monitor: New format for handlers argument types

Current handlers argument types, as defined in qemu-monitor.hx file,
are a sequence of chars where each one represents one argument type
of the command handler. The number of chars is also used to know how
many arguments a given handler accepts.

This commit defines a new format, which makes mandatory the use of
a name for each argument.

For example, do_eject() command handler is currently defined as:

{ "eject", "-fB", do_eject, ... }

With the new format it becomes:

{ "eject", "force:-f,filename:B", do_eject, ... }

This way the Monitor will be capable of setting up a dictionary, using
each argument's name as the key and the argument itself as the value.

This commit also adds two new functions: key_get_info() and
next_arg_type(), both are used to parse the new format.

Currently key_get_info() consumes the 'key' part of the new format and
discards it, this way the current parsing code is not affected by this
change.

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

Add wrappers to functions used by the Monitor

Some functions exported to be used by the Monitor as command
handlers are also called in other places as regular functions.

When those functions got ported to use the Monitor dictionary
to pass argments, the callers will have to setup a dictionary
to be able to call them.

To avoid this problem, this commit add wrappers to those functions,
so that we change the wrapper to accept the dictionary, letting
the current functions as is.

The following wrappers are being added:

- do_help_cmd()
- do_pci_device_hot_remove()

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

Introduce QDict

QDict is a high-level dictionary data type that can be used to store a
collection of QObjects. A unique key is associated with only one
QObject.

The following functions are available:

- qdict_new()    Create a new QDict
- qdict_put()    Add a new 'key:object' pair
- qdict_get()    Get the QObject of a given key
- qdict_del()    Delete a 'key:object' pair
- qdict_size()   Return the size of the dictionary
- qdict_haskey() Check if a given 'key' exists

Some high-level helpers to operate on QStrings and QInts objects
are also provided.

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

Introduce QString

QString is a high-level data type that can be used to represent
C strings.

The following functions are available:

- qstring_from_str() Create a new QString
- qstring_get_str()  Get a pointer to the stored string

Note that qstring_get_str() is too low-level for a data type like
this, but it's interesting for quick read-only accesses.

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

Introduce QInt

QInt is a high-level data type that can be used to represent integers,
internally it stores an int64_t value.

The following functions are available:

- qint_from_int() Create a new QInt
- qint_get_int()  Get the stored integer

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

Introduce QObject

This commit introduces the qobject.h header file, it contains
basic QObject definitions and helper macros.

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

ide: move cmd+irq from IDEState to IDEBus.

These variables are per bus, not per drive.  Lets move them and
cleanup things a bit.  And fix the cmd migration bug for real.

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

ide: pass down DriveInfo instead of BlockDriverState

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

Support for multiple -monitor devices

Rebased version of Anthony's patch: Allow to specify more than one
monitor terminal via the -monitor command line switch. This is
particularly useful when libvirt or some other management tool already
occupies the primary monitor but you need another one for debugging.
The current clumsy workaround is to multiplex such additional terminals
over a qemu character device (e.g. -serial mon:<device>).

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

net: Fix send queue ordering

Ensure that packets enqueued for delayed delivery are dequeued in FIFO
order. At least one simplistic guest TCP/IP stack became unhappy due to
sporadically reordered packet streams.

At this chance, switch the send queue implementation to TAILQ.

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

do not issue ioctl from within the io thread

According to Documentation/kvm/api.txt, (and well, to common sense),
we should not be calling vcpu ioctls from within the iothread.
Since vcpu initialization issues a vcpu ioctl, move it a little bit
further in time to prevent it.

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

Delay sighandler_setup()

If we are using --serial telnet:0:5555,server  or similar, ^C will not
kill qemu.  We need to first connect using telnet, and the the ^C takes
effect.

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

Remove typedef for bool from eepro100.c

eepro100.c shouldn't have the need to do this in its local header file.

And I recently started getting this:

$ make -j3
...
  CC    x86_64-softmmu/eepro100.o
/home/amit/src/qemu/hw/eepro100.c:112: error: two or more data types
in declaration specifiers
/home/amit/src/qemu/hw/eepro100.c:112: warning: useless type name in
empty declaration
make[1]: *** [eepro100.o] Error 1

so just remove the typedef and include <stdbool.h> instead.

Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

add documentation for multi-core features

extend QEMU's internal help and man page to cover the recently
added multi-core feature.

Signed-off-by: Andre Przywara <andre.przywara@amd.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

piix3: use new vmstate infrastructure

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

i440fx: use new vmstate infrastructure

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

VMState: Fix sub-structs versioning

We can't check the version in a substruct, it is not stored anywhere

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

Update SaveVM versions

Now that we have all fields belonging to a PCIDevice, save each field
on the device that it belongs.  This means moving pci_irq_levels
from PCII440FXState to PIIX3State.
Old formats are loaded, but we only save on the new saner format.

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

pci_irq_levels[] belong to PIIX3State

With previous cleanups, now it is possible to put it where it belongs

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

Save irq_state into PCII440FXState

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

We can add piix3_dev now to PIIX3IrqState

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

Fold piix3_init() intto i440fx_init

i440fx_init will now work properly if we don't setup piix3

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

Introduce PIIX3IrqState for piix3 irq's state

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

Create PIIX3State instead of using PCIDevice for PIIX3

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

We want the argument pass to set_irq to be opaque

piix_pci want to pass more things that the pic

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

Move smm_enabled and isa_memory_mappings to PCII440FXState

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

Use PCII440FXState instead of generic PCIDevice

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

low_set_irq is not used anywhere

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

Split piix4 support from piix_pci.c

Now mips_malta uses piix4 and pc's use piix_pci definitions

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

piix4 don't use pci_irq_levels at all

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

microblaze: Trap on illegal load/store sizes.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

microblaze: Correct FPU decoding masks.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

microblaze: Compute masks for alignment checks at translation time.

Thanks to Blue Swirl for reporting.
Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

microblaze: Trap on bus accesses to unmapped areas.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

microblaze: MMU shows more respect to synthesis config.

The microblaze MMU can be synthesized in different configurations.
Have the MMU model show more respect to the chosen configuration.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

microblaze: Trap on unaligned data accesses.

Untested...

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

microblaze: Trap on divizions by zero.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

microblaze: Correct mfs into r0.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

microblaze: Catch illegal insns and privilege violations.

Raise illegal instruction exceptions when executing instructions that
require units not available on the particulare microblaze configuration.

Also trap priviliege violations made by userspace.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

microblaze: Add infrastructure for supporting hw exceptions.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

microblaze: Clear exception in dslot ESR bit if not in dslot.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

microblaze: Correct MMU debug log.

Signed-off-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>

Fix warning on x86_64

audio/esdaudio.c: In function 'qesd_thread_out':
audio/esdaudio.c:136: error: format '%d' expects type 'int', but
argument 3 has type 'ssize_t'
audio/esdaudio.c: In function 'qesd_thread_in':
audio/esdaudio.c:366: error: format '%d' expects type 'int', but
argument 3 has type 'ssize_t'

Signed-off-by: Kirill A. Shutemov <kirill@shutemov.name>
Signed-off-by: malc <av1474@comtv.ru>

Sparc32: port sun4c interrupt controller to VMState design

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

Sparc32: port timers to VMState design

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

Sparc32: timer field is never NULL

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

Add VMState support for ptimers

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

esp (sparc32) Extra scsi data.

> ESP: Message Accepted (12)
> ESP: Transfer status (sense=0)
> ESP: read reg[5]: 0x20
> ESP: read reg[4]: 0x07
> ESP: read reg[7]: 0x02
> Extra scsi data. Fatal error.

It looks like "Message Accepted" shouldn't write a response. At least
ESP_RFLAGS must definetely be 0. With the following fix OBP goes one
step further:

Signed-off-by: Artyom Tarasenko <atar4qemu@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

Fix compiler warnings

Starting with commit df7a86ed735eafefbd046c8cad7134652fe3f600,
mingw32 builds result in a compiler warning for dns_addr:

  CC    slirp/slirp.o
/home/stefan/src/qemu/savannah/qemu/slirp/slirp.c:50: warning: missing braces around initializer
/home/stefan/src/qemu/savannah/qemu/slirp/slirp.c:50: warning: (near initialization for ‘dns_addr.S_un’)

Removing the assignment fixes the warning without the need of special code
for mingw32 (and also saves some bytes in the resulting binary).

To fix another potential compiler warning, the missing 'static'
attribute was added.

The same changes were applied to dns_addr_time.

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>

Fix breakage due to __thread

Thread-local storage is not supported on all hosts.

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

Fix gcc 3 warning: comparison is always true due to limited range of data type

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

Fix gcc 3 warning about uninitialized variable

If nb_sectors is 0, cluster_offset will not be initialized.

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

Sparc32: port Slavio misc devices to VMState design

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

Sparc32: port ECC memory controller to VMState design

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

Sparc32: port IOMMU to VMState design

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

Sparc32: port DMA controller to VMState design

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

Sparc32: port TCX to VMState design

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

Sparc32: port interrupt controller to VMState design

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

Update OpenBIOS images to r569

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

Suppress kraxelisms

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

Fix Sparc64 breakage: add dummy ISA irqs

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

Fix merge of 59f2a78793b6d17634f39646d604e84af51e0919

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

Don't compile aio code if CONFIG_LINUX_AIO is undefined

This patch fixes linker errors when building QEMU without Linux AIO support.

It is based on suggestions from malc and Kevin Wolf.

Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>

Fix sb16 breakage due to unassigned s->irq

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

Fix formatting

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

Merge commit 'linux-user/linux-user-for-upstream' into staging

Relax pthreads check to avoid win32

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

e1000 cleanup

Remove un needed casts from void *.
Use DO_UPCAST() instead of blind casts

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

pcnet: Add pci_pcnet_{save/load} functions, then remove PCNetState pci_dev field

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

pcnet: remove useless casts This are casts to the very type of the function

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

pcnet: Change casts to DO_UPCAST() for PCIPCNetState

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

Add pci_cirrus_vga_{save/load} functions, then remove vga.pci_dev use

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

cirrus_vga: Change casts to DO_UPCAST() for PCICirrusVGAState

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

cirrus_vga: remove pointless cast from void *

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

Add pci_ne2000_{save/load} functions, then remove pci_dev NE2000State field

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

We need PCINE2000State for save/load functions

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

ne2000: Change casts to DO_UPCAST() for PCINE2000State

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

ne2000: pci_dev has this very value with the right type

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

ne2000: remove casts from void *

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

usb-ohci: Change casts to DO_UPCAST() for OHCIPCIState

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

Everything outside of vga.c should use VGACommonState

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

vga: Rename vga_state -> vga

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

Introduce vga_common_reset() to be able to typcheck vga_reset()

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

lsi53c895a: LSIState is a DeviceHost

Go figure.

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

lsi53c895a: rename PCIDevice field from pci_dev to dev (consistence)

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

lsi53c895a: use DO_UPCAST to cast from PCIDevice

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

lsi53c895a: remove pointless cast from void *

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

rtl8139: remove pointless cast from void *

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

rtl8139: Remove unneeded double indirection of PCIRTL8139State

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

eepro100: Remove unused indirection of PCIDevice

Once there, there is no way that we don't have a PCI Device at save/load time. Remove the check

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

eepro100: cast a void * makes no sense

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

eepro100: convert casts to DO_UPCAST()

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

slirp: Read host DNS config on demand

Currently the qemu user-mode networking stack reads the host DNS
configuration (/etc/resolv.conf or the Windows equivalent) only once
when qemu starts.  This causes name lookups in the guest to fail if the
host is moved to a different network from which the original DNS servers
are unreachable, a common occurrence when the host is a laptop.

This patch changes the slirp code to read the host DNS configuration on
demand, caching the results for at most 1 second to avoid unnecessary
overhead if name lookups occur in rapid succession.  On non-Windows
hosts, /etc/resolv.conf is re-read only if the file has been replaced or
if its size or mtime has changed.

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

slirp: Remove our_addr code

Three problems with our_addr:

- It's determined only once when qemu starts, but the address can change
(just like the DNS configuration can).

- It's supposed to be the IP address of a host network interface, but
there's no guarantee that gethostbyname(gethostname()) actually does
that: the host might be a laptop that has only a loopback interface up,
or the hostname might be localhost.localdomain, etc.

- It's useless at best: get_dns_addr() calls it, there's no reason to
send DNS requests to a different IP address if you're running a DNS
server on the host and resolv.conf points to 127.0.0.1.

These problems are easily solved by removing the code.

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

switch balloon initialization to -device.

With that patch applied "-balloon virtio,args" becomes a shortcut for
"-device virtio-balloon-pci,args".

Side effects:
 - ballon device gains support for id=<tag>.
 - ballon device is off by default now.
 - initialization order changes, which may in different pci slot
   assignment depending on the VM configuration.

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

ide: move code to hw/ide/

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

ide: add save/restore support for mmio

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

ide: add save/restore support for isa

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

ide: split away ide-microdrive.c

create ide-microdrive.c and place microdrive support there.
only build ide-microdrive support for platforms using it.

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

ide: split away ide-mmio.c

create ide-mmio.c and place mmio support there.
only build ide-mmio support for platforms using it.

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