Merge remote-tracking branch 'awilliam/ipxe' into staging

target-i386: switch to softfloat

This increase the correctness (precision, NaN values, corner cases) on
non-x86 machines, and add the possibility to handle the exception
correctly.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: fix constants wrt softfloat

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: fix helper_fprem() and helper_fprem1() wrt softfloat

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: fix logarithmic and trigonometric helpers wrt softfloat

Use the new CPU86_LDouble <-> double conversion functions to make logarithmic
and trigonometric helpers working with softfloat.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: add CPU86_LDouble <-> double conversion functions

Add functions to convert CPU86_LDouble to double and vice versa. They
are going to be used to implement logarithmic and trigonometric function
until softfloat implement them.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: replace approx_rsqrt and approx_rcp by softfloat ops

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: fix helper_fsqrt() wrt softfloat

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: fix helper_fdiv() wrt softfloat

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: fix helper_fxtract() wrt softfloat

With softfloat it's not possible to play with the overflow of an
unsigned value to get the 0 case partially correct. Use a special case
for that. Using a division to generate an infinity is the easiest way
that works for both softfloat and softfloat-native.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: fix helper_fbld_ST0() wrt softfloat

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: fix helper_fscale() wrt softfloat

Use the scalbn softfloat function to implement helper_fscale(). This
fixes corner cases (e.g. NaN) and makes a few more GNU libc math tests
to pass.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat-native: add float*_is_any_nan() functions

Add float*_is_any_nan() functions to match the softfloat API.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat-native: fix float*_scalbn() functions

float*_scalbn() should be able to take a status parameter. Fix that.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat: fix float*_scalnb() corner cases

float*_scalnb() were not taking into account all cases. This patch fixes
some corner cases:
- NaN values in input were not properly propagated and the invalid flag
  not correctly raised. Use propagateFloat*NaN() for that.
- NaN or infinite values in input of floatx80_scalnb() were not correctly
  detected due to a typo.
- The sum of exponent and n could overflow, leading to strange results.
  Additionally having int16 defined to int make that happening for a very
  small range of values. Fix that by saturating n to the maximum exponent
  range, and using an explicit wider type if needed.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat: add floatx80_compare*() functions

Add floatx80_compare() and floatx80_compare_quiet() functions to match
the softfloat-native ones.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat-native: add a few constant values

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat: add pi constants

Add a pi constant for float32, float64, floatx80. It will be used by
target-i386 and later by the trigonometric functions.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat: add floatx80 constants

Add floatx80 constants similarly to float32 or float64.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat: fix floatx80_is_infinity()

With floatx80, the explicit bit is set for infinity.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat: fix floatx80 handling of NaN

The floatx80 format uses an explicit bit that should be taken into account
when converting to and from commonNaN format.

When converting to commonNaN, the explicit bit should be removed if it is
a 1, and a default NaN should be used if it is 0.

When converting from commonNan, the explicit bit should be added.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

vmstate: port mac_dbdma

Signed-off-by: Juan Quintela <quintela@redhat.com>

mac_dbdma: create DBDMAState instead of passing one array around

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port piix4

Signed-off-by: Juan Quintela <quintela@redhat.com>

piix4: create PIIX4State

It only contains a PCIDevice by know, but it makes easy to use migration code

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port mac_nvram

Signed-off-by: Juan Quintela <quintela@redhat.com>

mac_nvram: size is a size, no need to be a target dependent type

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port nand

Signed-off-by: Juan Quintela <quintela@redhat.com>

nand: pin values are uint8_t

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port max111x

Signed-off-by: Juan Quintela <quintela@redhat.com>

max111x: input field is only used as uint8_t

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port pxa2xx_lcd

Signed-off-by: Juan Quintela <quintela@redhat.com>

pxa2xx_lcd: up field is used as a bool and migrated as an uint8_t

Signed-off-by: Juan Quintela <quintela@redhat.com>

pxa2xx_lcd: name anonymous struct

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: stellaris use unused for placeholder entries

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port stellaris gamepad

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port syborg_keyboard

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port syborg_serial

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port stellaris_adc

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port syborg_pointer

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port ppc4xx_pci

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port ppce500_pci

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port pxa2xx_pm

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port pxa2xx_mm

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port pxa2xx_cm

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port pxa2xx_i2s

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port stellaris gptm

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port cuda

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port heathrow_pic

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port pl022 ssp

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port stellaris sys

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port stellaris ssi bus

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port stellaris i2c

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port armv7m nvic

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port pl011

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port pxa2xx_keypad

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port syborg_rtc

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port pmtimer

It was a half conversion.  Finish it.
enabled can only get values of 0, 1 or 2, was declared as an int but
sent as an unint8_t, change its type.

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port sysborg_timer

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port arm_timer

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port arm sp804

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port mipsnet

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port m48t59

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port ads7846

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port adb_mouse

Signed-off-by: Juan Quintela <quintela@redhat.com>

vmstate: port adb_kbd

Signed-off-by: Juan Quintela <quintela@redhat.com>

target-arm: Set Invalid flag for NaN in float-to-int conversions

When we catch the special case of an input NaN in ARM float to int
helper functions, set the Invalid flag as well as returning the
correct result.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

hw/arm_boot.c: move initrd load address up to accommodate large kernels

Newer kernels are large enough that they can overlap the address
where qemu places the initrd. Move the initrd up so that there is
enough space for the kernel again.

Unfortunately it's not possible to automatically determine the
size of the kernel if it is compressed, so this is the best we
can do.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

Basic implementation of Sharp Zaurus SL-5500 collie PDA

Add very basic implementation of collie PDA emulation. The system lacks
LoCoMo and graphics/sound emulation. Linux kernel boots up to mounting
rootfs (theoretically it can be provided in pflash images).

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

Implement basic part of SA-1110/SA-1100

Basic implementation of DEC/Intel SA-1100/SA-1110 chips emulation.
Implemented:
 - IRQs
 - GPIO
 - PPC
 - RTC
 - UARTs (no IrDA/etc.)
 - OST reused from pxa25x

Everything else is TODO (esp. PM/idle/sleep!) - see the todo in the
hw/strongarm.c

Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

Remove unused function parameter from cpu_restore_state

The previous patch removed the need for parameter puc.
Is is now unused, so remove it.

Cc: Aurelien Jarno <aurelien@aurel32.net>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>

Remove unused function parameters from gen_pc_load and rename the function

Function gen_pc_load was introduced in commit
d2856f1ad4c259e5766847c49acbb4e390731bd4.
The only reason for parameter searched_pc was
a debug statement in target-i386/translate.c.

Parameter puc was needed by target-sparc until
commit d7da2a10402f1644128b66414ca8f86bdea9ae7c.

Remove searched_pc from the debug statement and remove both
parameters from the parameter list of gen_pc_load.

As the function name gen_pc_load was also misleading,
it is now called restore_state_to_opc. This new name
was suggested by Peter Maydell, thanks.

v2: Remove last parameter, too, and rename the function.

v3: Fix [] typo in target-arm/translate.c.
    Fix wrong SHA1 object name in commit message (copy+paste error).

Cc: Aurelien Jarno <aurelien@aurel32.net>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>

s390x: virtio machine storage keys

For emulation (and migration) we need to know about the guest's storage keys.
These are separate from actual RAM contents, so we need to allocate them in
parallel to RAM.

While touching the file, this patch also adjusts the hypercall function
to a new syntax that aligns better with tcg emulated code.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

s390x: Dispatch interrupts to KVM or the real CPU

The KVM interrupt injection path is non-generic for now. So we need to push
knowledge of how to inject a device interrupt using KVM into the actual device
code.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

s390x: Enable s390x-softmmu target

This patch adds some code paths for running s390x guest OSs without the
need for KVM.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

s390x: Prepare cpu.h for emulation

We need to add some more logic to the CPU description to leverage emulation
of an s390x CPU. This patch adds all the required helpers, fields in CPUState
and constant definitions required for user and system emulation.

Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

PXE: Refresh all PXE ROMs from the ipxe submodule

Add script to make this easy to repeat later.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

PXE: Use consistent naming for PXE ROMs

And add missing ROMs to tarbin build target.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

move helpers.h to helper.h

This provides a consistent naming scheme across all targets.

Signed-off-by: Lluís Vilanova <vilanova@ac.upc.edu>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

libcacard: fix opposite usage of isspace

Signed-off-by: Alon Levy <alevy@redhat.com>
Tested-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-mips: clear softfpu exception state for comparison instructions

MIPS FPU instructions should start with a clean softfpu status. This
is done for the most instructions, but not for comparison ones.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-mips: fix c.ps.* instructions

Contrary to cabs.ps.* instructions, c.ps.* should not compare the absolute
value of the operand, but directly the operands.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-mips: don't hardcode softfloat exception bits

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-mips: simplify FP comparisons

As the softfloat comparison functions already test for NaN, there is no
need to always call the float*_unordered*() functions.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-ppc: fix SPE comparison functions

efstst*() functions are fast SPE funtions which do not take into account
special values (infinites, NaN, etc.), while efscmp*() functions are
IEEE754 compliant.

Given that float32_*() functions are IEEE754 compliant, the efscmp*()
functions are correctly implemented, while efstst*() are not. This
patch reverse the implementation of this two groups of functions and
fix the comments. It also use float32_eq() instead of float32_eq_quiet()
as qNaNs should not be ignored.

Cc: Alexander Graf <agraf@suse.de>
Cc: Nathan Froyd <froydnj@codesourcery.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat: improve description of comparison functions

Make clear for all comparison functions which ones trigger an exception
for all NaNs, and which one only for sNaNs.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat: move float*_eq and float*_eq_quiet

I am not a big fan of code moving, but having the signaling version in
the middle of quiet versions and vice versa doesn't make the code easy
to read.

This patch is a simple code move, basically swapping locations of
float*_eq and float*_eq_quiet.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat: rename float*_eq_signaling() into float*_eq()

float*_eq_signaling functions have a different semantics than other
comparison functions. Fix that by renaming float*_quiet_signaling() into
float*_eq().

Note that it is purely mechanical, and the behaviour should be unchanged.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat: rename float*_eq() into float*_eq_quiet()

float*_eq functions have a different semantics than other comparison
functions. Fix that by first renaming float*_quiet() into float*_eq_quiet().

Note that it is purely mechanical, and the behaviour should be unchanged.
That said it clearly highlight problems due to this different semantics,
they are fixed later in this patch series.

Cc: Alexander Graf <agraf@suse.de>
Acked-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: fix CMPUNORDPS/D and CMPORDPS/D instructions

SSE instructions CMPUNORDPS/D and CMPORDPS/D do not trigger an invalid
exception if operands are qNANs.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-mips: use new float*_unordered*() functions

Use the new float*_unordered*() functions from softfloat instead of
redefining a private version.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-alpha: use new float64_unordered_quiet() function

Use float64_unordered_quiet() in helper_cmptun() instead of doing the
the comparison manually.

According to the "Alpha Compiler Writer's Guide", we should use the
_quiet version here, as CMPTUN and CMPTEQ should generate InvalidOp
for SNaNs but not for QNaNs.

Thanks to Peter Maydell <peter.maydell@linaro.org> and Richard
Henderson <rth@twiddle.net> for digging into the manuals.

Acked-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat-native: add float*_unordered_quiet() functions

Add float*_unordered_quiet() functions to march the softfloat versions.
As FPU status is not tracked with softfloat-native, they don't differ
from the signaling version.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat: add float*_unordered_{,quiet}() functions

Add float*_unordered() functions to softfloat, matching the softfloat-native
ones. Also add float*_unordered_quiet() functions to match the others
comparison functions.

This allow target-i386/ops_sse.h to be compiled with softfloat.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: add floatx_{add,mul,sub} and use them

Add floatx_{add,mul,sub} defines, and use them instead of using direct
C operations.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: use float unions from cpu-all.h

Use float unions from cpu-all.h instead of redefining new (wrong for arm)
ones in target-i386. This also allows building cpu-exec.o with softfloat.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

target-i386: use CPU_LDoubleU instead of a private union

Use CPU_LDoubleU in cpu_dump_state() instead of redefining a union for
doing the conversion.

Based on a patch from Laurent Vivier <laurent@vivier.eu>.

Cc: Laurent Vivier <laurent@vivier.eu>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

cpu-all.h: define CPU_LDoubleU

Add a CPU_LDoubleU type, matching the floatx80 definition and the long
double type on x86 hosts.

Based on a patch from Laurent Vivier <laurent@vivier.eu>.

Cc: Laurent Vivier <laurent@vivier.eu>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

softfloat: use GCC builtins to count the leading zeros

Softfloat has its own implementation to count the leading zeros. However
a lot of architectures have either a dedicated instruction or an
optimized to do that. When using GCC >= 3.4, this patch uses GCC builtins
instead of the handcoded implementation.

Note that I amware that QEMU_GNUC_PREREQ is defined in osdep.h and that
clz32() and clz64() are defined in host-utils.h, but I think it is better
to keep the softfloat implementation self contained.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>

usb: fix spelling errors in usb-linux.c

Signed-off-by: Brad Hards <bradh@frogmouth.net>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>