qemu-option: Fix uninitialized value in append_option_parameter

When dest is NULL, i.e. a new copy of the list is created, we don't get a
properly terminated list after the realloc. Initialize it as an empty list.

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

Change #define DEBUG to #define E1000_DEBUG in hw/e1000.c

Change #define DEBUG to #define E1000_DEBUG in hw/e1000.c to make
it possible to build QEMU with -DDEBUG

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

Remove unused DEBUG defines from hw/msix.c

Remove unused DEBUG defines from hw/msix.c to avoid having anything
define the word DEBUG without any additions such as MSIX_DEBUG.

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

Merge remote branch 'kwolf/for-anthony' into staging

Conflicts:
hw/pc.c

Merge remote branch 'qmp/for-anthony' into staging

OHCI address decoding fix

Ignore high address bits when PCI memory window is not mapped on a page
boundary.

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

Move stdbool.h

Move inclusion of stdbool.h to common header files, instead of including
in an ad-hoc manner.

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

monitor/QMP: Drop info hpet / query-hpet

This command was of minimal use before, now it is useless as the hpet
become a qdev device and is thus easily discoverable. We should
definitely not set query-hpet in QMP's stone, and there is also no good
reason to keep it for the interactive monitor.

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

hpet: Add MSI support

This implements the HPET capability of routing IRQs to the front-side
bus, aka MSI support. This feature can be enabled via the qdev property
"msi" and is off by default.

Note that switching it on can cause guests (at least Linux) to use the
HPET as timer instead of the LAPIC. KVM users should recall that only
the latter is currently available as fast in-kernel model.

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

hpet: Make number of timers configurable

One HPET block supports up to 32 timers. Allow to instantiate more than
the recommended and implemented minimum of 3. The number is configured
via the qdev property "timers". It is also saved/restored so that it
need not match between migration peers.

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

vmstate: Add VMSTATE_STRUCT_VARRAY_UINT8

Required for hpet.

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

hpet: Add support for level-triggered interrupts

By implementing this feature we can also remove a nasty way to kill qemu
(by trying to enable level-triggered hpet interrupts).

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

hpet: Drop static state

Instead of keeping a static reference around, pass the state to
hpet_enabled and hpet_get_ticks. All callers now have it at hand. Will
once allow to instantiate the HPET more than a single time.

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

hpet/rtc: Rework RTC IRQ replacement by HPET

Allow the intercept the RTC IRQ for the HPET legacy mode. Then push
routing to IRQ8 completely into the HPET. This allows to turn
hpet_in_legacy_mode() into a private function. Furthermore, this stops
the RTC from clearing IRQ8 even if the HPET is in control.

This patch comes with a side effect: The RTC timers will no longer be
stoppend when there is no IRQ consumer, possibly causing a minor
performance degration. But as the guest may want to redirect the RTC to
the SCI in that mode, it should normally disable unused IRQ source
anyway.

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

hpet: Start/stop timer when HPET_TN_ENABLE is modified

We have to update the qemu timer when the per-timer enable bit is
toggled, just like for HPET_CFG_ENABLE changes.

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

hpet: Convert to qdev

Register the HPET as a sysbus device and create it that way. As it can
route its IRQs to any ISA IRQ, we need to connect it to all 24 of them.
Once converted to qdev, we can move reset handler and vmstate
registration into its hands as well.

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

hpet: Move static timer field initialization

Properly initialize HPETTimer::tn and HPETTimer::state once during
hpet_init instead of (re-)writing them on every reset.

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

hpet: Silence warning on write to running main counter

Setting the main counter while the HPET is enabled may not be a good
idea of the guest, but it is supported and should, thus, not spam the
host console with warnings.

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

hpet: Coding style cleanups and some refactorings

This moves the private HPET structures into the C module, simplifies
some helper functions and fixes most coding style issues (biggest chunk
was improper switch-case indention). No functional changes.

Signed-off-by: Jan Kiszka <[email protected]>
Reviewed-by: Juan Quintela <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

hpet: Catch out-of-bounds timer access

Also prevent out-of-bounds write access to the timers but don't spam the
host console if it triggers.

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

configure: Fix evaluation of config-host.mak in create_config

Only match on true dir variable assignments, avoid generating garbage
due to the "# Configured with: ..." line which may contain "*dir=" as
well.

Signed-off-by: Jan Kiszka <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

tcg-s390: correctly detect s390 with a 64-bit kernel

Signed-off-by: Aurelien Jarno <[email protected]>

Compile OS specific files only once for all targets

OS specific files are not target dependent, so they can be compiled
once for all targets.

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

Move set_proc_name() to OS specific files.

Move handling to change process name to POSIX specific files
plus add a better error message to cover the case where the
feature isn't supported.

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Move line-buffering setup to OS specific files.

Move line-buffering setup to OS specific files.

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Make os_change_process_uid and os_change_root os-posix.c local

os_change_process_uid() and os_change_root() are now only called
from os-posix.c, so no need to keep win32 stubs for them.

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Move daemonize handling to OS specific files

Move daemonize handling from vl.c to OS specific files. Provide dummy
stubs for Win32.

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Move chroot handling to OS specific files.

Move chroot handling to OS specific files.

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Move runas handling from vl.c to OS specific files.

Move code to handle runas, ie. change of user id of QEMU process
to OS specific files and provide dummy stub for Win32.

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Introduce OS specific cmdline argument handling and move SMB arg to os-posix.c

Introduce OS specific cmdline argument handling by calling
os_parse_cmd_args() at the end of switch() statement. Move option
enum to qemu-options.h and have it included from os-posix.c and
os-win32.c in addition to vl.c.

In addition move SMB argument to os-posix.c

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Rename qemu-options.h to qemu-options.def

Rename qemu-options.h to qemu-options.def as it is not a header file
for general use and this leaves space for a proper qemu-options.h

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Move find_datadir to OS specific files.

This moves the win32 and POSIX versions of find_datadir() to OS
specific files, and removes some #ifdef clutter from vl.c

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Move main signal handler setup to os specificfiles.

Move main signal handler setup to os specific files.

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Rename os_setup_signal_handling() to os_setup_early_signal_handling()

Rename os_setup_signal_handling() to os_setup_early_signal_handling()

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Move win32 early signal handling setup to os_setup_signal_handling()

Move win32 early signal handling setup to os_setup_signal_handling()

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Introduce os-posix.c and create os_setup_signal_handling()

Introcuce os-posix.c and move posix specific signal handling
there.

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

vl.c: Move host_main_loop_wait() to OS specific files.

Move host_main_loop_wait() to OS specific files. Create
qemu-os-posix.h and provide empty inline for the POSIX case.

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Introduce os-win32.c and move polling functions from vl.c

This introduces os-win32.c. It is meant to carry win32 specific
functions thata are not relevant for all of QEMU as well as win32
versions of various pieces like signal handling etc.

Move win32 polling handler helper functions from vl.c to os-win32.c

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

Create qemu-os-win32.h and move WIN32 specific declarations there

Create qemu-os-win32.h for WIN32 specific declarations. Move polling
handling declaration into this file from sysemu.h

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

vl.c: Remove double include of netinet/in.h for Solaris

vl.c: netinet/in.h is already included once above for the

Signed-off-by: Jes Sorensen <[email protected]>
Acked-by: Andreas Faerber <[email protected]>
Acked-by: Juan Quintela <[email protected]>
Acked-by: Richard Henderson <[email protected]>
Signed-off-by: Blue Swirl <[email protected]>

NEON vldN optimization

When combining multiple values as part of a NEON array load, do explcit
shift/or rather than using gen_bfi.  This voids redundant mask
operations.

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

hxtool: Fix line number reporting on SQMP/EQMP errors

Signed-off-by: Jan Kiszka <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

remove unnecessary lookaheads

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

implement optional lookahead in json lexer

Not requiring one extra character when lookahead is not necessary
ensures that clients behave properly even if they, for example,
send QMP requests without a trailing newline.

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

add some tests for invalid JSON

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Luiz Capitulino <[email protected]>

json-streamer: Don't use qdict_put_obj()

It's not needed, use qobject_put() instead and get a cleaner code.

Signed-off-by: Luiz Capitulino <[email protected]>

json-lexer: Drop 'buf'

QString supports adding a single char, 'buf' is unneeded.

Signed-off-by: Luiz Capitulino <[email protected]>

check-qjson: Add more escape tests

While there make the fail_unless() calls print error messages.

IMPORTANT: The test for "\/" is failing, don't know why.

Signed-off-by: Luiz Capitulino <[email protected]>

qjson: Handle "\f"

It's valid JSON and should be handled.

Signed-off-by: Luiz Capitulino <[email protected]>

json-lexer: Handle missing escapes

The JSON escape sequence "\/" and "\\" are valid and should be
handled.

Signed-off-by: Luiz Capitulino <[email protected]>

json-lexer: Initialize 'x' and 'y'

The 'lexer' variable is passed by the caller, it can contain anything
(eg. garbage).

Signed-off-by: Luiz Capitulino <[email protected]>

tcg: fix DEF macro after commit c61aaf7a388c4ad95d8b546fdb9267dc01183317

Signed-off-by: Aurelien Jarno <[email protected]>

tcg-s390: Adjust compilation flags.

Force -m31/-m64 based on s390/s390x target.

Force -march=z990.  The TCG backend will always require the
long-displacement facility, so the compiler may as well make
use of that as well.

Signed-off-by: Richard Henderson <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

tcg-s390: Compute is_write in cpu_signal_handler.

Signed-off-by: Richard Henderson <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

s390x: Don't use a linker script for user-only.

The default placement of the application at 0x80000000 is fine,
and will avoid the default placement for most other guests.

Signed-off-by: Richard Henderson <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

s390x: Avoid _llseek.

There's no _llseek on s390x either.  Replace the existing
test for __x86_64__ with a functional test for __NR_llseek.

Signed-off-by: Richard Henderson <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

tcg-s390: Allocate the code_gen_buffer near the main program.

This allows the use of direct calls to the helpers,
and a direct branch back to the epilogue.

Signed-off-by: Richard Henderson <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

tcg-s390: Icache flush is a no-op.

Before gcc 4.2, __builtin___clear_cache doesn't exist, and
afterward the gcc s390 backend implements it as nothing.

Signed-off-by: Richard Henderson <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

tcg-i386: fix andi r, r, 0xff

Signed-off-by: Aurelien Jarno <[email protected]>

tcg-i386: remove use of _Bool that slipped code review

Signed-off-by: Aurelien Jarno <[email protected]>

esp: lower IRQ on soft reset

42f1ced228c9b616cfa2b69846025271618e4ef5 removed irq lowering
during reset. However, for chip reset command and DMA reset signal,
its actually the correct thing to do.

Lower IRQ on soft reset only.

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

qbus: fix memory leak in qbus_free()

BusState::name is allocated in qbus_create_inplace().
So it should be freed by qbus_free().

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

multiboot: compilation fix with DEBUG_MULTIBOOT enabled.

This patch fixes the following compilation errors in multiboot.c
when DEBUG_MULTIBOOT is defined.
Use TARGET_FMT_plx instead of %x for target_phys_addr_t.

  CC    i386-softmmu/multiboot.o
cc1: warnings being treated as errors
qemu/hw/multiboot.c: In function 'mb_add_mod':
qemu/hw/multiboot.c:121: error: format '%08x' expects type 'unsigned int', but argument 4 has type 'target_phys_addr_t'
qemu/hw/multiboot.c:121: error: format '%08x' expects type 'unsigned int', but argument 5 has type 'target_phys_addr_t'
qemu/hw/multiboot.c: In function 'load_multiboot':
qemu/hw/multiboot.c:279: error: format '%#x' expects type 'unsigned int', but argument 5 has type 'target_phys_addr_t'
qemu/hw/multiboot.c:307: error: format '%x' expects type 'unsigned int', but argument 3 has type 'target_phys_addr_t'
qemu/hw/multiboot.c:308: error: format '%x' expects type 'unsigned int', but argument 3 has type 'target_phys_addr_t'
make[1]: *** [multiboot.o] Error 1

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

vga-isa-mm: remove one #ifdef CONFIG_BOCHS_VBE.

remove one #ifdef CONFIG_BOCHS_VBE.
Call vga_init_vbe() instead.

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

main: allocate gui_timer only once.

fix memory leak.
there is no need to allocate more than one gui_timer.

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

Merge remote branch 'mst/for_anthony' into staging

cris: Break out image loading to hw/cris-boot.c.

Signed-off-by: Edgar E. Iglesias <[email protected]>

s390: Disassemble some general-instruction-extension insns.

The full general-instruction-extension facility was added to binutils
after the change to GPLv3.  This is not the entire extension, just
what we're using in TCG.

Signed-off-by: Richard Henderson <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

s390: Update disassembler to the last GPLv2 from binutils.

Signed-off-by: Richard Henderson <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

tcg-i386: Merge 64-bit generation.

Signed-off-by: Richard Henderson <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: display sysconfdir in summary

Signed-off-by: Aurelien Jarno <[email protected]>

move computation of tools and roms outside of config-host.mak generation

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: move directory defaults earlier

Unify with existing special-purpose configure code for win32.

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: ignore unknown --xyzdir options

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: introduce more --xyzdir options

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: expand ${prefix} in create_config

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: move all directory entries in config-host.mak close

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: unify handling of xyzdir variables

Making an xyzdir variable for each directory prepares for the next
patches introducing config-host.h defines and configure options for them.
It also fixes the problem where overriding prefix at "make install"
time did not override it for sysconfdir.

Removes some of the differences between sysconfdir and other variables,
the rest will go away later.

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

rename CONFIG_QEMU_PREFIX

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: introduce confdir and confsuffix

confsuffix was write-only, flesh it out.

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: introduce sysconfsuffix

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: delete duplicate create_config case stanza

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: dyngen is long time gone

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: avoid using expr

Just a personal preference against duplicating hieroglyphics.

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: bail out early on invalid -cpu option

It would fail later anyway.

Signed-off-by: Paolo Bonzini <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

configure: remove some bashisms

Signed-off-by: Aurelien Jarno <[email protected]>

target-mips: Fix compilation

TCGv t1 needs tcg_temp_free instead of tcg_temp_free_i32.

Cc: Nathan Froyd <[email protected]>
Cc: Aurelien Jarno <[email protected]>
Signed-off-by: Stefan Weil <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

hw: honor low bit in mipssim machine

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

linux-user: honor low bit of entry PC for MIPS

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

target-mips: add microMIPS exception handler support

Unlike MIPS16, microMIPS lets you choose the ISA mode for your exception
handlers.  The ISA mode is selectable via a user-writable CP0.Config3
flag.

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

target-mips: microMIPS ASE support

Add instruction decoding for the microMIPS ASE.  All we do is decode and
then forward to the existing gen_* routines.

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

target-mips: mips16 cleanups

Change code handling mips16-specific branches to use ISA-neutral special
opcodes.  Since there are several places where the delay slot
requirements for microMIPS branches differ from mips16 branches, using
opcodes is easier than checking hflags, then checking mips16
vs. microMIPS.

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

target-mips: refactor c{, abs}.cond.fmt insns

Move all knowledge about coprocessor-checking and register numbering
into the gen_cmp* helper functions.

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

target-mips: move FP FMT comments closer to the definitions

Signed-off-by: Aurelien Jarno <[email protected]>

target-mips: define constants for magic numbers

Add FMT_* constants for the floating-point format field in opcodes and
tweak a few places to use them.  Add enums for various invocations of
FOP and tweak gen_farith and its lone caller accordingly.

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

tcg: get rid of DEF2 in tcg-opc.h

Now that tcg-opc.h is only used in TCG code, get rid of DEF2 in
tcg-opc.h.

Signed-off-by: Aurelien Jarno <[email protected]>

tcg: get rid of copy_size in TCGOpDef

copy_size is a left-over from the dyngen era, remove it.

Signed-off-by: Aurelien Jarno <[email protected]>

Fix multiboot compilation

Commit dd4239d6574ca41c94fc0d0f77ddc728510ffc57 broke multiboot. It replaced the
instruction "rep insb (%dx), %es:(%edi)" by the binary output of
"addr32 rep insb (%dx), %es:(%di)".

Linuxboot calls the respective helper function in a code16 section. So the
original instruction was automatically translated to its "addr32" equivalent.
For multiboot, we're running in code32 so gcc didn't add the "addr32" which
breaks the instruction.

This patch splits that helper function in one which uses addr32 and one which
does not, so everyone's happy.

The good news is that nobody probably cared so far. The bundled multiboot.bin
binary was built before the change and is thus correct.

Please also put this patch into -stable.

Signed-off-by: Alexander Graf <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>

net: Fix hotplug with pci_add

The correct model type wasn't getting added when hotplugging nics with
pci_add.

Testcase: start VM with default nic type. In the qemu_monitor:

(qemu) pci_add auto nic model=virtio

This results in a nic hot-plug of the same nic type as the default.

This was broken in 5294e2c774f120e10b44652ac143abda356f44eb

Also changes the behaviour where no .init is defined for a
net_client_type. Previously, 0 was returned, which indicated the init
was successful and that 0 was the index into the nd_tables[] array.
Return -1, indicating unsuccessful init, in such a case.

Signed-off-by: Amit Shah <[email protected]>
Signed-off-by: Michael S. Tsirkin <[email protected]>

tcg: Make some tcg-target.c routines static.

Both tcg_target_init and tcg_target_qemu_prologue
are unused outside of tcg.c.

Signed-off-by: Richard Henderson <[email protected]>
Signed-off-by: Aurelien Jarno <[email protected]>