nbd: simplify the nbd_request and nbd_reply structs

These structs are never used to represent the bytes that go over the
network.  The big-endian network data is built into a uint8_t array
in nbd_{receive,send}_{request,reply}.  Remove the unused magic field,
reorder the struct to avoid holes, and remove the packed attribute.

Signed-off-by: Paolo Bonzini <[email protected]>

nbd: Don't use cpu_to_*w() functions

The cpu_to_*w() functions just compose a pointer dereference
with a byteswap. Instead use st*_p(), which handles potential
pointer misalignment and avoids the need to cast the pointer.

Signed-off-by: Peter Maydell <[email protected]>
Message-Id: <1465575342[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>

nbd: Don't use *_to_cpup() functions

The *_to_cpup() functions are not very useful, as they simply do
a pointer dereference and then a *_to_cpu(). Instead use either:
 * ld*_*_p(), if the data is at an address that might not be
   correctly aligned for the load
 * a local dereference and *_to_cpu(), if the pointer is
   the correct type and known to be correctly aligned

Signed-off-by: Peter Maydell <[email protected]>
Message-Id: <1465570836[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>

configure: Remove unused CONFIG_SIGEV_THREAD_ID switch

The CONFIG_SIGEV_THREAD_ID switch is unused since the related code
has been removed by commit 6d327171551a12b937c5718073b9848d0274c74d
("aio / timers: Remove alarm timers"), so it can safely be removed
nowadays.

Signed-off-by: Thomas Huth <[email protected]>
Message-Id: <1465571084[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>

avx2 configure: Use primitives in test

Use the avx2 primitives during the test, thus making sure that the
compiler and assembler could actually use avx2.

This also detects the failure case on gcc 4.8.x with -save-temps
and avoids the need for the gcc version check in cutils.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Message-Id: <1465557378[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>

Make avx2 configure test work with -O2

When configured with --extra-cflags=-O2 gcc optimised out the test
and the readelf failed the check leaving avx2 disabled.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Message-Id: <1465557378[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>

Makefile: Fix tag file generation targets

"ctags" produces a file named "tags", not "ctags". It doesn't look
reasonable to use phony target name as a file name to remove. Just use
exact file names to remove in "ctags" and "TAGS" target receipts.

Signed-off-by: Sergey Fedorov <[email protected]>
Signed-off-by: Sergey Fedorov <[email protected]>
Message-Id: <1465495115[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>

configure: Enable -Werror for MinGW builds, too

MinGW seems to compile currently without warnings, so it should
be safe to enable -Werror now for this environment, too.

Signed-off-by: Thomas Huth <[email protected]>
Message-Id: <1465373606[email protected]>
Tested-by: Stefan Hajnoczi <[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>

clean-includes: run it once more

Signed-off-by: Paolo Bonzini <[email protected]>

os-posix: include sys/mman.h

qemu/osdep.h checks whether MAP_ANONYMOUS is defined, but this check
is bogus without a previous inclusion of sys/mman.h.  Include it in
sysemu/os-posix.h and remove it from everywhere else.

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

configure: Remove unused CONFIG_ZERO_MALLOC setting

CONFIG_ZERO_MALLOC was only used in qemu-malloc.c and
this file has been removed with the following commit:

41a748265f4879b52b0e87ff9c93bed975163886
Remove qemu_malloc/qemu_free

So we don't need this configuration setting anymore.
This patch also removes the z_version variable, since
this is now also not needed anymore.

Signed-off-by: Thomas Huth <[email protected]>
Reviewed-by: Laurent Vivier <[email protected]>
Message-Id: <1465398683[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>

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

Block layer patches

# gpg: Signature made Thu 16 Jun 2016 15:01:27 BST
# gpg:                using RSA key 0x7F09B272C88F2FD6
# gpg: Good signature from "Kevin Wolf <[email protected]>"
# Primary key fingerprint: DC3D EB15 9A9A F95D 3D74  56FE 7F09 B272 C88F 2FD6

* remotes/kevin/tags/for-upstream: (39 commits)
  hbitmap: add 'pos < size' asserts
  iotests: Add test for oVirt-like storage migration
  iotests: Add test for post-mirror backing chains
  block/null: Implement bdrv_refresh_filename()
  block/mirror: Fix target backing BDS
  block: Allow replacement of a BDS by its overlay
  rbd:change error_setg() to error_setg_errno()
  iotests: 095: Clean up QEMU before showing image info
  block: Create the commit block job before reopening any image
  block: Prevent sleeping jobs from resuming if they have been paused
  block: use the block job list in qmp_query_block_jobs()
  block: use the block job list in bdrv_drain_all()
  block: Fix snapshot=on with aio=native
  block: Remove bs->zero_beyond_eof
  qcow2: Let vmstate call qcow2_co_preadv/pwrite directly
  block: Make bdrv_load/save_vmstate coroutine_fns
  block: Allow .bdrv_load/save_vmstate() to return 0/-errno
  block: Make .bdrv_load_vmstate() vectored
  block: Introduce bdrv_preadv()
  doc: Fix mailing list address in tests/qemu-iotests/README
  ...

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

Merge remote-tracking branch 'mreitz/tags/pull-block-for-kevin-2016-06-16' into queue-block

Block patches

# gpg: Signature made Thu Jun 16 15:21:35 2016 CEST
# gpg:                using RSA key 0x3BB14202E838ACAD
# gpg: Good signature from "Max Reitz <[email protected]>"
# Primary key fingerprint: 91BE B60A 30DB 3E88 57D1  1829 F407 DB00 61D5 CF40
#      Subkey fingerprint: 58B3 81CE 2DC8 9CF9 9730  EE64 3BB1 4202 E838 ACAD

* mreitz/tags/pull-block-for-kevin-2016-06-16:
  hbitmap: add 'pos < size' asserts
  iotests: Add test for oVirt-like storage migration
  iotests: Add test for post-mirror backing chains
  block/null: Implement bdrv_refresh_filename()
  block/mirror: Fix target backing BDS
  block: Allow replacement of a BDS by its overlay
  rbd:change error_setg() to error_setg_errno()
  iotests: 095: Clean up QEMU before showing image info
  block: Create the commit block job before reopening any image
  block: Prevent sleeping jobs from resuming if they have been paused
  block: use the block job list in qmp_query_block_jobs()
  block: use the block job list in bdrv_drain_all()

Signed-off-by: Kevin Wolf <[email protected]>

hbitmap: add 'pos < size' asserts

For now, fail in hbitmap_set on start + count > size will come from
hbitmap_set
  hb_count_between
    hbitmap_iter_init
      assert(pos < hb->size)

This patch adds such checks to set/get/reset functions of hbitmap.

Signed-off-by: Vladimir Sementsov-Ogievskiy <[email protected]>
Message-id: 1465924093[email protected]
Signed-off-by: Max Reitz <[email protected]>

iotests: Add test for oVirt-like storage migration

Signed-off-by: Max Reitz <[email protected]>
Message-id: 20160610185750[email protected]
Reviewed-by: Kevin Wolf <[email protected]>
Reviewed-by: Fam Zheng <[email protected]>
Signed-off-by: Max Reitz <[email protected]>

iotests: Add test for post-mirror backing chains

Signed-off-by: Max Reitz <[email protected]>
Message-id: 20160610185750[email protected]
Reviewed-by: Fam Zheng <[email protected]>
[[email protected]: Removed unnecessary imports]
Signed-off-by: Max Reitz <[email protected]>

block/null: Implement bdrv_refresh_filename()

The null block driver ignores any filename used for creating its BDSs,
which allows creating such BDSs even without any filename at all. In
that case, we currently construct a JSON filename when queried instead
of a plain "null-co://" or "null-aio://". This patch implements
bdrv_refresh_filename() to remedy this behavior.

Signed-off-by: Max Reitz <[email protected]>
Message-id: 20160610185750[email protected]
[[email protected]: Added commit message]
Signed-off-by: Max Reitz <[email protected]>

block/mirror: Fix target backing BDS

Currently, we are trying to move the backing BDS from the source to the
target in bdrv_replace_in_backing_chain() which is called from
mirror_exit(). However, mirror_complete() already tries to open the
target's backing chain with a call to bdrv_open_backing_file().

First, we should only set the target's backing BDS once. Second, the
mirroring block job has a better idea of what to set it to than the
generic code in bdrv_replace_in_backing_chain() (in fact, the latter's
conditions on when to move the backing BDS from source to target are not
really correct).

Therefore, remove that code from bdrv_replace_in_backing_chain() and
leave it to mirror_complete().

Depending on what kind of mirroring is performed, we furthermore want to
use different strategies to open the target's backing chain:

- If blockdev-mirror is used, we can assume the user made sure that the
  target already has the correct backing chain. In particular, we should
  not try to open a backing file if the target does not have any yet.

- If drive-mirror with mode=absolute-paths is used, we can and should
  reuse the already existing chain of nodes that the source BDS is in.
  In case of sync=full, no backing BDS is required; with sync=top, we
  just link the source's backing BDS to the target, and with sync=none,
  we use the source BDS as the target's backing BDS.
  We should not try to open these backing files anew because this would
  lead to two BDSs existing per physical file in the backing chain, and
  we would like to avoid such concurrent access.

- If drive-mirror with mode=existing is used, we have to use the
  information provided in the physical image file which means opening
  the target's backing chain completely anew, just as it has been done
  already.
  If the target's backing chain shares images with the source, this may
  lead to multiple BDSs per physical image file. But since we cannot
  reliably ascertain this case, there is nothing we can do about it.

Signed-off-by: Max Reitz <[email protected]>
Message-id: 20160610185750[email protected]
Reviewed-by: Kevin Wolf <[email protected]>
Reviewed-by: Fam Zheng <[email protected]>
Signed-off-by: Max Reitz <[email protected]>

block: Allow replacement of a BDS by its overlay

change_parent_backing_link() asserts that the BDS to be replaced is not
used as a backing file. However, we may want to replace a BDS by its
overlay in which case that very link should not be redirected.

For instance, when doing a sync=none drive-mirror operation, we may have
the following BDS/BB forest before block job completion:

  target

  base <- source <- BlockBackend

During job completion, we want to establish the source BDS as the
target's backing node:

          target
            |
            v
  base <- source <- BlockBackend

This makes the target a valid replacement for the source:

          target <- BlockBackend
            |
            v
  base <- source

Without this modification to change_parent_backing_link() we have to
inject the target into the graph before the source is its backing node,
thus temporarily creating a wrong graph:

  target <- BlockBackend

  base <- source

Signed-off-by: Max Reitz <[email protected]>
Message-id: 20160610185750[email protected]
Reviewed-by: Kevin Wolf <[email protected]>
Reviewed-by: Fam Zheng <[email protected]>
Signed-off-by: Max Reitz <[email protected]>

rbd:change error_setg() to error_setg_errno()

Ceph RBD block driver does not use error_setg_errno() where
it is possible to use. This patch replaces error_setg()
from error_setg_errno().

Signed-off-by: Vikhyat Umrao <[email protected]>
Message-id: 1462780319[email protected]
Reviewed-by: Josh Durgin <[email protected]>
Signed-off-by: Max Reitz <[email protected]>

iotests: 095: Clean up QEMU before showing image info

Signed-off-by: Fam Zheng <[email protected]>
Message-id: 1464944872[email protected]
Signed-off-by: Max Reitz <[email protected]>

block: Create the commit block job before reopening any image

If the base or overlay images need to be reopened in read-write mode
but the block_job_create() call fails then no one will put those
images back in read-only mode.

We can solve this problem easily by calling block_job_create() first.

Signed-off-by: Alberto Garcia <[email protected]>
Message-id: aa495045770a6f1a7cc5d408397a17c75097fdd8.1464346103[email protected]
Reviewed-by: Max Reitz <[email protected]>
Signed-off-by: Max Reitz <[email protected]>

block: Prevent sleeping jobs from resuming if they have been paused

If we pause a block job and drain its BlockDriverState we want that
the job remains inactive until we call block_job_resume() again.

However if we pause the job while it is sleeping then it will resume
when the sleep timer fires.

This patch prevents that from happening by checking if the job has
been paused after it comes back from sleeping.

Signed-off-by: Alberto Garcia <[email protected]>
Suggested-by: Kevin Wolf <[email protected]>
Message-id: 3d9011151512326b890d22bdab3530244ef349d7.1464346103[email protected]
Reviewed-by: Max Reitz <[email protected]>
Signed-off-by: Max Reitz <[email protected]>

block: use the block job list in qmp_query_block_jobs()

qmp_query_block_jobs() uses bdrv_next() to look for block jobs, but
this function can only find those in top-level BlockDriverStates.

This patch uses block_job_next() instead.

Signed-off-by: Alberto Garcia <[email protected]>
Message-id: a8b7e5497b7c1fa67c12fcceae1630d01c3b1f96.1464346103[email protected]
Reviewed-by: Max Reitz <[email protected]>
Signed-off-by: Max Reitz <[email protected]>

block: use the block job list in bdrv_drain_all()

bdrv_drain_all() pauses all block jobs by using bdrv_next() to iterate
over all top-level BlockDriverStates. Therefore the code is unable to
find block jobs in other nodes.

This patch uses block_job_next() to iterate over all block jobs.

Signed-off-by: Alberto Garcia <[email protected]>
Message-id: 55ee7d7d4a65c28aa1a1b28823897ef326f328e2.1464346103[email protected]
Reviewed-by: Max Reitz <[email protected]>
Signed-off-by: Max Reitz <[email protected]>

block: Fix snapshot=on with aio=native

snapshot=on creates a temporary overlay that is always opened with
cache=unsafe (the cache mode specified by the user is only for the
actual image file and its children). This means that we must not inherit
the BDRV_O_NATIVE_AIO flag for the temporary overlay because trying to
use Linux AIO with cache=unsafe results in an error.

Reproducer without this patch:

$ x86_64-softmmu/qemu-system-x86_64 -drive file=/tmp/test.qcow2,cache=none,aio=native,snapshot=on
qemu-system-x86_64: -drive file=/tmp/test.qcow2,cache=none,aio=native,snapshot=on: aio=native was
specified, but it requires cache.direct=on, which was not specified.

Signed-off-by: Kevin Wolf <[email protected]>

block: Remove bs->zero_beyond_eof

It is always true for open images now.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Fam Zheng <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>

qcow2: Let vmstate call qcow2_co_preadv/pwrite directly

We don't really want to go through the block layer in order to read from
or write to the vmstate in a qcow2 image. Doing so required a few ugly
hacks like saving and restoring the old image size (because writing to
vmstate offsets would increase the image size) or disabling the "reads
after EOF = zeroes" logic. When calling the right functions directly,
these hacks aren't necessary any more.

Note that .bdrv_vmstate_load/save() return 0 instead of the number of
bytes in case of success now.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Fam Zheng <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>

block: Make bdrv_load/save_vmstate coroutine_fns

This allows drivers to share code between normal I/O and vmstate
accesses.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Fam Zheng <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>

block: Allow .bdrv_load/save_vmstate() to return 0/-errno

The return value of .bdrv_load/save_vmstate() can be any non-negative
number in case of success now. It used to be bytes/-errno.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Fam Zheng <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>

block: Make .bdrv_load_vmstate() vectored

This brings it in line with .bdrv_save_vmstate().

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Fam Zheng <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>

block: Introduce bdrv_preadv()

We already have a byte-based bdrv_pwritev(), but the read counterpart
was still missing. This commit adds it.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Fam Zheng <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>

doc: Fix mailing list address in tests/qemu-iotests/README

The address of the mailing list is [email protected]
instead of [email protected]. And while we're
at it, also mention the qemu-block mailing list here.

Signed-off-by: Thomas Huth <[email protected]>
Signed-off-by: Kevin Wolf <[email protected]>

linux-aio: Cancel BH if not needed

linux-aio uses a BH in order to make sure that the remaining completions
are processed even in nested event loops of completion callbacks in
order to avoid deadlocks.

There is no need, however, to have the BH overhead for the first call
into qemu_laio_completion_bh() or after all pending completions have
already been processed. Therefore, this patch calls directly into
qemu_laio_completion_bh() in qemu_laio_completion_cb() and cancels
the BH after qemu_laio_completion_bh() has processed all pending
completions.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>

block: Don't enforce 512 byte minimum alignment

If block drivers say that they can do an alignment < 512 bytes, let's
just suppose they mean it. raw-posix used to be an offender with respect
to this, but it can actually deal with byte-aligned requests now.

The default is still 512 bytes for any drivers that only implement
sector-based interfaces, but it is 1 now for drivers that implement
.bdrv_co_preadv.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>

raw-posix: Implement .bdrv_co_preadv/pwritev

The raw-posix block driver actually supports byte-aligned requests now
on non-O_DIRECT images, like it already (and previously incorrectly)
claimed in bs->request_alignment.

For some block drivers this means that a RMW cycle can be avoided when
they write sub-sector metadata e.g. for cluster allocation.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>

raw-posix: Switch to bdrv_co_* interfaces

In order to use the modern byte-based .bdrv_co_preadv/pwritev()
interface, this patch switches raw-posix to coroutine-based interfaces
as a first step. In terms of semantics and performance, it doesn't make
a difference with the existing code whether we go from a coroutine to a
callback-based interface already in block/io.c or only in linux-aio.c

As there have been concerns in the past that this change may be a step
in the wrong direction with respect to a possible AIO fast path, the
old callback-based interface for linux-aio is left around and can be
reactivated when a fast path (e.g. directly from virtio-blk dataplane,
bypassing the whole block layer) is implemented.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>

block: Prepare bdrv_aligned_pwritev() for byte-aligned requests

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>

block: Prepare bdrv_aligned_preadv() for byte-aligned requests

This patch makes bdrv_aligned_preadv() ready to accept byte-aligned
requests. Note that this doesn't mean that such requests are actually
made. The caller still ensures that all requests are aligned to at least
512 bytes.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>

block: Byte-based bdrv_co_do_copy_on_readv()

In a first step to convert the common I/O path to work on bytes rather
than sectors, this converts the copy-on-read logic that is used by
bdrv_aligned_preadv().

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>

block: drop support for using qcow[2] encryption with system emulators

Back in the 2.3.0 release we declared qcow[2] encryption as
deprecated, warning people that it would be removed in a future
release.

  commit a1f688f4152e65260b94f37543521ceff8bfebe4
  Author: Markus Armbruster <[email protected]>
  Date:   Fri Mar 13 21:09:40 2015 +0100

    block: Deprecate QCOW/QCOW2 encryption

The code still exists today, but by a (happy?) accident we entirely
broke the ability to use qcow[2] encryption in the system emulators
in the 2.4.0 release due to

  commit 8336aafae1451d54c81dd2b187b45f7c45d2428e
  Author: Daniel P. Berrange <[email protected]>
  Date:   Tue May 12 17:09:18 2015 +0100

    qcow2/qcow: protect against uninitialized encryption key

This commit was designed to prevent future coding bugs which
might cause QEMU to read/write data on an encrypted block
device in plain text mode before a decryption key is set.

It turns out this preventative measure was a little too good,
because we already had a long standing bug where QEMU read
encrypted data in plain text mode during system emulator
startup, in order to guess disk geometry:

  Thread 10 (Thread 0x7fffd3fff700 (LWP 30373)):
  #0  0x00007fffe90b1a28 in raise () at /lib64/libc.so.6
  #1  0x00007fffe90b362a in abort () at /lib64/libc.so.6
  #2  0x00007fffe90aa227 in __assert_fail_base () at /lib64/libc.so.6
  #3  0x00007fffe90aa2d2 in  () at /lib64/libc.so.6
  #4  0x000055555587ae19 in qcow2_co_readv (bs=0x5555562accb0, sector_num=0, remaining_sectors=1, qiov=0x7fffffffd260) at block/qcow2.c:1229
  #5  0x000055555589b60d in bdrv_aligned_preadv (bs=bs@entry=0x5555562accb0, req=req@entry=0x7fffd3ffea50, offset=offset@entry=0, bytes=bytes@entry=512, align=align@entry=512, qiov=qiov@entry=0x7fffffffd260, flags=0) at block/io.c:908
  #6  0x000055555589b8bc in bdrv_co_do_preadv (bs=0x5555562accb0, offset=0, bytes=512, qiov=0x7fffffffd260, flags=<optimized out>) at block/io.c:999
  #7  0x000055555589c375 in bdrv_rw_co_entry (opaque=0x7fffffffd210) at block/io.c:544
  #8  0x000055555586933b in coroutine_thread (opaque=0x555557876310) at coroutine-gthread.c:134
  #9  0x00007ffff64e1835 in g_thread_proxy (data=0x5555562b5590) at gthread.c:778
  #10 0x00007ffff6bb760a in start_thread () at /lib64/libpthread.so.0
  #11 0x00007fffe917f59d in clone () at /lib64/libc.so.6

  Thread 1 (Thread 0x7ffff7ecab40 (LWP 30343)):
  #0  0x00007fffe91797a9 in syscall () at /lib64/libc.so.6
  #1  0x00007ffff64ff87f in g_cond_wait (cond=cond@entry=0x555555e085f0 <coroutine_cond>, mutex=mutex@entry=0x555555e08600 <coroutine_lock>) at gthread-posix.c:1397
  #2  0x00005555558692c3 in qemu_coroutine_switch (co=<optimized out>) at coroutine-gthread.c:117
  #3  0x00005555558692c3 in qemu_coroutine_switch (from_=0x5555562b5e30, to_=to_@entry=0x555557876310, action=action@entry=COROUTINE_ENTER) at coroutine-gthread.c:175
  #4  0x0000555555868a90 in qemu_coroutine_enter (co=0x555557876310, opaque=0x0) at qemu-coroutine.c:116
  #5  0x0000555555859b84 in thread_pool_completion_bh (opaque=0x7fffd40010e0) at thread-pool.c:187
  #6  0x0000555555859514 in aio_bh_poll (ctx=ctx@entry=0x5555562953b0) at async.c:85
  #7  0x0000555555864d10 in aio_dispatch (ctx=ctx@entry=0x5555562953b0) at aio-posix.c:135
  #8  0x0000555555864f75 in aio_poll (ctx=ctx@entry=0x5555562953b0, blocking=blocking@entry=true) at aio-posix.c:291
  #9  0x000055555589c40d in bdrv_prwv_co (bs=bs@entry=0x5555562accb0, offset=offset@entry=0, qiov=qiov@entry=0x7fffffffd260, is_write=is_write@entry=false, flags=flags@entry=(unknown: 0)) at block/io.c:591
  #10 0x000055555589c503 in bdrv_rw_co (bs=bs@entry=0x5555562accb0, sector_num=sector_num@entry=0, buf=buf@entry=0x7fffffffd2e0 "\321,", nb_sectors=nb_sectors@entry=21845, is_write=is_write@entry=false, flags=flags@entry=(unknown: 0)) at block/io.c:614
  #11 0x000055555589c562 in bdrv_read_unthrottled (nb_sectors=21845, buf=0x7fffffffd2e0 "\321,", sector_num=0, bs=0x5555562accb0) at block/io.c:622
  #12 0x000055555589c562 in bdrv_read_unthrottled (bs=0x5555562accb0, sector_num=sector_num@entry=0, buf=buf@entry=0x7fffffffd2e0 "\321,", nb_sectors=nb_sectors@entry=21845) at block/io.c:634
    nb_sectors@entry=1) at block/block-backend.c:504
  #14 0x0000555555752e9f in guess_disk_lchs (blk=blk@entry=0x5555562a5290, pcylinders=pcylinders@entry=0x7fffffffd52c, pheads=pheads@entry=0x7fffffffd530, psectors=psectors@entry=0x7fffffffd534) at hw/block/hd-geometry.c:68
  #15 0x0000555555752ff7 in hd_geometry_guess (blk=0x5555562a5290, pcyls=pcyls@entry=0x555557875d1c, pheads=pheads@entry=0x555557875d20, psecs=psecs@entry=0x555557875d24, ptrans=ptrans@entry=0x555557875d28) at hw/block/hd-geometry.c:133
  #16 0x0000555555752b87 in blkconf_geometry (conf=conf@entry=0x555557875d00, ptrans=ptrans@entry=0x555557875d28, cyls_max=cyls_max@entry=65536, heads_max=heads_max@entry=16, secs_max=secs_max@entry=255, errp=errp@entry=0x7fffffffd5e0) at hw/block/block.c:71
  #17 0x0000555555799bc4 in ide_dev_initfn (dev=0x555557875c80, kind=IDE_HD) at hw/ide/qdev.c:174
  #18 0x0000555555768394 in device_realize (dev=0x555557875c80, errp=0x7fffffffd640) at hw/core/qdev.c:247
  #19 0x0000555555769a81 in device_set_realized (obj=0x555557875c80, value=<optimized out>, errp=0x7fffffffd730) at hw/core/qdev.c:1058
  #20 0x00005555558240ce in property_set_bool (obj=0x555557875c80, v=<optimized out>, opaque=0x555557875de0, name=<optimized out>, errp=0x7fffffffd730)
        at qom/object.c:1514
  #21 0x0000555555826c87 in object_property_set_qobject (obj=obj@entry=0x555557875c80, value=value@entry=0x55555784bcb0, name=name@entry=0x55555591cb3d "realized", errp=errp@entry=0x7fffffffd730) at qom/qom-qobject.c:24
  #22 0x0000555555825760 in object_property_set_bool (obj=obj@entry=0x555557875c80, value=value@entry=true, name=name@entry=0x55555591cb3d "realized", errp=errp@entry=0x7fffffffd730) at qom/object.c:905
  #23 0x000055555576897b in qdev_init_nofail (dev=dev@entry=0x555557875c80) at hw/core/qdev.c:380
  #24 0x0000555555799ead in ide_create_drive (bus=bus@entry=0x555557629630, unit=unit@entry=0, drive=0x5555562b77e0) at hw/ide/qdev.c:122
  #25 0x000055555579a746 in pci_ide_create_devs (dev=dev@entry=0x555557628db0, hd_table=hd_table@entry=0x7fffffffd830) at hw/ide/pci.c:440
  #26 0x000055555579b165 in pci_piix3_ide_init (bus=<optimized out>, hd_table=0x7fffffffd830, devfn=<optimized out>) at hw/ide/piix.c:218
  #27 0x000055555568ca55 in pc_init1 (machine=0x5555562960a0, pci_enabled=1, kvmclock_enabled=<optimized out>) at /home/berrange/src/virt/qemu/hw/i386/pc_piix.c:256
  #28 0x0000555555603ab2 in main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) at vl.c:4249

So the safety net is correctly preventing QEMU reading cipher
text as if it were plain text, during startup and aborting QEMU
to avoid bad usage of this data.

For added fun this bug only happens if the encrypted qcow2
file happens to have data written to the first cluster,
otherwise the cluster won't be allocated and so qcow2 would
not try the decryption routines at all, just return all 0's.

That no one even noticed, let alone reported, this bug that
has shipped in 2.4.0, 2.5.0 and 2.6.0 shows that the number
of actual users of encrypted qcow2 is approximately zero.

So rather than fix the crash, and backport it to stable
releases, just go ahead with what we have warned users about
and disable any use of qcow2 encryption in the system
emulators. qemu-img/qemu-io/qemu-nbd are still able to access
qcow2 encrypted images for the sake of data conversion.

In the future, qcow2 will gain support for the alternative
luks format, but when this happens it'll be using the
'-object secret' infrastructure for getting keys, which
avoids this problematic scenario entirely.

Signed-off-by: Daniel P. Berrange <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Signed-off-by: Kevin Wolf <[email protected]>

block: Assert that flags are in range

Add a new BDRV_REQ_MASK constant, and use it to make sure that
caller flags are always valid.

Tested with 'make check' and with qemu-iotests on both '-raw'
and '-qcow2'; the only failure turned up was fixed in the
previous commit.

Signed-off-by: Eric Blake <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>
Signed-off-by: Kevin Wolf <[email protected]>

block: Avoid bogus flags during mirroring

Commit e253f4b8 converted mirroring from sector-based bdrv_aio_*
to byte-based blk_aio_*, but failed to account for the subtle
difference in signatures (the former takes a semi-redundant length,
the latter takes a flags parameter).  Since all of our flags are
currently smaller in size than BDRV_SECTOR_SIZE, it has no ill
effects until we either perform sub-sector mirroring, or we start
asserting that no unexpected flags are set.  I found it while
testing new asserts when qemu-iotests 132 started warning about an
unknown flag 0x200000.

Signed-off-by: Eric Blake <[email protected]>
Signed-off-by: Kevin Wolf <[email protected]>

qemu-img bench: Fix uninitialised writethrough mode

If no -t option is specified, bool writethrough stayed uninitialised.
Initialise it as false, which makes cache=writeback the default cache
mode.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Max Reitz <[email protected]>
Reviewed-by: Stefan Hajnoczi <[email protected]>

m25p80: fix test on blk_pread() return value

commit 243e6f69c129 ("m25p80: Switch to byte-based block access")
replaced blk_read() calls with blk_pread() but return values are
different.

Signed-off-by: Cédric Le Goater <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Signed-off-by: Kevin Wolf <[email protected]>

hmp: acquire aio_context in hmp_qemu_io

Acquire aio context before run command, this is mandatory for unit tests.

Signed-off-by: Vladimir Sementsov-Ogievskiy <[email protected]>
Signed-off-by: Denis V. Lunev <[email protected]>
CC: Kevin Wolf <[email protected]>
CC: Paolo Bonzini <[email protected]>
Signed-off-by: Kevin Wolf <[email protected]>

blockdev: clarify error on attempt to open locked tray

When opening a device with a locked tray, gives an error explaining the
device tray is locked and that the user should wait and try again. This
is less confusing than the previous error, which simply stated that the
tray was locked.

Signed-off-by: Colin Lord <[email protected]>
Signed-off-by: Kevin Wolf <[email protected]>

qcow2: Implement .bdrv_co_pwritev()

This changes qcow2 to implement the byte-based .bdrv_co_pwritev
interface rather than the sector-based old one.

As preallocation uses the same allocation function as normal writes, and
the interface of that function needs to be changed, it is converted in
the same patch.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Eric Blake <[email protected]>

qcow2: Use bytes instead of sectors for QCowL2Meta

In preparation for implementing .bdrv_co_pwritev in qcow2.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Eric Blake <[email protected]>

qcow2: Make copy_sectors() byte based

This will allow copy on write operations where the overwritten part of
the cluster is not aligned to sector boundaries.

Also rename the function because it has nothing to do with sectors any
more.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Eric Blake <[email protected]>

qcow2: Implement .bdrv_co_preadv()

Reading from qcow2 images is now byte granularity.

Most of the affected code in qcow2 actually gets simpler with this
change. The only exception is encryption, which is fixed on 512 bytes
blocks; in order to keep this working, bs->request_alignment is set for
encrypted images.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Eric Blake <[email protected]>

qcow2: Work with bytes in qcow2_get_cluster_offset()

This patch changes the units that qcow2_get_cluster_offset() uses
internally, without touching the interface just yet. This will be done
in another patch.

Signed-off-by: Kevin Wolf <[email protected]>
Reviewed-by: Eric Blake <[email protected]>

Merge remote-tracking branch 'remotes/amit-migration/tags/migration-for-2.7-4' into staging

Migration:

- Fixes for TLS series
- Postcopy: Add stats, fix, test case

# gpg: Signature made Thu 16 Jun 2016 05:40:09 BST
# gpg:                using RSA key 0xEB0B4DFC657EF670
# gpg: Good signature from "Amit Shah <[email protected]>"
# gpg:                 aka "Amit Shah <[email protected]>"
# gpg:                 aka "Amit Shah <[email protected]>"
# Primary key fingerprint: 48CA 3722 5FE7 F4A8 B337  2735 1E9A 3B5F 8540 83B6
#      Subkey fingerprint: CC63 D332 AB8F 4617 4529  6534 EB0B 4DFC 657E F670

* remotes/amit-migration/tags/migration-for-2.7-4:
  migration: rename functions to starting migrations
  migration: fix typos in qapi-schema from latest migration additions
  Postcopy: Check for support when setting the capability
  tests: fix libqtest socket timeouts
  test: Postcopy
  Postcopy: Add stats on page requests
  Migration: Split out ram part of qmp_query_migrate
  Postcopy: Avoid 0 length discards

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

migration: rename functions to starting migrations

Apply the following renames for starting incoming migration:

 process_incoming_migration -> migration_fd_process_incoming
 migration_set_incoming_channel -> migration_channel_process_incoming
 migration_tls_set_incoming_channel -> migration_tls_channel_process_incoming

and for starting outgoing migration:

 migration_set_outgoing_channel -> migration_channel_connect
 migration_tls_set_outgoing_channel -> migration_tls_channel_connect

Signed-off-by: Daniel P. Berrange <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Message-id: 1464776234[email protected]
Message-Id: <1464776234[email protected]>
Signed-off-by: Amit Shah <[email protected]>

migration: fix typos in qapi-schema from latest migration additions

Recent migration QAPI enhancements had a few spelling mistakes
and also incorrect version number in a few places.

Signed-off-by: Daniel P. Berrange <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Message-id: 1464776234[email protected]
Message-Id: <1464776234[email protected]>
Signed-off-by: Amit Shah <[email protected]>

Postcopy: Check for support when setting the capability

Knowing whether the destination host supports migration with
postcopy can be tricky.
The destination doesn't need the capability set, however
if we set it then use the opportunity to do the test and
tell the user/management layer early.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Message-id: 1465816605[email protected]
Message-Id: <1465816605[email protected]>
Signed-off-by: Amit Shah <[email protected]>

tests: fix libqtest socket timeouts

I kept getting timeouts and unix socket accept failures under high
load, the patch fixes it.

Signed-off-by: Andrea Arcangeli <[email protected]>
Reviewed-by: Marcel Apfelbaum <[email protected]>
Message-id: 1465816605[email protected]
Message-Id: <1465816605[email protected]>
Signed-off-by: Amit Shah <[email protected]>

test: Postcopy

This is a postcopy test (x86 only) that actually runs the guest
and checks the memory contents.

The test runs from an x86 boot block with the hex embedded in the test;
the source for this is:

...........

.code16
.org 0x7c00
.file "fill.s"
.text
.globl start
.type start, @function
start:             # at 0x7c00 ?
        cli
        lgdt gdtdesc
        mov $1,%eax
        mov %eax,%cr0  # Protected mode enable
        data32 ljmp $8,$0x7c20

.org 0x7c20
.code32
        # A20 enable - not sure I actually need this
        inb $0x92,%al
        or  $2,%al
        outb %al, $0x92

        # set up DS for the whole of RAM (needed on KVM)
        mov $16,%eax
        mov %eax,%ds

        mov $65,%ax
        mov $0x3f8,%dx
        outb %al,%dx

        # bl keeps a counter so we limit the output speed
        mov $0, %bl
mainloop:
        # Start from 1MB
        mov $(1024*1024),%eax
innerloop:
        incb (%eax)
        add $4096,%eax
        cmp $(100*1024*1024),%eax
        jl innerloop

        inc %bl
        jnz mainloop

        mov $66,%ax
        mov $0x3f8,%dx
        outb %al,%dx

jmp mainloop

        # GDT magic from old (GPLv2)  Grub startup.S
        .p2align        2       /* force 4-byte alignment */
gdt:
        .word   0, 0
        .byte   0, 0, 0, 0

        /* -- code segment --
         * base = 0x00000000, limit = 0xFFFFF (4 KiB Granularity), present
         * type = 32bit code execute/read, DPL = 0
         */
        .word   0xFFFF, 0
        .byte   0, 0x9A, 0xCF, 0

        /* -- data segment --
         * base = 0x00000000, limit 0xFFFFF (4 KiB Granularity), present
         * type = 32 bit data read/write, DPL = 0
         */
        .word   0xFFFF, 0
        .byte   0, 0x92, 0xCF, 0

gdtdesc:
        .word   0x27                    /* limit */
        .long   gdt                     /* addr */

/* I'm a bootable disk */
.org 0x7dfe
        .byte 0x55
        .byte 0xAA

...........

and that can be assembled by the following magic:
    as --32 -march=i486 fill.s -o fill.o
    objcopy -O binary fill.o fill.boot
    dd if=fill.boot of=bootsect bs=256 count=2 skip=124
    xxd -i bootsect

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Marcel Apfelbaum <[email protected]>
Message-id: 1465816605[email protected]
Message-Id: <1465816605[email protected]>
Signed-off-by: Amit Shah <[email protected]>

Postcopy: Add stats on page requests

On the source, add a count of page requests received from the
destination.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Reviewed-by: Denis V. Lunev <[email protected]>
Message-id: 1465816605[email protected]
Message-Id: <1465816605[email protected]>
Signed-off-by: Amit Shah <[email protected]>

Migration: Split out ram part of qmp_query_migrate

The RAM section of qmp_query_migrate is reasonably complex
and repeated 3 times.  Split it out into a helper.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Eric Blake <[email protected]>
Message-id: 1465816605[email protected]
Reviwed-by: Denis V. Lunev <[email protected]>
Message-Id: <1465816605[email protected]>
Signed-off-by: Amit Shah <[email protected]>

Postcopy: Avoid 0 length discards

The discard code in migration/ram.c would send request for
zero length discards in the case where no discards were needed.
It doesn't appear to have had any bad effect.

Signed-off-by: Dr. David Alan Gilbert <[email protected]>
Reviewed-by: Denis V. Lunev <[email protected]>
Message-id: 1465816605[email protected]
Message-Id: <1465816605[email protected]>
Signed-off-by: Amit Shah <[email protected]>

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

X86 queue, 2016-06-14 (v2)

# gpg: Signature made Tue 14 Jun 2016 22:29:04 BST
# gpg:                using RSA key 0x2807936F984DC5A6
# gpg: Good signature from "Eduardo Habkost <[email protected]>"
# Primary key fingerprint: 5A32 2FD5 ABC4 D3DB ACCF  D1AA 2807 936F 984D C5A6

* remotes/ehabkost/tags/x86-pull-request:
  target-i386: Consolidate calls of object_property_parse() in x86_cpu_parse_featurestr
  target-i386: Use cpu_generic_init() in cpu_x86_init()
  target-i386: Move xcc->kvm_required check to realize time
  target-i386: Remove assert(kvm_enabled()) from host_x86_cpu_initfn()
  target-i386: Move features logic that requires CPUState to realize time
  target-i386: Remove xlevel & hv-spinlocks option fixups
  target-i386: Implement CPUID[0xB] (Extended Topology Enumeration)
  pc: Add 2.7 machine
  target-i386: add Skylake-Client cpu model

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

target-i386: Consolidate calls of object_property_parse() in x86_cpu_parse_featurestr

Signed-off-by: Eduardo Habkost <[email protected]>
Reviewed-by: Igor Mammedov <[email protected]>
Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eduardo Habkost <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

target-i386: Use cpu_generic_init() in cpu_x86_init()

Now cpu_x86_init() does nothing more or less
than duplicating cpu_generic_init() logic.
So simplify it by using cpu_generic_init().

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eduardo Habkost <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

target-i386: Move xcc->kvm_required check to realize time

It will allow to drop custom cpu_x86_init() and use
cpu_generic_init() instead, reducing cpu_x86_create()
to a simple 3-liner.

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eduardo Habkost <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

target-i386: Remove assert(kvm_enabled()) from host_x86_cpu_initfn()

The code will be changed to allow creation of the CPU object and
report kvm_required errors only at realizefn, so we need to make
the instance_init function more flexible.

Reviewed-by: Igor Mammedov <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

target-i386: Move features logic that requires CPUState to realize time

Making x86_cpu_parse_featurestr() a pure convertor
of legacy feature string into global properties, needs
it to be called before a CPU instance is created so
parser shouldn't modify CPUState directly or access
it at all. Hence move current hack that directly pokes
into CPUState, to set/unset +-feats, from parser to
CPU's realize method.

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eduardo Habkost <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

target-i386: Remove xlevel & hv-spinlocks option fixups

The "fixup will be removed in future versions" warnings are
present since QEMU 1.7.0, at least, so users should have fixed
their scripts and configurations, already.

In the case of libvirt users, libvirt doesn't use the "xlevel"
option, and already rejects HyperV spinlock retry count < 0xFFF.

Reviewed-by: Eric Blake <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

target-i386: Implement CPUID[0xB] (Extended Topology Enumeration)

I looked at a dozen Intel CPU that have this CPUID and all of them
always had Core offset as 1 (a wasted bit when hyperthreading is
disabled) and Package offset at least 4 (wasted bits at <= 4 cores).

QEMU uses more compact IDs and it doesn't make much sense to change it
now.  I keep the SMT and Core sub-leaves even if there is just one
thread/core;  it makes the code simpler and there should be no harm.

Signed-off-by: Radim Krčmář <[email protected]>
Reviewed-by: Eduardo Habkost <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

pc: Add 2.7 machine

Signed-off-by: Igor Mammedov <[email protected]>
Reviewed-by: Eduardo Habkost <[email protected]>
Reviewed-by: Michael S. Tsirkin <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

target-i386: add Skylake-Client cpu model

Introduce Skylake-Client cpu mode which inherits the features from
Broadwell and supports some additional features that are: MPX,
XSAVEC, and XGETBV1.

Signed-off-by: Eduardo Habkost <[email protected]>
Signed-off-by: Xiao Guangrong <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Reviewed-by: Xiao Guangrong <[email protected]>
Signed-off-by: Eduardo Habkost <[email protected]>

Merge remote-tracking branch 'remotes/sstabellini/tags/xen-20160614-tag' into staging

Xen 2016/06/14

# gpg: Signature made Tue 14 Jun 2016 16:01:52 BST
# gpg:                using RSA key 0x894F8F4870E1AE90
# gpg: Good signature from "Stefano Stabellini <[email protected]>"
# Primary key fingerprint: D04E 33AB A51F 67BA 07D3  0AEA 894F 8F48 70E1 AE90

* remotes/sstabellini/tags/xen-20160614-tag:
  xen: Clean up includes
  xen/blkif: avoid double access to any shared ring request fields

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

Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20160614-2' into staging

target-arm queue:
 * add PMU support for virt machine under KVM
 * fix reset and migration of TTBCR(S)
 * add virt-2.7 machine type
 * QOMify various ARM devices
 * implement xilinx DisplayPort device
 * don't permit ARMv8-only Neon insns to work on ARMv7

# gpg: Signature made Tue 14 Jun 2016 16:01:45 BST
# gpg:                using RSA key 0x3C2525ED14360CDE
# gpg: Good signature from "Peter Maydell <[email protected]>"
# gpg:                 aka "Peter Maydell <[email protected]>"
# gpg:                 aka "Peter Maydell <[email protected]>"
# Primary key fingerprint: E1A5 C593 CD41 9DE2 8E83  15CF 3C25 25ED 1436 0CDE

* remotes/pmaydell/tags/pull-target-arm-20160614-2: (30 commits)
  target-arm: Don't permit ARMv8-only Neon insns on ARMv7
  arm: xlnx-zynqmp: Add xlnx-dp and xlnx-dpdma
  introduce xlnx-dp
  introduce xlnx-dpdma
  hw/i2c-ddc.c: Implement DDC I2C slave
  introduce dpcd module
  introduce aux-bus
  i2c: Factor our send() and recv() common logic
  i2c: implement broadcast write
  i2cbus: remove unused dev field
  hw/sd: QOM'ify pl181.c
  hw/dma: QOM'ify pxa2xx_dma.c
  hw/misc: QOM'ify mst_fpga.c
  hw/misc: QOM'ify exynos4210_pmu.c
  hw/misc: QOM'ify arm_l2x0.c
  hw/gpio: QOM'ify zaurus.c
  hw/gpio: QOM'ify pl061.c
  hw/gpio: QOM'ify omap_gpio.c
  hw/i2c: QOM'ify versatile_i2c.c
  hw/i2c: QOM'ify omap_i2c.c
  ...

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

target-arm: Don't permit ARMv8-only Neon insns on ARMv7

The Neon instructions VCVTA, VCVTM, VCVTN, VCVTP, VRINTA, VRINTM,
VRINTN, VRINTP, VRINTX, and VRINTZ were only introduced with ARMv8,
so they need a guard to make them UNDEF if the CPU only supports ARMv7.
(We got this right for all the other new-in-v8 insns, but forgot
it for these Neon 2-reg-misc ops.)

Reported-by: Christophe Lyon <[email protected]>
Tested-by: Christophe Lyon <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>
Message-id: 1465492511[email protected]

arm: xlnx-zynqmp: Add xlnx-dp and xlnx-dpdma

This adds the DP and the DPDMA to the Zynq MP platform.

Signed-off-by: KONRAD Frederic <[email protected]>
Reviewed-by: Peter Crosthwaite <[email protected]>
Reviewed-by: Alistair Francis <[email protected]>
Tested-By: Hyun Kwon <[email protected]>
Message-id: 1465833014[email protected]
Signed-off-by: Peter Maydell <[email protected]>

introduce xlnx-dp

This is the implementation of the DisplayPort.
It has an aux-bus to access dpcd and edid.

Graphic plane is connected to the channel 3.
Video plane is connected to the channel 0.
Audio stream are connected to the channels 4 and 5.

Signed-off-by: KONRAD Frederic <[email protected]>
Tested-By: Hyun Kwon <[email protected]>
Reviewed-by: Alistair Francis <[email protected]>
Message-id: 1465833014[email protected]
[PMM: fixed format strings]
Signed-off-by: Peter Maydell <[email protected]>

introduce xlnx-dpdma

This is the implementation of the DPDMA.

Signed-off-by: KONRAD Frederic <[email protected]>
Reviewed-by: Alistair Francis <[email protected]>
Tested-By: Hyun Kwon <[email protected]>
Message-id: 1465833014[email protected]
Signed-off-by: Peter Maydell <[email protected]>

hw/i2c-ddc.c: Implement DDC I2C slave

Implement an I2C slave which implements DDC and returns the
EDID data for an attached monitor.

Signed-off-by: Peter Maydell <[email protected]>
Reviewed-by: Alistair Francis <[email protected]>
Tested-by: Hyun Kwon <[email protected]>
Signed-off-by: KONRAD Frederic <[email protected]>
Message-id: 1465833014[email protected]

  - Rebased on the current master.
  - Modified for QOM.

Signed-off-by: KONRAD Frederic <[email protected]>
Reviewed-by: Alistair Francis <[email protected]>
Tested-By: Hyun Kwon <[email protected]>
[PMM: actually wire up the vmstate to dc->vmsd]
Signed-off-by: Peter Maydell <[email protected]>

introduce dpcd module

This introduces dpcd module.
It wires on a aux-bus and can be accessed by the driver to get lane-speed, etc.

Signed-off-by: KONRAD Frederic <[email protected]>
Reviewed-by: Alistair Francis <[email protected]>
Reviewed-by: Peter Crosthwaite <[email protected]>
Tested-By: Hyun Kwon <[email protected]>
Message-id: 1465833014[email protected]
Signed-off-by: Peter Maydell <[email protected]>

introduce aux-bus

This introduces a new bus: aux-bus.

It contains an address space for aux slaves devices and a bridge to an I2C bus
for I2C through AUX transactions.

Signed-off-by: KONRAD Frederic <[email protected]>
Tested-By: Hyun Kwon <[email protected]>
Reviewed-by: Alistair Francis <[email protected]>
Message-id: 1465833014[email protected]
Signed-off-by: Peter Maydell <[email protected]>

i2c: Factor our send() and recv() common logic

Most of the control flow logic between send and recv (error checking
etc) is the same. Factor this out into a common send_recv() API.
This is then usable by clients, where the control logic for send
and receive differs only by a boolean. E.g.

if (send)
   i2c_send(...):
else
   i2c_recv(...);

becomes:

i2c_send_recv(... , send);

Signed-off-by: Peter Crosthwaite <[email protected]>
Reviewed-by: Alistair Francis <[email protected]>
Signed-off-by: KONRAD Frederic <[email protected]>
Message-id: 1465833014[email protected]
Changes from FK:
  * Rebased on master.
  * Rebased on my i2c broadcast patch.
Signed-off-by: KONRAD Frederic <[email protected]>
Reviewed-by: Alistair Francis <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

i2c: implement broadcast write

This does a write to every slaves when the I2C bus get a write to address 0.

Signed-off-by: KONRAD Frederic <[email protected]>
Reviewed-by: Alistair Francis <[email protected]>
Reviewed-by: Peter Crosthwaite <[email protected]>
Tested-By: Hyun Kwon <[email protected]>
Message-id: 1465833014[email protected]
Signed-off-by: Peter Maydell <[email protected]>

i2cbus: remove unused dev field

The dev field in i2cbus is not used.
So just drop it.

Signed-off-by: KONRAD Frederic <[email protected]>
Reviewed-by: Alistair Francis <[email protected]>
Reviewed-by: Peter Crosthwaite <[email protected]>
Tested-By: Hyun Kwon <[email protected]>
Message-id: 1465833014[email protected]
Signed-off-by: Peter Maydell <[email protected]>

hw/sd: QOM'ify pl181.c

split the old SysBus init function into an instance_init
and a Device realize function

Signed-off-by: xiaoqiang zhao <[email protected]>
Message-id: 1465815255[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/dma: QOM'ify pxa2xx_dma.c

split the old SysBus init function into an instance_init
and a Device realize function

Signed-off-by: xiaoqiang zhao <[email protected]>
Message-id: 1465815255[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/misc: QOM'ify mst_fpga.c

Drop the old SysBus init function and use instance_init

Signed-off-by: xiaoqiang zhao <[email protected]>
Message-id: 1465815255[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/misc: QOM'ify exynos4210_pmu.c

Drop the old SysBus init function and use instance_init

Signed-off-by: xiaoqiang zhao <[email protected]>
Message-id: 1465815255[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/misc: QOM'ify arm_l2x0.c

Drop the old SysBus init function and use instance_init

Signed-off-by: xiaoqiang zhao <[email protected]>
Message-id: 1465815255[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/gpio: QOM'ify zaurus.c

Drop the old SysBus init function and use instance_init

Signed-off-by: xiaoqiang zhao <[email protected]>
Message-id: 1465815255[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/gpio: QOM'ify pl061.c

* Merge the pl061_initfn into pl061_init
* Drop the old SysBus init function and use instance_init

Signed-off-by: xiaoqiang zhao <[email protected]>
Message-id: 1465815255[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/gpio: QOM'ify omap_gpio.c

* Split the old SysBus init into an instance_init and
  DeviceClass::realize function
* Drop the SysBus init function

Signed-off-by: xiaoqiang zhao <[email protected]>
Message-id: 1465815255[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/i2c: QOM'ify versatile_i2c.c

Drop the old SysBus init function and use instance_init

Signed-off-by: xiaoqiang zhao <[email protected]>
Message-id: 1465815255[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/i2c: QOM'ify omap_i2c.c

* Split the omap_i2c_init into an instance_init and realize function
* Drop the old SysBus init function and use instance_init

Signed-off-by: xiaoqiang zhao <[email protected]>
Message-id: 1465815255[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/i2c: QOM'ify exynos4210_i2c.c

* Rename the exynos4210_i2c_realize to exynos4210_i2c_init
* Drop the old SysBus init function and use instance_init

Signed-off-by: xiaoqiang zhao <[email protected]>
Message-id: 1465815255[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/i2c: QOM'ify bitbang_i2c.c

Drop the old SysBus init function and use instance_init

Signed-off-by: xiaoqiang zhao <[email protected]>
Message-id: 1465815255[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/arm/virt: create the 2.7 machine type

Signed-off-by: Andrew Jones <[email protected]>
Message-id: 1465746713[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/arm/virt: introduce DEFINE_VIRT_MACHINE_AS_LATEST

Create two variants of DEFINE_VIRT_MACHINE. One, just called
DEFINE_VIRT_MACHINE, that does not set properties that only
the latest machine type should have, and another that does.
This will hopefully reduce potential for errors when adding
new versions.

Signed-off-by: Andrew Jones <[email protected]>
Message-id: 1465746713[email protected]
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Peter Maydell <[email protected]>

hw/arm/virt: introduce DEFINE_VIRT_MACHINE

Use DEFINE_VIRT_MACHINE to generate versioned machine type info.

Signed-off-by: Andrew Jones <[email protected]>
Reviewed-by: Peter Maydell <[email protected]>
Message-id: 1465746713[email protected]
Signed-off-by: Peter Maydell <[email protected]>

hw/arm/virt: separate versioned type-init code

Rename machvirt_info (which is specifically for 2.6 TypeInfo)
to machvirt_2_6_info, and separate the type registration of the
abstract machine type from the versioned type.

Signed-off-by: Andrew Jones <[email protected]>
Reviewed-by: Peter Maydell <[email protected]>
Message-id: 1465746713[email protected]
Signed-off-by: Peter Maydell <[email protected]>

target-arm: Fix reset and migration of TTBCR(S)

Commit 6459b94c26dd666badb3 broke reset and migration of the AArch32
TTBCR(S) register if the guest used non-LPAE page tables. This is
because the AArch32 TTBCR register definition is marked as ARM_CP_ALIAS,
meaning that the AArch64 variant has to handle migration and reset.
Although AArch64 TCR_EL3 doesn't need to care about the mask and
base_mask fields, AArch32 may do so, and so we must use the special
TTBCR reset and raw write functions to ensure they are set correctly.

This doesn't affect TCR_EL2, because the AArch32 equivalent of that
is HTCR, which never uses the non-LPAE page table variant.

Signed-off-by: Peter Maydell <[email protected]>
Reported-by: Pranith Kumar <[email protected]>
Reviewed-by: Sergey Fedorov <[email protected]>
Message-id: 1465488181[email protected]