Fix out of tree build broken by 791e08c753a9f9be3c3880b4ea83b6dfa4b6ccad

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

Fix typo WIN32 -> _WIN32

This was spotted by Juan Quintela

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

set SEARCH_PATH for the linker script from output of ld --verbose -v

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

generate LDFLAGS for *-linux-user and *-bsd-user in a single place in configure

Remove lots of duplicate code in the process

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

our build system don't support mips little endian linux-user or bsd-user

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Don't build option roms on Mac OS X

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

multiboot.bin is a generated file

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

use cc-option for optionrom

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Use quiet-command for building optionrom

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

move cc-option definition to rules.mak

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Generate config-host.h from config-host.mak

Generate CONFIG_AUDIO_DRIVERS.  Order is important here, because the
first driver in the list is the one used by default.

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

rename USE_NPTL to CONFIG_USE_NPTL

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

rename NEEDS_LIBSUNMATH to CONFIG_NEEDS_LIBSUNMATH

Once there fix a place where it was misspelled

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

define ENOTSUP the same that the other errors

aliguori: ENOTSUP is not 4096 universally, only on OpenBSD

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

test if xen is enabled only once

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

simplify fdt libs selection

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

simplify vde libs selection

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

simplify brlapi selection

Use same style that everythnig else

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

simplify curses library selection

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Define and use xen libs in a single place

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

indent with 2 spaces sdl tests

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

remove sdl_static. Just do the right thing if static is yes

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

target_softmmu is undefined at that point

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

instead of setup sdl_x11 just add -lX11 to sdl_libs

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Calculate sdl_libs and sdl_flags and use them everywhere

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

sdl_config value was always sdl-config

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

the else part of this test is obsolete We are testing for sdl = yes inside the else part of the test if sdl = yes

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Use CONFIG_POSIX to simplify Makefile

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Add CONFIG_POSIX

We need a way to define posix-like (a.k.a. no win32)

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

rename WORDS_BIGENDIAN to HOST_WORDS_BIGENDIAN

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

rename DEBUG_EXEC to CONFIG_DEBUG_EXEC

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

rename DEBUG_TCG to CONFIG_DEBUG_TCG

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

rename HAVE_BYTESWAP_H to CONFIG_BYTESWAP_H

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

rename HAVE_GPROF to TARGET_GPROF

Use was not consistent, in Makefile was TARGET_GPROF and in *h HAVE_GPROF

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

rename HAVE_MACHINE_BSWAP_H to CONFIG_MACHINE_BSWAP_H

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

rename HAVE_FNMATCH_H to CONFIG_FNMATCH

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

rename HAVE_IOVEC to CONFIG_IOVEC

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

rename HAVE_PREADV to CONFIG_PREADV

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

define MAP_ANONYMOUS if it is not defined in qemu-common.h

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

rename HOST_BSD to CONFIG_BSD

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

O_LARGEFILE is already set in qemu-common.h

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

fix indententation

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Refactor code to remove one #ifdef CONFIG_FDT

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Rename HAVE_FDT to CONFIG_FDT and define it also in Makefile

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Add -static in configure if needed

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

more specific config.mak can overwrote more general config.mak

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

move printing of config-host.mak variables to end of generation

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

make oss use OSS_LIBS

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

make fmod also use FMOD_{LIBS,CFLAGS}

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

remove CONFIG_ from BLUEZ_{LIBS,FLAGS}

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

remove CONFIG_ from VNC_SASL_{LIBS, FLAGS}

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

use same algorithm for testing and

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

remove CONFIG_ from VNC_TLS_{LIBS, FLAGS}

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

target_arch2 is redefined unconditionally later

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Remove SP_CFLAGS and SP_LDFLAGS

SP_CFLAGS and SP_LDFLAGS are only used as initial values for ARCH_CFLAGS/ARCH_LDFLAGS.  Call it directly ARCH_*.  Once there, use the same indentantion that the rest of the file

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

change HOST_SOLARIS to CONFIG_SOLARIS{_VERSION}

Signed-off-by: Juan Quintela <quintela@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Enable USE_DIRECT_JUMP on Win32.

Signed-off-by: Filip Navara <filip.navara@gmail.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Remove special Win32 code in vl.c that's no longer needed.

Signed-off-by: Filip Navara <filip.navara@gmail.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Remove setvbuf(<handle>, NULL, _IOLBF, 0) calls for Win32

On Win32 the setvbuf function requires the last parameter to be size between 2 and INT_MAX bytes, so the calls always failed. Since the whole point of the calls is to set line-buffered mode for the file handle and that's not supported on Win32 anyway, conditionally remove them.

Signed-off-by: Filip Navara <filip.navara@gmail.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Remove useless Win32 include files and unused function in net.c.

Signed-off-by: Filip Navara <filip.navara@gmail.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

tap-win32: Use correct headers.

Replace the usage of DDK headers with the SDK counterpart "winioctl.h".

Signed-off-by: Filip Navara <filip.navara@gmail.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

fix broken migration

While fixing migration with -S, commit
89befdd1a6b18215153b8976682d57b7d03d5782 broke the rest of us. Poor
glommer, with a poor family, spare him his life from this monstruosity.

Since the unconditional vm_start, not autostart was the villain, I'm putting
back autostart. Let me know if you prefer other solutions, it doesn't really matter,
doesn't really matter to me.

Any way the wind blows...

Signed-off-by: Glauber Costa <glommer@redhat.com>
CC: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Fix broken build

The only caller of on_vcpu() is protected by ifdef
KVM_CAP_SET_GUEST_DEBUG, so protect on_vcpu() too otherwise QEMU
may not to build.

Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

slirp: Use monotonic clock if available (v2)

Calling gettimeofday() to compute a time interval can cause problems if
the system clock jumps forwards or backwards; replace updtime() with
qemu_get_clock(rt_clock), which calls clock_gettime(CLOCK_MONOTONIC) if
it is available.

Also remove some useless macros.

Signed-off-by: Ed Swierk <eswierk@aristanetworks.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

slirp: Remove UDP protocol emulation (talk, cuseeme)

The UDP emulation code for talk has been commented out since the
beginning of time, and unless someone who runs CU-SeeMe on qemu with
user-mode networking can vouch that the special magic (a) is necessary
and (b) works, let's get rid of the code.

Signed-off-by: Ed Swierk <eswierk@aristanetworks.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

qemu: msix nit: clear msix_entries_nr on error

I don't think it's critical to do this, but it's
best to keep uninit and error recovery consistent.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

QEMU BOCHS bios patches to use maxcpus value.

Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Introduce -smp , maxcpus= flag to specify maximum number of CPUS.

Follow on patch will use it to determine the size of the MADT and
other BIOS tables.

Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Use Little Endian for Dirty Log

We currently use host endian long types to store information
in the dirty bitmap.

This works reasonably well on Little Endian targets, because the
u32 after the first contains the next 32 bits. On Big Endian this
breaks completely though, forcing us to be inventive here.

So Ben suggested to always use Little Endian, which looks reasonable.

We only have dirty bitmap implemented in Little Endian targets so far
and since PowerPC would be the first Big Endian platform, we can just
as well switch to Little Endian always with little effort without
breaking existing targets.

This is the userspace part of the patch. It shouldn't change anything
for existing targets, but help PowerPC.

It replaces my older patch called "Use 64bit pointer for dirty log".

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Use 64bit pointer for dirty log

Dirty logs currently get written with native "long" size. On little endian
it doesn't matter if we use uint64_t instead though, because we'd still end
up using the right bytes.

On big endian, this does become a bigger problem, so we need to ensure that
kernel and userspace talk the same language, which means getting rid of "long"
and using a defined size instead.

So I decided to use 64 bit types at all times. This doesn't break existing
targets but will in conjunction with a patch I'll send to the KVM ML make
dirty logs work with 32 bit userspace on 64 kernel with big endian.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

HPET fixes for reg writes

This patch addresses the problems found by Andriy Gapon:

- The code was incorrectly overwriting the high order 32
  bits of the timer and hpet config registers. This didn't show up
  in testing because linux and windows use hpet in legacy mode,
  where the high order 32 bits (advertising available interrupts)
  of the timer config register are ignored, and the high order 32
  bits of the hpet config register are reserved and unused.

- The mask for level-triggered interrupts was off by a bit. (hpet
  doesn't currently support level-triggered interrupts).

In addition, I removed some unused #defines, and corrected the ioapic
interrupt values advertised. I'd set this up early in hpet development
and never went back to correct it, and no bugs resulted since linux and
windows use hpet in legacy mode where available interrupts are ignored.

Signed-off-by: Beth Kon <eak@us.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

slirp: Fix guestfwd for incoming data

Unless a virtual server address was explicitly defined (which is
impossible with the legacy -net channel format), guestfwd did not
properly forwarded host->guest packets. This patch fixes it.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

switch -drive to QemuOpts.

Demo QemuOpts in action ;)

Implementing a alternative way to specify the filename should be
just a few lines of code now once we decided how the cmd line syntax
should look like.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

QemuOpts: framework for storing and parsing options.

This stores device parameters in a better way than unparsed strings.

New types:
  QemuOpt       -  one key-value pair.
  QemuOpts      -  group of key-value pairs, belonging to one
                   device, i.e. one drive.
  QemuOptsList  -  list of some kind of devices, i.e. all drives.

Functions are provided to work with these types.  The plan is that some
day we will pass around QemuOpts pointers instead of strings filled with
"key1=value1,key2=value2".

Check out the next patch to see all this in action ;)

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

qemu-option: factor out parse_option_size

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

qemu-option: factor out parse_option_bool

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

move parser functions from vl.c to qemu-option.c

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

kill drives_opt

cleanup pretty simliar to the drives_table removal patch:
 - drop the table and make a linked list out of it.
 - pass around struct pointers instead of table indices.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

add support for drive ids.

-drive accepts the new id= now, allowing to explicitely name your
drives.  They will show up with that name in "info block" if specified,
otherwise the existing namimg scheme is used to autogenerate one.

There is also a new function to lookup drives by name.  Not used yet.
The plan is to link disk drivers and drives using the drive id instead
of passing around pointers to BlockDriveState.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

kill drives_table

First step cleaning up the drives handling.  This one does nothing but
removing drives_table[], still it became seriously big.

drive_get_index() is gone and is replaced by drives_get() which hands
out DriveInfo pointers instead of a table index.  This needs adaption in
*tons* of places all over.

The drives are now maintained as linked list.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

qdev/pci: hook up i440fx.

Hook i44fx pcihost into sysbus.
Convert Host bridge and ISA bridge pci devices to qdev.
Tag as no-user.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

qdev: add -device command line option.

The -device switch is the users frontend to the qdev_device_add function
added by the previous patch.

Also adds a linked list where command line options can be saved.
Use it for the new -device and for the -usbdevice and -bt switches.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

qdev: bus walker + qdev_device_add()

This patch implements a parser and qdev tree walker for bus paths and
adds qdev_device_add on top of this.

A bus path can be:
  (1) full path, i.e. /i440FX-pcihost/pci.0/lsi/scsi.0
  (2) bus name, i.e. "scsi.0".  Best used together with id= to make
      sure this is unique.
  (3) relative path starting with a bus name, i.e. "pci.0/lsi/scsi.0"

For the (common) case of a single child bus being attached to a device
it is enougth to specify the device only, i.e. "pci.0/lsi" will be
accepted too.

qdev_device_add() adds devices and accepts bus= parameters to find the
bus the device should be attached to.  Without bus= being specified it
takes the first bus it finds where the device can be attached to (i.e.
first pci bus for pci devices, ...).

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

qdev: create default bus names.

Create a default bus name if none is passed to qbus_create().

If the parent device has DeviceState->id set it will be used to create
the bus name,. i.e. -device lsi,id=foo will give you a scsi bus named
"foo.0".

If there is no id BusInfo->name (lowercased) will be used instead, i.e.
-device lsi will give you a scsi bus named "scsi.0".

A scsi adapter with two scsi busses would have "scsi.0" and "scsi.1" or
"$id.0" and "$id.1" busses.  The numbers of the child busses are per
device, i.e. when adding two lsi adapters both will have a "*.0" child
bus.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

qdev/pci: use qdev_prop_pci_devfn

Put the new property into use.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

qdev/prop: add pci devfn property

So we can parse "$slot.$fn" strings into devfn numbers.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Add a pc-0.11 machine type and make the pc type an alias

The pc-0.11 type allows users of qemu-0.11 to use a machine type which
they know will remain compatible when the upgrade to qemu-0.12.

Management tools may choose to canonicalize the 'pc' machine type to
'pc-0.11' so that if the 'pc' alias changes target in future versions
of qemu, the machine type used will remain compatible.

Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Add machine type aliases

Add an 'alias' field to QEMUMachine and display it in the output of
'qemu -M ?' with an '(aliased to foo)' suffix.

Aliases can change targets in newer versions of qemu, so management tools
may choose canonicalize machine types to ensure that if a user chooses an
alias, that the actual machine type used will remain compatible in
future.

This is intended to mimic a symlink to a machine description file.

Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Add support for fd=name to tap and socket networking

This allows a program to initialize a host networking device using a
file descriptor passed over a unix monitor socket.

The program must first pass the file descriptor using SCM_RIGHTS
ancillary data with the getfd monitor command. It then may do
"host_net_add tap fd=name" to use the named file descriptor.

Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Add monitor_get_fd() command for fetching named fds

Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Add getfd and closefd monitor commands

Add monitor commands to support passing file descriptors via
SCM_RIGHTS.

getfd assigns the passed file descriptor a name for use with other
monitor commands.

closefd allows passed file descriptors to be closed. If a monitor
command actually uses a named file descriptor, closefd will not be
required.

Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Add SCM_RIGHTS support to unix socket character devices

If a file descriptor is passed via a message with SCM_RIGHTS ancillary
data on a unix socket, store the file descriptor for use in the
chr_read() handler. Close the file descriptor if it was not used.

The qemu_chr_get_msgfd() API provides access to the passed descriptor.

Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Make tcp_chr_read() use recvmsg()

Split out tcp_chr_recv() out of tcp_chr_read() and implement it on
non-win32 using recvmsg(). This is needed for a subsequent patch
which implements SCM_RIGHTS support.

Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

sparc64 really implement itlb/dtlb automatic replacement writes

- implement "used" bit in tlb translation entry
- mark tlb entry used if qemu code/data translation succeeds
- fold i/d mmu replacement writes code into replace_tlb_1bit_lru which
adds 1bit lru replacement algorithm; previously code tried to replace
first unlocked entry only

- extract more bitmasks to named macros
- add "immu" or "dmmu" type name to debugging output where appropriate

Signed-off-by: igor.v.kovalenko@gmail.com
--
Kind regards,
Igor V. Kovalenko

sparc64 name mmu registers and general cleanup

- add names to mmu registers, this helps understanding the code which
uses/modifies them.
- fold i/d mmu tlb entries tag and tte arrays into arrays of tlb entries
- extract demap_tlb routine (code duplication)
- extract replace_tlb routine (code duplication)

- flush qemu tlb translations when replacing sparc64 mmu tlb entries

I have no test case which demands flushing qemu translations,
and this patch should have no other visible changes to runtime.

Signed-off-by: igor.v.kovalenko@gmail.com
--
Kind regards,
Igor V. Kovalenko

Give the kernel more room

My self-built PPC kernel doesn't fit in the region reserved for
the kernel, so I can't use -kernel with it.

Let's just extend the region.

Signed-off-by: Alexander Graf <agraf@suse.de>

PPC: Round VGA BIOS size to page boundary

When giving KVM a slot of a size not on page boundary, it chokes. So let's
just round up the VGA BIOS size so nobody complains anymore and we don't need
to implement sub-page slots.

Required for booting a PPC guest in KVM.

Signed-off-by: Alexander Graf <agraf@suse.de>

Fix period initialization

Signed-off-by: malc <av1474@comtv.ru>

provide tests for pit in kernel and irqchip in kernel

KVM can have an in-kernel pit or irqchip. While we don't implement it
yet, having a way for test for it (that always returns zero) will allow us
to reuse code in qemu-kvm that tests for it.

Signed-off-by: Glauber Costa <glommer@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

qemu-thread: use pthread_equal

Fixes

qemu-thread.c: In function `qemu_thread_equal':
qemu-thread.c:161: error: invalid operands to binary ==

Use of pthread_equal suggested by Filip Navara.

Signed-off-by: Sebastian Herbszt <herbszt@gmx.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>