qemu.git
7 years agodocs/qcow2: Correct refcount_block_entries
Max Reitz [Tue, 2 Sep 2014 22:25:08 +0000 (00:25 +0200)] 
docs/qcow2: Correct refcount_block_entries

A refblock entry may have a different size than 16 bits, it may even be
smaller than a byte. Correct the refcount_block_entries calculation
accordingly.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Drop REFCOUNT_SHIFT
Max Reitz [Tue, 2 Sep 2014 22:25:07 +0000 (00:25 +0200)] 
qcow2: Drop REFCOUNT_SHIFT

With BDRVQcowState.refcount_block_bits, we don't need REFCOUNT_SHIFT
anymore.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoiotests: Add test for potentially damaging repairs
Max Reitz [Wed, 22 Oct 2014 12:09:43 +0000 (14:09 +0200)] 
iotests: Add test for potentially damaging repairs

There are certain cases where repairing a qcow2 image might actually
damage it further (or rather, where repairing it has in fact damaged it
further with the old qcow2 check implementation). This should not
happen, so add a test for these cases.

Furthermore, the repair function now repairs refblocks beyond the image
end by resizing the image accordingly. Add several tests for this as
well.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoiotests: Fix test outputs
Max Reitz [Wed, 22 Oct 2014 12:09:42 +0000 (14:09 +0200)] 
iotests: Fix test outputs

039, 060 and 061 all create images with referenced clusters having a
refcount of 0. Because previous commits changed handling of such errors,
these tests now have a different output. Fix it.

Furthermore, 060 created a refblock with a refcount greater than one
which now results in having to rebuild the refcount structure as well.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Clean up after refcount rebuild
Max Reitz [Wed, 22 Oct 2014 12:09:41 +0000 (14:09 +0200)] 
qcow2: Clean up after refcount rebuild

Because the old refcount structure will be leaked after having rebuilt
it, we need to recalculate the refcounts and run a leak-fixing operation
afterwards (if leaks should be fixed at all).

Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Rebuild refcount structure during check
Max Reitz [Wed, 22 Oct 2014 12:09:40 +0000 (14:09 +0200)] 
qcow2: Rebuild refcount structure during check

The previous commit introduced the "rebuild" variable to qcow2's
implementation of the image consistency check. Now make use of this by
adding a function which creates a completely new refcount structure
based solely on the in-memory information gathered before.

The old refcount structure will be leaked, however. This leak will be
dealt with in a follow-up commit.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Do not perform potentially damaging repairs
Max Reitz [Wed, 22 Oct 2014 12:09:39 +0000 (14:09 +0200)] 
qcow2: Do not perform potentially damaging repairs

If a referenced cluster has a refcount of 0, increasing its refcount may
result in clusters being allocated for the refcount structures. This may
overwrite the referenced cluster, therefore we cannot simply increase
the refcount then.

In such cases, we can either try to replicate all the refcount
operations solely for the check operation, basing the allocations on the
in-memory refcount table; or we can simply rebuild the whole refcount
structure based on the in-memory refcount table. Since the latter will
be much easier, do that.

To prepare for this, introduce a "rebuild" boolean which should be set
to true whenever a fix is rather dangerous or too complicated using the
current refcount structures. Another example for this is refcount blocks
being referenced more than once.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Fix refcount blocks beyond image end
Max Reitz [Wed, 22 Oct 2014 12:09:38 +0000 (14:09 +0200)] 
qcow2: Fix refcount blocks beyond image end

If the qcow2 check function detects a refcount block located beyond the
image end, grow the image appropriately. This cannot break anything and
is the logical fix for such a case.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Reuse refcount table in calculate_refcounts()
Max Reitz [Wed, 22 Oct 2014 12:09:37 +0000 (14:09 +0200)] 
qcow2: Reuse refcount table in calculate_refcounts()

We will later call calculate_refcounts multiple times, so reuse the
refcount table if possible.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Let inc_refcounts() resize the reftable
Max Reitz [Wed, 22 Oct 2014 12:09:36 +0000 (14:09 +0200)] 
qcow2: Let inc_refcounts() resize the reftable

Now that the refcount table can be passed around by reference, do that
for inc_refcounts() (and subsequently check_refcounts_l1() and
check_refcounts_l2()) and use it for resizing it when a cluster after
the image end is encountered.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Let inc_refcounts() return -errno
Max Reitz [Wed, 22 Oct 2014 12:09:35 +0000 (14:09 +0200)] 
qcow2: Let inc_refcounts() return -errno

As of a future patch, inc_refcounts() will have to throw errors which
are generally signaled by returning -errno. Therefore, let it return an
integer which is either 0 for success or -errno and handle the -errno
case in all callers.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Split fail code in L1 and L2 checks
Max Reitz [Wed, 22 Oct 2014 12:09:34 +0000 (14:09 +0200)] 
qcow2: Split fail code in L1 and L2 checks

Instead of printing out an error message, incrementing check_errors and
returning a fixed -errno, just do cleanups and return -ret, with ret set
by the code which threw the exception (jumped to the fail label).

Also, increment check_errors on error in check_refcounts_l2().

Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Use int64_t for in-memory reftable size
Max Reitz [Wed, 22 Oct 2014 12:09:33 +0000 (14:09 +0200)] 
qcow2: Use int64_t for in-memory reftable size

Use int64_t for the entry count of the in-memory refcount table
throughout the check functions.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Pull check_refblocks() up
Max Reitz [Wed, 22 Oct 2014 12:09:32 +0000 (14:09 +0200)] 
qcow2: Pull check_refblocks() up

Pull check_refblocks() before calculate_refcounts() so we can drop its
static declaration.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Use sizeof(**refcount_table)
Max Reitz [Wed, 22 Oct 2014 12:09:31 +0000 (14:09 +0200)] 
qcow2: Use sizeof(**refcount_table)

When implementing variable refcounts, we want to be able to easily find
all the places in qemu which are tied to a certain refcount order.
Replace sizeof(uint16_t) in the check code by sizeof(**refcount_table)
so we can later find it more easily.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Split qcow2_check_refcounts()
Max Reitz [Wed, 22 Oct 2014 12:09:30 +0000 (14:09 +0200)] 
qcow2: Split qcow2_check_refcounts()

Put the code for calculating the reference counts and comparing them
during qemu-img check into own functions.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Fix leaks in dirty images
Max Reitz [Wed, 22 Oct 2014 12:09:29 +0000 (14:09 +0200)] 
qcow2: Fix leaks in dirty images

When opening dirty images, qcow2's repair function should not only
repair errors but leaks as well.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: Calculate refcount block entry count
Max Reitz [Wed, 22 Oct 2014 12:09:28 +0000 (14:09 +0200)] 
qcow2: Calculate refcount block entry count

The size of a refblock entry is (in theory) variable; calculate
therefore the number of entries per refblock and the according bit shift
(1 << x == entry count) when opening an image.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: Add qemu_{,try_}blockalign0()
Max Reitz [Wed, 22 Oct 2014 12:09:27 +0000 (14:09 +0200)] 
block: Add qemu_{,try_}blockalign0()

These functions call their non-0-counterparts and then fill the
allocated buffer with 0 (if the allocation has been successful).

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: qemu-iotests change _supported_proto to file once more.
Peter Lieven [Mon, 20 Oct 2014 11:47:11 +0000 (13:47 +0200)] 
block: qemu-iotests change _supported_proto to file once more.

In preparation to possible automatic regression and performance
testing for the block layer I found that the iotests don't work
for all protocols anymore.

In commit 1f7bf7d0 I started to change supported protocols from
generic to file for various tests. Unfortunately, some tests
added in the meantime again carry generic protocol altough they
can only work with file because they require local file access.

The other way around for some tests that only support file I added
NFS protocol after confirming they work.

Signed-off-by: Peter Lieven <pl@kamp.de>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock/vdi: Use {DIV_,}ROUND_UP
Max Reitz [Tue, 21 Oct 2014 08:51:25 +0000 (10:51 +0200)] 
block/vdi: Use {DIV_,}ROUND_UP

There are macros for these operations, so make use of them.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoMAINTAINERS: add the image fuzzer to the block layer
Paolo Bonzini [Tue, 21 Oct 2014 22:29:58 +0000 (00:29 +0200)] 
MAINTAINERS: add the image fuzzer to the block layer

More work for the block device maintainers!

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoMAINTAINERS: qemu-iotests belongs to the block layer
Kevin Wolf [Tue, 21 Oct 2014 15:43:17 +0000 (17:43 +0200)] 
MAINTAINERS: qemu-iotests belongs to the block layer

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoMAINTAINERS: add aio to block layer
Paolo Bonzini [Tue, 21 Oct 2014 13:12:13 +0000 (15:12 +0200)] 
MAINTAINERS: add aio to block layer

AioContext falls under the block layer, mark it as such.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoMerge remote-tracking branch 'remotes/mdroth/tags/qga-pull-2014-10-22-tag' into staging
Peter Maydell [Wed, 22 Oct 2014 20:42:33 +0000 (21:42 +0100)] 
Merge remote-tracking branch 'remotes/mdroth/tags/qga-pull-2014-10-22-tag' into staging

qga: remove readdir_r usage and fix use-after-free

Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
# gpg: Signature made Wed 22 Oct 2014 13:56:19 BST using RSA key ID F108B584
# gpg: Can't check signature: public key not found

* remotes/mdroth/tags/qga-pull-2014-10-22-tag:
  qga: Rewrite code where using readdir_r

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agoMerge remote-tracking branch 'remotes/bkoppelmann/tags/pull-tricore-20141021' into...
Peter Maydell [Wed, 22 Oct 2014 17:43:35 +0000 (18:43 +0100)] 
Merge remote-tracking branch 'remotes/bkoppelmann/tags/pull-tricore-20141021' into staging

TriCore ABS, ABSB, B, BIT, BO instructions added

# gpg: Signature made Tue 21 Oct 2014 17:47:32 BST using RSA key ID 6B69CA14
# gpg: Good signature from "Bastian Koppelmann <kbastian@mail.uni-paderborn.de>"

* remotes/bkoppelmann/tags/pull-tricore-20141021:
  target-tricore: Add instructions of BO opcode format
  target-tricore: Add instructions of BIT opcode format
  target-tricore: Add instructions of B opcode format
  target-tricore: Add instructions of ABS, ABSB opcode format
  target-tricore: Cleanup and Bugfixes

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agoMerge remote-tracking branch 'remotes/cohuck/tags/s390x-20141021' into staging
Peter Maydell [Wed, 22 Oct 2014 17:41:38 +0000 (18:41 +0100)] 
Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20141021' into staging

add missing s390x files to MAINTAINERS

# gpg: Signature made Tue 21 Oct 2014 11:57:12 BST using RSA key ID C6F02FAF
# gpg: Good signature from "Cornelia Huck <huckc@linux.vnet.ibm.com>"
# gpg:                 aka "Cornelia Huck <cornelia.huck@de.ibm.com>"

* remotes/cohuck/tags/s390x-20141021:
  s390x: sweep up unmaintained files

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agoMerge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
Peter Maydell [Wed, 22 Oct 2014 15:39:49 +0000 (16:39 +0100)] 
Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging

Block patches

# gpg: Signature made Mon 20 Oct 2014 13:04:09 BST using RSA key ID C88F2FD6
# gpg: Good signature from "Kevin Wolf <kwolf@redhat.com>"

* remotes/kevin/tags/for-upstream: (28 commits)
  block: Make device model's references to BlockBackend strong
  block: Lift device model API into BlockBackend
  blockdev: Convert qmp_eject(), qmp_change_blockdev() to BlockBackend
  block/qapi: Convert qmp_query_block() to BlockBackend
  blockdev: Fix blockdev-add not to create DriveInfo
  blockdev: Drop superfluous DriveInfo member id
  pc87312: Drop unused members of PC87312State
  ide: Complete conversion from BlockDriverState to BlockBackend
  hw: Convert from BlockDriverState to BlockBackend, mostly
  virtio-blk: Rename VirtIOBlkConf variables to conf
  virtio-blk: Drop redundant VirtIOBlock member conf
  block: Rename BlockDriverCompletionFunc to BlockCompletionFunc
  block: Rename BlockDriverAIOCB* to BlockAIOCB*
  block: Eliminate DriveInfo member bdrv, use blk_by_legacy_dinfo()
  block: Merge BlockBackend and BlockDriverState name spaces
  block: Eliminate BlockDriverState member device_name[]
  block: Eliminate bdrv_iterate(), use bdrv_next()
  blockdev: Eliminate drive_del()
  block: Make BlockBackend own its BlockDriverState
  block: Code motion to get rid of stubs/blockdev.c
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agoMerge remote-tracking branch 'remotes/kraxel/tags/pull-usb-20141015-2' into staging
Peter Maydell [Wed, 22 Oct 2014 14:48:32 +0000 (15:48 +0100)] 
Merge remote-tracking branch 'remotes/kraxel/tags/pull-usb-20141015-2' into staging

usb: add high speed mouse & keyboard configuration

* remotes/kraxel/tags/pull-usb-20141015-2:
  xhci: remove dead code
  usb-hid: Add high speed keyboard configuration
  usb-hid: Add high speed mouse configuration
  usb-hid: Move descriptor decision to usb-hid initfn

Conflicts:
include/hw/i386/pc.h
[Fixed trivial merge conflict in the pc-2.1 property list]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agoMerge remote-tracking branch 'remotes/spice/tags/pull-spice-20141015-1' into staging
Peter Maydell [Wed, 22 Oct 2014 13:49:37 +0000 (14:49 +0100)] 
Merge remote-tracking branch 'remotes/spice/tags/pull-spice-20141015-1' into staging

qxl: keep going if reaching guest bug on empty area

# gpg: Signature made Wed 15 Oct 2014 11:45:37 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-20141015-1:
  qxl: keep going if reaching guest bug on empty area

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agoMerge remote-tracking branch 'remotes/kraxel/tags/pull-gtk-20141015-1' into staging
Peter Maydell [Wed, 22 Oct 2014 13:01:42 +0000 (14:01 +0100)] 
Merge remote-tracking branch 'remotes/kraxel/tags/pull-gtk-20141015-1' into staging

gtk: fix memory leak, add pause key support.

# gpg: Signature made Wed 15 Oct 2014 11:30:39 BST using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>"
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>"
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>"

* remotes/kraxel/tags/pull-gtk-20141015-1:
  gtk: add support for the Pause key
  gtk.c: Fix memory leak in gd_set_keycode_type()

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agoqga: Rewrite code where using readdir_r
zhanghailiang [Fri, 19 Sep 2014 03:09:10 +0000 (11:09 +0800)] 
qga: Rewrite code where using readdir_r

If readdir_r fails, error_setg_errno will reference the freed
pointer *dirpath*.

Moreover, readdir_r may cause a buffer overflow, using readdir instead.

Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
7 years agoMerge remote-tracking branch 'remotes/kraxel/tags/pull-console-20141015-1' into staging
Peter Maydell [Wed, 22 Oct 2014 12:08:43 +0000 (13:08 +0100)] 
Merge remote-tracking branch 'remotes/kraxel/tags/pull-console-20141015-1' into staging

configure: Prepend pixman and ftd flags to overrule system-provided ones

# gpg: Signature made Wed 15 Oct 2014 11:21:13 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-20141015-1:
  configure: Prepend pixman and ftd flags to overrule system-provided ones

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agoMerge remote-tracking branch 'remotes/lalrae/tags/mips-20141015' into staging
Peter Maydell [Wed, 22 Oct 2014 11:06:47 +0000 (12:06 +0100)] 
Merge remote-tracking branch 'remotes/lalrae/tags/mips-20141015' into staging

* remotes/lalrae/tags/mips-20141015: (28 commits)
  target-mips: Remove unused gen_load_ACX, gen_store_ACX and cpu_ACX
  target-mips/dsp_helper.c: Add ifdef guards around various functions
  target-mips/translate.c: Add ifdef guard around check_mips64()
  target-mips/op_helper.c: Remove unused do_lbu() function
  target-mips/dsp_helper.c: Remove unused function get_DSPControl_24()
  target-mips: fix broken MIPS16 and microMIPS
  target-mips/translate.c: Update OPC_SYNCI
  target-mips: define a new generic CPU supporting MIPS64 Release 6 ISA
  mips_malta: update malta's pseudo-bootloader - replace JR with JALR
  target-mips: remove JR, BLTZAL, BGEZAL and add NAL, BAL instructions
  target-mips: do not allow Status.FR=0 mode in 64-bit FPU
  target-mips: add new Floating Point Comparison instructions
  target-mips: add new Floating Point instructions
  softfloat: add functions corresponding to IEEE-2008 min/maxNumMag
  target-mips: add AUI, LSA and PCREL instruction families
  target-mips: add compact and CP1 branches
  target-mips: add ALIGN, DALIGN, BITSWAP and DBITSWAP instructions
  target-mips: Status.UX/SX/KX enable 32-bit address wrapping
  target-mips: move CLO, DCLO, CLZ, DCLZ, SDBBP and free special2 in R6
  target-mips: redefine Integer Multiply and Divide instructions
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agohw/i386/pc_q35.c: Avoid g_assert_cmpint() as it is not in glib 2.12
Peter Maydell [Fri, 17 Oct 2014 18:04:35 +0000 (19:04 +0100)] 
hw/i386/pc_q35.c: Avoid g_assert_cmpint() as it is not in glib 2.12

The function g_assert_cmpint() is not in glib 2.12, which is our current
minimum requirement. Rephrase the recently added assertion to avoid it.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Gonglei <arei.gonglei@huawei.com>
7 years agos390x: sweep up unmaintained files
Cornelia Huck [Mon, 20 Oct 2014 17:00:03 +0000 (19:00 +0200)] 
s390x: sweep up unmaintained files

Several s390x/kvm/ccw related files don't have an entry in MAINTAINERS:
Sort them into the appropriate sections.

Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Reviewed-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
7 years agoblock: Make device model's references to BlockBackend strong
Markus Armbruster [Tue, 7 Oct 2014 11:59:26 +0000 (13:59 +0200)] 
block: Make device model's references to BlockBackend strong

Doesn't make a difference just yet, but it's the right thing to do.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: Lift device model API into BlockBackend
Markus Armbruster [Tue, 7 Oct 2014 11:59:25 +0000 (13:59 +0200)] 
block: Lift device model API into BlockBackend

Move device model attachment / detachment and the BlockDevOps device
model callbacks and their wrappers from BlockDriverState to
BlockBackend.

Wrapper calls in block.c change from

    bdrv_dev_FOO_cb(bs, ...)

to

    if (bs->blk) {
        bdrv_dev_FOO_cb(bs->blk, ...);
    }

No change, because both bdrv_dev_change_media_cb() and
bdrv_dev_resize_cb() do nothing when no device model is attached, and
a device model can be attached only when bs->blk.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblockdev: Convert qmp_eject(), qmp_change_blockdev() to BlockBackend
Markus Armbruster [Tue, 7 Oct 2014 11:59:24 +0000 (13:59 +0200)] 
blockdev: Convert qmp_eject(), qmp_change_blockdev() to BlockBackend

Much more command code needs conversion.  I'm converting these now
because they're using bdrv_dev_* functions, which I'm about to lift
into BlockBackend.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock/qapi: Convert qmp_query_block() to BlockBackend
Markus Armbruster [Tue, 7 Oct 2014 11:59:23 +0000 (13:59 +0200)] 
block/qapi: Convert qmp_query_block() to BlockBackend

Much more command code needs conversion.  I start with this one
because it's using bdrv_dev_* functions, which I'm about to lift into
BlockBackend.

While there, give bdrv_query_info() internal linkage.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblockdev: Fix blockdev-add not to create DriveInfo
Markus Armbruster [Tue, 7 Oct 2014 11:59:22 +0000 (13:59 +0200)] 
blockdev: Fix blockdev-add not to create DriveInfo

blockdev_init() always creates a DriveInfo, but only drive_new() fills
it in.  qmp_blockdev_add() leaves it blank.  This results in a drive
with type = IF_IDE, bus = 0, unit = 0.  Screwed up in commit ee13ed1c.

Board initialization code looking for IDE drive (0,0) can pick up one
of these bogus drives.  The QMP command has to execute really early to
be visible.  Not sure how likely that is in practice.

Fix by creating DriveInfo in drive_new().  Block backends created by
blockdev-add don't get one.

Breaks the test for "has been created by qmp_blockdev_add()" in
blockdev_mark_auto_del() and do_drive_del(), because it changes the
value of dinfo && !dinfo->enable_auto_del from true to false.  Simply
test !dinfo instead.

Leaves DriveInfo member enable_auto_del unused.  Drop it.

A few places assume a block backend always has a DriveInfo.  Fix them
up.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblockdev: Drop superfluous DriveInfo member id
Markus Armbruster [Tue, 7 Oct 2014 11:59:21 +0000 (13:59 +0200)] 
blockdev: Drop superfluous DriveInfo member id

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agopc87312: Drop unused members of PC87312State
Markus Armbruster [Tue, 7 Oct 2014 11:59:20 +0000 (13:59 +0200)] 
pc87312: Drop unused members of PC87312State

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoide: Complete conversion from BlockDriverState to BlockBackend
Markus Armbruster [Tue, 7 Oct 2014 11:59:19 +0000 (13:59 +0200)] 
ide: Complete conversion from BlockDriverState to BlockBackend

Add a BlockBackend member to TrimAIOCB, so ide_issue_trim_cb() can use
blk_aio_discard() instead of bdrv_aio_discard().

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agohw: Convert from BlockDriverState to BlockBackend, mostly
Markus Armbruster [Tue, 7 Oct 2014 11:59:18 +0000 (13:59 +0200)] 
hw: Convert from BlockDriverState to BlockBackend, mostly

Device models should access their block backends only through the
block-backend.h API.  Convert them, and drop direct includes of
inappropriate headers.

Just four uses of BlockDriverState are left:

* The Xen paravirtual block device backend (xen_disk.c) opens images
  itself when set up via xenbus, bypassing blockdev.c.  I figure it
  should go through qmp_blockdev_add() instead.

* Device model "usb-storage" prompts for keys.  No other device model
  does, and this one probably shouldn't do it, either.

* ide_issue_trim_cb() uses bdrv_aio_discard() instead of
  blk_aio_discard() because it fishes its backend out of a BlockAIOCB,
  which has only the BlockDriverState.

* PC87312State has an unused BlockDriverState[] member.

The next two commits take care of the latter two.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agovirtio-blk: Rename VirtIOBlkConf variables to conf
Markus Armbruster [Tue, 7 Oct 2014 11:59:17 +0000 (13:59 +0200)] 
virtio-blk: Rename VirtIOBlkConf variables to conf

This is consistent with how VirtIOFOOConf variables are named
elsewhere, and makes blk available for BlockBackend variables.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agovirtio-blk: Drop redundant VirtIOBlock member conf
Markus Armbruster [Tue, 7 Oct 2014 11:59:16 +0000 (13:59 +0200)] 
virtio-blk: Drop redundant VirtIOBlock member conf

Commit 12c5674 turned it into a pointer to member blk.conf.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: Rename BlockDriverCompletionFunc to BlockCompletionFunc
Markus Armbruster [Tue, 7 Oct 2014 11:59:15 +0000 (13:59 +0200)] 
block: Rename BlockDriverCompletionFunc to BlockCompletionFunc

I'll use it with block backends shortly, and the name is going to fit
badly there.  It's a block layer thing anyway, not just a block driver
thing.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: Rename BlockDriverAIOCB* to BlockAIOCB*
Markus Armbruster [Tue, 7 Oct 2014 11:59:14 +0000 (13:59 +0200)] 
block: Rename BlockDriverAIOCB* to BlockAIOCB*

I'll use BlockDriverAIOCB with block backends shortly, and the name is
going to fit badly there.  It's a block layer thing anyway, not just a
block driver thing.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: Eliminate DriveInfo member bdrv, use blk_by_legacy_dinfo()
Markus Armbruster [Tue, 7 Oct 2014 11:59:13 +0000 (13:59 +0200)] 
block: Eliminate DriveInfo member bdrv, use blk_by_legacy_dinfo()

The patch is big, but all it really does is replacing

    dinfo->bdrv

by

    blk_bs(blk_by_legacy_dinfo(dinfo))

The replacement is repetitive, but the conversion of device models to
BlockBackend is imminent, and will shorten it to just
blk_legacy_dinfo(dinfo).

Line wrapping muddies the waters a bit.  I also omit tests whether
dinfo->bdrv is null, because it never is.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: Merge BlockBackend and BlockDriverState name spaces
Markus Armbruster [Tue, 7 Oct 2014 11:59:12 +0000 (13:59 +0200)] 
block: Merge BlockBackend and BlockDriverState name spaces

BlockBackend's name space is separate only to keep the initial patches
simple.  Time to merge the two.

Retain bdrv_find() and bdrv_get_device_name() for now, to keep this
series manageable.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: Eliminate BlockDriverState member device_name[]
Markus Armbruster [Tue, 7 Oct 2014 11:59:11 +0000 (13:59 +0200)] 
block: Eliminate BlockDriverState member device_name[]

device_name[] can become non-empty only in bdrv_new_root() and
bdrv_move_feature_fields().  The latter is used only to undo damage
done by bdrv_swap().  The former is called only by blk_new_with_bs().
Therefore, when a BlockDriverState's device_name[] is non-empty, then
it's been created with a BlockBackend, and vice versa.  Furthermore,
blk_new_with_bs() keeps the two names equal.

Therefore, device_name[] is redundant.  Eliminate it.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: Eliminate bdrv_iterate(), use bdrv_next()
Markus Armbruster [Tue, 7 Oct 2014 11:59:10 +0000 (13:59 +0200)] 
block: Eliminate bdrv_iterate(), use bdrv_next()

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblockdev: Eliminate drive_del()
Markus Armbruster [Tue, 7 Oct 2014 11:59:09 +0000 (13:59 +0200)] 
blockdev: Eliminate drive_del()

drive_del() has become a trivial wrapper around blk_unref().  Get rid
of it.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: Make BlockBackend own its BlockDriverState
Markus Armbruster [Tue, 7 Oct 2014 11:59:08 +0000 (13:59 +0200)] 
block: Make BlockBackend own its BlockDriverState

On BlockBackend destruction, unref its BlockDriverState.  Replaces the
callers' unrefs.

This turns the pointer from BlockBackend to BlockDriverState into a
strong reference, managed with bdrv_ref() / bdrv_unref().  The
back-pointer remains weak.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: Code motion to get rid of stubs/blockdev.c
Markus Armbruster [Tue, 7 Oct 2014 11:59:07 +0000 (13:59 +0200)] 
block: Code motion to get rid of stubs/blockdev.c

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: Connect BlockBackend and DriveInfo
Markus Armbruster [Tue, 7 Oct 2014 11:59:06 +0000 (13:59 +0200)] 
block: Connect BlockBackend and DriveInfo

Make the BlockBackend own the DriveInfo.  Change blockdev_init() to
return the BlockBackend instead of the DriveInfo.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: Connect BlockBackend to BlockDriverState
Markus Armbruster [Tue, 7 Oct 2014 11:59:05 +0000 (13:59 +0200)] 
block: Connect BlockBackend to BlockDriverState

Convenience function blk_new_with_bs() creates a BlockBackend with its
BlockDriverState.  Callers have to unref both.  The commit after next
will relieve them of the need to unref the BlockDriverState.

Complication: due to the silly way drive_del works, we need a way to
hide a BlockBackend, just like bdrv_make_anon().  To emphasize its
"special" status, give the function a suitably off-putting name:
blk_hide_on_behalf_of_do_drive_del().  Unfortunately, hiding turns the
BlockBackend's name into the empty string.  Can't avoid that without
breaking the blk->bs->device_name equals blk->name invariant.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: New BlockBackend
Markus Armbruster [Tue, 7 Oct 2014 11:59:04 +0000 (13:59 +0200)] 
block: New BlockBackend

A block device consists of a frontend device model and a backend.

A block backend has a tree of block drivers doing the actual work.
The tree is managed by the block layer.

We currently use a single abstraction BlockDriverState both for tree
nodes and the backend as a whole.  Drawbacks:

* Its API includes both stuff that makes sense only at the block
  backend level (root of the tree) and stuff that's only for use
  within the block layer.  This makes the API bigger and more complex
  than necessary.  Moreover, it's not obvious which interfaces are
  meant for device models, and which really aren't.

* Since device models keep a reference to their backend, the backend
  object can't just be destroyed.  But for media change, we need to
  replace the tree.  Our solution is to make the BlockDriverState
  generic, with actual driver state in a separate object, pointed to
  by member opaque.  That lets us replace the tree by deinitializing
  and reinitializing its root.  This special need of the root makes
  the data structure awkward everywhere in the tree.

The general plan is to separate the APIs into "block backend", for use
by device models, monitor and whatever other code dealing with block
backends, and "block driver", for use by the block layer and whatever
other code (if any) dealing with trees and tree nodes.

Code dealing with block backends, device models in particular, should
become completely oblivious of BlockDriverState.  This should let us
clean up both APIs, and the tree data structures.

This commit is a first step.  It creates a minimal "block backend"
API: type BlockBackend and functions to create, destroy and find them.

BlockBackend objects are created and destroyed exactly when root
BlockDriverState objects are created and destroyed.  "Root" in the
sense of "in bdrv_states".  They're not yet used for anything; that'll
come shortly.

A root BlockDriverState is created with bdrv_new_root(), so where to
create a BlockBackend is obvious.  Where these roots get destroyed
isn't always as obvious.

It is obvious in qemu-img.c, qemu-io.c and qemu-nbd.c, and in error
paths of blockdev_init(), blk_connect().  That leaves destruction of
objects successfully created by blockdev_init() and blk_connect().

blockdev_init() is used only by drive_new() and qmp_blockdev_add().
Objects created by the latter are currently indestructible (see commit
48f364d "blockdev: Refuse to drive_del something added with
blockdev-add" and commit 2d246f0 "blockdev: Introduce
DriveInfo.enable_auto_del").  Objects created by the former get
destroyed by drive_del().

Objects created by blk_connect() get destroyed by blk_disconnect().

BlockBackend is reference-counted.  Its reference count never exceeds
one so far, but that's going to change.

In drive_del(), the BB's reference count is surely one now.  The BDS's
reference count is greater than one when something else is holding a
reference, such as a block job.  In this case, the BB is destroyed
right away, but the BDS lives on until all extra references get
dropped.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock: Split bdrv_new_root() off bdrv_new()
Markus Armbruster [Tue, 7 Oct 2014 11:59:03 +0000 (13:59 +0200)] 
block: Split bdrv_new_root() off bdrv_new()

Creating an anonymous BDS can't fail.  Make that obvious.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agonbd: Fix filename generation
Max Reitz [Wed, 8 Oct 2014 17:55:15 +0000 (19:55 +0200)] 
nbd: Fix filename generation

Export names may be used with nbd+unix, too, fix nbd_refresh_filename()
accordingly. Also, for nbd+tcp, the documented path schema is
"nbd://host[:port]/export", so use it. Furthermore, as can be seen from
that schema, the port is optional.

That makes six single cases for how the filename can be formatted; it is
not easy to generalize these cases without the resulting statement being
completely unreadable, thus there is simply one snprintf() per case.

Finally, taking the options from BDRVNBDState::socket_opts is wrong,
because those will not contain the export name. Just use
BlockDriverState::options instead.

Reported-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock/raw-posix: use seek_hole ahead of fiemap
Tony Breeds [Thu, 25 Sep 2014 23:14:12 +0000 (09:14 +1000)] 
block/raw-posix: use seek_hole ahead of fiemap

try_fiemap() uses FIEMAP_FLAG_SYNC which has a significant performance
impact.

Prefer seek_hole() over fiemap() to avoid this impact where possible.
seek_hole is more widely used and, arguably, has potential to be
optimised in the kernel.

Reported-By: Michael Steffens <michael_steffens@posteo.de>
Signed-off-by: Tony Breeds <tony@bakeyournoodle.com>
Cc: Kevin Wolf <kwolf@redhat.com>
Cc: Markus Armbruster <armbru@redhat.com>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Cc: Max Reitz <mreitz@redhat.com>
Cc: Pádraig Brady <pbrady@redhat.com>
Cc: Eric Blake <eblake@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoblock/raw-posix: Fix disk corruption in try_fiemap
Tony Breeds [Thu, 25 Sep 2014 23:14:11 +0000 (09:14 +1000)] 
block/raw-posix: Fix disk corruption in try_fiemap

Using fiemap without FIEMAP_FLAG_SYNC is a known corrupter.

Add the FIEMAP_FLAG_SYNC flag to the FS_IOC_FIEMAP ioctl.  This has
the downside of significantly reducing performance.

Reported-By: Michael Steffens <michael_steffens@posteo.de>
Signed-off-by: Tony Breeds <tony@bakeyournoodle.com>
Cc: Kevin Wolf <kwolf@redhat.com>
Cc: Markus Armbruster <armbru@redhat.com>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Cc: Max Reitz <mreitz@redhat.com>
Cc: Pádraig Brady <pbrady@redhat.com>
Cc: Eric Blake <eblake@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agoqcow2: fix leak of Qcow2DiscardRegion in update_refcount_discard
Zhang Haoyu [Sat, 11 Oct 2014 08:35:43 +0000 (16:35 +0800)] 
qcow2: fix leak of Qcow2DiscardRegion in update_refcount_discard

When the Qcow2DiscardRegion is adjacent to another one referenced by "d",
free this Qcow2DiscardRegion metadata referenced by "p" after
it was removed from s->discards queue.

Signed-off-by: Zhang Haoyu <zhanghy@sangfor.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
7 years agotarget-tricore: Add instructions of BO opcode format
Bastian Koppelmann [Fri, 26 Sep 2014 19:36:09 +0000 (20:36 +0100)] 
target-tricore: Add instructions of BO opcode format

Add instructions of BO opcode format.
Add microcode generator functions gen_swap, gen_ldmst.
Add microcode generator functions gen_st/ld_preincr, which write back the address after the memory access.
Add helper for circular and bit reverse addr mode calculation.
Add sign extended bitmask for BO_OFF10 field.

Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
Reviewed-by: Richard Henderson <rth@twiddle.net>
7 years agotarget-tricore: Add instructions of BIT opcode format
Bastian Koppelmann [Fri, 26 Sep 2014 19:34:22 +0000 (20:34 +0100)] 
target-tricore: Add instructions of BIT opcode format

Add instructions of BIT opcode format.
Add microcode generator functions gen_bit_1/2op to do 1/2 bit operations on the last bit.

Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
Reviewed-by: Richard Henderson <rth@twiddle.net>
7 years agotarget-tricore: Add instructions of B opcode format
Bastian Koppelmann [Fri, 26 Sep 2014 19:28:02 +0000 (20:28 +0100)] 
target-tricore: Add instructions of B opcode format

Add instructions of B opcode format.

Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
Reviewed-by: Richard Henderson <rth@twiddle.net>
7 years agotarget-tricore: Add instructions of ABS, ABSB opcode format
Bastian Koppelmann [Fri, 26 Sep 2014 19:26:31 +0000 (20:26 +0100)] 
target-tricore: Add instructions of ABS, ABSB opcode format

Add instructions of ABS, ABSB opcode format.
Add microcode generator functions for ld/st of two 32bit reg as one 64bit value.
Add microcode generator functions for ldmst and swap.
Add helper ldlcx, lducx, stlcx and stucx.

Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
Reviewed-by: Richard Henderson <rth@twiddle.net>
7 years agotarget-tricore: Cleanup and Bugfixes
Bastian Koppelmann [Fri, 26 Sep 2014 19:04:51 +0000 (20:04 +0100)] 
target-tricore: Cleanup and Bugfixes

Move FCX loading of save_context_ to caller functions, for STLCX, STUCX insn to use those functions.
Move FCX storing of restore_context_ to caller functions, for LDLCX, LDUCX insn to use those functions.
Remove do_raise_exception function, which caused clang to emit a warning.
Fix: save_context_lower now saves a[11] instead of PSW.
Fix: MASK_OP_ABSB_BPOS starting at wrong offset.

Signed-off-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
Reviewed-by: Richard Henderson <rth@twiddle.net>
7 years agoglib: add compatibility interface for g_strcmp0()
Gonglei [Thu, 16 Oct 2014 10:59:37 +0000 (18:59 +0800)] 
glib: add compatibility interface for g_strcmp0()

This patch fixes compilation errors when building against glib < 2.16.0
due to the missing g_strcmp0() function.

Suggested-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Message-id: 1413457177-10132-1-git-send-email-arei.gonglei@huawei.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agoMerge remote-tracking branch 'remotes/kraxel/tags/pull-vga-20141015-1' into staging
Peter Maydell [Thu, 16 Oct 2014 08:26:14 +0000 (09:26 +0100)] 
Merge remote-tracking branch 'remotes/kraxel/tags/pull-vga-20141015-1' into staging

vga-pci: add qext region to mmio
vga: Remove unused arrays dmask4 and dmask16

# gpg: Signature made Wed 15 Oct 2014 10:12:06 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-20141015-1:
  hw/display/vga: Remove unused arrays dmask4 and dmask16
  vga-pci: add qext region to mmio

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agoMerge remote-tracking branch 'remotes/kraxel/tags/pull-bootindex-20141015-1' into...
Peter Maydell [Wed, 15 Oct 2014 20:03:28 +0000 (21:03 +0100)] 
Merge remote-tracking branch 'remotes/kraxel/tags/pull-bootindex-20141015-1' into staging

allow changing bootorder via monitor at runtime,
by making bootindex a writable qom property.

* remotes/kraxel/tags/pull-bootindex-20141015-1: (34 commits)
  bootindex: change fprintf to error_report
  bootindex: delete bootindex when device is removed
  bootindex: move calling add_boot_device_patch to bootindex setter function
  ide: add calling add_boot_device_patch in bootindex setter function
  nvma: ide: add bootindex to qom property
  usb-storage: add bootindex to qom property
  virtio-blk: alias bootindex property explicitly for virt-blk-pci/ccw/s390
  block: remove bootindex property from qdev to qom
  virtio-blk: add bootindex to qom property
  ide: add bootindex to qom property
  scsi: add bootindex to qom property
  isa-fdc: remove bootindexA/B property from qdev to qom
  redirect: remove bootindex property from qdev to qom
  vfio: remove bootindex property from qdev to qom
  pci-assign: remove bootindex property from qdev to qom
  host-libusb: remove bootindex property from qdev to qom
  virtio-net: alias bootindex property explicitly for virt-net-pci/ccw/s390
  net: remove bootindex property from qdev to qom
  usb-net: add bootindex to qom property
  vmxnet3: add bootindex to qom property
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agoglib: add compatibility interface for g_get_monotonic_time()
Stefan Hajnoczi [Wed, 15 Oct 2014 12:29:30 +0000 (14:29 +0200)] 
glib: add compatibility interface for g_get_monotonic_time()

This patch fixes compilation errors when building against glib <2.28.0
due to the missing g_get_monotonic_time() function.

The compilation error in tests/libqos/virtio.c was introduced in commit
70556264a89a268efba1d7e8e341adcdd7881eb4 ("libqos: use microseconds
instead of iterations for virtio timeout").

Add a simple g_get_monotonic_time() implementation to glib-compat.h
based on code from vhost-user-test.c.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
[Igor: add G_TIME_SPAN_SECOND, include glib-compat.h in libqtest.h]
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agoxhci: remove dead code
Gerd Hoffmann [Tue, 14 Oct 2014 06:40:22 +0000 (08:40 +0200)] 
xhci: remove dead code

Reported-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agousb-hid: Add high speed keyboard configuration
Jan Vesely [Tue, 30 Sep 2014 02:21:11 +0000 (22:21 -0400)] 
usb-hid: Add high speed keyboard configuration

Signed-off-by: Jan Vesely <jano.vesely@gmail.com>
[ kraxel: fixup compat property to apply to 2.1 & older ]

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agousb-hid: Add high speed mouse configuration
Jan Vesely [Tue, 30 Sep 2014 02:21:10 +0000 (22:21 -0400)] 
usb-hid: Add high speed mouse configuration

Signed-off-by: Jan Vesely <jano.vesely@gmail.com>
[ kraxel: fixup compat property to apply to 2.1 & older ]

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agoMerge remote-tracking branch 'remotes/juanquintela/tags/migration/20141015' into...
Peter Maydell [Wed, 15 Oct 2014 10:55:53 +0000 (11:55 +0100)] 
Merge remote-tracking branch 'remotes/juanquintela/tags/migration/20141015' into staging

migration/next for 20141015

# gpg: Signature made Wed 15 Oct 2014 09:21:54 BST using RSA key ID 5872D723
# gpg: Can't check signature: public key not found

* remotes/juanquintela/tags/migration/20141015:
  migration: catch unknown flag combinations in ram_load
  qemu-file: Move stdio implementation to qemu-file-stdio.c
  qemu-file: Move unix and socket implementations to qemu-file-unix.c
  qemu-file: Use qemu_file_is_writable() on stdio_fclose()
  qemu-file: Make qemu_file_is_writable() non-static
  qemu-file: Add copyright header to qemu-file.c
  vmstate: Allow dynamic allocation for VBUFFER during migration
  block/migration: Disable cache invalidate for incoming migration
  Tests: QEMUSizedBuffer/QEMUBuffer
  QEMUSizedBuffer based QEMUFile

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
7 years agoconfigure: Prepend pixman and ftd flags to overrule system-provided ones
Jan Kiszka [Thu, 2 Oct 2014 08:03:55 +0000 (10:03 +0200)] 
configure: Prepend pixman and ftd flags to overrule system-provided ones

Other packages may provide includes for pixman as well if the host has a
devel package installed. So add ours to the front to unsure that the
right version is used.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agohw/display/vga: Remove unused arrays dmask4 and dmask16
Peter Maydell [Fri, 10 Oct 2014 19:44:29 +0000 (20:44 +0100)] 
hw/display/vga: Remove unused arrays dmask4 and dmask16

Following cleanup of the vga device code in commit d2e043a8041,
the arrays dmask4 and dmask16 are now unused. gcc doesn't warn
about this, but clang does; remove them.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agovga-pci: add qext region to mmio
Gerd Hoffmann [Tue, 23 Sep 2014 10:45:56 +0000 (12:45 +0200)] 
vga-pci: add qext region to mmio

Add a qemu extented register range to the standard vga mmio bar.
Right nowe there are two registers:  One readonly register returning the
size of the region (so we can easily add more registers there if needed)
and one endian control register, so guests (especially ppc) can flip
the framebuffer endianness as they need it.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
7 years agoqxl: keep going if reaching guest bug on empty area
Marc-André Lureau [Fri, 10 Oct 2014 17:05:11 +0000 (19:05 +0200)] 
qxl: keep going if reaching guest bug on empty area

Xorg server hangs when using xfig and typing a text with space:
 #0  qxl_wait_for_io_command (qxl=<value optimized out>) at qxl_io.c:47
 #1  0x00007f826a49a299 in qxl_download_box (surface=0x221d030, x1=231, y1=259,
     x2=<value optimized out>, y2=<value optimized out>) at qxl_surface.c:143

       while (!(ram_header->int_pending & QXL_INTERRUPT_IO_CMD))
         usleep (1);

The QXL driver is calling QXL_IO_UPDATE_AREA with an empty area. This
is a guest bug. The call is async and no ack is sent back on guest
bug, so the X server will hang. The driver should be improved to avoid
this situation and also to abort on QXL_INTERRUPT_ERROR. This will be
a different patch series for the driver. However, it is simple enough
to keep qemu running on empty areas update, which is what this patch
provides.

https://bugzilla.redhat.com/show_bug.cgi?id=1151363

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agousb-hid: Move descriptor decision to usb-hid initfn
Jan Vesely [Tue, 30 Sep 2014 02:21:09 +0000 (22:21 -0400)] 
usb-hid: Move descriptor decision to usb-hid initfn

Signed-off-by: Jan Vesely <jano.vesely@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agogtk: add support for the Pause key
Martin Decky [Tue, 16 Sep 2014 14:04:40 +0000 (16:04 +0200)] 
gtk: add support for the Pause key

Special handing of the Pause key. Implemented in a similar way as in
ui/sdl.c.

Signed-off-by: Martin Decky <martin@decky.cz>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agogtk.c: Fix memory leak in gd_set_keycode_type()
Chen Fan [Tue, 2 Sep 2014 06:33:59 +0000 (14:33 +0800)] 
gtk.c: Fix memory leak in gd_set_keycode_type()

 this memory leak is introduced by the original
 commit 3158a3482b0093e41f2b2596fba50774ea31ae08

 valgrind out showing:
 ==14553== 21,459 (72 direct, 21,387 indirect) bytes in 1 blocks are definitely
 lost in loss record 8,055 of 8,082
 ==14553==    at 0x4A06BC3: calloc (vg_replace_malloc.c:618)
 ==14553==    by 0x80DBFBC: XkbGetKeyboardByName (in /usr/lib64/libX11.so.6.3.0)
 ==14553==    by 0x40C704: gtk_display_init (gtk.c:1798)
 ==14553==    by 0x1AEDC1: main (vl.c:4480)

Signed-off-by: Chen Fan <chen.fan.fnst@cn.fujitsu.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agobootindex: change fprintf to error_report
Gonglei [Tue, 7 Oct 2014 08:00:38 +0000 (16:00 +0800)] 
bootindex: change fprintf to error_report

The function may be called by qmp command, we should
report error message to the caller.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agobootindex: delete bootindex when device is removed
Gonglei [Tue, 7 Oct 2014 08:00:37 +0000 (16:00 +0800)] 
bootindex: delete bootindex when device is removed

Device should be removed from global boot list when
it is hot-unplugged.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agobootindex: move calling add_boot_device_patch to bootindex setter function
Gonglei [Tue, 7 Oct 2014 08:00:36 +0000 (16:00 +0800)] 
bootindex: move calling add_boot_device_patch to bootindex setter function

On this way, we can assure the new bootindex take effect
during vm rebooting.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agoide: add calling add_boot_device_patch in bootindex setter function
Gonglei [Tue, 7 Oct 2014 08:00:35 +0000 (16:00 +0800)] 
ide: add calling add_boot_device_patch in bootindex setter function

On this way, we can assure the new bootindex take effect
during vm rebooting. Meanwhile set the initial value of
bootindex to -1.

Because ide devcies's unit property maybe
do not initialize when set_bootindex function is called,
so that we don't know its suffix. So we have to save the
call add_boot_device_path() on ide realize/init function.
When we want to change bootindex during vm rebooting, we
can call it in setter function.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agonvma: ide: add bootindex to qom property
Gonglei [Tue, 7 Oct 2014 08:00:34 +0000 (16:00 +0800)] 
nvma: ide: add bootindex to qom property

At present, nvma cannot boot. However, it provides already
a bootindex property, so change bootindex to qom for nvma
device, but not call add_boot_device_path.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agousb-storage: add bootindex to qom property
Gonglei [Tue, 7 Oct 2014 08:00:33 +0000 (16:00 +0800)] 
usb-storage: add bootindex to qom property

Add a qom property with the same name 'bootindex',
when we remove it form qdev property, things will
continue to work just fine, and we can use qom features
which are not supported by qdev property.

Because usb-storage rely on scsi-disk which is created
in usb_msg_realize_storage(), so we should store the SCSIDevice
pointer in MSDState struct. Only in this way, we can change
the global boot_order_list when we want to change the bootindex
during vm rebooting by calling object_property_set_int(Object(SCSIDevice),).

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agovirtio-blk: alias bootindex property explicitly for virt-blk-pci/ccw/s390
Gonglei [Tue, 7 Oct 2014 08:00:32 +0000 (16:00 +0800)] 
virtio-blk: alias bootindex property explicitly for virt-blk-pci/ccw/s390

Since the "bootindex" property is a QOM property and not a qdev property
now, we must alias it explicitly for virtio-blk-pci, as well as CCW and
s390-virtio.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agoblock: remove bootindex property from qdev to qom
Gonglei [Tue, 7 Oct 2014 08:00:31 +0000 (16:00 +0800)] 
block: remove bootindex property from qdev to qom

Remove bootindex form qdev property to qom, things will
continue to work just fine, and we can use qom features
which are not supported by qdev property.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agovirtio-blk: add bootindex to qom property
Gonglei [Tue, 7 Oct 2014 08:00:30 +0000 (16:00 +0800)] 
virtio-blk: add bootindex to qom property

Add a qom property with the same name 'bootindex',
when we remove it form qdev property, things will
continue to work just fine, and we can use qom features
which are not supported by qdev property.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agoide: add bootindex to qom property
Gonglei [Tue, 7 Oct 2014 08:00:29 +0000 (16:00 +0800)] 
ide: add bootindex to qom property

Add a qom property with the same name 'bootindex',
when we remove it form qdev property, things will
continue to work just fine, and we can use qom features
which are not supported by qdev property.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agoscsi: add bootindex to qom property
Gonglei [Tue, 7 Oct 2014 08:00:28 +0000 (16:00 +0800)] 
scsi: add bootindex to qom property

Add a qom property with the same name 'bootindex',
when we remove it form qdev property, things will
continue to work just fine, and we can use qom features
which are not supported by qdev property.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agoisa-fdc: remove bootindexA/B property from qdev to qom
Gonglei [Tue, 7 Oct 2014 08:00:27 +0000 (16:00 +0800)] 
isa-fdc: remove bootindexA/B property from qdev to qom

Remove bootindexA/B form qdev property to qom, things will
continue to work just fine, and we can use qom features
which are not supported by qdev property.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agoredirect: remove bootindex property from qdev to qom
Gonglei [Tue, 7 Oct 2014 08:00:26 +0000 (16:00 +0800)] 
redirect: remove bootindex property from qdev to qom

Remove bootindex form qdev property to qom, things will
continue to work just fine, and we can use qom features
which are not supported by qdev property.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agovfio: remove bootindex property from qdev to qom
Gonglei [Tue, 7 Oct 2014 08:00:25 +0000 (16:00 +0800)] 
vfio: remove bootindex property from qdev to qom

Remove bootindex form qdev property to qom, things will
continue to work just fine, and we can use qom features
which are not supported by qdev property.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agopci-assign: remove bootindex property from qdev to qom
Gonglei [Tue, 7 Oct 2014 08:00:24 +0000 (16:00 +0800)] 
pci-assign: remove bootindex property from qdev to qom

Remove bootindex form qdev property to qom, things will
continue to work just fine, and we can use qom features
which are not supported by qdev property.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
7 years agohost-libusb: remove bootindex property from qdev to qom
Gonglei [Tue, 7 Oct 2014 08:00:23 +0000 (16:00 +0800)] 
host-libusb: remove bootindex property from qdev to qom

Remove bootindex form qdev property to qom, things will
continue to work just fine, and we can use qom features
which are not supported by qdev property.

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>