user: Split set_mem_table for postcopy

Split the set_mem_table routines in both qemu and libvhost-user
because the postcopy versions are going to be quite different
once changes in the later patches are added. However, this patch
doesn't produce any functional change, just the split.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

vhost+postcopy: Transmit 'listen' to slave

Notify the vhost-user slave on reception of the 'postcopy-listen'
event from the source.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Reviewed-by: Peter Xu <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

vhost+postcopy: Register shared ufd with postcopy

Register the UFD that comes in as the response to the 'advise' method
with the postcopy code.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

postcopy: Allow registering of fd handler

Allow other userfaultfd's to be registered into the fault thread
so that handlers for shared memory can get responses.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Peter Xu <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

libvhost-user: Open userfaultfd

Open a userfaultfd (on a postcopy_advise) and send it back in
the reply to the qemu for it to monitor.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

libvhost-user: Support sending fds back to qemu

Allow replies with fds (for postcopy)

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

vhost-user: Add 'VHOST_USER_POSTCOPY_ADVISE' message

Wire up a notifier to send a VHOST_USER_POSTCOPY_ADVISE
message on an incoming advise.

Later patches will fill in the behaviour/contents of the
message.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

postcopy: Add vhost-user flag for postcopy and check it

Add a vhost feature flag for postcopy support, and
use the postcopy notifier to check it before allowing postcopy.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Peter Xu <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

postcopy: Add notifier chain

Add a notifier chain for postcopy with a 'reason' flag
and an opportunity for a notifier member to return an error.

Call it when enabling postcopy.

This will initially used to enable devices to declare they're unable
to postcopy and later to notify of devices of stages within postcopy.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Peter Xu <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

postcopy: use UFFDIO_ZEROPAGE only when available

Use a flag on the RAMBlock to state whether it has the
UFFDIO_ZEROPAGE capability, use it when it's available.

This allows the use of postcopy on tmpfs as well as hugepage
backed files.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Peter Xu <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

qemu_ram_block_host_offset

Utility to give the offset of a host pointer within a RAMBlock
(assuming we already know it's in that RAMBlock)

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Peter Xu <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

migrate: Update ram_block_discard_range for shared

The choice of call to discard a block is getting more complicated
for other cases.   We use fallocate PUNCH_HOLE in any file cases;
it works for both hugepage and for tmpfs.
We use the DONTNEED for non-hugepage cases either where they're
anonymous or where they're private.

Care should be taken when trying other backing files.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Peter Xu <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

Makefile: add target to print generated files

This is helpful for automatic code analysis.

Signed-off-by: Michael S. Tsirkin <[email protected]>

test/acpi-test-data: add ACPI tables for dimmpxm test

Reviewers can use ACPI tables in this patch to run
test_acpi_{piix4,q35}_tcg_dimm_pxm cases.

Signed-off-by: Haozhong Zhang <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

tests/bios-tables-test: add test cases for DIMM proximity

QEMU now builds one SRAT memory affinity structure for each PC-DIMM
and NVDIMM device presented at boot time with the proximity domain
specified in the device option 'node', rather than only one SRAT
memory affinity structure covering the entire hotpluggable address
space with the proximity domain of the last node.

Add test cases on PC and Q35 machines with 4 proximity domains, and
one PC-DIMM and one NVDIMM attached to the 2nd and 3rd proximity
domains respectively. Check whether the QEMU-built SRAT tables match
with the expected ones.

The following ACPI tables need to be added for this test:
  tests/acpi-test-data/pc/APIC.dimmpxm
  tests/acpi-test-data/pc/DSDT.dimmpxm
  tests/acpi-test-data/pc/NFIT.dimmpxm
  tests/acpi-test-data/pc/SRAT.dimmpxm
  tests/acpi-test-data/pc/SSDT.dimmpxm
  tests/acpi-test-data/q35/APIC.dimmpxm
  tests/acpi-test-data/q35/DSDT.dimmpxm
  tests/acpi-test-data/q35/NFIT.dimmpxm
  tests/acpi-test-data/q35/SRAT.dimmpxm
  tests/acpi-test-data/q35/SSDT.dimmpxm
New APIC and DSDT are needed because of the multiple processors
configuration. New NFIT and SSDT are needed because of NVDIMM.

Signed-off-by: Haozhong Zhang <[email protected]>
Suggested-by: Igor Mammedov <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

hw/acpi-build: build SRAT memory affinity structures for DIMM devices

ACPI 6.2A Table 5-129 "SPA Range Structure" requires the proximity
domain of a NVDIMM SPA range must match with corresponding entry in
SRAT table.

The address ranges of vNVDIMM in QEMU are allocated from the
hot-pluggable address space, which is entirely covered by one SRAT
memory affinity structure. However, users can set the vNVDIMM
proximity domain in NFIT SPA range structure by the 'node' property of
'-device nvdimm' to a value different than the one in the above SRAT
memory affinity structure.

In order to solve such proximity domain mismatch, this patch builds
one SRAT memory affinity structure for each DIMM device present at
boot time, including both PC-DIMM and NVDIMM, with the proximity
domain specified in '-device pc-dimm' or '-device nvdimm'.

The remaining hot-pluggable address space is covered by one or multiple
SRAT memory affinity structures with the proximity domain of the last
node as before.

Signed-off-by: Haozhong Zhang <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

qmp: distinguish PC-DIMM and NVDIMM in MemoryDeviceInfoList

It may need to treat PC-DIMM and NVDIMM differently, e.g., when
deciding the necessity of non-volatile flag bit in SRAT memory
affinity structures.

A new field 'nvdimm' is added to the union type MemoryDeviceInfo for
such purpose. Its type is currently PCDIMMDeviceInfo and will be
updated when necessary in the future.

It also fixes "info memory-devices"/query-memory-devices which
currently show nvdimm devices as dimm devices since
object_dynamic_cast(obj, TYPE_PC_DIMM) happily cast nvdimm to
TYPE_PC_DIMM which it's been inherited from.

Signed-off-by: Haozhong Zhang <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Igor Mammedov <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

pc-dimm: make qmp_pc_dimm_device_list() sort devices by address

Make qmp_pc_dimm_device_list() return sorted by start address
list of devices so that it could be reused in places that
would need sorted list*. Reuse existing pc_dimm_built_list()
to get sorted list.

While at it hide recursive callbacks from callers, so that:

  qmp_pc_dimm_device_list(qdev_get_machine(), &list);

could be replaced with simpler:

  list = qmp_pc_dimm_device_list();

* follow up patch will use it in build_srat()

Signed-off-by: Haozhong Zhang <[email protected]>
Reviewed-by: Igor Mammedov <[email protected]>
Acked-by: David Gibson <[email protected]> for ppc part
Reviewed-by: Bharata B Rao <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

hw/pci: remove obsolete PCIDevice->init()

All PCI devices are now QOM'ified.

Signed-off-by: Philippe Mathieu-Daudé <[email protected]>
Reviewed-by: Marcel Apfelbaum <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

standard-headers: update virtio_net.h

include speed/duplex fields

Signed-off-by: Michael S. Tsirkin <[email protected]>

tests: acpi: don't read all fields in test_acpi_fadt_table()

there is no point to read fields here but not actually
checking them so drop it and read only header + dsdt/facs
addresses since it's needed later to fetch that tables.

With this cleanup we can get rid of AcpiFadtDescriptorRev3/
ACPI_FADT_COMMON_DEF which have no users left.

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eric Auger <[email protected]>
Tested-by: Eric Auger <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

virt_arm: acpi: reuse common build_fadt()

Extend generic build_fadt() to support rev5.1 FADT
and reuse it for 'virt' board, it would allow to
phase out usage of AcpiFadtDescriptorRev5_1 and
later ACPI_FADT_COMMON_DEF.

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eric Auger <[email protected]>
Tested-by: Eric Auger <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

acpi: move build_fadt() from i386 specific to generic ACPI source

It will be extended and reused by follow up patch for ARM target.

PS:
Since it's generic function now, don't patch FIRMWARE_CTRL, DSDT
fields if they don't point to tables since platform might not
provide them and use X_ variants instead if applicable.

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eric Auger <[email protected]>
Tested-by: Eric Auger <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

pc: acpi: use build_append_foo() API to construct FADT

build_append_foo() API doesn't need explicit endianness
conversions which eliminates a source of errors and
it makes build_fadt() look like declarative definition of
FADT table in ACPI spec, which makes it easy to review.
Also it allows easily extending FADT to support other
revisions which will be used by follow up patches
where build_fadt() will be reused for ARM target.

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eric Auger <[email protected]>
Tested-by: Eric Auger <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

pc: acpi: isolate FADT specific data into AcpiFadtData structure

move FADT data initialization out of fadt_setup() into dedicated
init_fadt_data() that will set common for pc/q35 values in
AcpiFadtData structure and acpi_get_pm_info() will complement
it with pc/q35 specific values initialization.

That will allow to get rid of fadt_setup() and generalize
build_fadt() so it could be easily extended for rev5 and
reused by ARM target.

While at it also move facs/dsdt/xdsdt offsets from build_fadt()
arg list into AcpiFadtData, as they belong to the same dataset.

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eric Auger <[email protected]>
Tested-by: Eric Auger <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

acpi: move ACPI_PORT_SMI_CMD define to header it belongs to

ACPI_PORT_SMI_CMD is alias for APM_CNT_IOPORT,
so make it really one instead of duplicating its value.

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eric Auger <[email protected]>
Tested-by: Eric Auger <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

acpi: add build_append_gas() helper for Generic Address Structure

it will help to add Generic Address Structure to ACPI tables
without using packed C structures and avoid endianness
issues as API doesn't need an explicit conversion.

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eric Auger <[email protected]>
Tested-by: Eric Auger <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

acpi: reuse AcpiGenericAddress instead of Acpi20GenericAddress

Drop duplicate in form of Acpi20GenericAddress and reuse
AcpiGenericAddress.

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eric Auger <[email protected]>
Tested-by: Eric Auger <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

pc: replace pm object initialization with one-liner in acpi_get_pm_info()

next patch will need it before it gets to piix4/lpc branches
that initializes 'obj' now.

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eric Auger <[email protected]>
Reviewed-by: Philippe Mathieu-Daudé <[email protected]>
Tested-by: Eric Auger <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

acpi: remove unused acpi-dsdt.aml

SeaBIOS blob which is currently shipped with QEMU
doesn't need acpi-dsdt.aml nor is able to use it
and code that loaded it in QEMU was removed by
(commit 9fb7aaaf4c "pc: drop external DSDT loading")
in 2013.

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Gerd Hoffmann <[email protected]>
Reviewed-by: Eric Auger <[email protected]>
Tested-by: Eric Auger <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

virtio-net: add linkspeed and duplex settings to virtio-net

Although linkspeed and duplex can be set in a linux guest via 'ethtool -s',
this requires custom ethtool commands for virtio-net by default.

Introduce a new feature flag, VIRTIO_NET_F_SPEED_DUPLEX, which allows
the hypervisor to export a linkspeed and duplex setting. The user can
subsequently overwrite it later if desired via: 'ethtool -s'.

Linkspeed and duplex settings can be set as:
'-device virtio-net,speed=10000,duplex=full'

where speed is [0...INT_MAX], and duplex is ["half"|"full"].

Signed-off-by: Jason Baron <[email protected]>
Cc: "Michael S. Tsirkin" <[email protected]>
Cc: Jason Wang <[email protected]>
Cc: [email protected]
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

virtio-net: use 64-bit values for feature flags

In prepartion for using some of the high order feature bits, make sure that
virtio-net uses 64-bit values everywhere.

Signed-off-by: Jason Baron <[email protected]>
Cc: "Michael S. Tsirkin" <[email protected]>
Cc: Jason Wang <[email protected]>
Cc: [email protected]
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

scripts/update-linux-headers: add ethtool.h and update to 4.16.0-rc4

A subsequent patch to add support for setting linkspeed/duplex in
virtio-net, requires a few definitions from ethtool.h, which ends up
pulling in kernel.h and sysinfo.h as well.

Signed-off-by: Jason Baron <[email protected]>
Cc: "Michael S. Tsirkin" <[email protected]>
Cc: Jason Wang <[email protected]>
Cc: [email protected]

Merge remote-tracking branch 'remotes/ehabkost/tags/python-next-pull-request' into staging

Python queue, 2018-03-12

# gpg: Signature made Mon 12 Mar 2018 22:10:36 GMT
# gpg:                using RSA key 2807936F984DC5A6
# gpg: Good signature from "Eduardo Habkost <[email protected]>"
# Primary key fingerprint: 5A32 2FD5 ABC4 D3DB ACCF  D1AA 2807 936F 984D C5A6

* remotes/ehabkost/tags/python-next-pull-request:
  device-crash-test: Use 'python' binary
  qmp.py: Encode json data before sending
  qemu.py: Use items() instead of iteritems()
  device-crash-test: New known crashes

Signed-off-by: Peter Maydell <[email protected]>

Merge remote-tracking branch 'remotes/ehabkost/tags/x86-next-pull-request' into staging

x86 queue, 2018-03-12

* Intel Processor Trace support
* KVM_HINTS_DEDICATED

# gpg: Signature made Mon 12 Mar 2018 19:58:39 GMT
# gpg:                using RSA key 2807936F984DC5A6
# gpg: Good signature from "Eduardo Habkost <[email protected]>"
# Primary key fingerprint: 5A32 2FD5 ABC4 D3DB ACCF  D1AA 2807 936F 984D C5A6

* remotes/ehabkost/tags/x86-next-pull-request:
  i386: Add support to get/set/migrate Intel Processor Trace feature
  i386: Add Intel Processor Trace feature support
  target-i386: add KVM_HINTS_DEDICATED performance hint

Signed-off-by: Peter Maydell <[email protected]>

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

docker patches

# gpg: Signature made Mon 12 Mar 2018 17:25:57 GMT
# gpg:                using RSA key CA35624C6A9171C6
# gpg: Good signature from "Fam Zheng <[email protected]>"
# Primary key fingerprint: 5003 7CB7 9706 0F76 F021  AD56 CA35 624C 6A91 71C6

* remotes/famz/tags/staging-pull-request:
  tests: make docker-test-debug@fedora run sanitizers

Signed-off-by: Peter Maydell <[email protected]>

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

# gpg: Signature made Mon 12 Mar 2018 16:01:16 GMT
# gpg:                using RSA key 9CA4ABB381AB73C8
# gpg: Good signature from "Stefan Hajnoczi <[email protected]>"
# gpg:                 aka "Stefan Hajnoczi <[email protected]>"
# Primary key fingerprint: 8695 A8BF D3F9 7CDA AC35  775A 9CA4 ABB3 81AB 73C8

* remotes/stefanha/tags/block-pull-request:
  block: make BDRV_POLL_WHILE() re-entrancy safe

Signed-off-by: Peter Maydell <[email protected]>

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

# gpg: Signature made Mon 12 Mar 2018 15:59:54 GMT
# gpg:                using RSA key 9CA4ABB381AB73C8
# gpg: Good signature from "Stefan Hajnoczi <[email protected]>"
# gpg:                 aka "Stefan Hajnoczi <[email protected]>"
# Primary key fingerprint: 8695 A8BF D3F9 7CDA AC35  775A 9CA4 ABB3 81AB 73C8

* remotes/stefanha/tags/tracing-pull-request:
  trace: only permit standard C types and fixed size integer types
  trace: remove use of QEMU specific types from trace probes
  trace: include filename when printing parser error messages
  simpletrace: fix timestamp argument type
  log-for-trace.h: Split out parts of log.h used by trace.h

Signed-off-by: Peter Maydell <[email protected]>

device-crash-test: Use 'python' binary

Now the script works with Python 3, so we can use the 'python'
binary provided by the system.

Signed-off-by: Eduardo Habkost <[email protected]>
Message-Id: <20180312185503[email protected]>
Reviewed-by: Daniel P. Berrangé <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

qmp.py: Encode json data before sending

On Python 3, json.dumps() return a str object, which can't be
sent directly through a socket and must be encoded into a bytes
object.  Use .encode('utf-8'), which will work on both Python 2
and Python 3.

Signed-off-by: Eduardo Habkost <[email protected]>
Message-Id: <20180312185503[email protected]>
Reviewed-by: Daniel P. Berrangé <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

qemu.py: Use items() instead of iteritems()

items() is less efficient on Python 2.x, but makes the code work
on both Python 2 and Python 3.

Cc: Lukáš Doktor <[email protected]>
Cc: Philippe Mathieu-Daudé <[email protected]>
Cc: Cleber Rosa <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>
Message-Id: <20180312185503[email protected]>
Reviewed-by: Daniel P. Berrangé <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

device-crash-test: New known crashes

We are not running the script on "make check" yet, and additional
bugs were introduced recently in the tree.

Whitelist the new crashes while we investigate, to allow us to
run device-crash-test on "make check" as soon as possible to
prevent new bugs.

Cc: Pavel Pisa <[email protected]>
Cc: John Snow <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>
Message-Id: <20180309202827[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

Merge remote-tracking branch 'remotes/kraxel/tags/usb-20180312-pull-request' into staging

usbredir: reorder fields in USBRedirDevice to reduce padding

# gpg: Signature made Mon 12 Mar 2018 11:05:19 GMT
# gpg:                using RSA key 4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg:                 aka "Gerd Hoffmann <[email protected]>"
# gpg:                 aka "Gerd Hoffmann (private) <[email protected]>"
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/usb-20180312-pull-request:
  usbredir: reorder fields in USBRedirDevice to reduce padding

Signed-off-by: Peter Maydell <[email protected]>

i386: Add support to get/set/migrate Intel Processor Trace feature

Add Intel Processor Trace related definition. It also add
corresponding part to kvm_get/set_msr and vmstate.

Signed-off-by: Chao Peng <[email protected]>
Signed-off-by: Luwei Kang <[email protected]>
Message-Id: <1520182116[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

i386: Add Intel Processor Trace feature support

Expose Intel Processor Trace feature to guest.

To make Intel PT live migration safe and get same CPUID information
with same CPU model on diffrent host. CPUID[14] is constant in this
patch. Intel PT use EPT is first supported in IceLake, the CPUID[14]
get on this machine as default value. Intel PT would be disabled
if any machine don't support this minial feature list.

Signed-off-by: Chao Peng <[email protected]>
Signed-off-by: Luwei Kang <[email protected]>
Message-Id: <1520182116[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

target-i386: add KVM_HINTS_DEDICATED performance hint

Add KVM_HINTS_DEDICATED performance hint, guest checks this feature bit
to determine if they run on dedicated vCPUs, allowing optimizations such
as usage of qspinlocks.

Cc: Paolo Bonzini <[email protected]>
Cc: Radim Krčmář <[email protected]>
Cc: Eduardo Habkost <[email protected]>
Signed-off-by: Wanpeng Li <[email protected]>
Message-Id: <1518185725[email protected]>
[ehabkost: Renamed property to kvm-hint-dedicated]
Signed-off-by: Eduardo Habkost <[email protected]>

Merge remote-tracking branch 'remotes/kraxel/tags/vga-20180312-pull-request' into staging

7cdc61becd vga: fix region calculation

# gpg: Signature made Mon 12 Mar 2018 10:59:24 GMT
# gpg:                using RSA key 4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg:                 aka "Gerd Hoffmann <[email protected]>"
# gpg:                 aka "Gerd Hoffmann (private) <[email protected]>"
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/vga-20180312-pull-request:
  vga: fix region calculation

Signed-off-by: Peter Maydell <[email protected]>

tests: make docker-test-debug@fedora run sanitizers

Since --enable-debug no longer enable sanitizers, we need explicit
--enable-sanitizers.

llvm package is required for llvm-symbolizer, to get symbols in
backtraces.

Add make V=1 to get details about failing tests.

Signed-off-by: Marc-André Lureau <[email protected]>
Message-Id: <20180312120849[email protected]>
Signed-off-by: Fam Zheng <[email protected]>

Merge remote-tracking branch 'remotes/kraxel/tags/audio-20180312-pull-request' into staging

modules: use gmodule-export.
audio: add driver registry, enable module builds.

# gpg: Signature made Mon 12 Mar 2018 10:42:19 GMT
# gpg:                using RSA key 4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg:                 aka "Gerd Hoffmann <[email protected]>"
# gpg:                 aka "Gerd Hoffmann (private) <[email protected]>"
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/audio-20180312-pull-request:
  audio/sdl: build as module
  audio/pulseaudio: build as module
  audio/oss: build as module
  audio/alsa: build as module
  build: enable audio modules
  audio: add module loading support
  audio: add driver registry
  modules: use gmodule-export

Signed-off-by: Peter Maydell <[email protected]>

Merge remote-tracking branch 'remotes/kraxel/tags/ui-20180312-pull-request' into staging

gtk,spice: add dmabuf support.
sdl,vnc,gtk: bugfixes.
ui/qapi: add device ID and head parameters to screendump.
build: try improve handling of clang warnings.

# gpg: Signature made Mon 12 Mar 2018 09:13:28 GMT
# gpg:                using RSA key 4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg:                 aka "Gerd Hoffmann <[email protected]>"
# gpg:                 aka "Gerd Hoffmann (private) <[email protected]>"
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/ui-20180312-pull-request:
  qapi: Add device ID and head parameters to screendump
  spice: add cursor_dmabuf support
  spice: add scanout_dmabuf support
  spice: drop dprint() debug logging
  vnc: deal with surface NULL pointers
  ui/gtk-egl: add cursor_dmabuf support
  ui/gtk-egl: add scanout_dmabuf support
  ui/gtk: use GtkGlArea on wayland only
  ui/opengl: Makefile cleanup
  ui/gtk: group gtk.mo declarations in Makefile
  ui/gtk: make GtkGlArea usage a runtime option
  sdl: workaround bug in sdl 2.0.8 headers
  make: switch language file build to be gtk module aware
  build: try improve handling of clang warnings

Signed-off-by: Peter Maydell <[email protected]>

Merge remote-tracking branch 'remotes/dgilbert/tags/pull-migration-20180309a' into staging

Migration pull 2018-03-09

# gpg: Signature made Fri 09 Mar 2018 17:52:46 GMT
# gpg:                using RSA key 0516331EBC5BFDE7
# gpg: Good signature from "Dr. David Alan Gilbert (RH2) <[email protected]>"
# Primary key fingerprint: 45F5 C71B 4A0C B7FB 977A  9FA9 0516 331E BC5B FDE7

* remotes/dgilbert/tags/pull-migration-20180309a:
  tests: Silence migration-test 'bad' test
  migration: fix applying wrong capabilities
  migration/block: rename MAX_INFLIGHT_IO to MAX_IO_BUFFERS
  migration/block: reset dirty bitmap before read in bulk phase
  migration: do not transfer ram during bulk storage migration
  migration: fix minor finalize leak

Signed-off-by: Peter Maydell <[email protected]>

Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20180309' into staging

target-arm queue:
 * i.MX: Add i.MX7 SOC implementation and i.MX7 Sabre board
 * Report the correct core count in A53 L2CTLR on the ZynqMP board
 * linux-user: preliminary SVE support work (signal handling)
 * hw/arm/boot: fix memory leak in case of error loading ELF file
 * hw/arm/boot: avoid reading off end of buffer if passed very
   small image file
 * hw/arm: Use more CONFIG switches for the object files
 * target/arm: Add "-cpu max" support
 * hw/arm/virt: Support -machine gic-version=max
 * hw/sd: improve debug tracing
 * hw/sd: sdcard: Add the Tuning Command (CMD 19)
 * MAINTAINERS: add Philippe as odd-fixes maintainer for SD

# gpg: Signature made Fri 09 Mar 2018 17:24:23 GMT
# gpg:                using RSA key 3C2525ED14360CDE
# gpg: Good signature from "Peter Maydell <[email protected]>"
# gpg:                 aka "Peter Maydell <[email protected]>"
# gpg:                 aka "Peter Maydell <[email protected]>"
# Primary key fingerprint: E1A5 C593 CD41 9DE2 8E83  15CF 3C25 25ED 1436 0CDE

* remotes/pmaydell/tags/pull-target-arm-20180309: (25 commits)
  MAINTAINERS: Add entries for SD (SDHCI, SDBus, SDCard)
  sdhci: Fix a typo in comment
  sdcard: Add the Tuning Command (CMD19)
  sdcard: Display which protocol is used when tracing (SD or SPI)
  sdcard: Display command name when tracing CMD/ACMD
  sdcard: Do not trace CMD55, except when we already expect an ACMD
  hw/arm/virt: Support -machine gic-version=max
  hw/arm/virt: Add "max" to the list of CPU types "virt" supports
  target/arm: Make 'any' CPU just an alias for 'max'
  target/arm: Add "-cpu max" support
  target/arm: Move definition of 'host' cpu type into cpu.c
  target/arm: Query host CPU features on-demand at instance init
  arm: avoid heap-buffer-overflow in load_aarch64_image
  arm: fix load ELF error leak
  hw/arm: Use more CONFIG switches for the object files
  aarch64-linux-user: Add support for SVE signal frame records
  aarch64-linux-user: Add support for EXTRA signal frame records
  aarch64-linux-user: Remove struct target_aux_context
  aarch64-linux-user: Split out helpers for guest signal handling
  linux-user: Implement aarch64 PR_SVE_SET/GET_VL
  ...

Signed-off-by: Peter Maydell <[email protected]>

trace: only permit standard C types and fixed size integer types

Some trace backends will compile code based on the declared trace
events. It should not be assumed that the backends can resolve any QEMU
specific typedefs. So trace events should restrict their argument
types to the standard C types and fixed size integer types. Any complex
pointer types can be declared as "void *" for purposes of trace events,
since nothing will be dereferencing these pointer arguments.

Signed-off-by: Daniel P. Berrangé <[email protected]>
Message-id: 20180308155524[email protected]
Signed-off-by: Stefan Hajnoczi <[email protected]>

trace: remove use of QEMU specific types from trace probes

Any compound structs / unions / etc, should always be declared as
'void *' pointers, since it cannot be assumed that trace backends
are able to resolve QEMU typedefs.

Signed-off-by: Daniel P. Berrangé <[email protected]>
Message-id: 20180308155524[email protected]
Signed-off-by: Stefan Hajnoczi <[email protected]>

trace: include filename when printing parser error messages

Improves error messages from:

  ValueError: Error on line 72: need more than 1 value to unpack

To

  ValueError: Error at /home/berrange/src/virt/qemu/trace-events:72:
    need more than 1 value to unpack

Signed-off-by: Daniel P. Berrangé <[email protected]>
Reviewed-by: Philippe Mathieu-Daudé <[email protected]>
Message-id: 20180306154650[email protected]
Signed-off-by: Stefan Hajnoczi <[email protected]>

simpletrace: fix timestamp argument type

The timestamp argument to a trace event method is documented as follows:

  The method can also take a timestamp argument before the trace event
  arguments:

    def runstate_set(self, timestamp, new_state):
        ...

  Timestamps have the uint64_t type and are in nanoseconds.

In reality methods with a timestamp argument actually receive a tuple
like (123456789,) as the timestamp argument.  This is due to a bug in
simpletrace.py.

This patch unpacks the tuple so that methods receive the correct
timestamp argument type.

Signed-off-by: Stefan Hajnoczi <[email protected]>
Message-id: 20180222163901[email protected]
Signed-off-by: Stefan Hajnoczi <[email protected]>

log-for-trace.h: Split out parts of log.h used by trace.h

A persistent build problem we see is where a source file
accidentally omits the #include of log.h. This slips through
local developer testing because if you configure with the
default (log) trace backend trace.h will pull in log.h for you.
Compilation fails only if some other backend is selected.

To make this error cause a compile failure regardless of
the configured trace backend, split out the parts of log.h
that trace.h requires into a new log-for-trace.h header.
Since almost all manual uses of the log.h functions will
use constants or functions which aren't in log-for-trace.h,
this will let us catch missing #include "qemu/log.h" more
consistently.

Signed-off-by: Peter Maydell <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Richard Henderson <[email protected]>
Message-id: 20180213140029[email protected]
Signed-off-by: Stefan Hajnoczi <[email protected]>

block: make BDRV_POLL_WHILE() re-entrancy safe

Nested BDRV_POLL_WHILE() calls can occur.  Currently
assert(!wait_->wakeup) fails in AIO_WAIT_WHILE() when this happens.

This patch converts the bool wait_->need_kick flag to an unsigned
wait_->num_waiters counter.

Nesting works correctly because outer AIO_WAIT_WHILE() callers evaluate
the condition again after the inner caller completes (invoking the inner
caller counts as aio_poll() progress).

Reported-by: "fuweiwei (C)" <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>
Message-id: 20180307124619[email protected]
Cc: Paolo Bonzini <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

vga: fix region calculation

Typically the scanline length and the line offset are identical.  But
in case they are not our calculation for region_end is incorrect.  Using
line_offset is fine for all scanlines, except the last one where we have
to use the actual scanline length.

Fixes: CVE-2018-7550
Reported-by: Ross Lagerwall <[email protected]>
Signed-off-by: Gerd Hoffmann <[email protected]>
Reviewed-by: Prasad J Pandit <[email protected]>
Tested-by: Ross Lagerwall <[email protected]>
Message-id: 20180309143704[email protected]

usbredir: reorder fields in USBRedirDevice to reduce padding

Changing the current ordering saves 8 bytes per entry in x86_64.

Signed-off-by: zhenwei.pi <[email protected]>
Message-id: 1520318781[email protected]
Signed-off-by: Gerd Hoffmann <[email protected]>

audio/sdl: build as module

Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 20180306074053[email protected]

audio/pulseaudio: build as module

Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 20180306074053[email protected]

audio/oss: build as module

Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 20180306074053[email protected]

audio/alsa: build as module

Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 20180306074053[email protected]

build: enable audio modules

Add audio/ to common-obj-m variable.

Also run both audio and ui variables through unnest-vars.
This avoids sdl.mo (exists in both audio/ and ui/) name clashes.

Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 20180306074053[email protected]

audio: add module loading support

Make audio_driver_lookup() try load the module in case it doesn't find
the driver in the registry.  Also load all modules for -audio-help, so
the help output includes the help text for modular audio drivers.

Signed-off-by: Gerd Hoffmann <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Message-id: 20180306074053[email protected]

audio: add driver registry

Add registry for audio drivers, using the existing audio_driver struct.
Make all drivers register themself.  The old list of audio_driver struct
pointers is now a list of audio driver names, specifying the priority
(aka probe order) in case no driver is explicitly asked for.

Signed-off-by: Gerd Hoffmann <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Message-id: 20180306074053[email protected]

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

Block layer patches

# gpg: Signature made Fri 09 Mar 2018 15:09:20 GMT
# gpg:                using RSA key 7F09B272C88F2FD6
# gpg: Good signature from "Kevin Wolf <[email protected]>"
# Primary key fingerprint: DC3D EB15 9A9A F95D 3D74  56FE 7F09 B272 C88F 2FD6

* remotes/kevin/tags/for-upstream: (56 commits)
  qemu-iotests: fix 203 migration completion race
  iotests: Tweak 030 in order to trigger a race condition with parallel jobs
  iotests: Skip test for ENOMEM error
  iotests: Mark all tests executable
  iotests: Test creating overlay when guest running
  qemu-iotests: Test ssh image creation over QMP
  qemu-iotests: Test qcow2 over file image creation with QMP
  block: Fail bdrv_truncate() with negative size
  file-posix: Fix no-op bdrv_truncate() with falloc preallocation
  ssh: Support .bdrv_co_create
  ssh: Pass BlockdevOptionsSsh to connect_to_ssh()
  ssh: QAPIfy host-key-check option
  ssh: Use QAPI BlockdevOptionsSsh object
  sheepdog: Support .bdrv_co_create
  sheepdog: QAPIfy "redundancy" create option
  nfs: Support .bdrv_co_create
  nfs: Use QAPI options in nfs_client_open()
  rbd: Use qemu_rbd_connect() in qemu_rbd_do_create()
  rbd: Assign s->snap/image_name in qemu_rbd_open()
  rbd: Support .bdrv_co_create
  ...

Signed-off-by: Peter Maydell <[email protected]>

modules: use gmodule-export

As we want qemu symbols be exported to modules we should use the
gmodule-export-2.0 pkg-config instead of gmodule-2.0.

Cc: Marc-André Lureau <[email protected]>
Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 20180308085301[email protected]

qapi: Add device ID and head parameters to screendump

QEMU's screendump command can only take dumps from the primary display.
When using multiple VGA cards, there is no way to get a dump from a
secondary card or other display heads yet. So let's add a 'device' and
a 'head' parameter to the HMP and QMP commands to be able to specify
alternative devices and heads with the screendump command, too.

Reviewed-by: Daniel P. Berrangé <[email protected]>
Acked-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Gerd Hoffmann <[email protected]>
Signed-off-by: Thomas Huth <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Message-id: 1520267868[email protected]
Signed-off-by: Gerd Hoffmann <[email protected]>

spice: add cursor_dmabuf support

Add support for cursor dmabufs.  qemu has to render the cursor for
that, so in case a cursor is present qemu allocates a new dmabuf, blits
the scanout, blends in the pointer and passes on the new dmabuf to
spice-server.  Without cursor qemu continues to simply pass on the
scanout dmabuf as-is.

Signed-off-by: Gerd Hoffmann <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Message-id: 20180308090618[email protected]

spice: add scanout_dmabuf support

Add support for scanout dmabufs.  Just
pass them through to spice-server.

Signed-off-by: Gerd Hoffmann <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Message-id: 20180308090618[email protected]

spice: drop dprint() debug logging

Some calls are deleted, some are converted into tracepoints.

Signed-off-by: Gerd Hoffmann <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>
Message-id: 20180308090618[email protected]

vnc: deal with surface NULL pointers

Secondary displays in multihead setups are allowed to have a NULL
DisplaySurface.  Typically user interfaces handle this by hiding the
window which shows the display in question.

This isn't an option for vnc though because it simply hasn't a concept
of windows or outputs.  So handle the situation by showing a placeholder
DisplaySurface instead.  Also check in console_select whenever a surface
is preset in the first place before requesting an update.

This fixes a segfault which can be triggered by switching to an unused
display (via vtrl-alt-<nr>) in a multihead setup, for example using
-device virtio-vga,max_outputs=2.

Cc: Christian Borntraeger <[email protected]>
Signed-off-by: Gerd Hoffmann <[email protected]>
Reviewed-by: Philippe Mathieu-Daudé <[email protected]>
Tested-by: Christian Borntraeger <[email protected]>
Message-id: 20180308161803[email protected]

ui/gtk-egl: add cursor_dmabuf support

Add support for cursor dmabufs to gtk-egl.  Just blend in the cursor
(if we have one) when rendering the dmabuf.

Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 20180306090951[email protected]

ui/gtk-egl: add scanout_dmabuf support

Add support for dmabuf scanouts to gtk-egl.

Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 20180306090951[email protected]

ui/gtk: use GtkGlArea on wayland only

For dma-buf support we need a egl context.  The gtk x11 backend uses glx
contexts though.  We can't use the GtkGlArea widget on x11 because of
that, so use our own gtk-egl code instead.  wayland continues to use
the GtkGlArea widget.

Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 20180306090951[email protected]

ui/opengl: Makefile cleanup

With gtk.mo bits moved away we don't need the ifeq any more.
Also add missing opengl libs for some objects.

Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 20180306090951[email protected]

ui/gtk: group gtk.mo declarations in Makefile

Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 20180306090951[email protected]

ui/gtk: make GtkGlArea usage a runtime option

Compile in both gtk-egl and gtk-gl-area, then allow to choose at runtime
instead of compile time which opengl variant we want use.

Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 20180306090951[email protected]

sdl: workaround bug in sdl 2.0.8 headers

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=892087

Signed-off-by: Gerd Hoffmann <[email protected]>
Reviewed-by: Daniel P. Berrangé <[email protected]>
Message-id: 20180307154258[email protected]

make: switch language file build to be gtk module aware

Now that gtk support builds as a module, CONFIG_GTK changed from
y to m. Adjust Makefile correspondingly.

Signed-off-by: Bruce Rogers <[email protected]>
Message-id: 20180307155517[email protected]
Signed-off-by: Gerd Hoffmann <[email protected]>

build: try improve handling of clang warnings

This patch disables the pragma diagnostic -Wunused-but-set-variable for
clang in util/coroutine-ucontext.c.

This in turn allows us to remove it from the configure check, so the
CONFIG_PRAGMA_DIAGNOSTIC_AVAILABLE will succeed for clang.

With that in place clang builds (linux) will use -Werror by default,
which breaks the build due to warning about unaligned struct members.

Just turning off this warning isn't a good idea as it indicates
portability problems.  So make it a warning again, using
-Wno-error=address-of-packed-member.  That way it doesn't break the
build but still shows up in the logs.

Now clang builds qemu without errors.  Well, almost.  There are some
left in the rdma code.  Leaving that to the rdma people.  All others can
use --disable-rdma to workarounds this.

Cc: Peter Maydell <[email protected]>
Signed-off-by: Gerd Hoffmann <[email protected]>
Message-Id: <20180309135945[email protected]>
Reviewed-by: Thomas Huth <[email protected]>
Reviewed-by: Philippe Mathieu-Daudé <[email protected]>

Merge remote-tracking branch 'remotes/vivier/tags/m68k-for-2.12-pull-request' into staging

# gpg: Signature made Fri 09 Mar 2018 14:54:33 GMT
# gpg:                using RSA key F30C38BD3F2FBE3C
# gpg: Good signature from "Laurent Vivier <[email protected]>"
# gpg:                 aka "Laurent Vivier <[email protected]>"
# gpg:                 aka "Laurent Vivier (Red Hat) <[email protected]>"
# Primary key fingerprint: CD2F 75DD C8E3 A4DC 2E4F  5173 F30C 38BD 3F2F BE3C

* remotes/vivier/tags/m68k-for-2.12-pull-request:
  target/m68k: implement ftentox
  target/m68k: implement ftwotox
  target/m68k: implement fetox
  target/m68k: implement flog2
  target/m68k: implement flog10
  target/m68k: implement flogn
  target/m68k: implement flognp1
  target/m68k: define floatx80_move()

Signed-off-by: Peter Maydell <[email protected]>

tests: Silence migration-test 'bad' test

In 2c9bb29703c I added a migration test that purposely fails;
unfortunately it prints a copy of the failure message to stderr
which makes the output a bit messy.

Hide stderr for that test.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Message-Id: <20180306173042[email protected]>
Reviewed-by: Peter Xu <[email protected]>
Tested-by: Peter Xu <[email protected]>
Signed-off-by: Dr. David Alan Gilbert <[email protected]>

migration: fix applying wrong capabilities

When setting migration capabilities via QMP/HMP, we'll apply them even
if the capability check failed.  Fix it.

Fixes: 4a84214ebe ("migration: provide migrate_caps_check()", 2017-07-18)
Signed-off-by: Peter Xu <[email protected]>
Message-Id: <20180305094938[email protected]>
Reviewed-by: Dr. David Alan Gilbert <[email protected]>
Signed-off-by: Dr. David Alan Gilbert <[email protected]>

migration/block: rename MAX_INFLIGHT_IO to MAX_IO_BUFFERS

this actually limits (as the original commit mesage suggests) the
number of I/O buffers that can be allocated and not the number
of parallel (inflight) I/O requests.

Signed-off-by: Peter Lieven <[email protected]>
Message-Id: <1520507908[email protected]>
Reviewed-by: Juan Quintela <[email protected]>
Signed-off-by: Dr. David Alan Gilbert <[email protected]>

migration/block: reset dirty bitmap before read in bulk phase

Reset the dirty bitmap before reading to make sure we don't miss
any new data.

Cc: [email protected]
Signed-off-by: Peter Lieven <[email protected]>
Message-Id: <1520507908[email protected]>
Reviewed-by: Juan Quintela <[email protected]>
Signed-off-by: Dr. David Alan Gilbert <[email protected]>

migration: do not transfer ram during bulk storage migration

this patch makes the bulk phase of a block migration to take
place before we start transferring ram. As the bulk block migration
can take a long time its pointless to transfer ram during that phase.

Signed-off-by: Peter Lieven <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>
Message-Id: <1520507908[email protected]>
Reviewed-by: Juan Quintela <[email protected]>
Signed-off-by: Dr. David Alan Gilbert <[email protected]>

migration: fix minor finalize leak

Spotted thanks to ASAN:
QTEST_QEMU_BINARY=x86_64-softmmu/qemu-system-x86_64 tests/migration-test -p /x86_64/migration/bad_dest

==30302==ERROR: LeakSanitizer: detected memory leaks

Direct leak of 48 byte(s) in 1 object(s) allocated from:
    #0 0x7f60efba1a38 in __interceptor_calloc (/lib64/libasan.so.4+0xdea38)
    #1 0x7f60eef3cf75 in g_malloc0 ../glib/gmem.c:124
    #2 0x55ca9094702c in error_copy /home/elmarco/src/qemu/util/error.c:203
    #3 0x55ca9037a30f in migrate_set_error /home/elmarco/src/qemu/migration/migration.c:1139
    #4 0x55ca9037a462 in migrate_fd_error /home/elmarco/src/qemu/migration/migration.c:1150
    #5 0x55ca9038162b in migrate_fd_connect /home/elmarco/src/qemu/migration/migration.c:2411
    #6 0x55ca90386e41 in migration_channel_connect /home/elmarco/src/qemu/migration/channel.c:81
    #7 0x55ca9038335e in socket_outgoing_migration /home/elmarco/src/qemu/migration/socket.c:85
    #8 0x55ca9083dd3a in qio_task_complete /home/elmarco/src/qemu/io/task.c:142
    #9 0x55ca9083d6cc in gio_task_thread_result /home/elmarco/src/qemu/io/task.c:88
    #10 0x7f60eef37317 in g_idle_dispatch ../glib/gmain.c:5552
    #11 0x7f60eef3490b in g_main_dispatch ../glib/gmain.c:3182
    #12 0x7f60eef357ac in g_main_context_dispatch ../glib/gmain.c:3847
    #13 0x55ca90927231 in glib_pollfds_poll /home/elmarco/src/qemu/util/main-loop.c:214
    #14 0x55ca90927420 in os_host_main_loop_wait /home/elmarco/src/qemu/util/main-loop.c:261
    #15 0x55ca909275fa in main_loop_wait /home/elmarco/src/qemu/util/main-loop.c:515
    #16 0x55ca8fc1c2a4 in main_loop /home/elmarco/src/qemu/vl.c:1942
    #17 0x55ca8fc2eb3a in main /home/elmarco/src/qemu/vl.c:4724
    #18 0x7f60e4082009 in __libc_start_main (/lib64/libc.so.6+0x21009)

Indirect leak of 45 byte(s) in 1 object(s) allocated from:
    #0 0x7f60efba1850 in malloc (/lib64/libasan.so.4+0xde850)
    #1 0x7f60eef3cf0c in g_malloc ../glib/gmem.c:94
    #2 0x7f60eef3d1cf in g_malloc_n ../glib/gmem.c:331
    #3 0x7f60eef596eb in g_strdup ../glib/gstrfuncs.c:363
    #4 0x55ca90947085 in error_copy /home/elmarco/src/qemu/util/error.c:204
    #5 0x55ca9037a30f in migrate_set_error /home/elmarco/src/qemu/migration/migration.c:1139
    #6 0x55ca9037a462 in migrate_fd_error /home/elmarco/src/qemu/migration/migration.c:1150
    #7 0x55ca9038162b in migrate_fd_connect /home/elmarco/src/qemu/migration/migration.c:2411
    #8 0x55ca90386e41 in migration_channel_connect /home/elmarco/src/qemu/migration/channel.c:81
    #9 0x55ca9038335e in socket_outgoing_migration /home/elmarco/src/qemu/migration/socket.c:85
    #10 0x55ca9083dd3a in qio_task_complete /home/elmarco/src/qemu/io/task.c:142
    #11 0x55ca9083d6cc in gio_task_thread_result /home/elmarco/src/qemu/io/task.c:88
    #12 0x7f60eef37317 in g_idle_dispatch ../glib/gmain.c:5552
    #13 0x7f60eef3490b in g_main_dispatch ../glib/gmain.c:3182
    #14 0x7f60eef357ac in g_main_context_dispatch ../glib/gmain.c:3847
    #15 0x55ca90927231 in glib_pollfds_poll /home/elmarco/src/qemu/util/main-loop.c:214
    #16 0x55ca90927420 in os_host_main_loop_wait /home/elmarco/src/qemu/util/main-loop.c:261
    #17 0x55ca909275fa in main_loop_wait /home/elmarco/src/qemu/util/main-loop.c:515
    #18 0x55ca8fc1c2a4 in main_loop /home/elmarco/src/qemu/vl.c:1942
    #19 0x55ca8fc2eb3a in main /home/elmarco/src/qemu/vl.c:4724
    #20 0x7f60e4082009 in __libc_start_main (/lib64/libc.so.6+0x21009)

Signed-off-by: Marc-André Lureau <[email protected]>
Message-Id: <20180306170959[email protected]>
Reviewed-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Philippe Mathieu-Daudé <[email protected]>
Signed-off-by: Dr. David Alan Gilbert <[email protected]>

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

# gpg: Signature made Fri 09 Mar 2018 13:19:02 GMT
# gpg:                using RSA key 9CA4ABB381AB73C8
# gpg: Good signature from "Stefan Hajnoczi <[email protected]>"
# gpg:                 aka "Stefan Hajnoczi <[email protected]>"
# Primary key fingerprint: 8695 A8BF D3F9 7CDA AC35  775A 9CA4 ABB3 81AB 73C8

* remotes/stefanha/tags/block-pull-request:
  vl: introduce vm_shutdown()
  virtio-scsi: fix race between .ioeventfd_stop() and vq handler
  virtio-blk: fix race between .ioeventfd_stop() and vq handler
  block: add aio_wait_bh_oneshot()
  virtio-blk: dataplane: Don't batch notifications if EVENT_IDX is present
  README: Fix typo 'git-publish'
  block: Fix qemu crash when using scsi-block

Signed-off-by: Peter Maydell <[email protected]>

MAINTAINERS: Add entries for SD (SDHCI, SDBus, SDCard)

After spending months studying all the different SD Specifications
from the SD Association, voluntarily add myself as maintainer
for the SD code.

Signed-off-by: Philippe Mathieu-Daudé <[email protected]>
Reviewed-by: Peter Maydell <[email protected]>
Message-id: 20180309153654[email protected]
Signed-off-by: Peter Maydell <[email protected]>