sdk/emulator/qemu.git
12 years agoMerge remote-tracking branch 'aneesh/for-upstream' into staging
Anthony Liguori [Wed, 8 Feb 2012 13:23:55 +0000 (07:23 -0600)]
Merge remote-tracking branch 'aneesh/for-upstream' into staging

* aneesh/for-upstream:
  hw/9pfs: Remove O_NOATIME flag from 9pfs open() calls in readonly mode
  hw/9pfs: Update MAINTAINERS file
  fsdev: Fix parameter parsing for proxy helper
  hw/9pfs: Fix crash when mounting with synthfs
  hw/9pfs: Preserve S_ISGID
  hw/9pfs: Add new security model mapped-file.

12 years agoMerge remote-tracking branch 'sweil/w32' into staging
Anthony Liguori [Wed, 8 Feb 2012 13:23:23 +0000 (07:23 -0600)]
Merge remote-tracking branch 'sweil/w32' into staging

* sweil/w32:
  w32: Initialise critical section before starting thread (fix #922131)
  w32: Build windows and console executables

12 years agoRestore consistent formatting
malc [Tue, 7 Feb 2012 18:11:04 +0000 (22:11 +0400)]
Restore consistent formatting

Signed-off-by: malc <av1474@comtv.ru>
12 years agow32: Initialise critical section before starting thread (fix #922131)
Stefan Weil [Tue, 31 Jan 2012 06:14:15 +0000 (07:14 +0100)]
w32: Initialise critical section before starting thread (fix #922131)

This patch was contributed by Bogdan Harjoc. I added some assertions.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Weil <sw@weilnetz.de>
12 years agow32: Build windows and console executables
Stefan Weil [Thu, 22 Dec 2011 10:18:53 +0000 (11:18 +0100)]
w32: Build windows and console executables

System emulation executables with SDL are typically windows
executables. Sometimes console executables are more useful,
so create both variants if linker option -mwindows was detected.

v2:
This version uses QEMU_PROGW / QEMU_PROG instead of QEMU_PROG / QEMU_PROGC.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
12 years agos390x: fix qom-ification fall-out
Anthony Liguori [Mon, 6 Feb 2012 17:07:18 +0000 (11:07 -0600)]
s390x: fix qom-ification fall-out

Tested-by: Andreas Faerber <afaerber@suse.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoKVM: Fix breakages of QOM conversion
Jan Kiszka [Fri, 3 Feb 2012 21:06:47 +0000 (22:06 +0100)]
KVM: Fix breakages of QOM conversion

KVM APIC and PIC require instance sizes.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agousb-redir: convert to QOM
Anthony Liguori [Fri, 3 Feb 2012 23:39:57 +0000 (17:39 -0600)]
usb-redir: convert to QOM

This was missed due to the fact that it's in the top level and it uses
'struct DeviceInfo' instead of 'DeviceInfo' for some strange reason.

Tested-by: Benoît Canet <benoit.canet@gmail.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoFix previous commit
Blue Swirl [Sat, 4 Feb 2012 17:29:00 +0000 (17:29 +0000)]
Fix previous commit

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agousb-bsd: convert to QOM
Anthony Liguori [Fri, 3 Feb 2012 23:41:59 +0000 (17:41 -0600)]
usb-bsd: convert to QOM

Simple enough.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agovga: improve documentation
Blue Swirl [Sun, 29 Jan 2012 17:29:12 +0000 (17:29 +0000)]
vga: improve documentation

Add links to chipset docs and FreeVGA site.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agovga: use constants from vga.h
Blue Swirl [Sun, 29 Jan 2012 17:02:07 +0000 (17:02 +0000)]
vga: use constants from vga.h

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agovga.h: remove unused stuff and reformat
Blue Swirl [Sun, 29 Jan 2012 15:29:02 +0000 (15:29 +0000)]
vga.h: remove unused stuff and reformat

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoAdd vga.h unmodified from Linux
Blue Swirl [Sun, 29 Jan 2012 15:23:50 +0000 (15:23 +0000)]
Add vga.h unmodified from Linux

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agovga: move Cirrus VGA template to its own file
Blue Swirl [Wed, 25 Jan 2012 16:10:44 +0000 (16:10 +0000)]
vga: move Cirrus VGA template to its own file

Standard VGA does not use vga_draw_cursor_line_* functions.
Move the template to cirrus_vga_template.h.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agomemory: change dirty getting API to take a size
Blue Swirl [Sun, 22 Jan 2012 16:38:21 +0000 (16:38 +0000)]
memory: change dirty getting API to take a size

Instead of each device knowing or guessing the guest page size,
just pass the desired size of dirtied memory area.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoMerge branch 'linux-user-for-upstream' of git://git.linaro.org/people/rikuvoipio...
Blue Swirl [Sat, 4 Feb 2012 12:18:36 +0000 (12:18 +0000)]
Merge branch 'linux-user-for-upstream' of git://git.linaro.org/people/rikuvoipio/qemu

* 'linux-user-for-upstream' of git://git.linaro.org/people/rikuvoipio/qemu:
  linux-user: Fix sa_flags byte swaps for mips
  linux-user: Define TARGET_QEMU_ESIGRETURN for mips64
  linux-user: Define TARGET_QEMU_ESIGRETURN for mipsn32
  linux-user: Add default configs for mips64[el]
  linux-user: Add default-configs for mipsn32[el]
  linux-user: Implement *listxattr syscalls
  linux-user/syscall.c: Implement f and l versions of set/get/removexattr
  linux-user: Allow NULL value pointer in setxattr and getxattr
  linux-user: fix wait* syscall status returns
  linux-user/strace.c: Correct errno printing for mmap etc
  linux-user: fix QEMU_STRACE=1 segfault
  linux-user: add SO_PEERCRED support for getsockopt
  linux-user/main.c: Add option to user-mode emulation so that user can specify log file name
  linux-user: fake /proc/self/auxv
  linux-user: fake /proc/self/stat
  linux-user: fake /proc/self/maps
  linux-user: add open() hijack infrastructure
  linux-user: save auxv length
  linux-user: stack_base is now mandatory on all targets

12 years agow32: Fix build with new net bridge code
Stefan Weil [Sat, 4 Feb 2012 08:24:46 +0000 (09:24 +0100)]
w32: Fix build with new net bridge code

Commit a7c36ee4920ea3acc227a0248dd161693f207357 added code for a net
bridge and explicitly said that "this is very Linux centric".

Indeed, compilation failed for w32, so the bridge code is now
conditional. Hosts which don't support it can simply remove the
definition of CONFIG_NET_BRIDGE.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
12 years agoMerge branch 'for-upstream' of git://repo.or.cz/qemu/agraf
Blue Swirl [Sat, 4 Feb 2012 09:37:55 +0000 (09:37 +0000)]
Merge branch 'for-upstream' of git://repo.or.cz/qemu/agraf

* 'for-upstream' of git://repo.or.cz/qemu/agraf: (21 commits)
  PPC: E500: Populate L1CFG0 SPR
  PPC: e500mc: Enable processor control
  PPC: E500: Implement msgsnd
  PPC: E500: Implement msgclr
  PPC: Enable doorbell excp handlers
  PPC: Add CPU feature for processor control
  PPC: E500: Add doorbell defines
  PPC: E500: Add some more excp vectors
  KVM: Fix compilation on non-x86
  PPC: booke206: move avail check to tlbwe
  PPC: booke206: Check for TLB overrun
  PPC: booke206: Implement tlbilx
  PPC: booke206: Check for min/max TLB entry size
  PPC: booke: add tlbnps handling
  PPC: booke206: allow NULL raddr in ppcmas_tlb_check
  PPC: rename msync to msync_4xx
  PPC: e500: msync is 440 only, e500 has real sync
  PPC: e500mc: add missing IVORs to bitmap
  PPC: Add IVOR 38-42
  PPC: KVM: Update HIOR code to new interface
  ...

12 years agoFix build breakage from last commit.
Anthony Liguori [Fri, 3 Feb 2012 17:22:35 +0000 (11:22 -0600)]
Fix build breakage from last commit.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agos390x: fix -drive in the absence of aliases
Anthony Liguori [Fri, 3 Feb 2012 17:13:30 +0000 (11:13 -0600)]
s390x: fix -drive in the absence of aliases

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agocontainer: make a decendent of Object
Anthony Liguori [Fri, 23 Dec 2011 15:08:05 +0000 (09:08 -0600)]
container: make a decendent of Object

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
---
v1 -> v2
 - Add license (Paolo)

12 years agoobject: sure up reference counting
Anthony Liguori [Fri, 23 Dec 2011 14:47:39 +0000 (08:47 -0600)]
object: sure up reference counting

Now we have the following behavior:

1) object_new() returns an object with ref = 1
2) object_initialize() does not increase the reference count (ref may be 0).
3) object_deref() will finalize the object when ref = 0.  it does not free the
   memory associated with the object.
4) both link and child properties correctly set the reference count.

The expected usage is the following:

1) child devices should generally be created via object_initialize() using
   memory from the parent device.  Adding the object as a child property will
   take ownership of the object and tie the child's life cycle to the parent.

2) If a child device is created via qdev_create() or some other form of
   object_new(), there must be an object_delete() call in the parent device's
   finalize function.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoinfo qdm: do not require a parent_bus to be set
Anthony Liguori [Fri, 23 Dec 2011 14:16:58 +0000 (08:16 -0600)]
info qdm: do not require a parent_bus to be set

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqdev: implement cleanup logic in finalize
Anthony Liguori [Fri, 23 Dec 2011 14:38:56 +0000 (08:38 -0600)]
qdev: implement cleanup logic in finalize

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqom: accept any compatible type when setting a link property
Anthony Liguori [Fri, 23 Dec 2011 14:35:43 +0000 (08:35 -0600)]
qom: accept any compatible type when setting a link property

Links had limited utility before as they only allowed a concrete type to be
specified.  Now we can support abstract types and interfaces which means it's
now possible to have a link<PCIDevice>.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqom: move properties from qdev to object
Anthony Liguori [Mon, 30 Jan 2012 14:55:55 +0000 (08:55 -0600)]
qom: move properties from qdev to object

This is mostly code movement although not entirely.  This makes properties part
of the Object base class which means that we can now start using Object in a
meaningful way outside of qdev.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqdev: nuke qdev_init_chardev()
Anthony Liguori [Thu, 22 Dec 2011 21:29:25 +0000 (15:29 -0600)]
qdev: nuke qdev_init_chardev()

I'm sure the intentions were good here, but there's no reason this should be in
qdev.  Move it to qemu-char where it belongs.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqdev: split out UI portions into a new function
Anthony Liguori [Thu, 22 Dec 2011 21:24:20 +0000 (15:24 -0600)]
qdev: split out UI portions into a new function

qdev-monitor.c deals with the -device, device_add, and info qdm/qtree
interfaces.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqdev: refactor away qdev_create_from_info
Anthony Liguori [Thu, 22 Dec 2011 21:14:27 +0000 (15:14 -0600)]
qdev: refactor away qdev_create_from_info

Note that the FIXME gets fixed in series 4/4.  We need to convert BusState to
QOM before we can make parent_bus a link.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqdev: split out common init to instance_init
Anthony Liguori [Thu, 22 Dec 2011 21:06:37 +0000 (15:06 -0600)]
qdev: split out common init to instance_init

This gets us closer to being able to object_new() a qdev type and have a
functioning object verses having to call qdev_create().

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqom: add new command to search for types
Anthony Liguori [Thu, 22 Dec 2011 20:40:54 +0000 (14:40 -0600)]
qom: add new command to search for types

This adds a command that allows searching for types that implement a property.
This allows you to do things like search for all available PCIDevices.  In the
future, we'll also have a standard interface for things with a BlockDriverState
property that a PCIDevice could implement.

This will enable search queries like, "any type that implements the BlockDevice
interface" which would allow management tools to present available block devices
without having to hard code device names.  Since an object can implement
multiple interfaces, one device could act both as a BlockDevice and a
NetworkDevice.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqdev: remove baked in notion of aliases (v2)
Anthony Liguori [Thu, 22 Dec 2011 17:05:00 +0000 (11:05 -0600)]
qdev: remove baked in notion of aliases (v2)

Limit them to the device_add functionality.  Device aliases were a hack based
on the fact that virtio was modeled the wrong way.  The mechanism for aliasing
is very limited in that only one alias can exist for any device.

We have to support it for the purposes of compatibility but we only need to
support it in device_add so restrict it to that piece of code.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
---
v1 -> v2
 - Use a table for aliases (Paolo)

12 years agoqdev: kill off DeviceInfo
Anthony Liguori [Fri, 9 Dec 2011 21:02:56 +0000 (15:02 -0600)]
qdev: kill off DeviceInfo

It is no longer used in the tree since everything is done natively through
QEMU Object Model.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqdev: register all types natively through QEMU Object Model
Anthony Liguori [Thu, 8 Dec 2011 03:34:16 +0000 (21:34 -0600)]
qdev: register all types natively through QEMU Object Model

This was done in a mostly automated fashion.  I did it in three steps and then
rebased it into a single step which avoids repeatedly touching every file in
the tree.

The first step was a sed-based addition of the parent type to the subclass
registration functions.

The second step was another sed-based removal of subclass registration functions
while also adding virtual functions from the base class into a class_init
function as appropriate.

Finally, a python script was used to convert the DeviceInfo structures and
qdev_register_subclass functions to TypeInfo structures, class_init functions,
and type_register_static calls.

We are almost fully converted to QOM after this commit.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqdev: kill off DeviceInfo list
Anthony Liguori [Wed, 1 Feb 2012 15:34:28 +0000 (09:34 -0600)]
qdev: kill off DeviceInfo list

Teach the various bits of code that need to walk through available devices to
do so via QOM.

Signed-off-by: Anthony Liguori
12 years agoqom: allow object_class_foreach to take additional parameters to refine search
Anthony Liguori [Thu, 22 Dec 2011 20:11:53 +0000 (14:11 -0600)]
qom: allow object_class_foreach to take additional parameters to refine search

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqdev: refactor device creation to allow bus_info to be set only in class
Anthony Liguori [Fri, 9 Dec 2011 18:08:01 +0000 (12:08 -0600)]
qdev: refactor device creation to allow bus_info to be set only in class

As we use class_init to set class members, DeviceInfo no longer holds this
information.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqdev: allow classes to overload qdev functions
Anthony Liguori [Fri, 9 Dec 2011 17:22:58 +0000 (11:22 -0600)]
qdev: allow classes to overload qdev functions

This allows us to drop per-Device registration functions by allowing the
class_init functions to overload qdev methods.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqdev: remove info from class
Anthony Liguori [Fri, 9 Dec 2011 17:06:57 +0000 (11:06 -0600)]
qdev: remove info from class

Now DeviceInfo is no longer used after object construction.  All of the
relevant members have been moved to DeviceClass.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoqdev: make DeviceInfo private
Anthony Liguori [Fri, 9 Dec 2011 16:51:49 +0000 (10:51 -0600)]
qdev: make DeviceInfo private

Introduce accessors and remove any code that directly accesses DeviceInfo
members.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agousb: separate out legacy usb registration from type registration
Anthony Liguori [Thu, 8 Dec 2011 20:56:53 +0000 (14:56 -0600)]
usb: separate out legacy usb registration from type registration

Type registeration is going to get turned into a QOM call so decouple the
legacy support.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agousb-hid: simplify class initialization a bit
Anthony Liguori [Sun, 4 Dec 2011 22:13:14 +0000 (16:13 -0600)]
usb-hid: simplify class initialization a bit

We can probably model USBHidDevice as a base class to get even better code
sharing but for now, just use a common function to initialize the common class
members.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agolinux-user: Fix sa_flags byte swaps for mips
Andreas Färber [Thu, 29 Dec 2011 15:55:43 +0000 (16:55 +0100)]
linux-user: Fix sa_flags byte swaps for mips

sa_flags is uint32_t for mips{,n32,64}, so don't use tswapal().

edited by Riku Voipio: likewise on alpha

Reported-by: Khansa Butt <khansa@kics.edu.pk>
Suggested-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Cc: Ehsan Ul Haq <ehsan.ulhaq@kics.edu.pk>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: Define TARGET_QEMU_ESIGRETURN for mips64
Andreas Färber [Thu, 29 Dec 2011 15:55:42 +0000 (16:55 +0100)]
linux-user: Define TARGET_QEMU_ESIGRETURN for mips64

Copied from mips/syscall.h.

Signed-off-by: Khansa Butt <khansa@kics.edu.pk>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: Define TARGET_QEMU_ESIGRETURN for mipsn32
Andreas Färber [Thu, 29 Dec 2011 15:55:41 +0000 (16:55 +0100)]
linux-user: Define TARGET_QEMU_ESIGRETURN for mipsn32

Copied from mips/syscall.h.

Signed-off-by: Ulrich Hecht <uli@suse.de>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: Add default configs for mips64[el]
Andreas Färber [Thu, 29 Dec 2011 15:55:40 +0000 (16:55 +0100)]
linux-user: Add default configs for mips64[el]

Prepares for mips64[el]-linux-user targets.

Signed-off-by: Khansa Butt <khansa@kics.edu.pk>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: Add default-configs for mipsn32[el]
Andreas Färber [Thu, 29 Dec 2011 15:55:39 +0000 (16:55 +0100)]
linux-user: Add default-configs for mipsn32[el]

Prepares for mipsn32[el]-linux-user targets.

Signed-off-by: Ulricht Hecht <uli@suse.de>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: Implement *listxattr syscalls
Peter Maydell [Wed, 14 Dec 2011 15:37:19 +0000 (15:37 +0000)]
linux-user: Implement *listxattr syscalls

Implement listxattr, flistxattr and llistxattr syscalls.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user/syscall.c: Implement f and l versions of set/get/removexattr
Peter Maydell [Wed, 14 Dec 2011 15:37:18 +0000 (15:37 +0000)]
linux-user/syscall.c: Implement f and l versions of set/get/removexattr

Implement the f and l versions (operate on fd, don't follow links)
of the setxattr, getxattr and removexattr syscalls.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: Allow NULL value pointer in setxattr and getxattr
Peter Maydell [Wed, 14 Dec 2011 15:37:17 +0000 (15:37 +0000)]
linux-user: Allow NULL value pointer in setxattr and getxattr

It's valid to pass a NULL value pointer to setxattr, so don't
fail this case EFAULT.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: fix wait* syscall status returns
Alexander Graf [Wed, 23 Nov 2011 23:44:43 +0000 (00:44 +0100)]
linux-user: fix wait* syscall status returns

When calling wait4 or waitpid with a status pointer and WNOHANG, the
syscall can potentially not modify the status pointer input. Now if we
have guest code like:

  int status = 0;
  waitpid(pid, &status, WNOHANG);
  if (status)
     <breakage>

then we have to make sure that in case status did not change we actually
return the guest's initialized status variable instead of our own uninitialized.
We fail to do so today, as we proxy everything through an uninitialized status
variable which for me ended up always containing the last error code.

This patch fixes some test cases when building yast2-core in OBS for ARM.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user/strace.c: Correct errno printing for mmap etc
Peter Maydell [Mon, 21 Nov 2011 12:21:19 +0000 (12:21 +0000)]
linux-user/strace.c: Correct errno printing for mmap etc

Correct the printing of errnos for syscalls which are handled
via print_syscall_ret_addr (mmap, mmap2, brk, shmat): errnos
are returned as negative returned values at this level, not
via the host 'errno' variable.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: fix QEMU_STRACE=1 segfault
Alexander Graf [Mon, 21 Nov 2011 11:04:07 +0000 (12:04 +0100)]
linux-user: fix QEMU_STRACE=1 segfault

While debugging some issues with QEMU_STRACE I stumbled over segmentation
faults that were pretty reproducible. Turns out we tried to treat a
normal return value as errno, resulting in an access over array boundaries
for the resolution.

Fix this by allowing failure to resolve invalid errnos into strings.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: add SO_PEERCRED support for getsockopt
Akos PASZTORY [Mon, 14 Nov 2011 13:09:49 +0000 (15:09 +0200)]
linux-user: add SO_PEERCRED support for getsockopt

Signed-off-by: Akos PASZTORY <akos.pasztory@gmail.com>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user/main.c: Add option to user-mode emulation so that user can specify log...
陳韋任 [Tue, 8 Nov 2011 09:46:44 +0000 (17:46 +0800)]
linux-user/main.c: Add option to user-mode emulation so that user can specify log file name

  QEMU linux user-mode's default log file name is "/tmp/qemu.log". In order to
change the log file name, user need to modify the source code then recompile
QEMU. This patch allow user use "-D logfile" option to specify the log file
name.

Signed-off-by: Chen Wen-Ren <chenwj@iis.sinica.edu.tw>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: fake /proc/self/auxv
Alexander Graf [Wed, 2 Nov 2011 19:23:26 +0000 (20:23 +0100)]
linux-user: fake /proc/self/auxv

Gtk tries to read /proc/self/auxv to find its auxv table instead of
taking it from its own program memory space.

However, when running with linux-user, we see the host's auxv which
clearly exposes wrong information. so let's instead expose the guest
memory backed auxv tables via /proc/self/auxv as well.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: fake /proc/self/stat
Alexander Graf [Wed, 2 Nov 2011 19:23:25 +0000 (20:23 +0100)]
linux-user: fake /proc/self/stat

The boehm gc finds the program's stack starting pointer by
checking /proc/self/stat. Unfortunately, so far it reads
qemu's stack pointer which clearly is wrong.

So let's instead fake the file so the guest program sees the
right address.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: fake /proc/self/maps
Alexander Graf [Wed, 2 Nov 2011 19:23:24 +0000 (20:23 +0100)]
linux-user: fake /proc/self/maps

glibc's pthread_attr_getstack tries to find the stack range from
/proc/self/maps. Unfortunately, /proc is usually the host's /proc
which means linux-user guests see qemu's stack there.

Fake the file with a constructed maps entry that exposes the guest's
stack range.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: add open() hijack infrastructure
Alexander Graf [Wed, 2 Nov 2011 19:23:23 +0000 (20:23 +0100)]
linux-user: add open() hijack infrastructure

There are a number of files in /proc that expose host information
to the guest program. This patch adds infrastructure to override
the open() syscall for guest programs to enable us to on the fly
generate guest sensible files.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agolinux-user: save auxv length
Alexander Graf [Sat, 28 Jan 2012 19:12:14 +0000 (21:12 +0200)]
linux-user: save auxv length

We create our own AUXV segment on stack and save a pointer to it.
However we don't save the length of it, so any code that wants to
do anything useful with it later on has to walk it again.

Instead, let's remember the length of our AUXV segment. This
simplifies later uses by a lot.

(edited by Riku to apply to qemu HEAD)

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
12 years agolinux-user: stack_base is now mandatory on all targets
Riku Voipio [Sat, 28 Jan 2012 20:00:17 +0000 (22:00 +0200)]
linux-user: stack_base is now mandatory on all targets

Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
12 years agoPPC: E500: Populate L1CFG0 SPR
Alexander Graf [Tue, 31 Jan 2012 02:46:55 +0000 (03:46 +0100)]
PPC: E500: Populate L1CFG0 SPR

When running Linux on e500 with powersave-nap enabled, Linux tries to
read out the L1CFG0 register and calculates some things from it. Passing
0 there ends up in a division by 0, resulting in -1, resulting in badness.

So let's populate the L1CFG0 register with reasonable defaults. That way
guests aren't completely confused.

Reported-by: Shrijeet Mukherjee <shm@cumulusnetworks.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: e500mc: Enable processor control
Alexander Graf [Tue, 31 Jan 2012 02:20:31 +0000 (03:20 +0100)]
PPC: e500mc: Enable processor control

The e500mc implements Embedded.Processor Control, so enable it and
thus enable guests to IPI each other. This makes -smp work with -cpu
e500mc.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: E500: Implement msgsnd
Alexander Graf [Tue, 31 Jan 2012 02:19:23 +0000 (03:19 +0100)]
PPC: E500: Implement msgsnd

This patch implements the msgsnd instruction. It is part of the
Embedded.Processor Control specification and allows one CPU to
IPI another CPU without going through an interrupt controller.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: E500: Implement msgclr
Alexander Graf [Tue, 31 Jan 2012 02:18:35 +0000 (03:18 +0100)]
PPC: E500: Implement msgclr

This patch implements the msgclr instruction. It is part of the
Embedded.Processor Control specification and clears pending doorbell
interrupts on the current CPU.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: Enable doorbell excp handlers
Alexander Graf [Tue, 31 Jan 2012 02:17:24 +0000 (03:17 +0100)]
PPC: Enable doorbell excp handlers

We already had all the code available to have doorbell exceptions
be handled properly. It was just disabled.

Enable it, so we can rely on it.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: Add CPU feature for processor control
Alexander Graf [Tue, 31 Jan 2012 02:13:30 +0000 (03:13 +0100)]
PPC: Add CPU feature for processor control

We're soon going to implement processor control features. Add the
feature flag, so we're well prepared.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: E500: Add doorbell defines
Alexander Graf [Tue, 31 Jan 2012 02:11:32 +0000 (03:11 +0100)]
PPC: E500: Add doorbell defines

We're going to introduce doorbell instructions (called processor
control in the spec) soon. Add some defines for easier patch
readability later.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: E500: Add some more excp vectors
Alexander Graf [Tue, 31 Jan 2012 02:09:58 +0000 (03:09 +0100)]
PPC: E500: Add some more excp vectors

Our EXCP list is getting outdated. By now, 3 new exception vectors have
been introduced. Update the list so we have everything at one place.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoKVM: Fix compilation on non-x86
Alexander Graf [Wed, 25 Jan 2012 17:28:05 +0000 (18:28 +0100)]
KVM: Fix compilation on non-x86

Commit 84b058d broke compilation for KVM on non-x86 targets, which
don't have KVM_CAP_IRQ_ROUTING defined.

Fix by not using the unavailable constant when it's not around.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: booke206: move avail check to tlbwe
Alexander Graf [Wed, 25 Jan 2012 16:06:30 +0000 (17:06 +0100)]
PPC: booke206: move avail check to tlbwe

We can have TLBs that only support a single page size. This is defined
by the absence of the AVAIL flag in TLBnCFG. If this is the case, we
currently write invalid size info into the TLB, but override it on
internal fault.

Let's move the check over to tlbwe, so we don't have the AVAIL check in
the hotter fault path.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: booke206: Check for TLB overrun
Alexander Graf [Wed, 25 Jan 2012 15:27:26 +0000 (16:27 +0100)]
PPC: booke206: Check for TLB overrun

Our internal helpers to fetch TLB entries were not able to tell us
that an entry doesn't even exist. Pass an error out if we hit such
a case to not accidently pass beyond the TLB array.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: booke206: Implement tlbilx
Alexander Graf [Fri, 20 Jan 2012 03:09:15 +0000 (04:09 +0100)]
PPC: booke206: Implement tlbilx

The PowerPC 2.06 BookE ISA defines an opcode called "tlbilx" which is used
to flush TLB entries. It's the recommended way of flushing in virtualized
environments.

So far we got away without implementing it, but Linux for e500mc uses this
instruction, so we better add it :).

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: booke206: Check for min/max TLB entry size
Alexander Graf [Fri, 20 Jan 2012 03:07:51 +0000 (04:07 +0100)]
PPC: booke206: Check for min/max TLB entry size

When setting a TLB entry, we need to check if the TLB we're putting it in
actually supports the given size. According to the 2.06 PowerPC ISA, a
value that's out of range can either be redefined to something implementation
dependent or we can raise an illegal opcode exception. We do the latter.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: booke: add tlbnps handling
Alexander Graf [Sat, 21 Jan 2012 03:45:46 +0000 (04:45 +0100)]
PPC: booke: add tlbnps handling

When using MAV 2.0 TLB registers, we have another range of TLB registers
available to read the supported page sizes from.

Add SPR definitions for those and add a helper function that we can use
to receive such a bitmap even when using MAV 1.0.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: booke206: allow NULL raddr in ppcmas_tlb_check
Alexander Graf [Fri, 20 Jan 2012 03:06:18 +0000 (04:06 +0100)]
PPC: booke206: allow NULL raddr in ppcmas_tlb_check

We might want to call the tlb check function without actually caring about
the real address resolution. Check if we really should write the value
back.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: rename msync to msync_4xx
Alexander Graf [Sat, 21 Jan 2012 02:48:58 +0000 (03:48 +0100)]
PPC: rename msync to msync_4xx

The msync instruction as defined today is only valid on 4xx cores, not
on e500 which also supports msync, but treats it the same way as sync.

Rename it to reflect that it's 4xx only.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: e500: msync is 440 only, e500 has real sync
Alexander Graf [Thu, 19 Jan 2012 18:51:50 +0000 (19:51 +0100)]
PPC: e500: msync is 440 only, e500 has real sync

The e500 CPUs don't use 440's msync which falls on the same opcode IDs,
but instead use the real powerpc sync instruction. This is important,
since the invalid mask differs between the two.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: e500mc: add missing IVORs to bitmap
Alexander Graf [Thu, 19 Jan 2012 17:57:26 +0000 (18:57 +0100)]
PPC: e500mc: add missing IVORs to bitmap

E500mc supports IVORs 36-41. Add them to the support mask. Drop SPE
support too.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoPPC: Add IVOR 38-42
Alexander Graf [Thu, 19 Jan 2012 18:31:51 +0000 (19:31 +0100)]
PPC: Add IVOR 38-42

Our code only knows IVORs up to 37. Add the new ones defined in ISA 2.06
from 38 - 42.

Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Andreas Färber <afaerber@suse.de>
12 years agoPPC: KVM: Update HIOR code to new interface
Alexander Graf [Fri, 6 Jan 2012 03:02:24 +0000 (04:02 +0100)]
PPC: KVM: Update HIOR code to new interface

Unfortunately the HIOR setting code slipped into upstream QEMU
before it was pulled into upstream KVM. And since Murphy is always
right, comments on the patches only emerged on the pull request
leading to changes in the interface.

So here's an update to the HIOR setting. While at it, I also relaxed
it a bit since for HV KVM we can already run fine without and 3.2
works just fine with HV KVM but when not setting HIOR. We will only
need this when running PAPR in PR KVM.

Since we accidently changed the ABI and API along the way, we have
to update the underlying kernel headers together with the code that
uses it to not break bisectability.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoKVM: Update headers (except HIOR mess)
Alexander Graf [Fri, 20 Jan 2012 13:41:12 +0000 (14:41 +0100)]
KVM: Update headers (except HIOR mess)

This patch is basically what ./scripts/update-linux-headers.sh against
upstream KVM's next branch outputs except that all the HIOR bits are
removed. These we have to update with the code that uses them.

Signed-off-by: Alexander Graf <agraf@suse.de>
12 years agoChange license from GPLv2 to GPLv2+
Stefan Weil [Fri, 27 Jan 2012 16:40:43 +0000 (17:40 +0100)]
Change license from GPLv2 to GPLv2+

This file only contains code from Red Hat, so it can use GPLv2+.
Tested with `git blame -M -C net/checksum.c`.

Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoAdd support for net bridge
Corey Bryant [Thu, 26 Jan 2012 14:42:27 +0000 (09:42 -0500)]
Add support for net bridge

The most common use of -net tap is to connect a tap device to a bridge.  This
requires the use of a script and running qemu as root in order to allocate a
tap device to pass to the script.

This model is great for portability and flexibility but it's incredibly
difficult to eliminate the need to run qemu as root.  The only really viable
mechanism is to use tunctl to create a tap device, attach it to a bridge as
root, and then hand that tap device to qemu.  The problem with this mechanism
is that it requires administrator intervention whenever a user wants to create
a guest.

By essentially writing a helper that implements the most common qemu-ifup
script that can be safely given cap_net_admin, we can dramatically simplify
things for non-privileged users.  We still support existing -net tap options
as a mechanism for advanced users and backwards compatibility.

Currently, this is very Linux centric but there's really no reason why it
couldn't be extended for other Unixes.

A typical invocation would be similar to one of the following:

  qemu linux.img -net bridge -net nic,model=virtio

  qemu linux.img -net tap,helper="/usr/local/libexec/qemu-bridge-helper"
                 -net nic,model=virtio

  qemu linux.img -netdev bridge,id=hn0
                 -device virtio-net-pci,netdev=hn0,id=nic1

  qemu linux.img -netdev tap,helper="/usr/local/libexec/qemu-bridge-helper",id=hn0
                 -device virtio-net-pci,netdev=hn0,id=nic1

The default bridge that we attach to is br0.  The thinking is that a distro
could preconfigure such an interface to allow out-of-the-box bridged networking.

Alternatively, if a user wants to use a different bridge, a typical invocation
would be simliar to one of the following:

  qemu linux.img -net bridge,br=qemubr0 -net nic,model=virtio

  qemu linux.img -net tap,helper="/usr/local/libexec/qemu-bridge-helper --br=qemubr0"
                 -net nic,model=virtio

  qemu linux.img -netdev bridge,br=qemubr0,id=hn0
                 -device virtio-net-pci,netdev=hn0,id=nic1

  qemu linux.img -netdev tap,helper="/usr/local/libexec/qemu-bridge-helper --br=qemubr0",id=hn0
                 -device virtio-net-pci,netdev=hn0,id=nic1

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Richa Marwaha <rmarwah@linux.vnet.ibm.com>
Signed-off-by: Corey Bryant <coreyb@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoAdd cap reduction support to enable use as SUID
Corey Bryant [Thu, 26 Jan 2012 14:42:26 +0000 (09:42 -0500)]
Add cap reduction support to enable use as SUID

The ideal way to use qemu-bridge-helper is to give it an fscap of using:

 setcap cap_net_admin=ep qemu-bridge-helper

Unfortunately, most distros still do not have a mechanism to package files
with fscaps applied.  This means they'll have to SUID the qemu-bridge-helper
binary.

To improve security, use libcap to reduce our capability set to just
cap_net_admin, then reduce privileges down to the calling user.  This is
hopefully close to equivalent to fscap support from a security perspective.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Richa Marwaha <rmarwah@linux.vnet.ibm.com>
Signed-off-by: Corey Bryant <coreyb@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoAdd access control support to qemu bridge helper
Corey Bryant [Thu, 26 Jan 2012 14:42:25 +0000 (09:42 -0500)]
Add access control support to qemu bridge helper

We go to great lengths to restrict ourselves to just cap_net_admin as an OS
enforced security mechanism.  However, we further restrict what we allow users
to do to simply adding a tap device to a bridge interface by virtue of the fact
that this is the only functionality we expose.

This is not good enough though.  An administrator is likely to want to restrict
the bridges that an unprivileged user can access, in particular, to restrict
an unprivileged user from putting a guest on what should be isolated networks.

This patch implements an ACL mechanism that is enforced by qemu-bridge-helper.
The ACLs are fairly simple whitelist/blacklist mechanisms with a wildcard of
'all'.  All users are blacklisted by default, and deny takes precedence over
allow.

An interesting feature of this ACL mechanism is that you can include external
ACL files.  The main reason to support this is so that you can set different
file system permissions on those external ACL files.  This allows an
administrator to implement rather sophisticated ACL policies based on
user/group policies via the file system.

As an example:

/etc/qemu/bridge.conf root:qemu 0640

 allow br0
 include /etc/qemu/alice.conf
 include /etc/qemu/bob.conf
 include /etc/qemu/charlie.conf

/etc/qemu/alice.conf root:alice 0640
 allow br1

/etc/qemu/bob.conf root:bob 0640
 allow br2

/etc/qemu/charlie.conf root:charlie 0640
 deny all

This ACL pattern allows any user in the qemu group to get a tap device
connected to br0 (which is bridged to the physical network).

Users in the alice group can additionally get a tap device connected to br1.
This allows br1 to act as a private bridge for the alice group.

Users in the bob group can additionally get a tap device connected to br2.
This allows br2 to act as a private bridge for the bob group.

Users in the charlie group cannot get a tap device connected to any bridge.

Under no circumstance can the bob group get access to br1 or can the alice
group get access to br2.  And under no cicumstance can the charlie group
get access to any bridge.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Richa Marwaha <rmarwah@linux.vnet.ibm.com>
Signed-off-by: Corey Bryant <coreyb@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoAdd basic version of bridge helper
Corey Bryant [Thu, 26 Jan 2012 14:42:24 +0000 (09:42 -0500)]
Add basic version of bridge helper

This patch adds a helper that can be used to create a tap device attached to
a bridge device.  Since this helper is minimal in what it does, it can be
given CAP_NET_ADMIN which allows qemu to avoid running as root while still
satisfying the majority of what users tend to want to do with tap devices.

The way this all works is that qemu launches this helper passing a bridge
name and the name of an inherited file descriptor.  The descriptor is one
end of a socketpair() of domain sockets.  This domain socket is used to
transmit a file descriptor of the opened tap device from the helper to qemu.

The helper can then exit and let qemu use the tap device.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Richa Marwaha <rmarwah@linux.vnet.ibm.com>
Signed-off-by: Corey Bryant <coreyb@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agohw/vmmouse.c: Disable vmmouse after reboot
Gerhard Wiesinger [Wed, 25 Jan 2012 20:04:14 +0000 (21:04 +0100)]
hw/vmmouse.c: Disable vmmouse after reboot

Bugfix after reboot when vmmouse was enabled and another OS which uses e.g. PS/2
mouse.

Details:
When a guest activated the vmmouse followed by a reboot the vmmouse was still
enabled and the PS/2 mouse was therefore unsusable. When another guest is then
booted without vmmouse support (e.g. PS/2 mouse) the mouse is not working.

Reason is that VMMouse has priority and disables all other mouse entities
and therefore must be disabled on reset.

Testscenario:
1.) Boot e.g. OS with VMMouse support (e.g. Windows with VMMouse tools)
2.) reboot
3.) Boot e.g. OS without VMMouse support (e.g. DOS) => PS/2 mouse doesn't work
     any more. Fixes that issue.

Testscenario 2 by Jan Kiszka <jan.kiszka@siemens.com>:
Confirm that this patch fixes a real issue. Setup: qemu.git,
opensuse 11.4 guest, SDL graphic, system_reset while guest is using the
vmmouse. Without the patch, the vmmouse become unusable after the
reboot. Also, the mouse stays in absolute mode even before X starts again.

Fixed by:
Disabling the vmmouse in its reset handler.

Tested-by: Andreas F=E4rber <afaerber@suse.de>
Signed-off-by: Gerhard Wiesinger <lists@wiesinger.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agokeep the PID file locked for the lifetime of the process
Laszlo Ersek [Fri, 27 Jan 2012 13:34:05 +0000 (14:34 +0100)]
keep the PID file locked for the lifetime of the process

The lockf() call in qemu_create_pidfile() aims at ensuring mutual
exclusion. We shouldn't close the pidfile on success (as introduced by
commit 1bbd1592), because that drops the lock as well [1]:

    "File locks shall be released on first close by the locking process
    of any file descriptor for the file."

Coverity may complain again about the leaked file descriptor; let's
worry about that later.

v1->v2:
- add reference to 1bbd1592
- explain the intentional fd leak in the source

[1] http://pubs.opengroup.org/onlinepubs/9699919799/functions/lockf.html

Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agomain-loop: For tools, initialize timers as part of qemu_init_main_loop()
Michael Roth [Sat, 21 Jan 2012 17:13:53 +0000 (11:13 -0600)]
main-loop: For tools, initialize timers as part of qemu_init_main_loop()

In some cases initializing the alarm timers can lead to non-negligable
overhead from programs that link against qemu-tool.o. At least,
setting a max-resolution WinMM alarm timer via mm_start_timer() (the
current default for Windows) can increase the "tick rate" on Windows
OSs and affect frequency scaling, and in the case of tools that run
in guest OSs such has qemu-ga, the impact can be fairly dramatic
(+20%/20% user/sys time on a core 2 processor was observed from an idle
Windows XP guest).

This patch doesn't address the issue directly (not sure what a good
solution would be for Windows, or what other situations it might be
noticeable), but it at least limits the scope of the issue to programs
that "opt-in" to using the main-loop.c functions by only enabling alarm
timers when qemu_init_main_loop() is called, which is already required
to make use of those facilities, so existing users shouldn't be
affected.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agomain-loop: Fix SetEvent() on uninitialized handle on win32
Michael Roth [Sat, 21 Jan 2012 01:08:27 +0000 (19:08 -0600)]
main-loop: Fix SetEvent() on uninitialized handle on win32

The __attribute__((constructor)) init_main_loop() automatically get
called if qemu-tool.o is linked in. On win32, this leads to
a qemu_notify_event() call which attempts to SetEvent() on a HANDLE that
won't be initialized until qemu_init_main_loop() is manually called,
breaking qemu-tools.o programs on Windows at runtime.

This patch checks for an initialized event handle before attempting to
set it, which is analoguous to how we deal with an unitialized
io_thread_fd in the posix implementation.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agooptionroms: Silence intermediate file removal
Jan Kiszka [Mon, 30 Jan 2012 10:27:33 +0000 (11:27 +0100)]
optionroms: Silence intermediate file removal

The build process of optionroms spits out an "rm ..." line. Moreover, it
removes all .o files that can be handy for debugging purposes. So
disable automatic intermediate removal.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agosdl: Limit sdl_grab_end in handle_activation to Windows hosts
Jan Kiszka [Tue, 31 Jan 2012 12:45:31 +0000 (13:45 +0100)]
sdl: Limit sdl_grab_end in handle_activation to Windows hosts

There are scenarios on Linux with some SDL versions where
handle_activation is continuous invoked with state = SDL_APPINPUTFOCUS
and gain = 0 while we grabbed the input. This causes a ping-pong when we
grab the input after an absolute mouse entered the window.

As this sdl_grab_end was once introduced to work around a Windows-only
issue (0294ffb9c8), limit it to that platform.

CC: Erik Rull <erik.rull@rdsoftware.de>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agosdl: Grab input on end of non-absolute mouse click
Jan Kiszka [Tue, 31 Jan 2012 12:45:30 +0000 (13:45 +0100)]
sdl: Grab input on end of non-absolute mouse click

By grabbing the input already on button down, we leave the button in
that state for the host GUI. Thus it takes another click after releasing
the input again to synchronize the mouse button state.

Avoid this by grabbing on button up.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoRevert "Handle SDL grabs failing (Mark McLoughlin)"
Jan Kiszka [Tue, 31 Jan 2012 12:45:29 +0000 (13:45 +0100)]
Revert "Handle SDL grabs failing (Mark McLoughlin)"

This reverts commit 6bb816031f8bc0aafc3476e6dfa4293ee3a5f106.

SDL_WM_GrabInput does not reliably bail out if grabbing is impossible.
So if we get here, we already lost and will block. But this can no
longer happen due to the check in sdl_grab_start. So this patch became
obsolete.

Conflicts:

sdl.c

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agosdl: Fix block prevention of SDL_WM_GrabInput
Jan Kiszka [Tue, 31 Jan 2012 12:45:28 +0000 (13:45 +0100)]
sdl: Fix block prevention of SDL_WM_GrabInput

Consistently check for SDL_APPINPUTFOCUS before trying to grab the input
focus. Just checking for SDL_APPACTIVE doesn't work. Moving the check to
sdl_grab_start allows for some consolidation.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agosdl: Do not grab mouse on mode switch while in background
Jan Kiszka [Tue, 31 Jan 2012 12:45:27 +0000 (13:45 +0100)]
sdl: Do not grab mouse on mode switch while in background

When the mouse mode changes to absolute while the SDL windows is not in
focus, refrain from grabbing the input. It would steal from some other
window.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agoImprove default machine options usability
Jan Kiszka [Fri, 27 Jan 2012 18:55:43 +0000 (19:55 +0100)]
Improve default machine options usability

So far we overwrite the machine options completely with defaults if no
accel=value is provided. More user friendly is to fill in only
unspecified options. The new qemu_opts_set_defaults enables this.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
12 years agomc146818rtc: Use lost_tick_policy property
Jan Kiszka [Mon, 23 Jan 2012 19:15:12 +0000 (20:15 +0100)]
mc146818rtc: Use lost_tick_policy property

Allow to configure the MC146818 RTC via the new lost tick policy
property and replace rtc_td_hack with this mechanism.

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