OSDN Git Service

qmiga/qemu.git
9 years agolinux-user: Convert blkpg to use a special subop handler
Alexander Graf [Fri, 22 Aug 2014 11:15:50 +0000 (13:15 +0200)]
linux-user: Convert blkpg to use a special subop handler

The blkpg ioctl can take different payloads depending on the opcode in
its payload structure. Create a new special ioctl handler that can only
deal with partition style ones for now.

This patch fixes running parted for me.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
9 years agolinux-user: Enable epoll_pwait syscall for ARM
Peter Maydell [Fri, 19 Sep 2014 01:35:20 +0000 (02:35 +0100)]
linux-user: Enable epoll_pwait syscall for ARM

We have support for the epoll_pwait syscall, but it wasn't enabled for
ARM guests because we hadn't defined the syscall number; correct this
deficiency.

Reported-by: Dave Flogeras <dflogeras2@gmail.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
9 years agogdbstub: Allow target CPUs to specify watchpoint STOP_BEFORE_ACCESS flag
Peter Maydell [Fri, 12 Sep 2014 18:04:17 +0000 (19:04 +0100)]
gdbstub: Allow target CPUs to specify watchpoint STOP_BEFORE_ACCESS flag

GDB assumes that watchpoint set via the gdbstub remote protocol will
behave in the same way as hardware watchpoints for the target. In
particular, whether the CPU stops with the PC before or after the insn
which triggers the watchpoint is target dependent. Allow guest CPU
code to specify which behaviour to use. This fixes a bug where with
guest CPUs which stop before the accessing insn GDB would manually
step forward over what it thought was the insn and end up one insn
further forward than it should be.

We set this flag for the CPU architectures which set
gdbarch_have_nonsteppable_watchpoint in gdb 7.7:
ARM, CRIS, LM32, MIPS and Xtensa.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Tested-by: Max Filippov <jcmvbkbc@gmail.com>
Tested-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Tested-by: Michael Walle <michael@walle.cc> (for lm32)
Message-id: 1410545057-14014-1-git-send-email-peter.maydell@linaro.org

9 years agoMerge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging
Peter Maydell [Mon, 6 Oct 2014 09:59:55 +0000 (10:59 +0100)]
Merge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging

# gpg: Signature made Sat 04 Oct 2014 21:24:46 BST using RSA key ID 81AB73C8
# gpg: Good signature from "Stefan Hajnoczi <stefanha@redhat.com>"
# gpg:                 aka "Stefan Hajnoczi <stefanha@gmail.com>"

* remotes/stefanha/tags/block-pull-request: (23 commits)
  blockdev-test: Test device_del after drive_del
  blockdev-test: Factor out some common code into helpers
  blockdev-test: Simplify by using g_assert_cmpstr()
  blockdev-test: Clean up bogus drive_add argument
  blockdev-test: Use single rather than double quotes in QMP
  drive_del-test: Merge of qdev-monitor-test, blockdev-test
  iotests: qemu-img info output for corrupt image
  qapi: Add corrupt field to ImageInfoSpecificQCow2
  iotests: Use _img_info
  util: Emancipate id_wellformed() from QemuOpts
  q35/ahci: Pick up -cdrom and -hda options
  qtest/bios-tables: Correct Q35 command line
  ide: Update ide_drive_get to be HBA agnostic
  pc/vl: Add units-per-default-bus property
  blockdev: Allow overriding if_max_dev property
  blockdev: Orphaned drive search
  qemu-iotests: Fix supported cache modes for 052
  make check-block: Use default cache modes
  Modify qemu_opt_rename to realize renaming all items in opts
  vmdk: Fix integer overflow in offset calculation
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoblockdev-test: Test device_del after drive_del
Markus Armbruster [Thu, 2 Oct 2014 14:51:36 +0000 (16:51 +0200)]
blockdev-test: Test device_del after drive_del

Executed in this order, drive_del and device_del's automatic drive
deletion take notoriously tricky special paths.

[Fixed "an device" -> "a device" typo as requested by Eric Blake.
--Stefan]

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1412261496-24455-7-git-send-email-armbru@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoblockdev-test: Factor out some common code into helpers
Markus Armbruster [Thu, 2 Oct 2014 14:51:35 +0000 (16:51 +0200)]
blockdev-test: Factor out some common code into helpers

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1412261496-24455-6-git-send-email-armbru@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoblockdev-test: Simplify by using g_assert_cmpstr()
Markus Armbruster [Thu, 2 Oct 2014 14:51:34 +0000 (16:51 +0200)]
blockdev-test: Simplify by using g_assert_cmpstr()

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1412261496-24455-5-git-send-email-armbru@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoblockdev-test: Clean up bogus drive_add argument
Markus Armbruster [Thu, 2 Oct 2014 14:51:33 +0000 (16:51 +0200)]
blockdev-test: Clean up bogus drive_add argument

The first argument should be a PCI address, which pci-addr=auto isn't.
Doesn't really matter, as drive_add ignores its first argument when
its second argument has if=none.  Clean it up anyway.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1412261496-24455-4-git-send-email-armbru@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoblockdev-test: Use single rather than double quotes in QMP
Markus Armbruster [Thu, 2 Oct 2014 14:51:32 +0000 (16:51 +0200)]
blockdev-test: Use single rather than double quotes in QMP

QMP accepts both single and double quotes.  This is the only test
using double quotes.  They need to be quoted in C strings.  Replace
them by single quotes.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1412261496-24455-3-git-send-email-armbru@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agodrive_del-test: Merge of qdev-monitor-test, blockdev-test
Markus Armbruster [Thu, 2 Oct 2014 14:51:31 +0000 (16:51 +0200)]
drive_del-test: Merge of qdev-monitor-test, blockdev-test

Each of qdev-monitor-test and blockdev-test has just one test case,
and both are about drive_del.

[Extended copyright from 2013 to 2013-2014 as requested by Eric Blake.
--Stefan]

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1412261496-24455-2-git-send-email-armbru@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoiotests: qemu-img info output for corrupt image
Max Reitz [Tue, 30 Sep 2014 19:31:29 +0000 (21:31 +0200)]
iotests: qemu-img info output for corrupt image

The "corrupt" entry in the format-specific information section should be
"true".

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1412105489-7681-4-git-send-email-mreitz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoqapi: Add corrupt field to ImageInfoSpecificQCow2
Max Reitz [Tue, 30 Sep 2014 19:31:28 +0000 (21:31 +0200)]
qapi: Add corrupt field to ImageInfoSpecificQCow2

Just like lazy-refcounts, this field will be present iff the qcow2
compat level is 1.1 (or probably any future revision).

As expected, this breaks some tests due to the new field present in
qemu-img info output; so fix their output accordingly.

Suggested-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1412105489-7681-3-git-send-email-mreitz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoiotests: Use _img_info
Max Reitz [Tue, 30 Sep 2014 19:31:27 +0000 (21:31 +0200)]
iotests: Use _img_info

qemu-img info should only be used directly if the format-specific
information or the name of the format is relevant (some tests explicitly
test format-specific information; test 082 uses qcow2-specific settings
to test the qemu-img interface); otherwise, tests should always use
_img_info instead.

Test 082 was touched only partially. It does test the qemu-img
interface; however, its invocations of qemu-img info are not real tests
but rather verifications, so if format-specific information is not
important for the test, there is no reason not to use _img_info. In
contrast to directly invoking qemu-img info, "qcow2" is replaced by
"IMGFMT"; but as "qcow2" is only mentioned once in test 082 (in
_supported_fmt), I consider this an improvement.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1412105489-7681-2-git-send-email-mreitz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoutil: Emancipate id_wellformed() from QemuOpts
Markus Armbruster [Tue, 30 Sep 2014 11:59:30 +0000 (13:59 +0200)]
util: Emancipate id_wellformed() from QemuOpts

IDs have long spread beyond QemuOpts: not everything with an ID
necessarily goes through QemuOpts.  Commit 9aebf3b is about such a
case: block layer names are meant to be well-formed IDs, but some of
them don't go through QemuOpts, and thus weren't checked.  The commit
fixed that the straightforward way: rename the internal QemuOpts
helper id_wellformed() to qemu_opts_id_wellformed() and give it
external linkage.

Instead of using it directly in block.c, the commit adds wrapper
bdrv_is_valid_name(), probably to hide the connection to QemuOpts.

Go one logical step further: emancipate IDs from QemuOpts.  Rename the
function back to id_wellformed(), and put it in another file.  While
there, clean up its value to bool.  Peel off the bdrv_is_valid_name()
wrapper.

[Replaced stray return 0 with return false to match bool returns used
elsewhere in id_wellformed().
--Stefan]

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoq35/ahci: Pick up -cdrom and -hda options
John Snow [Wed, 1 Oct 2014 18:19:29 +0000 (14:19 -0400)]
q35/ahci: Pick up -cdrom and -hda options

This patch implements the backend for the Q35 board
for us to be able to pick up and use drives defined
by the -cdrom, -hda, or -drive if=ide shorthand options.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Message-id: 1412187569-23452-7-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoqtest/bios-tables: Correct Q35 command line
John Snow [Wed, 1 Oct 2014 18:19:28 +0000 (14:19 -0400)]
qtest/bios-tables: Correct Q35 command line

If the Q35 board types are to begin recognizing
and decoding syntactic sugar for drive/device
declarations, then workarounds found within
the qtests suite need to be adjusted to prevent
any test failures after the fix.

bios-tables-test improperly uses this cli:
-drive file=etc,id=hd -device ide-hd,drive=hd

Which will create a drive and device due to
the lack of specifying if=none. Then, it will
attempt to create a second device and fail.

This patch corrects this test to always use
the full, non-sugared -device/-drive syntax
for both PC and Q35.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Message-id: 1412187569-23452-6-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoide: Update ide_drive_get to be HBA agnostic
John Snow [Wed, 1 Oct 2014 18:19:27 +0000 (14:19 -0400)]
ide: Update ide_drive_get to be HBA agnostic

Instead of duplicating the logic for the if_ide
(bus,unit) mappings, rely on the blockdev layer
for managing those mappings for us, and use the
drive_get_by_index call instead.

This allows ide_drive_get to work for AHCI HBAs
as well, and can be used in the Q35 initialization.

Lastly, change the nature of the argument to
ide_drive_get so that represents the number of
total drives we can support, and not the total
number of buses. This will prevent array overflows
if the units-per-default-bus property ever needs
to be adjusted for compatibility reasons.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Message-id: 1412187569-23452-5-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agopc/vl: Add units-per-default-bus property
John Snow [Wed, 1 Oct 2014 18:19:26 +0000 (14:19 -0400)]
pc/vl: Add units-per-default-bus property

This patch adds the 'units_per_default_bus' property which
allows individual boards to declare their desired
index => (bus,unit) mapping for their default HBA, so that
boards such as Q35 can specify that its default if_ide HBA,
AHCI, only accepts one unit per bus.

This property only overrides the mapping for drives matching
the block_default_type interface.

This patch also adds this property to *all* past and present
Q35 machine types. This retroactive addition is justified
because the previous erroneous index=>(bus,unit) mappings
caused by lack of such a property were not utilized due to
lack of initialization code in the Q35 init routine.

Further, semantically, the Q35 board type has always had the
property that its default HBA, AHCI, only accepts one unit per
bus. The new code added to add devices to drives relies upon
the accuracy of this mapping. Thus, the property is applied
retroactively to reduce complexity of allowing IDE HBAs with
different units per bus.

Examples:

Prior to this patch, all IDE HBAs were assumed to use 2 units
per bus (Master, Slave). When using Q35 and AHCI, however, we
only allow one unit per bus.

-hdb foo.qcow2 would become index=1, or bus=0,unit=1.
-hdd foo.qcow2 would become index=3, or bus=1,unit=1.
-drive file=foo.qcow2,index=5 becomes bus=2,unit=1.

These are invalid for AHCI. They now become, under Q35 only:

-hdb foo.qcow2 --> index=1, bus=1, unit=0.
-hdd foo.qcow2 --> index=3, bus=3, unit=0.
-drive file=foo.qcow2,index=5 --> bus=5,unit=0.

The mapping is adjusted based on the fact that the default IF
for the Q35 machine type is IF_IDE, and units-per-default-bus
overrides the IDE mapping from its default of 2 units per bus
to just 1 unit per bus.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Message-id: 1412187569-23452-4-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoblockdev: Allow overriding if_max_dev property
John Snow [Wed, 1 Oct 2014 18:19:25 +0000 (14:19 -0400)]
blockdev: Allow overriding if_max_dev property

The if_max_devs table as in the past been an immutable
default that controls the mapping of index => (bus,unit)
for all boards and all HBAs for each interface type.

Since adding this mapping information to the HBA device
itself is currently unwieldly from the perspective of
retrieving this information at option parsing time
(e.g, within drive_new), we consider the alternative
of marking the if_max_devs table mutable so that
later configuration and initialization can adjust the
mapping at will, but only up until a drive is added,
at which point the mapping is finalized.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-id: 1412187569-23452-3-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoblockdev: Orphaned drive search
John Snow [Wed, 1 Oct 2014 18:19:24 +0000 (14:19 -0400)]
blockdev: Orphaned drive search

When users use command line options like -hda, -cdrom,
or even -drive if=ide, it is up to the board initialization
routines to pick up these drives and create backing
devices for them.

Some boards, like Q35, have not been doing this.
However, there is no warning explaining why certain
drive specifications are just silently ignored,
so this function adds a check to print some warnings
to assist users in debugging these sorts of issues
in the future.

This patch will not warn about drives added with if_none,
for which it is not possible to tell in advance if
the omission of a backing device is an issue.

A warning in these cases is considered appropriate.

Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-id: 1412187569-23452-2-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoqemu-iotests: Fix supported cache modes for 052
Kevin Wolf [Tue, 30 Sep 2014 11:27:10 +0000 (13:27 +0200)]
qemu-iotests: Fix supported cache modes for 052

The requirement for this test case is really "no O_DIRECT", because the
temporary snapshot for BDRV_O_SNAPSHOT is created in /tmp, which often
is a tmpfs.

Commit f210a83c ('qemu-iotests: Add _default_cache_mode and
_supported_cache_modes') turned the restriction into writethrough-only,
but that's not really necessary.

Allow to run the test for any non-O_DIRECT cache modes, and use the
global default of writeback if no cache mode is specified.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-id: 1412076430-11623-3-git-send-email-kwolf@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agomake check-block: Use default cache modes
Kevin Wolf [Tue, 30 Sep 2014 11:27:09 +0000 (13:27 +0200)]
make check-block: Use default cache modes

When qemu-iotests only gave a choice between cache=none and
cache=writethrough, we picked cache=none because it was the option that
would complete the test in finite time. Some tests could only work for
one of the two options and would be skipped with cache=none, but that
was an acceptable trade-off at the time.

Today, however, qemu-iotests is a bit more flexible than that and you
can specify any of the cache modes supported by qemu. The default is
writeback, like in qemu, which is fast and (unlike cache=none) compatible
with any host filesystem. Test cases that have specific requirements for
the cache mode can also specify a different default.

In order to get a fast test run that works everywhere and doesn't skip
tests that need a different cache mode, not specifying any cache mode
and instead relying on the default is the best we can do today.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-id: 1412076430-11623-2-git-send-email-kwolf@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoModify qemu_opt_rename to realize renaming all items in opts
Jun Li [Wed, 24 Sep 2014 05:45:27 +0000 (13:45 +0800)]
Modify qemu_opt_rename to realize renaming all items in opts

Add realization of rename all items in opts for qemu_opt_rename.
e.g:
When add bps twice in command line, need to rename all bps to
throttling.bps-total.

This patch solved following bug:
Bug 1145586 - qemu-kvm will give strange hint when add bps twice for a drive
ref:https://bugzilla.redhat.com/show_bug.cgi?id=1145586

[Resolved conflict with commit 5abbf0ee4d87c695deb1c3fca9bb994b93a3e3be
("block: Catch simultaneous usage of options and their aliases").  Check
for simultaneous use first, and then loop over all options.
--Stefan]

Signed-off-by: Jun Li <junmuzi@gmail.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1411537527-16715-1-git-send-email-junmuzi@gmail.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agovmdk: Fix integer overflow in offset calculation
Fam Zheng [Tue, 23 Sep 2014 01:56:21 +0000 (09:56 +0800)]
vmdk: Fix integer overflow in offset calculation

This fixes the bug introduced by commit c6ac36e (vmdk: Optimize cluster
allocation).

$ ~/build/master/qemu-io /stor/vm/arch.vmdk -c 'write 2G 1k'
write failed: Invalid argument

Reported-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Message-id: 1411437381-11234-1-git-send-email-famz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoblock: Drop superfluous conditionals around qemu_opts_del()
Markus Armbruster [Mon, 29 Sep 2014 14:07:55 +0000 (16:07 +0200)]
block: Drop superfluous conditionals around qemu_opts_del()

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1411999675-14533-1-git-send-email-armbru@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agossh: Don't crash if either host or path is not specified.
Richard W.M. Jones [Mon, 29 Sep 2014 08:06:22 +0000 (09:06 +0100)]
ssh: Don't crash if either host or path is not specified.

$ ./qemu-img create -f qcow2 overlay \
    -b 'json: { "file.driver":"ssh",
                "file.host":"localhost",
                "file.host_key_check":"no" }'
qemu-img: qobject/qdict.c:193: qdict_get_obj: Assertion `obj != ((void *)0)' failed.
Aborted

A similar crash also happens if the file.host field is omitted.

https://bugzilla.redhat.com/show_bug.cgi?id=1147343

Bug found and reported by Jun Li.

Signed-off-by: Richard W.M. Jones <rjones@redhat.com>
Reviewed-by: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agosnapshot: fix referencing wrong variable in while loop in do_delvm
Zhang Haoyu [Mon, 29 Sep 2014 08:38:02 +0000 (16:38 +0800)]
snapshot: fix referencing wrong variable in while loop in do_delvm

The while loop variabal is "bs1",
but "bs" is always passed to bdrv_snapshot_delete_by_id_or_name.
Broken in commit a89d89d, v1.7.0.

Signed-off-by: Zhang Haoyu <zhanghy@sangfor.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9 years agoMerge remote-tracking branch 'remotes/kraxel/tags/pull-input-20141002-1' into staging
Peter Maydell [Thu, 2 Oct 2014 14:01:48 +0000 (15:01 +0100)]
Merge remote-tracking branch 'remotes/kraxel/tags/pull-input-20141002-1' into staging

input monitor patches: fix send-key release ordering
and new input-send-event command

# gpg: Signature made Thu 02 Oct 2014 09:10:44 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-input-20141002-1:
  add input-send-event command
  input: fix send-key monitor command release event ordering

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoMerge remote-tracking branch 'remotes/kraxel/tags/pull-console-20141002-1' into staging
Peter Maydell [Thu, 2 Oct 2014 12:23:56 +0000 (13:23 +0100)]
Merge remote-tracking branch 'remotes/kraxel/tags/pull-console-20141002-1' into staging

pixman: fix qemu_default_pixman_format (32bpp non-native endian)

# gpg: Signature made Thu 02 Oct 2014 08:47:20 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-console-20141002-1:
  pixman: fix qemu_default_pixman_format (32bpp non-native endian)

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoMerge remote-tracking branch 'remotes/kraxel/tags/pull-vga-20141002-1' into staging
Peter Maydell [Thu, 2 Oct 2014 11:28:50 +0000 (12:28 +0100)]
Merge remote-tracking branch 'remotes/kraxel/tags/pull-vga-20141002-1' into staging

vga: cleanups, prepare for endianness switching

# gpg: Signature made Thu 02 Oct 2014 08:10:49 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-20141002-1:
  vga: Add endian to vmstate
  vga: Make fb endian a common state variable
  vga: Rename vga_template.h to vga-helpers.h
  vga: Remove some "should be done in BIOS" comments
  cirrus: Remove non-32bpp cursor drawing
  vga: Simplify vga_draw_blank() a bit
  vga: Remove rgb_to_pixel indirection
  vga: Separate LE and BE conversion functions
  vga: Remove remainder of old conversion cruft
  vga: Start cutting out non-32bpp conversion support

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoadd input-send-event command
Marcelo Tosatti [Tue, 30 Sep 2014 21:10:17 +0000 (18:10 -0300)]
add input-send-event command

Which allows specification of absolute/relative,
up/down and console parameters.

Suggested by Gerd Hoffman.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
9 years agoinput: fix send-key monitor command release event ordering
Gerd Hoffmann [Fri, 26 Sep 2014 08:02:16 +0000 (10:02 +0200)]
input: fix send-key monitor command release event ordering

commit 2e377f1730d06deafb3e3ef6cf88792de4a6f4df changed the ordering
of the release events as side effect.  Some guests are not happy with
that and don't recognise ctrl-alt-del any more.  This patch restores
the old last-pressed first-released behavior.

Cc: Amos Kong <akong@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
9 years agoMerge remote-tracking branch 'remotes/bonzini/tags/for-upstream' into staging
Peter Maydell [Tue, 30 Sep 2014 15:45:34 +0000 (16:45 +0100)]
Merge remote-tracking branch 'remotes/bonzini/tags/for-upstream' into staging

This update brings dataplane to virtio-scsi (NOT
yet 100% thread-safe, though, which makes it really, really
experimental.  It also brings asynchronous cancellation to
the SCSI subsystem and implements it in virtio-scsi.  This
is a pretty important feature.  Almost all the work here
was done by Fam Zheng.

I also included the virtio refcount fixes from Gonglei,
because they had a small conflict with virtio-scsi dataplane.

This pull request is using the new subkey 4E6B09D7.

# gpg: Signature made Tue 30 Sep 2014 12:31:02 BST using RSA key ID 4E6B09D7
# gpg: Good signature from "Paolo Bonzini <pbonzini@redhat.com>"
# gpg:                 aka "Paolo Bonzini <bonzini@gnu.org>"

* remotes/bonzini/tags/for-upstream: (39 commits)
  block/iscsi: handle failure on malloc of the allocationmap
  util: introduce bitmap_try_new
  virtio-scsi: Handle TMF request cancellation asynchronously
  scsi: Introduce scsi_req_cancel_async
  scsi: Introduce scsi_req_cancel_complete
  scsi: Drop SCSIReqOps.cancel_io
  scsi: Unify request unref in scsi_req_cancel
  scsi-generic: Handle canceled request in scsi_command_complete
  scsi: Drop scsi_req_abort
  virtio-scsi: Process ".iothread" property
  virtio-scsi: Call bdrv_io_plug/bdrv_io_unplug in cmd request handling
  virtio-scsi: Batched prepare for cmd reqs
  virtio-scsi: Two stages processing of cmd request
  virtio-scsi: Add migration state notifier for dataplane code
  virtio-scsi: Hook up with dataplane
  virtio-scsi-dataplane: Code to run virtio-scsi on iothread
  virtio-scsi: Add VirtIOSCSIVring in VirtIOSCSIReq
  virtio-scsi: Add 'iothread' property to virtio-scsi
  virtio: add a wrapper for virtio-backend initialization
  virtio-9p: fix virtio-9p child refcount in transports
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoMerge remote-tracking branch 'remotes/kraxel/tags/pull-audio-20140930-1' into staging
Peter Maydell [Tue, 30 Sep 2014 14:03:27 +0000 (15:03 +0100)]
Merge remote-tracking branch 'remotes/kraxel/tags/pull-audio-20140930-1' into staging

ac97: register reset via qom

# gpg: Signature made Tue 30 Sep 2014 12:32:29 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-audio-20140930-1:
  ac97: register reset via qom

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoMerge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging
Peter Maydell [Tue, 30 Sep 2014 12:09:39 +0000 (13:09 +0100)]
Merge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging

pci, pc, virtio, misc bugfixes

A bunch of bugfixes.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
# gpg: Signature made Mon 29 Sep 2014 17:59:57 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:
  vl: Adjust the place of calling mlockall to speedup VM's startup
  pc-dimm: Don't check dimm->node when there is non-NUMA config
  pci-hotplug-old: avoid losing error message
  Revert "virtio-pci: fix migration for pci bus master"
  loader: g_realloc(p, 0) frees and returns NULL, simplify

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agovga: Add endian to vmstate
Benjamin Herrenschmidt [Mon, 23 Jun 2014 03:57:41 +0000 (13:57 +1000)]
vga: Add endian to vmstate

Include the endian state in the migration stream as an optional
subsection which we only include when the endian isn't the default,
thus enabling backward compatibility of the common case.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Changes by kraxel:
 * Remove bochs dispi interface changes.  We'll do that in
   a different way to make sure we don't conflict with
   possible future bochs dispi interface changes.
 * keep live migration bits.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
9 years agovga: Make fb endian a common state variable
Benjamin Herrenschmidt [Sun, 6 Jul 2014 23:50:12 +0000 (09:50 +1000)]
vga: Make fb endian a common state variable

And initialize it based on target endian

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
9 years agovga: Rename vga_template.h to vga-helpers.h
Benjamin Herrenschmidt [Mon, 23 Jun 2014 01:46:06 +0000 (11:46 +1000)]
vga: Rename vga_template.h to vga-helpers.h

It's no longer a template, we only instanciate the file once.

Keep it a #included file so the functions remain static.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
9 years agovga: Remove some "should be done in BIOS" comments
Benjamin Herrenschmidt [Sun, 22 Jun 2014 08:28:05 +0000 (18:28 +1000)]
vga: Remove some "should be done in BIOS" comments

Not all platforms have a VGA BIOS, powerpc typically relies on
using the DISPI interface to initialize the card.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
9 years agocirrus: Remove non-32bpp cursor drawing
Benjamin Herrenschmidt [Sun, 22 Jun 2014 01:04:24 +0000 (11:04 +1000)]
cirrus: Remove non-32bpp cursor drawing

We only draw cursor on non-shared surfaces (so it seems...) and
these are always 32bpp

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
9 years agovga: Simplify vga_draw_blank() a bit
Benjamin Herrenschmidt [Sun, 22 Jun 2014 01:03:49 +0000 (11:03 +1000)]
vga: Simplify vga_draw_blank() a bit

The test for surface_bits_per_pixel() isn't necessary anymore,
the 8bpp case never happens.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
9 years agovga: Remove rgb_to_pixel indirection
Benjamin Herrenschmidt [Sun, 22 Jun 2014 01:00:50 +0000 (11:00 +1000)]
vga: Remove rgb_to_pixel indirection

We always use rgb_to_pixel32 nowadays.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
9 years agovga: Separate LE and BE conversion functions
Benjamin Herrenschmidt [Sun, 6 Jul 2014 23:48:28 +0000 (09:48 +1000)]
vga: Separate LE and BE conversion functions

Provide different functions for converting from an LE vs a BE
framebuffer. We cannot rely on the simple cases always being
shared surfaces since cirrus will need to always shadow for
cursor emulation, so we need the full set of functions to
be able to later handle runtime switching.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>\
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
9 years agovga: Remove remainder of old conversion cruft
Benjamin Herrenschmidt [Sat, 21 Jun 2014 05:51:52 +0000 (15:51 +1000)]
vga: Remove remainder of old conversion cruft

All the macros used to generate different versions of vga_template.h
are now unnecessary, take them all out and remove the _32 suffix from
most functions.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
9 years agovga: Start cutting out non-32bpp conversion support
Benjamin Herrenschmidt [Sat, 21 Jun 2014 05:39:31 +0000 (15:39 +1000)]
vga: Start cutting out non-32bpp conversion support

Nowadays, we either share a surface with the host, or we create
a 32bpp ARGB console surface.

So we only need to draw/convert to 32bpp, enabling us to remove
all but one instance of vga_template.h inclusion (to be further
cleaned up), rgb_to_pixel_* etc...

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
9 years agopixman: fix qemu_default_pixman_format (32bpp non-native endian)
Gerd Hoffmann [Mon, 15 Sep 2014 06:33:37 +0000 (08:33 +0200)]
pixman: fix qemu_default_pixman_format (32bpp non-native endian)

Bug breaks SDL display of bigendian guests on little endian hosts.

Reported-by: BALATON Zoltan <balaton@eik.bme.hu>
Reported-by: Valentin Manea <valentin.manea@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
9 years agoblock/iscsi: handle failure on malloc of the allocationmap
Peter Lieven [Tue, 30 Sep 2014 07:09:12 +0000 (09:09 +0200)]
block/iscsi: handle failure on malloc of the allocationmap

Signed-off-by: Peter Lieven <pl@kamp.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agoutil: introduce bitmap_try_new
Peter Lieven [Tue, 30 Sep 2014 07:09:11 +0000 (09:09 +0200)]
util: introduce bitmap_try_new

regular bitmap_new simply aborts if the memory allocation fails.
bitmap_try_new returns NULL on failure and allows for proper
error handling.

Signed-off-by: Peter Lieven <pl@kamp.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-scsi: Handle TMF request cancellation asynchronously
Fam Zheng [Tue, 30 Sep 2014 03:40:23 +0000 (11:40 +0800)]
virtio-scsi: Handle TMF request cancellation asynchronously

For VIRTIO_SCSI_T_TMF_ABORT_TASK and VIRTIO_SCSI_T_TMF_ABORT_TASK_SET,
use scsi_req_cancel_async to start the cancellation.

Because each tmf command may cancel multiple requests, we need to use a
counter to track the number of remaining requests we still need to wait
for.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agoscsi: Introduce scsi_req_cancel_async
Fam Zheng [Sun, 28 Sep 2014 01:48:00 +0000 (09:48 +0800)]
scsi: Introduce scsi_req_cancel_async

Devices will call this function to start an asynchronous cancellation. The
bus->info->cancel will be called after the request is canceled.

Devices will probably need to track a separate TMF request that triggers this
cancellation, and wait until the cancellation is done before completing it. So
we store a notifier list in SCSIRequest and in scsi_req_cancel_complete we
notify them.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agoscsi: Introduce scsi_req_cancel_complete
Fam Zheng [Thu, 25 Sep 2014 02:20:47 +0000 (10:20 +0800)]
scsi: Introduce scsi_req_cancel_complete

Let the aio cb do the clean up and notification job after scsi_req_cancel, in
preparation for asynchronous cancellation.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agoscsi: Drop SCSIReqOps.cancel_io
Fam Zheng [Wed, 24 Sep 2014 08:27:55 +0000 (16:27 +0800)]
scsi: Drop SCSIReqOps.cancel_io

The only two implementations are identical to each other, with nothing specific
to device: they only call bdrv_aio_cancel with the SCSIRequest.aiocb.

Let's move it to scsi-bus.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agoscsi: Unify request unref in scsi_req_cancel
Fam Zheng [Wed, 24 Sep 2014 08:27:54 +0000 (16:27 +0800)]
scsi: Unify request unref in scsi_req_cancel

Before, scsi_req_cancel will take ownership of the canceled request and unref
it. We did this because we didn't know whether AIO CB will be called or not
during the cancelling, so we set the io_canceled flag before calling it, and
skip unref in the potentially called callbacks, which is not very nice.

Now, bdrv_aio_cancel has a stricter contract that the completion callbacks are
always called, so we can remove the checks of req->io_canceled and just unref
it in callbacks.

It will also make implementing asynchronous cancellation easier.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agoscsi-generic: Handle canceled request in scsi_command_complete
Fam Zheng [Wed, 24 Sep 2014 08:27:53 +0000 (16:27 +0800)]
scsi-generic: Handle canceled request in scsi_command_complete

Now that we always called the cb in bdrv_aio_cancel, let's make scsi-generic
callbacks check io_canceled flag similarly to scsi-disk.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agoscsi: Drop scsi_req_abort
Fam Zheng [Wed, 24 Sep 2014 08:27:52 +0000 (16:27 +0800)]
scsi: Drop scsi_req_abort

The only user of this function is spapr_vscsi.c. We can convert to
scsi_req_cancel plus adding a check in vscsi_request_cancelled.

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
[Drop prototype. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agoMerge remote-tracking branch 'remotes/rth/tags/tcg-next-201400729' into staging
Peter Maydell [Tue, 30 Sep 2014 10:52:41 +0000 (11:52 +0100)]
Merge remote-tracking branch 'remotes/rth/tags/tcg-next-201400729' into staging

tcg updates

# gpg: Signature made Mon 29 Sep 2014 19:58:04 BST using RSA key ID 4DD0279B
# gpg: Good signature from "Richard Henderson <rth7680@gmail.com>"
# gpg:                 aka "Richard Henderson <rth@redhat.com>"
# gpg:                 aka "Richard Henderson <rth@twiddle.net>"

* remotes/rth/tags/tcg-next-201400729:
  tcg: Always enable TCGv type checking
  qemu/compiler: Define QEMU_ARTIFICIAL
  tcg-aarch64: Use 32-bit loads for qemu_ld_i32
  tcg-sparc: Use UMULXHI instruction
  tcg-sparc: Rename ADDX/SUBX insns
  tcg-sparc: Use ADDXC in setcond_i64
  tcg-sparc: Fix setcond_i32 uninitialized value
  tcg-sparc: Use ADDXC in addsub2_i64
  tcg-sparc: Support addsub2_i64

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agoMerge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20140929' into...
Peter Maydell [Tue, 30 Sep 2014 10:02:06 +0000 (11:02 +0100)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20140929' into staging

target-arm:
 * more EL2/EL3 preparation work
 * don't handle c15_cpar changes via tb_flush()
 * fix some unused function warnings in ARM devices
 * build the GDB XML for 32 bit CPUs into qemu-*-aarch64
 * implement guest breakpoint support

# gpg: Signature made Mon 29 Sep 2014 19:25:37 BST using RSA key ID 14360CDE
# gpg: Good signature from "Peter Maydell <peter.maydell@linaro.org>"

* remotes/pmaydell/tags/pull-target-arm-20140929:
  target-arm: Add support for VIRQ and VFIQ
  target-arm: Add IRQ and FIQ routing to EL2 and 3
  target-arm: A64: Emulate the SMC insn
  target-arm: Add a Hypervisor Trap exception type
  target-arm: A64: Emulate the HVC insn
  target-arm: A64: Correct updates to FAR and ESR on exceptions
  target-arm: Don't take interrupts targeting lower ELs
  target-arm: Break out exception masking to a separate func
  target-arm: A64: Refactor aarch64_cpu_do_interrupt
  target-arm: Add SCR_EL3
  target-arm: Add HCR_EL2
  target-arm: Don't handle c15_cpar changes via tb_flush()
  hw/input/tsc210x.c: Delete unused array tsc2101_rates
  hw/display/pxa2xx_lcd.c: Remove unused function pxa2xx_dma_rdst_set
  hw/intc/imx_avic.c: Remove unused function imx_avic_set_prio()
  hw/display/blizzard.c: Delete unused function blizzard_rgb2yuv
  configure: Build GDB XML for 32 bit ARM CPUs into qemu aarch64 binaries
  target-arm: Implement handling of breakpoint firing
  target-arm: Implement setting guest breakpoints

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agovirtio-scsi: Process ".iothread" property
Fam Zheng [Tue, 23 Sep 2014 07:49:30 +0000 (15:49 +0800)]
virtio-scsi: Process ".iothread" property

We are ready, now let's effectively enable dataplane.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-scsi: Call bdrv_io_plug/bdrv_io_unplug in cmd request handling
Fam Zheng [Tue, 23 Sep 2014 07:49:29 +0000 (15:49 +0800)]
virtio-scsi: Call bdrv_io_plug/bdrv_io_unplug in cmd request handling

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-scsi: Batched prepare for cmd reqs
Fam Zheng [Tue, 23 Sep 2014 07:49:28 +0000 (15:49 +0800)]
virtio-scsi: Batched prepare for cmd reqs

Queue the popped requests while calling
virtio_scsi_handle_cmd_req_prepare(), then submit them after all
prepared.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-scsi: Two stages processing of cmd request
Fam Zheng [Tue, 23 Sep 2014 07:49:27 +0000 (15:49 +0800)]
virtio-scsi: Two stages processing of cmd request

Mechanical change, in preparation for bdrv_io_plug/bdrv_io_unplug.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-scsi: Add migration state notifier for dataplane code
Fam Zheng [Tue, 23 Sep 2014 07:49:26 +0000 (15:49 +0800)]
virtio-scsi: Add migration state notifier for dataplane code

Similar to virtio-blk-dataplane, we stop the iothread while migration
starts and restart it when migration finishes.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-scsi: Hook up with dataplane
Fam Zheng [Tue, 23 Sep 2014 07:49:25 +0000 (15:49 +0800)]
virtio-scsi: Hook up with dataplane

This enables the virtio-scsi-dataplane code by setting the iothread
in virtio-scsi device, and makes any function that is called by
back from dataplane to cooperate with the caller: they need to be
vring/iothread aware when handling the requests and using scsi devices
on the bus.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-scsi-dataplane: Code to run virtio-scsi on iothread
Fam Zheng [Tue, 23 Sep 2014 07:49:24 +0000 (15:49 +0800)]
virtio-scsi-dataplane: Code to run virtio-scsi on iothread

This implements the core part of dataplane feature of virtio-scsi.

A few fields are added in VirtIOSCSICommon to maintain the dataplane
status. These fields are managed by a new source file:
virtio-scsi-dataplane.c.

Most code in this file will run on an iothread, unless otherwise
commented as in a global mutex context, such as those functions to
start, stop and setting the iothread property.

Upon start, we set up guest/host event notifiers, in a same way as
virtio-blk does. The handlers then pop request from vring and call into
virtio-scsi.c functions to process it. So we need to make sure make all
those called functions work with iothread, too.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-scsi: Add VirtIOSCSIVring in VirtIOSCSIReq
Fam Zheng [Wed, 24 Sep 2014 07:21:41 +0000 (15:21 +0800)]
virtio-scsi: Add VirtIOSCSIVring in VirtIOSCSIReq

Move VirtIOSCSIReq to header and add one field "vring" as a wrapper
structure of Vring, VirtIOSCSIVring.

This is necessary for coming dataplane code that runs uses vring on
iothread.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-scsi: Add 'iothread' property to virtio-scsi
Fam Zheng [Wed, 6 Aug 2014 05:35:06 +0000 (13:35 +0800)]
virtio-scsi: Add 'iothread' property to virtio-scsi

Similar to this property in virtio-blk for dataplane, add it as a QOM
link in virtio-scsi and an alias in virtio-scsi-pci and virtio-scsi-ccw,
in order to assign an iothread to the device.

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio: add a wrapper for virtio-backend initialization
Gonglei [Tue, 30 Sep 2014 06:10:38 +0000 (14:10 +0800)]
virtio: add a wrapper for virtio-backend initialization

For better code sharing, add a helper function that handles
reference counting of the virtio backend for virtio proxy devices.

Cc: Cornelia Huck <cornelia.huck@de.ibm.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-9p: fix virtio-9p child refcount in transports
Gonglei [Tue, 30 Sep 2014 06:10:37 +0000 (14:10 +0800)]
virtio-9p: fix virtio-9p child refcount in transports

object_initialize() leaves the object with a refcount of 1.
object_property_add_child() adds its own reference which is
dropped again when the property is deleted.

The upshot of this is that we always have a refcount >= 1. Upon
unplug the virtio-9p child is not finalized!

Drop our reference after the child property has been added to the
parent.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-9p: use aliases instead of duplicate qdev properties
Gonglei [Tue, 30 Sep 2014 06:10:36 +0000 (14:10 +0800)]
virtio-9p: use aliases instead of duplicate qdev properties

virtio-9p-pci all duplicate the qdev properties of their
V9fsState child. This approach does not work well with
string or pointer properties since we must be careful
about leaking or double-freeing them.

Use the QOM alias property to forward property accesses to the
V9fsState child.  This way no duplication is necessary.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-balloon: fix virtio-balloon child refcount in transports
Gonglei [Tue, 30 Sep 2014 06:10:35 +0000 (14:10 +0800)]
virtio-balloon: fix virtio-balloon child refcount in transports

object_initialize() leaves the object with a refcount of 1.
object_property_add_child() adds its own reference which is dropped
again when the property is deleted.

The upshot of this is that we always have a refcount >= 1.  Upon hot
unplug the virtio-balloon child is not finalized!

Drop our reference after the child property has been added to the
parent.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-rng: fix virtio-rng child refcount in transports
Gonglei [Tue, 30 Sep 2014 06:10:34 +0000 (14:10 +0800)]
virtio-rng: fix virtio-rng child refcount in transports

object_initialize() leaves the object with a refcount of 1.
object_property_add_child() adds its own reference which is dropped
again when the property is deleted.

The upshot of this is that we always have a refcount >= 1.  Upon hot
unplug the virtio-rng child is not finalized!

Drop our reference after the child property has been added to the
parent.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-rng: use aliases instead of duplicate qdev properties
Gonglei [Tue, 30 Sep 2014 06:10:33 +0000 (14:10 +0800)]
virtio-rng: use aliases instead of duplicate qdev properties

virtio-rng-{pci, s390, ccw} all duplicate the
qdev properties of their VirtIORNG child.
This approach does not work well with string or pointer
properties since we must be careful about leaking or
double-freeing them.

Use the QOM alias property to forward property accesses to the
VirtIORNG child.  This way no duplication is necessary.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-serial: fix virtio-serial child refcount in transports
Gonglei [Tue, 30 Sep 2014 06:10:32 +0000 (14:10 +0800)]
virtio-serial: fix virtio-serial child refcount in transports

object_initialize() leaves the object with a refcount of 1.
object_property_add_child() adds its own reference which is dropped
again when the property is deleted.

The upshot of this is that we always have a refcount >= 1.  Upon hot
unplug the virtio-serial child is not finalized!

Drop our reference after the child property has been added to the
parent.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-serial: use aliases instead of duplicate qdev properties
Gonglei [Tue, 30 Sep 2014 06:10:31 +0000 (14:10 +0800)]
virtio-serial: use aliases instead of duplicate qdev properties

virtio-serial-{pci, s390, ccw} all duplicate the
qdev properties of their VirtIOSerial child.
This approach does not work well with string or pointer
properties since we must be careful about leaking or
double-freeing them.

Use the QOM alias property to forward property accesses to the
VirtIOSerial child.  This way no duplication is necessary.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio/vhost-scsi: fix virtio-scsi/vhost-scsi child refcount in transports
Gonglei [Tue, 30 Sep 2014 06:10:30 +0000 (14:10 +0800)]
virtio/vhost-scsi: fix virtio-scsi/vhost-scsi child refcount in transports

object_initialize() leaves the object with a refcount of 1.
object_property_add_child() adds its own reference which is dropped
again when the property is deleted.

The upshot of this is that we always have a refcount >= 1.  Upon hot
unplug the virtio-scsi/vhost-scsi child is not finalized!

Drop our reference after the child property has been added to the
parent.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio/vhost-scsi: use aliases instead of duplicate qdev properties
Gonglei [Tue, 30 Sep 2014 06:10:29 +0000 (14:10 +0800)]
virtio/vhost-scsi: use aliases instead of duplicate qdev properties

{virtio, vhost}-scsi-{pci, s390, ccw} all duplicate the
qdev properties of their VirtIOSCSI/VHostSCSI child.
This approach does not work well with string or pointer
properties since we must be careful about leaking or
double-freeing them.

Use the QOM alias property to forward property accesses to the
VirtIOSCSI/VHostSCSI child. This way no duplication is necessary.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-net: fix virtio-net child refcount in transports
Gonglei [Tue, 30 Sep 2014 06:10:28 +0000 (14:10 +0800)]
virtio-net: fix virtio-net child refcount in transports

object_initialize() leaves the object with a refcount of 1.
object_property_add_child() adds its own reference which is dropped
again when the property is deleted.

The upshot of this is that we always have a refcount >= 1.  Upon hot
unplug the virtio-net child is not finalized!

Drop our reference after the child property has been added to the
parent.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agovirtio-net: use aliases instead of duplicate qdev properties
Gonglei [Tue, 30 Sep 2014 06:10:27 +0000 (14:10 +0800)]
virtio-net: use aliases instead of duplicate qdev properties

virtio-net-pci, virtio-net-s390, and virtio-net-ccw all duplicate the
qdev properties of their VirtIONet child. This approach does not work
well with string or pointer properties since we must be careful about
leaking or double-freeing them.

Use the QOM alias property to forward property accesses to the
VirtIONet child.  This way no duplication is necessary.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agotcg: Always enable TCGv type checking
Richard Henderson [Tue, 16 Sep 2014 16:51:46 +0000 (09:51 -0700)]
tcg: Always enable TCGv type checking

Instead of using structures, which imply some amount of overhead
on certain ABIs, use pointer types.

This actually reduces the size of the binaries vs a NON-debug
build on ppc64 and x86_64, due to a reduction in the number of
sign-extension insns.

Signed-off-by: Richard Henderson <rth@twiddle.net>
9 years agoqemu/compiler: Define QEMU_ARTIFICIAL
Richard Henderson [Tue, 16 Sep 2014 16:47:32 +0000 (09:47 -0700)]
qemu/compiler: Define QEMU_ARTIFICIAL

The combination of always_inline + artificial allows tiny inline
functions to be written that do not interfere with debugging.
In particular, gdb will not step into an artificial function.

The always_inline attribute was introduced in gcc 4.2,
and the artificial attribute was introduced in gcc 4.3.

Signed-off-by: Richard Henderson <rth@twiddle.net>
9 years agotcg-aarch64: Use 32-bit loads for qemu_ld_i32
Richard Henderson [Tue, 2 Sep 2014 20:59:47 +0000 (13:59 -0700)]
tcg-aarch64: Use 32-bit loads for qemu_ld_i32

The "old" qemu_ld opcode did not specify the size of the result,
and so we had to assume full register width.  With the new opcodes,
we can narrow the result.

Signed-off-by: Richard Henderson <rth@twiddle.net>
9 years agotcg-sparc: Use UMULXHI instruction
Richard Henderson [Wed, 6 Aug 2014 19:32:40 +0000 (12:32 -0700)]
tcg-sparc: Use UMULXHI instruction

Signed-off-by: Richard Henderson <rth@twiddle.net>
9 years agotcg-sparc: Rename ADDX/SUBX insns
Richard Henderson [Wed, 6 Aug 2014 19:16:04 +0000 (12:16 -0700)]
tcg-sparc: Rename ADDX/SUBX insns

The pre-v9 ADDX/SUBX insns were renamed ADDC/SUBC for v9.
Standardizing on the v9 name makes things less confusing.

Signed-off-by: Richard Henderson <rth@twiddle.net>
9 years agotcg-sparc: Use ADDXC in setcond_i64
Richard Henderson [Wed, 6 Aug 2014 19:11:25 +0000 (12:11 -0700)]
tcg-sparc: Use ADDXC in setcond_i64

Similar to the ADDC tricks we use in setcond_i32.

Signed-off-by: Richard Henderson <rth@twiddle.net>
9 years agotcg-sparc: Fix setcond_i32 uninitialized value
Richard Henderson [Wed, 6 Aug 2014 19:09:56 +0000 (12:09 -0700)]
tcg-sparc: Fix setcond_i32 uninitialized value

We failed to swap c1 and c2 correctly for NE c2 == 0.

Signed-off-by: Richard Henderson <rth@twiddle.net>
9 years agotcg-sparc: Use ADDXC in addsub2_i64
Richard Henderson [Wed, 6 Aug 2014 18:48:48 +0000 (11:48 -0700)]
tcg-sparc: Use ADDXC in addsub2_i64

On T4 and newer Sparc chips we have an add-with-carry insn
that takes its input from %xcc instead of %icc.

Signed-off-by: Richard Henderson <rth@twiddle.net>
9 years agotcg-sparc: Support addsub2_i64
Richard Henderson [Thu, 6 Mar 2014 20:49:02 +0000 (12:49 -0800)]
tcg-sparc: Support addsub2_i64

Signed-off-by: Richard Henderson <rth@twiddle.net>
9 years agoMerge remote-tracking branch 'remotes/spice/tags/pull-spice-20140929-1' into staging
Peter Maydell [Mon, 29 Sep 2014 18:28:15 +0000 (19:28 +0100)]
Merge remote-tracking branch 'remotes/spice/tags/pull-spice-20140929-1' into staging

add and use graphic_console_set_hwops

# gpg: Signature made Mon 29 Sep 2014 11:18:37 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-20140929-1:
  qxl: use graphic_console_set_hwops
  console: add graphic_console_set_hwops

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Add support for VIRQ and VFIQ
Edgar E. Iglesias [Mon, 29 Sep 2014 17:48:51 +0000 (18:48 +0100)]
target-arm: Add support for VIRQ and VFIQ

This only implements the external delivery method via the GIC.

Acked-by: Greg Bellows <greg.bellows@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1411718914-6608-12-git-send-email-edgar.iglesias@gmail.com
[PMM: adjusted following cpu-exec refactoring]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Add IRQ and FIQ routing to EL2 and 3
Edgar E. Iglesias [Mon, 29 Sep 2014 17:48:51 +0000 (18:48 +0100)]
target-arm: Add IRQ and FIQ routing to EL2 and 3

Reviewed-by: Greg Bellows <greg.bellows@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1411718914-6608-11-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: A64: Emulate the SMC insn
Edgar E. Iglesias [Mon, 29 Sep 2014 17:48:50 +0000 (18:48 +0100)]
target-arm: A64: Emulate the SMC insn

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1411718914-6608-10-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Add a Hypervisor Trap exception type
Edgar E. Iglesias [Mon, 29 Sep 2014 17:48:50 +0000 (18:48 +0100)]
target-arm: Add a Hypervisor Trap exception type

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1411718914-6608-9-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: A64: Emulate the HVC insn
Edgar E. Iglesias [Mon, 29 Sep 2014 17:48:50 +0000 (18:48 +0100)]
target-arm: A64: Emulate the HVC insn

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1411718914-6608-8-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: A64: Correct updates to FAR and ESR on exceptions
Edgar E. Iglesias [Mon, 29 Sep 2014 17:48:50 +0000 (18:48 +0100)]
target-arm: A64: Correct updates to FAR and ESR on exceptions

Not all exception types update both FAR and ESR.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Greg Bellows <greg.bellows@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1411718914-6608-7-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Don't take interrupts targeting lower ELs
Edgar E. Iglesias [Mon, 29 Sep 2014 17:48:49 +0000 (18:48 +0100)]
target-arm: Don't take interrupts targeting lower ELs

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Greg Bellows <greg.bellows@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1411718914-6608-6-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Break out exception masking to a separate func
Edgar E. Iglesias [Mon, 29 Sep 2014 17:48:49 +0000 (18:48 +0100)]
target-arm: Break out exception masking to a separate func

Reviewed-by: Greg Bellows <greg.bellows@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1411718914-6608-5-git-send-email-edgar.iglesias@gmail.com
[PMM: updated to account for recent cpu-exec refactoring]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: A64: Refactor aarch64_cpu_do_interrupt
Edgar E. Iglesias [Mon, 29 Sep 2014 17:48:49 +0000 (18:48 +0100)]
target-arm: A64: Refactor aarch64_cpu_do_interrupt

Introduce new_el and new_mode in preparation for future patches
that add support for taking exceptions to and from EL2 and 3.
No functional change.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1411718914-6608-4-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Add SCR_EL3
Edgar E. Iglesias [Mon, 29 Sep 2014 17:48:49 +0000 (18:48 +0100)]
target-arm: Add SCR_EL3

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1411718914-6608-3-git-send-email-edgar.iglesias@gmail.com
[PMM: apply offsetoflow32() to correct regdef]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Add HCR_EL2
Edgar E. Iglesias [Mon, 29 Sep 2014 17:48:48 +0000 (18:48 +0100)]
target-arm: Add HCR_EL2

Reviewed-by: Greg Bellows <greg.bellows@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 1411718914-6608-2-git-send-email-edgar.iglesias@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9 years agotarget-arm: Don't handle c15_cpar changes via tb_flush()
Peter Maydell [Mon, 29 Sep 2014 17:48:48 +0000 (18:48 +0100)]
target-arm: Don't handle c15_cpar changes via tb_flush()

At the moment we try to handle c15_cpar with the strategy of:
 * emit generated code which makes assumptions about its value
 * when the register value changes call tb_flush() to throw
   away the now-invalid generated code
This works because XScale CPUs are always uniprocessor, but
it's confusing because it suggests that the same approach can
be taken for other registers. It also means we do a tb_flush()
on CPU reset, which makes multithreaded linux-user binaries
even more likely to fail than would otherwise be the case.

Replace it with a combination of TB flags for the access
checks done on cp0/cp1 for the XScale and iwMMXt instructions,
plus a runtime check for cp2..cp13 coprocessor accesses.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1411056959-23070-1-git-send-email-peter.maydell@linaro.org