s390-ccw.img: Enhance drain_irqs().

- Use tpi + tsch to get interrupts.
- Return an error if the irb indicates problems.

Signed-off-by: Cornelia Huck <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

s390-ccw.img: Detect devices with stsch.

stsch is the canonical way to detect devices. As a bonus, we can
abort the loop if we get cc 3, and we need to check only the valid
devices (dnv set).

Signed-off-by: Cornelia Huck <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

s390-ccw.img: Fix compile warning in s390 ccw virtio code

Lets fix this gcc warning:

virtio.c: In function ‘vring_send_buf’:
virtio.c:125:35: error: operation on ‘vr->next_idx’ may be undefined
[-Werror=sequence-point]

Signed-off-by: Christian Borntraeger <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

s390-ccw.img: Take care of the elf->img transition

We have to call strip with s390-ccw.elf as input and
s390-ccw.img as output

Signed-off-by: Christian Borntraeger <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

s390-ccw.img: build s390-ccw rom on s3900 system by default

Lets build the s390-ccw rom if on s390. Also fix the separate build
folder case.

Signed-off-by: Christian Borntraeger <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

s390-ccw.img: replace while loop with a disabled wait on s390 bios

dont waste cpu power on an error condition. Lets stop the guest
with a disabled wait.

Signed-off-by: Christian Borntraeger <[email protected]>
Signed-off-by: Alexander Graf <[email protected]>

S390: ccw firmware: Add Makefile

This patch adds a makefile, so we can build our ccw firmware. Also
add the resulting binaries to .gitignore, so that nobody is annoyed
they might be in the tree.

Signed-off-by: Alexander Graf <[email protected]>

S390: ccw firmware: Add bootmap interpreter

On s390, there is an architected boot map format that we can read to
boot a certain entry off the disk. Implement a simple reader for this
that always boots the first (default) entry.

Signed-off-by: Alexander Graf <[email protected]>

S390: ccw firmware: Add glue header

Like all great programs, we have to call between different functions in
different object files. And all of them need a common ground of defines.

Provide a file that provides these defines.

Signed-off-by: Alexander Graf <[email protected]>

S390: ccw firmware: Add virtio device drivers

In order to boot, we need to be able to access a virtio-blk device through
the CCW bus. Implement support for this.

Signed-off-by: Alexander Graf <[email protected]>

S390: ccw firmware: Add sclp output

In order to communicate with the user, we need an I/O mechanism that he
can read. Implement SCLP ASCII support, which happens to be the default
in the s390 ccw machine.

This file is missing read support for now. It can only print messages.

Signed-off-by: Alexander Graf <[email protected]>

S390: ccw firmware: Add main program

This C file is the main driving piece of the s390 ccw firmware. It
provides a search for a workable block device, sets it as the default
to boot off of and boots from it.

Signed-off-by: Alexander Graf <[email protected]>

S390: ccw firmware: Add start assembly

We want to write most of our code in C, so add a small assembly
stub that jumps straight into C code for us to continue booting.

Signed-off-by: Alexander Graf <[email protected]>

S390: IPL: Use different firmware for different machines

We have a virtio-s390 and a virtio-ccw machine in QEMU. Both use vastly
different ways to do I/O. Having the same firmware blob for both doesn't
really make any sense.

Instead, let's parametrize the firmware file name, so that we can have
different blobs for different machines.

Signed-off-by: Alexander Graf <[email protected]>

S390: IPL: Support ELF firmware

Our firmware blob is always a raw file that we load at a fixed address today.
Support loading an ELF blob instead that we can map high up in memory.

This way we don't have to be so conscious about size constraints.

Signed-off-by: Alexander Graf <[email protected]>

S390: Make IPL reset address dynamic

We can have different load addresses for different blobs we boot with.
Make the reset IP dynamic, so that we can handle things more flexibly.

Signed-off-by: Alexander Graf <[email protected]>

gtk: refactor menu creation

Signed-off-by: Anthony Liguori <[email protected]>

Merge remote-tracking branch 'stefanha/block' into staging

# By Liu Yuan (4) and others
# Via Stefan Hajnoczi
* stefanha/block:
  sheepdog: fix loadvm operation
  sheepdog: resend write requests when SD_RES_READONLY is received
  sheepdog: add helper function to reload inode
  sheepdog: add SD_RES_READONLY result code
  sheepdog: cleanup find_vdi_name
  rbd: Fix use after free in rbd_open()
  block: Disable driver-specific options for 1.5
  sheepdog: implement .bdrv_co_is_allocated()
  sheepdog: use BDRV_SECTOR_SIZE
  sheepdog: add discard/trim support for sheepdog
  block/ssh: Require libssh2 >= 1.2.8.

Message-id: 1366976682[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

Merge remote-tracking branch 'stefanha/trivial-patches' into staging

# By Ed Maste (2) and others
# Via Stefan Hajnoczi
* stefanha/trivial-patches:
  bsd-user: Track change in FreeBSD SYSCTL(9) types
  virtio: Fix compilation without CONFIG_VHOST_SCSI
  qemu-doc: Option -ignore-environment removed.
  s390x: use CONFIG_INT128 to detect __uint128_t
  linux-user: fix compile error due to stray colon at end of #ifdef line

Message-id: 1366975563[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

sheepdog: fix loadvm operation

Currently the 'loadvm' opertaion works as following:
1. switch to the snapshot
2. mark current working VDI as a snapshot
3. rely on sd_create_branch to create a new working VDI based on the snapshot

This works not the same as other format as QCOW2. For e.g,

qemu > savevm # get a live snapshot snap1
qemu > savevm # snap2
qemu > loadvm 1 # This will steally create snap3 of the working VDI

Which will result in following snapshot chain:

base <-- snap1 <-- snap2 <-- snap3
          ^
          |
      working VDI

snap3 was unnecessarily created and might be annoying users.

This patch discard the unnecessary 'snap3' creation. and implement
rollback(loadvm) operation to the specified snapshot by
1. switch to the snapshot
2. delete working VDI
3. rely on sd_create_branch to create a new working VDI based on the snapshot

The snapshot chain for above example will be:

base <-- snap1 <-- snap2
          ^
          |
      working VDI

Cc: [email protected]
Cc: MORITA Kazutaka <[email protected]>
Cc: Kevin Wolf <[email protected]>
Cc: Stefan Hajnoczi <[email protected]>
Signed-off-by: Liu Yuan <[email protected]>
Reviewed-by: MORITA Kazutaka <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

sheepdog: resend write requests when SD_RES_READONLY is received

When a snapshot is taken from out side of qemu (e.g. qemu-img
snapshot), write requests to the current vdi return SD_RES_READONLY.
In this case, the sheepdog block driver needs to update the current
inode to the latest one and resend the write requests.

Signed-off-by: MORITA Kazutaka <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

sheepdog: add helper function to reload inode

This adds a helper function to update the current inode state with the
specified vdi object.

Signed-off-by: MORITA Kazutaka <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

sheepdog: add SD_RES_READONLY result code

Sheepdog returns SD_RES_READONLY when qemu sends write requests to the
snapshot vdi.  This adds the result code and makes sd_strerror() print
its error reason.

Signed-off-by: MORITA Kazutaka <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

sheepdog: cleanup find_vdi_name

This makes 'filename' and 'tag' constant variables, and renames
'for_snapshot' to 'lock' to clear how it works.

Signed-off-by: MORITA Kazutaka <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

rbd: Fix use after free in rbd_open()

Commit a9ccedc3 frees the QemuOpts for the driver-specific options
immediately, even though it still needs the filename string that is
contained there. This doesn't work. Move the deletion of the QemuOpts to
the end of the function where its content isn't needed any more.

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

block: Disable driver-specific options for 1.5

We don't want to commit to the API yet before everything is worked out.
Disable it for the 1.5 release. This commit is meant to be reverted
after the 1.5 release.

The disabling of the driver-specific options is achieved by applying the
old checks while parsing the command line.

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

sheepdog: implement .bdrv_co_is_allocated()

Cc: MORITA Kazutaka <[email protected]>
Cc: Kevin Wolf <[email protected]>
Cc: Stefan Hajnoczi <[email protected]>
Signed-off-by: Liu Yuan <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

sheepdog: use BDRV_SECTOR_SIZE

Cc: MORITA Kazutaka <[email protected]>
Cc: Kevin Wolf <[email protected]>
Cc: Stefan Hajnoczi <[email protected]>
Signed-off-by: Liu Yuan <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

sheepdog: add discard/trim support for sheepdog

The 'TRIM' command from VM that is to release underlying data storage for
better thin-provision is already supported by the Sheepdog.

This patch adds the TRIM support at QEMU part.

For older Sheepdog that doesn't support it, we return 0(success) to upper layer.

Cc: MORITA Kazutaka <[email protected]>
Cc: Kevin Wolf <[email protected]>
Cc: Stefan Hajnoczi <[email protected]>
Cc: Paolo Bonzini <[email protected]>
Signed-off-by: Liu Yuan <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

block/ssh: Require libssh2 >= 1.2.8.

libssh2 >= 1.2.8 is required to enable this block device (because
that version introduced the libssh2_session_handshake call).

Change the test to use pkg-config exclusively.  If the user requests
--enable-libssh2 and the minimum version is not available, then the
following error is displayed:

  $ ./configure --enable-libssh2

  ERROR: libssh2 >= 1.2.8 required for --enable-libssh2

If --enable-libssh2 is not specified, then the feature is silently
disabled if sufficiently new libssh2 is not available.

Signed-off-by: Stefan Hajnoczi <[email protected]>

bsd-user: Track change in FreeBSD SYSCTL(9) types

Originally from Garrett Cooper in FreeBSD PR ports/155558
http://www.freebsd.org/cgi/query-pr.cgi?pr=155558

Signed-off-by: Ed Maste <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

virtio: Fix compilation without CONFIG_VHOST_SCSI

Signed-off-by: Ed Maste <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

qemu-doc: Option -ignore-environment removed.

Has been removed in commit fc9c54124d134dbd76338a92a91804dab2df8166.

Signed-off-by: Thomas Schwinge <[email protected]>
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

s390x: use CONFIG_INT128 to detect __uint128_t

Target s390x uses ad-hoc macro magic to guess if the compiler
supports the GCC extension __uint128_t.  This patch uses the
the dedicated macro CONFIG_INT128 defined by configure instead.

This fixes compilation with the CIL source code analyzer, which
uses GCC as a preprocessor but does not support __uint128_t.

Signed-off-by: Gabriel Kerneis <[email protected]>
Reviewed-by: Richard Henderson <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

linux-user: fix compile error due to stray colon at end of #ifdef line

Remove a stray colon from the end of a #ifdef line. Some versions
of gcc complain about this:
 linux-user/syscall.c: In function ‘do_syscall’:
 linux-user/syscall.c:7606:28: error: extra tokens at end of #ifdef directive [-Werror]

Signed-off-by: Peter Maydell <[email protected]>
Acked-By: Riku Voipio <[email protected]>
Signed-off-by: Stefan Hajnoczi <[email protected]>

microblaze: Add internal base vectors reg

Configurable at CPU synthesis/instantiation.

Reviewed-by: Peter Crosthwaite <[email protected]>
Signed-off-by: Edgar E. Iglesias <[email protected]>

Makefile: Use QEMU_FLAGS for DTC compilation

Build DTC as QEMU is built. Fixes the issue with mingw build which can't
handles DTC's default CFLAGS.

Signed-off-by: Peter Crosthwaite <[email protected]>
Signed-off-by: Edgar E. Iglesias <[email protected]>

Makefile: Don't build shared libfdt

The submodule DTC should just build and use DTC as static (the whole
motivation for using submod is lack of widespread distro support).

The .so causes build failures on some platforms, so adjust the make
target to just build the static lib.

Signed-off-by: Peter Crosthwaite <[email protected]>
Signed-off-by: Edgar E. Iglesias <[email protected]>

Merge remote-tracking branch 'luiz/queue/qmp' into staging

# By Amos Kong
# Via Luiz Capitulino
* luiz/queue/qmp:
  monitor: introduce query-command-line-options

Message-id: 1366922656[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

qemu-char: Set foo_tag = 0 when returning FALSE from callbacks

While reviewing some patches I found this problem where tcp_chr_accept
does not clear listen_tag when returning FALSE, leading to a double
g_source_remove of the underlying source. Not really a problem unless the id
gets re-used in between, but still something we should fix.

While at it I've also reviewed all the other code in qemu-char.c for
similar problems and found that pty_chr_timer has the same problem.

Cc: Anthony Liguori <[email protected]>
Signed-off-by: Hans de Goede <[email protected]>
Message-id: 1366890782[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

Remove unnecessary FreeBSD #include

sys/param.h was included to define __FreeBSD_version, but the conditional
using it was removed by commit d05ef160453e98546a4197496dc8a3cb2defac53
(Brad Smith, "Allow clock_gettime() monotonic clock to be utilized on more
OS's"), so the include is no longer needed here.

Signed-off-by: Ed Maste <[email protected]>
Reviewed-by: Peter Maydell <[email protected]>
Message-id: 1366906631[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

console: add dummy surface for guests without graphic card

So users get a notification instead of a crash.

Signed-off-by: Gerd Hoffmann <[email protected]>

console: zap ds arg from register_displaychangelistener

We don't have multiple DisplayStates any more,
so passing it in as argument is not needed.

Signed-off-by: Gerd Hoffmann <[email protected]>

console: switch ppm_save to qemu_open

... so it works with fdset.

Signed-off-by: Gerd Hoffmann <[email protected]>

console: add qemu_console_lookup_by_device

Look up the QemuConsole for a given device, using the new link.

Signed-off-by: Gerd Hoffmann <[email protected]>

console: add device link to QemuConsoles

So it is possible to figure which qemu console displays which device.

Signed-off-by: Gerd Hoffmann <[email protected]>

console: qom-ify QemuConsole

Just the minimal bits to turn QemuConsoles into Objects.

Signed-off-by: Gerd Hoffmann <[email protected]>

monitor: introduce query-command-line-options

Libvirt has no way to probe if an option or property is supported,
This patch introduces a new qmp command to query command line
option information. hmp command isn't added because it's not needed.

Signed-off-by: Amos Kong <[email protected]>
CC: Luiz Capitulino <[email protected]>
CC: Osier Yang <[email protected]>
CC: Anthony Liguori <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>
Reviewed-by: Eric Blake <[email protected]>

virtio-ccw: Check indicators location.

If a guest neglected to register (secondary) indicators but still runs
with notifications enabled, we might end up writing to guest zero;
avoid this by checking for valid indicators and only writing to the
guest and generating an interrupt if indicators have been setup.

Cc: [email protected]
Signed-off-by: Cornelia Huck <[email protected]>

pc: Kill the "use flash device for BIOS unless KVM" misfeature

Use of a flash memory device for the BIOS was added in series "[PATCH
v10 0/8] PC system flash support", commit 4732dca..1b89faf, v1.1.

Flash vs. ROM is a guest-visible difference.  Thus, flash use had to
be suppressed for machine types pc-1.0 and older.  This was
accomplished by adding a dummy device "pc-sysfw" with property
"rom_only":

* Non-zero rom_only means "use ROM".  Default for pc-1.0 and older.
* Zero rom_only means "maybe use flash".  Default for newer machines.

Not only is the dummy device ugly, it was also retroactively added to
the older machine types!  Fortunately, it's not guest-visible (thus no
immediate guest ABI breakage), and has no vmstate (thus no immediate
migration breakage).  Breakage occurs only if the user unwisely
enables flash by setting rom_only to zero.  Patch review FAIL #1.

Why "maybe use flash"?  Flash didn't (and still doesn't) work with
KVM.  Therefore, rom_only=0 really means "use flash, except when KVM
is enabled, use ROM".  This is a Bad Idea, because it makes enabling/
disabling KVM guest-visible.  Patch review FAIL #2.

Aside: it also precludes migrating between KVM on and off, but that's
not possible for other reasons anyway.

Fix as follows:

1. Change the meaning of rom_only=0 to mean "use flash, no ifs, buts,
or maybes" for pc-i440fx-1.5 and pc-q35-1.5.  Don't change anything
for older machines (to remain bug-compatible).

2. Change the default value from 0 to 1 for these machines.
Necessary, because 0 doesn't work with KVM.  Once it does, we can flip
the default back to 0.

3. Don't revert the retroactive addition of device "pc-sysfw" to older
machine types.  Seems not worth the trouble.

4. Add a TODO comment asking for device "pc-sysfw" to be dropped once
flash works with KVM.

Net effect is that you get a BIOS ROM again even when KVM is disabled,
just like for machines predating the introduction of flash.

To get flash instead, use "--global pc-sysfw.rom_only=0".

Signed-off-by: Markus Armbruster <[email protected]>
Message-id: 1365780303[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

pc: Split pc_init_pci_1_0() off pc_init_pci_1_2()

Just to make the next commit easier to review.

Signed-off-by: Markus Armbruster <[email protected]>
Message-id: 1365780303[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

pc: Inline pc_init_pci_1_3() into pc_init_pci_1_2()

Just to make the commit after next easier to review.

Signed-off-by: Markus Armbruster <[email protected]>
Message-id: 1365780303[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

Merge remote-tracking branch 'alon/libcacard_ccid.1' into staging

# By Alon Levy (15) and others
# Via Alon Levy
* alon/libcacard_ccid.1: (28 commits)
  libcacard/cac: change big switch functions to single return point
  dev-smartcard-reader: empty implementation for Mechanical (fail correctly)
  libcacard: move atr setting from macro to function
  libcacard/vreader: add debugging messages for apdu
  dev-smartcard-reader: copy atr protocol to ccid parameters
  dev-smartcard-reader: change default protocol to T=0
  dev-smartcard-reader: define structs for CCID_Parameter internals
  ccid-card-passthru, dev-smartcard-reader: add debug environment variables
  ccid-card-passthru: add atr check
  libcacard: change default ATR
  dev-smartcard-reader: reuse usb.h definitions
  dev-smartcard-reader: support windows guest
  dev-smartcard-reader: remove aborts (never triggered, but just in case)
  dev-smartcard-reader: nicer debug messages
  dev-smartcard-reader: white space fixes
  libcacard: remove default libcoolkey loading
  libcacard: remove sql: prefix
  libcacard: teach vscclient to use GMainLoop for portability
  libcacard: vscclient to use QemuThread for portability
  libcacard: split vscclient main() from socket reading
  ...

Message-id: 921423767.1475937.1366790789930[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

build: include config-{, all-}devices.mak after defining CONFIG_SOFTMMU and CONFIG_USER_ONLY

Moving the inclusions closer to Makefile, and before rules.mak, makes
Makefile and Makefile.target more consistent with each other.

Signed-off-by: Paolo Bonzini <[email protected]>
Tested-by: Markus Armbruster <[email protected]>
Message-id: 1366102238[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio: drop unused function prototypes.

This removes the unused prototypes in virtio.h.

Signed-off-by: KONRAD Frederic <[email protected]>
Message-id: 1366791683[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio: cleanup: init and exit function.

This clean the init and the exit functions and rename virtio_common_cleanup
to virtio_cleanup.

Signed-off-by: KONRAD Frederic <[email protected]>
Message-id: 1366791683[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio: remove virtiobindings.

This remove virtio-bindings, and use class instead.

Signed-off-by: KONRAD Frederic <[email protected]>
Message-id: 1366791683[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio: remove the function pointer.

This remove the function pointer in VirtIODevice, and use only
VirtioDeviceClass function pointer.

Signed-off-by: KONRAD Frederic <[email protected]>
Message-id: 1366791683[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio-pci: cleanup.

This remove the init, exit functions as they are no longer used.

Signed-off-by: KONRAD Frederic <[email protected]>
Message-id: 1366791683[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio-bus: make virtio_x_bus_new static.

virtio_x_bus_new are only used in file scope.
So this make them static.

Signed-off-by: KONRAD Frederic <[email protected]>
Message-id: 1366791683[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio-bus: add new functions.

This add two functions:
   * virtio_bus_set_vdev_config.
   * virtio_bus_set_vdev_feature.

Needed by virtio-ccw.

Signed-off-by: KONRAD Frederic <[email protected]>
Message-id: 1366791683[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio-rng: cleanup: use QOM casts.

As the virtio-rng-pci, virtio-rng-s390 and virtio-rng-ccw are
switched to the new API, we can use QOM casts.

Signed-off-by: KONRAD Frederic <[email protected]>
Reviewed-by: Andreas Färber <[email protected]>
Reviewed-by: Andreas Färber <[email protected]>
Tested-by: Cornelia Huck <[email protected]>
Acked-by: Amit Shah <[email protected]>
Message-id: 1366790881[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio-rng: cleanup: remove qdev field.

The qdev field is no longer needed, just drop it.

Signed-off-by: KONRAD Frederic <[email protected]>
Tested-by: Cornelia Huck <[email protected]>
Acked-by: Amit Shah <[email protected]>
Message-id: 1366790881[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio-rng: cleanup: init and exit functions.

This remove old init and exit function as they are no longer needed.

Signed-off-by: KONRAD Frederic <[email protected]>
Tested-by: Cornelia Huck <[email protected]>
Acked-by: Amit Shah <[email protected]>
Message-id: 1366790881[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio-rng-ccw: switch to the new API.

Here the virtio-rng-ccw is modified for the new API. The device
virtio-rng-pci extends virtio-pci. It creates and connects a virtio-rng-device
during the init. The properties are not changed.

Signed-off-by: KONRAD Frederic <[email protected]>
Tested-by: Cornelia Huck <[email protected]>
Acked-by: Amit Shah <[email protected]>
Message-id: 1366790881[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio-rng-s390: switch to the new API.

Here the virtio-rng-s390 is modified for the new API. The device
virtio-rng-s390 extends virtio-s390-device as before. It creates and
connects a virtio-rng during the init. The properties are not modified.

Signed-off-by: KONRAD Frederic <[email protected]>
Tested-by: Cornelia Huck <[email protected]>
Acked-by: Amit Shah <[email protected]>
Message-id: 1366790881[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio-rng-pci: switch to the new API.

Here the virtio-rng-pci is modified for the new API. The device
virtio-rng-pci extends virtio-pci. It creates and connects a virtio-rng-device
during the init. The properties are not changed.

The virtio_pci_reset function, is removed as no longer used.
The virtio_pci_rst function, is renamed virtio_pci_reset.

Signed-off-by: KONRAD Frederic <[email protected]>
Tested-by: Cornelia Huck <[email protected]>
Acked-by: Amit Shah <[email protected]>
Message-id: 1366790881[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio-rng: add virtio-rng device.

Create virtio-rng-device which extends virtio-device, so it can be connected on
virtio-bus.

Signed-off-by: KONRAD Frederic <[email protected]>
Tested-by: Cornelia Huck <[email protected]>
Acked-by: Amit Shah <[email protected]>
Message-id: 1366790881[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

virtio-rng: don't use pointer for configuration.

The configuration field must not be a pointer as it will be used for
virtio-rng properties. So *conf is replaced by conf.

Signed-off-by: KONRAD Frederic <[email protected]>
Tested-by: Cornelia Huck <[email protected]>
Acked-by: Amit Shah <[email protected]>
Message-id: 1366790881[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

input: introduce keyboard handler list

Add a linked list of keyboard handlers.  Added handlers will go
to the head of the list.  Removed handlers will be zapped from
the list.  The head of the list will be used for events.

This fixes the keyboard-dead-after-usb-kbd-unplug issue, key events
will be re-routed to the ps/2 kbd instead of being discarded.

[ v2: fix cut+paste bug found my Markus ]

Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 1366798118[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

input: make QEMUPutLEDEntry + QEMUPutMouseEntry private

There is no need for anybody outside ui/input.c to access the
struct elements.  Move the definitions, leaving only the typedefs
in the header files.

Signed-off-by: Gerd Hoffmann <[email protected]>
Message-id: 1366798118[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

ich9: kill cmos_s3

Signed-off-by: Hu Tao <[email protected]>
Reviewed-by: Paolo Bonzini <[email protected]>
Message-id: 1366799842[email protected]
Signed-off-by: Anthony Liguori <[email protected]>

libcacard/cac: change big switch functions to single return point

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

dev-smartcard-reader: empty implementation for Mechanical (fail correctly)

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

libcacard: move atr setting from macro to function

Only because qemu's checkpatch complains about it.

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

libcacard/vreader: add debugging messages for apdu

Using g_debug with log domain libcacard

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

dev-smartcard-reader: copy atr protocol to ccid parameters

Adds todos.

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

dev-smartcard-reader: change default protocol to T=0

We don't support T=1 so we shouldn't advertise it by default.

Two independent changes:
* Default ATR
 sets T=0. This gets overwritten by the client provided ATR later.
* Class descriptor
 changes dwAdvertise dwProtocols.PPPP to 0x1 and dwProtocols.RRRR=0 per spec.

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

dev-smartcard-reader: define structs for CCID_Parameter internals

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

ccid-card-passthru, dev-smartcard-reader: add debug environment variables

Introduces a new utility function: parse_debug_env to avoid code
duplication.

This overrides whatever debug value is set on the corresponding devices
from the command line, and is meant to ease the usage with any
management stack. For libvirt you can set environment variables by
extending the dom namespace, i.e:

<domain type='kvm' id='3' xmlns:qemu='http://libvirt.org/schemas/domain/qemu/1.0'>
  <qemu:commandline>
    <qemu:env name='QEMU_CCID_PASSTHRU_DEBUG' value='4'/>
    <qemu:env name='QEMU_CCID_DEBUG' value='4'/>
  </qemu:commandline>
</domain>

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

ccid-card-passthru: add atr check

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

libcacard: change default ATR

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

dev-smartcard-reader: reuse usb.h definitions

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

dev-smartcard-reader: support windows guest

By not advertising USB wakeup support (which we don't).

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

dev-smartcard-reader: remove aborts (never triggered, but just in case)

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

dev-smartcard-reader: nicer debug messages

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

dev-smartcard-reader: white space fixes

Signed-off-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

libcacard: remove default libcoolkey loading

Use only the modules defined in the NSS database.

Signed-off-by: Marc-André Lureau <[email protected]>
Reviewed-by: Alon Levy <[email protected]>

libcacard: remove sql: prefix

For some reason, with sql:/ prefix, the PKCS11 modules are not loaded.

This patch goes on top of Alon smartcard series.

Signed-off-by: Marc-André Lureau <[email protected]>
Reviewed-by: Alon Levy <[email protected]>

libcacard: teach vscclient to use GMainLoop for portability

This version handles non-blocking sending and receiving from the
socket.

Signed-off-by: Marc-André Lureau <[email protected]>
Reviewed-by: Alon Levy <[email protected]>

libcacard: vscclient to use QemuThread for portability

Signed-off-by: Marc-André Lureau <[email protected]>
Reviewed-by: Alon Levy <[email protected]>

libcacard: split vscclient main() from socket reading

Signed-off-by: Marc-André Lureau <[email protected]>
Reviewed-by: Alon Levy <[email protected]>

libcacard: fix mingw64 cross-compilation

Compile and link with version.lo

Signed-off-by: Marc-André Lureau <[email protected]>
Reviewed-by: Alon Levy <[email protected]>

build-sys: must link with -fstack-protector

It is needed to give that flag to the linker as well, but latest
libtool 2.4.2 still swallows that argument, so let's pass it with
libtool -Wc argument.

qemu-1.4.0/stubs/arch-query-cpu-def.c:6: undefined reference to `__stack_chk_guard'

Signed-off-by: Marc-André Lureau <[email protected]>
Reviewed-by: Alon Levy <[email protected]>

util: move socket_init() to osdep.c

vscclient needs to call socket_init() for portability.
Moving to osdep.c since it has no internal dependency.

Signed-off-by: Marc-André Lureau <[email protected]>
Reviewed-by: Alon Levy <[email protected]>

libcacard: use system config directory for nss db on win32

It's a bit nicer to look for default database under
CSIDL_COMMON_APPDATA\pki\nss rather that /etc/pki/nss.

Signed-off-by: Marc-André Lureau <[email protected]>
Reviewed-by: Alon Levy <[email protected]>

ccid: declare DEFAULT_ATR table to be "static const"

Signed-off-by: Jim Meyering <[email protected]>
Reviewed-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

ccid: make backend_enum_table "static const" and adjust users

Signed-off-by: Jim Meyering <[email protected]>
Reviewed-by: Alon Levy <[email protected]>
Reviewed-by: Marc-André Lureau <[email protected]>

ccid-card-emul: do not crash if backend is not provided

Program received signal SIGSEGV, Segmentation fault.
__strcmp_sse42 () at ../sysdeps/x86_64/multiarch/strcmp-sse42.S:164
164               movdqu        (%rsi), %xmm2
(gdb) bt
    at /home/elmarco/320g/src/qemu/hw/ccid-card-emulated.c:477
    at /home/elmarco/320g/src/qemu/hw/ccid-card-emulated.c:503

Signed-off-by: Marc-André Lureau <[email protected]>
Reviewed-by: Alon Levy <[email protected]>

libcacard: correct T0 historical bytes size

The VCARD_ATR_PREFIX macro adds a prefix of 6 characters only.

pcsc_scan was complaining before the patch:

+ Historical bytes: 56 43 41 52 44 5F 4E 53 53
 ERROR! ATR is truncated: 2 byte(s) is/are missing

Signed-off-by: Marc-André Lureau <[email protected]>
Reviewed-by: Alon Levy <[email protected]>