block: Improve message for device name clashing with node name

Suggested-by: Benoit Canet <benoit.canet@nodalink.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qemu-nbd: Destroy the BlockDriverState properly

Match the bdrv_new() with a bdrv_unref(), just to be tidy.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: Keep DriveInfo alive until BlockDriverState dies

If the BDS's refcnt > 0, drive_del() destroys the DriveInfo, but not
the BDS.  This can happen in three places:

* Device model destruction during unplug: blockdev_auto_del()

* Xen IDE unplug: pci_piix3_xen_ide_unplug()

* drive_del command when no device model is attached: do_drive_del()

The other callers of drive_del are on error paths where refcnt == 1.

If the user somehow manages to plug in a device model using a BDS that
has gone through drive_del(), the legacy configuration passed in
DriveInfo doesn't reach the device model, and automatic deletion on
unplug doesn't work.  Worse, some device models such as scsi-disk
crash when DriveInfo doesn't exist.

This is theoretical; I didn't research an actual reproducer. The problem
was introduced when we replaced DriveInfo reference counting by BDS
reference counting in commit a94a3fa..fa510eb.

Fix by keeping DriveInfo alive until its BDS dies.

This affects qemu_drive_opts: now you can't reuse the same ID for new
drive options until the BDS dies.  Before, you could, but since the
code always attempts to create a BDS with the same ID next, the
enclosing operation "create a new drive" failed anyway.  Different
error path, same result.

Unfortunately, the fix involves use of blockdev.c stuff from block.c,
which is a layering violation.  Fortunately, my forthcoming
BlockBackend work will get rid of it again.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

blockdev: Disentangle BlockDriverState and DriveInfo creation

blockdev_init() mixes up BlockDriverState and DriveInfo initialization
Finish the BlockDriverState job before starting to mess with
DriveInfo.  Easier on the eyes.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

blkdebug: show an error for invalid event names

It is easy to typo a blkdebug configuration and waste a lot of time
figuring out why no rules are matching.

Push the Error** down into add_rule() so we can report an error when the
event name is invalid.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

Merge remote-tracking branch 'remotes/mcayland/tags/qemu-sparc-signed' into staging

tcx: Implement hardware acceleration

# gpg: Signature made Tue 23 Sep 2014 22:52:34 BST using RSA key ID AE0F321F
# gpg: Can't check signature: public key not found

* remotes/mcayland/tags/qemu-sparc-signed:
  tcx: Implement hardware acceleration

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

virtio: Fix wrong type cast from pointer to long

Compiler warning (w32, w64):

include/hw/virtio/virtio_ring.h:142:26: warning:
 cast from pointer to integer of different size [-Wpointer-to-int-cast]

When sizeof(long) < sizeof(void *), this is not only a warning but a
real program error.

Add also missing blanks in the same statement.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1411536002-14088-1-git-send-email-sw@weilnetz.de
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/awilliam/tags/vfio-pci-for-qemu-20140923.0' into staging

Endian updates to re-fix cross endian host and guest and
enable the same for ROM loading (Alexey)

# gpg: Signature made Tue 23 Sep 2014 18:03:03 BST using RSA key ID 3BB08B22
# gpg: Can't check signature: public key not found

* remotes/awilliam/tags/vfio-pci-for-qemu-20140923.0:
  vfio: make rom read endian sensitive
  Revert "vfio: Make BARs native endian"

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

tcx: Implement hardware acceleration

The S24/TCX framebuffer is a mildly accelerated video card with
blitter, stippler and hardware cursor.

* Solaris and NetBSD 6.x use all the hardware acceleration features
* The Xorg driver (used by Linux) can use the hardware cursor only

This patch implements hardware acceleration in both 8 bit and 24 bit
modes. It is based on the NetBSD driver sources and from tests with
Solaris.

Signed-off-by: Olivier Danet <odanet@caramail.com>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>

slirp: udp: fix NULL pointer dereference because of uninitialized socket

When guest sends udp packet with source port and source addr 0,
uninitialized socket is picked up when looking for matching and already
created udp sockets, and later passed to sosendto() where NULL pointer
dereference is hit during so->slirp->vnetwork_mask.s_addr access.

Fix this by checking that the socket is not just a socket stub.

This is CVE-2014-3640.

Signed-off-by: Petr Matousek <pmatouse@redhat.com>
Reported-by: Xavier Mehrenberger <xavier.mehrenberger@airbus.com>
Reported-by: Stephane Duverger <stephane.duverger@eads.net>
Reviewed-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael Tokarev <mjt@tls.msk.ru>
Message-id: 20140918063537.GX9321@dhcp-25-225.brq.redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/kraxel/tags/pull-usb-20140923-1' into staging

usb: enable hotplug, switch to realize, ohci tracing, misc fixes.

# gpg: Signature made Tue 23 Sep 2014 12:42:29 BST using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>"
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>"
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>"

* remotes/kraxel/tags/pull-usb-20140923-1: (26 commits)
  usb: tag standalone ehci as hotpluggable
  usb: tag standalone uhci as hotpluggable
  usb: tag xhci as hotpluggable
  usb-serial: only check speed once at realize time
  usb-bus: introduce a wrapper function to check speed
  usb-bus: remove "init" from USBDeviceClass struct
  usb-mtp: convert init to realize
  usb-redir: convert init to realize
  usb-audio: convert init to realize
  dev-wacom: convert init to realize
  dev-hid: convert init to realize
  usb-ccid: convert init to realize
  dev-serial: convert init to realize
  dev-bluetooth: convert init to realize
  dev-uas: using error_report instead of fprintf
  dev-uas: convert init to realize
  dev-storage: usring error_report instead of fprintf/printf
  dev-storage: convert init to realize
  usb-hub: convert init to realize
  libusb: using error_report instead of fprintf
  ...

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

Merge remote-tracking branch 'remotes/borntraeger/tags/s390x-20140923' into staging

s390x/kvm: some fixes and cleanups

1. sclp: get of of duplicate defines
2. ccw: implement and fix handling of some special cases

# gpg: Signature made Tue 23 Sep 2014 13:10:47 BST using RSA key ID B5A61C7C
# gpg: Can't check signature: public key not found

* remotes/borntraeger/tags/s390x-20140923:
  s390x/css: catch ccw sequence errors
  s390x/css: support format-0 ccws
  s390x: remove duplicate defines in SCLP code

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

s390x/css: catch ccw sequence errors

We must not allow chains of more than 255 ccws without data transfer.

Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Jens Freimann <jfrei@linux.vnet.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>

s390x/css: support format-0 ccws

Add support for format-0 ccws in channel programs. As a format-1 ccw
contains the same information as format-0 ccws, only supporting larger
addresses, simply convert every ccw to format-1 as we walk the chain.

Reviewed-by: David Hildenbrand <dahi@linux.vnet.ibm.com>
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Jens Freimann <jfrei@linux.vnet.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>

s390x: remove duplicate defines in SCLP code

Let's get rid of these duplicate defines.

Signed-off-by: Jens Freimann <jfrei@linux.vnet.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>

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

# gpg: Signature made Mon 22 Sep 2014 12:41:59 BST using RSA key ID 81AB73C8
# gpg: Good signature from "Stefan Hajnoczi <stefanha@redhat.com>"
# gpg:                 aka "Stefan Hajnoczi <stefanha@gmail.com>"

* remotes/stefanha/tags/block-pull-request: (59 commits)
  block: Always compile virtio-blk dataplane
  vring: Better error handling if num is too large
  virtio: Import virtio_vring.h
  async: aio_context_new(): Handle event_notifier_init failure
  block: vhdx - fix reading beyond pointer during image creation
  block: delete cow block driver
  block/archipelago: Fix typo in qemu_archipelago_truncate()
  ahci: Add test_identify case to ahci-test.
  ahci: Add test_hba_enable to ahci-test.
  ahci: Add test_hba_spec to ahci-test.
  ahci: properly shadow the TFD register
  ahci: add test_pci_enable to ahci-test.
  ahci: Add test_pci_spec to ahci-test.
  ahci: MSI capability should be at 0x80, not 0x50.
  ahci: Adding basic functionality qtest.
  layout: Add generators for refcount table and blocks
  fuzz: Add fuzzing functions for entries of refcount table and blocks
  docs: List all image elements currently supported by the fuzzer
  qapi/block-core: Add "new" qcow2 options
  qcow2: Add overlap-check.template option
  ...

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

usb: tag standalone ehci as hotpluggable

Add a flag to EHCIPCIInfo saying whenever the controller supports
companions or not.  Make sure we only allow registering companions for
ehci versions supporting that.  Enable pci hotplug for the ehci
variants not supporting companions.

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

usb: tag standalone uhci as hotpluggable

uhci hostadapters in companion setups can't be hotplugged.  So leave
hotplug disabled for all ich9 variants (which are already tagged with
unplug = true in the info struct).  For the other variants we'll enable
hotplug and remove the companion setup properties.

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

usb: tag xhci as hotpluggable

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

usb-serial: only check speed once at realize time

Whatever the chardev is open or not, we should assure
the speed is matched each other. So, call usb_check_attach()
check speed. And then pass &error_abort at all calls to
usb_device_attach().

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

usb-bus: introduce a wrapper function to check speed

In this way, we can check speed directly, don't need
call usb_device_attach(), which has other conditions,
such as checking the chardev is open.

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

usb-bus: remove "init" from USBDeviceClass struct

All usb-bus devices are realized by realize(),
remove init callback function from USBDeviceClass struct.

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

usb-mtp: convert init to realize

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

usb-redir: convert init to realize

In this way, all the implementations now use
error_setg instead of qerror_report for reporting error.

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

usb-audio: convert init to realize

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

dev-wacom: convert init to realize

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

dev-hid: convert init to realize

In this way, all the implementations now use
error_setg instead of error_report for reporting error.

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

usb-ccid: convert init to realize

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

dev-serial: convert init to realize

In this way, all the implementations now use
error_setg instead of error_report for reporting error.

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

dev-bluetooth: convert init to realize

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

dev-uas: using error_report instead of fprintf

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

dev-uas: convert init to realize

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

dev-storage: usring error_report instead of fprintf/printf

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

dev-storage: convert init to realize

In this way, all the implementations now use
error_setg instead of error_report for reporting error.

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

usb-hub: convert init to realize

In this way, all the implementations now use
error_setg instead of error_report for reporting error.

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

libusb: using error_report instead of fprintf

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

libusb: convert init to realize

In this way, all the implementations now use
error_setg instead of error_report for reporting error.

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

usb-net: convert init to realize

meanwhile, qerror_report_err() is a transitional interface to
help with converting existing HMP commands to QMP. It should
not be used elsewhere.

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

usb-bus: convert USBDeviceClass init to realize

Add "realize/unrealize" in USBDeviceClass, which has errp
as a parameter. So all the implementations now use
error_setg instead of error_report for reporting error.

Note: this patch still keep "init" in USBDeviceClass, and
call kclass->init in usb_device_realize(), avoid breaking
git bisect. After realize all usb devices, will be removed.

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

ohci: Convert fprint/DPRINTF/print to traces

This converts many kinds of debug prints to traces.

This implements packets logging to avoid unnecessary calculations if
usb_ohci_td_pkt_short/usb_ohci_td_pkt_long is not enabled.

This makes OHCI errors (such as "DMA error") invisible by default.

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

Merge remote-tracking branch 'remotes/mjt/tags/trivial-patches-2014-09-22' into staging

trivial patches for 2014-09-22

# gpg: Signature made Mon 22 Sep 2014 09:10:03 BST using RSA key ID A4C3D7DB
# 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: 6F67 E18E 7C91 C5B1 5514  66A7 BEE5 9D74 A4C3 D7DB

* remotes/mjt/tags/trivial-patches-2014-09-22:
  arch_init: Setting QEMU_ARCH enum straight
  pc: Add missing 'static' attribute
  block: allow creation of fixed vhdx images
  vl: Print maxmem in hex format for error message
  configure: trivial fixes
  xen-hvm.c: Always return -1 when failure occurs in xen_hvm_init()
  rdma: Fix incorrect description in comments
  Fix typos and misspellings in comments
  qemu-char: Permit only a single "stdio" character device

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

usb-storage: fix possible memory leak and missing error message

When scsi_bus_legacy_add_drive() return NULL, meanwhile err will
be not NULL, which will casue memory leak and missing error message.

Cc: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

vfio: make rom read endian sensitive

All memory regions used by VFIO are LITTLE_ENDIAN and they
already take care of endiannes when accessing real device BARs
except ROM - it was broken on BE hosts.

This fixes endiannes for ROM BARs the same way as it is done
for other BARs.

This has been tested on PPC64 BE/LE host/guest in all possible
combinations including TCG.

Signed-off-by: Nikunj A Dadhania <nikunj@linux.vnet.ibm.com>
[aik: added commit log]
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

Revert "vfio: Make BARs native endian"

This reverts commit c40708176a6b52b73bec14796b7c71b882ceb102.

The resulting code wrongly assumed target and host endianness are
the same which is not always the case for PPC64.

[aw: or potentially any host supporting VFIO and TCG]

Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

block: Always compile virtio-blk dataplane

Dataplane doesn't depend on linux-aio any more, so we don't need the
compiling condition now.

Configure options are kept but just print a message.

Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1410329871-28885-4-git-send-email-famz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

vring: Better error handling if num is too large

To be more consistent inside this function.

Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1410329871-28885-3-git-send-email-famz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

virtio: Import virtio_vring.h

This header has no further dependencies. It only has some stable data
types and primitive functions, so we can copy it to include/hw/virtio in
order to allow vring code (and its user virtio-blk dataplane) to be
built unconditionally, even for cross compiling.

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1410329871-28885-2-git-send-email-famz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

async: aio_context_new(): Handle event_notifier_init failure

On a system with a low limit of open files the initialization
of the event notifier could fail and QEMU exits without printing any
error information to the user.

The problem can be easily reproduced by enforcing a low limit of open
files and start QEMU with enough I/O threads to hit this limit.

The same problem raises, without the creation of I/O threads, while
QEMU initializes the main event loop by enforcing an even lower limit of
open files.

This commit adds an error message on failure:

 # qemu [...] -object iothread,id=iothread0 -object iothread,id=iothread1
 qemu: Failed to initialize event notifier: Too many open files in system

Signed-off-by: Chrysostomos Nanakos <cnanakos@grnet.gr>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

block: vhdx - fix reading beyond pointer during image creation

In vhdx_create_metadata(), we allocate 40 bytes to entry_buffer for
the various metadata table entries.  However, we write out 64kB from
that buffer into the new file.  Only write out the correct 40 bytes.

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

block: delete cow block driver

This patch removes support for the cow file format.

Normally we do not break backwards compatibility but in this case there
is no impact and it is the most logical option.  Extraordinary claims
require extraordinary evidence so I will show why removing the cow block
driver is the right thing to do.

The cow file format is the disk image format for Usermode Linux, a way
of running a Linux system in userspace.  The performance of UML was
never great and it was hacky, but it enjoyed some popularity before
hardware virtualization support became mainstream.

QEMU's block/cow.c is supposed to read this image file format.
Unfortunately the file format was underspecified:

1. Earlier Linux versions used the MAXPATHLEN constant for the backing
   filename field.  The value of MAXPATHLEN can change, so Linux
   switched to a 4096 literal but QEMU has a 1024 literal.

2. Padding was not used on the header struct (both in the Linux kernel
   and in QEMU) so the struct layout varied across architectures.  In
   particular, i386 and x86_64 were different due to int64_t alignment
   differences.  Linux now uses __attribute__((packed)), QEMU does not.

Therefore:

1. QEMU cow images do not conform to the Linux cow image file format.

2. cow images cannot be shared between different host architectures.

This means QEMU cow images are useless and QEMU has not had bug reports
from users actually hitting these issues.

Let's get rid of this thing, it serves no purpose and no one will be
affected.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-id: 1410877464-20481-1-git-send-email-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

block/archipelago: Fix typo in qemu_archipelago_truncate()

Fix a typo introduced by 94c80a438c85f2c19698547fbb115ea46d80c5f1

Signed-off-by: Chrysostomos Nanakos <cnanakos@grnet.gr>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

ahci: Add test_identify case to ahci-test.

Utilizing all of the bring-up code in pci_enable and hba_enable,
this test issues a simple IDENTIFY command via the HBA and retrieves
the response via the PIO receive mechanisms of the HBA.

Bugs: The DPS interrupt (Descriptor Processed Status) does not
currently get set. This will need to be adjusted in a future
patch series when the AHCI DMA pathways are reworked to allow
the feature, which may be utilized by OSX guests.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1408643079-30675-9-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

ahci: Add test_hba_enable to ahci-test.

This test engages the HBA functionality and initializes
values to sane defaults to allow for minimal HBA functionality.

Buffers are allocated and pointers are updated to allow minimal
I/O commands to complete as expected. Error registers and responses
are sanity checked for specification adherence.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1408643079-30675-8-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

ahci: Add test_hba_spec to ahci-test.

Add a test routine that checks the boot-up values of the HBA
configuration memory space against the AHCI 1.3 specification
and Intel ICH9 data sheet (for Q35 machines) for adherence and
sane values.

The HBA is not yet engaged or put into the idle state.

[Replaced g_assert_false(...) with g_assert(!...) for glib <2.38
compatibility, reported by Peter Maydell <peter.maydell@linaro.org>.
--Stefan]

Signed-off-by: John Snow <jsnow@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1408643079-30675-7-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

ahci: properly shadow the TFD register

In a real AHCI device, several S/ATA registers are mirrored or shadowed
within the AHCI register set. These registers are not updated
synchronously for each read access, but are instead updated after a
Device-to-Host Register FIS packet is received. The D2H FIS contains
the values from these registers on the device.

In QEMU, by reaching directly into the device to grab these bits before
they are "sent," we may introduce race conditions where unexpected
values are present "before they are sent" which could cause issues for
some guests, particularly if an attempt is made to read the PxTFD
register prior to enabling the port, where incorrect values will be read.

This patch also addresses the boot-time values for the PxTFD and PxSIG
registers to bring them in line with the AHCI 1.3 specification.

Lastly, several fields (PxTFD, PxSIG and PxSACT) are read-only,
and any attempts to write to them should be ignored.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1408643079-30675-6-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

ahci: add test_pci_enable to ahci-test.

This adds a test wherein we engage the PCI AHCI
device and ensure that the memory region for the
HBA functionality is now accessible.

Under Q35 environments, additional PCI configuration
is performed to ensure that the HBA functionality
will become usable.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1408643079-30675-5-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

ahci: Add test_pci_spec to ahci-test.

Adds a specification adherence test for AHCI
where the boot-up values for the PCI configuration space
are compared against the AHCI 1.3 specification.

This test does not itself attempt to engage the device.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1408643079-30675-4-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

ahci: MSI capability should be at 0x80, not 0x50.

In the Intel ICH9 data sheet, the MSI capability offset
in the PCI configuration space for ICH9 AHCI devices is
specified to be 0x80.

Further, the PCI capability pointer should always point
to 0x80 in ICH9 devices, despite the fact that AHCI 1.3
specifies that it should be pointing to PMCAP (Which in
this instance would be 0x70) to maintain adherence to
the Intel data sheet specifications and real observed behavior.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1408643079-30675-3-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

ahci: Adding basic functionality qtest.

Currently, there is no qtest to test the functionality of
the AHCI functionality present within the Q35 machine type.

This patch adds a skeleton for an AHCI test suite,
and adds a simple sanity-check test case where we
identify that the AHCI device is present, then
disengage the virtual machine.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1408643079-30675-2-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

layout: Add generators for refcount table and blocks

Refcount structures are placed in clusters randomly selected from all
unallocated host clusters.

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Maria Kustova <maria.k@catit.be>
Reviewed-by: Fam Zheng <famz@redhat.com>
Message-id: 7e2f38608db6fba2da53997390b19400d445c45d.1408450493.git.maria.k@catit.be
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

fuzz: Add fuzzing functions for entries of refcount table and blocks

Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Maria Kustova <maria.k@catit.be>
Message-id: c9f4027b6f401c67e9d18f94aed29be445e81d48.1408450493.git.maria.k@catit.be
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

docs: List all image elements currently supported by the fuzzer

Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Maria Kustova <maria.k@catit.be>
Message-id: cb71485d0f55d1d8401eebaead8324eb78673060.1408450493.git.maria.k@catit.be
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qapi/block-core: Add "new" qcow2 options

qcow2 supports more than four options by now, add the new options
(overlap check mode and metadata cache size)

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1408557576-14574-5-git-send-email-mreitz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qcow2: Add overlap-check.template option

Being able to set the overlap-check option to a string and then refine
it via the overlap-check.* options is a nice idea for the command line
but does not work so well for non-flattened dicts. In that case, one can
only specify either but not both, so add a field to overlap-check.*
which does the same as directly specifying overlap-check but can be used
in conjunction with the other fields in non-flattened dicts.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1408557576-14574-4-git-send-email-mreitz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qapi: Allow enums in anonymous unions

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1408557576-14574-3-git-send-email-mreitz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qcow2: Fix leak of QemuOpts in qcow2_open()

Currently, the QemuOpts object opts is leaked if anything fails from its
creation up to and including the image repair block. Fix this by freeing
that object in the fail path.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Message-id: 1408557576-14574-2-git-send-email-mreitz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

hmp: fix memory leak at hmp_info_block_jobs()

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-id: 1410874615-14292-1-git-send-email-arei.gonglei@huawei.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

image-fuzzer: Trivial readability and formatting improvements

Signed-off-by: Maria Kustova <maria.k@catit.be>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

iotests: Add more tests for qcow2 corruption

Add tests for unaligned L1/L2/reftable entries and non-fatal corruption
reports.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1409926039-29044-6-git-send-email-mreitz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qcow2: Check L1/L2/reftable entries for alignment

Offsets taken from the L1, L2 and refcount tables are generally assumed
to be correctly aligned. However, this cannot be guaranteed if the image
has been written to by something different than qemu, thus check all
offsets taken from these tables for correct cluster alignment.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1409926039-29044-5-git-send-email-mreitz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qcow2: Use qcow2_signal_corruption() for overlaps

Use the new function in case of a failed overlap check.

This changes output in case of corruption, so adapt iotest 060's
reference output accordingly.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Message-id: 1409926039-29044-4-git-send-email-mreitz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qcow2: Add qcow2_signal_corruption()

Add a helper function for easily marking an image corrupt (on fatal
corruptions) while outputting an informative message to stderr and via
QAPI.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Message-id: 1409926039-29044-3-git-send-email-mreitz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qapi/block: Add "fatal" to BLOCK_IMAGE_CORRUPTED

Not every BLOCK_IMAGE_CORRUPTED event must be fatal; for example, when
reading from an image, they should generally not be. Nonetheless, even
an image only read from may of course be corrupted and this can be
detected during normal operation. In this case, a non-fatal event should
be emitted, but the image should not be marked corrupt (in accordance to
"fatal" set to false).

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1409926039-29044-2-git-send-email-mreitz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qapi: Sort items in BlockdevOptions definition

Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1410415798-20673-4-git-send-email-famz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qapi: Sort BlockdevDriver enum data list

Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1410415798-20673-3-git-send-email-famz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

block: Introduce "null" drivers

This is an analogue to Linux null_blk. It can be used for testing or
benchmarking block device emulation and general block layer
functionalities such as coroutines and throttling, where disk IO is not
necessary or wanted.

Use null-aio:// for AIO version, and null-co:// for coroutine version.

[Resolved conflict with Fam's async bdrv_aio_cancel() series:
1. Drop .bdrv_aio_cancel() since it is now done by block.c
2. Rename qemu_aio_release() to qemu_aio_unref()
--Stefan]

Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Benoît Canet <benoit.canet@nodalink.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1410415798-20673-2-git-send-email-famz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

aio-win32: avoid out-of-bounds access to the events array

If ret is WAIT_TIMEOUT and there was an event returned by select(),
we can write to a location after the end of the array.  But in
that case we can retry the WaitForMultipleObjects call with the
same set of events, so just move the event[ret - WAIT_OBJECT_0]
assignment inside the existin conditional.

Reported-by: TeLeMan <geleman@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: TeLeMan <geleman@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qdev-monitor: fix segmentation fault on qdev_device_help()

Normally, qmp_device_list_properties() may return NULL when
a device haven't special properties excpet Object and DeviceState
properties, such as virtio-balloon-device.

We just need check local_err instead of prop_list.

Example:

Segmentation fault (core dumped)

The backtrace as below:

Program received signal SIGSEGV, Segmentation fault.
0x00005555559af1a8 in error_get_pretty (err=0x0) at util/error.c:152
152         return err->msg;
(gdb) bt
    func=0x55555574a6ca <device_help_func>, opaque=0x0, abort_on_failure=0) at util/qemu-option.c:1072

Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

block: Rename qemu_aio_release -> qemu_aio_unref

Suggested-by: Benoît Canet <benoit.canet@irqsave.net>
Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

block: Drop AIOCBInfo.cancel

Now that all the implementations are converted to asynchronous version
and we can emulate synchronous cancellation with it. Let's drop the
unused member.

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

ide: Convert trim_aiocb_info.cancel to .cancel_async

We know that either bh is scheduled or ide_issue_trim_cb will be called
again, so we just set i, j and ret to the right values. In both cases,
ide_trim_bh_cb will be called.

Also forward the cancellation to the iocb->aiocb which we get from
bdrv_aio_discard.

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

win32-aio: Drop win32_aiocb_info.cancel

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

sheepdog: Convert sd_aiocb_info.cancel to .cancel_async

Also drop the now unused SheepdogAIOCB.finished field. Note that this
aio is internal to sheepdog driver and has NULL cb and opaque, and
should be unused at all.

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

rbd: Drop rbd_aiocb_info.cancel

And also drop the now unused "cancelled" field.

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

quorum: Convert quorum_aiocb_info.cancel to .cancel_async

Before, we cancel all the child requests with bdrv_aio_cancel, then free
the acb..

Now we just kick off asynchronous cancellation of child requests and
return, we know quorum_aio_cb will be called later, so in the end
quorum_aio_finalize will take care of calling the caller's cb.

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

quorum: fix quorum_aio_cancel()

For a fifo read pattern, we only have one running aio (possible other cases that
has less number than num_children in the future), so we need to check if
.acb is NULL against bdrv_aio_cancel() to avoid segfault.

Cc: Eric Blake <eblake@redhat.com>
Cc: Benoit Canet <benoit@irqsave.net>
Cc: Kevin Wolf <kwolf@redhat.com>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Liu Yuan <namei.unix@gmail.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

qed: Drop qed_aiocb_info.cancel

Also drop the now unused ->finished field.

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

curl: Drop curl_aiocb_info.cancel

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

blkverify: Drop blkverify_aiocb_info.cancel

Also the finished pointer is not used any more.

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

blkdebug: Drop blkdebug_aiocb_info.cancel

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

archipelago: Drop archipelago_aiocb_info.cancel

The cancelled flag is no longer useful. Later the request will complete
as before, and cb will be called.

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

iscsi: Convert iscsi_aiocb_info.cancel to .cancel_async

Also drop the unused field "canceled".

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

dma: Convert dma_aiocb_info.cancel to .cancel_async

Just forward the request to bdrv_aio_cancel_async.

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

linux-aio: Convert laio_aiocb_info.cancel to .cancel_async

Just call io_cancel (2), if it fails, it means the request is not
canceled, so the event loop will eventually call
qemu_laio_process_completion.

In qemu_laio_process_completion, change to call the cb unconditionally.
It is required by bdrv_aio_cancel_async.

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

thread-pool: Convert thread_pool_aiocb_info.cancel to cancel_async

The .cancel_async shares the same the first half with .cancel: try to
steal the request if not submitted yet. In this case set the elem to
THREAD_DONE status and ret to -ECANCELED, which means
thread_pool_completion_bh will call the cb with -ECANCELED.

If the request is already submitted, do nothing, as we know the normal
completion will happen in the future.

Testing code update:

Before, done_cb is only called if the request is already submitted by
thread pool. Now done_cb is always called, even before it is submitted,
because we emulate bdrv_aio_cancel with bdrv_aio_cancel_async. So also
update the test criteria accordingly.

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

block: Drop bdrv_em_aiocb_info.cancel

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

block: Drop bdrv_em_co_aiocb_info.cancel

Also drop the now unused ->done pointer.

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

block: Add bdrv_aio_cancel_async

This is the async version of bdrv_aio_cancel, which doesn't block the
caller. It guarantees that the cb is called either before returning or
some time later.

bdrv_aio_cancel can base on bdrv_aio_cancel_async, later we can convert
all .io_cancel implementations to .io_cancel_async, and the aio_poll is
the common logic. In the end, .io_cancel can be dropped.

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

block: Add refcnt in BlockDriverAIOCB

This will be useful in synchronous cancel emulation with
bdrv_aio_cancel_async.

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

ide/ahci: Check for -ECANCELED in aio callbacks

Before, bdrv_aio_cancel will either complete the request (like normal)
and call CB with an actual return code, or skip calling the request (for
example when the IO req is not submitted by thread pool yet).

We will change bdrv_aio_cancel to do it differently: always call CB
before return, with either [1] a normal req completion ret code, or [2]
ret == -ECANCELED. So the callers' callback must accept both cases. The
existing logic works with case [1], but not [2].

The simplest transition of callback code is do nothing in case [2], just
as if the CB is not called by the bdrv_aio_cancel() call.

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>