sdk/emulator/qemu.git
13 years agoMerge remote branch 'mst/for_anthony' into staging
Anthony Liguori [Fri, 17 Dec 2010 14:21:29 +0000 (08:21 -0600)]
Merge remote branch 'mst/for_anthony' into staging

13 years agoWatchdog: disable watchdog timer when hard-rebooting a guest.
Richard W.M. Jones [Fri, 24 Sep 2010 15:08:06 +0000 (16:08 +0100)]
Watchdog: disable watchdog timer when hard-rebooting a guest.

This commit causes the watchdog timer to be reset when a guest is
hard-rebooted.

The failure case previously was as follows:

  (a) guest boots, watchdog is enabled

  (b) guest does a reset eg:
        echo 'b' > /proc/sysrq-trigger
    (note that an ordinary /sbin/reboot wouldn't hit this case
    since as the watchdog daemon is shut down, the daemon would
    properly disable the watchdog device)

  (c) the reboot takes longer than the remaining time on the
    watchdog

  (d) the watchdog therefore fires during the reboot

  (e) probably the VM would just reboot again at this point which
    is pretty benign, but it could depend on the action that the
    user had selected for the watchdog

Now we use the qdev reset function to register a reset handler
which disables the timer.  Note the handler is called _either_
just after init _or_ when the guest reboots.

In the i6300esb case there is a small refactoring of the code so
that the device's internal state is now fully restored to defaults
on a reboot.

Signed-off-by: Richard W.M. Jones <rjones@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoPass boot device list to firmware.
Gleb Natapov [Wed, 8 Dec 2010 11:35:09 +0000 (13:35 +0200)]
Pass boot device list to firmware.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoAdd notifier that will be called when machine is fully created.
Gleb Natapov [Wed, 8 Dec 2010 11:35:08 +0000 (13:35 +0200)]
Add notifier that will be called when machine is fully created.

Action that depends on fully initialized device model should register
with this notifier chain.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoAdd bootindex for option roms.
Gleb Natapov [Wed, 8 Dec 2010 11:35:07 +0000 (13:35 +0200)]
Add bootindex for option roms.

Extend -option-rom command to have additional parameter ,bootindex=.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoChange fw_cfg_add_file() to get full file path as a parameter.
Gleb Natapov [Wed, 8 Dec 2010 11:35:06 +0000 (13:35 +0200)]
Change fw_cfg_add_file() to get full file path as a parameter.

Change fw_cfg_add_file() to get full file path as a parameter instead
of building one internally. Two reasons for that. First caller may need
to know how file is named. Second this moves policy of file naming out
from fw_cfg. Platform may want to use more then two levels of
directories for instance.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoAdd bootindex parameter to net/block/fd device
Gleb Natapov [Wed, 8 Dec 2010 11:35:05 +0000 (13:35 +0200)]
Add bootindex parameter to net/block/fd device

If bootindex is specified on command line a string that describes device
in firmware readable way is added into sorted list. Later this list will
be passed into firmware to control boot order.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoAdd get_fw_dev_path callback to scsi bus.
Gleb Natapov [Wed, 8 Dec 2010 11:35:04 +0000 (13:35 +0200)]
Add get_fw_dev_path callback to scsi bus.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoAdd get_fw_dev_path callback for usb bus.
Gleb Natapov [Wed, 8 Dec 2010 11:35:03 +0000 (13:35 +0200)]
Add get_fw_dev_path callback for usb bus.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoRecord which USBDevice USBPort belongs too.
Gleb Natapov [Wed, 8 Dec 2010 11:35:02 +0000 (13:35 +0200)]
Record which USBDevice USBPort belongs too.

Ports on root hub will have NULL here. This is needed to reconstruct
path from device to its root hub to build device path.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoAdd get_fw_dev_path callback for pci bus.
Gleb Natapov [Wed, 8 Dec 2010 11:35:01 +0000 (13:35 +0200)]
Add get_fw_dev_path callback for pci bus.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoAdd get_fw_dev_path callback for system bus.
Gleb Natapov [Wed, 8 Dec 2010 11:35:00 +0000 (13:35 +0200)]
Add get_fw_dev_path callback for system bus.

Prints out mmio or pio used to access child device.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoAdd get_fw_dev_path callback to IDE bus.
Gleb Natapov [Wed, 8 Dec 2010 11:34:59 +0000 (13:34 +0200)]
Add get_fw_dev_path callback to IDE bus.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoStore IDE bus id in IDEBus structure for easy access.
Gleb Natapov [Wed, 8 Dec 2010 11:34:58 +0000 (13:34 +0200)]
Store IDE bus id in IDEBus structure for easy access.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoAdd get_fw_dev_path callback to ISA bus in qdev.
Gleb Natapov [Wed, 8 Dec 2010 11:34:57 +0000 (13:34 +0200)]
Add get_fw_dev_path callback to ISA bus in qdev.

Use device ioports to create unique device path.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoKeep track of ISA ports ISA device is using in qdev.
Gleb Natapov [Wed, 8 Dec 2010 11:34:56 +0000 (13:34 +0200)]
Keep track of ISA ports ISA device is using in qdev.

Store all io ports used by device in ISADevice structure.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoIntroduce new BusInfo callback get_fw_dev_path.
Gleb Natapov [Wed, 8 Dec 2010 11:34:55 +0000 (13:34 +0200)]
Introduce new BusInfo callback get_fw_dev_path.

New get_fw_dev_path callback will be used for build device path usable
by firmware in contrast to qdev qemu internal device path.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoIntroduce fw_name field to DeviceInfo structure.
Gleb Natapov [Wed, 8 Dec 2010 11:34:54 +0000 (13:34 +0200)]
Introduce fw_name field to DeviceInfo structure.

Add "fw_name" to DeviceInfo to use in device path building. In
contrast to "name" "fw_name" should refer to functionality device
provides instead of particular device model like "name" does.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agomonitor: implement x86 info mem for PAE and long modes
Blue Swirl [Sat, 11 Dec 2010 18:56:27 +0000 (18:56 +0000)]
monitor: implement x86 info mem for PAE and long modes

'info mem' didn't show correct information for PAE mode and
x86_64 long mode.

Fix by implementing the output for missing modes.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agomonitor: implement x86 info tlb for PAE and long modes
Blue Swirl [Sat, 11 Dec 2010 18:56:24 +0000 (18:56 +0000)]
monitor: implement x86 info tlb for PAE and long modes

'info tlb' didn't show correct information for PAE mode and
x86_64 long mode.

Implement the missing modes. Also print NX bit for PAE and long modes.
Fix off-by-one error in 32 bit mode mask.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agowdt_i6300esb: register a reset function
Bernhard Kohl [Wed, 8 Dec 2010 14:59:55 +0000 (15:59 +0100)]
wdt_i6300esb: register a reset function

The device shall set its default hardware state after each reset.
This includes that the timer is stopped which is especially important
if the guest does a reboot independantly of a watchdog bite. I moved
the initialization of the state variables completely from the init
to the reset function which is called right after init during the
first boot and afterwards during each reboot.

Signed-off-by: Bernhard Kohl <bernhard.kohl@nsn.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoisa-bus.c: use hw_error instead of fprintf
Tristan Gingold [Fri, 3 Dec 2010 11:05:03 +0000 (12:05 +0100)]
isa-bus.c: use hw_error instead of fprintf

Minor clean-up in isa-bus.c.  Using hw_error is more consistent.
There is a difference however: hw_error dumps the cpu state.

Signed-off-by: Tristan Gingold <gingold@adacore.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agousb_ohci: Always use little endian
Alexander Graf [Wed, 8 Dec 2010 11:05:50 +0000 (12:05 +0100)]
usb_ohci: Always use little endian

This patch replaces explicit bswaps with endianness hints to the
mmio layer.

Because we don't depend on the target endianness anymore, we can also
move the driver over to Makefile.objs.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoisa_mmio: Always use little endian
Alexander Graf [Wed, 8 Dec 2010 11:05:49 +0000 (12:05 +0100)]
isa_mmio: Always use little endian

This patch converts the ISA MMIO bridge code to always use little endian mmio.
All bswap code that existed was only there to convert from native cpu
endianness to little endian ISA devices.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoheathrow_pic: Declare as little endian
Alexander Graf [Wed, 8 Dec 2010 11:05:48 +0000 (12:05 +0100)]
heathrow_pic: Declare as little endian

This patch replaces explicit bswaps with endianness hints to the
mmio layer.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agortl8139: Declare as little endian
Alexander Graf [Wed, 8 Dec 2010 11:05:47 +0000 (12:05 +0100)]
rtl8139: Declare as little endian

This patch replaces explicit bswaps with endianness hints to the
mmio layer.

Because we don't depend on the target endianness anymore, we can also
move the driver over to Makefile.objs.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoopenpic: Replace explicit byte swap with endian hints
Alexander Graf [Wed, 8 Dec 2010 11:05:46 +0000 (12:05 +0100)]
openpic: Replace explicit byte swap with endian hints

This patch replaces explicit bswaps with endianness hints to the
mmio layer.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoppc4xx_pci: Declare as little endian
Alexander Graf [Wed, 8 Dec 2010 11:05:45 +0000 (12:05 +0100)]
ppc4xx_pci: Declare as little endian

This patch replaces explicit bswaps with endianness hints to the
mmio layer.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoversatile_pci: Declare as little endian
Alexander Graf [Wed, 8 Dec 2010 11:05:44 +0000 (12:05 +0100)]
versatile_pci: Declare as little endian

This patch replaces explicit bswaps with endianness hints to the
mmio layer.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoprep: Declare as little endian
Alexander Graf [Wed, 8 Dec 2010 11:05:43 +0000 (12:05 +0100)]
prep: Declare as little endian

This patch replaces explicit bswaps with endianness hints to the
mmio layer.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoe1000: Make little endian
Alexander Graf [Wed, 8 Dec 2010 11:05:42 +0000 (12:05 +0100)]
e1000: Make little endian

The e1000 has compatibility code to handle big endianness which makes it
mandatory to be recompiled on different targets.

With the generic mmio endianness solution, there's no need for that anymore.
We just declare all mmio to be little endian and call it a day.

Because we don't depend on the target endianness anymore, we can also
move the driver over to Makefile.objs.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agouninorth: Get rid of bswap
Alexander Graf [Wed, 8 Dec 2010 11:05:41 +0000 (12:05 +0100)]
uninorth: Get rid of bswap

There's no need to bswap once we correctly set the mmio to be little endian.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agopci-host: Delegate bswap to mmio layer
Alexander Graf [Wed, 8 Dec 2010 11:05:40 +0000 (12:05 +0100)]
pci-host: Delegate bswap to mmio layer

The only reason we have bswap versions of the pci host code is that
most pci host devices are little endian. The ppc e500 is the only
odd one here, being big endian.

So let's directly pass the endianness down to the mmio layer and not
worry about it on the pci host layer.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agodbdma: Make little endian
Alexander Graf [Wed, 8 Dec 2010 11:05:39 +0000 (12:05 +0100)]
dbdma: Make little endian

The device is only used on big endian systems, but always byte swaps. That's
a very good indicator that it's actually a little endian device ;-).

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoMake simple io mem handler endian aware
Alexander Graf [Wed, 8 Dec 2010 11:05:38 +0000 (12:05 +0100)]
Make simple io mem handler endian aware

As an alternative to the 3 individual handlers, there is also a simplified
io mem hook function. To be consistent, let's add an endianness parameter
there too.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoAdd endianness as io mem parameter
Alexander Graf [Wed, 8 Dec 2010 11:05:37 +0000 (12:05 +0100)]
Add endianness as io mem parameter

As stated before, devices can be little, big or native endian. The
target endianness is not of their concern, so we need to push things
down a level.

This patch adds a parameter to cpu_register_io_memory that allows a
device to choose its endianness. For now, all devices simply choose
native endian, because that's the same behavior as before.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoexec: introduce endianness swapped mmio
Alexander Graf [Wed, 8 Dec 2010 11:05:36 +0000 (12:05 +0100)]
exec: introduce endianness swapped mmio

The way we're currently modeling mmio is too simplified. We assume that
every device has the same endianness as the target CPU. In reality,
most devices are little endian (all PCI and ISA ones I'm aware of). Some
are big endian (special system devices) and a very little fraction is
target native endian (fw_cfg).

So instead of assuming every device to be native endianness, let's move
to a model where the device tells us which endianness it's in.

That way we can compile the devices only once and get rid of all the ugly
swap will be done by the underlying layer.

For the same of readability, this patch only introduces the helper framework
but doesn't allow the registering code to set its endianness yet.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agonoaudio: fix return value for read()
Michael Walle [Wed, 8 Dec 2010 23:34:51 +0000 (00:34 +0100)]
noaudio: fix return value for read()

Read should return bytes instead of samples.

Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: malc <av1474@comtv.ru>
13 years agopci/aer: factor out common code
Michael S. Tsirkin [Wed, 8 Dec 2010 08:46:28 +0000 (17:46 +0900)]
pci/aer: factor out common code

Same logic is used to assert interrupts
and send msix messages, so add a static functin for this.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
13 years agopci/aer: remove dead code
Michael S. Tsirkin [Wed, 8 Dec 2010 08:46:27 +0000 (17:46 +0900)]
pci/aer: remove dead code

Remove some unused variables and return values.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
13 years agopci/aer: fix interrupt on config write
Michael S. Tsirkin [Wed, 8 Dec 2010 08:46:26 +0000 (17:46 +0900)]
pci/aer: fix interrupt on config write

config write handling for aer seems broken:
For example, it won't clear a level interrupt
when command register is set to 0.

Make it match the spec: level should equal
the logical or of enabled bits, msi only
be sent when the logical or changes.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
13 years agopci/aer: fix error injection
Michael S. Tsirkin [Wed, 8 Dec 2010 08:46:25 +0000 (17:46 +0900)]
pci/aer: fix error injection

Fix the injection logic upon aer message to follow 6.2.4.1.2 more
closely: specifically only send an msi interrupt when the logical or of
the enabled bits changed, not when a bit which was previously clear
becomes set.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
13 years agoMakefile: make msix/msi depend on CONFIG_PCI
Michael S. Tsirkin [Wed, 8 Dec 2010 08:46:24 +0000 (17:46 +0900)]
Makefile: make msix/msi depend on CONFIG_PCI

Possible now that pci is not depending on these.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
13 years agopci: untangle pci/msi dependency
Michael S. Tsirkin [Wed, 8 Dec 2010 08:46:23 +0000 (17:46 +0900)]
pci: untangle pci/msi dependency

msi depends on pci but pci should not depend on msi.
The only dependency we have is a recent addition
of pci_msi_ functions, IMO they add little enough to
open-code in the small number of users.

Follow-up patches add more cleanups.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
13 years agopci: make command SERR bit writable
Isaku Yamahata [Fri, 26 Nov 2010 12:01:41 +0000 (21:01 +0900)]
pci: make command SERR bit writable

pcie aer needs SERR bit to be writable, and the PCI spec requires
this as well.  For compatibility, introduce compat global property
command_serr_enable and make this bit readonly for a pre 0.14 pc
machine.

Signed-off-by: Isaku Yamahata <yamahata@valinux.co.jp>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
13 years agovirtio-net: stop/start bh when appropriate
Michael S. Tsirkin [Mon, 22 Nov 2010 17:52:30 +0000 (19:52 +0200)]
virtio-net: stop/start bh when appropriate

Avoid sending out packets, and modifying
memory, when VM is stopped.
Add assert statements to verify this does not happen.

Avoid scheduling bh when vhost-net is started.

Stop bh when driver disabled bus mastering
(we must not access memory after this).

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Jason Wang <jasowang@redhat.com>
13 years agovirtio-net: don't dma while vm is stopped
Michael S. Tsirkin [Mon, 22 Nov 2010 17:52:19 +0000 (19:52 +0200)]
virtio-net: don't dma while vm is stopped

DMA into memory while VM is stopped makes it
hard to debug migration (consequitive saves
result in different files).
Fixing this completely is a large effort,
this patch does this for virtio-net.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Jason Wang <jasowang@redhat.com>
13 years agomigration/savevm: no need to flush requests
Michael S. Tsirkin [Mon, 22 Nov 2010 17:52:26 +0000 (19:52 +0200)]
migration/savevm: no need to flush requests

There's no need to flush requests after vmstop
as vmstop does it for us automatically now.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Jason Wang <jasowang@redhat.com>
13 years agocpus: flush all requests on each vm stop
Michael S. Tsirkin [Mon, 22 Nov 2010 17:52:22 +0000 (19:52 +0200)]
cpus: flush all requests on each vm stop

Flush all requests once we have stopped all
cpus and devices.
Make sure disk is in consistent state.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Jason Wang <jasowang@redhat.com>
Acked-by: Marcelo Tosatti <mtosatti@redhat.com>
13 years agonet/sock: option to specify local address
Mike Ryan [Wed, 1 Dec 2010 19:16:47 +0000 (11:16 -0800)]
net/sock: option to specify local address

Add an option to specify the host IP to send multicast packets from,
when using a multicast socket for networking. The option takes an IP
address and sets the IP_MULTICAST_IF socket option, which causes the
packets to use that IP's interface as an egress.

This is useful if the host machine has several interfaces with several
virtual networks across disparate interfaces.

Signed-off-by: Mike Ryan <mikeryan@ISI.EDU>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
13 years agoppc: kvm: fix signedness warning
Alexander Graf [Thu, 25 Nov 2010 07:20:46 +0000 (08:20 +0100)]
ppc: kvm: fix signedness warning

I get a warning on a signed comparison with an unsigned variable, so
let's make the variable signed and be happy.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Edgar E. Iglesias <edgar@axis.com>
13 years agoSpeedup 'tb_find_slow' by using the same heuristic as during memory page lookup
Kirill Batuzov [Thu, 2 Dec 2010 13:12:46 +0000 (16:12 +0300)]
Speedup 'tb_find_slow' by using the same heuristic as during memory page lookup

Move the last found TB to the head of the list so it will be found more quickly next time it will be looked for.

Signed-off-by: Kirill Batuzov <batuzovk@ispras.ru>
Signed-off-by: Pavel Yushchenko <pau@ispras.ru>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoRemove unused spin_trylock() function
Peter Maydell [Wed, 1 Dec 2010 19:44:38 +0000 (19:44 +0000)]
Remove unused spin_trylock() function

Remove the spin_trylock() function, as it is not used anywhere,
and is not even implemented if CONFIG_USE_NPTL is defined.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agodarwin-user: Use GCC_FMT_ATTR (format checking)
Stefan Weil [Mon, 15 Nov 2010 20:00:48 +0000 (21:00 +0100)]
darwin-user: Use GCC_FMT_ATTR (format checking)

The redundant forward declaration of qerror in machload.c
is removed because it should be taken from qemu.h.

Please note that this patch is untested because
I have no matching environment to compile it.

Cc: Blue Swirl <blauwirbel@gmail.com>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoaudio: Use GCC_FMT_ATTR (format checking)
Stefan Weil [Mon, 15 Nov 2010 19:54:12 +0000 (20:54 +0100)]
audio: Use GCC_FMT_ATTR (format checking)

Cc: Blue Swirl <blauwirbel@gmail.com>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agotarget-sparc: Use fprintf_function (format checking)
Stefan Weil [Mon, 15 Nov 2010 18:58:41 +0000 (19:58 +0100)]
target-sparc: Use fprintf_function (format checking)

This change was missing in commit
9a78eead0c74333a394c0f7bbfc4423ac746fcd5.

Cc: Blue Swirl <blauwirbel@gmail.com>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years ago*-dis: Replace fprintf_ftype by fprintf_function (format checking)
Stefan Weil [Mon, 15 Nov 2010 18:39:43 +0000 (19:39 +0100)]
*-dis: Replace fprintf_ftype by fprintf_function (format checking)

This patch adds more printf format checking.

Additional modifications were needed for this code change:

* alpha-dis.c: The local definition of MAX conflicts with
  a previous definition from osdep.h, so add an #undef.

* dis-asm.h: Add include for fprintf_function (qemu-common.h).
  The standard (now redundant) includes are removed.

* mis-dis.c: The definition of ARRAY_SIZE is no longer needed
  and must be removed (conflict with previous definition from
  qemu-common.h).

* sh4-dis.c: Remove some unneeded forward declarations.

Cc: Blue Swirl <blauwirbel@gmail.com>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoFix mingw32 and OpenBSD warnings
Blue Swirl [Sat, 4 Dec 2010 17:37:35 +0000 (17:37 +0000)]
Fix mingw32 and OpenBSD warnings

ffsl() is not universally available, so there are these warnings
on both mingw32 and OpenBSD:
/src/qemu/hw/pcie_aer.c: In function 'pcie_aer_update_log':
/src/qemu/hw/pcie_aer.c:399: warning: implicit declaration of function 'ffsl'

Since status field in PCIEAERErr is uint32_t, we can just use ffs() instead.

Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
13 years agoMerge branch 'linux-user-for-upstream' of git://gitorious.org/qemu-maemo/qemu
Edgar E. Iglesias [Sat, 4 Dec 2010 03:18:28 +0000 (04:18 +0100)]
Merge branch 'linux-user-for-upstream' of git://gitorious.org/qemu-maemo/qemu

* 'linux-user-for-upstream' of git://gitorious.org/qemu-maemo/qemu:
  linux-user: fix mips and ppc to use UID16
  update binfmt conf
  linux-user: fix compiler error on nptl
  ARM: linux-user: Restore iWMMXT state from ucontext on sigreturn
  ARM: linux-user: Expose iWMMXT registers to signal handlers
  ARM: linux-user: Restore VFP state from ucontext on sigreturn
  ARM: linux-user: Expose VFP registers to signal handlers
  ARM: Expose vfp_get_fpscr() and vfp_set_fpscr() to C code
  ARM: linux-user: Correct size of padding in target_ucontext_v2
  target-sparc: remove unused functions cpu_lock(), cpu_unlock()
  ARM: enable XScale/iWMMXT in linux-user mode
  linux-user: Translate getsockopt level option
  linux-user: remove unnecessary local from __get_user(), __put_user()
  linux-user: fix memory leaks with NPTL emulation
  linux-user: mmap_reserve() not controlled by RESERVED_VA
  [PATCH] target-arm: remove unused functions cpu_lock(), cpu_unlock()

13 years agoexec: Remove debugging fprintf() that slipped into qemu_ram_alloc_from_ptr()
Stefan Hajnoczi [Fri, 3 Dec 2010 17:09:01 +0000 (17:09 +0000)]
exec: Remove debugging fprintf() that slipped into qemu_ram_alloc_from_ptr()

Remove the debugging fprintf() slipped in via the following commit:

    commit b2e0a138e77245290428a7d599a929e2e1bfe510
    Author: Michael S. Tsirkin <mst@redhat.com>
    Date:   Mon Nov 22 19:52:34 2010 +0200

        migration: stable ram block ordering

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
13 years agolinux-user: fix mips and ppc to use UID16
Martin Mohring [Sun, 7 Nov 2010 19:31:57 +0000 (20:31 +0100)]
linux-user: fix mips and ppc to use UID16

Signed-off-by: Martin Mohring <martin.mohring@5edatasoft.com>
Signed-off-by: Jan-Simon Möller <jsmoeller@linuxfoundation.org>
Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
13 years agoupdate binfmt conf
Riku Voipio [Wed, 1 Dec 2010 12:53:10 +0000 (14:53 +0200)]
update binfmt conf

1) dont register i386 qemu on x86_64 host
2) widen sparc and arm match
3) add sh4, based on patch by David Kozub <zub@linux.fjfi.cvut.cz>

Rest based on patch by Jan-Simon Möller <jsmoeller@linuxfoundation.org>

13 years agolinux-user: fix compiler error on nptl
Riku Voipio [Fri, 26 Nov 2010 14:21:34 +0000 (16:21 +0200)]
linux-user: fix compiler error on nptl

Some compilers detect that new_stack isnt used after dd75d784

Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
13 years agoARM: linux-user: Restore iWMMXT state from ucontext on sigreturn
Peter Maydell [Wed, 24 Nov 2010 15:20:08 +0000 (15:20 +0000)]
ARM: linux-user: Restore iWMMXT state from ucontext on sigreturn

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
13 years agoARM: linux-user: Expose iWMMXT registers to signal handlers
Peter Maydell [Wed, 24 Nov 2010 15:20:07 +0000 (15:20 +0000)]
ARM: linux-user: Expose iWMMXT registers to signal handlers

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
13 years agoARM: linux-user: Restore VFP state from ucontext on sigreturn
Peter Maydell [Wed, 24 Nov 2010 15:20:06 +0000 (15:20 +0000)]
ARM: linux-user: Restore VFP state from ucontext on sigreturn

Restore the VFP registers from the ucontext on return from a signal
handler in linux-user mode. This means that signal handlers cannot
accidentally corrupt the interrupted code's VFP state, and allows
them to deliberately modify the state via the ucontext structure.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
13 years agoARM: linux-user: Expose VFP registers to signal handlers
Peter Maydell [Wed, 24 Nov 2010 15:20:05 +0000 (15:20 +0000)]
ARM: linux-user: Expose VFP registers to signal handlers

For ARM linux-user mode signal handlers, fill in the ucontext with
VFP register contents in the same way that the kernel does. We only
do this for v2 format sigframe (2.6.12 and above); this is actually
bug-for-bug compatible with the older kernels, which don't save and
restore VFP registers either.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
13 years agoARM: Expose vfp_get_fpscr() and vfp_set_fpscr() to C code
Peter Maydell [Wed, 24 Nov 2010 15:20:04 +0000 (15:20 +0000)]
ARM: Expose vfp_get_fpscr() and vfp_set_fpscr() to C code

Expose the vfp_get_fpscr() and vfp_set_fpscr() functions to C
code as well as generated code, so we can use them to read and
write the FPSCR when saving and restoring VFP registers across
signal handlers in linux-user mode.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
13 years agoARM: linux-user: Correct size of padding in target_ucontext_v2
Peter Maydell [Wed, 24 Nov 2010 15:20:03 +0000 (15:20 +0000)]
ARM: linux-user: Correct size of padding in target_ucontext_v2

The padding in the target_ucontext_v2 is defined by the size of
the target's sigset_t type, not the host's. (This bug only causes
problems when we start using the uc_regspace[] array to expose
VFP registers to userspace signal handlers.)

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
13 years agotarget-sparc: remove unused functions cpu_lock(), cpu_unlock()
Peter Maydell [Fri, 19 Nov 2010 13:54:39 +0000 (13:54 +0000)]
target-sparc: remove unused functions cpu_lock(), cpu_unlock()

13 years agoARM: enable XScale/iWMMXT in linux-user mode
Peter Maydell [Fri, 19 Nov 2010 15:36:47 +0000 (15:36 +0000)]
ARM: enable XScale/iWMMXT in linux-user mode

In linux-user mode, the XScale/iWMMXT coprocessors must be enabled
at reset so that we can run code that uses these instructions.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
13 years agolinux-user: Translate getsockopt level option
Jamie Lentin [Fri, 26 Nov 2010 13:04:08 +0000 (15:04 +0200)]
linux-user: Translate getsockopt level option

n setsockopt, the socket level options are translated to the hosts'
architecture before the real syscall is called, e.g.
TARGET_SO_TYPE -> SO_TYPE. This patch does the same with getsockopt.

Tested on a x86 host emulating MIPS.  Without it:-

$ grep getsockopt host.strace
31311 getsockopt(3, SOL_SOCKET, 0x1007 /* SO_??? */, 0xbff17208,
0xbff17204) = -1 ENOPROTOOPT (Protocol not available)

With:-

$ grep getsockopt host.strace
25706 getsockopt(3, SOL_SOCKET, SO_ERROR, [0], [4]) = 0

Whitespace cleanup: Riku Voipio

Signed-off-by: Jamie Lentin <jm@lentin.co.uk>
Signed-off-by: Riku Voipio <riku.voipio@iki.fi>
13 years agolinux-user: remove unnecessary local from __get_user(), __put_user()
Peter Maydell [Mon, 8 Nov 2010 18:13:58 +0000 (18:13 +0000)]
linux-user: remove unnecessary local from __get_user(), __put_user()

Remove an unnecessary local variable from the __get_user() and
__put_user() macros. This avoids confusing compilation failures
if the name of the local variable ('size') happens to be the
same as the variable the macro user is trying to read/write.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
13 years agolinux-user: fix memory leaks with NPTL emulation
Nathan Froyd [Fri, 29 Oct 2010 14:48:57 +0000 (07:48 -0700)]
linux-user: fix memory leaks with NPTL emulation

Running programs that create large numbers of threads, such as this
snippet from libstdc++'s pthread7-rope.cc:

  const int max_thread_count = 4;
  const int max_loop_count = 10000;
  ...
  for (int j = 0; j < max_loop_count; j++)
    {
      ...
      for (int i = 0; i < max_thread_count; i++)
pthread_create (&tid[i], NULL, thread_main, 0);

      for (int i = 0; i < max_thread_count; i++)
pthread_join (tid[i], NULL);
    }

in user-mode emulation will quickly run out of memory.  This is caused
by a failure to free memory in do_syscall prior to thread exit:

          /* TODO: Free CPU state.  */
          pthread_exit(NULL);

The first step in fixing this is to make all TaskStates used by QEMU
dynamically allocated.  The TaskState used by the initial thread was
not, as it was allocated on main's stack.  So fix that, free the
cpu_env, free the TaskState, and we're home free, right?

Not exactly.  When we create a thread, we do:

        ts = qemu_mallocz(sizeof(TaskState) + NEW_STACK_SIZE);
        ...
        new_stack = ts->stack;
        ...
        ret = pthread_attr_setstack(&attr, new_stack, NEW_STACK_SIZE);

If we blindly free the TaskState, then, we yank the current (host)
thread's stack out from underneath it while it still has things to do,
like calling pthread_exit.  That causes problems, as you might expect.

The solution adopted here is to let the C library allocate the thread's
stack (so the C library can properly clean it up at pthread_exit) and
provide a hint that we want NEW_STACK_SIZE bytes of stack.

With those two changes, we're done, right?  Well, almost.  You see,
we're creating all these host threads and their parent threads never
bother to check that their children are finished.  There's no good place
for the parent threads to do so.  Therefore, we need to create the
threads in a detached state so the parent thread doesn't have to call
pthread_join on the child to release the child's resources; the child
does so automatically.

With those three major changes, we can comfortably run programs like the
above without exhausting memory.  We do need to delete 'stack' from the
TaskState structure.

Signed-off-by: Nathan Froyd <froydnj@codesourcery.com>
Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
13 years agolinux-user: mmap_reserve() not controlled by RESERVED_VA
amateur [Tue, 14 Sep 2010 05:22:34 +0000 (13:22 +0800)]
linux-user: mmap_reserve() not controlled by RESERVED_VA

mmap_reserve() should be called only when RESERVED_VA is enabled.
Otherwise, unmaped virtual address space will never be reusable. This
bug will exhaust virtual address space in extreme conditions.

Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
13 years ago[PATCH] target-arm: remove unused functions cpu_lock(), cpu_unlock()
Peter Maydell [Fri, 19 Nov 2010 13:54:44 +0000 (13:54 +0000)]
[PATCH] target-arm: remove unused functions cpu_lock(), cpu_unlock()

Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
13 years agoFix build
Anthony Liguori [Thu, 2 Dec 2010 20:41:59 +0000 (14:41 -0600)]
Fix build

msix.o and msi.o get pulled into the build unconditionally for QMP.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
13 years agoMerge remote branch 'mst/for_anthony' into staging
Anthony Liguori [Thu, 2 Dec 2010 20:16:40 +0000 (14:16 -0600)]
Merge remote branch 'mst/for_anthony' into staging

13 years agovhost: Fix address calculation in vhost_dev_sync_region()
Jason Wang [Sat, 27 Nov 2010 14:05:07 +0000 (22:05 +0800)]
vhost: Fix address calculation in vhost_dev_sync_region()

We still need advance address even we find there's no dirty pages in
current chunk.

Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
13 years agomigration: allow rate > 4g
Michael S. Tsirkin [Tue, 23 Nov 2010 17:05:54 +0000 (19:05 +0200)]
migration: allow rate > 4g

I'd like to disable bandwidth limit or make it very high,
Use int64_t all over to make values >= 4g work.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Jason Wang <jasowang@redhat.com>
13 years agomigration: stable ram block ordering
Michael S. Tsirkin [Mon, 22 Nov 2010 17:52:34 +0000 (19:52 +0200)]
migration: stable ram block ordering

This makes ram block ordering under migration stable, ordered by offset.
This is especially useful for migration to exec, for debugging.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Jason Wang <jasowang@redhat.com>
13 years agotcg-ia64: Fix warning in qemu_ld.
Richard Henderson [Mon, 22 Nov 2010 22:57:58 +0000 (14:57 -0800)]
tcg-ia64: Fix warning in qemu_ld.

The usermode version of qemu_ld doesn't used mem_index,
leading to set-but-not-used warnings.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Edgar E. Iglesias <edgar@axis.com>
13 years agotcg-ia64: Fix address compilation in qemu_st.
Richard Henderson [Mon, 22 Nov 2010 22:57:57 +0000 (14:57 -0800)]
tcg-ia64: Fix address compilation in qemu_st.

A typo in the usermode address calculation path; R3 used where R2 needed.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Edgar E. Iglesias <edgar@axis.com>
13 years agotcg-ia64: Fix tlb read error for 32-bit targets.
Richard Henderson [Mon, 22 Nov 2010 22:57:56 +0000 (14:57 -0800)]
tcg-ia64: Fix tlb read error for 32-bit targets.

Use ld4 not ld8 for reading the tlb of 32-bit targets.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Edgar E. Iglesias <edgar@axis.com>
13 years agotcg-ia64: Implement qemu_ld32.
Richard Henderson [Mon, 22 Nov 2010 22:57:55 +0000 (14:57 -0800)]
tcg-ia64: Implement qemu_ld32.

The port was not properly merged following
86feb1c860dc38e9c89e787c5210e8191800385e

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Edgar E. Iglesias <edgar@axis.com>
13 years agotcg-ia64: Provide default GUEST_BASE.
Richard Henderson [Mon, 22 Nov 2010 22:57:54 +0000 (14:57 -0800)]
tcg-ia64: Provide default GUEST_BASE.

Fix compilation error when GUEST_BASE is not defined.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Edgar E. Iglesias <edgar@axis.com>
13 years agotcg: Fix default definition of divu_i32 and remu_i32.
Richard Henderson [Mon, 22 Nov 2010 22:57:53 +0000 (14:57 -0800)]
tcg: Fix default definition of divu_i32 and remu_i32.

The arguments to tcg_gen_helper32 for these functions were not
updated correctly in rev 2bece2c88331f024a46527634e3dd91c71d22141.

Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Edgar E. Iglesias <edgar@axis.com>
13 years agoMerge remote branch 'origin/master' into pci
Michael S. Tsirkin [Wed, 1 Dec 2010 05:11:51 +0000 (07:11 +0200)]
Merge remote branch 'origin/master' into pci

Conflicts:
Makefile.objs
hw/virtio.c

13 years agoMerge remote branch 'kwolf/for-anthony' into staging
Anthony Liguori [Tue, 30 Nov 2010 21:25:34 +0000 (15:25 -0600)]
Merge remote branch 'kwolf/for-anthony' into staging

13 years agoMerge remote branch 'qmp/for-anthony' into staging
Anthony Liguori [Tue, 30 Nov 2010 21:24:26 +0000 (15:24 -0600)]
Merge remote branch 'qmp/for-anthony' into staging

13 years agoUse a Linux-style MAINTAINERS file
Anthony Liguori [Thu, 9 Sep 2010 19:51:31 +0000 (14:51 -0500)]
Use a Linux-style MAINTAINERS file

I make no claims that this is accurate or exhaustive but I think it's a
reasonable place to start.

As the file mentions, the purpose of this file is to give contributors
information about who they can go to with questions about a particular piece of
code or who they can ask for review.

If you sign up for a piece of code and indicate that it's Maintained or
Supported, please be prepared to be responsive to questions about that
subsystem.

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

v1 -> v2
 - Sort alphabetically
 - Copy in instructions from linux MAINTAINERS
 - Fix entries based on review feedback

13 years agoide: Reset current_addr after stopping DMA
Kevin Wolf [Fri, 26 Nov 2010 15:36:16 +0000 (16:36 +0100)]
ide: Reset current_addr after stopping DMA

Whenever SSBM is reset in the command register all state information is lost.
Restarting DMA means that current_addr must be reset to the base address of the
PRD table. The OS is not required to change the base address register before
starting a DMA operation, it can reuse the value it wrote for an earlier
request.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
13 years agoide: Ignore double DMA transfer starts/stops
Kevin Wolf [Fri, 26 Nov 2010 15:44:53 +0000 (16:44 +0100)]
ide: Ignore double DMA transfer starts/stops

You can only start a DMA transfer if it's not running yet, and you can only
cancel it if it's running.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
13 years agoide: Set bus master inactive on error
Kevin Wolf [Fri, 26 Nov 2010 15:47:42 +0000 (16:47 +0100)]
ide: Set bus master inactive on error

BMIDEA in the status register must be cleared on error. This makes FreeBSD
respond (more) correctly to I/O errors.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
13 years agoide: Factor ide_dma_set_inactive out
Kevin Wolf [Fri, 26 Nov 2010 15:31:37 +0000 (16:31 +0100)]
ide: Factor ide_dma_set_inactive out

Several places that stop a DMA transfer duplicate this code. Factor it out into
a common function.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
13 years agoAdd pcnet-pci.c
Paul Brook [Sat, 27 Nov 2010 11:56:02 +0000 (11:56 +0000)]
Add pcnet-pci.c

Add file missing from last commit.

Signed-off-by: Paul Brook <paul@codesourcery.com>
13 years agoSplit out common pcnet code
Paul Brook [Sat, 27 Nov 2010 11:23:34 +0000 (11:23 +0000)]
Split out common pcnet code

The core pcnet emulation code is used by both the PCI "pcnet" device
and the SPARC "lance" device.  Split the common code frm the PCI code so
that that can be configures independantly.

Signed-off-by: Paul Brook <paul@codesourcery.com>
13 years agoRemove PCI from sparc32 target
Paul Brook [Sat, 27 Nov 2010 10:33:55 +0000 (10:33 +0000)]
Remove PCI from sparc32 target

None of the (current) sparc32 machines have a PCI bus, so remove the PCI
code from these configs.

Signed-off-by: Paul Brook <paul@codesourcery.com>
13 years agoDetect missing config includes
Paul Brook [Sat, 27 Nov 2010 00:43:04 +0000 (00:43 +0000)]
Detect missing config includes

Terminate make_device_config.sh if the awk command fails.
Typically this means a missing file.

Signed-off-by: Paul Brook <paul@codesourcery.com>
13 years agoFix previous commit
Paul Brook [Sat, 27 Nov 2010 00:34:15 +0000 (00:34 +0000)]
Fix previous commit

Fix breakage from previous commit (missing pci.mak, and incorrect
include in default-configs/s390x-softmmu.mak).

Signed-off-by: Paul Brook <paul@codesourcery.com>