constify drive_get_by_id arg

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

QemuOpts: switch over -device.

Make -device switch use the QemuOpts framework.
Everything should continue to work like it did before.

New: "-set device.$id.$property=$value" works.

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

QemuOpts: add -set option

One use case will be file for drives (no filename quoting issues), i.e.

-drive id=test,if=virtio
-set drive.test.file=/vmdisk/test-virtio.img

It will work for any other option (assuming handled by QemuOpts) though.
Except for id= for obvious reasons ;).

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

QemuOpts: create qemu-config.h

Move drive option description there.
Rename it, give it a qemu_ prefix.

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

QemuOpts: make the drive id actually show up in "info block".

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

QemuOpts: qemu_opts_parse: fix id= parsing

We can't use get_param_value(), it can't handle parameters without
'=' in there.  Examples not working because of that:

  -device foo,id=bar
  -device file=/path/image,format=qcow2,snapshot,id=disk0

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

QemuOpts: add some functions

qemu_opt_foreach: loop over all QemuOpts entries.
qemu_opts_id: return QemuOpts id.

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

Fix symfind.

this patch fixes an issue in symfind.

Assume you have the following symbols:

Address  Size
0045bca0 00000080 T s0
0045bd20 00000112 T s1

You'll notice that s1 is s0 + size.

So the current symfind will find that address 0045bd20 belongs to s0
instead of s1.

Laurent

Signed-off-by: Laurent Desnogues <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>
Message-Id:

configure: let it find xen headers and libs

Attached patch lets configure find xen headers and libs
with --extra-cflags and --extra-ldlfags option.

Signed-off-by: Christoph Egger <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>
Message-Id:

tolower -> qemu_tolower

Use qemu_tolower() instead of tolower().
Fixes warning on NetBSD.

Signed-off-by: Christoph Egger <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>
Message-Id:

Do not try to invoke shebang scripts directly (NFS issues)

Signed-off-by: malc <[email protected]>

Use qemu_irq for system_powerdown

Signed-off-by: Blue Swirl <[email protected]>

Sparc32: use qemu_irq for system_powerdown

Signed-off-by: Blue Swirl <[email protected]>

Sparc32: move sparc32_dma init to sun4m.c

Also connect ESP and Lance reset signals to DMA.

Signed-off-by: Blue Swirl <[email protected]>

Sparc32: remove VRAM and NVRAM sizes from hwdef

Signed-off-by: Blue Swirl <[email protected]>

Sparc32: remove IRQ numbers from hwdef

Signed-off-by: Blue Swirl <[email protected]>

Sparc32: move intbit_to_level table back to slavio_intctl.c

Signed-off-by: Blue Swirl <[email protected]>

Sparc32: move device instantiation to sun4m.c

Signed-off-by: Blue Swirl <[email protected]>

Sparc32: Refactor slavio timer

Signed-off-by: Blue Swirl <[email protected]>

ppc_newworld: configure screen size from QEMU command line options

Use the FW_CFG interface to send user requested screen size and depth to
OpenBIOS like 7f1aec5f93382eef75920899f4065613aeaf02a2 for ppc_oldworld.

Signed-off-by: Blue Swirl <[email protected]>

Sparc64: configure screen size from QEMU command line options

Use the FW_CFG interface to send user requested screen size and depth to
OpenBIOS like 7f1aec5f93382eef75920899f4065613aeaf02a2 for ppc_oldworld.

Signed-off-by: Blue Swirl <[email protected]>

ppc_oldworld: configure screen size from qemu command line options

This patch uses the FW_CFG interface to send user requested screen size
and depth to openbios.

Signed-off-by: Laurent Vivier <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Sparc64: replace tsptr with helper routine

tl and tsptr of members sparc64 cpu state must be changed
simultaneously to keep trap state window in sync with current
trap level. Currently translation of store to tl does not change
tsptr, which leads to corrupt trap state on corresponding
trap level.

This patch removes tsptr from sparc64 cpu state and replaces
all uses with call to helper routine.

Changes v0->v1:
- reimplemented helper routine with tcg generator
- on cpu reset trap type and pstate are populated with power-on reset
values, including tl=maxtl

Signed-off-by: [email protected]
Signed-off-by: Blue Swirl <[email protected]>

linux-user: make FUTEX_* calls honor timeout parameter

Signed-off-by: Nathan Froyd <[email protected]>
Signed-off-by: malc <[email protected]>

enable NPTL for ppc-linux-user targets in configure

Enabling support for ppc64-linux-user should be easy enough to do later.

Signed-off-by: Nathan Froyd <[email protected]>
Signed-off-by: malc <[email protected]>

linux-user: handle POWERPC_EXCP_STCX

We handle conditional stores as an exception so we can ensure that no
other thread is changing memory out from underneath us.

Signed-off-by: Nathan Froyd <[email protected]>
Signed-off-by: malc <[email protected]>

target-ppc: add exceptions for conditional stores

Signed-off-by: Nathan Froyd <[email protected]>
Signed-off-by: malc <[email protected]>

target-ppc: retain l{w,d}arx loaded value

We do this so we can check on the corresponding stc{w,d}x. whether the
value has changed.  It's a poor man's form of implementing atomic
operations and is valid only for NPTL usermode Linux emulation.

Signed-off-by: Nathan Froyd <[email protected]>
Signed-off-by: malc <[email protected]>

target-ppc: add cpu_set_tls

Signed-off-by: Nathan Froyd <[email protected]>
Signed-off-by: malc <[email protected]>

target-ppc: fix cpu_clone_regs

We only need to make sure that the clone syscall looks like it
succeeded, not clobber 60% of the register set.

Signed-off-by: Nathan Froyd <[email protected]>
Signed-off-by: malc <[email protected]>

Fix Sparse warning about missing prototype

Signed-off-by: Blue Swirl <[email protected]>

Fix Sparse warning about "expression using sizeof on a function"

Signed-off-by: Blue Swirl <[email protected]>

Add missing "static"

Signed-off-by: Blue Swirl <[email protected]>

More NULL pointer fixes

Signed-off-by: Blue Swirl <[email protected]>

Fix Sparse warnings: "Using plain integer as NULL pointer"

Signed-off-by: Blue Swirl <[email protected]>

PPC: convert Uni-north to qdev: also fixes Mac99 machine crash

Signed-off-by: Blue Swirl <[email protected]>

PPC: convert Grackle to qdev

Signed-off-by: Blue Swirl <[email protected]>

Option rom makefile fixes

Fix toplevel option rom makefile rules.

Signed-off-by: Paul Brook <[email protected]>

Save/restore ARMv6 MMU state

Correctly save/restore ARMV6 MMU state.

Signed-off-by: Paul Brook <[email protected]>

esp: fix interrupt register read

Read of interrupt register should clear it and also sequence step and status.

Signed-off-by: Blue Swirl <[email protected]>

sparc64 flush pending conditional evaluations before exposing cpu state

If translation block is interrupted by e.g. mmu exception
we need to compute conditional flags for inclusion into
saved cpu state. Otherwise after return from trap
conditional instructions would use stale psr/xcc data.

Signed-off-by: [email protected]
--
Kind regards,
Igor V. Kovalenko

Fix SDL zooming with pl110 (cf. d3ffcafe25b5966b351ea6100160c2156688f22f)

Signed-off-by: Blue Swirl <[email protected]>

e1000.c doesn't properly emulate EERD and ICS registers

Once again, the emulation of the EERD and ICS registers in e1000.c is
incorrect. Nobody has noticed this before because none of the Intel-written
e1000 drivers use these registers, and all of the independently written open
source drivers copy Intel's example, so they don't use them either.
Regardless, these registers are documented in the programmer's manuals, and
their emulated behavior doesn't match the verified behavior of real hardware,
so any software that does use them doesn't function correctly.

-Bill

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

qemu-option: fix parse_option_number().

It works much better when parse_option_number actually
returns the number parsed ...

Common breakage resulting from this bug is that
'qemu -hda foo.img -cdrom bar.iso' stops working
(cdrom isn't there).

Cc: Avi Kivity <[email protected]>
Signed-off-by: Gerd Hoffmann <[email protected]>
Signed-off-by: Anthony Liguori <[email protected]>

qdev: convert all vga devices.

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

qdev/core: add monitor command to list all drivers

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

qdev: factor out qdev_print_devinfo.

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

Fix VM state change handlers running out of order

When a VM state change handler changes VM state, other VM state change
handlers can see the state transitions out of order.

bmdma_map(), scsi_disk_init() and virtio_blk_init() install VM state
change handlers to restart DMA.  These handlers can vm_stop() by
running into a write error on a drive with werror=stop.  This throws
the VM state change handler callback into disarray.  Here's an example
case I observed:

0. The virtual IDE drive goes south.  All future writes return errors.

1. Something encounters a write error, and duly stops the VM with
   vm_stop().

2. vm_stop() calls vm_state_notify(0).

3. vm_state_notify() runs the callbacks in list vm_change_state_head.
   It contains ide_dma_restart_cb() installed by bmdma_map().  It also
   contains audio_vm_change_state_handler() installed by audio_init().

4. audio_vm_change_state_handler() stops audio stuff.

5. User continues VM with monitor command "c".  This runs vm_start().

6. vm_start() calls vm_state_notify(1).

7. vm_state_notify() runs the callbacks in vm_change_state_head.

8. ide_dma_restart_cb() happens to come first.  It does its work, runs
   into a write error, and duly stops the VM with vm_stop().

9. vm_stop() runs vm_state_notify(0).

10. vm_state_notify() runs the callbacks in vm_change_state_head.

11. audio_vm_change_state_handler() stops audio stuff.  Which isn't
   running.

12. vm_stop() finishes, ide_dma_restart_cb() finishes, step 7's
   vm_state_notify() resumes running handlers.

13. audio_vm_change_state_handler() starts audio stuff.  Oopsie.

Fix this by moving the actual write from each VM state change handler
into a new bottom half (suggested by Gleb Natapov).

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

vnc: fix copyrect screen corruption

When sending a copyrect command to the vnc client, we must also update
the local server surface.  Otherwise the server's and the client's idea
of the screen content run out of sync and screen updates don't work
correctly.

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

Remove the virtio-{blk, console}-pci-0-10 device types

These are now unused.

However, perhaps the idea is that when we add -device, they will be
useful? In that case, we should add virtio-net-pci-0-10 too.

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

Remove the pc-0-10 machine type

We have the pc-0.10 machine type now which does exactly the same
thing.

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

fix migration to obey -S

Since migration returns right away, starting the VM right
after calling qemu_start_incoming_migration is wrong even
if -S is not passed.  We have to do this after migration
has completed.

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

alsa: add host suspend/resume support

Both input and output streams may be in SND_PCM_STATE_SUSPENDED
after the host is suspended and resumed, meaning "Hardware is
suspended".  snd_pcm_readi() and snd_pcm_writei() will return
-ESTRPIPE if called while the stream is in this state.

Call snd_pcm_resume() to enable audio output and capture after
host resume.

Signed-off-by: Bjørn Mork <[email protected]>
Signed-off-by: malc <[email protected]>

Define ENOTSUP anywhere it isn't defined

I confused ENOTSUP with ENOTSUPP.  Juan's original patch was correct.

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

Revert "Fake dirty loggin when it's not there"

This reverts commit bd8367761236cd5c435598aeb2f1b8240c09b059.

PPC should just implement dirty logging so we can avoid all the fall-out from
this changeset.

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

Fix out of tree build broken by 791e08c753a9f9be3c3880b4ea83b6dfa4b6ccad

Signed-off-by: Blue Swirl <[email protected]>

Fix typo WIN32 -> _WIN32

This was spotted by Juan Quintela

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

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

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

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

Remove lots of duplicate code in the process

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

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

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

Don't build option roms on Mac OS X

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

multiboot.bin is a generated file

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

use cc-option for optionrom

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

Use quiet-command for building optionrom

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

move cc-option definition to rules.mak

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

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

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

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

rename USE_NPTL to CONFIG_USE_NPTL

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

rename NEEDS_LIBSUNMATH to CONFIG_NEEDS_LIBSUNMATH

Once there fix a place where it was misspelled

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

define ENOTSUP the same that the other errors

aliguori: ENOTSUP is not 4096 universally, only on OpenBSD

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

test if xen is enabled only once

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

simplify fdt libs selection

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

simplify vde libs selection

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

simplify brlapi selection

Use same style that everythnig else

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

simplify curses library selection

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

Define and use xen libs in a single place

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

indent with 2 spaces sdl tests

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

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

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

target_softmmu is undefined at that point

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

instead of setup sdl_x11 just add -lX11 to sdl_libs

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

Calculate sdl_libs and sdl_flags and use them everywhere

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

sdl_config value was always sdl-config

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

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

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

Use CONFIG_POSIX to simplify Makefile

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

Add CONFIG_POSIX

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

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

rename WORDS_BIGENDIAN to HOST_WORDS_BIGENDIAN

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

rename DEBUG_EXEC to CONFIG_DEBUG_EXEC

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

rename DEBUG_TCG to CONFIG_DEBUG_TCG

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

rename HAVE_BYTESWAP_H to CONFIG_BYTESWAP_H

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

rename HAVE_GPROF to TARGET_GPROF

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

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

rename HAVE_MACHINE_BSWAP_H to CONFIG_MACHINE_BSWAP_H

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

rename HAVE_FNMATCH_H to CONFIG_FNMATCH

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

rename HAVE_IOVEC to CONFIG_IOVEC

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

rename HAVE_PREADV to CONFIG_PREADV

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

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

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

rename HOST_BSD to CONFIG_BSD

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

O_LARGEFILE is already set in qemu-common.h

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

fix indententation

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

Refactor code to remove one #ifdef CONFIG_FDT

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

Rename HAVE_FDT to CONFIG_FDT and define it also in Makefile

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

Add -static in configure if needed

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