xen_disk: fix io accounting

bdrv_acct_done was called unconditional. But in case the ioreq has no
segments there is no matching bdrv_acct_start call. This could lead to
bogus accounting values.

Found by code inspection.

Signed-off-by: Olaf Hering <olaf@aepfle.de>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>

Call pci_piix3_xen_ide_unplug from unplug_disks

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>

Merge remote-tracking branch 'remotes/kraxel/tags/pull-usb-3' into staging

- xhci improvements and fixes.
- uhci bugfix.
- cleanups.

# gpg: Signature made Tue 18 Feb 2014 15:48:10 GMT 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-usb-3:
  xhci: use DPRINTF() instead of fprintf(stderr, ...)
  xhci: switch debug printf to tracepoint
  xhci iso: allow for some latency
  xhci iso: fix time calculation
  uhci: invalidate queue on device address changes
  xhci: fix overflow in usb_xhci_post_load
  usb: Remove magic constants from device bmAttributes

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

Merge remote-tracking branch 'remotes/rth/tcg-next' into staging

* remotes/rth/tcg-next:
  tcg/i386: Use SHLX/SHRX/SARX instructions
  tcg/i386: Use ANDN instruction
  tcg/i386: Add tcg_out_vex_modrm
  tcg/i386: Move TCG_CT_CONST_* to tcg-target.c
  disas/i386: Disassemble ANDN/SHLX/SHRX/SHAX
  tcg/optimize: Add more identity simplifications
  tcg/optimize: Optmize ANDC X,Y,Y to MOV X,0
  tcg/optimize: Simply some logical ops to NOT
  tcg/optimize: Handle known-zeros masks for ANDC
  tcg/optimize: add known-zero bits compute for load ops
  tcg/optimize: improve known-zero bits for 32-bit ops
  tcg/optimize: fix known-zero bits optimization
  tcg/optimize: fix known-zero bits for right shift ops
  tcg-arm: The shift count of op_rotl_i32 is in args[2] not args[1].
  TCG: Fix 32-bit host allocation typo

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

Merge remote-tracking branch 'remotes/afaerber/tags/qom-devices-for-peter' into staging

QOM infrastructure fixes and device conversions

* QTest cleanups and test cases for PCI NICs
* NAND fix for "info qtree"
* Cleanup and extension of QOM machine tests
* IndustryPack test cases and conversion to QOM realize
* I2C cleanups
* Cleanups of legacy qdev properties

# gpg: Signature made Mon 17 Feb 2014 22:15:37 GMT using RSA key ID 3E7E013F
# gpg: Good signature from "Andreas Färber <afaerber@suse.de>"
# gpg:                 aka "Andreas Färber <afaerber@suse.com>"

* remotes/afaerber/tags/qom-devices-for-peter: (49 commits)
  qtest: Include system headers before user headers
  qapi: Refine human printing of sizes
  qdev: Use QAPI type names for properties
  qdev: Add enum property types to QAPI schema
  block: Handle "rechs" and "large" translation options
  qdev: Remove hex8/32/64 property types
  qdev: Remove most legacy printers
  qdev: Use human mode in "info qtree"
  qapi: Add human mode to StringOutputVisitor
  qdev: Inline qdev_prop_parse()
  qdev: Legacy properties are just strings
  qdev: Legacy properties are now read-only
  qdev: Remove legacy parsers for hex8/32/64
  qdev: Sizes are now parsed by StringInputVisitor
  qapi: Add size parser to StringInputVisitor
  qtest: Don't segfault with invalid -qtest option
  ipack: Move IndustryPack out of hw/char/
  ipoctal232: QOM parent field cleanup
  ipack: QOM parent field cleanup for IPackDevice
  ipack: QOM parent field cleanup for IPackBus
  ...

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

Merge remote-tracking branch 'remotes/qmp-unstable/queue/qmp' into staging

* remotes/qmp-unstable/queue/qmp:
  monitor: Add object_add class argument completion.
  monitor: Add object_del id argument completion.
  monitor: Add device_add device argument completion.
  monitor: Add device_del id argument completion.
  qmp: expose list of supported character device backends
  Use error_is_set() only when necessary
  QMP: allow JSON dict arguments in qmp-shell
  hmp: migrate command (without -d) now blocks correctly

Conflicts:
blockdev.c

[PMM: resolved trivial conflict in blockdev.c]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

xhci: use DPRINTF() instead of fprintf(stderr, ...)

So we don't spam stderr with (guest-triggerable) messages by default.

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

xhci: switch debug printf to tracepoint

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

xhci iso: allow for some latency

Allow the scheduled transfer time be a bit behind, to
compensate for latencies.  Without this xhci will wait
way to often for the mfindex wraparound, assuming the
scheduled time is in the future just because qemu is
a bit behind in processing the iso transfer requests.

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

xhci iso: fix time calculation

Frameid specifies frames not microframes, so we
need to shift it to get the microframe index.

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

uhci: invalidate queue on device address changes

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

xhci: fix overflow in usb_xhci_post_load

Found by Coverity.

Reported-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

usb: Remove magic constants from device bmAttributes

Replace magic constants in device bmAttributes with symbolic ones
from Linux kernel ch9.h

Signed-off-by: Pantelis Koukousoulas <pktoss@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

qtest: Include system headers before user headers

It is dangerous to include user headers before system headers since user
macros can affect system headers.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

monitor: Add object_add class argument completion.

Signed-off-by: Hani Benhabiles <hani@linux.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>

monitor: Add object_del id argument completion.

Signed-off-by: Hani Benhabiles <hani@linux.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>

monitor: Add device_add device argument completion.

Signed-off-by: Hani Benhabiles <hani@linux.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>

monitor: Add device_del id argument completion.

Signed-off-by: Hani Benhabiles <hani@linux.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>

qmp: expose list of supported character device backends

Introduce 'query-chardev-backends' QMP command which lists all
supported character device backends.

Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>

Use error_is_set() only when necessary

error_is_set(&var) is the same as var != NULL, but it takes
whole-program analysis to figure that out.  Unnecessarily hard for
optimizers, static checkers, and human readers.  Dumb it down to
obvious.

Gets rid of several dozen Coverity false positives.

Note that the obvious form is already used in many places.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>

QMP: allow JSON dict arguments in qmp-shell

qmp-shell hides the QMP wire protocol JSON encoding from the user.  Most
of the time this is helpful and makes the command-line human-friendly.

Some QMP commands take a dict as an argument.  In order to express this
we need to revert back to JSON notation.

This patch allows JSON dict arguments in qmp-shell so commands like
blockdev-add and nbd-server-start can be invoked:

  (QEMU) blockdev-add options={"driver":"file","id":"drive1",...}

Note that spaces are not allowed since str.split() is used to break up
the command-line arguments first.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>

hmp: migrate command (without -d) now blocks correctly

This patch fixes a timing issue that migrate command (without -d) does not
block in some cases.

The original version of hmp.c:hmp_migrate_status_cb checks if the
migration status is 'active' or not to detect the completion of a migration.

However, if this function is executed when the migration status is stil
'setup' (the status before 'active'), migration command returns
immediately even if the user does not specify -d option.

Signed-off-by: Soramichi Akiyama <akiyama@nii.ac.jp>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>

tcg/i386: Use SHLX/SHRX/SARX instructions

These three-operand shift instructions do not require the shift count
to be placed into ECX.  This reduces the number of mov insns required,
with the mere addition of a new register constraint.

Don't attempt to get rid of the matching constraint, as that's impossible
to manipulate with just a new constraint.  In addition, constant shifts
still need the matching constraint.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Richard Henderson <rth@twiddle.net>

tcg/i386: Use ANDN instruction

Note that the optimizer cannot simplify ANDC X,Y,C to AND X,Y,~C
so we must handle constants in the implementation of andc.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>

tcg/i386: Add tcg_out_vex_modrm

Prepare for emitting BMI insns which require VEX encoding.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Richard Henderson <rth@twiddle.net>

tcg/i386: Move TCG_CT_CONST_* to tcg-target.c

These are not needed by users of tcg-target.h.  No need to recompile
when we adjust them.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Richard Henderson <rth@twiddle.net>

disas/i386: Disassemble ANDN/SHLX/SHRX/SHAX

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Richard Henderson <rth@twiddle.net>

tcg/optimize: Add more identity simplifications

Recognize 0 operand to andc, and -1 operands to and, orc, eqv.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Richard Henderson <rth@twiddle.net>

tcg/optimize: Optmize ANDC X,Y,Y to MOV X,0

Like we already do for SUB and XOR.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Richard Henderson <rth@twiddle.net>

tcg/optimize: Simply some logical ops to NOT

Given, of course, an appropriate constant.  These could be generated
from the "canonical" operation for inversion on the guest, or via
other optimizations.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Richard Henderson <rth@twiddle.net>

tcg/optimize: Handle known-zeros masks for ANDC

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Richard Henderson <rth@twiddle.net>

tcg/optimize: add known-zero bits compute for load ops

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Richard Henderson <rth@twiddle.net>

tcg/optimize: improve known-zero bits for 32-bit ops

The shl_i32 op might set some bits of the unused 32 high bits of the
mask. Fix that by clearing the unused 32 high bits for all 32-bit ops
except load/store which operate on tl values.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Richard Henderson <rth@twiddle.net>

tcg/optimize: fix known-zero bits optimization

Known-zero bits optimization is a great idea that helps to generate more
optimized code. However the current implementation only works in very few
cases as the computed mask is not saved.

Fix this to make it really working.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Richard Henderson <rth@twiddle.net>

tcg/optimize: fix known-zero bits for right shift ops

32-bit versions of sar and shr ops should not propagate known-zero bits
from the unused 32 high bits. For sar it could even lead to wrong code
being generated.

Cc: qemu-stable@nongnu.org
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Richard Henderson <rth@twiddle.net>

tcg-arm: The shift count of op_rotl_i32 is in args[2] not args[1].

It's this that should be subtracted from 0x20 when converting to a right rotate.

Cc: qemu-stable@nongnu.org
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Richard Henderson <rth@twiddle.net>

Fix QEMU build on OpenBSD on x86 archs

This resolves the build issue with building the ROMs on OpenBSD on x86 archs.
As of OpenBSD 5.3 the compiler builds PIE binaries by default and thus the
whole OS/packages and so forth. The ROMs need to have PIE disabled.
Check in configure whether the compiler supports the flags for disabling
PIE, and if it does then use them for building the ROMs. This fixes the
following buildbot failure:

>From the OpenBSD buildbots..
  Building optionrom/multiboot.img
ld: multiboot.o: relocation R_X86_64_16 can not be used when making a shared object; recompile with -fPIC

Signed-off by: Brad Smith <brad@comstyle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

TCG: Fix 32-bit host allocation typo

The second half register of a 64-bit temp on a 32-bit host
was allocated with the wrong base_type.

The base_type of the second half register is never checked,
but for consistency it should be the same as the first half.

Signed-off-by: Richard Henderson <rth@twiddle.net>

libvixl: fix 64bit constants usage

Since commit 999b53ec8794f203964db3ecf939a3da5c4bc843:
 Author: Claudio Fontana <claudio.fontana@linaro.org>
 Date:   Wed Feb 5 17:27:28 2014 +0000

    disas: Implement disassembly output for A64

    Use libvixl to implement disassembly output in debug
    logs for A64, for use with both AArch64 hosts and targets.

disas/libvixl/ contains functions which uses 64bit constants
without using appropriate suffixes, which fails on 32bits.

Fix this by using ULL suffix.

Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/mjt/tags/trivial-patches-2014-02-15' into staging

trivial patches for 2014-02-15

# gpg: Signature made Sat 15 Feb 2014 12:10:46 GMT using RSA key ID 74F0C838
# gpg: Good signature from "Michael Tokarev <mjt@tls.msk.ru>"
# gpg:                 aka "Michael Tokarev <mjt@corpit.ru>"
# gpg:                 aka "Michael Tokarev <mjt@debian.org>"
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 6EE1 95D1 886E 8FFB 810D  4324 457C E0A0 8044 65C5
#      Subkey fingerprint: E190 8639 3B10 B51B AC2C  8B73 5253 C5AD 74F0 C838

* remotes/mjt/tags/trivial-patches-2014-02-15:
  char/serial: Fix emptyness check
  gitignore: anchor all ignored names
  vl: trim includes
  vl: remove old, long-unused defines
  net: declare struct iovec in checksum.h to fix compiler warning
  linux-user: refactor do_socketcall()
  configure: add hints to a remedy for feature_not_found errors
  configure: add hint of libfdt to DTC dependency not found message
  sparc/leon3: Initialize stack pointer
  misc: Fix case Qemu -> QEMU

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

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

Block pull request

# gpg: Signature made Fri 14 Feb 2014 17:26:30 GMT using RSA key ID 81AB73C8
# gpg: Good signature from "Stefan Hajnoczi <stefanha@redhat.com>"
# gpg:                 aka "Stefan Hajnoczi <stefanha@gmail.com>"
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 8695 A8BF D3F9 7CDA AC35  775A 9CA4 ABB3 81AB 73C8

* remotes/stefanha/tags/block-pull-request:
  block: Open by reference will try device then node_name.
  block: Relax bdrv_lookup_bs constraints.
  blockdev: Fix wrong usage of QDECREF causing snapshoted quorum to crash on close.
  block: mirror - use local_err to avoid NULL errp
  qemu-iotests: Don't run 005 on vmdk split formats
  block: qemu-iotests - add vhdx log replay tests for qemu-img
  block: qemu-iotests - fix test 070 (vhdx)
  block: Don't throw away errno via error_setg
  block: Add notes to iSCSI's .bdrv_open and .bdrv_reopen_prepare
  blockdev: Remove 'type' parameter from blockdev_init()
  sdhci: Drop unnecessary #include

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

Merge remote-tracking branch 'remotes/jliu/or32-ld-st' into staging

* remotes/jliu/or32-ld-st:
  target-openrisc: Use new qemu_ld/st opcodes

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

Merge remote-tracking branch 'remotes/jovanovic/mips-ufrp' into staging

* remotes/jovanovic/mips-ufrp:
  target-mips: add user-mode FR switch support for MIPS32r5
  target-mips: add support for CP0_Config5
  target-mips: add support for CP0_Config4
  target-mips: add CPU definition for MIPS32R5

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

char/serial: Fix emptyness check

This was guarding against a full fifo rather than an empty fifo when
popping. Fix.

Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>

gitignore: anchor all ignored names

by default, patterns/names in .gitignore are applied
recursively to all subdirectories.  So any name mentioned
in .gitignore is ignored in all subdirectores.  This is good
for, say. object files (*.o), but not good for particular
names which should be ignored only in one directory.  For
example, qemu-img.1 file is generated in the top directory,
and it should be ignored only there, not in some subdir.

At first, this might not matter much, but we have lots of
examples already where it actually does not help at all.
For example, top-level .gitignore ignores a file/dir named
"patches" (which is very questionable by itself), but it
is applied recursively, so git also ignores, for example,
debian/patches/ which should not be ignored.

So anchor all the names where appropriate.  .gitignore
should be cleaned up further, which will be addressed in
a subsequent patch.

Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>

vl: trim includes

Over time, lots of stuff moved from vl.c into separate
files.  But include statements has never been cleaned,
and they continue to carry lots of anymore-unused stuff.

Remove includes which are not relevant for vl.c anymore.
Apparently there are more includes like this, because
many are included from qemu-common.h and the like, or,
for example, I don't see were we use win32-specific
stuff in vl.c (so that maybe #include <windows.h> might
be removed too).

Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>

vl: remove old, long-unused defines

Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>

net: declare struct iovec in checksum.h to fix compiler warning

The checksum calculation header exports a function that refers to
struct iov defined in iov.h.  Without including the former, build
fails like this:

  In file included from hw/net/fsl_etsec/rings.c:24:0:
  include/net/checksum.h:51:31: error: ‘struct iovec’ declared inside parameter list [-Werror]
  include/net/checksum.h:51:31: error: its scope is only this definition or declaration, which is probably not what you want [-Werror]

Mention struct iovec there.

Reported-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>

linux-user: refactor do_socketcall()

Refactor do_socketcall() to do argument conversion/checking first,
according to a lookup table (which call has how many args) and
by calling the right function second with ready-to-go arguments.

This ensures that all arguments are handled as abi_long, according
to socketcall prototype, and simplifies argument handling alot too.

Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>

configure: add hints to a remedy for feature_not_found errors

Modify feature_not_found to accept an optional second parameter to be
printed after the generic feature not found error.

Modify most calls to feature_not_found to provide hints as to the
packages that may be missing. The few calls remaining without a remedy
are ones I couldn't work out how to remedy myself.

Signed-off-by: Stewart Smith <stewart@linux.vnet.ibm.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>

configure: add hint of libfdt to DTC dependency not found message

Most distros package it as libfdt, and mentioning libfdt here makes it
much easier to find the package you're missing.

Signed-off-by: Stewart Smith <stewart@linux.vnet.ibm.com>
Reviewed-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>

sparc/leon3: Initialize stack pointer

A lot of real world LEON3 systems are shipped with the GRMON boot
loader.  This boot loader initializes the stack pointer with the end of
RAM address.  The application can use this to detect the RAM size of a
particular board variant.

Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
Reviewed-by: Fabien Chouteau <chouteau@adacore.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>

misc: Fix case Qemu -> QEMU

Signed-off-by: Stefan Weil <sw@weilnetz.de>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>

qapi: Refine human printing of sizes

This fixes several bugs or shortcomings of the previous pretty-printer.
In particular:

* use PRIu64 instead of casting to long long

* the exact value is included too

* the correct unit of measure (MiB, GiB, etc.) is used.  PiB and EiB
are added too.

* due to an off-by-one error, 512*2^30 was printed as 0.500MiB rather than
512MiB.  floor(log2(val)) is equal to 63 - clz(val), while the code used 64.

* The desired specification is %g rather than %f, which always uses three
decimals in the current code.  However %g would switch to scientific
notation when the integer part is >= 1000 (e.g. 1000*2^30).  To keep the
code simple, switch to the higher power when the integer part is >= 1000;
overflow is avoided by using frexp instead of clz.

Suggested-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

qdev: Use QAPI type names for properties

Use "drive", "chr", etc. only for legacy_name (which shows up
in -device foo,? output).

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

qdev: Add enum property types to QAPI schema

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

block: Handle "rechs" and "large" translation options

Sure, CHS translation is an obscure topic, and legacy options for
hard-disk geometries are obscure as well.  But since QEMU does nothing
with it except telling the BIOS, and since there "large" and "rechs"
are listed in the enums, parsing them seems to be the bare minimum.

Acked-by: Stefan Hajnoczi <stefanha@gmail.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

qdev: Remove hex8/32/64 property types

Replace them with uint8/32/64.

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

qdev: Remove most legacy printers

Their functionality is either aesthetic only (e.g. on/off vs. true/false)
or obtained by the "human mode" of StringOutputVisitor.

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

qdev: Use human mode in "info qtree"

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

qapi: Add human mode to StringOutputVisitor

This will be used by "info qtree".  For numbers it prints both the
decimal and hex values.  For sizes it rounds to the nearest power
of 2^10.  For strings, it puts quotes around the string and separates
NULL and empty string.

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

qdev: Inline qdev_prop_parse()

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

qdev: Legacy properties are just strings

prop->info->legacy_name is still used by "-device foo,?".

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

qdev: Legacy properties are now read-only

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

qdev: Remove legacy parsers for hex8/32/64

The hexNN property types have not been accepting values not prefixed
by "0x" since QEMU 1.2.  Parse those values as decimals now.

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

qdev: Sizes are now parsed by StringInputVisitor

Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

qapi: Add size parser to StringInputVisitor

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

qtest: Don't segfault with invalid -qtest option

This prints an error message, instead of core dump, when "-qtest"
option value is invalid, e.g.:

    $ ./x86_64-softmmu/qemu-system-x86_64 -qtest unknown
        qemu-system-x86_64: Failed to initialize device for qtest:
        "unknown"

Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

ipack: Move IndustryPack out of hw/char/

Move the header defining an IPackBus and IPackDevice base class into
a new include/ directory and move their implementation and a
PCI-IndustryPack bridge out of hw/char/ directory into a new hw/ipack/.

Acked-by: Alberto Garcia <agarcia@igalia.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

ipoctal232: QOM parent field cleanup

Clean up accesses to IPOctalState::dev field and rename it.

Acked-by: Alberto Garcia <agarcia@igalia.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

ipack: QOM parent field cleanup for IPackDevice

Rename the IPackDevice::qdev field to avoid accidental use.

Acked-by: Alberto Garcia <agarcia@igalia.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

ipack: QOM parent field cleanup for IPackBus

Clean up the only user of IPackBus::qbus field and rename it.

Acked-by: Alberto Garcia <agarcia@igalia.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

ipack: Convert to QOM realize

Acked-by: Alberto Garcia <agarcia@igalia.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

tests: Add ipoctal232 qtest

Acked-by: Alberto Garcia <agarcia@igalia.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

tests: Add tpci200 qtest

Acked-by: Alberto Garcia <agarcia@igalia.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

tests: Add virtio-net qtest

Cc: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

tests: Add ne2000 qtest

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

block: Open by reference will try device then node_name.

Since we introduced node_name for named bs of the graph modify the opening by
reference to use it as a fallback.

This patch also enforce the separation of the device id and graph node
namespaces.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

block: Relax bdrv_lookup_bs constraints.

The following patch will reuse bdrv_lookup_bs in order to open images by
references so the rules of usage of bdrv_lookup_bs must be relaxed a bit.

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

blockdev: Fix wrong usage of QDECREF causing snapshoted quorum to crash on close.

As bdrv_open() documentation states:
"The reference to the QDict belongs to the block layer
 * after the call (even on failure), so if the caller intends to reuse the
 * dictionary, it needs to use QINCREF() before calling bdrv_open."

the optional options dict will not be reused after bdrv_open() and should
belong to the block layer so remove the extra QDECREF(options).

Signed-off-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

block: mirror - use local_err to avoid NULL errp

When starting a block job, commit_active_start() relies on whether *errp
is set by mirror_start_job.  This allows it to determine if the mirror
job start failed, so that it can clean up any changes to open flags from
the bdrv_reopen().  If errp is NULL, then it will not be able to
determine if mirror_start_job failed or not.

To avoid this, use a local Error variable, and then propagate the error
(if any) to errp.

Reported-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Jeff Cody <jcody@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qemu-iotests: Don't run 005 on vmdk split formats

There would be too many extents that VMDK driver can't open all of them:

    005 0s ... - output mismatch (see 005.out.bad)
    --- 005.out     2013-12-24 09:27:27.608181030 +0800
    +++ 005.out.bad 2014-02-13 10:00:15.282184557 +0800
    @@ -4,10 +4,10 @@
     Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=5368709120000

     small read
    -read 4096/4096 bytes at offset 1024
    -4 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
    +qemu-io: can't open device /tmp/qemu-iotests/t.vmdk: Could not open '/tmp/qemu-iotests/t-s1016.vmdk': Too many open files
    +no file open, try 'help open'

     small write
    -wrote 4096/4096 bytes at offset 8192
    -4 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
    +qemu-io: can't open device /tmp/qemu-iotests/t.vmdk: Could not open '/tmp/qemu-iotests/t-s1016.vmdk': Too many open files
    +no file open, try 'help open'
     *** done

So disable the two subformats.

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

block: qemu-iotests - add vhdx log replay tests for qemu-img

VHDX logs can now be replayed via 'qemu-img check -r all'.  Add
tests to verify that the log replay is successful when using qemu-img.

Signed-off-by: Jeff Cody <jcody@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

block: qemu-iotests - fix test 070 (vhdx)

VHDX test 070 failed, due to different output from qemu-io / qemu
when opening an image read-only that contains a log file.  Filter
the output, and update the expected results to match the correct
output.

Signed-off-by: Jeff Cody <jcody@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

block: Don't throw away errno via error_setg

There are a handful of places in the block layer where a failure path
has a valid -errno value, yet error_setg() is used.  Those instances
should instead use error_setg_errno(), to preserve as much error
information as possible.

This patch replaces those instances with error_setg_errno(), so that
errno is passed up the stack in the error message.

Reported-By: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Jeff Cody <jcody@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

tests: Add eepro100 qtest

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

tests: Add pcnet qtest

Test PCI only for now.

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

tests: Add rtl8139 qtest

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

tests: Add vmxnet3 qtest

Note that this will emit a warning:
[vmxnet3][WR][vmxnet3_peer_has_vnet_hdr]: Peer has no virtio extension.
Task offloads will be emulated.

Reviewed-by: Dmitry Fleytman <dmitry@daynix.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

tests: Add e1000 qtest

Cc: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

i2c: Drop FROM_I2C_SLAVE() macro

We now use type-specific QOM cast macros instead.

Reviewed-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

twl92230: QOM'ify

Replace usages of FROM_I2C_SLAVE() and direct parent field accesses with
QOM cast macro. Rename parent field.

Add missing braces while at it.

Reviewed-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

ds1338: QOM'ify

Replace usages of FROM_I2C_SLAVE() with QOM cast macro.
Rename parent field.

Reviewed-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

lm832x: QOM'ify

Replace usages of FROM_I2C_SLAVE() and direct parent field accesses with
QOM cast macro. Rename parent field.

Reviewed-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

max7310: QOM'ify

Replace FROM_I2C_SLAVE() usages with QOM cast macro.
Rename parent field.

Reviewed-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

ssd0303: QOM'ify

Replace usages of FROM_I2C_SLAVE() with QOM cast macro.
Rename parent field.

Reviewed-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

wm8750: QOM'ify

Replace usages of FROM_I2C_SLAVE() and direct parent field accesses with
QOM cast macro. Rename parent field.

Reviewed-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

z2: QOM'ify AER915

Replace usages of FROM_I2C_SLAVE() with QOM cast macro.
Rename the parent field. Reuse the type constant in z2_init().

Reviewed-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
[AF: Use TYPE_AER915 in z2_init() too]
Signed-off-by: Andreas Färber <afaerber@suse.de>

tosa: QOM'ify DAC

Replace usages of FROM_I2C_SLAVE() with QOM cast macro.
Rename parent field. Use type constant in tosa_tg_init().

Reviewed-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>

pxa2xx: QOM'ify I2C slave

Replace usages of FROM_I2C_SLAVE() and direct parent field accesses with
QOM cast macro. Rename parent field to assure we caught all. Reuse type
constant in pxa2xx_i2c_init().

Add some missing braces while at it.

Signed-off-by: Andreas Färber <afaerber@suse.de>